The British Journal of Criminology

Miller,, Vincent;Hayward, Keith, J

doi: 10.1093/bjc/azy017pmid: N/A

Abstract This paper considers the recent phenomenon of the vehicle-ramming attack (VRA): i.e. the act of purposely driving a vehicle into pedestrians and populated vehicles. It documents the recent (2015–2017) rise in the prevalence of ramming attacks and how these incidents challenge some of the assumptions we have about terrorism and its causes. Typically, criminologists and terrorist scholars tend to focus on either the ‘psychology’ of individual terrorists or wider structural or ethno-political issues, such as religion, ideological doctrine or the role of terrorist organizations in converting and recruiting people to violence. This paper will adopt a different position, one which focusses less on structure and individual psychology, and more on the act itself, as something that is not merely an expression of an individual or an ideology, but something that has a lure and force all of its own, as something that travels through our contemporary mediascape, to be internalized and imitated by an increasingly varied set of subjects with varying motivations, psychologies, ideologies and circumstantial backgrounds. ‘A complex weapon makes the strong stronger, while a simple weapon—so long as there is no answer to it—gives claws to the weak’. (Orwell, 1945) Introduction Shortly after midnight on 19 June 2017, Darren Osbourne, a heavy-drinking, 47-year-old resident of Pentwyn, Cardiff, deliberately swerved a rented white Luton panel van into a crowd of people outside the Finsbury Park Mosque, North London. The victims, a group of Muslim men who had just left the Mosque after attending Ramadan night-time prayers, were gathered together to assist a passer-by who had collapsed on the pavement outside. Even without this distraction, it is doubtful that the mosque-goers could have avoided the attack; such was the innocuous ubiquity of Osbourne’s indiscriminate but deadly weapon. Osbourne’s intentions, however, were anything but innocuous.1 After his automotive missile came to a halt against some nearby bollards, he leapt from the cab and, wild-eyed and sweating, shouted ‘I’m going to kill all Muslims—I did my bit’, before declaring that his actions were revenge ‘for London Bridge’ (Mortimer 2017).2 In the days following the attack, the press struggled to make sense of Osbourne’s rapid descent into extremism. Certainly, his biography provided few clues as to what had motivated him to undertake the attack. We do know he was deeply affected by recent Islamic State-inspired terrorist attacks—including two involving vehicle-ramming incidents.3 But what is equally clear is that, prior to his decision to attack the mosque, Osbourne had no history of—or even interest in—violent extremism. Indeed, described variously by his sister, Nicola, as someone who was ‘not interested in terrorism’ or even politics in general (‘He wouldn’t even know who the Prime Minister is’), and by his drinking partners as ‘a loveable mentalist’, the picture of Osbourne that surfaced was of a ‘shouty’ but normal ‘family man’ (Ward et al. 2017)—‘Troubled, but not a racist’ (Mortimer 2017). Even those who disliked Osbourne appeared shocked by his actions; one of his more forthcoming Pentwyn neighbours commenting ‘He’s always been a complete c***, but this is really surprising’ (ibid). Such quotes are typical of the narrative of bewilderment that quickly emerges in the wake of egregious (Western) terrorist incidents. For decades, it was standard reportorial protocol for journalists to write that serial killers were always ‘quiet neighbours’ who ‘kept themselves to themselves’. In recent years, a similar cliché now commonly depicts terrorists as ‘non-descript’, ‘normal’, even ‘fun-loving’ individuals who exhibited no prior evidence of the violence to come. Such reportage is not, however, without foundation. One of the most robust and oft-repeated research findings within the discipline of terrorism studies is that, rather than suffering from any profound psychopathology, ‘the outstanding characteristic of terrorists is their normality’ (Crenshaw 1981:390; Corrado 1981; Silke 1998). This ‘ordinariness’ finding was a necessary rejoinder to the common misconception of terrorists as ‘crazies’ or psychopaths, but at the same time it marked the end of the search for anything as identifiable as a distinct terrorist ‘personality type’.4 This does not mean that terrorism scholars have failed to produce any valuable insights into terrorist behaviour. On the contrary, there exists a veritable mass of excellent psycho-sociological scholarship on the terrorist actor (see e.g. Horgan 2005; Post 2007; McCauley and Moskalenko 2011; Gill 2016; and Cottee and Hayward 2011 for one of the author’s own efforts). Instead, it is simply to point out that it remains difficult to extrapolate out from the individual, psycho-biographical level to make wider points that can be used to improve security and enhance detection. This problem is especially apparent with vehicle ramming, a phenomenon marked by a surprising diversity of perpetrators (in terms of their ideologies, biographies and geographical locations). It is largely for this reason that, in the analysis of the vehicle-ramming attack (VRA) that follows, we have chosen to focus not on the perpetrator per se, but on a very different, and much-neglected, aspect of terrorism: the terrorist act itself. In this case, the specific phenomenon of purposely driving a vehicle into groups of civilians or police/military personnel. In criminology, of course, there is nothing new about training attention exclusively on the criminal act. In fact, since the 1980s, situational crime preventionists have done precisely that, downplaying any interest in offender motivation and stressing instead the importance of the criminal event itself and the situational factors that influence its commission (e.g. Cornish and Clarke 1986). The explicit goal of this ‘administrative criminology’, then and now, remains beguilingly straightforward: to implement practical, cost-effective crime control measures that will bring about a quantifiable reduction in the overall rate of specific crimes. Nothing wrong with this of course, but our rationale for focusing attention on the criminal/terrorist event does not stem from a situational perspective—although obviously, we hope that the theoretical analysis of VRAs presented here will help preventionists better combat the phenomenon in the future. Instead, our interest in vehicle ramming is due to its growing appeal as a simplistic but potent means of doing terrorism—of performing an act of violence against civilians for political or symbolic purposes. Vehicle ramming is a cheap but extremely effective way of inducing fear, promoting an ideological message, or just simply wreaking destruction for any number of personal reasons, suicide included. It is perhaps for this reason that ramming attacks are now being conducted by such a wide array of perpetrators. Yet the fact that VRAs are easy to carry out and impossible to predict does not explain why they have become so prevalent over the last year or so. It is this particular aspect of the vehicle-ramming event that interests us. Why, given the mass availability of automobiles for many decades around the world, have these motorised crimes suddenly been adopted by everyone from committed Islamic State-affiliated jihadists, to ‘lone-wolf’ Palestinians and other Muslims with no previous links to organized terrorist groups, to anti-Muslim extremists, to American right-wing Christians, to unbalanced members of the public at large? In a bid to answer this question, the paper will proceed in three parts. Part one provides some historical and geographical context in the form of a short genealogy of VRAs that traces (and statistically documents) two recent steep rises in ramming incidents; a first pulse that took place in Jerusalem and the West Bank between 2015 and 2016 and a second ongoing ‘wave of vehicular terror’ that started in 2016 and is currently affecting Europe and North America. In part two, we address the most common extant explanations for the rise of global ramming attacks. For example, we consider the publication of a series of articles in Al Qaeda and Islamic State online propaganda magazines that discuss and actively encourage VRAs in the name of Jihad. However, as our analysis will show, though relevant to the situation, the idea that these two publications can somehow account for the recent spike in VRAs (as is often suggested in the media) is too reductive. If, as we suggest, standard rationalisations for VRAs are problematic, what other factors might have contributed to the rapid increase in what was once an extremely rare form of violence? We set ourselves this task in the final part of the article. Inspired by the social theory of Gabriel Tarde (2012/1895, 1903, 2010/1888), part three introduces the concepts of ‘contagion’, ‘imitation’ and the ‘mimetic’ and assesses their value in terms of understanding the current spate of VRAs. More specifically, we assert that, as a result of the intensely connected and networked nature of contemporary society, at least some of the explanatory focus should now be trained on the terrorist act itself as something with a seductive appeal and force all of its own, as something that travels through our contemporary ‘mediascape’, to be internalized and imitated by a varied array of subjects who in turn are animated by a diverse set of motivations, psychologies and ideologies. After a general consideration of Tarde’s imitation-based approach to the social world, the final part divides into three brief subsections that each illustrate how his thinking can be used to understand some of the prominent features associated with the VRA: (1) the sudden, wave-like growth and diffusion of the act of vehicle ramming; (2) its enactment by a diverse set of (predominantly) ‘lone-actor’ terrorists; and (3) the influence of contemporary mediascapes in the sudden lurch towards political violence that is a common feature of vehicle rammers. In conclusion, we assert that Tarde’s work is likely to be of considerable use when it comes to analysing subsequent trend-like pulses in terrorist modus operandi. A brief history of the vehicle-ramming attack In his typically engaging book Buda’s Wagon: A Brief History of the Car Bomb, Mike Davis documents what he describes as ‘the irreversible globalization of car-bombing’. Likening the ‘steep, almost exponential’ rise of vehicle-born bombing in the final third of the twentieth century to ‘an implacable virus’ that inserted itself into a ‘host society’, Davis claims that ‘the most dramatic impact of the car bomb has been precisely its enfranchisement of marginal actors in modern history’ (2007:11). Although the bloody violence and territorial reach associated with the car bomb far exceeds that of the VRA, it is difficult not to also note some striking similarities. To start with, like its more explosive counterpart, VRAs also seem to have something of a virus-like quality. For example, in the last four years, the VRA has transitioned from being a relatively rare occurrence, to become, by 2016, the most lethal form of terror attack in Western countries, claiming just over half of all terrorism-related deaths in the West that year.5 (Importantly, these figures do not include the growing number of ramming incidents in North America and Europe that have been classified as ‘non-terrorist’ in origin). The two terror tactics are also similar in that, by transforming a bland, everyday object into a lethal, semi-strategic weapon, they empower marginal actors by providing them with the means to strike at the heart of urban centres and sow fear in the wider society. Finally, although both techniques are frequently employed by well-organized terrorist groups, neither method is exclusive to terrorism (nor as we have seen above, to Islamist terrorism). Instead, both acts have a history of being adopted, adapted and transformed by groups and perpetrators coming from a variety of religious and non-religious backgrounds and ethnicities—including so-called ‘lone-wolf’ actors, as exemplified by the Osbourne case above.6 Given such similarities, it would seem instructive to follow Davis’s lead and undertake a brief history of the VRA. To understand more about the nature of VRAs, we created a database of 125 such attacks from 1999 to the present. The primary source for this database was the National Consortium for the Study of Terrorism and Responses to Terrorism Global Terrorism Database (https://www.start.umd.edu/gtd/). However, due to certain limitations,7 we supplemented the sample with information drawn from the RAND Database of Worldwide Terrorism Incidents (https://www.rand.org/nsrd/projects/terrorism-incidents.html), as well as further examples from our own media searches. Inclusion in the database required the following criteria to be met: (1) A vehicle had to be used as a weapon on pedestrians or populated vehicles. (2) The vehicle had to be the primary weapon. Any other weapons used (knives, firearms) could only be used when the vehicle was disabled and not part of an ‘armed assault’ or ‘car bomb’. (3) The incident was not part of a kidnapping attempt. (4) Pedestrian casualties were intended, and not part of a chase, evasion or accident. (5) The vehicle was not used primarily for demolition. This yielded a database of 125 attacks from 1 January 1999 to 31 December 2017, and a pattern demonstrated in Figures 1 and 2 below: Fig. 1 View largeDownload slide Total vehicle-ramming attacks worldwide, six-month intervals 1999–2017 Fig. 1 View largeDownload slide Total vehicle-ramming attacks worldwide, six-month intervals 1999–2017 Fig. 2 View largeDownload slide Vehicle-ramming attacks by region 1999–2017 Fig. 2 View largeDownload slide Vehicle-ramming attacks by region 1999–2017 Here is the distribution separated for location of either Western Europe/North America, Middle East/North Africa, Rest of the world: The figures demonstrate a sporadic history of VRAs in different parts of the world up until 2014. At that point, a rise in ramming incidents occurs, beginning with a wave of attacks in the Middle East (almost exclusively in Israel and the West Bank, see Figure 2), which, by the end of 2015, becomes quite dramatic. This is followed by a second spike starting in 2015 across North America and Western Europe (Figure 2). Let us look at each in turn. ‘The intifada of the individuals’ When examined more closely, the 2015 spike in ramming attacks in Israel and the West Bank can be seen as part of a larger wave of violence known as ‘The wave of terror’ (Ostrovsky 2015), or ‘The intifada of the individuals’ (Ackerman 2015; Nashashibi 2015; David 2016; Pfeffer 2016). In Israel, military officials now use both appellations to describe a spontaneous escalation in unsophisticated, ‘lone-wolf’ DIY attacks by Palestinians on Israelis that involved knives and other sharp-edged instruments (cleavers, screwdrivers, etc.), but also—more unusually—vehicular ramming. Indeed, during this time, ramming itself went from a rare occurrence to the second most common form of attack in Israel and the West Bank (Eglash et al 2016). Commentary on both sides of the conflict attributed this unforeseen ‘wave’ of terror to a well-established list of causes: poor economic conditions among Palestinian youth, lack of progress in the peace negotiations and conflicts over Israeli access to the Temple Mount. But, although initially there was an all-too-familiar feel to both the attacks and the explanatory commentary, it quickly emerged that certain aspects of this latest pulse of attacks marked it out as different from previous intifadas. To start with, many of the attacks during this period were carried out by so-called ‘lone-wolf’ assailants; individuals with no previous history of violent political behaviour and no established association with existing terrorist organizations. In the words of Nashashibi (2015), such is the level of the Palestinians’ exasperation, ‘they are rising up as individuals rather than as a collective mass (out of necessity, not choice), without the relative safety of numbers and the backing of their leaders’. Second, it appeared that the attackers’ motivation for violence was also more varied and personal than in other periods of the conflict. In other words, as Pfeffer makes clear in the following quotation, though all expressed frustration with the plight of Palestine, the triggering factor was often as much personal as it was political or ideological: The attackers have a social media presence, usually on Facebook but also on Twitter and Instagram. Analysis of their behaviour online has yielded profiles of potential attackers and, in many cases, also the additional motivations for a suicide attack. Often these are personal and non-ideological reasons such as debt or forced marriage. In many cases, Israeli security services believe the real motivation behind attacks was the desire for “suicide by IDF”, attaining the status of martyrdom and sparing family embarrassment. (Pfeffer 2016) These two points are empirically substantiated in a recent study by Perry et al. (2017). Drawing on a database supplied by the Israeli Security Agency, Perry and his colleagues examined 62 VRAs perpetrated by ‘lone terrorists’ in Israel and the West Bank between January 2000 and March 2016. They point out that none of the 62 acts were committed by a terrorist organization and that only 14.5% of the sample had previously engaged in terrorist activity. They also stressed that, while over half of the perpetrators declared a nationalistic or religious motivation for their attack (suggesting that they were affected by the political situation around them), in many cases, political events shrouded or were used to justify other unrelated personal grievances.8 Finally, as Ackerman makes clear, many of the attacks associated with ‘the intifada of the individuals’ were unsophisticated, unpredictable and, interestingly, very hastily planned, using simple ‘DIY’ weapons (typically knives and vehicles). These types of terrorist attacks cannot be detected ahead of time, as there is no planning or strategy behind them. As a result, the security forces have no way of preventing them. In these cases, intelligence gathering does not give us an advantage… When lone-wolf attacks succeed, other individuals think to themselves, “Hey, I can do that too.” (Ackerman 2015) Importantly, it is not simply that more successful Israeli security measures have created a situation where only knives and cars are available (as is sometimes claimed, see the section on From nothing to something? Extant explanations). Instead, it is symptomatic of the style of the attack, one lacking in organization and planning (Beaumont 2015). A style that is perhaps inspiring simply because it is so easy to repeat. Ripple effect: beyond the West Bank and into the West Although VRAs and other related forms of violence associated with the intifada of the individuals peaked in 2015 (and actually started to decline in Israel and the West Bank by the end of 2016), we can clearly see from Figure 2 that, by late 2016, ramming incidents had started to spread to other regions—most notably Europe and North America. For example, by the end of 2017 (a year not currently covered by the Global Terrorism Database), there had already been many more VRAs in Western cities (32) than in the West Bank and Jerusalem (10). In large part, this increase in Western attacks can be attributed to a spate of spectacular and deadly terrorist actions by Islamic State-inspired supporters. Consider, for example, the following high-profile VRAs undertaken by jihadists in the last 12 months of the sample: • 19 December 2016: failed asylum-seeker, Anis Amri, steers a hijacked HGV into a Christmas market in Breitscheidplatz, Berlin, killing 12, injuring 56. • 22 March 2017: British Muslim Khalid Masood drives his car into pedestrians on Westminster Bridge, London, killing five, injuring fifty. • 7 April 7 2017: Uzbek asylum-seeker Rakhmat Akilov attacks a pedestrianized shopping street in Stockholm, Sweden, killing five, injuring fifty. • 3 June 2017: three Muslim men drive a van into pedestrians on London Bridge before attacking passers-by in a nearby market with knives, killing eight, injuring 48. • 8 August 2017: Hamou Benlatreche plows his car into a military patrol in Levallois-Perret, Paris, injuring six soldiers. • 17 August 2017: a jihadist cell undertakes twin VRAs in Barcelona and Cambrils, Spain. Fifteen people were killed and over 130 were injured. • 31 October 2017: Sayfullo Saipov drove a ‘Home Depot’ truck onto bike paths in Manhattan, deliberately targeting bicyclists and pedestrians, before finally driving into a school bus. Eight people were killed, eleven injured. However, it should not be forgotten that alongside these horrendous attacks a growing number of ‘non-terrorist’ ramming events also took place in Western cities in 2017—a great many of which, just like in Palestine and Israel, were either spontaneous, chaotically planned or undertaken by people with no previous affiliation with terrorist networks. Once again, when listed, these events make for chilling reading: • 20 January 2017: Dimitrious Garagasoulas, a 26-year-old Australian, stabbed his brother (allegedly for being gay), then went on a driving rampage across Melbourne, killing 6 people, wounding 30. • 25 February: a German man armed with a knife drove a rented car into a square in Heidelberg, killing one, injuring two. • 18 May: a US military veteran drove his car into pedestrians in Times Square, New York City in an apparent ‘suicide by cop’ attempt, killing one, injuring twenty. • 29 July: one person was killed and four injured when a drunk man deliberately careered into a group of pedestrians in Helsinki, Finland. • 14 August: Eric Patterson crashed his car into a pizzeria terrace in a small shopping area outside Paris, France, killing one and injuring thirteen. • 21 August: a man with a history of petty crime deliberately rammed pedestrians waiting at two separate bus stops in Marseilles, France with a stolen white van, killing one, injuring one. • 10 November: a man deliberately drove his car into a group of Chinese students in Blagnac, France, injuring three. • 2 December: five people were injured in south-west London when a rental car rammed pedestrians after an altercation between occupants of the vehicle and the pedestrians. Such incidents illustrate that, not only has vehicle ramming spread beyond the Israeli-Palestinian conflict and into Western cities, but it is now also a favoured practice of Jihadists, anti-Islamists, right-wing Christians, and unbalanced members of the public at large. As Youself Munayyer, Executive Director for the US Campaign for Palestinian Rights, suggested in an interview: The tools that people use to carry out attacks don’t have a religion, trucks don’t have an ideology. They just require somebody to drive them… This is a question of utility and opportunity and tactics more than it is about ideologies. Israeli settlers have used their cars to run over Palestinian civilians in the West Bank. It’s a tactic that can be used by lots of different people. (Munayyer, Cited in Moore 2017) This is certainly true, but it does not account for why VRAs have emerged as such a popular option at this particular moment in history. To understand this development, we must look elsewhere. From nothing to something? Extant explanations In the face of the spate of VRAs that have occurred in Western cities since 2016, three online articles have frequently been held up in the press as both the inspiration behind and the catalyst for this new terrorist modus operandi. In the Fall 2010 edition of Inspire, the official magazine of Al Qaeda in the Arabian Peninsula, a feature article appeared entitled ‘The ultimate mowing machine’. The title page depicted a Ford F-Series ‘Super Duty’ pickup truck. However, in the Inspire article the original Ford slogan—‘The ultimate towing machine’—had been changed in a bid to encourage the ‘mowing down’ of the enemies of Allah.9 ‘The ultimate mowing machine’ was followed in 2016 and 2017 by two similar articles in Rumiyah, an online magazine used by the Islamic State for propaganda and recruitment purposes (‘Just terror tactics’, Issue 3, November 2016; and ‘Just terror tactics: truck attacks’ Issue 9, May 2017). All three articles encouraged vehicle ramming as an action in which lone-actor jihadists could easily and without suspicion obtain (preferably large) vehicles and then target festivals, crowded city centres and other ‘pedestrian only’ areas for maximum destructive effect. This seemingly straightforward explanation is problematic for a number of reasons. To start with, when one examines the publication dates of these articles more closely, their release bears little significance to the actual pattern of attacks as shown in Figure 3 below. For example, if ‘The ultimate mowing machine’ was so influential, why did it take more than four years for the first high-profile VRAs to occur in Western cities? Indeed, further analysis reveals that the original Inspire article may have itself been inspired by a March 2006 incident in which a man, wanting to ‘avenge the deaths of Muslims around the world’, drove a rented Jeep Cherokee SUV into a crowd of people in a busy pedestrian area at the University of North Carolina, Chapel Hill, injuring nine. Interestingly, this attack was followed just a few weeks later by another SUV ramming incident, this time of pedestrians in San Francisco, California, by a schizophrenic Afghani immigrant.10 Furthermore, although there was a discernible pulse of VRAs following the publication of the ‘Just terror attacks’ article in Rumiyah in November 2016,11 this article actually appeared after both the wave of ramming incidents in Israel and the West Bank, and, importantly, Europe’s most deadly and most publicised VRA—the July 2016 Bastille Day attack in Nice, France, which killed 87 and injured 433.12 Likewise, the May 2017 article was published after highly mediated attacks in Berlin, Stockholm and London. None of this is to suggest that these three now infamous articles have not played an important part in promoting the concept of the VRA, only that they should not be seen as causal.13 If the ‘The ultimate mowing machine’ and the two subsequent Rumiyah articles were merely an articulation of an ongoing VRA wave rather than, as some have suggested, actually precipitating or causing it, how else might we account for the current popularity of this mode of attack? Fig. 3 View largeDownload slide Ramming attacks by three-month intervals, 2010–2017 Fig. 3 View largeDownload slide Ramming attacks by three-month intervals, 2010–2017 A second extant explanation relates to situational developments associated with the so-called ‘security hypothesis’ (Farrell et al. 2011). In relation to the VRA, the argument would go as follows: because situational crime preventionists have successfully ‘reduced the rewards of terrorism’ by physically target-hardening structures and situations that traditionally have been the focus of terrorist attacks (the most obvious example here being the implementation of airport scanning equipment to combat terrorist “skyjackings”), contemporary terrorists have shifted their attention to ‘softer targets’ (e.g. unprotected pedestrians) and more mundane methods (e.g. vehicle ramming). No doubt this argument has merit. Certainly, proponents of the situational approach argue that the shift away from complex, so-called ‘expeditionary terrorism’ (Kilcullen 2009:32) (such as the 1993 and 2001 World Trade Centre attacks) towards more ‘parochial, devolved, amateurish, ugly, and uncinematic’ forms of terrorism (Cottee, 2017) is evidence of the effectiveness of Crime Prevention through Environmental Design (CPtED) and related preventative concepts such as Newman and Clarke’s (2008) ‘EVIL DONE’ vulnerability index (Freilich and Newman 2009).14 Yet on closer inspection, there is a problem. If targeted preventative measures against terrorism have been in position for decades (starting with the aforementioned control and regulation of airports in the 1970s, continuing with urban counter-terrorist measures such as those deployed against the Irish Republican Army in Northern Ireland and London in the 1980s and 1990s, and culminating in the type of technically-sophisticated target-hardening that has come to define the post-9/11 era cityscape), why is it only in the last two years that VRAs have become commonplace? If the two most prevalent explanations for the rise of the VRA are unsatisfactory, how else might we account for this burgeoning phenomenon? Interestingly, there does exist one final, albeit almost entirely ignored, explanation—not of VRAs per se, but of previous modes of terrorism that also appeared to follow a wave/trend-like pattern. Consider, for example, the skyjacking epidemic of the late-1960s and early-1970s, a phenomenon that, as Brendan Koerner suggests, has striking similarities with today’s ramming situation. The perpetrators of these crimes often said they were acting to support one of the era’s fashionable political causes—the Black Power movement, for example, or the crusade to end the Vietnam War. But if you scratched beneath the surface, you often found people in desperate straits—people like Roger Holder, a PTSD-afflicted Vietnam vet who ostensibly hijacked a Western Airlines jet to Algeria as part of a convoluted plot to win the freedom of American political activist Angela Davis but was also keen to avoid a looming court date for fraud. Or Paul Joseph Sini, an alcoholic loner who claimed to be an affiliate of the Irish Republican Army but who really hijacked Air Canada Flight 812 and demanded a $1.5 million ransom because he was sick of feeling worthless… When interviewed in prison, many hijackers confessed that they’d become intrigued after viewing news footage of stolen planes; when they committed their own crimes, they were often mindful of the fact that their exploits would be aired to millions. (Koerner 2016) The contagion-like nature of skyjacking and other similarly ‘fashionable’ terrorist practices have not passed without comment. Holden (1986) stressed this very theme in his article on the subject, whereas Jenkins (1981) demonstrated similar effects with regard to embassy takeovers in the 1970s. More generally, Midlarsky et al. (1980) found a contagion pattern in the spread of ‘terrorism’ itself, as well as with specific terrorist actions, from nationalist movements in the developing world to radical groups in Western Europe. However, in terms of the argument we will develop in the remainder of the article, perhaps, the most important paper on ‘the contagion effect’ is not on trends in terrorist practice per se, but on the (‘copycat’) relationship between heavy media coverage of high-profile killings and subsequent clusters of violent crimes. Berkowitz and Macauley (1971) established a relationship between several of the more notorious crimes of 1960s America, including the JFK assassination, the Richard Speck killing of eight nurses in Chicago in 1966 and the mass shooting at the University of Texas that same year, and rises in violent crimes nationwide. All these studies, to a greater or lesser degree, evoke something that we feel might help explain the current spate of VRAs. This work, for shorthand purposes can be incorporated under the umbrella term ‘contagion theory’, an unsung field of study which posits that, much like biological contagion, if one person is exposed to a phenomenon (either in person or through the media), this can serve as the stimulus for the imitative social actions of another, thereby explaining the subsequent (sudden) spatial or temporal clustering of certain social phenomena. It is to this body of work that we now turn. Gabriel Tarde, Contagion Theory and the Vehicle-Ramming Attack as a Mimetic Phenomenon What is notable, theoretically speaking, about ‘exposure’ or ‘contagion’ as a causal phenomenon is that it shifts the analysis and explanation of behaviour away from ‘subjects’ in themselves to emphasize instead the relational aspects of interaction, exposure and the flow of behaviours between persons. As Marsden (2005) notes, this shift in emphasis away from an autonomous individual subject is one reason mainstream social science has never really taken contagion theory seriously, as any approach that seeks to explain adult behavioural patterns via ‘social replication’ undermines the traditional, Western, Cartesian understanding of the human subject as an agent defined solely by individual intentionality/evaluation. ‘Social contagion’ is thus incompatible with the idea of the self-contained, rational calculating actor. Such criticism has also been voiced within terrorism studies specifically, with the likes of Picard (1986) dismissing the concept of social contagion as a ‘dangerous idea backed by dubious science’. Yet, might it be time for terrorist researchers and criminologists to do what other branches of the social sciences are now doing and reconsider contagion-like arguments in light of the Internet age?15 Certainly, if we consider the speed, intensity and scope of the wave-like spread of VRAs over the last two years, it would seem that, of the currently available explanations, the ‘contagion approach’, centred as it is on networked technologies and imitative processes associated with ‘the viral’, has much to offer. Indeed, in recent years, the imitative and mimetic processes associated with contagion have been reconsidered by ‘digital culture’ scholars interested in how information is exchanged and disseminated through the networked technologies of the Internet. Most obviously, this body of work has extended the biological metaphor via the use of the terms ‘viral’ and ‘meme’ (e.g. Sampson 2012; Guadagno et al. 2013; Shifman 2014).16 Work on ‘contagion’ has its roots in turn-of-the-century ‘crowd’ theorists such as Gabriel Tarde and Gustav le Bon. Tarde (1903, 2010/1888, 2012/1895), in particular, is noteworthy here, as his work proved to be extremely influential in the early development of sociology, psychology and (lest we forget) criminology (Bierne 1993), both in Europe and North America (Leys, 1993; Kinnunen 1996). More recently, Tarde’s work has found favour not just with scholars examining the viral or mimetic nature(s) of the Internet (Sampson 2012; Burgess et al 2017), but also with sociologists of ‘affect’ (Blackman 2007), Actor-Network Theorists involved in innovation, diffusion, and science and technology studies (Latour 2002), and researchers interested in the spread of new social and political movements (Marrero-Guillamón 2013). For Tarde (1903), imitation is a (perhaps the) basic process of the social world. According to his thesis, imitation occurs as the by-product of human co-existence and is a construct he used to develop a view of society and social actors that privileges neither individual ‘psychology’ nor deterministic social structures as the primary initiates of social action. Later work on ‘mimesis’ by Walter Benjamin also resonated with Tarde, with the German philosopher noting: ‘There is perhaps not a single one of his [sic] higher functions in which his [sic] mimetic faculty does not play a decisive role’ (Benjamin 1999/1933: 720). Importantly, imitation here does not refer to a process of simple observation and mimicry, but to the general openness of human subjectivity to the influence of the surrounding social world—i.e. the ability to affect and be affected, consciously and unconsciously, by others (Blackman 2007, 2012). Accordingly, Tarde’s ‘inter-psychological’ model of subjectivity concentrated on how subjects internalize, almost in the manner of hypnotic ‘suggestion’, the ideas, desires, beliefs, emotions and thus behaviours, of others, and then either reject, replicate or adapt (innovate) them to suit their own individual circumstances. By emphasizing imitation and interaction, Tarde focused attention on the contingent and fluid nature of the social as a continual process. At the same time, in place of the concept of ‘individual’, Tarde proposed a social ontology of the self as a kind of ‘socialized monad’ (Candea 2016: 12); a subject that is a reflection (or microcosm) of the social whole and thus inherently open to change through the influence of external factors (Schmid 2004). So, while ‘individual’ preferences, cognitions and decisions exist because of varying biographies of exposure, all of these important elements of the decision-making process (and subsequent behaviour) are internalized from others as part of the process of social interaction and imitation. Thus, social beings are ‘social’ because they internalize the external. In Tarde’s own words, social actors deeply influence—or, more accurately, ‘mutually possess’—one another through diverse interactive and imitative processes (Sampson 2012).17 Thus, a society or culture can be seen as being made up of a ‘career of imitations’ that has been created through a constant set of individual interactions over time (Marsden 2000). Tarde’s concept of imitative innovation helped him explain both the idea (and the origins) of social similarity and (at the same time) the sometimes rapid spread of new ideas, practices, beliefs, tactics and affects through populations. In this sense, his work should be seen as an attempt to explain both social stability and social change without relying on static/binary notions of the ‘individual’ vs ‘the social’, agency vs structure or subject vs object. Tarde’s ideas provide an important theoretical contextualization to research which emphasizes the role of ‘exposure’, ‘contagion’ or ‘mimesis’ in the spread of social phenomena. More specifically in terms of our overall hypothesis, we argue that Tarde’s theory of imitation can help us better explain VRAs, not only as ‘expressions’ of either structural/ideological or psychological factors, but as social actions that have an energy and force all of their own. In the following three short subsections, we suggest that Tarde’s thinking can provide a useful starting point for thinking about certain peculiar aspects of VRAs: i.e. their ‘wave-like’ pattern of temporal clustering (which inevitably makes us question why these attacks suddenly became so popular despite the ready availability of cars and trucks around the world for decades), the diversity of perpetrators involved in the attacks, and the often spontaneous nature of the incidents. Vehicle ramming as an imitative wave Although clearly very different social phenomena, VRAs and online events such as viral media campaigns, memes and different forms of online activism/awareness-raising may not be as different as one might think. Both exhibit the same wave-like distribution or temporal clustering—a pattern of inactivity, a rapid increase in take-up, interest or diffusion (usually on a profoundly international scale) and finally a longer, but still rapid, drop-off in participation. For example, this typical pattern of wave-like temporal clustering was demonstrated when seventeen million people participated in the ‘Ice bucket challenge’, an imitative meme that captured global attention and participation for a brief period in 2014. Such patterned spontaneity would doubtless have fascinated Tarde. In his day, he was struck by how quickly innovations, ideas, fashions, technologies and crimes could be diffused throughout a population or a crowd. Moreover, he realized the importance of the new communication technologies of his time—newspapers, railways, the telegraph, etc.—in promoting the increasingly rapid transmission of ideas across dispersed populations (Barry and Thrift 2007). It was these technologies, especially the media, that turned localized ‘crowds’ into dispersed ‘publics’, subsequently transforming an imitative process that was once spatially clustered, into a ‘contagion without contact’ (Tarde 1901: 11, cited in Gibbs 2008) in which the suggestibility of publics was demonstrated by temporal clustering of social phenomena. This process of diffusion, exposure, connection and influence became the contemporary social fact for Tarde, and these ‘pulses’ or ‘waves’ of innovation, change or action are what he sought to explain in his work. Social things, which maintain and perpetuate themselves by the individual consciousness’s through which they evolve, are like an ocean wave, which crosses innumerable molecules and seems to animate them even while living from their force (Tarde 2010/1888:120). Tarde was not interested in the ‘subject’ so much as with the interactive processes that (he felt) modified individuals, and thus created similarities among them. He was, therefore, not focused on ‘individual’ persons per se, nor with societal structures, but with the force of micro-actions (be it ideas, fashions, behaviours, communications, emotions) which, he claimed, move through social life not unlike the flow of an ocean wave (Sampson 2012). Although waves have their origins in individual events, they also gain their own momentum and force, a force which can be dissipated or amplified, and can even merge with other waves. Tarde saw the events of social life as analogous: having their origins in the creativity of specific individual actions, but also gaining their own momentum—what he called an ‘imitative radiation’ that created ‘lines of force which traverse the individual person’ (Barry 2005: 54). This suggests a view of social phenomena themselves as vitalistic forces that emerge through relationality and encounter. Such an ‘imitative radiation’ is apparent, for example, in Towers et al.’s (2015) recent study of mass killings and school shootings in the United States. Applying a ‘contagion model’ to recent datasets of firearm incidents that left four or more people killed, Towers et al. concluded that there is ‘significant evidence’ that these incidents ‘are incented by similar events in the immediate past’—with the ‘temporary increase in probability’ lasting 13 days. This ‘contagion effect’, they argue, is due in large part to intense cycles of media coverage of these shootings that prolong and diffuse the event among dispersed populations. It is a sentiment echoed in recent commentary on VRAs: The fact that this has been broadcast, the attacks in Nice and Berlin have been shown so many times in the international media, simply had the effect of drawing the attention of people that this is an effective instrument… People copy, people imitate, you see it coming and going… (Uzi Arad, Israel’s National Security Council, quoted in Moore 2017) These waves of imitative radiation are evident both internationally (where global ramming incidents have gone from insignificance to over 40 per year within two years) and at a more localized, micro-level. For example, on the 21 December 2014, 11 civilians were injured in Dijon, France, when a mentally unstable ‘40-year-old man of Arab origin’ used a Transit-style van as a weapon in five parts of the city in the space of 30 minutes. Within 24 hours, a Frenchman, with a history of petty crime, alcoholism and mental health issues drove his van into shoppers at a Christmas market in Nantes, injuring ten. He then stabbed himself 13 times in the chest with a knife. Authorities believe he had no political or religious motive, but was directly inspired by the Dijon incident the previous evening (Lichfield 2014). Likewise, the inspiration for Osbourne’s June 2017 attack on the Finsbury Mosque can be traced (in large part) to his desire for revenge for the London Bridge attack 16 days before. Interestingly, the Osbourne attack was then itself copied just four days later by Marek Zakarocki, a Polish immigrant to the United Kingdom and ‘Britain First’ supporter, who allegedly shouted ‘I am going to kill a Muslim. I’m doing it for Britain’, before twice driving his white van at the owner of a London curry house. Six days later, a Frenchman, also wanting to avenge recent Islamic State attacks in Paris, was arrested after driving his car into a crowd in front of a mosque in a Paris suburb. At the most basic explanatory level, then, Tarde’s ideas provide an interesting, tentative framework for thinking about the rapid rise and spread of VRAs from an exceedingly rare practice just three years ago, to something now occurring with alarming frequency in countries across the globe. However, these examples not only highlight the wave-like clustering of VRAs, but also the diversity of the perpetrators involved in terms of their motives, ideologies and individual circumstances. In the next section, we demonstrate how such diversity can also be explained using Tardian theory. Imitation, anti-structure and the diversity of VRA perpetrators According to Tarde (1903), it is the imitation process, rather than social structures, that provides the basis for both the socialization and the actions of individuals. Hence, the social is not a fixed entity, but is continually (re)made through transmissions/processes and the resultant actions of imitation, opposition and adaptation (Burgess et al. 2017). This ‘anti-structural’ and ‘anti-individual’ approach to the origins of human behaviour is useful when we consider the diverse motivations of vehicle rammers. Previously, we observed how VRAs spread (in short order) from the Palestinian conflict to jihadist attacks in Western cities, ultimately encompassing a diverse array of attackers with a variety of motives. We similarly noted that even within the narrower confines of the ‘Intifada of Individuals’, it was clear from commentary on both sides that the Palestinian perpetrators had little in common in terms of their motivation (David 2016; Pfeffer 2016). Figure 4 illustrates this spread of perpetrators over time from an almost exclusively Palestinian practice, to jihadists in Western cities, to people form a variety of backgrounds and motives. Fig. 4 View largeDownload slide VRA perpetrators by group (2010–2017) Fig. 4 View largeDownload slide VRA perpetrators by group (2010–2017) Many, but not all, VRA perpetrators were inspired by some sense of injustice towards either Palestinians or Muslims. Others were not Muslims at all but instead wanted to avenge jihadist attacks. Some, but not all, suffered from a history of mental health problems. Some had histories of petty violence, some did not. Others merely wanted to kill themselves. Focusing more closely on 2017 sharply illustrates this diversity. Of the 42 VRAs, we catalogued in that year, ten were perpetrated by Palestinians in Israel and the West Bank and another ten were attributed to jihadists. Four incidents were related to the ramming of (left-wing) protestors; four attributed to ‘revenge’ against Muslims; five appeared to be spontaneous responses to previous altercations or disputes and nine committed by those with either no clear motivation, mental health issues or aspirations of suicide. This is notable because when we conceptualize a ‘wave’ of ramming attacks, the tendency is to gloss over the act itself and to look instead at its ‘cause’, either in terms of individual psychology or the social structures and ideologies that surround the individual. Too often in the analysis of terrorism, the act itself is rarely considered relevant, apart from being a means either to demonstrate an ideological goal, or to express the will of an individual, sane or otherwise. This, in turn, leads one to see ‘cause’ in a shared ideology, or a common individual alienation or mental illness. Of course, these are relevant, but why not also consider the material act itself? Given the eclecticism of vehicle rammers in terms of personal/ideological motivation, it seems reasonable to pay attention to the one thing that unites their actions—the attack mode: the ‘style’, the instruments (cars, trucks, knives) involved, and the exposure to similar actions conducted by others through mass media and the Internet. To see the attack exclusively as an expression of individual psychology, or larger ideological/structural ‘social facts’, is to potentially miss out on the power of the act itself to inspire and to be worth imitating, either meticulously or more spontaneously. Vehicle ramming and emotional contagion: mimesis and the contemporary mediascape As discussed above, in terms of their biographies, relationships to organized terrorist groups and lack of planning, a considerable number of contemporary vehicle rammers contradict the established blueprint of the ideologically committed, well-organized terrorist actor. These observations resonate with the findings of the Israeli National institute for Security Studies which suggested that the perpetrators of the ‘Intifada of Individuals’ were not typical of previous ‘intifadas’ in that they were largely ‘lone wolf’ assailants, often with no previous history of political violence and no established links with existing terrorist organizations. Instead, they were diverse individuals driven by different personal circumstances who were ‘inspired by the media (the internet, television, and the press), which provided them with illustrations of the injustice that Israel commits against the Palestinians and sparked them to take action’ (David 2016). This is particularly prescient to a contemporary digital media culture in which the notion of ‘exposure’ to the acts of others has become extremely complicated. Although contagion research in the 1970s demonstrated how the proliferation of broadcast media brought with it a new age of copycat crimes and terrorist inspiration, the move to a decentralised, networked, interactive mediascape has created an even more complicated, uncontrollable and immersive media experience. Such environments provide almost continual, instantaneous exposure to the acts and experiences of others around the world. Castells (1996) famously referred to this technologically enabled, networked existence of continual connection beyond one’s own physical locality as ‘the space of flows’, and it is reasonable to suggest that processes like ‘contagion’ become markedly more significant in a space of flows that allows for so many vectors of ‘exposure’, so many acts to imitate, and so many emotions with which to resonate and rationalize one’s action. Not without relevance here is recent work that examines so-called ‘emotional contagion’ (Neumann and Strack 2000; Howard and Gengler 2001), which demonstrates how exposure to the emotional states of others, affects one’s own emotional mood or position. The most notorious illustration of this was the Facebook ‘emotional contagion’ experiment of 2012 (Kramer et al 2014), which found that people’s moods closely mirrored the moods of others around them, even in non-proximal, online environments. Indeed, Gibbs (2008) sees social media, in particular, with its technological capacities for connection and social influence as ‘the locus par excellence of imitation which imagines itself to be original’ (ibid: 138) and in that sense contributes to the production of an affective ‘mimetic field’, which again can be demonstrated in the intensive, wave-like nature of much online social phenomena. Such a ‘mimetic field’ for car-based violence has also been articulated in the US in recent years through several popular memes. ‘Run them over’ was a saying which spread through comments sections on YouTube videos, blogposts and news stories depicting confrontations between frustrated drivers and ‘Black lives matter’ protestors on the streets of American cities. In February 2017, the ‘All lives splatter’ meme (Figure 5) emerged out of the ‘Take America back’ Facebook page and was widely disseminated through right-wing websites (Grabar 2017). Similar sentiments and incidents18 were replicated in a cluster of YouTube videos all encouraging or celebrating the ramming of left-wing protesters in both the run-up to and the aftermath of the Charlottesville VRA in August 2017.19 In total, we found six incidents of vehicle ramming involving the targeting of protestors in 2016 and 2017. Fig. 5 View largeDownload slide “All lives splatter” internet meme Fig. 5 View largeDownload slide “All lives splatter” internet meme The haphazard, lone-operator nature of VRAs and the often sudden lurch towards violence, both highlight the contingent nature of attacker’s identities and the usefulness of Tarde’s more fluid notion of the individual as monad—something always in formation as it internalizes the external world. Such thinking inevitably differentiates Tarde’s ‘imitative’ approach from rational choice or media ‘effects’ conceptions that rely on a self-enclosed, ‘Cartesian’ notion of the subject. More problematically in terms of prevention and risk-assessment, this view of the late modern terrorist as an unbounded subject continually open to internalizing and emulating external influences and acts, does much to undermine the static notions of ‘identity’ behind the psychologically constructed ‘personality types’ popular with twentieth-century terrorism researchers. For some, this detour into digital culture studies and contagion theory will seem unnecessary. However, to adopt such a position would be to ignore the fact that the entire history of twentieth-century terrorism has itself been the subject of a wave-like metaphor. Here, we refer to Rapoport’s (2001) much-cited categorization of four historical waves of modern terrorism.20 According to Rapoport, each ‘wave’ had a distinct ‘energy’ which influenced the formation of new groups under particular ideologies. Indeed, if we consider J.D Simon’s (2011) recent augmentation of the Rapoport framework—i.e. his emerging ‘fifth wave’ of terrorism—even more parallels are apparent. Most obviously, Simon asserts that the next wave of terror will be undertaken not by highly coordinated terrorist groups with clearly demarcated political or religious objectives, but by ‘lone-operator’ terrorists. Moreover, these actors will plan their acts and distribute their message via digital technologies; hence, Simon’s description of this new era as a ‘technological’ wave whose ‘energy’ will come from its ability to spread information, tactics, awareness, radical ideas and perhaps even emotional contagion through the diverse vectors of the Internet. This being the case, it would seem that many of the sociological theories currently being used to understand memes, virals and other digital phenomena are likely to become increasingly relevant to any analysis of new trends and patterns in late modern terrorism. Conclusion The very same night Osbourne committed his attack outside the Finsbury Park Mosque, Djaziri Adam Lotfi, an IS supporter on a French terrorist watch list, ploughed his car into Gendarmerie on the Champs Elysees. A few days later, in an attack that closely mirrors the Osbourne incident, a disgruntled Frenchman was arrested while attempting to drive his car into a crowd near a Mosque in suburban Paris. Like Osbourne, he too was seeking to avenge recent Islamic State attacks. Such incidents illustrate the appeal of vehicle ramming as a means of symbolic violence, a form of attack as likely to be employed by a ‘remote-controlled’ Islamic State supporter as by an organizationally unaffiliated, disgruntled loner. In short, just as car bombing brought about the ‘enfranchisement of marginal actors in history’ (Davis 2007: 11), vehicle ramming has further democratized terrorism to the point where, as Cottee (2017) suggests, it now ‘just requires a willingness to kill and die. Indeed, just about any village idiot can aspire to become a martyr to a cause that politically makes no real sense and that they barely understand’. But even though VRAs have transformed the Western terrorist act into something more akin to a misdirected ‘crime of passion’ than an act of ‘expeditionary terrorism’, it is likely that the current spate of VRAs will eventually subside just as other wave or trend-like terrorist phenomena have done in the past. This is not to suggest that the latest ‘fifth’ wave of terrorism is petering out. On the contrary, our point is that, even if we do see a marked decline in vehicle ramming (as was the case with ‘the intifada of individuals’), another equally tawdry and devolved form of violence will quickly emerge to take its place. Indeed, our media-saturated, networked society makes such waves not just likely, but inevitable. Not only will the next democratized terrorist methodology be widely available to view and download on an array of digital formats, but more worryingly, online ‘filter bubbles’ and other forms of personalized digital isolation will continue to intensify the ongoing conflation of personal and political grievances that is the product of online hate speech and weaponized identity politics. In previous decades, terrorists required both a viable (successful) tactical exemplar and a warrant from a credible/worthy organization before embarking on an action. Today, the former is only a meme away and the latter no longer needed. Such a situation demands that criminologists and terrorism scholars play closer attention to developing terrorist trends. But, documenting the frequency and diffusion patterns of whatever practice follows in the wake of vehicle ramming will not be enough. We must also take a page from Tarde’s book and acknowledge that certain high-profile acts have a relational energy and imitative force all of their own. Only then, when we have given the mimetic power of terrorism its due causal weight, will we be truly capable of combating it. References Ackerman , L . ( 2015 ), ‘ Has A Third Intifada Begun ?’, The Jerusalem Post , 15 October, available online at https://www.jpost.com/Opinion/Has-a-third-intifada-begun-426125. Barry , A . ( 2005 ), ‘ Pharmaceutical Matters: The Invention of Informed Materials ’, Theory, Culture and Society , 22 : 51 – 69 . Google Scholar Crossref Search ADS Barry , A. and Thrift , N . ( 2007 ), ‘ Gabriel Tarde: Imitation, Invention and Economy ’, Economy and Society , 36 : 509 – 25 . Google Scholar Crossref Search ADS Beaumont , P . ( 2015 ), ‘ What’s Driving the Young Lone Wolves Who are Stalking the Streets of Israel ?’, The Guardian , 18 October, available online at https://www.theguardian.com/world/2015/oct/18/knife-intifada-palestinian-israel-west-bank. Benjamin , W . (1999/1933) , “ On the Mimetic Faculty, ” in M . Jennings , H. Eiland , and G. Smith eds., Selected Writings , 1926 – 34 , Belknap Press . Berkowitz , L. and Macaulay , J . ( 1971 ), ‘ The Contagion of Criminal Violence ’, Sociometry , 238 – 60 . Bierne , P . ( 1993 ), Inventing Criminology , SUNY Press . Blackman , L . ( 2007 ), ‘ Reinventing Psychological Matters: The Importance of the Suggestive Realm of Tarde’s Ontology ’, Economy and Society , 36 : 574 – 96 . Google Scholar Crossref Search ADS —— ( 2012 ), Immaterial Bodies , Sage . Blackmore , S . ( 2000 ), The Meme Machine , Oxford University Press . Burgess , A. Miller , V. and Moore , S . ( 2017 ), ‘ Prestige, Performance and Social Pressure in Viral Challenge Memes ’, Sociology , doi: 10.1177/0038038516680312. Candea , M . ( 2016 ), The Social After Gabriel Tarde , Routledge . Castells , M . ( 1996 ), The Information Age , Vol. I . Blackwell . Christakis , N. and Fowler , J . ( 2007 ), ‘ The Spread of Obesity in a Large Social Network Over 32 Years ’, New England Journal of Medicine , 357 : 370 – 79 . Google Scholar Crossref Search ADS PubMed Cornish , D. B. and Clarke , R. V. G . ( 1986 ), The Reasoning Criminal: Rational Choice Perspectives on Offending , Springer-Verlag . Corrado , R . ( 1981 ), ‘ A Critique of the Mental Disorder Perspective of Political Violence ’, International Journal of Law and Psychiatry , 4 : 1 – 47 . Google Scholar Crossref Search ADS PubMed Cottee , S . ( 2017 ), ‘ The Islamic State’s Shock-and-bore Terrorism ’, Foreign Policy , 27 July, available online at http://foreignpolicy.com/2017/07/27/islamic-state-shock-and-bore-terrorism-al-qaeda-9-11/ [Accessed: 05/12/2017]. Cottee , S. and Hayward , K. J . ( 2011 ), ‘ Terrorist (e)motives: The Existential Attractions of Terrorism ’, Studies in Conflict and Terrorism , 34 : 963 – 86 . Google Scholar Crossref Search ADS Crenshaw , M . ( 1981 ), ‘ The Causes of Terrorism ’, Comparative Politics , 13 : 379 – 99 . Google Scholar Crossref Search ADS David , A. B . ( 2016 ), ‘ Six Months of Terrorist Attacks by Lone Individuals: To Be Continued ’, Israeli Institute for National Security Studies , Insight No. 806. 20/03/2016, available online at http://www.inss.org.il/publication/six-months-of-terrorist-attacks-by-lone-individuals-to-be-continued/ [ Accessed: 06/12/2017 ]. Davis , M . ( 2007 ), Buda’s Wagon , Verso . Dawkins , R . ( 1976 ), The Selfish Gene , Oxford University Press . Eglash , R. Booth , W. and Cameron , D . ( 2016 ), ‘ “A New Kind of Terrorism” in Israel ’, The Washington Post , 20 September, available online at https://www.washingtonpost.com/graphics/world/israel-palestine-deaths/ [ Accessed: 08/09/2017 ]. Farrell , G. Tseloni , A. Mailley , J. and Tilley , N . ( 2011 ), ‘ The Crime Drop and the Security Hypothesis ’, Journal of Research in Crime and Delinquency , 48 : 147 – 75 . Google Scholar Crossref Search ADS FireArmsResQ ( 2017 ), ‘Trump protester gets hit by car 6 times. Darwin awards. Run over bill’ , 05 February, available online at https://www.youtube.com/watch?v=fkvHP6nKIcU. Freilich , J. D. and Newman , G. R . ( 2009 ), Reducing Terrorism Through Situational Crime Prevention , Criminal Justice Press . ‘GEKRONS’ ( 2017 ), ‘Protesters getting hit by cars’ , 13 May, available online at https://www.youtube.com/watch?v=ivUkN6cofK0. Gibbs , A . ( 2008 ), ‘ Panic! Affect Contagion, Mimesis and Suggestion in the Social Field ’, Cultural Studies Review , 14 : 130 . Gill , P . ( 2016 ), Lone-Actor Terrorists , Routledge . Gould , M. Jamieson , P. and Romer , D . ( 2003 ), ‘ Media Contagion and Suicide Among the Young ’, American Behavioral Scientist , 46 : 1269 – 84 . Google Scholar Crossref Search ADS Grabar , H . ( 2017 ), “ ‘Run Them Down’: Driving into Crowds of Protesters was a Right Wing Fantasy Long Before the Violence in Charlottesville ”, Slate , 14 August, available online at http://www.slate.com/articles/business/metropolis/2017/08/driving_into_crowds_of_protesters_was_a_right_wing_fantasy_long_before_charlottesville.html [ Accessed: 01/12/2017 ]. Guadagno , R. E. Rempala , D. M. Murphy , S. and Okdie , B. M . ( 2013 ), ‘ What Makes A Video Go Viral? An Analysis of Emotional Contagion and Internet Memes ’, Computers in Human Behavior , 29 : 2312 – 19 . Google Scholar Crossref Search ADS Hamm , M. and Spaaj , R . ( 2017 ), The Age of Lone Wolf Terrorism , Columbia University Press . Holden , R . ( 1986 ), ‘ The Contagiousness of Aircraft Hijacking ’, American Journal of Sociology , 91 : 874 – 904 . Google Scholar Crossref Search ADS Horgan , J . ( 2005 ), The Psychology of Terrorism , Routledge . Howard , D. J. , and Gengler , C . ( 2001 ), ‘ Emotional Contagion Effects on Product Attitudes ’, Journal of Consumer Research , 28 : 189 – 201 . Google Scholar Crossref Search ADS Ibrahim , Y . ( 2010 ), ‘The ultimate mowing machine’ Inspire , Autumn 2010 : 54 – 5 , available online at https://azelin.files.wordpress.com/2010/10/inspire-magazine-2.pdf [ Accessed: 12/06/2017 ]. Jenkins , B . ( 1981 ), Embassies Under Siege: A Review of 48 Embassy Takeovers , 1971 – 80 . No. RAND/R-2651-RC. RAND CORP SANTA MONICA CA . Jenkins , B . ( 2017 ), ‘ Vehicular Terrorism: Weighing the Benefits, and Worth, of Prevention ’, The RAND Blog , available online at www.RAND.org. KFVS ( 2016 ) ‘Driver plows through ‘Black Lives matter’ protesters in southern IL. KFVS television , 11 June, available online at http://www.kfvs12.com/story/32413138/driver-plows-through-black-lives-matter-protesters-in-southern-il [ Accessed 05/12/2017 ]. Kilcullen , D . ( 2009 ), The Accidental Guerrilla , C. Hurst and Co . Kinnunen , J . ( 1996 ), ‘ Gabriel Tarde as a Founding Father of Innovation Diffusion Research ’. Acta Sociologica , 39 : 431 – 42 . Google Scholar Crossref Search ADS Koerner , B . ( 2016 ), ‘Why is ISIS Winning the Social Media War?’ , April, available online at Wired.com. Kramer , A. D. Guillory , J. E. and Hancock , J. T . ( 2014 ), ‘ Experimental Evidence of Massive-scale Emotional Contagion Through Social Networks ’, Proceedings of the National Academy of Sciences , 111 : 8788 – 90 . Google Scholar Crossref Search ADS Latour , B . ( 2002 ), ‘ Gabriel Tarde and the End of the Social ’, in P. Joyce , ed., The Social in Question , Routledge . Lazo , L . ( 2017 ), ‘ Vehicles as Weapons of Terror: US Cities on Alert as Attacks Hit the West ’, Stars and Stripes , 09 July, available online at https://www.stripes.com/news/us/vehicles-as-weapons-of-terror-us-cities-on-alert-as-attacks-hit-the-west-1.477364 [ Accessed: 08/08/2017 ]. Leys , R . ( 1993 ), Mead’s Voices: Imitation as Foundation, or, the Struggle Against Mimesis . Critical Inquiry , 19 : 277 – 307 . Google Scholar Crossref Search ADS Lichfield , J . ( 2014 ), ‘ France Gripped by Fear at Christmas After Third Street Attack in Three Days ’, The Independent , 23 December, available online at http://www.independent.co.uk/news/world/europe/france-gripped-by-fear-at-christmas-after-third-street-attack-in-three-days-9943109.html [ Accessed: 12/10/2017 ]. Marsden , P . ( 1998 ), ‘ Memetics and Social Contagion: Two Sides of the Same Coin ’, Journal of Memetics-Evolutionary Models of Information Transmission , 2 : 171 – 85 . —— ( 2000 ), ‘ Forefathers of Memetics: Gabriel Tarde and the Laws of Imitation ’, Journal of Memetics – Evolutionary Models of Information Transmission , 4 . Available online at http://cfpm.org/jom-emit/2000/vol4/marsden_p.html. Marrero-Guillamon , I . ( 2013 ), ‘ Actor-Network Theory, Gabriel Tarde and the Study of an Urban Social Movement: The Case of Can Ricart, Barcelona ’, Qualitative Sociology , 36 ( 4 ): 403 – 421 . Google Scholar Crossref Search ADS Massumi , B . ( 1995 )‘ The Autonomy of Affect ’, Cultural Critique , 31 : 83 – 109 . Google Scholar Crossref Search ADS Mccauley , C. and Moskalenko , S . ( 2011 ), Friction: How Radicalization Happens to them and us , Oxford University Press . Midlarsky , M. I. Crenshaw , M. and Yoshida , F . ( 1980 ), ‘ Why Violence Spreads: The Contagion of International Terrorism ’, International Studies Quarterly , 24 : 262 – 298 . Google Scholar Crossref Search ADS Moore , J . ( 2017 ), ‘ After London, Nice and Berlin Vehicle Attacks, What Can Cities Do to Prevent More ?’ Newsweek , 23 March, available online at http://www.newsweek.com/after-london-nice-and-berlin-what-can-cities-do-prevent-car-ramming-attacks-572692 [ Accessed: 12/08/2017 ]. Mortimer , C . ( 2017 ), Darren Osborne: Family of Man Arrested After Finsbury Park Mosque Terror Attack Says He is ‘Troubled’ But ‘Not Racist’ , The Independent.co.uk, June 19th, available online at https://www.independent.co.uk/news/uk/crime/finsbury-park-mosque-terror-attack-muslims-darren-osborne-van-driver-family-neighbour-troubled-a7798256.html. Nashashibi , S . ( 2015 ), ‘ Israel’s Cruelty has Unleashed an Intifada of Individuals ’, The National (UAE Edition) , 31 October, available online at https://www.thenational.ae/opinion/israel-s-cruelty-has-unleashed-an-intifada-of-individuals-1.67173 [ Accessed: 07/08/2017 ]. Newman , G. and Clarke , R . ( 2008 ), Policing Terrorism: An Executive’s Guide , Office of Community Oriented Policing Services . Neumann , R. and Strack , F . ( 2000 ), ‘ “Mood Contagion”: The Automatic Transfer of Mood Between Persons ’, Journal of Personality and Social Psychology , 79 ( 2 ): 211 – 223 . Google Scholar Crossref Search ADS PubMed Orwell , G . ( 1945 ), You and the Atomic Bomb , Tribune , October 19, 1945 . Ostrovsky , A . ( 2015 ), ‘ Why is The World Ignoring a Wave of Terror in Israel ?’, The Telegraph , 09 October, available online at http://www.telegraph.co.uk/news/worldnews/middleeast/israel/11921994/Why-is-the-world-ignoring-a-wave-of-terror-in-Israel.html [ Accessed: 09/09/2017 ]. Patten , S. B. and Arboleda-Florez , J . ( 2004 ), ‘ Epidemic Theory and Group Violence ’, Social Psychiatry and Psychiatric Epidemiology , 39 : 853 – 56 . Google Scholar Crossref Search ADS PubMed Perry , S. Hasisi , B. and Perry , G . ( 2017 ), ‘ Who is the Lone Terrorist? A Study of Vehicle-Borne Attackers in Israel and the West Bank ’, Studies in Conflict and Terrorism , 40: 1–15. doi: https://doi.org/10.1080/1057610X.2017.1348101 Pfeffer , A . ( 2016 ), ‘ This is the Age of the Individual’s Intifada ’, The Jewish Chronicle , 16 June, available online at https://www.thejc.com/comment/analysis/this-is-the-age-of-the-individual-s-intifada-1.58851. Picard , R. G . ( 1986 ), ‘ News Coverage as the Contagion of Terrorism: Dangerous Charges Backed by Dubious Science ’, Media Coverage of Terrorism , Sage . Post , J. M . ( 2007 ), The Mind of the Terrorist , Palgrave-Macmillan . ‘RANDY’ ( 2017 ), ‘Anti-Trump protesters hit by cars compilation NEW CLIPS’ , 9 August, available online at https://www.youtube.com/watch?v=idjGM6l9wEA. Rapoport , D. C . ( 2001 ), ‘ The Fourth Wave: September 11 in the History of Terrorism ’, Current History , 100 : 419 . Sampson , T. D . ( 2012 ), Virality . University of Minnesota Press . Schmid , H. B . ( 2004 ), ‘ Evolution by Imitation: Gabriel Tarde and the Limits of Memetics ’, Distinktion: Scandinavian Journal of Social Theory , 5 ( 2 ): 103 – 118 . Google Scholar Crossref Search ADS Schuurman , B. Lindekilde , L. Malthaner , S. O’connor , F. Gill , P and Bouhana , N . ( 2018 ), ‘ End of the Lone Wolf: The Typology that Should not have been ’, Studies in Conflict & Terrorism , doi: https://doi.org/10.1080/1057610X.2017.1419554 Sedgwick , M . ( 2010 ), ‘ The Concept of Radicalization as a Source of Confusion ’, Terrorism and Political Violence , 22 : 479 – 94 . Google Scholar Crossref Search ADS Shifman , L . ( 2014 ), Memes in Digital Culture , MIT . Silke , A . ( 1998 ), ‘ Cheshire-cat Logic: The Recurring Theme of Terrorist Abnormality in Psychological Research ’, Psychology, Crime, and Law , 4 : 51 – 69 . Google Scholar Crossref Search ADS Simon J. D . ( 2011 ), ‘ Technology and Lone Operator Terrorism: Prospects for a Fifth Wave of Global Terrorism ’, in J. Rosenfeld , ed., Terrorism, Identity and Legacy , Routledge . Tarde , G . ( 1901 ), L’Opinion et la foule , Félix Alcan . —— ( 1903 ), The Laws of Imitation , Henry, Holt and Co . —— ( 2010/1888 ), Gabriel Tarde on Communication and Social Influence , University of Chicago Press . —— ( 2012/1895 ), Monadology and Sociology , Re Press . Towers , S. Gomez-Lievano , A. Khan , M. Mubayi , A. and Castillo-Chavez , C . ( 2015 ), ‘ Contagion in Mass Killings and School Shootings ’, PLoS one , 10 : e0117259 . Google Scholar Crossref Search ADS PubMed Ward , V. , Evans , M. and Furness , H . ( 2017 ), ‘Who is Darren Osbourne? Everything We Know About the Finsbury Park Mosque Suspect’ , The Telegraph, 28 July, available online at: http://www.telegraph.co.uk/news/0/darren-osborne-everything-know-finsbury-park-mosque-suspect/ [Accessed on 07/06/2017 ]. Wiggins , B. E. and Bowers , G. B . ( 2015 ), ‘ Memes as Genre: A Structurational Analysis of the Memescape ’, New Media and Society , 17 : 1886 – 1906 . Google Scholar Crossref Search ADS Footnotes 1 One person was killed and 20 injured in the Osbourne attack. 2 On 3 June 2017, three Islamic State supporters drove a rented van into pedestrians on London Bridge. After their van was disabled, the men made their way to a nearby market and carried out a frenzied knife attack that left eight dead and 48 injured. 3 In the days leading up to the Mosque incident, Osbourne had allegedly ‘hurled insults at his Asian neighbour’s 12-year-old son’, and the night prior to the attack he was thrown out of his local pub for making threats against Muslims. 4 This tendency to focus not on the ‘why question’ but on the ‘how question’ is now very marked within terrorism studies. In particular, the issue of process (as opposed to a specific personality trait or biographical ‘turning point’) is seen as especially important when it comes to ‘radicalization’ and ‘de-radicalization’; discourses that, as Sedgwick (2010) has pointed out, emerged as a sort of ‘holy grail of national security research’ in the aftermath of the July 7/7 London bombings. 5 Terrorist casualties in Western countries in 2016: vehicular: 99 killed, 502 injured; bombing: 32 killed, 330 injured; shooting: 51 killed, 61 injured; stabbing: 7 killed, 27 injured (Lazo 2017). 6 The term ‘lone wolf’ is a subject of debate (Schuurman et al. 2018). However, we adopt the position of Hamm and Spaaj (2017:14) who define lone wolf terrorism as ‘terrorist actions carried out by lone individuals, as opposed to those carried out on the part of terrorist organizations or state bodies’. Accordingly, ‘The lone wolf is solitary by nature… although his or her radicalization to action may be spurred by violent images, incendiary books, manifestos, and fatwas’ (ibid:5). 7 Detailed records of incidents are not provided before 1999, and the database does not include 2017. In addition, the database only includes incidents that have been judged to be ‘terrorism’. For our purposes, we needed to also include VRAs involving other motivations, such as mental illness, etc. 8 Aside from the fact that most of the attackers were under 25 years of age, Perry et al (2017) concluded that, because of the diversity of the sample, a profile of lone-actor terrorists in Jerusalem/the West Bank was not possible. For example, only 28.5% had prior criminal convictions, and only 12.9% were said to be suffering from psychological or mental health issues. 9 The author also suggested welding steel blades to the front of a pickup truck to cause further damage to pedestrians on the streets of Western cities, adding later: ‘This is one of many ways to implement this idea. You may modify it and add or subtract it accordingly to what is suitable to your particular conditions… Tell the world why you did it…’ (Ibrahim 2010:54). 10 Our sample reveals a sporadic history of ramming attacks dating back to 1999 (the last year of reliable records from the GTD), which includes VRAs by both Palestinians and Israeli Settlers as early as 2001. 11 For example, on 28 November 2016, Abdul Razak Ali Artan rammed his vehicle into a group of students and then stabbed bystanders on the Ohio State University campus in Columbus, Ohio. At least 11 people were injured before Artan was shot and killed by a nearby police officer. ISIL subsequently claimed responsibility for the incident. 12 The driver of the white 19-tonne Renault Midlum truck used in Nice was Mohamed Lahouaiej- Bouhlel, a 31-year-old man of Tunisian origin. Once again, the Islamic State would claim responsibility. 13 Here, it is important to consider the ‘caliphate’ effect. After the United States started bombing ISIS positions in September 2014, Abu Mohammed al-Adnani called (in numerous pronouncements) for attacks on the West, famously declaring: ‘If you kill a disbelieving American or European… or any other disbeliever from the disbelievers waging war, including the citizens of the countries that entered into a coalition against the Islamic State, then rely upon Allah, and kill him in any manner or way, however it may be… Smash his head with a rock, or slaughter him with a knife, or run him over with your car, or throw him down from a high place, or choke him or poison him’ (our emphasis). It goes without saying that this edict, emanating from a prominent figure in a (then) nascent nation state, was an important hinge-moment for many jihadists post-2014. The perpetrator of the July 2016 Nice VRA, Mohamed Lahouaiej-Bouhlel, e.g., was very likely influenced by al-Adnani’s brutishly practical edict. However, what is important here (at least temporally) is how the dramatic fallout of the Nice attack cascaded through the global mediascape, amplifying the magic of this mode of attack. The Rumiyah articles should thus be seen as part of this radiation effect rather than the cause. 14 Such claims must be tempered by the Achilles heel of CPtED: displacement; a point born out in a recent commentary on VRAs by terrorist scholar Brian Jenkins (2017). After outlining ‘Ten potential mitigation measures’ to combat motorized assaults he concludes: ‘it is not clear that any of the suggested potential security measures would prevent a determined terrorist behind the wheel from driving a bit further to get roughly the same results’. 15 Recent years have seen a significant revival of interest in contagion theory. See, e.g., Gould et al (2003) and Marsden (1998) on suicide; Christakis and Fowler (2007) on obesity; Patten and Arboleda-Florez (2004) on violent group behaviour; Howard and Gengler (2001) on the ‘emotional contagions’ of mood and aggression. 16 ‘Memes’ are the socio-cultural equivalent of the biological/evolutionary concept of the gene (Dawkins 1976). The term is used to describe cultural messages which, through the creative and participatory nature of Web 2.0, proliferate across social media, while continually being adapted, parodied and reworked (Shifman 2014; Wiggins and Bowers 2015). 17 Such thinking chimes with recent Deleuzian-inspired work on affect/emotion in sociology and the humanities (Massumi 1995; Blackman 2007; Blackmore 2000). 18 The Charlottesville ramming was preceded by at least three ramming incidents targeting BLM protestors: 6 January 2015, in South Minneapolis, MN; 8 July 2016 outside the Ferguson police station, St Louis, MO; 11 July 2016 in Carbondale, IL. In this last incident, the driver stopped in front of the protesters, emerged from his vehicle and shouted ‘All lives matter, not blacks, all lives’, before returning to his vehicle and driving into the protesters (KFVS 2016). 19 Some examples include: FireArmsResQ (2017); Randy (2017); Gekrons (2017). 20 The anarchist wave (late nineteenth and early twentieth century); the anti-colonialist wave (1920s–1960s); the new-left wave (1960s–1980); and the religious wave (1979–present). © The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). All rights reserved. For permissions, please e-mail: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Titeca,, Kristof

doi: 10.1093/bjc/azy009pmid: N/A

Abstract This article examines illegal ivory trade in Uganda, which constitutes a major transport route through which ivory exits Africa. The analysis is based on empirical data collected among illegal ivory traders between 2012 and 2017. The findings unpack the notion of illegal ivory trade as ‘transnational organized crime’, by showing its reliance on local and regional connections, in which ‘nodes’ are crucial. These nodes can be both traders (such as middlemen), and locations (such as border towns), connecting these various levels. In doing so, it shows how this trade functions in a decentralized and loose fashion. There are clear power differences between the traders, which is explained through the kind of connections with government officials. Introduction From 2007 onwards, illegal ivory trade became a global and urgent problem, largely fuelled by increasing demand and consumption in East and South East Asia (Milliken et al. 2016a: 11, 24; Underwood et al. 2013: 6). Particularly since 2010, the illicit ivory trade has been at record high levels (Milliken et al. 2016b: 3). Similar trends can be seen when looking at other indicators, such as poaching and ivory prices: there have been increasing levels of poaching since 2006, with a peak in 2011 (UNEP et al. 2013: 11, 33). It is estimated that between 2010 and 2012, 100,000 elephants were killed in Africa (Wittmeyer et al. 2014). How is this booming trade organized? Academic, policy- and media-reports dominantly explain this trade through the involvement of transnational crime—reports refer, respectively, to ‘ballooning criminal networks that allow this transnational crime to operate’ (Wasser et al. 2015: 87), ‘international crime syndicates’ (EIA 2014: 9, 15) or the ‘undeniable’ fact that ‘commercial poaching and trafficking are run by multi-national criminal gangs’ (McCann 2017). These results strongly resonate with international policy: new CITES regulations—the UN Convention on International Trade in Endangered Species—now consider poaching and wildlife trafficking as a form of transnational organized crime (Biegus and Bueger 2017: 29). These statements are largely based on confiscation data, which feature prominently in CITES reports and other policy studies, academic studies and press articles. Concretely, these data show how large raw ivory shipments of over 100 kg are the driving force behind the increased illegal ivory trade (Milliken et al. 2016a: 8; Milliken et al. 2016b). For example, in 2009, 2010 and 2011, there, respectively, were 8, 9 and 17 seizures over 800 kg, for a total of almost 61 tonnes of ivory (UNEP et al. 2013: 46). These kind of large ivory seizures are unprecedented. Based on these data, analyses are quasi-unanimous: seizures of large movement of ‘ivory represent(s) the involvement of international criminal syndicates in the trade operating through sophisticated networks that link Africa with Asia’ (UNEP et al. 2013: 46). In other words, international organized crime, and existing networks in the trade in drugs, arms and humans, have come to infiltrate and dominate the ivory trade (Clarke and Babic 2016: 57). Other research, based on DNA analysis of confiscated ivory, presents similar results highlighting how ‘crime syndicates were targeting specific populations for intense exploitation, hitting them hard and fast to satisfy the demands of a particular consignment’ (Wasser et al. 2008: 1070). Yet, how is this transnational organized crime organized? The above dominant statements on ‘ballooning criminal networks’, ‘crime syndicates’ and ‘gangs’ remain only that: statements, without an explanation or empirical analysis on what these look like or how this trade is organized. A major gap in the academic literature is therefore in-depth research on the organization of illegal ivory trade, particularly on the recently increased trade within Africa: little is known on how illegal ivory is traded through the continent, i.e. up to the moment it exits Africa. This gap is acknowledged and highlighted in the policy literature. The UN office on Drugs and Crime (UNODC 2013: 48), for example, argues how ‘A major research gap is the groups, routes and methodology used to move ivory within Africa, between the poaching sites and the ports of exit’. Milliken (2014: 13) argues for more understanding of the structure of the ivory trade, and in particular understanding on how ‘criminal operations are collecting, consolidating and shipping ivory between regions’, while other studies specifically argue to target traders (Somerville 2016). This article engages with these research gaps: by analysing the organization of illegal ivory trade through Uganda, it aims to better understand the role of transnational organized crime in the illegal ivory trade. In doing so, and in line with Runhovde (2017; 2), this article aims to deconstruct the ‘over-homogenized discourse surrounding transnational organized crime’, in order to actually ‘understand the complexity of the trade and to respond to it efficiently, more precision about local actors, activities and context is needed’. Uganda is chosen as a case study. Since 2007, Uganda has increasingly established itself as a major transit point for ivory. CITES (2016) calls Uganda a country of ‘primary concern’ in the illicit ivory trade: it is listed as one of the ten countries worldwide ‘linked to the greatest illegal ivory trade flows since 2012’ (CITES 2016: 25) and serves as ‘an important entrepôt/export centre in East Africa with clear links to Central African ivory trade flows’ (CITES 2016: 15). Methodology Crucial in understanding this organization was in-depth research among illegal ivory traders in Uganda (and, to a lesser extent, in the Democratic Republic of Congo [DRC]), with whom were semi-structured interviews conducted between 2012 and 2017. First, this research was part of a long-term research engagement with a group of illegal cross-border traders in the Uganda–DRC border region, with whom I have been interacting since 2005 (e.g. Titeca 2006). This research initially paid no attention to illegal ivory trade, until this started becoming more and more present in their trading activities. As will be shown in the text, from around 2009 onwards, ivory was becoming an increasingly important commodity in the Uganda–DRC borderlands and in the traders’ activities. Second, between 2012 and 2017, specific interviews with ivory traders were conducted on two specific locations, which are key-nodes in the illegal ivory trade: the Ugandan–DRC borderlands (in particular the border towns Arua in Uganda and Aru in DRC) and the Ugandan capital Kampala. More specifically, 11 ivory traders were followed throughout these years. Trust was crucial in this process (and more generally in conducting research on illegal trade, Ellis and MacGaffey 1996). Given that I have been conducting research on illegal trade since 2005, this meant that I had built up a reputation as being ‘reliable’—particularly among traders in the Uganda–DRC borderlands. This facilitated introduction to illegal ivory traders, as well as the interviews themselves. In addition, interviews were conducted with law enforcement officials, journalists, NGOs and analysts who are working on this issue. Seventy-six interviews were conducted in total, and forty-nine interviews were conducted with 11 ivory traders. At the time of writing (2017–18), this is the first study primarily based on empirical research with illegal ivory traders. The article is structured in the following way: the next section discusses the literature on transnational (dis)organized crime and how these findings apply to the illegal wildlife trade, in particular the ivory trade. The third section discusses the activities of the ivory traders in Uganda and the power differentials among these traders. The fourth section concludes. Transnational (dis)organized crime and illegal wildlife trade (Dis)organized crime Definitions of organized crime differ considerably in their conceptualizations and are often vague or overly broad (Finckenauer 2005: 68). Central to understanding organized crime is the extent to which it is ‘organized’ (Finckenauer 2005: 60; Hagan 2006: 133) and the different structures through which it organizes itself. Generally, there is a consensus that organisations with well-structured hierarchies—the ‘mafia model’—have become an exception and have been replaced by crime which is much more ‘disorganized’ (Reuter 1985) consisting of ‘loosely affiliated networks of criminals who coalesce around certain criminal opportunities’ and which is ‘much more amorphous, free floating, and flatter, and thus lacking in a rigid hierarchy’ (Finckenauer 2005: 65–6). For example, the literature on organized crime in Europe shows how large hierarchical structures are no longer common, but how these groups are instead characterized by loose affiliations and less obvious chains of command, such as flexible and changeable networks, or small and ephemeral enterprises (Paoli and Fijnaut 2004: 607–10). Why is this the case? Similar to legal, legitimate businesses, the kind of structure utilized in organized crime is influenced by contextual factors such as the dynamic character of the environment or the competitors in this particular market (Southerland and Potter 1993). For example, the organization of human smuggling and heroin trafficking is influenced by market conditions and operational requirements (Zhang and Chin, 2003: 478). They are organized into ‘small-sized groups’ that are able to ‘adapt well to ever-changing market constraints and uncertainties. They have little vertical bureaucratic structure, even though their operations may involve highly specialized tasks. They ally with only those with valuable resources to contribute to their objective of making a profit’ (Zhang and Chin 2003: 478). The latter point (i.e. the objective of making a profit) illustrates how the structure of organized crime is also related to the kind of criminal activities these actors are engaged in: organizations with a primary interest in rule enforcement and dispute settlement will have a different structure from organizations who primarily want to make money. The latter is flexible and fluid by nature, rather than vertically and horizontally integrated organizations. Instead, they ‘flexibly cooperate in changing combinations of individuals and small groups within larger transnational networks of criminally exploitable ties’ (Von Lampe 2014: 86). Moreover, the high demands on communication, and the risks associated with it, constitute an important downside of strongly hierarchical organizations (Zaitch 2002). This has been shown in other fields, such as drug trafficking: research shows how this is primarily organized by ‘small groups of loosely linked entrepreneurs rather than large, highly structured criminal syndicates’ (Zaitch 2002; Natarajan 2006: 171), allowing these loose networks to ‘quickly react to shifting market conditions’ (Bichler et al. 2017: 2). How does this apply to environmental crime and more specifically illegal wildlife trade? This is discussed in the next section. Illegal wildlife trade The term ‘transnational environmental crime’ is used to study the transnational character of environmental crimes. Transnational environmental crime is mostly defined as a crime with cross-border transference and an international or global dimension. Specifically, these are crimes related to pollution or crimes against wildlife, such as illegal wildlife trade (White 2011). The value chain of the latter category is very often transnational, involving, for example, supply from Africa (such as ivory) and demand from other parts in the world (such as Asia or Europe). Much of the findings on the organization of illegal wildlife trade, therefore, analyse its transnational character. The concepts of ‘transnational crime’ and ‘organized crime’, to a large degree, overlap, but are not always the same. Organized crime that does not cross a border is an example of this (White 2011). This section aims to understand the organization of a specific transnational environmental crime, illegal wildlife trade. More specifically, it aims to understand the extent to which the above insights into organized (transnational) crime apply to illegal wildlife trade. A first strand of research on illegal wildlife trade shows the strong involvement of ‘traditional’ organized crime, i.e. the interest of vertically organized ‘mafia-style groups’ and ‘major international smuggling rings’ (Zimmerman 2003: 1668). This has also been shown for transnational ivory trade. A research report (Vira and Ewing 2014: 9), for example, argues how a ‘global criminal enterprise’ is controlling ivory trade from poaching up to the final consumer. Subsistence and artisanal poaching has been ‘co-opted or crowded out by an illicit commercial trade that is monopolized by organized crime’ (Vira and Ewing 2014: 10). Milliken (2014: 23, 18) argues how ‘transnational criminal organized organizations’, which are ‘typically led by African-based Asian nationals, are directly involved in the procurement and illegal movement of rhino horn out of Africa to markets in Asia’. Bergenas and Knight (2015: 124) describe ivory poaching as ‘methodic and militarized’, while ivory trade is considered to be controlled by sophisticated networks of ‘Asian-based criminal syndicates’. Specifically on Tanzania, a research report argues how ivory trade is dominated by ‘international criminal syndicates’ and ‘sophisticated criminal networks’ (EIA 2014: 9, 15). A second stream of research shows the dominance of loosely organized networks instead of ‘traditional’ vertically organized syndicates. It is shown how most environmental crime is committed by ‘loosely organized networks of individuals’ (Hayman and Brack 2011: 7), or small groups organized for a short period of time (Wright 2011: 338). Wyatt (2013: 87–9) shows how illegal wildlife traders are not part of hierarchical networks, but operate more in informal networks, while Van Uhm (2016b) shows how the illegal trade in Barbara Macaques is based on semi-structured, flexible, yet sophisticated trading networks. A third stream argues that the focus on organized crime is misguided and is ‘not the dominant picture as some journalists and researchers would like to believe’ (Pires and Moreto 2011: 105). Rather than organized crime as such, it is the participation of corrupt officials that fuels this trade. Naylor (2004: 263) calls the involvement of organized crime a ‘recurrent fable’ and a ‘tale’, in reality run by ‘complicit politicians and corrupt functionaries’. However, empirical data on this are rare and thin. Hübschle (2016) is helpful: she argues how the label of ‘transnational organized crime’ is of limited use, as the illegal wildlife trade is driven by ‘insider trading’, i.e. corruption and collusion between state officials and wildlife traffickers: it is a ‘business enterprise facilitated by a multitude of diverse actors with close, limited or no links to “organized crime”’ (Hübschle 2016: 195). Runhovde (2017) looks at Uganda. Based on interviews with law enforcement officers, she does not find indications of organized crime in the illegal ivory trade, but of a range of individuals with various organizational capacities. However, little analysis or data are presented on these individuals. Moreover, it is unclear to which extent this conclusion is the outcome of methodological bias, rather than an analysis of the illegal trade: the study is based on interviews with law enforcement officers and not on illegal ivory traders.1 Yet, are all these viewpoints mutually exclusive? Van Uhm (2016a), for example, shows how the degree of organization and structure of the illegal wildlife trade widely varies, ranging from opportunistic poachers and traders to highly organized poaching and trading networks (Van Uhm 2016a: 263). Others argue how ‘depending on the species and region of the world, “organized” can simply mean anything from three individuals who are loosely organized together to a vast criminal enterprise that compromises all stages of the wildlife trade (i.e. vertically integrated organizations)’ (Pires and Moreto 2011: 104). A central element is that illegal wildlife involves different steps: a multi-level chain typically involving poachers, middlemen and markets (Pires and Moreto 2011: 104). It is perfectly possible that each step in the chain has different actors who are involved with different levels of organization. For example, Pires and Clarke (2011: 315) argue that while poaching might be a largely opportunistic activity with little links with organized crime, this becomes different after the initial act of poaching, when ‘greater organization becomes apparent in subsequent stages that involve a variety of middlemen, processing centres and markets’. Specific research on ivory trade suggest similar dynamics, in which local actors are responsible up to the point when ‘ivory is consolidated and hidden inside a container, while Asian and other organized crime groups control the supply chain from containerization all the way through the shipping and transport systems to market’ (Vira and Ewing 2014: 14). The relationship between ‘local’ and ‘transnational’ trade helps to further unpack this differentation, which is discussed in the next section. Connectivity, nodes and transnational trade Much research on organized crime has shown how this no longer fixed to one location or limited to having a strictly regional character, but instead has ‘evolved into an international or transnational character complete with regional and global alliances’ (Hill 2005: 47). Transnational crime can be defined as ‘the illicit procurement, transportation, and distribution of commodities across international borders’ (Warchol et al. 2003: 1). In other words, organized crime has expanded its scope of operation on a global level, partnering with actors located in other countries or regions to gain access to new markets. The discussion on the organization of transnational organized crime largely mirrors the debate on organized crime. While some reference is made towards the involvement of ‘traditional’ mob-style organizations (UNODC 2002), most of the literature emphasizes the loose character of transnational organized crime: at the heart of transnational crime are temporary arrangements of ever-changing actors, in which particular groups or individuals act as hubs, allowing social networks organizing crime allowing to operate in ‘infinite variety of deviant collaborations’ (Hobbs and Dunnighan 1998: 299, 296). Transnational crime is therefore organized through ‘constant mediations and re-negotiations that are duplicated as perpetually mutating social systems’ (Hobbs 1998: 417; Hobbs and Dunnighan 1998: 298) It is the connectedness, created through the degree of connectivity between groups and individuals, which constitutes the structural in ‘organized crime’, operating on a transnational scale. In other words, connectivity is key in this debate: it are particular connections between different levels of crime that make a particular crime transnational. Writing on the Mediterranean, Horden and Purcell (2000) use connectivity to describe the ways in which micro-regions cohere, ‘both internally and also one with another—in aggregates that may range in size from small clusters to something approaching the entire Mediterranean’ (Horden and Purcell 2000: 123). The Mediterranean can, therefore, be considered as a zone of ‘intense topographical fragmentation, overlaid by a kaleidoscope of human micro-ecologies, which are in turn densely interconnected’ (Horden 2012: 28). In this way, connectivity refers to the way in which micro-regions are both internally connected and externally related to each other. The literature on inter- and trans-Saharan trade is specifically useful in understanding these connections. When discussing Saharan trade, the focus has particularly been on transnational trade, with little attention to the importance of intra-regional and local trade in its history. Inter-Saharan trade has always been considered less important than trans-Saharan trade: the Sahara itself was considered a ‘blank, ready to be washed by cultural “tides” from elsewhere’ (Horden 2012: 26). However, as particularly the work of Judith Scheele has shown, the dominance of long-distance trade in the Sahara is an ‘illusion’, but depends on local exchange and inter-Saharan trade. (Scheele 2010: 282; Scheele 2011: 145). Concretely, transnational trade largely relies on regional and local infrastructure, for access to transport, manpower and networks of protection. Trans-Saharan trade external to the Sahara was therefore rare, as it remained dependent on local alliances and pre-established social networks, in which regional traders, for example, act as intermediaries (Scheele 2010: 298; Scheele 2011: 145, 160). Transnational and local trade are therefore very much embedded in each other: the difference between these categories is one of ‘scale rather than of kind, as the one is already and necessarily contained in the other: places appear as nodes of particular density in overlapping networks of connectivity’ (Scheele 2010: 283). Saharan trade should in this context be considered as a ‘network of interdependent sub-systems, of short-, medium- and long-distance exchange’, of which ‘trans-Saharan traffic was only one part of this network’ (Wilson 2012: 409). The local, regional and transnational levels are in this way closely connected with each other. Nodes play a central role in these connections: they are the points in which these various scales come together and in which connections take place. In nodes, the ‘local’ has to be understood with references to wider scales: the local and/or regional is connected with the transnational, and vice-versa nodes often act as entrepôt points: points were goods are stored, sold, traders meet, break up and reform, connecting local, regional and transnational economies (Wilson 2012: 414). Border towns are a good example of nodes. Connections and nodes do not only manifest itself in localities, but also through persons: it particularly are middlemen who play a central role in connecting various networks and scales. They are the ‘critical nodes that intermediate flows in a directed networks’ (Sims and Gilles 2016). Similar to border towns, middlemen are therefore able to connect different scales: the local, regional and transnational trade. This article builds further on these insights: in order to understand transnational crime, it argues how a number of issues are crucial. First, the article argues how, similar to the trans-Saharan trade, the understanding of transnational crime (and in particular transnational illegal ivory trade) needs an understanding of local and regional illegal ivory trade: transnational ivory trade relies on regional trade. Second, connections are crucial in linking these various levels. This principally happens in nodes, and in particular by traders acting as middlemen: it are these traders that connect the various scales of this illegal trade. However, the article adds a third and crucial element: it argues how the power of these connections is crucial. In order to do so, specific attention needs to be given to the leverage of the traders’ various connections. For example, it will be shown some traders limit their activities to the ‘micro-region’ of the Uganda–DRC borderlands, whereas other trades operating from Kampala are able to work directly with foreign exporters and are able to charge higher prices. They are able to do so because of their more powerful connections with state operatives. In other words, a better understanding of the connections in place allows a better understanding of the nature of transnational crime: these connections not only integrate different scales, but also allow to understand the various degrees of power. Ivory traders in Uganda Individual ivory traders are the central actors in the illegal ivory trade in Uganda. They are the ones smuggling the goods across borders, looking for a market or for supply. In this article, the term ‘trader’, rather than the terms ‘middlemen’, ‘broker’, ‘trafficker’ or ‘smuggler’, will be used, as all traders involved perform a variety of functions: on occasions, they smuggle goods across the borders; on other occasions, they arrange physical transport within a particular country (i.e. not crossing borders or smuggling). On certain occasions, they simply link up people and do not touch the ivory; on other occasions, they are only responsible for the storage of ivory. Sometimes, they buy ivory themselves and look for buyers; on other occasions, they are simply taking a commission on ivory delivered from one actor to another. This section follows the line of argument(s) as outlined earlier: first, the article shows the importance of regional trade for transnational illegal ivory trade. Second, it shows how traders and their connections are central. Third, and importantly, it shows the various degrees of power of the traders’ connections, which is particularly related to access to government officials. In order to do so, the article looks in detail at how these connections develop themselves in key-nodes of the trade: we start by looking at the Ugandan border town of Arua, which is a central entry-point for the ivory trade, and then look at the Ugandan capital Kampala, which is a central exit-point out of the country. The Ugandan border town of Arua as a nodal point The first nodal point this article focuses on is the Ugandan border town of Arua, bordering the DRC and South Sudan. This area is a key nodal point in the regional trade: much ivory originating from North-Eastern DRC, South Sudan or the Central African Republic (CAR) passes through this border town and its twin (border) town Aru on the Congolese side of the border. In the next subsections, the ivory supply, smuggling and sale of the local traders are discussed (Figure 1). Fig. 1 View largeDownload slide Uganda and Arua. Source: Map reprinted from Titeca and De Herdt (2011). Ref. ‘Figure 1 Map of Uganda, showing the border regions’. Fig. 1 View largeDownload slide Uganda and Arua. Source: Map reprinted from Titeca and De Herdt (2011). Ref. ‘Figure 1 Map of Uganda, showing the border regions’. Ivory supply Historically, the border town Arua has been a regional smuggle centre, facilitated by the presence of three borders (Uganda, DRC, South Sudan), weak border infrastructure, close connections between ethnic groups on various sides of the border, refugee streams and different tax regimes (Meagher 1990; Titeca 2009). Starting in the late 70s and early 80s, this trade has led to the establishment of a range of what are locally called ‘tycoons’ or ‘Arua boys’ in Uganda: Ugandan traders active in the illegal cross-border trade, dealing in a wide portfolio of illegal goods such as petrol, minerals, sugar, clothes, batteries and so on (Titeca 2012). Historically, traders in the border town have acted as middlemen, largely between the transnational, regional and local levels: goods such as petrol and cigarettes were smuggled in from neighbouring countries, such as the DRC or South Sudan, and sold to local traders at a cheaper price. Minerals were also sold to international traders, who came to the border town. Ivory was present as a commodity, but rare: very few Ugandan traders had access to this commodity and were trading in it. This changed from around 2009–10 onwards, when large quantities of ivory reached the area. This ivory came from all neighbouring countries in the region: South Sudan, Central African Republic, but particularly from the DRC. During those years, the Garamba National Park—which is near the Ugandan and South Sudanese borders—became increasingly militarized and led to increased levels of poaching. While this park traditionally has been a site of poaching, the impact on the ivory trade and elephant levels remained limited (Titeca 2013). This changed in the late 2000s (particularly 2008–10) when a whole range of armed actors started poaching elephants in the park: the rebel group, the Lord’s Resistance Army (LRA), South Sudanese and Sudanese poachers, individual Congolese soldiers and individual Ugandan soldiers (as shown below). Suddenly, the border region had a major supply of ivory. As a Ugandan trader summarized: ‘before 2008, I didn’t know of ivory. I was trading in other things across the border. But from this time onwards, there was plenty of it’.2 A variety of actors were supplying this ivory: first, Congolese traders, who buy ivory from Congolese poachers, mostly from in or around Garamba National Park. They bring this ivory to various locations on the border, but particularly to one of the two border towns (mostly on the Congolese side), where they sell it to the Ugandan traders. In doing so, this supply was largely relying on this historical trading network. Second, some of this ivory originating from Garamba (which is poached by South Sudanese or Congolese poachers) first passes through South Sudan, in which South Sudanese traders are the ones offering the ivory to the Ugandan traders. A third possibility is a collaboration between Ugandan truck drivers and Ugandan soldiers. Since late 2008, the Ugandan army was involved in a military operation against the LRA on this territory, first in the DRC, but also in South Sudan, and later on in CAR. The Ugandan military have a history of smuggling natural resources from neighbouring countries, in particular Congo (Titeca 2011). The anti-LRA operations did not lead to the structural involvement of the Ugandan army, but individual Ugandan soldiers tried to profit from their exposure to natural resources, and particularly ivory. Civilian trucks, supplying the military with, for example, food or medicines, played an important role in this: after having supplied the military with these goods, the truck drivers in collaboration with individual soldiers, return with other goods such as timber or ivory, which were smuggled across the border.3 As one of these truck drivers summarized: We were civilian trucks hired by the military. What we did: we hid the ivory in the trucks. We were not being controlled, because we were working for the military. (…) The number of kilograms varied: sometimes 100 kilograms, sometimes less. We didn’t always find ivory. Sometimes we did find, sometimes we did not. (…) It was up to 2011 we went to Garamba. (…) How ivory was bought: soldiers collaborated with locals, who were selling it. And these locals delivered it to their—how do you call this?—their barracks there.4 Not all traders were able to collaborate with the military: instead, they smuggled the ivory themselves across the border. How is it smuggled across the border? Historically, the illegal trading networks of the border region have been smuggling goods across the border in two main ways: first, through smuggling routes that avoid border posts altogether. In this ‘ant-trade’, these goods are smuggled in small quantities, largely on bicycles, and then stored until a sufficiently large quantity is achieved. For example, an estimated 40,000 kg of sugar passed daily through these smuggling routes in the second half of the 2000s (Titeca 2012). A second possibility is that goods are smuggled through the border points, by hiding and/or under-declaring them (i.e. not paying tax on a substantial amount of the goods crossing the border). A large percentage of goods pass through Uganda’s customs points in this way. For example, research of the Uganda Bureau of Statistics and the Bank of Uganda (who had placed surveyors at the border points) shows how, in 2008, informal exports out of Uganda were worth US$1.35 billion, while formal exports were worth US$1.7 billion (UBOS 2009). In other words: more goods were leaving Uganda untaxed or smuggled, than taxed—showing the fluidity of these official border points (Titeca and Flynn 2014). Ivory is smuggled across the border in the same two ways: a first option is that smuggling roads are used (given the bad state of these roads, mainly on bicycles, but also with cars) and stored in storage locations, such as shops, small warehouses or traders’ homes. As a trader summarized We take it [ivory, KT] through the porous borders. You hire a team; they are based a border post. These are not the official border; but through panya roads [smuggling road, KT]. This can be on foot, on a bicycle, and so on. All of it in a small quantity. It is then loaded in a car.5 A second possibility is that ivory is smuggled through border points, either through hiding the ivory in various ways or by bribing the customs officials. For example, sometimes ivory is hidden in busses, as the following trader explains: I have an arrangement with the ‘turnboy’, the boy who opens and closes the booth of the busses. (…) They create a hidden space in the booth, which cannot be seen. When the bus goes to the washing bay upon arrival, they take the ivory. Sometimes the owner is not aware of this.6 Other traders rely on their relations with customs officials to smuggle the goods: traders have developed personal relations with these officials, facilitating this trade. Getting ivory from Congo is the easiest! I know all the border people. Al the Congolese at the border, they are friends. If it’s necessary, I give them 50 or 100 USD. (…)In all these things, personality is the most important! People know me at the border. I succeed in this because I have an easy personality.7 In total, seven of these traders were followed throughout the course of the research. For all of these traders, ivory is only one of the commodities that is traded in these networks. In other words, the traders have the infrastructure to move a variety of commodities, of which ivory is only one. As mentioned earlier, for these interviewed traders, ivory is a fairly ‘new’ commodity, which recently became part of their portfolio of goods, but became integrated into the historically embedded trading networks. How do traders sell it? The involved traders not only smuggle goods across the borders but also stock the ivory: similar to their other commodities, they store the ivory in and around the border town. This allows them to store the ivory in case of increased security controls, collect bigger quantities upon demands of buyers, wait for more favourable prices and so on. They also sell the ivory to traders who come to the border town of Arua. These outside traders mainly originate from the capital Kampala. Some of the interviewed traders are in regular contact with a number of these Kampala traders and call them when they have sufficient supply of ivory. Other outside traders ‘show up’ in Arua and start looking for ivory. Before ivory started arriving on a large scale in a border region, these buyers were absent in the region. During the interviews, a wide range of nationalities were mentioned as ivory buyers, such as Chinese, Korean, Pakistani, Turkish and South Africans.8 Connections and their limits The Uganda–DRC border region can be considered a ‘micro-region’ (Horden and Purcell 2000), which is connected to external regions by the above-mentioned traders: they play an important role as middlemen, connecting various scales of the illegal ivory trade. The border town of Arua acts as a central node connecting traders within the micro-region (i.e. across the Uganda–DRC border), but also towards outside actors: the traders connect the historically embedded regional trading networks and the transnational level. By selling the ivory to outside actors, and particularly to international actors, the ivory is fed into the international market. The traders perform a range of functions that are necessary for the ivory to be moved: buying, smuggling and selling on. Rather than being characterized by a hierarchical structure of an organized crime group, these local ivory traders operate in loose and flexible networks. Within these networks, the traders’ connections are crucial, and particularly the power of these connections. On the one hand, their illegal trade is mainly based on connections with lower-level government officials: custom officials who allow their goods to be passed, local security officials who inform them about checkpoints and so on. Second, they know the local context very well: they know whom to hire as transporters and whom to sell to—including outside traders. These connections allow goods to cross the border, allows them to stock their goods and allows them to sell the ivory. However, the nature of these connections has its limits: transport and trade outside of this ‘micro-region’ is particularly difficult. For example, transporting ivory to Kampala involves a number of risks, for which the traders do not have sufficiently powerful connections: connections are needed more, and particularly higher-level officials, to avoid confiscation on the way to Kampala—which involves bypassing several police checkpoints. Moreover, selling in Kampala also involves knowledge of the market there, which basically involves knowing reliable buyers: throughout the course of the research, many traders had been cheated in one way or the other, in which they, for example, sold ivory to another trader, but never received the money. The quote from the following trader is an illustration of this dynamic: In my ivory dealings I’ve been cheated before. I gave ivory to a man, and he told me: when I reached Pakwatch, the army confiscated everything: they opened the booth, and confiscated everything. He had only given me 1000 USD, and he still owed me 15,000 USD! (…) After this, I even lost interest in the business, and I lost hope. (….) Q: Did you believe him—that everything was confiscated? A: No, he was lying. Because he knew that there are certain things, if you take them, you can’t go to the police.9 Selling ivory requires knowledge of the Kampala ivory market: throughout the course of the research, a number of the local traders tried to expand their reach to Kampala, but encountered major problems in doing so. For example, one trader in this border region, from the town of Arua, summarized the situation as the following: For you as a local to go to Kampala, it is risky. And you don’t know your way there. So we don’t do this. (…) These traders [from the capital Kampala, KT] had created the image that it’s very difficult to sell ivory there, that there is no market! It are these traders who know the market in Kampala. The locals from here don’t know the market there. And the government had started confiscating ivory. So the Arua boys here [the local traders, KT] had enough of it. They did not want anything to do with this [trading in Kampala, KT]. And the traders who know the market in Kampala, they don’t want to share the market with their colleagues: they do not want to tell us where we can sell this.10 Similar examples were given by other traders, who encountered problems either in transporting ivory (such as confiscation) or in finding a market. In other words, for many Arua traders, their activities are limited: their networks are clearly located in the border region and do not reach the capital Kampala, where they do not know the market and are not able to sell the ivory. As a result, many traders could not sell ivory in the capital Kampala, although they could make more profit there: Kampala is the last nodal point before it exits the continent through the airport or is traded further towards Indian Ocean Ports—and has consistently higher ivory prices. For example, between 2010 and 2011, the ivory selling price was between 100 and 120 USD per kilogram in Arua and around 150 USD per kilogram in Kampala. More generally, my field research shows how, between 2010 and 2017, significantly more profit can be made in Kampala: ivory prices are consistently between 30 and 70 USD per kilogram more expensive in Kampala. In sum, many traders limit their activities to their ‘micro-region’: their connections are mainly embedded in this region. In the next section, we look at the activities of traders in another key-node in the illegal ivory trade: Uganda’s capital Kampala, which acts as a major exit-point for ivory, out of Uganda. Connections with government officials: Kampala as a nodal point for illegal trade The connections of traders are not uniform, but differ in power: these differences are principally based on the contacts with the government officials. As shown above, most of the Arua traders base their activities on connections with lower-level government officials based in the region (such as customs officials). Other traders have connections with higher-level government officials, and in particular security officials. These connections give an increased level of security at various points of the trading chain: when crossing the border, when transporting throughout the country, on exit points, and so on. It also allows for access to information on security operations, and interventions in case of a problem, such as the potential release of confiscated goods. A few of these traders are based in Arua, but they are mainly based in Uganda’s capital, Kampala. The city is a nodal point for regional illegal trade: it is a key collection point for illegal commodities such as ivory, from where it transits to the national airport (in Entebbe) or to Kenya (and more particularly the port of Mombasa). Operating in these areas is particularly risky, as the airport is a particularly sensitive point and exporting ivory to Kenya involves longer distances (and more police checkpoints). Connections with high-level government officials allow to navigate these difficulties. Importantly, it also gives access to a wide(r) range of sources (which in turn allows for bigger stocks of ivory). Let us look at the example of Adam (a pseudonym), who is an ivory trader based in Kampala. He has been trading in illegal ivory for about 15 years, and throughout the years, he has been receiving ivory from a range of different sources, with whom he remains in (ir)regular contact: The Tanzanian Wildlife Authority: corruption with Tanzanian government officials is a major problem for wildlife in Tanzania (Brockington 2008) and allows confiscated ivory to re-enter the market—among others, in the hands of Adam. In his words: ‘I know this guy who used to deal with the Tanzania wildlife authority. So they would steal it from the stock, or from poachers, and they would bring it here. (…) They come here, because here the law enforcement department is much easier to manipulate! You always take the route which is the easiest to manipulate’.11 On a number of occasions, Adam went to buy this ivory on the Tanzanian–Ugandan border. Officers from the Ugandan Wildlife Authority (UWA) have been engaged in a number of corruption scandals, equally allowing ivory to enter the market (Tenywa 2014, 2015). Adam frequently collaborates with a trader who used to be a UWA official and who was a regular source. As this person explained: ‘when I was working with UWA, we were systematically selling ivory. When I was working in [national park XY] we did not declare everything which we confiscated. For example, if we confiscated 10 pieces, we declared 2. The rest we were selling’.12 Burundian traders: these traders, bringing over ivory from Bujumbura, largely bought their ivory in Eastern DRC. Adam himself claimed that this ivory mainly originated from rebel groups in Eastern DRC. Individual Ugandan soldiers, who bring it from the Central African Republic. As will be discussed below, Adam has close connections with army officials, allowing him to fairly easily tap into this supply stream. The above-mentioned ‘Arua boys’: Adam has been organizing transports from the Uganda-DRC border region to Kampala. Four traders such as Adam were followed for this research project: two of those (including Adam) had a wide variety of sources; two others were mainly relying on the military. From these various sources, the traders were transporting ivory to Kampala in a number of ways: either it is smuggled directly to the capital, in which the ivory is hidden in various ways (in busses, trucks, cars). Another option is for the traders to go and pick it up at border points, for which they are able to mobilize the necessary contacts. The traders can either smuggle the goods themselves to Kampala or purely act as a middleman, in which they pass on the ivory towards another trader (and pocket the profit). Traders can also rely on military transport infrastructure to do so. These traders have a variety of nationalities: there are not only the Ugandan traders, but there is also a particularly significant presence of West African traders, from, for example, Nigeria, Guinee, Sierra Leone or Côte d'Ivoire. All Kampala traders mentioned personal links with these government actors, particularly army officials: one trader was groomed in the illegal trade business through an army commander who was a family friend, another trader went to high school with an army commander, another traders’ best friend is the brother of a high-level army commander and so on. These connections, in turn, allow operations to be conducted on a bigger scale. Adam, the illegal ivory trader mentioned earlier, is a good example: his area of operation was virtually the whole of the country, from border zones with Tanzania or Congo, to operating in the airport. Key for his operations is his connections with politico-military elites. Particularly his friendship with one army commander (a family friend) proved to be crucial in his introduction to the illegal business: We used to chill at his place when I was at university. We have a drink, and in this way, I knew him; and he introduced me in the business. He was a shrude guy: he only knew the language of money. And he had the cover to make things moving. Also the white people, whom we call ‘investors’, they all knew the big guys. The mainly go to the big guys, they particularly go to the big guys, and they tell the commander what they are looking for. At the same time, many Congolese came here, and in this way business was created.13 Hanging out with the commander exposed him to a range of contacts and business opportunities. This happened at the time of the Congolese wars, when Ugandan army officers were profiting from Congo’s natural resources, in particular minerals (Titeca 2011). While his close connection with the above army official introduced him to the trade, and remained important, this introduction allowed him to build up his own network, involving other traders, buyers and so on. In doing so, he expanded his network and started dealing in a variety of commodities—not only minerals and drugs, but also ivory. Similar stories were told by the other Kampala traders. Importantly, it is through these connections with government officials that traders such as Adam were able to handle this diversity of sources and trade over such a big territory: different from the localized engagements of the Arua traders,14 high-level connections allow transport over a bigger territory. Through these connections, Adam—and other traders with more powerful connections—act as nodes between different scales. First, they have connections with a number of micro-regions: not only with the Uganda–Congo border region, but also with other border regions, and with a steady supply from various government agencies. Second, they are able to connect these various micro-regions to the transnational level, by selling to transnational traders. The contacts of the traders with the international traders vary: a few of them are regularly meeting the international recipients and many of them do not. While the latter were aware of dealing international traders, they were not aware of the eventual recipient. As one trader summarized ‘I never see the person for whom the ivory is eventually intended; I always see an intermediary. So there’s a whole chain of people (…). The final buyer is the Chinese, Thai or Russian. But I never see the person on the meeting: they all have national agents’.15 In some cases, more powerful traders are involved in having the ivory pass through customs and/or airport. For example, one trader primarily dealt with the airport: I depend on the airport. When there suddenly is a window of opportunity: then it can suddenly leave. They call and say: it can leave the country now. It are the airport agents who make sure this happens. They tell me: there’s now place on a plane.16 Another trader highlighted how It used to be a high-level military officer who would help things pass through the airport. (…) So they’re in charge of that business, and they would make sure the plane is delayed for 5 minutes, and the ivory would get in board.17 Connections and power In sum, connectedness is crucial in this trade, as well as particular nodes providing these connections. These nodes can be both places and persons: the article highlighted the Ugandan border town of Arua and the capital Kampala as central localities in the illegal ivory trade, in which various scales come together. They each constitutes nodes, bringing together illegal ivory trade from a variety of sources, and acting as a locality to store the ivory, bring together various traders and so on. Also, the traders act as nodes, in which their connections bridge various levels and scales: both the traders in Arua and Kampala connect with transnational traders. While both are linking up the regional with the transnational level, the power of their connections in doing so differs, particularly their internal connections: the connections of less powerful traders (such as the majority of the traders in Arua) are with lower-level, largely street-level government officials. While this allows goods to cross borders, and allows them to sell to international traders, it does not do more than that. It makes their activities largely limited to the micro-region in which they are based. Other traders (such as the Kampala traders we discussed) who have connections with higher-level officials have a number of advantages from these connections: they are able to operate on a bigger scale, make bigger profit and so on. They face less risks, and are better equipped, to operate on a larger territory, allowing them to collect supplies from various regions. They can also transport larger quantities (as they, e.g., can be in a position to rely on army trucks) and are able to store larger quantities (as they are better informed about potential confiscations). A last quote to illustrate this dynamic, in which one of the Kampala traders explains how ivory is traded, with the help of a high-ranked military official: A Congolese contacted me here in Kampala, and told me about the ivory he had. It was stuck at the border in Congo. I contacted [high-ranking army official]. The Congolese had 3 pieces of ivory as a sample, and they told me they had a ton. The problem was bringing it over; and [high-ranking army official] was the one able to offer protection. So [high-ranking army official] came with the money; and he gave me a truck with soldiers. I negotiated with the traders 20 USD per kg, and I told [high-ranking army official] 50 USD per kg. So I made a lot of money then.18 Conclusions My findings, based on the research among illegal ivory traders in Uganda, aimed to unpack the transnational organization of the illegal ivory trade. Empirical understanding of the trade’s organization is limited: a range of academic literature, policy reports and media reports suggest that a transnational organization—a ‘criminal syndicate’ or ‘gang’—is responsible for this trade. My findings suggest that, similar to the research on human smuggling or drug trade (Zhang and Chin 2003; Natarajan 2006), ivory traffickers form ad hoc alliances and flexible networks, in which individual traders are the thriving force behind this trade: they constitute nodes at particular points of the value chain (such as the Uganda–DRC borderlands or the capital Kampala). Instead of relying on a group structure, they form flexible coalitions in order to buy, sell or transport the ivory. A crucial explanation for this are the market conditions of ivory, which are dynamic and fragmented: supply and demand are unpredictable, and a variety of dangers loom, such as confiscation, lack of market and being cheated by other traders. In these conditions, the traders’ personality and connections are central throughout the trade: personal ties with customs officials in order to cross the border, personal ties with army officials for supply, protection and sale, personal connections with other traders in order to avoid cheating and so on. This does not mean that the level of (dis)organization is uniform in the ivory trade. The article has showed the various degrees of power of the traders’ connections: the better connected with higher-level officials, the more advantages for a trader. More particularly, it gives them a higher degree of security and predictability in their activities: not only in supply and sale points, but also in transporting goods through important points such as customs, where they are able to rely on connections with high(er)-level government officials. This, in turn, allows them to transport and store larger quantities of ivory. The degree of security and predictability of traders with lower-level connections are more limited, largely limiting their actions to their ‘micro-regions’. When the literature on illegal ivory trade emphasises the dominance of international syndicates—what Hübschle (2016: 208) calls the idea of a ‘foreign-dominated parasitic conspiracy’—important realities get overlooked. Concretely, this research has shown how, first, ivory trade is to a large extent organized and experienced locally and regionally. Transnational ivory trade mainly relies on historically embedded trading networks, such as the regional trade along the Uganda–DRC border. Second, it is the traders’ (unequal) connections linking the regional and transnational scales, which are central in this trade. In other words, the transnational is not found in global syndicates, but in the various interactions between these different local and global traders, inside and between different countries (Columb 2017). References Bergenas , J. and Knight , A . ( 2015 ), ‘ Green Terror: Environmental Crime and Illicit Financing ’, SAIS Review of International Affairs , 35 : 119 – 31 . Google Scholar Crossref Search ADS Bichler , G. , Malm , A. and Cooper , T . ( 2017 ), ‘ Drug Supply Networks: A Systematic Review of the Organizational Structure of Illicit Drug Trade ’, Crime Science , 6 : 1 – 23 . Google Scholar Crossref Search ADS Biegus , O. and Bueger , C . ( 2017 ), ‘ Poachers and Pirates ’, South African Crime Quarterly , 60 : 29 – 36 . Brockington , D . ( 2008 ), ‘ Corruption, Taxation and Natural Resource Management in Tanzania ’, Journal of Development Studies , 44 : 103 – 26 . Google Scholar Crossref Search ADS Cakaj , L. and Lezhnev , S . ( 2017 ), Deadly Profit: Illegal Wildlife Trafficking through Uganda and South Sudan . The Enough Project . Clarke , A. and Babic , A . ( 2016 ), ‘ Wildlife Trafficking Trends in Sub-Saharan Africa ’, in OECD , ed., Illicit Trade: Converging Criminal Networks . OECD Reviews of Risk Management Policies, OECD Publishing , 57–78. Columb , S . ( 2017 ), ‘ Excavating the Organ Trade: An Empirical Study of Organ Trading Networks in Cairo, Egypt ’, British Journal of Criminology , 57 : 1301 – 21 . Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES) ( 2016 ), ‘ Report on the Elephant Trade Information System ’, in Seventeenth Meeting of the Conference of the Parties . CITES . Ellis , S. and MacGaffey , J . ( 1996 ), ‘ Research on Sub-Saharan Africa’s Unrecorded International Trade: Some Methodological and Conceptual Problems ’, African Studies Review , 39 : 19 – 41 . Google Scholar Crossref Search ADS Environmental Investigation Agency (EIA) ( 2014 ), Vanishing Point: Criminality, Corruption and the Devastation of Tanzania’s Elephants . EIA . Finckenauer , J . ( 2005 ), ‘ Problems of Definition: What is Organized Crime ?’, Trends in Organized Crime , 8 : 63 – 83 . Google Scholar Crossref Search ADS Hagan , F . ( 2006 ), ‘ “Organized Crime” and “Organized Crime”: Indeterminate Problems of Definition ’, Trends in Organized Crime , 9 : 127 – 37 . Google Scholar Crossref Search ADS Hayman , G. and Brack , D . ( 2011 ), International Environmental Crime . The Royal Institute of International Affairs . Hill , C . ( 2005 ), ‘ Measuring Transnational Crime ’, in P. Reichel , ed., Handbook of Transnational Crime and Justice . Sage . Hobbs , D . ( 1998 ), ‘ Going Down the Glocal: The Local Context of Organized Crime ’, The Howard Journal , 37 : 407 – 22 . Google Scholar Crossref Search ADS Hobbs , D. and Dunnighan , C . ( 1998 ), ‘ Glocal Organized Crime: Context and Pretext ’, in V. Ruggiero , N. South and I. Taylor , eds, The New European Criminology: Crime and Social Order in Europe . Routledge . Horden , P . ( 2012 ), ‘ Situations Both Alike? Connectivity, the Mediterranean, the Sahara ’, in J. McDougall and J. Scheele , eds, Saharan Frontiers: Space and Mobility in Northwest Africa . Indiana University Press . Horden , P. and N. Purcell ( 2000 ), The Corrupting Sea . Oxford University Press . Hübschle , A . ( 2016 ), ‘ Security Coordination in an Illegal Market: The Transnational Trade in Rhinoceros Horn ’, Politikon , 43 : 193 – 214 . Google Scholar Crossref Search ADS McCann , N . ( 2017 ), ‘ Attacks on “Militarized Conservation” Are Naïve ’, Mongabay , 10 October, available online at https://news.mongabay.com/2017/10/attacks-on-militarized-conservation-are-naive-commentary (accessed on 12 October 2017 ). Meagher , K . ( 1990 ), ‘ The Hidden Economy: Informal and Parallel Trade in Northwestern Uganda ’, Review of African Political Economy , 47 : 64 – 83 . Google Scholar Crossref Search ADS Milliken , T . ( 2014 ), Illegal Trade in Ivory and Rhino Horn . USAID and TRAFFIC . Milliken , T. , Underwood , F. M. , Burn , R. W. and Sangalakula , L . ( 2016a ), The Elephant Trade Information System (ETIS) and the Illicit Trade in Ivory: A Report to the 17th Meeting of the Conference of the Parties to CITES . TRAFFIC . —— ( 2016b ), Addendum to the Elephant Trade Information System (ETIS) and the Illicit Trade in Ivory: A Report to the 17th Meeting of the Conference of the Parties to CITES . TRAFFIC . Natarajan , M . ( 2006 ), ‘ Understanding the Structure of a Large Heroin Distribution Network: A Quantitative Analysis of Qualitative Data ’, Journal of Quantitative Criminology , 22 : 171 – 92 . Google Scholar Crossref Search ADS Naylor , R. T . ( 2004 ), ‘ The Underworld of Ivory ’, Crime, Law & Social Change , 42 : 261 – 95 . Google Scholar Crossref Search ADS Paoli , L. and Fijnaut , C . ( 2004 ), ‘ Comparative Synthesis of PartII ’, in L. Paoli and C. Fijnaut , eds, Organized Crime in Europe: Concepts, Patterns and Control Policies in the European Union and Beyond . Springer . Pires , S. and Clarke , R . ( 2011 ), ‘ Sequential Foraging, Itinerant Fences and Parrot Poaching in Bolivia ’, British Journal of Criminology , 51 : 314 – 35 . Google Scholar Crossref Search ADS Pires , S. and Moreto , W . ( 2011 ), ‘ Preventing Wildlife Crimes: Solutions that Can Overcome the “Tragedy of the Commons” ’, European Journal on Criminal Policy and Research , 17 : 101 – 23 . Google Scholar Crossref Search ADS Reuter , P . ( 1985 ), Disorganized Crime . MIT Press . Runhovde , S . ( 2017 ), ‘ Merely a Transit Country? Examining the Role of Uganda ’, Trends in Organized Crime , available online at https://link.springer.com/article/10.1007%2Fs12117-016-9299-7. Scheele , J . ( 2010 ), ‘ Traders, Saints and Irrigation: Reflections on Saharan Connectivity ’, Journal of African History , 51 : 281 – 300 . Google Scholar Crossref Search ADS —— ( 2011 ), ‘ Circulations Marchandes au Sahara: Entre Licite et Illicite ’, Hérodote , 3 : 143 – 62 . Sims , O. and Gilles , R . ( 2016 ), ‘ Middlemen and Contestation in Directed Networks ’, December 2016. Working paper, available online at https://arxiv.org/pdf/1612.08884.pdf Somerville , K . ( 2016 ), Ivory: Power and Poaching in Africa . Hurst Publishers . Southerland , M. and Potter , G. W . ( 1993 ), ‘ Applying Organization Theory to Organized Crime ’, Journal of Contemporary Criminal Justice , 9 : 251 – 67 . Google Scholar Crossref Search ADS Tenywa , G . ( 2014 ), ‘ Sh3B Ivory Goes Missing from UWA ’, New Vision , 11 November. —— ( 2015 ), ‘ UWA Ivory Scandal Sucks in Army, Police ’, New Vision , 2 February. Titeca , K . ( 2006 ), ‘ Les OPEC boys en Ouganda, Trafiquants de Pétrole et Acteurs Politiques ’, Politique Africaine , 103 : 143 – 59 . Google Scholar Crossref Search ADS —— ( 2009 ), ‘ The Changing Cross-Border Trade Dynamics Between North-Western Uganda, North-Eastern Congo and Southern Sudan ’, Crisis States Working Paper 63, Series 2. London School of Economics and Political Science . —— ( 2011 ), ‘ Access to Resources and Predictability in Armed Rebellion: The “Force Armée du Peuple Congolais” Short-lived “Monaco” in Eastern Congo ’, Afrika Spectrum , 2 : 43 – 70 . —— ( 2012 ), ‘ Tycoons and Contraband: Informal Cross-Border Trade in West Nile, North-Western Uganda ’, Journal of Eastern African Studies , 6 : 47 – 63 . Crossref Search ADS —— ( 2013 ), ‘ Out of Garamba, into Uganda: Poaching and Trade of Ivory in Garamba National Park and LRA-Affected Areas in Congo ’, IOB Analysis and Policy Briefs No. 5. University of Antwerp . Titeca , K. and De Herdt , T . ( 2011 ), ‘ Regulation, Cross-Border Trade and Practical Norms in West Nile, North-Western Uganda ’, Africa , 80 : 573 – 94 . Google Scholar Crossref Search ADS Titeca , K. and Flynn , R . ( 2014 ), ‘ “Hybrid Governance,” Legitimacy and (Il)legality in the Informal Cross-border Trade in Panyimur, North-West Uganda ’, African Studies Review , 57 : 71 – 91 . Google Scholar Crossref Search ADS Uganda Bureau of Statistics (UBOS) ( 2009 ), The Informal Cross Border Trade Survey Report 2008 . UBOS and Bank of Uganda . Underwood , F. , Burn , R. and Milliken , T . ( 2013 ), ‘ Dissecting the Illegal Ivory Trade: An Analysis of Ivory Seizures Data ’, PLoS One , 8 : 1 – 12 . UNEP, CITES, IUCN, TRAFFIC ( 2013 ), Elephants in the Dust – The African Elephant Crisis . United Nations Environment Programme . United Nations Office on Drugs and Crime (UNODC) ( 2002 ), Results of a Pilot Survey of Forty Selected Organized Criminal Groups in Sixteen Countries . UNODC . —— ( 2013 ), Transnational Organized Crime in Eastern Africa . UNODC . Van Uhm , D . ( 2016a ), The Illegal Wildlife Trade . Springer . —— ( 2016b ), ‘ Monkey Business: The Illegal Trade in Barbary Macaques ’, Journal of Trafficking, Organized Crime and Security , 2 : 36 – 49 . Vira , V. and Ewing , T . ( 2014 ), ‘ Ivory’s Curse ’, C4ADS , April 2014. Von Lampe , K . ( 2014 ), ‘ Transnational Organized Crime in Europe ’, in J. S. Albanese and P. Reichel , eds, Transnational Organized Crime: An Overview from Six Continents , 75 – 92 . Sage . Warchol , G. , Zupan , L. and Clack , W . ( 2003 ), ‘ Transnational Criminality ’, International Criminal Justice Review , 13 : 1 – 26 . Google Scholar Crossref Search ADS Wasser , S. K. , Brown , L. , Mailand , C. , Mondol , S. , Clark , W. , Laurie , C. and Weir , B. S . ( 2015 ), ‘ Genetic Assignment of Large Seizures of Elephant Ivory Reveals Africa’s Major Poaching ’, Science , 349 : 84 – 7 . Google Scholar Crossref Search ADS PubMed Wasser , S. K. , Clark , W. J. , Drori , O. , Kisamo , E. S. , Mailand , C. , Mutayoba , B. and Stephens , M . ( 2008 ), ‘ Combating the Illegal Trade in African Elephant Ivory with DNA Forensics ’, Conservation Biology , 22 : 1065 – 71 . Google Scholar Crossref Search ADS PubMed White , R . ( 2011 ), Transnational Environmental Crime: Toward an Eco-global Criminology . Routledge . Wilson , A . ( 2012 ), ‘ Saharan Trade in the Roman Period ’, Azania: Archaeological Research in Africa , 47 : 409 – 49 . Google Scholar Crossref Search ADS Wittemyer, G., Northrup, J.M., Blanc, J., Douglas-Hamilton, I., Omondi, P. and Burnham, K.P. (2014) ‘Illegal Killing for Ivory Drives Global Decline in African Elephants’, Proceedings of the National Academy of Sciences of the United States of America 111: 13117–13121 . Wright , G . ( 2011 ), ‘ Conceptualising and Combating Transnational Environmental Crime ’, Trends in Organized Crime , 4 : 332 – 46 . Google Scholar Crossref Search ADS Wyatt , T . ( 2013 ), Wildlife Trafficking: A Deconstruction of the Crime, the Victims, and the Offenders . Palgrave Macmillan . Zaitch , D . ( 2002 ), Trafficking Cocaine: Colombian Drug Entrepreneurs in the Netherlands . Kluwer Law International . Zhang , S. and Chin , K . ( 2003 ), ‘ The Declining Significance of Triad Societies in Transnational Illegal Activities ’, British Journal of Criminology , 43 : 469 – 88 . Google Scholar Crossref Search ADS Zimmerman , M . ( 2003 ), ‘ The Black Market for Wildlife ’, Vanderbilt Journal of Transnational Law , 36 : 1657 – 90 . Footnotes 1 Moreover, the study is based on one-off interviews with law enforcement officers, which may explain the ‘conflicting accounts’ (Runhovde 2017: 15) given by the respondents: trust is essential in this kind of sensitive research on illegal commodities, particularly given the strong involvement of law enforcement officers in this trade (Ellis and MacGaffey 1996). 2 Interview Ugandan trader, border region, 18 May 2012. 3 The timber trade was, for example, confirmed by an internal UN report (MONUSCO Ituri Brigade Military Daily Sitrep Period Covered: 31800B JAN 10 to 011800B FEB 10, p. 11). The ivory trade is mentioned in a more recent policy report (Cakaj and Lezhnev 2017). 4 Interview trader, Arua, 12 November 2015. Similar stories are reported in recent reports, i.e. how army trucks are smuggling in a variety of goods, including timber and ivory, and how government officials are not allowed to check their trucks (Cakaj and Lezhnev 2017). 5 Interview ivory trader, Arua, 21 May 2012. 6 Interview ivory trade, Kampala, 29 January 2014. 7 Interview ivory trader, border region, 14 November 2015. 8 For example, a trader in the border region argued how ‘there are Chinese who come here; they know have someone looking for market. Also South Africans are here, looking. (…) It is us, Arua boys, who are the ones selling to them’ (interview ivory trader, Arua, 2 November 2015). 9 Interview ivory trader, Arua, 11 January 2016. 10 Interview ivory trader, Arua, 12 November 2015. 11 Interview ivory trader, Kampala, 17 January 2017. 12 Interview former UWA official, Kampala, 17 November 2015. 13 Interview ivory trader, Kampala, 01 March 2017. 14 These traders have a much more localized engagement with government officials: they know the custom authorities at the border, which they pay something from time to time, and with whom they have developed relations allowing ivory to pass through the border. 15 Interview ivory trader, Kampala, 29 January 2014. 16 Interview ivory trader, Kampala, 29 January 2014. 17 Interview ivory trader, Kampala, 26 January 2017. The airport came back in many of the traders’ accounts, as well as in interviews with law enforcement officials, journalists and analysts: all of these actors point at the increasing importance of Entebbe airport as an exit-point for ivory. In the words of a trader, when discussing the regional interest in the airport: ‘All these traders prefer to trade ivory through Entebbe, because things are more flexible here’ (interview ivory trader, Kampala, 17 November 2015). 18 Interview ivory trader, Kampala, 29 January 2014. © The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). All rights reserved. For permissions, please e-mail: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Mercan, Boran, Ali

doi: 10.1093/bjc/azy028pmid: N/A

Abstract Much of criminology has recently tended to replace the sociological concept of professional criminal or ‘good thief’ by the psychological concept of ‘expertise’ that draws attention to the cognitive processing of individualized offenders who automatically become aware of burglary-related cues. Setting out the experiences of a group of ex-offenders in the capital of Turkey, this article proposes the dispositional theory of action in understanding the modus operandi of burglary and further argues that professional burglary is a collective form of action made possible by the combination of bodily and mental dispositions capitalized within an internal organization of burglary. This article not only reveals the intrinsic operational logic of burglary in a non-Western context but also takes the first step in arousing scholarly interest to study the modus operandi of professional burglary specifically in Turkey. Introduction The news media regularly reports on the professionality of certain kinds of criminals and in such circumstances tends to reflect on the subtlety and dexterity of offenders in the commission of crime. In recent years, the Hatton Garden jewellery heist has been a notable example of professional burglary in Britain. Here, four older and experienced offenders planned and executed a burglary with skill and meticulous detail. The process included a number of steps: neutralizing an alarm system, disabling a lift in order to clamber down to a safe-deposit vault in the basement, drilling through a vault wall a full half metre thick, and finally taking a haul of gold, cash and 73 safe-deposit boxes containing jewellery (The Guardian, 15 January 2016). More than 40 years ago, Shover accounted for the characteristics of a good burglar and the modus operandi of professional burglary respectively in two groundbreaking articles. In the first article ‘Structures and Careers in Burglary’ (1972), inspired by Sutherland’s (1937) The Professional Thief, Shover defined good burglars as being technically competent, holding a reputation for personal integrity, specialized in burglary and having a certain degree of success determined by the amount of money stolen and the time done in prison (Shover 1972: 541). In picking up right participants for prison interviews, Shover (1972: 542) operationalized a good burglar as someone making big scores, entering a place by cutting a hole in the roof or wall and safecracking by the methods of drilling or burning. In this typology, the technical competency apparently took hold of the primary concern in operationalizing the construct of a good burglar. A year later in another article ‘The Social Organization of Burglary’ (1973), Shover this time demonstrated how the technical competency is put to work and that a successful burglary relies on a collective action by an internal and external cooperation if it is to be practiced professionally. The internal organization has to do with the division of labour within burglary crews and underlies how the crew operates while pulling off the scores, whereas the external organization refers to the interaction between burglars and a wider milieu with which they have to interact for receiving ‘tips’, fencing stolen goods and securing legal protection (Shover 1973: 501). In this line, much of criminology attempted to define burglars with regard to their level of technical competency. Many typologies were produced: the ‘organised offender’ (Irwin 1970), ‘roundman’ (Letkemann 1973), ‘able criminal’ (Mack 1975), ‘high-level’ (Maguire and Bennett 1982), ‘searcher’ (Bennett and Wright 1984) and ‘suburban burglar’ (Rengert and Wasilchick 2000). Although much of criminology has shifted from sociology of crime as work to cognitive psychology of crime as rational choice (Clarke and Cornish 1985), this has not prevented criminologists from studying professional burglary despite under different terminologies: The concept of ‘good thief’ (Shover 1973) has tended to be replaced by the term ‘expert’ (Nee 2015). Psychological studies have explored ‘expertise’ and ‘skilled offenders’ from the quasi-experimental analyses on the cognitive processing of offenders, by focusing on the phases of casing, targeting, breaking and entering, search and departure (Nee and Taylor 1988; Logie et al. 1992; Wright et al. 1995; Nee and Meenaghan 2006; Nee and Ward 2015). The Western criminology has been systematically studying the phenomenon of professional burglary for many decades, but the Turkish criminology has not properly paid attention to the subject yet. However, professional crime is a longstanding phenomenon, and professional forms of burglary are quite widespread in Turkey, too. Some recent examples would be illustrative: A mainstream newspaper report under the headline ‘Burglary Tours with Ladder, Break in 80 Houses in 9 Provinces [Merdivenle Hırsızlık Turu, 9 İlde 80 Eve Girdi]’ tells the story that a burglar breaks in houses by the help of a ladder he carries to climb up the first floors of the apartment blocks (Hürriyet Ankara, 31 December 2017). Also numerous daily reports cover the incidents of burglary crews and the police operations that reveal how organized the crews are in housebreaking: ‘Butterfly Operation on Burglary Crew [Hırsızlık Çetesine Kelebek Operasyonu]’ (Hürriyet Ankara, 28 December 2017) and ‘Eight Arrests in Goose Operation, They Break in by Breaking the [Window] Bars [Kaz Operasyonunda Sekiz Tutuklama, Demirleri Kırıp Girdiler]’ (Hürriyet Ankara, 25 April 2018). The successive police crackdowns demonstrate that burglary crews operate according to a certain division of labour and employ various methods in breaking and entering. Some time ago, a currently closed, pro-Islamist conservative newspaper published a particularly detailed story on the life of professional criminals. The story revealed the ‘Professional secrets of burglars! [Ev hırsızlarından meslek sırları!]’: Burglars always case a property before breaking and entering: throwing a ball at windows, checking the mail box and the footwear in front of the door, and pressing the bell are just a few of their techniques. They carry tools such as screw-drivers, stiff plastic cards (to force latch-type locks), crowbars, and adjustable spanners to open steel doors. Meanwhile, a plaster is taped over the security peephole of the opposite neighbour’s door. The lock mechanism, having been dismantled before entering, is replaced and then locked once the burglars are inside. Moreover, the interviewees reveal that the professional burglary is carried out by two to three individuals; one is always left as a lookout. Once entering, burglars move directly to the bedroom before ransacking the whole house. (Zaman, 11 August 2013; my translation) However, although there are countless journalistic information on burglary, there is a lack of social research to examine the modus operandi of burglary in Turkey. How do burglars organize the division of labour? What techniques do they make use of? How do they select the target? What cues do they pay attention to in the target-selection process? Is there any similarity with the findings of Western burglary studies, or can one mention a specificity of the Turkish case? If so, what are these? This article sets out to respond to such questions by examining the internal organization of a desisted burglary crew (1998–2003) through life stories and in-depth interviews in Ankara, the capital of Turkey. In doing so, the article focuses on the crew’s technique of burglary called askı [hanging] in the Turkish criminal subculture and analytically maps out their organization of everyday life, preparation, selection of neighbourhood and flat, division of labour, and the entering, breaking and departure phases of their burgling activities. However, in contrast to the recent experimental-cognitive analysis of expertise, the modus operandi of professional burglary will be constructed within the frame of dispositional theory of action (Bourdieu 1977; 1990; Wacquant 2016). The logic behind the dispositional examination rests on the observation that a successful internal organization of burglary is not only something embedded in the habitual, gratuitous and everyday life practices of a group of offenders but also carried out in putting collective criminal capital to work. Thus, this article intends to make three contributions to the literature of burglary: First, a professional form of burglary does not have to necessarily include a rational deliberation in deciding moves and actions at the scene of crime; instead, much of practice thrives unconsciously and as part of everyday life habits. Second, a successful burglary needs a cooperation and coordination as the literature emphasizes but also it requires a collective configuration of different skills performed by each member of the crew. Finally, the findings of this article not only reveal the operational logic of burglary in a non-Western context but also take the first step in arousing scholarly interest to study the modus operandi of professional burglary specifically in Turkey. In what follows, the importance of the dispositional theory of action will be explained in relation to the cognitive psychology’s approach to the mental script of offenders, then the method and sampling clarified and finally data analysed. Dispositional Theory of Burglary Burglars are often argued to display a bounded rationality in criminal decision-making as the researchers charted out the patterns in which offenders attend to some cues and prioritize some of them in offending (Cromwell et al. 1991; Shover and Honaker 1992; Tunnell 1992; Wright and Decker 1994). The recent cognitive psychological studies have claimed that continuous performances of burglary produce technical expertise, depending on the formation of long-term schemata that generate competence and skills in target selection, entry and crime commission (Clare 2011). Expert offenders decide to offend within a cognitive script by which to pre-consciously scan and interpret the environmental offence-related cues and act automatically, ensuing from the swift cognitive evaluation process (Nee and Ward 2015). For instance, offenders tend to act rationally with a certain deliberation in the target-selection process and are able to find cues that precipitate whether to decide break in or not. In this regard, expertise in crime is no way different from any expertise in any other legitimate fields (Nee and Meenaghan 2006; Nee 2015). The cognitive approach doubtlessly makes a significant contribution to the literature because situational crime prevention strategies depart from the results of cognitive psychological experiments that focus on much of rational aspects of offenders’ decisions in casing and target selecting (Brantingham and Brantingham 1993; 2004; Bernasco and Nieuwbeerta 2005; Clare 2011; Nee et al. 2015). Nee and Ward, however, state that ‘the act of offending seems strongly governed by habitual, gratuitous and largely unconscious decision-making’ (2015: 5). Stressing the unconscious nature of offending practice, this declaration of cognitive psychologists refers us back to the role of social practices, unconsciously mediated between the conduit of habitus and a given capital endowment of the agent (Bourdieu 1977; 1990; Bourdieu and Wacquant 1992). Underlying Bourdieu’s theory of social practice within a more systematic ground, the dispositional theory of action introduces the tripartite conceptual tools of habitus, capital and field (Wacquant 2016). These then guide an exploration of the socially individuated nature of human practice and helps us chart how bodily and mental dispositions are sedimented (habitus) and how individuals are equipped with certain dispositions as an embodied asset (capital) and ranked amongst one another in order to form a structured conflict (field) (Wacquant 2004; 2011). From within the frame of the dispositional theory of criminal action, becoming a professional criminal or expert refers to the formation of criminal habitus in which the subject acquires criminal dispositions and transforms into an adept agent of criminal fields, endowed with criminal capital. I elsewhere discussed the formation of burglary-related dispositions out of a group of offenders’ unconscious, game-like activities (Mercan 2018). Wacquant’s (2014) dissection of habitus into cognitive (the categories of perception), conative (proprioceptive capacities, sensorimotor skills and kinesthetic dexterities) and affective (libidinal and cathectic) components helped us map out how each offender gained the bodily and mental dispositions to select an appropriate flat and burgle it successfully. However, in that analysis, the operational principles of criminal habitus were out of the discussion. Here two points can be made about the way in which bodily and mental dispositions are put to work and operationalized, or come to be capital asset over the entire course of burglary. First, the criminal habituation process, by repeated exposure to the same experience, covers many dimensions of everyday life. It comprises what time the crew should depart, where they should go to ‘work’ and finally how they should break and enter. Deliberating on the wealth of the target neighbourhood, deciding which flat should be hit and instant recognition of burglary-related cues seem to be an accumulated result of whole-life experiences, ranging from unconsciously witnessing the middle-class lifestyle in apartment buildings to various encounters with upper class people at work and school (Mercan 2018: 4–9). Insofar as the criminal habitus feeds from all-encompassing life experiences, criminal practices are by and large determined by the routine activities of residents (Cohen and Felson 1979). Professionalism thus occurs through the organization of everyday life of the crews and comes to do with how they prepare themselves in relation to the routine activities of neighbourhood inhabitants. Second, the most important part of criminal habitus formation is that it has a collective character, which reflects on the way that the dispositions are put into practice. The cognitive, conative and affective parts of criminal habitus are all the product of collective training. [E]very component involved in the forging of habitus is quintessentially collective: the categories of perception are discerned and taught through joint activities; the skills are learned by observing and honed by acting in concert with members; the desires are aroused and channelled toward their proper objects in repeated interaction with other participants sharing the illusio specific to the universe studied. And the welding of perceptual, kinematic, and cathectic components of habitus into a coherent working ensemble is also carried out collectively in practice through mimesis and osmosis. (Wacquant 2014: 126) As such, the practice of burglary-related dispositions is also a collective activity in which all crew members come from the same socio-class background, habituate the same practices and desire the same monetary rewards. In this study, the modus operandi of burglary appears professionalized insofar as it requires a collective organization and overcomes deterrents due to a systematic division of labour. The Western literature of burglary has stressed the working patterns of burglars as an individual or, at least, dual enterprise in the form of crews (see Letkemann 1973; Rengert and Wasilchick 2000; Mawby 2001). In Turkey, apart from journalistic accounts portraying burglary as a collective activity as demonstrated earlier, there has not been any criminological research that reveals whether burglars work individually or in crew. However, our study shows that the crew enjoyed the turning of their collective force into economic capital. Different physiques and different skills in different parts of the criminal division of labour constitute a transcendental force that goes beyond skills unique to each individual. The collective criminal capital of the crews appears to be easily converted into economic gain as long as the crew overcome deterrents—such as window bars and upper floors—by utilizing a combination of unique individual skills. To this perspective, before I explain the modus operandi of burglary conducted by the crew, I will clarify methodological issues. Methods Data for this study were drawn out of field notes and unstructured interviews with a group of desisted offenders called the crew. The acquaintance with the crew dates back to my early street socialization in the neighbourhood once we used to live in. Despite holding a quite flexible membership structure at earlier times, the crew consisted of five stable members when it became professionalized. These were Muzo, Reddy, Maddy, Lordy and Alex who have mostly a lower-class, rural migrant background, whose parents (except Lordy and Alex’s) used to work as a doorkeeper in a middle-class, affluent neighbourhood of Ankara. They were all male, dropped out the school at very early ages, went in prison more than once (except Alex) and specialized in burglary. Their age varies from 35 to 40. The members of the crew were actively involved in burglary in 1998–2003 but then terminated their criminal career. For more than the 17 years of contact with the crew (1998–2015) despite some intervals, I was able to meet and talk with the crew members. Relying on the already-established rapport, each did not hesitate to participate in a criminological study. The interviews with the crew members were conducted in a one-to-one conversation and all of them allowed me to record the conversations. Although the participants were allowed to talk freely using their own concepts and terminology, a semi-structured method was also put into practice with regard to recording the technical part of crime commission. Some questions were submitted in advance for this: (1) How is burglary organized? (2) What kind of division of labour occurs? (3) What are the techniques of breaking and entering? (4) Which flats are targeted? However, direct participant observation and interview methods pose ethical and, more importantly, legal problems in the exploration of the modus operandi of burglary. To overcome this problem, ‘staged activity analysis’ was employed as an alternative strategy in which the offenders ‘are asked to reconstruct and simulate their past burglaries as nearly as possible in the same manner in which they were originally committed’ (Cromwell et al. 1991: 15–16). By virtue of using this method in interviews, the major legal and ethical problems were overcome. In this technique, ‘extensive interviews’ and ‘ride alongs’ were carried out with informants asked to recall and evaluate motivations and strategies, skills and techniques necessary to burglarize a residential site, as well as the drama of previous burglaries (p. 16). All interviews took place over a 7-month period in 2014–2015. The length of these interviews varied from 12 minutes to 5 hours. Recorded conversations were later transcribed in Turkish verbatim. Transcripts were coded and translated into English where necessary. The Modus Operandi of ‘Hanging’ The crew developed and applied the technique called as askı [hanging] or closed door [kapalı] in the Turkish criminal subculture. The crew members were young men in their mid of 20s in the late 1990s, but each had already known how to break in well by the early 2000s. In this process, the phenomenon of burglary appears in their narrative directly to do with the everyday life practices, which comprises not only the specific course of time allocated to housebreaking but also before and after burgling activities, and even the organization of whole day. Given the fact that the ‘before’ and ‘after’ of burglary have not been investigated enough regarding offenders’ perspective and behaviour (Nee 2015), the crew’s before and after activities appear very important to organize their criminal activities. As of 1998, the crew regulated and organized its daily activities regarding their ‘crimetime’ (Rengert and Wasilchick 2000). Our normal meeting was about 12–1 in the afternoon. We used to hang out about 12pm to 1pm. Well, we were going to eat something if people were up to it. Well, we’re eating the cheapest or the most expensive things [depending on the money in their pockets]. After that, women was a big reason [to go hanging out]. Q: When did you go out in the evening? About 5 to 7pm. In the meantime, around 5 to 7pm, everything used to stop! Everything used to stop and in that time, we’re going to work. (Maddy) In 1998, the crew was acting collectively. Muzo was a fan of football games and used to play matches almost all day at the park in the neighbourhood, where he also worked as a gardener for a short while. The members of the crew thus frequented Muzo’s lodge and grouped around there until he finished for the day. In 1998–1999, the daily hang-outs of the crew spatially revolved around Commercial High School and Lower Park in the neighbourhood, and later the bachelor pad rented in for sexual advantages and socialization purposes. The crew would be either around the stone benches near the school or playing billiards at the café, or staying at the pad until the light turned to dusk at 5 pm during the winter. In the years 2000–2003, however, the daily hang-outs gave away to meeting up only before nightly ‘work’ due to their various convictions and aging. During the day, Muzo pretended to be working in jobs if asked. Muzo was lying to his family that he was distributing newspapers in the morning; therefore, every morning he used to get up very early and leave the house. He was driving far away to the outskirts of the city, parking his car and sleeping in it. In fact, he spent the day playing football at various parks. Alex often joined these matches after school or when truanting, the latter of which was more common. An interim member Atom said he was running errands for his father’s painting job. Lordy was killing time around the parks with peers in the neighbourhood by boozing or going to nightclubs if he had money. Reddy was always at home by the time he was due to go to ‘work’ as he enjoyed a regular nightlife, often with Lordy, until the morning. The crew came together for work and split up after the job. The organization of daily life regarding specific burglary time, 5–8 pm during winter, did not vary for 5 years and would hold a central status for each member’s time schedule despite the crew members going in and out of prison. The crew time to ‘work’ can never be considered as independent from the routine activities of the capital’s population (Cohen and Felson 1979; Felson and Boba 2010). For all members living in the middle class, bureaucratic neighbourhood, individual accounts suggest that the best possible time for breaking was just after sunset at 5 pm until 8 pm during the winter. The reason behind this timing was due to the fact that Ankara is the capital of Turkey whose population mostly works in the state institutions and service sector. Thus, the interval between 5 pm and 8 pm is a time where civil servants and public officials are on their way back home while all the houses and flats are vacant. The crew’s choice of time was the result of witnessing to such a bureaucratic, middle-class lifestyle as the most members of the crew were from doorkeeper families. In the summer, the working time was postponed until sunset at 8:30 pm and went on until 11:30 pm because they waited for the sunset. The shift in the crew’s timing in the summer depends on their preference of ‘working’ in the darkness and detecting which/whether or not flat is empty according to the lights switched on/off. The crew’s temporal decision to offend was affected by the darkness as they believed it covered them effectively and the lights give clue about occupancy. Preparation The crew set out to move when darkness fell. The crew was usually ready with four to five members around locations either in the park or bachelor pad when the hour came. At the time they had no car, a possible vehicle might be a taxi or dolmuş [bus] to the preferred burglary area. Once Muzo had his own car, it served as the crew’s crime transport around the city and facilitated moving into different neighbourhoods. Muzo used to take the car and pick up the rest of the crew. In earlier times, the crew did not take on any extra equipment, neither did they wear gloves. In the first instance, the only person who had criminal record was Maddy: He had not known at the time how to prevent fingerprints when touching a flat surface by wrapping his hand up in his shirt or tucking his hands into his jacket pocket while holding something. Once most of the members gained experience in criminal justice procedure, the crew subsequently began carrying gloves, having become aware of crime scene techniques. Criminal justice and prison experiences reinforced the crew’s habitus, cognitively directing them to attend to the immediate environment and take necessary precautions. For instance, where they were without gloves, the crew used socks to cover their hands. The sock method was particularly useful in the spring and summer as wearing gloves would draw suspicion. These socks were hidden in the car’s glove box in the central console next to the driver’s seat. The other complementary equipment was a long metal screwdriver and a sort of crowbar the brand of which is İzeltaş (a Turkish brand). The screwdriver was often placed together with the socks in the partition of the driver door, more usually, it was buried in the boot of the car just under the spare tyre. Frustrated with past experiences and giving direction to the future, criminal habitus as a generative capacity produced many different methods and techniques to secure themselves for work; this doubtlessly rendered their actions more professional. Selection of target neighbourhood Once everyone got in the car and all the equipment was ready, Muzo used to decide where to go, with some other crew members only occasionally joining the conversation: We get in the car, where are we going to go tonight? I say ‘let’s go there’.… We get in the car and go there. (Muzo) Muzo decides where to go as the leader of the crew. Where they go appears to be an arbitrary decision taken on the spur of the moment. The only factor considered is not going to the same neighbourhood on successive days. Their favourite places were the upper-middle class neighbourhoods in Çankaya province where they mostly landed big scores of cash and gold. Therefore, the crew mainly moved around these areas and safeguarded themselves by changing their route frequently. What is crucial is that the crew travelled by car and could therefore access entirely new and different areas, and were not confined to their neighbourhood where they know almost all the inhabitants—and where the inhabitants knew them. Deciding which street to target was as arbitrary as the choice of area. Muzo pulled over wherever he felt was appropriate. Each took up his gloves or socks as well as the screwdriver before leaving. All identity cards and phones were left in the vehicle. Usually, the screwdriver was carried by the youngest person present, tasked to perform lookout duties, but this responsibility varied as Muzo sometimes undertook it. Once getting out of the car, each attempted to act quite normally as if residents of the destined neighbourhood. This stage is where social skills are put into drama (Letkemann 1973; Åkerström 1985). Relaxed, calm, unsuspicious behaviours were performed; tough accents and postures were muted. The crew’s capacity to be socially adaptable is an example of ‘code-switching’ depending on circumstances (Anderson 1999), which is a critical part of their criminal capital. The crew would move a few streets away from the car. The more they distanced themselves from the car, the safer it became for them, as the contents of the vehicle—and the car itself—might reveal both their connection with one another and true identities. Selection of target flat Casing the target domicile begins with a short walk past it on the pavement, but it never took longer than 5 minutes. The aim is to plunge into one of the backyards as soon as possible. Once pulling over and diving into the street, the biggest failure of people is walking in the open. We never walk down the street but, if any, make it on the pavement. Why? If a patrol car appears far ahead, then we get down behind a car. If you walk in the open, the patrol can see you from miles away. Never appearing in the street, we immediately dash into the backyards. Then we explore a flat. (Muzo) The form of multistorey houses [apartman] appears to be a unique product of the urbanization process in Turkey. The architectural type of apartment buildings are exemplified by four to five storeys with back gardens without ornamental and decorative design. Multistorey buildings are built to a regular pattern in orderly rows with a certain distance between each other and have often green spaces and gateways. The crew used narrow alleys between buildings as a route into back gardens. Here which alley or gateway is appropriate is determined by instantaneous reasoning (Nee and Meenaghan 2006; Nee and Ward 2015) as to ease of transition and the topographical conditions at which the apartment building is located. Regarding the narratives of participants, it seems not to have taken much time to decide which route should be taken. The experiences of numerous past break-ins have already provided the crew with a mental schema (Bourdieu 1990), which are called ‘automatic scripts’ (Cromwell et al. 1991; Wright and Decker 1994; Nee and Meenaghan 2006), wherein they can measure danger, relevancy and ease. The crew’ script of burglary was developed in experiencing the same type of building but only different regarding the topographical conditions (Mercan 2018). The aim of the crew was to reach backyards at which point they begin to select a target. Mental criminal schema guides the crew to make appropriate moves depending on the degree of difficulty in reaching the backyard. Gaining access to the back of the apartment buildings could take place in various ways: simply walking through the front doorway if and when possible, quickly jumping into the front garden and running into the back, or climbing a wall and creeping through shrubbery. Briefly, the topographical conditions resolve which move is to be made. The crucial point in making any move is just to remain invisible and silent. Once the crew reached the backyard, the target search was launched; this stage is where the ‘commando moves’ begin. Shover notes that burglars likened the offence to ‘military commando operations’ (1991: 91–2). More than being metaphorical, the crew’s behaviour comes to resemble a military operation. Under Muzo’s guidance, the rest of the crew used to jump walls, leap over barbed wire and pass through shrubs and hedges, if necessary. There are two rationales for using commando moves. First, without going out into the open, the crew intended to remain undetected throughout the target-selection process. Second, the searching process often took place in the hilly, fenced, rugged terrain of the backyards. To exemplify, while the crew was casing an empty flat, one member might see a more appropriate target in the next garden. If there were no direct entrance but a long private fence enclosing the backyard, then they would have to jump over it. The same applies to another form of barrier, such as locked gates or barbed wire atop fences, which entail them to either creep or pass through. Mostly, the crew members preferred to wear tracksuits and trainers, which allow them to act comfortably and to escape at speed in the event of danger. As a result, commando operations indicate to a necessary capital of burglary: physical criminal capital (Steffensmeier and Ulmer 2005). All the crew members had athletic bodily traits: slim and muscled enough to pass through narrow fences and run away quickly and strong enough to lift anything heavy such as a safe. A fit body is a sine qua non for ‘commando moves’ in the backyards until a proper target is selected. The actual target-selection process of the crew can be explained using three risk cues: ‘surveillability’, which refers to the visibility of the target to neighbours or passers-by; ‘occupancy’, which indicates whether or not residents are present in the target home and ‘accessibility’, which denotes how easily the home can be broken into and whether or not the home includes locks, burglar alarms, bars and dogs (Cromwell et al. 1991: 35–7; Wright and Decker 1994: 85–99). As the crew worked in the backyards of buildings, the crew attended to whether: (1) the light is switched on or off, indicating whether or not the residents are at home—occupancy; (2) flat is on the ground floor or upper floors—accessibility and observability; (3) there are bars protecting the windows and doors or not—accessibility. Exploring a suitable flat thus depends on some criteria: In the closed-system, you break in all together. You can stroll in gardens and explore. You reason: five windows is one flat. Kitchen, lounge and three rooms. You identify it first. Then you check if the light is switched on. (Lordy) Well, [it] depends on what the facade of the flat looks [out onto]. It depends on if it’s empty. We often prefer flats facing the backyard, or garden-floor flats [ground floor], or upper flats just above the garden floors. Or if you can climb or can easily run away, the second floors might be done as well, but it is quite rare because you can’t make it easily from the second, because you have to make it down to the first floor, only then jump over the soil. Of course, difficult. That’s why you prefer flats you can quickly hit and run away. (Alex) Those flats that were vacant and facing the backyard preferably at the garden (ground) floor seem to have been the crew’s predominant choice of target as those flats remain directly unobservable due to their location. The crew used to hit the ground floor targets in its early years. However, in time, the crew began hitting the upper floors; this was due the growing courage and self-confidence of its members as well as their physical criminal capital (see Steffensmeier and Ulmer 2005: 129). Their encroachment to the upper floors also stemmed from the crew’s preference of stealing cash and jewellery. They disregarded goods and items such as TV, stereo and furniture; the aim was only on extracting cash and gold. With this aim, climbing up and down became easier for the intruders searching of things light in weight but heavy in value. Thus, increasing criminal insight and self-confidence professionalized the crew in a way that allowed them to target upper floors, taking more risks. Division of labour The crew placed and organized themselves once the preferable target had been selected. Muzo tasked each member with a strategic position depending on the physical conditions of the backyard: I leave two or three lookouts. The other two are getting in. I sometimes go in as a third. But I often used to stay on the outside. I used to go on entering in later periods. But I usually didn’t enter. I’m telling my bro’; there was no mobile. One whistle is ‘watch out’, two whistles ‘someone coming up’, three whistles ‘run away’. That was the code. We attended to it very much, regardless of whatever happens. Then I left one at one corner, the other for another corner, me strolling around the building … Alex was a lookout, Lordy and Reddy intruders, Maddy sometimes intruder, but often on the outside. (Muzo) In the organization of division of labour, Muzo played a coordinating role between the lookouts and intruders, moving in and out of the house. Maddy also occupied the same position playing a combined role inside and out but mainly staying outside with Alex, who was specifically tasked as lookout. The crew members started carrying two mobile phones as of 2002–2003. One would be in the intruder’s pocket while the other was ready in the lookout’s hand who called the intruder in the event of danger. The deployment of the crew was strategically organized: Dangerous spots are obvious. As you’re in the rear of the building, there is always a gate straight down to the garden, always a doorkeeper flat [looking out or opening onto the garden]. You put the men called ‘lookout’ at these dangerous spots. There might be experts among them. You put them [there] to watch more dangerous spots where someone may just walk in. (Lordy) In this organization, the position of lookouts becomes quite important. Alex, the lookout of the crew, mentions a specific mental schema that a lookout should have, which ‘other’ people cannot simply possess: When you’re at it, you choose places people don’t prefer to look at. Some people wouldn’t prefer to look at some places [that are ugly, smelly or dirty]. For instance, while walking down the street, you look right or left but never see some spots. There are some blind spots; after walking past before those spots twice or three times, you never look at them. A good lookout is the one who knows those places well and stands there. To do his job well, he has to know them. Where are those places? So simple. Where do people dislike glancing at subconsciously? One never wants to look at a dump, or dirty spot, let’s say a ‘someone-coming-for-a-pee’ spot. That’s why you can wait in a dark corner, or obscure nooks and crannies. People never [do more than] glance at them while going past. Those who will look at these nooks and crannies are mostly the ones whose mentality is different from that of others. Well, mostly ones who did burglary before looking at these spots. I mean those having done lookout. (Alex) Alex’s comment on the necessary capacity of a lookout demonstrates one subtlety of being a professional burglar: being mentally selective in perceiving and selecting crucial spots. The intuition of ‘where to look and how to hide’, as Alex’s account suggests, is a result of a trained capacity, practiced over and over, directed by Muzo’s instruction of ‘seeing without being seen’. It designates an intrinsic capital of the lookout and starkly divides the perspective of a professional burglar/lookout from that of an amateur or ‘ordinary’ person. The importance of this trained skill functions to protect the whole crew and prevents ‘blowing the score’, while the intruder is inside the residence calmly searching for booty. Breaking Once the position of lookouts is strategically defined, the phase of breaking and entering starts. In the case of the crew, the methods do not involve sophisticated forms such as deactivating alarm systems and neutralizing dogs as the Western literature on burglary has observed (Scarr 1973; Bennet and Wright 1984; Åkerström 1985; Cromwell, et al. 1991; Wright and Decker 1994, Rengert and Wasilchick 2000; Nee 2015). The architectural and spatial design of buildings, however, has far-reaching impacts on the perceptions and decisions of the crew to offend (Brantingham and Brantingham 2004; Bernasco and Nieuwbeerta 2005). The break-in method of askı depends on the architectural style and form of deterrents of the multistorey apartment buildings. The form of entry and necessary capital depends sequentially on whether: (1) the flat is at ground level or on an upper floor; (2) there are window/door bars or not; (3) the window/door is wooden or PVC. First, if the flat targeted is located on the first or upper floor, the crew would look for a natural route using drainpipes, balcony railings and bars installed on ground-floor windows. The crew confirm that these bars or grilles are intended to protect the flat but often offer a path to the upper floors. If these bars allow them to climb to the first- and second-floor balconies, then they could readily reach the upper floors. However, varying degrees of physical and cultural criminal capital individually determine who of the members would be the climber and therefore intruder. Atom (the interim member) and Lordy come forward as the usual intruders leading the way and having higher levels of criminal capital. ‘Heart’, or courage, is the primary determiner (see also Steffensmeier and Ulmer 2005: 129). Without courage, entering a flat is not possible as Atom and Lordy indicate. However, as their accounts suggest, there is more to being an intruder than courage. Here pulling one’s body weight up using the window grilles and balcony railings is another critical part of physical criminal capital that allows the intruders to overcome any of these physical deterrents. Nonetheless, as with the commando moves, not everyone can do it easily. It requires one to be able to pull all bodily weight up, hold on to the bars and, when necessary, pass through window transoms. Due to their physical capability, Atom and Lordy, and much later, when the former two were in prison, to some extent Reddy and Muzo, undertook the intruder position. Differences in individual physical traits seem to have played a decisive role in the division of labour of the crew. However, the internal organization is not simply determined by individual capitals. To exemplify, should the flat selected be above 2.5 m above the ground without any ‘natural passageway’, then Atom or Lordy, whose bodies were slim and flexible enough, tried to reach the upper floors. By mounting the shoulders or leaning on the backs of physically more solid members, Muzo and Alex, the intruders try to climb the balcony. They used to forge a body escalator so that the intruders could grab the window poles or balcony bars to climb up to the flat. The crew were able to form a human scaffold using various acrobatic moves, if and where necessary. Here, collective criminal capital was put into practice by all the members together. Second, Lordy, Atom and later Muzo, as an intruder, took over the break-in process and used a screwdriver. Although it seems a quite individual operation, the type of deterrents faced requires the collective force of the crew members: Look! You know what we’re doing? We’re breaking [in]! Q: Alright, but how did you break in? We’re pulling the iron bars like that. We’re going out to work three to four guys. Pulling the window bars [high pitch]! Pull and leave it! Pull and leave it! (Maddy) If the flat was at the ground level in the backyard and the window was secured by an iron frame installed into the concrete surface of the building, all members would come together, brace their feet against the wall and dislodge the whole frame by pulling from the middle lower part of it. However, as they explain, the iron frame needs not be removed completely, because dismantling it fully is often very noisy. If some bars are twisted enough, then the intruders, Atom and Lordy, could go through the space opened. If not, it can be pulled off entirely, as getting rid of it completely would make the escape easier. This was the other moment when the individual strengths of the crew members turned into collective capital. After this operation, the lookouts Alex and Maddy return to their positions. Third, the final deterrent is whether or not there is a padlock on the window and what type of window it is they encounter. The intruder might come across a vault lock on both door and window bars. He gets through this security measure using a long İzeltaş-brand screwdriver, inserting it through the shackle and applying strong pressure diagonally until the metals break loose from the core. Muzo could easily and quickly tackle these locks due to his bodily power and also help Lordy or Atom to split them. Windows usually have a lock in the middle, right? I mean wooden windows. You can open it through prying up the one lock in the middle … PVC windows contain three to four round locks. All of them are embedded in the opening side of the window. Because PVC is very flexible plastic stuff, it is enough to displace them by slightly prying them up with a screwdriver. That’s assuming that the PVC window opens the moment you’ve pushed it by inserting the screwdriver through either above, or below, or middle of the window and its frame. The same goes for PVC doors, whereas wooden doors give you difficulty as it takes longer. If there is a key back on the inside facade of the door and if it’s locked, you can reach to the key and open it by breaking the corner of the window as little as you could stick your hand in. If it’s not locked well, then you can already open [it] and enter. (Alex) Prying or jimmying open the door and window appear to be the same technique of breaking long observed by the Western literature of criminology (Shover 1973; Repetto 1974; Walsh 1980; Bennet and Wight 1984; Cromwell et al. 1991; Wright and Decker 1994; Nee and Taylor 2000; Nee 2015). With PVC, a window and door can be opened simply by prying the screwdriver between the window/door and its frame, whereas a wooden window can be opened by crudely pushing on its frame. However, wooden doors are not simply opened by a crude kick-in method. Instead, it requires the intruder to cut the glass. In this technique, first, part of the window glass close to the key is inscribed using the tip of the screwdriver, drawing an arc from edge to edge of the frame. Second, the intruder strikes on the portion of the glass enclosed within the arc and frame, cleanly breaking this portion of glass away from the main body of the pane. The partially broken glass creates an opening large enough for a hand to reach in and remove the key. Doubtless, this method is trickier and needs some extra time to complete successfully and quietly. Entering Having removed all deterrents and opened the window or door, Atom (1999–2001) and Lordy (1998–2002), later Reddy (2001–2003), used to enter the flat as the primary intruder. The number of intruders inside would usually be limited to one or two, and would never go beyond three, because Muzo also sometimes accompanied them as a coordinator between inside and outside. Immediately before searching for booty, the intruders would silently investigate the house to ensure no one is at home and lock the front door by securing the hasp over the loop of the fastening: You have entered and started checking out rooms. If nobody is inside, then first lock the exit door to prevent the family from entering the home and to be able to hear that sound when they attempt to open the door. Well, you ransack the home then. (Lordy) The priority was to find cash and jewellery, and the search starts in the master bedroom: First, the master bedroom will be usually ransacked because the head of home is staying there. Jewellery is often in there. You give a look at the lounge and kitchen as well, but it is often less preferred. Kids’ rooms are usually not searched. Once realising it is an adolescent room, you do it then because a girl might have jewellery as well … you ransack every bedroom in the house. You search everywhere. You look at the lounge too. You can search in anywhere you feel you will find something valuable. There are ones having extracted money out of a freezer in the fridge. Just have a think. I took it out of a shirt from a wardrobe. The guy had hidden it by giving it the shape of a gift. It even comes out from under beds. (Alex) The master bedroom is the primary location of search in relevance to the finding of Western literature (Bennet and Wright 1984; Cromwell et al. 1991; Wright and Decker 1994; Clare 2011; Nee 2015). The crew realized big scores so often from the master bedroom, as Alex says, either from the dressing table or wardrobe. While looting the house, the intruders switched the light on to give the appearance that the occupants were inside. As the accounts put it, the intruders expedited the search process inside the dwelling, taking no longer than necessary. Their priority was cash and jewellery—except for mobile phones, laptops, televisions and other electronic devices were not what they were looking for. During 1998–2000, very often cash was found, so the crew didn’t need to look for gold and jewellery. Between 2000 and 2003, they stress that their attention turned to finding jewellery as the cash scores declined. Departure After comprehensively ransacking the home, the entry point was often used for departure. If a good deal of cash and jewellery had been extracted, the crew ended the night’s work. However, if the quantity of score was lower than expected, such as a single fine necklace or gold rings or a small amount of cash, then the crew continued to search for another target using the same method and techniques illustrated earlier. The average number of flats the crew ‘hit’ might amount to a succession of five to six per night. The collective division of labour provided the crew with ‘social facilitation’ reducing perceived risk and hitting the appropriate targets successively (Cromwell et al. 1991: 70). As long as the intruders returned empty handed from the flats, the crew kept going, but as a rule this would not often go beyond 9:30–10 pm. The accounts suggest that the crew often continued burgling in the same neighbourhood originally targeted as long as the neighbourhood stayed calm and no job was blown. If homeowners returned and neighbours became aware of the thefts, the crew quickly changed their location by car. Once finished, Muzo took everyone home, or they all would go back to the bachelor pad. The open-door method In the summer, the break-in method differs from that used in winter. The open-door [açık kapı] is put into practice during the summer months when the temperature is quite hot, and people leave their doors and windows open during the night. The technique rests on the ability to sneak through open windows or doors and creep into rooms without awakening sleepers. The intruder collects purses and wallets while the household is in a deep sleep. The schedule can start at 11 pm, but it is preferably carried out when all people are heavily asleep around 3–4 am. For this reason, on some nights during many summers the crew did not sleep. The modus operandi of open door resembles the method used by ‘cat burglars’ who are called ‘climbing thieves’, targeting windows on upper storeys left open at night (Walsh 1980: 31). The job carried much ‘prestige’ as it requires a higher criminal capital to climb up the pipes, pillars and balconies of multistorey buildings, wearing black clothes and appropriate shoes for slippery surfaces in pp. 31–3. Similarly, the open-door method was undertaken by only Atom and Lordy. They normally went out to work after dark at 8.30–9 pm in the summer; using the askı or closed-door [kapalı] method, it was not uncommon for Lordy or Atom—once realizing somebody was inside asleep—converted from the askı to the açık kapı method, while the others were waiting for them outside. As the closed-door work would typically extend late into the night, Atom or Lordy began looking for doors left open. They then targeted first and second storeys whose balcony doors were left open and to which they could climb. The open door was much more profitable than the closed system but also riskier as it might place the intruder in confrontation with the resident. However, it provides the opportunity for accessing cash directly. A lookout might be useful but is not necessary as most people are asleep after midnight: You should be calm at that moment. Well as if being quite normal, climbing like it’s your home, going past rooms to the saloon. Act as if it was your home! The logic is the same. Climb, enter the room; this is the bedroom, creep out; opposite is the bathroom, next to it, another room facing the kitchen, and there is a door opening out of the building.… It depends on a person’s character; you’ve got to be like that it’s usual. Easy and relaxed. While confronting a problem, you must be relaxed, you’re able to deal with it easily. This is not something like you can teach – like do this, hold that, or enter the door; like if you do that, they never hear you. There is not something like that! Well, you can develop by yourself.… Always the same logic. Enter, be quiet, athletic, be all ears, then do it, and achieve your goal! It all depends on that outright. (Lordy) Lordy was the most experienced and daring intruder of the crew. Beyond physical abilities such as climbing and pulling his body weight up, Lordy also points out a mental capacity having developed over time: being relaxed and acting as if the flat entered was the intruder’s home. It seems that such a perception, rather than being a trait of personality as Lordy emphasizes, ensues from the sedimentation and naturalization of certain dispositions after having been practiced numerous times. Countless repetition of criminal practice, the open-door, normalizes an extraordinary state of feeling in a way as to make it a rare asset in the subfield of burglary. Given the fact that habitus is the naturalization of social conditions and conditionings in practice, the internalization of burglary-related dispositions seems to neutralize not only guilt (Sykes and Matza 1957) but also fear and anxiety emanating from criminal action (Tunnell 1992: 82). Thus, the other necessary part of criminal capital is the skill of how to handle emotions (Wright and Decker 1994: 110). That refers to the higher criminal capital of some crew members and the higher level of professionalism of the crew in the burglary. Concluding Remarks: The Collective Professionalism of the Crew Apparently, the crew as a homogeneous unit practiced a specific burglary technique continuously and systematically with a certain division of labour and under the coordination of a leader, travelling far afield and combining individual physical capitals in a collective manner. During operations, their bodily and mental dispositions acquired early on first turn into collective criminal capital at work, and then economic capital, should a score be made. It is demonstrated earlier that when the crew were offending, they systematically and instantly carried out the stages of burglary in a collective division of labour. The lookouts and intruders automatically recognize the offence-related cues about the tasks each fulfils in the division of labour. Deterrents such as window bars and reaching upper floors are overcome by collective bodily capital. Here individual cognitive expertise, acquired by repeated exposure to the same practice, thereby inscribed in criminal habitus, turns into collective professionalism in action: Were we professional? Yes, very professional. We did this job for more than four, five years. Four to five years is not a little time, bro’! To be professional, you start a job, work for a year, and learn every minute detail or any shit to do with that, let’s say, in a private sector. In the second year, you’re getting to be a master. When it comes to the third year, you’re much better, that is, you start thinking you’re better than anyone else. But fourth and fifth is the thing for you. That work has got routine then. It comes to you so simple. You try something different, taking different scores. This [burglary] was like that. In the fifth [year], we used to go every evening like going out to buy some bread! (Alex) Trained break-in practices of the crew created a disciplined division of labour that prevented them from being caught. Countless break-ins for 5 years made burglary an ordinary practice in their everyday life. Burglary thus became an alternative source of income and thus an occupational activity. The unconsciously evolved practices turned out to be a deliberate professionalized performance. On the other hand, the findings have also demonstrated that the topographical structure of neighbourhoods and multistorey apartment buildings influence the way of organizing burglary, target-selecting process and the types of deterrents in the metropolitan cities of Turkey. The intrinsic architecture of apartment buildings and environmental conditions also seems to overdetermine the modus operandi of housebreaking in a non-Western context. However, it has appeared that there are many similarities between the Western context and the non-Western Turkey in terms of the risk cues considered in the target selection, the use of screwdriver and the primary location of search. The author completed MSc Social and Political Theory in Birkbeck, University of London, then received PhD in Sociology from University of Kent at Canterbury. He is currently Research Fellow in the Faculty of Political Science in Ankara University. References Åkerström , M . ( 1985 ), Crooks and Squares: Lifestyles of Thieves and Addicts in Comparison to Conventional People . Transaction . Anderson , E . ( 1999 ), Code of the Street: Decency, Violence, and the Moral Life of the Inner City . W. W. Norton & Company . Bennett , T. and Wright , R. T . ( 1984 ), Burglars on Burglary: Prevention and the Offender . Ashgate Publishing . Bernasco W. and Nieuwbeerta P . ( 2005 ), ‘ How Do Residential Burglars Select Target Areas ?’, British Journal of Criminology , 45 : 295 – 315 . Google Scholar Crossref Search ADS Bourdieu , P . ( 1977 ), Outline of a Theory of Practice . Cambridge University Press . —— ( 1990 ), The Logic of Practice . Stanford University Press . Bourdieu , P. and Wacquant , L. J . ( 1992 ), An Invitation to Reflexive Sociology . University of Chicago Press . Brantingham , P. L. and Brantingham , P. J . ( 1993 ), ‘ Nodes, Paths and Edges: Considerations on the Complexity of Crime and the Physical Environment ’, Journal of Environmental Psychology , 13 : 3 – 28 . Google Scholar Crossref Search ADS —— ( 2004 ), ‘ Computer Simulation as a Tool for Environmental Criminologists ’, Security Journal , 17 : 21 – 30 . Clare , J . ( 2011 ), ‘ Examination of Systematic Variations in Burglars’ Domain-Specific Perceptual and Procedural Skills ’, Psychology, Crime & Law , 17 : 199 – 214 . Google Scholar Crossref Search ADS Clarke , R. V. G. and Cornish , D. B . ( 1985 ), ‘ Modelling Offenders’ Decisions: A Framework for Policy and Research’ , in M. Tonry and N. Morris , eds, Crime and Justice: An Annual Review of Research , 147 – 85 . University of Chicago Press . Cohen , L. E. and Felson , M . ( 1979 ), ‘ Social Change and Crime Rate Trends: A routine Activity Approach ’, American Sociological Review , 44 : 588 – 608 . Google Scholar Crossref Search ADS Cromwell , P. , Olson , J. and Avary , D . ( 1991 ), Breaking and Entering: An Ethnographic Analysis of Burglary . SAGE . Felson , M. and Boba , R. L . ( 2010 ), Crime and Everyday Life , 4 th edn. SAGE . Irwin , J . ( 1970 ), The Felon . Prentice Hall . Letkemann , P . ( 1973 ), Crime as Work . Prentice-Hall . Logie , R. , Wright , R. and Decker , S . ( 1992 ), ‘ Recognition Memory Performance and Residential Burglary ’, Applied Cognitive Psychology , 6 : 109 – 23 . Google Scholar Crossref Search ADS Mack , J . ( 1975 ), The Crime Industry . Saxon House . Maguire , M. and Bennett , T . ( 1982 ), Burglary in a Dwelling: The Offence, the Offender and the Victim . Heinemann . Mawby , R. I . ( 2001 ), Burglary . Willan Publishing . Mercan , B. A . ( 2018 ), ‘ The Making of Professional Criminal in Turkey ’, Ethnography , doi.org/ https://doi.org/10.1177/1466138118779604 . Nee , C . ( 2015 ), ‘ Understanding Expertise in Burglars: From Pre-Conscious Scanning to Action and Beyond ’, Aggression and Violent Behavior , 20 : 53 – 61 . Google Scholar Crossref Search ADS Nee , C. and Meenaghan , A . ( 2006 ), ‘ Expert Decision Making in Burglars ’, British Journal of Criminology , 46 : 935 – 49 . Google Scholar Crossref Search ADS Nee , C. and Taylor , M . ( 1988 ), ‘ Residential Burglary in the Republic of Ireland: A Situational Perspective ’, The Howard Journal of Criminal Justice , 27 : 105 – 16 . Google Scholar Crossref Search ADS Nee , C. and Ward , T . ( 2015 ), ‘ Expertise and Offending: Themes for Future Research ’, Aggression and Violent Behavior , 20 : 92 – 3 . Google Scholar Crossref Search ADS Nee , C. , White , M. , Woolford , K. , Pascu , T. , Barker , L. and Wainwright , L . ( 2015 ), ‘ New Methods for Examining Expertise in Burglars in Natural and Simulated Environments: Preliminary Findings ’, Psychology, Crime & Law , 21 : 507 – 13 . Google Scholar Crossref Search ADS Nee , C. and Taylor , M . ( 2000 ), ‘ Examining Burglars’ Target Selection: Interview, Experiment or Ethnomethodology ?’, Psychology, Crime & Law , 6 : 45 – 60 . Google Scholar Crossref Search ADS Rengert , G. and Wasilchick , J . ( 2000 ), Suburban Burglary: A Tale of Two Suburbs . Charles Thomas . Repetto , T. A . ( 1974 ), Residential Crime . Ballinger Press . Scarr , H. A . ( 1973 ), Patterns of Burglary . Government Printing Office . Shover , N . ( 1972 ), ‘ Structures and Careers in Burglary ’, The Journal of Criminal Law, Criminology, and Police Science , 63 : 540 – 9 . Google Scholar Crossref Search ADS —— ( 1973 ), ‘ The Social Organization of Burglary ’, Social Problems , 20 : 499 – 514 . Crossref Search ADS —— ( 1991 ), ‘ Burglary ’, Crime and Justice , 14 : 73 – 113 . Crossref Search ADS Shover , N. and Honaker , D . ( 1992 ), ‘ The Socially Bounded Decision Making of Persistent Property Offenders ’, The Howard Journal of Criminal Justice , 31 : 276 – 93 . Google Scholar Crossref Search ADS Steffensmeier , D. J. and Ulmer , J. T . ( 2005 ), Confessions of a Dying Thief: Understanding Criminal Careers and Illegal Enterprise . Aldine Transaction, a Division of Transaction Publishers . Sutherland , E. H . ( 1937 ), The Professional Thief . University of Chicago Press . Sykes , G. M. and Matza , D . ( 1957 ), ‘ Techniques of Neutralization: A Theory of Delinquency ’, American Sociological Review , 22 : 664 – 70 . Google Scholar Crossref Search ADS Tunnell , K. D . ( 1992 ), Choosing Crime: The Criminal Calculus of Property Offenders . Nelson-Hall Publishers . Wacquant , L. J . ( 2004 ), Body & Soul . Oxford University Press . —— ( 2011 ), ‘ Habitus as Topic and Tool: Reflections on Becoming a Prizefighter ’, Qualitative Research in Psychology , 8 : 81 – 92 . Crossref Search ADS —— ( 2014 ), ‘ Putting Habitus in its place: Rejoinder to the symposium ’, Body & Society , 20 : 118 – 39 . Crossref Search ADS —— ( 2016 ), ‘ A Concise Genealogy and Anatomy of Habitus ’, The Sociological Review , 64 : 64 – 72 . Crossref Search ADS Walsh , D . ( 1980 ), Break-Ins: Burglary from Private Houses . Constable . Wright , R. and Decker , S . ( 1994 ), Burglars on the Job: Street Life and Residential Breakins . Northeastern University Press . Wright , R. , Logie , R. H. and Decker , S. H . ( 1995 ), ‘ Criminal Expertise and Offender Decision Making: An Experimental Study of the Target Selection Process in Residential Burglary ’, Journal of Research in Crime and Delinquency , 32 : 39 – 53 . Google Scholar Crossref Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). All rights reserved. For permissions, please e-mail: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Xu,, Jianhua;Jiang,, Anli

doi: 10.1093/bjc/azy018pmid: N/A

Abstract Using data collected from systematic social observation, participant observation, interviews, and content analysis of media reports in Guangzhou, this article studies violent confrontation between China’s para-police chengguan (urban management officers) and street vendors. We find that most violent confrontations occur between street vendors and civilian staff—auxiliary chengguan—rather than between street vendors and sworn officers—official chengguan. We further reveal that the unequal power structure within the chengguan system shapes the division of labour between official chengguan and auxiliary chengguan, resulting in most of the ‘dirty work’ of street-level law enforcement being conducted by the latter, the second-class staff in the system. The research contributes to our understanding of pluralised policing and how police civilianization affects a convergence of violence within the underclass as both auxiliary chengguan and street vendors are recruited from urban poor. Introduction One of the key features of contemporary social life is the reach of markets and market logic into areas that were traditionally governed by nonmarket norms (Sandel 2013: 7). Once mainly regarded as public goods and provided by governments, policing and security are increasingly becoming commodities to be purchased on the market (Loader 1999; Diphoorn 2016). The public police are actively engaged in market activities to sell their own services and apply market logic of cost-saving in managing police force, a trend termed as marketization of the public police by scholars (Loader et al. 2014). On the one hand, along with rapid increase of private policing and security, some services provided by public police become commodified. Not only the police may charge extra fees in some areas (e.g., shopping malls) or some special events (e.g., sports, concerts) for extra patrol in the form of user-pay policing (Crawford and Lister 2006), off-duty police officers may also work for private employees with the full panoply of police powers in the practice of moonlighting (Lippert and Walby 2013). To generate funds for the extra budget, the police may also accept commercial sponsorship (Grabosky 2007; Xu 2013), and in some controversial cases, commercial advertisements can even appear on police cars (Sandel 2013: 195). On the other hand, the market logics of cost-saving and efficiency are widely adopted by public police in their daily work. Two most obvious practices include police outsourcing some key service areas to private sectors (White 2014) and the civilianization of police forces (Davis et al. 2014; Cosgrove 2015). Mainly with the rationale of cost-effectiveness, the civilianization of policing has reached an unprecedented level of over 30–40 per cent in many Western countries (including the United States, the United Kingdom and Canada) where neo-liberal thinking has triumphed since the 1980s (Davis et al. 2014; Kiedrowski et al. 2017; Lumsden and Black 2018). Despite various benefits of civilianization, there are several concerns including the poor quality, the low morale and high turnover rate of civilian staff, and the unequal power relationship between civilian staff and sworn officers. It should also be noted that we know little about how police civilianization occurs in former socialist countries such as China where the spread of market logic into social life came much later than its Western counterparts. Using the case study of China’s para-police chengguan, this article examines how police civilianization occurs in China, and with what consequences. Chengguan (i.e., urban management 城管) officers are China’s para-police who are responsible for the enforcement of urban regulations, and one of their main tasks is to regulate street vending in urban China (Hanser 2016). Through the study of violent confrontations between chengguan and street vendors in Guangzhou, one of China’s megacities, this article examines how the civilianization of para-police chengguan affects the occurrence of violent confrontation between street vendors and auxiliary chengguan (协管 xieguan, civilian staff) and, by extension, the violence within the underclass. On the one hand, the article enriches our understanding of unequal power structure between civilian staff and sworn officers and how this structural force shapes the concentration of violence within the underclass. On the other hand, this article extends current knowledge of pluralized policing and the para-police (Rigakos 1999) in the police extended family. Police Civilianization and Its Concerns Police civilianization refers to the practice of hiring nonsworn officers to replace or augment sworn personnel in police force, mainly with the purpose of reducing cost and improving service (Maguire and King 2004). Although there was nothing novel in recruiting civilians in the police department, the percentage of civilians in the police force remained at a very low level in most police forces until the 1980s when commodification, marketization and privatization of policing took shape under the influence of neo-liberal thinking (Jones and Newburn 2002). Currently, there is nearly one civilian employee for every two sworn officers in most police forces in the Western countries. In the United States, civilians in the police force have grown from 2.7 per cent in 1937 (Maguire and King 2004) to 7.5 per cent in 1950s and to over 30 per cent nowadays (Davis et al. 2014). In Canada by 15 May 2015, 29 per cent of the employees (or 28,368 staff) in the police force were civilians (Kiedrowski et al. 2017). In the United Kingdom, by 31 March 2016, among 200,922 total police employees, there were 61,668 (31 per cent) police staff and 11,043 (5.5 per cent) Police Community Support Officers (PCSO) (Home Office 2016). Both police staff and PCSO are civilian officers while the former are non-uniformed personnel who work mainly at the backstage of law enforcement and the latter are uniformed nonsworn officers who primarily work on the street to deal with minor offenses and antisocial behaviours (O’Neill 2014). Altogether nearly 40 per cent of British police personnel are civilian officers. In earlier days, civilians were hired mainly on the rationale of special skills rather than saving cost. However, cost-effectiveness became the predominant reason for the ever-increasing civilianization when neo-liberal thinking spread over the police (Loader 1999). The operations of the police are increasingly following a ‘business-paradigm’ to save cost (Ayling et al. 2009). When the demand for police work is constantly increasing but the expansion in budgets cannot catch up with the pace of social needs, and even worse when the police face severe austerity cuts (Lumsden and Black 2018), the extensive hiring of civilians to save costs becomes an inevitable choice. Civilians cost much less than their sworn counterparts, usually around 60–70 per cent of that of sworn officers (Davis et al. 2014). In some cases, the cost can be as low as one-half to one-third (Forst 2000). Indeed, empirical findings also reveal that police forces experiencing more budget constraints witnessed higher level of civilianization of their forces (Crank 1989). In addition to cost-effectiveness, advocates of police civilianization also emphasize that the practice can make police work more efficient and effective. They argue that the hiring of civilians can free the police from administrative, clerical and routine tasks to increase police professionalism by focusing on the core function of law enforcement for which coercion, arrest and other special police skills are needed (Matrofski 1990; McCarty and Skogan 2013). In addition, while the police are usually trained as generalists, civilians can bring special skills in forensics, intelligence analysis, information technology, training, budgeting, human resource and research (Davis et al. 2014). Civilians are also easier to lay off flexibly when there are budget constraints as they are contract-based (Alderden and Skogan 2014). However, there are various concerns surrounding police civilianization. The first is about the poor quality of many civilian officers. Although some civilian staff are equipped with special skills, many others may be less competent in the work as they are usually recruited from the bottom end of the labour market. The British PCSO were cynically regarded as ‘plastic police’ by some critiques (Paskell 2007). Similar to private security guards, the poor quality may also make civilian personnel more likely to become deviant in their work (Diphoorn 2016). The second concern relates to low morale and the high turnover rate of civilian staff. Civilians usually lack opportunities for promotion and career development. The turnover rate among them is usually high and the long-term retention of highly skilled civilian staff is particularly difficult (Alderden and Skogan 2014). The third concern is about the integration of the civilian staff and sworn officers. While some sworn officers may worry that civilians can take away a number of easy jobs with fewer risks (such as office work) which they regard as a reward or assignment for the sick, the injured, the aged and pregnant officers (Kiedrowski et al. 2017), others are concerned about the possibility of civilians’ low commitment to the police force (Dick and Metcalfe 2001). More importantly, civilian staff may not be regarded as genuine and valuable members by the sworn officers and be treated as second-class citizens in the police organization (Alderden and Skogan 2014). The unequal positions and power structure within the police organization deserve serious sociological inquiries. In the existing literature, some scholars examined how civilian staff work as emotional labour in the force control room (Lumsden and Black 2018), whereas others compared different levels of job burnout and commitment between civilian staff and sworn officers (McCarty and Skogan 2013; Alderden and Skogan 2014). Generally speaking, empirical research about police civilianization is rather limited. In particular, we know little about how the unequal power between civilian staff and sworn officers plays out and with what consequences. This article provides one of the very first empirical examinations of China’s police civilianization. Specifically, through studying China’s para-police chengguan, it reveals the unequal power relations between official chengguan (sworn officers) and auxiliary chengguan (civilian staff). It further reveals how such an unequal power relation contributes to a concentration of violence within the underclass as both auxiliary chengguan and street vendors are recruited from the urban poor. Interpersonal Violence Within the Underclass It has been widely recognized that offenders and victims often come from similar backgrounds. Interpersonal violence usually involves males attacking males, the young attacking the young, the ethnic minority attacking the ethnic minority, immigrants attacking immigrants and the underclass attacking the underclass (Downes et al. 2017: 287–304). Indeed, as Jock Young once vividly pointed out: What does your likely murderer look like? If you pick up a mirror and look into it, you will see the image of your most likely attacker. He will be of the same class as you, of the same ethnic group, probably the same age, a member of your own social circle – dressing like you, with the same accent and habits. Despite all the talk about intra-racial attack, he will be the same color as you (cited in Rock 1998: 16). Empirical research has consistently shown that in the United States violent crimes were much more likely to be concentrated within the African-American part of the population (Anderson 1999; Bruce 2000). In Australia, violence was also more likely to occur among aboriginal people (Polk 1994: 58). Similar findings were found in China where a report revealed that a socially disadvantaged group, rural-to-urban migrant workers, accounted for over 80 per cent of police arrests in violent crime incidents. They were also the major targets of homicide, aggravated assault, robbery and rape (Xu 2007). Although earlier scholarship tend to argue that the distinct subculture of violence among the underclass might be the main reason for the intra-class convergence of violence as people influenced by the violent subculture are more likely to resort to violence in their daily life to maintain dignity and power or to solve conflicts (Wolfgang and Ferracuti 1967), it has now been widely acknowledged that that concentration of violence is highly affected by social structural factors, such as the poverty, unemployment, segregation or discrimination experienced in everyday life (Blau and Blau 1982; Massey and Denton 1993; Venkatesh 2008). Lower class boys may lack the skills to succeed in schools compared to those from the middle class, therefore they develop their own criteria to succeed, such as being tough and possessing fighting skills, something which seems easier to achieve (Cohen 1955). Youth in the disadvantaged neighbourhoods may find it hard to achieve economic success because of a lack of legitimate means, and resort to violence to attain status accordingly. Some lower class people are not only permanently excluded from the job market due to deindustrialization in many Western societies, but are also excluded from the protection of the criminal justice system (Young 1999; Reiman and Leighton 2016). In some disadvantaged neighbourhoods, the underclass has increasingly become the target of over-policing, and their lives become centred on the avoidance of being caught by the police (Goffman 2014). The limitations of the police may also encourage more violence as offenders are not regularly punished (Desmond 2017). Where the underclass is often over-policed but under-protected by the police, they may become estranged from civil society and find themselves having to solve the conflict by themselves, and violence is always a readily available option (Anderson 1999; Venkatesh 2008; Kirk and Papachristos 2011). While the existing literature has explored how social structural forces make the underclass population more likely to use violence, they largely focus on offenders. The convergence of opportunities for the underclass to interact with each other, and lead to possible violence among them, is often taken for granted and unexamined. However, it is important to explore what social structural factors contribute to the convergence of likely offenders and possible victims within the underclass (Cohen and Felson 1979). This paper addresses this matter by studying how police civilianization in China contributes to the convergence of auxiliary chengguan and street vendors on the street, and by extension, affects the intra-class concentration of violence. Chengguan Violence and the ‘Rotten Apple’ of Auxiliary Chengguan China’s chengguan, officially known as the Urban Administration and Law Enforcement Bureau or simply the Urban Management Bureau, is a para-police government department tasked with enforcing non-criminal urban administrative regulations (Zang 2017). One fundamental difference between chengguan and the police is that chengguan have no arrest power, which is an essential element of police work. Chinese chengguan have had a relatively short history, having been established against a backdrop of soaring problems of urban management caused by the rapid urbanization of the past three decades (Hanser 2016). The first chengguan force was founded in 1997 in Beijing. By 2015, a total of 3,074 counties and cities in China had created their own chengguan departments and 455,000 personnel were employed. Official statistics showed that while half of them were sworn officers, another half were fixed-term (1–3 years contract) auxiliary chengguan (Liu 2016). The auxiliary chengguan are uniformed nonsworn civilian staff, similar to PCSO in the United Kingdom. The level of civilianization is much higher than that in Western countries. There is a huge difference between the two forces in terms of job security, power and compensations. Generally speaking, the cost of hiring an auxiliary chengguan is just roughly one-third of that for an official chengguan. Auxiliary chengguan become the second-class citizens in the system. They have no independent power to enforce the law but can call upon the support of official chengguan’s law enforcement. Since its inception, chengguan have become increasingly omnipresent in domestic governance. For instance, the number of tasks for Guangzhou chengguan’s law enforcement has increased exponentially from 5 in 1997 to 321 in 2015. The most commonly known tasks include the regulation of street vending, unauthorized construction, noise and air pollution. Regulating street vending, however, is probably the most widely known task of chengguan as it frequently involves the use of excessive violence in law enforcement (Human Rights Watch 2012). A report described chengguan’s primary law enforcement methods in controlling street vendors: ‘confiscation of goods, kicking vendors’ stands, throwing (vendors’) goods to the ground, gang-style beatings, and triad-style protection fee collection’ (Zhang 2011). An internal training manual of Beijing chengguan on how to use violence in law enforcement went viral online, which instructed chengguan to ‘leave no blood on the face, leave no trace of injury on the body, and keep away from witnesses’ when beating vendors (Zhang 2009). As a result, Chinese chengguan has been regarded as the most unpopular government department (People.cn 2014). A survey indicated that less than one-third of citizens were satisfied with chengguan law enforcement (Guangzhou Public Opinion Research Center 2013). Media coverage on chengguan violence against street vendors or bystanders has been widespread. For instance, on 17 July 2013, a male street vendor was beaten to death and his wife was seriously injured by six chengguan when selling watermelons in Hubei province (Ouyang 2013). On 22 April 2014, a bystander was attacked by a group of chengguan as he took some photos of the scene where chengguan were driving vendors away in Zhejiang province. Later on, chengguan even overturned an ambulance that came to rescue an injured bystander (Zhu 2014). On 16 May 2012, a street vendor was stabbed four times with a knife by a chengguan in Guangzhou (Jiang 2012). These cases are just a tip of the iceberg of mounting conflict between Chinese chengguan and street vendors or bystanders. In Guangzhou city, it was reported that there were over 2,626 violent confrontations between chengguan and street vendors from 2005 to 2009 and more than 1,679 chengguan were physically injured (Table 1). A search of ‘chengguan violence’ (城管暴力 chengguan baoli) in Google or Baidu in 2016 both generated more than one million reports. Table 1 Chengguan’s violent confrontations with vendors in Guangzhou from media reports Year Number of violent confrontation Number of chengguan injured 2004 236 2005 700+ 330+ 2009 (1–10) 230 86 2005–2009 2626 1679 2013 (1–8) 119 99 2014 390 Year Number of violent confrontation Number of chengguan injured 2004 236 2005 700+ 330+ 2009 (1–10) 230 86 2005–2009 2626 1679 2013 (1–8) 119 99 2014 390 View Large Table 1 Chengguan’s violent confrontations with vendors in Guangzhou from media reports Year Number of violent confrontation Number of chengguan injured 2004 236 2005 700+ 330+ 2009 (1–10) 230 86 2005–2009 2626 1679 2013 (1–8) 119 99 2014 390 Year Number of violent confrontation Number of chengguan injured 2004 236 2005 700+ 330+ 2009 (1–10) 230 86 2005–2009 2626 1679 2013 (1–8) 119 99 2014 390 View Large In chengguan violence towards street vendors, the civilian staff auxiliary chengguan have always been identified as the culprits by the government. The following case in Guangzhou is illuminating. On an early morning in May 2012, several chengguan came to a market prohibiting a vendor from selling vegetables on the sidewalk. Soon after, verbal quarrelling escalated into physical violence. A chengguan officer picked up a knife from another nearby vendor and assaulted the vendor four times, resulting in the vendor’s skull being exposed. The vendor’s possessions were also smashed to pieces on the ground. In response to this case, one government official remarked that: ‘The offender is an auxiliary chengguan. He is a temporarily-hired staff, not an official chengguan’ (Jiang 2012). In another case, a male bystander was beaten up by a chengguan as he was mistakenly believed to be shooting a video of the present scene of law enforcement. One official from the Urban Management Bureau responded: Three perpetrators will be punished. Their identities have been identified, and they all are temporary employees (auxiliary chengguan) with low education…we will strengthen the education and management of temporary employees to avoid similar incidents in the future. (Chen and Kong 2013) While the government often attributes the chengguan violence to the so-called ‘low quality’ of auxiliary chengguan, others argue that these temporary employees might be just the scapegoats for official chengguan (Qi 2013; Ji 2014). A survey indicated that 76.8 per cent of respondents believed that temporarily hired auxiliary chengguan ‘take a fall’ for official chengguan (Lan 2013). However, it remains uncertain that whether auxiliary chengguan are the majority of perpetrators or simply the scapegoats for official chengguan. In addition, when the state attributes these violent events to the ‘low quality’ among auxiliary chengguan, the constant use of violence in law enforcement appears to be depicted officially as incidents caused by individual ‘rotten apples’ in the chengguan force (Sherman 1978). However, the deep root of the problems of chengguan system has not been exposed. That is, how the civilianization of the para-police chengguan system affects the production of violence. Research Questions, Data and Method The primary purpose of this study is to examine the pattern of chengguan violence towards street vendors and the underlying social structural reasons for the production of chengguan violence. To achieve this goal, the following questions are addressed: (1) To what extent are the auxiliary chengguan the major perpetrators of violence towards street vendors? Or are they simply the scapegoats of official chengguan as many commentators suggest? The answer to this question is likely to reveal the extent to which violence occurs within the underclass, as both street vendors and auxiliary chengguan are commonly recruited from the urban poor; (2) if most violent confrontations occur between auxiliary chengguan and street vendors, how does the difference in the power wielded by the official chengguan and auxiliary chengguan contribute to the observed pattern of violence? Addressing these questions will shed light on some negative consequences of police civilianization. All data for this study were collected from Guangzhou, a megacity located in south China with a population of around 17 million. Multiple strategies were used in data collection during the research team’s (consisting of two researchers and an assistant) three years of fieldwork in Guangzhou. The first set of data came from systematic social observation (SSO) of chengguan law enforcement. SSO proves to be a powerful method in criminology (Reiss 1971; Sampson and Raudenbush 1999). However, compared with survey, interview and participant observation, SSO is an underdeveloped tool (Xu 2009, 2013; Xu et al. 2013). In this study, we developed this method by systematically observing 112 instances of chengguan law enforcement during the summer of 2014. The research team coded the following variables for analysis: time, date and place of law enforcement, duration of law enforcement (how many minutes the law enforcement lasted), number of official chengguan, number of auxiliary chengguan, number of street vendors on the street, incidence of verbal violence from chengguan and incidence of physical violence from chengguan. Throughout, our main purpose was to identify risk factors for the occurrence of chengguan violence. The second part was devoted to participant observation. The team spent altogether eight months on the street of Guangzhou observing street vending and chengguan law enforcement from July to October 2014, and July to August in 2015 and 2016. In particular, the second author also worked as a street vendor for a whole month in September 2014, either independently or working with other street vendors (Figure 1). Other members of the research team, including the first author and the research assistant, engaged in participant observation from time to time. These experiences enabled us to become familiar with street vendors and develop a deeper understanding of their work nature, risks, challenges and daily life. It also provided us with many opportunities to interact with chengguan. Fig. 1 View largeDownload slide Hundreds of street vendors selling goods on a sidewalk in Guangzhou in 2014 Fig. 1 View largeDownload slide Hundreds of street vendors selling goods on a sidewalk in Guangzhou in 2014 The third part of our data collection involved in-depth semi-structured interviews with chengguan and street vendors. Making use of the team’s extensive network established during a decade of research in Guangzhou, we first interviewed 32 official chengguan at the district level, and 20 official chengguan and 7 auxiliary chengguan from street-level law enforcement branches. These interviews were mainly conducted in a relatively formal setting, usually in a meeting room in chengguan departments. To diversify our source of information, we also interviewed an additional 4 official chengguan and 13 auxiliary chengguan in informal settings, mostly on the street in a relaxed way after the researchers had become familiar with them during the lengthy process of street observation. Altogether, we interviewed 76 chengguan. Besides chengguan, 31 street vendors were also interviewed about their experience of interactions with chengguan in the late stage of fieldwork. Numerous informal interviews were also conducted during our observations. After the first interview in 2014, we kept in regular contact with ten vendors, mainly through social media on Wechat (the most popular cellphone app in China), about their recent experience of street vending (or other jobs after several of them had quit street vending and started other businesses). The fourth part was devoted to the data mining of media reports of chengguan violence in Guangzhou. ‘Chengguan or street vendors’ plus ‘violence or conflict’ were used as keywords to search news reports in Wisenews, a database containing most of the newspapers in Mainland China, Hong Kong and Macau. This search was limited to reports about mainland China from 2011 to 2014. Since some news items may not be published by newspapers but can be reported by online media, the above-mentioned keywords were searched in Baidu (the Chinese version of Google) as well. News from different sources was triangulated to ensure reliability and pieced together to get a full picture of each case. Altogether, 37 cases of violent conflicts between chengguan and street vendors in Guangzhou were identified. Although the sample may only reflect a tip of the iceberg of all confrontations, and although media reports tend to be pro-government due to the heavy censorship by Chinese government (Xu 2015), some factual variables could be identified for analysis: time, place, type of violence, number, age and gender of street vendors, number of official chengguan, the number of auxiliary chengguan, reported reasons for violent confrontation, physical injury level of street vendors, physical injury level of chengguan and the consequence of confrontation. Among the 37 cases, 25 had detailed information about whether it was official chengguan or auxiliary chengguan who were involved in violent confrontations with street vendors. Lastly, this research was informed by the first author’s experience of living for over one decade in Guangzhou as he witnessed the daily public performance of the ‘cat and mouse’ game being played between chengguan and street vendors. The interview data and field notes were analyzed through open, axil and thematic coding following the approach of grounded theory (Charmaz 2014) with the assistance of Nvivo 11.0. Violence Within the Underclass: Confrontations Mainly Occur Between Auxiliary Chengguan and Street Vendors We have argued that while Chinese government sources and official media attribute the occurrence of chengguan violence to the low quality of auxiliary chengguan and tend to individualize the problem of violence, some critical commentators pointed out that auxiliary chengguan might simply be the scapegoats for official chengguan (Qi 2013; Zhang 2013). The underlying implication of this argument is that it is official chengguan, not auxiliary chengguan, who perpetrate the violence towards street vendors. To unravel this puzzle, we sought to determine when violent confrontations actually occurred between chengguan and street vendors, and the extent to which both official chengguan and auxiliary chengguan were involved. Of the 37 cases we collected from Wisenews and Baidu about violent confrontations between chengguan and street vendors in Guangzhou from 2011 to 2014, 25 cases contained detailed information about whether it was official chengguan or auxiliary chengguan who were involved. Among these 25 cases, auxiliary chengguan were involved in using violence in every instance, whereas official chengguan were only involved in using violence in two cases (see Figure 2). It was quite clear that auxiliary chengguan were by far the more common perpetrators of violence on street vendors, and they were not the scapegoats of official chengguan as some critical commentators suggested (although auxiliary chengguan could have been deployed to do the dirty work that official chengguan were unwilling to perform). The finding is also confirmed by our interview with chengguan officers (see the analysis in the final part of “Official Chengguan: A Risk Factor of Chengguan Violence” for details). Fig. 2 View largeDownload slide The involvement of official chengguan and auxiliary chengguan in violent confrontations with street vendors in Guangzhou from media report Fig. 2 View largeDownload slide The involvement of official chengguan and auxiliary chengguan in violent confrontations with street vendors in Guangzhou from media report Chengguan may also get injured in violent confrontations. Indeed, it was reported that over 1,679 chengguan were physically injured between 2005 and 2009 (Table 1) in Guangzhou. The official data did not report the distribution of victims between official chengguan and auxiliary chengguan. The limited cases in Guangzhou reported from media also did not contain enough data to figure out the pattern. However, the data from extensive interviews with chengguan and street vendors indicated that auxiliary chengguan were the major victims. Besides, our analysis of more systematic nationwide data also demonstrated that auxiliary chengguan were the major victims when there were confrontations with street vendors. A search of media reports through Wisenews and Baidu about chengguan–vendor confrontations which resulted in serious injury or death of chengguan across China from 2006 to 2017 generated 74 cases that contained detailed information regarding the identity of victims. It was found that among these cases, nearly 80 per cent of victims were auxiliary chengguan (Table 2). Table 2 The distribution of death and severe injury between official chengguan and auxiliary chengguan from 74 reported cases 2006–2017 in China Official chengguan Auxiliary chengguan Total Death 4 7 11 Severe injury 22 87 109 Total 26 (21.67%) 94 (78.33%) 120 (100%) Official chengguan Auxiliary chengguan Total Death 4 7 11 Severe injury 22 87 109 Total 26 (21.67%) 94 (78.33%) 120 (100%) View Large Table 2 The distribution of death and severe injury between official chengguan and auxiliary chengguan from 74 reported cases 2006–2017 in China Official chengguan Auxiliary chengguan Total Death 4 7 11 Severe injury 22 87 109 Total 26 (21.67%) 94 (78.33%) 120 (100%) Official chengguan Auxiliary chengguan Total Death 4 7 11 Severe injury 22 87 109 Total 26 (21.67%) 94 (78.33%) 120 (100%) View Large One possible explanation for the fact that the majority of violent confrontations occur between street vendors and auxiliary chengguan, rather than between street vendors and official chengguan, could be the composition of official chengguan and auxiliary chengguan in the overall chengguan force. If auxiliary chengguan disproportionately outnumbered official chengguan, it might be reasonable to expect that more auxiliary chengguan would be involved in violence than official chengguan. However, the composition of the chengguan force is only part of the story. Figure 3 showed that in Guangzhou, there were 3,196 official chengguan and 6,000 auxiliary chengguan in 2015. The number of auxiliary chengguan was almost twice as many as official chengguan. In other words, while auxiliary chengguan accounted for two-thirds of the overall chengguan force, they participated in all violent confrontations. While official chengguan accounted for one-third among the overall chengguan force, they only participated in eight per cent of cases in which violence was used on street vendors. In comparison, the likelihood of violent confrontation between auxiliary chengguan and street vendors was eleven times higher than that between official chengguan and street vendors. All in all, the violence between auxiliary chengguan and street vendors can hardly be explained away by the composition of the chengguan force. Fig. 3 View largeDownload slide The proportion of official chengguan and auxiliary chengguan in overall chengguan force Fig. 3 View largeDownload slide The proportion of official chengguan and auxiliary chengguan in overall chengguan force The Convergence of Violence: A Routine Activity Perspective While the official discourse attributes the chengguan violence to the ‘low quality’ of auxiliary chengguan, we argue that the incidence of violence between auxiliary chengguan and street vendors is largely to be explained by the convergence of opportunities to encounter each other on the street (Cohen and Felson 1979): Auxiliary chengguan are in fact much more likely to interact with street vendors on the street than are official chengguan. Figure 4 showed the data from our SSO of chengguan law enforcement on the street. In the course of 112 episodes of law enforcement, 336 chengguan officers were observed on the street. Of that number, 326 of them were auxiliary chengguan, and only 10 were official chengguan. In other words, in street-level law enforcement, 97 per cent of the force were auxiliary chengguan and only three per cent were official chengguan. Taking the composition of the overall chengguan force into consideration, auxiliary chengguan were 17 times more likely than official chengguan to work on the street. The high opportunity for auxiliary chengguan and vendors to encounter each other on the street inescapably led to the greater possibility of violent confrontation between the two. Fig. 4 View largeDownload slide The proportion of official chengguan and auxiliary chengguan in street law enforcement from systematic social observation data (n = 336) Fig. 4 View largeDownload slide The proportion of official chengguan and auxiliary chengguan in street law enforcement from systematic social observation data (n = 336) That observed division of labour between official chengguan and auxiliary chengguan was confirmed and elaborated by our interviews. When asked how often they work on the street, an official chengguan said: Not so often, we (chengguan) seldom patrol on the street. Auxiliary chengguan do the patrol now. Actually, we already have too much administrative work, lots of meetings, preparing for supervision visit from upper level government departments…. (An official chengguan from a street-level law enforcement branch, Haizhu District) His experience echoes that of his colleagues from other districts: Nowadays, except for special operations, without some newly emerged or urgent tasks, we will not go on to the street for law enforcement...Sometimes we may have big operations and then I will have to join to write out a ticket, take pictures, and video-tape the whole law enforcement process. But for daily work, not big operations, just patrol, there is no plan to write a ticket, we don’t go out.…I may probably go for street work only once a month. (An official chengguan in Panyu District Bureau) An auxiliary chengguan further explained that: Every day when we start work, we first report our duty to the office, and then go to our own assigned sites to monitor or our own assigned area to patrol. As for official chengguan, they usually stay in the office. Occasionally, they may pay a short visit to us, have a look and then leave. (An auxiliary chengguan from a street-level law enforcement branch, Tianhe District) In a nutshell, instead of emphasizing the low quality of individual auxiliary chengguan, our study demonstrates that the convergence of violence between auxiliary chengguan and street vendors was largely shaped by the routine (working) activity of auxiliary chengguan, itself a result of the unequal division of labour between official chengguan and auxiliary chengguan. There is still a question to be asked about such a clearly unequal division of labour: Why for most of the time do the official chengguan tend to stay away from the street? Or, how is the high convergence of auxiliary chengguan and street vendors on the street socially produced? Police Civilianization in China and the ‘Dirty Work’ of Auxiliary Chengguan While police civilianization reaches an unprecedented level of over 30–40 per cent in Western free-market economies, some anecdotal data show that in China over half of the police force are civilian staff and in some big cities the percentage is even higher (Wang 2015). The official statistics collected during the course of our research in Guangzhou show that apart from 31,000 public police officers, there are another 30,000 contract-based, temporarily hired, and low-paid auxiliary police. The civilianization level of Guangzhou’s para-police chengguan is even higher with two-thirds of the force are auxiliary chengguan (Figure 3). One of the main concerns of police civilianization is the unequal power relationship between civilian staff and sworn officers (Alderden and Skogan 2014). In the case of China’s para-police chengguan, the civilian staff auxiliary chengguan are undoubtedly second-class citizens in the force, which can be clearly demonstrated by their differential treatment not only in the allocation of benefits, welfare and power but also in the division of labour between the two groups (Table 3). Table 3 The power structure of chengguan system Official chengguan Auxiliary chengguan Salary Around RMB 6,000 Around RMB 2,000 Job nature Tenure Short-term Position Civil servant Not civil servant Power Holding power, giving order No power, receiving order Daily work Office-based Street-based, dirty work Official chengguan Auxiliary chengguan Salary Around RMB 6,000 Around RMB 2,000 Job nature Tenure Short-term Position Civil servant Not civil servant Power Holding power, giving order No power, receiving order Daily work Office-based Street-based, dirty work View Large Table 3 The power structure of chengguan system Official chengguan Auxiliary chengguan Salary Around RMB 6,000 Around RMB 2,000 Job nature Tenure Short-term Position Civil servant Not civil servant Power Holding power, giving order No power, receiving order Daily work Office-based Street-based, dirty work Official chengguan Auxiliary chengguan Salary Around RMB 6,000 Around RMB 2,000 Job nature Tenure Short-term Position Civil servant Not civil servant Power Holding power, giving order No power, receiving order Daily work Office-based Street-based, dirty work View Large First, although the salary awarded to chengguan varies across different districts in Guangzhou, it is commonly observed that the compensation for official chengguan is around three times as large as that for auxiliary chengguan. In Guangzhou, the average monthly salary of an auxiliary chengguan was around RMB1 2,000 [about £230 or US $300] whereas it was around RMB 6,000 [about £700 or US $900] for an ordinary official chengguan in 2015. Our analysis of about 56 advertisements recruiting auxiliary chengguan in the first half of 2015 revealed that the average monthly starting salary was as low as RMB 1,800 [or £210 or US $270], just about the level of minimum wage in Guangzhou. An auxiliary chengguan complained: In Tianhe district, there are about 200 people in the Chebei street law enforcement branch, 150 of whom are auxiliary chengguan. Official chengguan are civil servants. Government pays their salary, about three times more than the salary of auxiliary chengguan…auxiliary chengguan are temporary employees, who do the same job but are paid less .…Besides, official chengguan have very good welfare agreements, while there is nothing for auxiliary chengguan… (An auxiliary chengguan from Chebei Street-level Law Enforcement Branch, Tianhe District) Second, there is a defining difference in levels of job security. While official chengguan are civil servants with job tenure, auxiliary chengguan are hired on a fixed-term basis, usually with a one year to three years’ contract, but they could be fired by local governments at will. An auxiliary chengguan expressed how they were treated in a campaign period when a zero-tolerance policy towards vendors would be expected: After our team leader (an official chengguan) came back from the meeting (to receive tasks), he scolded us and threatened that we would be fired if we did not perform well. We work overtime every day and suffer from a lack of sleep and psychological pressure.… We could not leave earlier for a single minute before 11:00 p.m. Those who leave earlier would be fired. (An auxiliary chengguan from Huangcun Street Law Enforcement Branch, Tianhe District) Third, there is a huge power difference between the two groups. Within the chengguan system, official chengguan are often regarded as the ‘bosses’ who are in charge whereas auxiliary chengguan are subordinate and subject to the control of official chengguan. The following narratives from an auxiliary chengguan are illuminating: To tell you the truth, official chengguan is a much better job than auxiliary chengguan. Official chengguan earn much more money. As formal staff, they are the core of our team. Besides, they have more power, good welfare and all kinds of subsidies. Nevertheless, they do not have too much work. Nobody could supervise them whatever they do…. The official chengguan can decide who is on the night shift and whether they come or not. It is okay even if they do not show up for work. Nobody could supervise them. (An auxiliary chengguan from Yangcheng Garden Street-level Law Enforcement Branch, Tianhe District) A 50-year-old auxiliary chengguan with nearly two decades of working experience complained to us when we met him on the street in 2014: I have to work for 15 hours a day, from 7:00a.m. to 10:00p.m. With over-time working allowance, I can now earn around RMB 100 per day, otherwise, only a bit over RMB 1,000 per month. I also have to work during weekends for extra payment. We got complaints for the chaotic situation on the street, and our team leader asked us to work extra time. We are even not able to leave the supervised spots when having meals. In terms of high working pressure, our former team leader would greet us by saying ‘you have been working hard, well done’, then I will feel much better. However, our current team leader always criticizes us for not doing our jobs well. Even if we have driven vendors to small lanes, he is still not satisfied. There are about 90 plus auxiliary chengguan in our branch and their monthly salary is less than RMB 2,000 while the team leader, the official chengguan, has a monthly income of over RMB 10,000. This huge gap makes me feel very bad, and most importantly, he always scolds us. We don’t even have some psychological comforting for such a hard job. We are suffering four levels of pressure. Firstly, vendors may say dirty words to us, and even to our family numbers, particularly to our mothers and our children. I feel very bad and upset about that. Secondly, bystanders may also bad-mouth by claiming we are bad and heartless people. Thirdly, some local residents may also complain to us that we are not able to do our job well to leave the street so chaotic. Fourthly and the most depressing experience is that we are always criticized by our boss, the official chengguan, even if we have been working very hard for this thankless job. (An auxiliary chengguan from Shisanhang Street-level Law Enforcement Branch, Yuexie District) Fourth, due to the unequal power enjoyed by official and auxiliary chengguan, daily street patrol and street law enforcement—tedious and less desirable tasks—are primarily conducted by auxiliary chengguan. Figure 5 shows that while official chengguan accounted for 35 per cent in the overall force, most of them work in management positions at district- and city-level bureaus. Our fieldwork data reveal that at the street-level law enforcement departments, official chengguan only took up 15 per cent, and in the actual street law enforcement, official chengguan were even less involved, only three per cent. Fig. 5 View largeDownload slide The composition of chengguan in the overall force (official statistics), street-level chengguan branch (calculated from 22 street-level law enforcement branches), and actual street law enforcement (SSO data) Fig. 5 View largeDownload slide The composition of chengguan in the overall force (official statistics), street-level chengguan branch (calculated from 22 street-level law enforcement branches), and actual street law enforcement (SSO data) In sum, law enforcement against street vendors could reasonably be regarded as ‘dirty work’ as it is physically disgusting and dangerous, socially degrading and morally challenging (Hughes 1958: 319; Löfstrand et al. 2015). The civilianization of the para-policing chengguan force has translated auxiliary chengguan into second-class officers responsible for most of the ‘dirty work’ of street-level law enforcement. Official Chengguan: A Risk Factor of Chengguan Violence So far, this research has demonstrated that violent confrontation tends to occur between auxiliary chengguan and street vendors, and that the unequal power structure within the chengguan force makes the official chengguan far less likely to work on the street, and that that, in turn, reduces their likelihood of engaging in confrontations with the vendors. While the official discourse attributes chengguan’s violence towards ‘low quality’ of auxiliary chengguan, the underlining implication is that official chengguan are of ‘high quality’, who may well have reduced the occurrence of violent confrontation with vendors had they been on the street. Nonetheless, our SSO data showed that the appearance of official chengguan on the street is a risk factor, instead of a containing factor, for chengguan violence. Out of 112 instance of chengguan’s street law enforcement, we observed the use of verbal violence (shouting, abusive wording) eight times and physical violence (kicking and hitting) on two occasions. All this violence was conducted by auxiliary chengguan. However, the logistic regression analysis revealed that the occurrence of verbal and physical violence was positively related to the appearance of official chengguan on the street (p < 0.05) (available from authors). That is, when official chengguan were on the street, it was more likely that auxiliary chengguan would use violence. There are several possible explanations. First, official chengguan seldom went on the street for law enforcement. When they did appear on the street, it usually meant they were participating in operations or campaign days. On these occasions, more strict law enforcement would be expected and thus the risks of confrontations would be raised. A street vendor remarked that: Most of time it is auxiliary chengguan who work on the street and patrol. Official chengguan in white uniforms only come out on special days, such as campaign days. On those days chengguan are very strict and we can hardly sell things on the street. (A street vendor W) Second, when confrontation occurs, official chengguan tend to stay away while ordering auxiliary chengguan to deal with the conflict. The following description is an excerpt from our field notes in 2016: It was about eight o’clock at night. A group of chengguan (two official chengguan and seven auxiliary chengguan) drove an electric car towards a crowd of vendors. Among them, an auxiliary chengguan took a video camera to record the whole process of law enforcement. Seeing the arrival of chengguan, the vendors started to flee in all directions while packing their stuff. In the middle of the vendors, there was a young man, not tall, around 20-year old, selling barbecue chicken. He was caught by two auxiliary chengguan as he had to turn the fire off before fleeing but his way was blocked by other vendors. The auxiliary chengguan intended to confiscate his tricycle (which he used to sell chicken) but he grabbed it firmly to resist. He was alone, weak and unable to defeat the two auxiliary chengguan. The two auxiliary chengguan pushed the young vendor to the ground and got his tricycle. Seeing his tricycle was about to be taken away, the young vendor became mad. He took a knife which he used to cut chicken and shouted ‘return the tricycle to me, return the tricycle to me’. The crowd started to gather around. Meanwhile, two official chengguan stood nearby and watched the process while having their hands crossed behind. After seeing the vendor waving his knife, one of them made a phone call. The auxiliary chengguan who was responsible for video-taping came closer to record the scenario of the vendor’s waving of the knife. More people gathered, and the situation became tense. The two auxiliary chengguan stopped taking the tricycle away and the vendor was surrounded by other auxiliary chengguan. They shouted to him: ‘What are you doing? Don’t make trouble. Take it easy.’ At this moment, one auxiliary chengguan suddenly grabbed the vendor from behind and took away the knife. The vendor could not move and other auxiliary chengguan rushed to knock the vendor down to the ground and started beating and kicking him while shouting ‘You Mother-F**ker, how dare you threaten us!’. The auxiliary chengguan who was recording the video was summoned by an official chengguan to stop video-taping of the beating. The beating lasted for 3–4 minutes. (Field notes, 28 March 2016, in Panyu District) When interviewed by a journalist, one auxiliary chengguan elaborated why it is always the auxiliary chengguan who are involved in direct confrontations: We call ourselves removers. It totally makes sense that why auxiliary chengguan often have conflicts with street vendors. It is because official chengguan just write tickets but never directly take away street vendors’ items2. The most dangerous work is done by us. We are the ‘hatchet men’ .…Confiscation is manual work, and official chengguan have a sense of superiority. They think it is our job.… And the official chengguan is our leader when coming out with us. How could we let our leader do such kind of work? (Zhang 2013). Third, auxiliary chengguan might also need to perform more aggressively when official chengguan were present. An auxiliary chengguan said: Official chengguan do not directly ask us to use violence, but you cannot be inactive…when other auxiliary chengguan start confiscation, you have to follow up… (An auxiliary chengguan Li) Sometimes, auxiliary chengguan may also feel they are institutionally backed when official chengguan are around. Therefore, they are more aggressive in law enforcement. An auxiliary chengguan said: We do whatever the official chengguan ask us to do… and the official chengguan will protect us. It means we have the official chengguan to protect us so that we need not be worried. It means if you get into trouble, they could promise that you will not be punished. We are temporary staff and we only try our best unless they promise us our actions will have no negative consequences, just like the police and auxiliary police … Once in a campaign, a barbecue vendor resisted our order. The official chengguan said ‘deal with him (搞定他 gao ding ta)’, so we take everything away from him. (An auxiliary chengguan from Tianhenan street-level law enforcement branch) When such institutional backing is not available, they feel frustrated. An auxiliary chengguan said: Once, a father and son who sell pork were asked to go away by an auxiliary chengguan. Then, they put knives on one the auxiliary chengguan’s neck and it seemed a terrible fighting was going to occur. Fortunately, another auxiliary chengguan came to persuade them and stopped the conflict. When I reported it to my team leader, an official chengguan said we must drive the street vendors away even if we would have to put ourselves in danger. He said we could call for help from the police. We should not be afraid of the street vendors. However, he was not concerned about our safety. I think we would have been dead long before the policemen arrived. I was very sad and angry about his words. (An auxiliary chengguan works in Shisanhang, Yuexiu District) In sum, while the government attributes the chengguan violence to the individual officers of auxiliary chengguan, this study demonstrates that the concentration of violence between auxiliary chengguan and street vendors is largely shaped by the division of labour within in chengguan system in which most of the undesirable ‘dirty work’ of street-level law enforcement is conducted by the second-class officers auxiliary chengguan. Therefore, the unequal power structure resulted from commodification of the para-police chengguan force shapes the concentration of violence within the underclass. Conclusion Police civilianization, albeit celebrated by some, has attracted various criticisms, and the unequal power relationship within the police system is one of them (Dick and Metcalfe 2001). Although the recent development of police civilianization may put some nonsworn officers in oversight positions (the elected civilian Police Crime Commissioners in the United Kingdom can set objects, allocate funds and hold police accountable with considerable power since 2010 (Lumsden and Black 2018)), the vast majority of civilian staff work as supportive roles for the public police, and in some cases they become second-class citizens in the police force. By studying violent confrontations between para-police chengguan and street vendors, this article reveals how police civilianization has produced unequal power relations within the para-police chengguan system in China. The unprivileged auxiliary chengguan have to transact most of the ‘dirty work’ of street-level law enforcement when confronting with street vendors. It is an unequal division of labour that makes the auxiliary chengguan far more likely to encounter street vendors in public space, leading to the possibility of confrontation and violence between the two. While official chengguan stay away from street law enforcement most of the time, their occasional appearance on the street actually brings ‘more’ violence to vendors. Since both auxiliary chengguan and street vendors are recruited from the urban poor, the article further illustrates how police civilianization unintentionally shapes the concentration of interpersonal violence within the underclass. My argument not only extends our understanding of the possible negative consequence of police civilianization but reveals a new macrostructural mechanism unexpectedly affecting the concentration of violence within the underclass. It further enriches our knowledge of the pluralization of policing and research in para-police in particular (Rigakos 1999). Future research can examine how police civilianization occurs in other former socialist countries where the adaptation of market logic comes much later than that in Western democracies. It will also be interesting to explore other forms of social inequality caused by the marketization of the public police. Funding Multi-Year Research Grant (MYRG2015-00163-FSS, MYRG2015-00039-FSS, MYRG2017-00115-FSS and MYRG2018-00109-FSS), University of Macau. Acknowledgements We are very much indebted to the following professors for their constructive comments and suggestions on various earlier versions of the article: Paul Rock, Borge Bakken, Steven Messner, Sheldon Zhang, Peter Grabosky, Maggy Lee, Nabo Chen, Zhidong Hao, Ivan Sun, Liqun Cao, Peng Wang, Vincent Cheng and two anonymous reviewers. The writing of the paper was partially completed during Jianhua Xu’s visit as a visiting scholar at RegNet, Australian National University during June to August, 2017. Footnotes 1 1 RMB = $ 0.153 or £0.128 at 2018 prices 2 Chengguan have power to confiscate vendors’ vending vehicles, tools and goods. Theoretically, upon paying fine, usually RMB 200, the confiscated stuff will be returned. In practice, vendors seldom get them back due to complicated bureaucracy or frustrated former experience. References Alderden , M. and Skogan , W. G . ( 2014 ), ‘ The Place of Civilians in Policing ’, Policing: An International Journal of Police Strategies & Management , 37 : 259 – 84 . Google Scholar Crossref Search ADS Anderson , E . ( 1999 ), Code of the Street: Decency, Violence, and the Moral Life of the Inner City . W.W. Norton . Ayling , J. , Grabosky , P. N. , Shearing , C. D. and Cambridge University Press ( 2009 ), Lengthening the Arm of the Law: Enhancing Police Resources in the Twenty-First Century . Cambridge University Press . Blau , J. R. and Blau , P. M . ( 1982 ), ‘ The Cost of Inequality: Metropolitan Structure and Violent Crime ’, American Sociological Review , 47 : 114 – 29 . Google Scholar Crossref Search ADS Bruce , M. A . ( 2000 ), ‘ Violence among African Americans ’, Journal of Contemporary Criminal Justice , 16 : 171 – 93 . Google Scholar Crossref Search ADS Charmaz , K . ( 2014 ), Constructing Grounded Theory . Sage . Chen , K. and Kong , Y . ( 2013 ), ‘ Three Chengguan in Fuzhou Broke Other’s Bones ’, China News , available online at http://www.chinanews.com/sh/2013/12–27/5672123.shtml. Cohen , A. K . ( 1955 ), Delinquent Boys: The Culture of the Gang . Free Press . Cohen , L. E. and Felson , M . ( 1979 ), ‘ Social Change and Crime Rate Trends: A Routine Activity Approach ’, American Sociological Review , 44 : 588 – 608 . Google Scholar Crossref Search ADS Cosgrove , F. M . ( 2015 ), ‘ “I Wannabe a Copper”: The Engagement of Police Community Support Officers with the Dominant Police Occupational Culture ’, Criminology & Criminal Justice , 16 : 119 – 38 . Google Scholar Crossref Search ADS Crank , J. P . ( 1989 ), ‘ Civilianization in Small and Medium Police Departments in Illinois, 1973–1986 ’, Journal of Criminal Justice , 17 : 167 – 77 . Google Scholar Crossref Search ADS Crawford , A. and Lister , S . ( 2006 ), ‘ Additional Security Patrols in Residential Areas: Notes from the Marketplace ’, Policing and Society , 16 : 164 – 88 . Google Scholar Crossref Search ADS Davis , R. C. , Lombardo , M. E. , Woods , D. J. , Koper , C. and Hawkins , C . ( 2014 ) Civilian Staff in Policing-an Assessment of the 2009 Byrne Civilian Hiring Program . The US Department of Justice . Desmond , M . ( 2017 ), Evicted: Poverty and Profit in the American City . Penguin Books . Dick , G. and Metcalfe , B . ( 2001 ), ‘ Managerial Factors and Organisational Commitment: A Comparative Study of Police Officers and Civilian Staff ’, International Journal of Public Sector Management , 14 : 111 – 28 . Google Scholar Crossref Search ADS Diphoorn , T. G . ( 2016 ), ‘ Twilight Policing: Private Security Practices in South Africa ’, British Journal of Criminology , 56 : 313 – 31 . Google Scholar Crossref Search ADS Downes , D. M. , Rock , P. E. and McLaughlin , E . ( 2017 ), Understanding Deviance: A Guide to the Sociology of Crime and Rule-Breaking . Oxford University Press . Forst , B . ( 2000 ), ‘ The Privatization and Civilianization of Policing ’, Criminal Justice , 24 : 19 – 78 . Goffman , A . ( 2014 ), On the Run: Fugitive Life in an American City . The University of Chicago Press . Grabosky , P. N . ( 2007 ), ‘ Private Sponsorship of Public Policing ’, Police Practice and Research , 8 : 5 – 16 . Google Scholar Crossref Search ADS Guangzhou Public Opinion Research Center ( 2013 ), ‘ Citizens Think Low of Chengguan’s Law Enforcement ’, Research Report , available online at http://www.c-por.org/index.php?c=news&a=ztdetail&id=2838&pid=57 . Hanser , A . ( 2016 ), ‘ Street Politics: Street Vendors and Urban Governance in China ’, The China Quarterly , 226 : 363 – 82 . Google Scholar Crossref Search ADS Home Office ( 2016 ), ‘ Police Workforce, England and Wales, 31 March 2016 ’, Home Office , available online at https://www.gov.uk/government/statistics/police-workforce-england-and-wales-31-march-2016 . Hughes , E. C . ( 1958 ), Men and Their Work . Free Press . Human Rights Watch ( 2012 ), ‘Beat Him, Take Everything Away: Abuses by China’s Chengguan Para-Police’ , Report, Printed in the United States of America , available online at https://www.hrw.org/news/2012/05/16/beat-him-take-everything-away-abuses-chinas-chengguan-para-police-report-press . Ji , F . ( 2014 )‘ How Can Temporary Employees Be Scapegoats of Official Chengguan ?’, Hubei Daily , available online at http://focus.cnhubei.com/original/201404/t2892649.shtml. Jiang , F . ( 2012 )‘ Guangzhou Chengguan Knifed a Street Vendor and Officials Said the Offender Was a Temporary Employee ’, Wangyi News Website , available online at http://news.163.com/12/0516/20/81LCPPVJ00011229.html. Jones , T. and Newburn , T . ( 2002 ), ‘ The Transformation of Policing? Understanding Current Trends in Policing Systems ’, The British Journal of Criminology , 42 : 129 – 46 . Google Scholar Crossref Search ADS Kiedrowski , J. , Ruddell , R. and Petrunik , M . ( 2017 ), ‘ Police Civilianisation in Canada: A Mixed Methods Investigation ’, Policing and Society . doi: https://doi.org/10.1080/10439463.2017.1281925 , available online at https://www.tandfonline.com/doi/citedby/10.1080/10439463.2017.1281925?scroll=top&needAccess=true . Kirk , D. S. and Papachristos , A. V . ( 2011 ), ‘ Cultural Mechanisms and the Persistence of Neighborhood Violence ’, American Journal of Sociology , 116 : 1190 – 233 . Google Scholar Crossref Search ADS Lan , Y . ( 2013 ), ‘90 Percent of Respndents Said They Believed Temporary Employee Are Scapegoats’ available online at http://www.gd.chinanews.com/2013/2013-06-08/2/252187.shtml. Lippert , R. K. and Walby , K . ( 2013 ), ‘ Police Moonlighting Revisited: The Case of ‘Pay Duty’ in Three Canadian Police Services ’, Policing: A Journal of Policy and Practice , 7 : 370 – 8 . Google Scholar Crossref Search ADS Liu , H . ( 2016 ), ‘ How Many Chengguan in China ?’, Urban Management , available online at http://www.sohu.com/a/77246457_355337. Loader , I . ( 1999 ), ‘ Consumer Culture and the Commodification of Policing and Security ’, Sociology-the Journal of the British Sociological Association , 33 : 373 – 92 . Loader , I. , Goold , B. and Thumala , A . ( 2014 ), ‘ The Moral Economy of Security ’, Theoretical Criminology , 18 : 469 – 88 . Google Scholar Crossref Search ADS Löfstrand , C. H. , Loftus , B. and Loader , I . ( 2015 ), ‘ Doing ‘Dirty Work’: Stigma and Esteem in the Private Security Industry ’, European Journal of Criminology , 13 : 297 – 314 . Google Scholar Crossref Search ADS Lumsden , K. and Black , A . ( 2018 ), ‘ Austerity Policing, Emotional Labour and the Boundaries of Police Work: An Ethnography of a Police Force Control Room in England ’, The British Journal of Criminology , 58 : 606 – 23 . Google Scholar Crossref Search ADS Maguire , E. R. and King , W. R . ( 2004 ), ‘ Trends in the Policing Industry ’, The Annals of the American Academy of Political and Social Science , 593 : 15 – 41 . Google Scholar Crossref Search ADS Massey , D. S. and Denton , N. A . ( 1993 ), American Apartheid: Segregation and the Making of the Underclass . Harvard University Press . Matrofski , S. D . ( 1990 ), ‘ The Prospects of Change in Police Patrol: A Decade in Review ’, American Journal of Police , 9 : 1 – 79 . McCarty , W. P. and Skogan , W. G . ( 2013 ), ‘ Job-Related Burnout among Civilian and Sworn Police Personnel ’, Police Quarterly , 16 : 66 – 84 . Google Scholar Crossref Search ADS O’Neill , M . ( 2014 ), ‘ Ripe for the Chop or the Public Face of Policing? Pcsos and Neighbourhood Policing in Austerity ’, Policing , 8 : 265 – 73 . Google Scholar Crossref Search ADS Ouyang , G . ( 2013 )‘ A Street Vendor Was Beaten to Death by Chengguan in Linwu,Hubei Province ’, Wangyi News , available online at http://news.163.com/photoview/00AP0001/36360.html#p=940SPP9D00AP0001&from=tj_review. Paskell , C . ( 2007 ), ‘ ‘Plastic Police’or’community Support’?The Role of Police Cimmunity Support Officers within Low-Income Neighbourhoods ’, European Urban and Regional Studies , 14 : 349 – 61 . Google Scholar Crossref Search ADS People.cn ( 2014 ), ‘Bluebook Pointed out Five Image Crisis of Officials’ , available online at http://politics.people.com.cn/n/2014/0528/c1001-25076827.html. Polk , K . ( 1994 ), When Men Kill: Scenarios of Masculine Violence . Cambridge University Press . Qi , Y . ( 2013 )‘ Temporary Employee Should Not Be the Scapegoat ’, Northeast News , available online at http://news.nen.com.cn/system/2013/04/24/010336109.shtml. Reiman , J. and Leighton , P . ( 2016 ) The Rich Get Richer and the Poor Get Prison: Ideology, Class, and Criminal Justice . Routledge . Reiss , A. J . ( 1971 ), ‘ Systematic Observation of Natural Social Phenomena ’, Sociological Methodology , 3 : 3 – 33 . Google Scholar Crossref Search ADS Rigakos , G. S . ( 1999 ), ‘ Hyperpanoptics as Commodity: The Case of the Parapolice ’, The Canadian Journal of Sociology , 24 : 381 – 409 . Google Scholar Crossref Search ADS Rock , P. E . ( 1998 ), After Homicide: Practical and Political Responses to Bereavement . Clarendon Press . Sampson , R. J. and Raudenbush , S. W . ( 1999 ), ‘ Systematic Social Observation of Public Spaces: A New Look at Disorder in Urban Neighborhoods ’, The American Journal of Sociology , 105 : 603 – 51 . Google Scholar Crossref Search ADS Sandel , M. J . ( 2013 ), What Money Can’t Buy: The Moral Limits of Markets . Penguin Books . Sherman , L. W . ( 1978 ), Scandal and Reform: Controlling Police Corruption . University of California Press . Venkatesh , S. A . ( 2008 ), Gang Leader for a Day: A Rogue Sociologist Crosses the Line . Allen Lane . Wang , W . ( 2015 ), ‘ The Hard Life of Auxiliary Police ’, People’s Daily , Febuary 11, 17. White , A . ( 2014 ), ‘ Post-Crisis Policing and Public–Private Partnerships ’, British Journal of Criminology , 54 : 1002 – 22 . Google Scholar Crossref Search ADS Wolfgang , M. E. and Ferracuti , F . ( 1967 ), The Subculture of Violence: Towards an Integrated Theory in Criminology . Tavistock Publications . Xu , J . ( 2007 )‘ Migrant Offenders and Victims of Violent Crime in China ’, Paper presented in “Crime, Law & Justice in Chinese Societies: Global Challenges & Local Response, an International Conference” , Chinese University of Hong Kong, Hong Kong , March 16–18 . —— ( 2009 ), ‘ The Robbery of Motorcycle Taxi Drivers in China: A Lifestyle/Routine Activity Perspective and Beyond ’, The British Journal of Criminology , 49 : 491 – 512 . Crossref Search ADS —— ( 2013 ), ‘ Police Accountability and the Commodification of Policing in China: A Study of Police/Business Posters in Guangzhou ’, British Journal of Criminology , 53 : 1093 – 117 . Crossref Search ADS —— ( 2015 ), ‘ Claims-Makers Versus Non-Issue-Makers: Media and the Social Construction of Motorcycle Ban Problems in China ’, Qualitative Sociology Review , 11 : 122 – 41 . Xu , J. , Laidler , K. J. and Lee , M . ( 2013 ), ‘ Doing Criminological Ethnography in China: Opportunities and Challenges ’, Theoretical Criminology , 17 : 271 – 9 . Google Scholar Crossref Search ADS Young , J . ( 1999 ), The Exclusive Society: Social Exclusion, Crime and Difference in Late Modernity . Sage . Zang , X . ( 2017 ), ‘ How Cohesive Is the Chinese Bureaucracy? A Case Study of Street-Level Bureaucrats in China ’, Public Administration and Development , 37 : 217 – 26 . Google Scholar Crossref Search ADS Zhang , D . ( 2009 )‘ Netizen Exposes Textbook on ‘Beating without Leaving Blood’ ’, Southern Metropolis Daily , April 22, A32. Zhang , H . ( 2011 )‘ The Problem of Chengguan System and Its Violation of Human Rights ’, Rights Defense Website , available online at http://wqw2010.blogspot.com/2011/11/blog-post_5808.html. —— ( 2013 )‘ A Young Man Revealed the Difficult Life as a Chengguan Temporary Worker ’, Hua Shang Bao , available online at http://news.qq.com/a/20130727/000565.htm, sina news. Zhu , X . ( 2014 ) ‘CCTV Exposed the Video Which Recorded Cang Nan Chengguan Attacking Street Vendor; Thousands of People Gathered Because of Rumor’ , available online at http://news.cjn.cn/shxw/201404/t2464663_1.htm. © The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). All rights reserved. For permissions, please e-mail: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Brunton-Smith, Ian; Bullock, Karen

doi: 10.1093/bjc/azy012pmid: N/A

Abstract Established in England and Wales in the context of the neo-liberal governments of the 1980s and promoted through the New Local agenda of New Labour and beyond, Neighbourhood Watch (NW) is a primary means through which the state and citizens may co-produce crime control. However, whether citizens have the time or inclination to co-produce is debated, and it is generally believed that NW proliferates in advantaged, low crime rate areas that need it least. Drawing on analysis of the Crime Survey for England and Wales (CSEW) (1988–2010/11), this article examines long-term trends in participation in NW. It examines the spread of NW, how household support for NW fluctuates once established and the changing importance of some of the key household drivers of participation in NW. It then assesses the extent to which NW schemes are concentrated in more affluent areas, showing that this is moderated by crime risk. Introduction The ‘co-production’ of public services ‘Co-production’ is a cornerstone of public policy reform in England and Wales as it is across the globe (Osborne et al. 2016: 639). This provides a model for the agents of the state and citizens to come together to contribute to the design and/or delivery of a public service or public good (Mattson 1986; Ostrom 1996; Pestoff 2006; Bovaird 2007; Parrado et al. 2013; Brandsen and Honingh 2015; Voorberg et al. 2015; Alford and Yates 2016; Van Eijk and Steen 2016; Van Eijk et al. 2017). Co-production may take multiple forms involving a wide range of actors and activities (Brandsen and Honingh 2015; Alford and Yates 2016; Loeffler and Bovaird 2016; Van Eijk and Steen 2016). But however it is configured, it represents a ‘process through which inputs from individuals who are not “in” the same organization are transformed into goods and services’ (Ostrom 1996: 1037). As such, it implies an interaction between public agents and citizens that goes beyond consultation or listening to citizens or responding to citizens’ concerns towards some kind of active, collective production of tangible outputs or outcomes (Mattson 1986; Ostrom 1996; Alford and Yates 2016). Active voluntary citizen involvement appears to be the critical dimension of all co-production research (Mattson 1986: 52). The concept of co-production, central to many areas of contemporary public administration, purports to offer a cost-effective way to both increase efficiency and enhance the effectiveness of public services (Mattson 1986; Ostrom 1996; Pestoff 2009; Voorberg et al. 2015; Osborne et al. 2016; Van Eijk et al. 2017). It gained traction during the New Public Management reforms of the 1980s (Pestoff 2009; Osborne et al. 2016) and continued to be highlighted as a means of solving the major demographical, political and economic challenges facing Western democratic governments in the 21st century (Pestoff 2009). It has also been identified as essential for sustaining provision in European welfare systems in the face of budget constraints (Pestoff 2006). However, since co-production differs from the ‘traditional model’ of public service—where public officials are exclusively charged with responsibility for designing and providing services to citizens (Pestoff 2006)—it has also been seen to offer a means of democratizing public services through promoting citizen empowerment and participation (Ostrom 1996; Pestoff 2009) (see also Lowndes and Pratchett 2011; Stoker 2006). Citizens, crime control and Neighbourhood Watch Citizens have long played a role in the enactment of crime control and the ways that they have done so are diverse. In days gone by, communities were expected to join the ‘hue and cry’, and householders were expected take turns volunteering as watchmen and presenting offenders at court (Rawlings 2002). Today, citizens in England and Wales may volunteer as special constables, may form street patrols and may, through collaborating with the police to identify and resolve crime problems, contribute to community policing (Skogan 2008; Tilley 2008; Bullock 2014). Yet while there clearly is a long history of citizen involvement in crime control and the roles played are wide ranging, contemporary calls for citizens in England and Wales to co-produce can be situated within the neo-liberal Conservative administrations of the 1980s and 1990s, which stressed the desirability of rolling back the state, improving the efficiency of public services and promoting volunteering and community/citizen action (Crawford 1999; Bullock 2014). It was into this political and economic context that Neighbourhood Watch (NW) became established in England and Wales. NW brings neighbours together to act as the ‘eyes and ears’ of the police through watching out for suspicious behaviours in their neighbourhoods and reporting them, and to protect their properties and those of their neighbours (Rosenbaum 1987; 1988; Bennett 1990; Laycock and Tilley 1995; Gresham et al. 2004; Bullock 2014). NW appealed at once to the politically Conservative governments of the 1980s, concerned about maintaining standards of behaviour, and to libertarian instincts regarding the importance of voluntarism, the role of the ‘active citizen’ and the minimal role of the state (Hope 1995: 43). The NW movement grew very quickly in this climate (Rosenbaum 1987; 1988; Skogan 1988; 1989; Bennett 1990; Hope 1995; Bullock 2014) and was to become one of the largest voluntary organizations in England and Wales and one of the largest single organized crime prevention activities in the world (Bennett et al. 2008). Co-production has been promoted as a way of increasing the effectiveness of public services (Mattson 1986; Ostrom 1996; Pestoff 2009; Voorberg et al. 2015; Osborne et al. 2016; Van Eijk et al. 2017), and reflecting these wider trends, citizen involvement in policing was promoted as cost-effective response to the problem of crime (Skogan 2008; Tilley 2008; Bullock 2014). Equally, citizen involvement in the governance and delivery of crime control from the 1970s onwards was perceived to be an effective response to various crises. Community policing styles were endorsed in the context of contested relationships between the police and communities—characterized by a disconnection between agents of the public police and citizens—and by troubled relationships between the police and minority ethnic communities (Scarman 1981; Skogan 2008; Tilley 2008; Bullock 2014). And NW itself was ‘promoted as evidence of a break with a past which had been marked by fractured relations with the community, a remote police service and styles of policing which had contributed to major confrontations with the public, including street disorders’ (McConville and Shepherd 1992: 1). The promotion of the citizen as a co-producer of public services continued unabated into the 1990s and beyond. New Labour, through their ‘New Localism’ agenda, sought to embed the citizen ever more firmly within the oversight and delivery of public services and to redistribute power from central bureaucrats to local structures and, through the processes of double devolution, to citizens themselves (Lowndes and Pratchett 2011; Stoker 2006). Central to this narrative were ‘active citizens’ (Lowndes and Pratchett 2011; Stoker 2006). Devolving power to ‘active citizens’ would improve effectiveness and generate new democratic accountabilities and scrutiny (Home Office 2004; see also Bullock 2014). In respect to policing governance this was usually linked to Neighbourhood Policing, a form of community policing, which aimed to engender greater democratic accountability through facilitating citizens involvement in determining how their communities are policed and ensuring that police listen and act on community concerns (Home Office 2004; Quinton and Morris 2008; Bullock 2014). Themes of devolving power to the responsible and ‘active citizen’ and imploring them to come together with state agents to address crime problems and to engender broader cultures of responsibility and mutuality have continued in the post-New Labour era into the Conservative-Liberal Coalition (2010–15) and the Conservative administration which followed it (2015+) (see e.g. Cabinet Office 2010). In respect to crime control, community policing continued to play a central role in the pursuit of direct citizen oversight of policing. And democratic accountability took a new twist with the introduction of elected officials (police and crime commissioners) who serve to orient police services around the needs of citizens and communities and offer the prospect of indirectly holding the service to account (see e.g. Bullock 2014; Lister and Rowe 2015). Aims and contribution While there has been much theoretical interest in the value of co-production, empirical studies of the key drivers of co-production are comparatively rare, with a general reliance on small-scale case studies (Parrado et al. 2013; Van Eijk and Steen 2016; Brandsen and Honingh 2015). And despite its world-wide influence and position as a primary way through which the state and citizens may co-produce crime control, important questions remain about the drivers of participation in NW and the ways that participation is distributed on the basis of household and neighbourhood factors. Existing studies have indicated that it is difficult to both establish and sustain NW activities (Mukherjee and Wilson 1987; Rosenbaum 1988; Laycock and Tilley 1995) and have painted a mixed picture of the drivers of participation. But it has become conventionally understood that NW is more easily implemented in low crime rate, relatively affluent areas where citizens own their own homes (McConville and Shepherd 1992; Laycock and Tilley 1995; Hope 2000). NW then, may actually be regressive in nature, further concentrating crime control in the hands of those that need it least. This raises important questions about the distribution of security, the use of state resources and social justice (Rosenbaum 1987; Skogan 1988; Laycock and Tilley 1995; Hope 2000; Bullock 2014). Similarly, the ‘club goods’ associated with NW—the security marking, the watchful neighbours—may provide powerful assurances about potential risks, addressing particularly the needs of risk-averse property-owners (Hope 2000: 160). A number of empirical studies documented the proliferation of NW in the early 1980s (e.g. Husain 1988; Bennett 1989; Laycock and Tilley 1995; Yarwood and Edwards 1995), but few studies of NW have been conducted since the 1990s (Bennett et al. 2008). However, understanding the nature of the relationship between citizens and the structures of the state is arguably of greater importance than ever before. Since the establishment of the first English NW scheme in the 1980s, the relationship between public policing and citizens has been transformed. And the recent promotion of citizen involvement goes hand in hand with the political focus of successive British governments on ‘austerity’ following the global economic crisis of 2008, which has led to significant cuts to police funding (HM Treasury 2010; HMIC 2017) and the retrenchment of the public police, reducing the visibility of uniformed officers in local communities (HMIC 2017). But the efficacy of such approaches to crime control obviously relies on the nature of citizens’ relationship with the state, or at least with the police, and their ability and willingness to organize themselves at the local level. We use data from the Crime Survey for England and Wales (CSEW) to examine patterns of participation in NW in England and Wales since its inception in the neo-liberal governments of the early 1980s through to the end of the New Labour Era in 2010.1 Combining CSEW data with information from the census, we then use multilevel models to explore the extent that household drivers of participation in NW have changed since its introduction, and the role that area crime and disadvantage may play. In so doing, we consider the vexed issue of whether NW takes root in areas where it is least needed, i.e., in low crime and affluent areas. We find falling rates of participation since the early 1990s, with one-in-ten households in England and Wales part of NW by 2010–11. Participation is less likely in disadvantaged areas, but more likely in high-crime areas. We also find that crime rates moderate the effect of disadvantage on participation in NW. Disadvantaged areas with a higher level of crime are more likely to operate an NW scheme than similar areas with a low level of crime, but both are less likely to have a scheme in operation than areas with a low level of disadvantage. Thus, our findings challenge any assumptions about there being a straightforward linear relationship between crime rates and NW or disadvantage and NW. Instead, we conclude that citizens will participate in NW where the ‘conditions are right’. Literature review Though arrangements may vary in practice, the term co-production suggests that outcomes relevant to the provision of public services are being produced by the activity of public service agents together with the voluntary activities of citizens or groups of citizens (Mattson 1986; Pestoff 2006; Parrado et al. 2013; Alford and Yates 2016). In respect of NW, co-production primarily involves agents of the public police coming together with citizens or groups of citizens. The public police provide practical support to shore up the operation of NW (usually their time or in-kind support) and provide symbolic support as citizen groups seem to demand the legitimacy that police sponsorship confers (Gresham et al. 2004; Bullock 2014). The formal organization of citizens is not necessary for co-production, but organizations may facilitate co-ordination between citizens and public agencies and so enhance the levels of co-production (Pestoff 2006). Indeed, NW is supported by national infrastructure (see e.g. Gresham et al. 2004). However, co-productive relationships and activities are various and complex, and their contribution may be more or less allied to the core task of a public service (Brandsen and Honingh 2015). NW incorporates diverse activities more or less clearly linked to the core police task (Laycock and Tilley 1995; Gresham et al. 2004). Co-productive relationships and activities also generally create a mixture of public and private value (Alford and Yates 2016). Similarly, an aim of NW is to protect individual private properties and people (e.g. through provision of property marking or personal alarms) while reducing overall public risk (e.g. through reducing the opportunities for crime within a neighbourhood) (Rosenbaum 1987; 1988; Laycock and Tilley 1995). The reasons why citizens co-produce are varied (Alford 2009; Pestoff 2012; Verschuere et al. 2012; Alford and Yates 2016). While it is often held that citizens co-produce to receive something tangible or meaningful in return, citizens actually co-produce for a variety of extrinsic and intrinsic reasons (Alford 2009; Verschuere et al. 2012). Citizens will not, however, spontaneously co-produce simply because benefits could be achieved (Ostrom 1996: 1082). If co-producing is straightforward (Alford 2009; Pestoff 2012; Verschuere et al. 2012; Van Eijk and Steen 2016), or where an issue is salient or meaningful (Loeffler and Bovaird 2016; Van Eijk and Steen 2016), citizens are more likely to act. There also needs to be attention to organizational issues such as training public officials and establishing appropriate institutional arrangements (e.g. management, oversight and processes) (Ostrom 1996; Alford 2009; Pestoff 2012; Verschuere et al. 2012; Loeffler and Bovaird 2016; Voorberg et al. 2015). Understanding citizen’s needs and motivations is essential, as is the provision of information and incentives, and the clarification of expectations and outcomes (Ostrom 1996; Alford 2009). Neighbourhood-level factors also play a role, with the higher levels of social capital, denser networks and the presence of voluntary opportunities all increasing co-production (Voorberg et al. 2015; Van Eijk and Steen 2016; see also Putnam 2000). The presence of self-efficacy—a belief that one can make a meaningful difference—is also often identified as a stimulator of co-production (Parrado et al. 2013; Alford and Yates 2016; Loeffler and Bovaird 2016), as are high levels of trust and satisfaction with government (Loeffler and Bovaird 2016; Van Eijk and Steen 2016). Studies of the drivers of participation in NW reveal mixed results. While some studies demonstrate that crime and disorder may motivate citizens to participate (Lavrakas and Herz 1982; Skogan 1989; Pattavina et al. 2006), others find the opposite. High rates of crime may deter participation in NW for a number of reasons. They may generate feelings of powerlessness, increase social isolation and undermine any belief that residents have that they are able to collectively control crime (Skogan 1988). Or high-crime rates may generate suspicion among residents, weakening the trust needed to motivate them to work together to implement crime prevention interventions (Henig 1984; Hourihan 1987; Rosenbaum 1987; Skogan 1988; Hope 1995; Laycock and Tilley 1995). Especially where members are asked to exchange information about themselves, keep surveillance of and/or report on strangers (or neighbours) (Henig 1984; Hope 1995). The relationship between crime and participation in NW may not be linear (Husain 1988; Skogan 1989). For example, Skogan (1989) outlines a situation where residents in both high and low crime rate areas may be deterred from participating—in the former, the intervention may seem too mild to deal with the problem and with the latter there is no need—but in areas of average crime rates motivation may be high. Existing research tends to identify a negative relationship between neighbourhood disadvantage and membership of NW (Shernock 1986; Hope 1988; Husain 1988; Yarwood and Edwards 1995). This is believed to reflect both reduced opportunities for participation in more disadvantaged communities and that residents of poorer areas are less likely to be involved even when the opportunity is available (Hope 1995). However, Skogan (1989) argues that while there may be an effect of disadvantage when it comes to the opportunity to participate—there are more opportunities in better-off areas—this effect is diminished because people from better-off areas have fewer reasons to participate. By this reasoning, neighbourhood disadvantage must be considered in conjunction with the objective need for NW in the local area, as reflected, for example, by the crime rate. Pattavina et al. (2006) make a similar observation in explaining their finding that participation in community crime prevention was not highest in the wealthiest, lowest crime rate areas. Instead, the factors affecting citizen participation in crime prevention varied by neighbourhood crime risk levels. In high-crime neighbourhoods, residents who felt like part of the neighbourhood, minority residents and residents who believed that the police get to know residents were more likely to participate. In low- to moderate-risk neighbourhoods, a more complex web of predictors fuelled participation, with additional differences based on home ownership and previous victimization (Pattavina et al. 2006: 228). It is often assumed that citizens’ beliefs about the police are related to their willingness to engage in anti-crime measures at the neighbourhood level (Frank et al. 1996). Studies have suggested that participation in anti-crime initiatives, including NW, is facilitated where residents have favourable opinions towards the police and believe that the police do a good job (Shernock 1986; Bennett 1989; Yarwood and Edwards 1995). However, other studies find no evidence of differential participation once other factors were controlled for (e.g. Frank et al. 1996). Bennett (1989) found that participants in NW were more likely than non-participants to view the police as helpful, believed that the police were doing a good job, and were pleased with police performance. But participation was not associated with knowing a police officer by name, feeling the police understood the problems of residents living in their area or among those who had made proactive contact with the police over the last year. This suggests that the nature of the relationship with the police may be playing a role in shaping participation. Indeed, Reisig (2007) found that citizens who judged police practices to be fair and respectful were more willing to participate in community crime prevention initiatives. This finding held across areas with low, moderate and high levels of property crime. Pattavina et al. (2006) also found that participation in crime prevention activities was less to do with perceptions of the effectiveness of the police and more to do with the development of personal relationships between the police and residents in these areas. North American research tends to find that participation in organized anti-crime initiatives can be understood as an extension of ‘civic mindedness’ and wider participation in community organizations (Rosenbaum 1988; Skogan 1988; Bennett 1989, and see studies by DuBow and Podolefsky 1982; Lavrakas and Herz 1982; Shernock 1986). Bennett (1989) also found participants in NW were more involved in their community than non-participants, leading him to conclude that participation might be a function of both crime and wider community participation. So while the salience of crime provides the external, environmental impetus for the development of community crime prevention programs, this may be more a necessary rather than a sufficient condition for their origin (Lavrakas and Herz 1982). The degree of social cohesion—measured by factors such as whether citizens have confidence in the ability of others to combat crime, felt they were friends with their neighbours and reported territorial attitudes or supportive environments—has also been linked with participation in NW (Lavrakas and Herz 1982; Hope 1988; Frank et al. 1996). Skogan (1989) found a tendency for less cohesive communities to employ more formal responses to crime, in part because in these areas informal capacity to problem-solve and organize against crime was weak. Others have reported mixed evidence. Bennett (1989), for example, reported significant differences in one of his research sites but not in the other. In addition, Pattavina et al. (2006) revealed a link between one indicator of social cohesion (residents feel like they are part of the neighbourhood) and participation but not another (rely on neighbours for help). While studies that have examined the relationship between ‘attachment’ to the community and participation in NW reveal mixed findings, it has often been shown that members of NW have a ‘stake in the community’ (Shernock 1986; Frank et al. 1996; Pattavina et al. 2006). Especially important has been long-term residency in an area, home ownership and housing type—with those living in flats often less likely to participate (Lavrakas and Herz 1982; Henig 1984; Hope 1988; Husain 1988; Bennett 1989; Webb 1994; Frank et al. 1996). This has generally been explained by the belief that homeowners have greater commitment to maintaining standards within an area because of incentives to maintain house prices and their greater freedom and responsibility for taking prevention measures (Hope 1995: 45). There are also socio-demographic differences in participation in NW. But while most research points to NW members being older (Shernock 1986; Webb 1994; Yarwood and Edwards 1995), even this is not universally agreed. For example, Lavrakas and Herz (1982) find participants tended to be aged less than 50 years, and Bennett (1989) reported no differences in participation on the basis of age. The evidence regarding the role played by ethnicity and gender is also mixed. Reviews of US-based studies have suggested that Black residents are more likely to participate (e.g. Bennett 1989), while other studies demonstrate that NW members are more commonly White (Henig 1984; Shernock 1986; Webb 1994; Pattavina et al. 2006) and some find no differences between ethnic groups. Similarly, while some studies have shown that NW members are more likely to be men (Hope 1988; Husain 1988; Yarwood and Edwards 1995), others have found participants are more likely to be women (Webb 1994). Data To assess the changing levels of NW in England and Wales and explore the key household and neighbourhood factors that remain important for participation in the co-production of NW, we use data from the CSEW. This is a nationally representative survey of residents’ experiences of crime and victimization. It has fielded questions about NW to a random subsample of respondents since 1988, allowing us to provide an overall picture of the changing prevalence of NW over the last three decades. The questions were omitted from the survey in 1998, 2002/03–2003/04 and 2007/08–2008/09, which limits our ability to provide a complete time trend. The survey has consistently maintained a response rate of more than 70%, achieving a response of 75% in 2010–11 (ONS 2012). Long-term trends in Neighbourhood Watch proliferation in England and Wales Changing levels of NW proliferation since 1988 are examined using annual weighted estimates of three survey items: 1. Is there a Neighbourhood Watch Scheme currently operating in this area that covers your address? Yes, No, Don’t know2 2. Is your household currently a member of the scheme? Yes, No, Don’t know 3. Would your household join a scheme if there were one in this area? Yes, No, Don’t know For each survey year, the percentage of households reporting that the area is part of an NW scheme, levels of household participation (overall, and restricted to areas where a scheme was in operation) and the percentage of residents reporting that they would like to belong to an NW scheme if it were available, are collated. We then use time-series cross-sectional models (Gelman and Hill 2007), with observations grouped in survey year, to examine whether overall trends in the prevalence of NW are consistent across household groups. Comparatively, few measures in the CSEW have been included in a consistent way across all survey waves, so we restrict our focus to differences based on accommodation type (detached, semi-detached, terraced, flats or other), household income, housing tenure (owner occupied, private rented, social rented) and reported levels of safety at home at night. We also include controls for respondent gender, age and length of residence in the local area. All household characteristics are interacted with survey year to enable assessment of whether observed differences in levels of participation have changed over time. Models also include a quadratic function of year to account for potential non-linearities. Results Throughout the 1990s, there was a general increase in the number of NW schemes in operation in England and Wales, with more than 30% of areas identified as part of a scheme by 2000, up from less than 20% of areas in 1988 (Figure 1, top panel). The number of areas operating NW then remained stable throughout the 2000s and no visible increase following the 2008 financial crisis.3 However, while the number of areas with a scheme appears to have remained comparatively stable, the number of households opting to be part of a scheme has fallen steadily (Figure 1, middle panel). Approximately 40% of households in eligible areas were part of NW by 2010–11, down from more than 80% in 1988. The main reason for not being a part of a scheme (reported by more than one quarter of respondents) was that they had not been asked to join. Other frequently noted reasons were being too busy, or not yet getting round to it, and not knowing how to join or having insufficient information about it. Only 6% said they were not interested, and just 1% did not think they were effective (for full results, see Appendix Table A1). The comparatively low levels of participation do not, however, appear to be reflective of a lack of demand, with the number of people indicating that they would join a scheme if one was available remaining relatively stable and more than 70% since 1988 (Figure 1, bottom panel). Of course, the high number of people indicating that they would like to join a scheme may in part be reflective of a social desirability bias, with respondents using this item to offset their existing lack of involvement (Nunnally 1978). Fig. 1 Open in new tabDownload slide Trends in participation in Neighbourhood Watch (NW). Respondent reports that area is part of an NW scheme (top panel); household a member of NW (middle panel); would join an NW scheme if one was available (bottom panel) Fig. 1 Open in new tabDownload slide Trends in participation in Neighbourhood Watch (NW). Respondent reports that area is part of an NW scheme (top panel); household a member of NW (middle panel); would join an NW scheme if one was available (bottom panel) Table 1 includes results from the repeated cross-sectional models, confirming the initial growth of NW and subsequent declines. Residents of semi-detached properties, terraces and flats are less likely to report being in an NW area than those living in detached accommodation (Model 1). Wealthier households, owner-occupiers and long-term residents are also more likely to report being in an NW area. These differences are generally stable over time, although there is moderate evidence that the income gap has reduced, and there has also been a small increase in reported NW prevalence among those living in social housing. A similar picture is evident when household membership of NW is considered (Model 2), although there is evidence of further declines in membership among semi-detached and terraced accommodation. Finally, we note lower odds of future NW enrolment among resident living in flats, those living in rental accommodation and those who report feeling safe at home after dark (Model 3). Conversely, women are more likely to indicate they would join a scheme, as are those with higher levels of income. Table 1 Multilevel repeated cross-sectional analysis of changing participation in Neighbourhood Watch (1988–2010/11) . Model 1: NW in area . Model 2: House part of NW . Model 3: Would join NW . OR . SE . OR . SE . OR . SE . Year 1.11** 0.03 1.08** 0.02 1.00 0.02 Year2 0.995** 0.00 0.996** 0.00 1.00 0.00 Accommodation (ref.: detached) Semi-detached 0.81** 0.05 0.79** 0.05 0.92 0.08 Terraced 0.57** 0.04 0.51** 0.04 0.89 0.08 Flat 0.46** 0.04 0.33** 0.04 0.72** 0.07 Other 0.79 0.11 0.60** 0.10 0.77** 0.12 Semi-detached × year 0.99 0.00 0.98** 0.00 0.99 0.01 Terraced × year 1.00 0.00 0.99* 0.01 0.99* 0.01 Flat × year 1.01 0.01 1.01 0.01 0.98* 0.01 Other × year 1.00 0.01 1.00 0.01 0.98 0.01 Income (ref.: under £20,000) £20,000+ 1.53** 0.08 1.65** 0.09 1.41** 0.10 £20,000+ × year 0.99** 0.00 0.98** 0.00 0.99 0.00 Tenure (ref.: owner) Private rented 0.73** 0.07 0.76* 0.09 0.50** 0.05 Social rented 0.55** 0.03 0.51** 0.04 0.63** 0.04 Private rented × year 1.01 0.01 0.99 0.01 1.01 0.01 Social rented × year 1.03** 0.00 1.01* 0.01 0.99 0.01 Feel very safe at home 0.95 0.04 1.02 0.05 0.84** 0.05 Feel very safe at home × year 1.01* 0.00 1.00 0.00 1.00 0.00 Female 1.03 0.02 1.09** 0.03 1.21** 0.03 Age (centred) 1.01** 0.00 1.02** 0.00 0.998** 0.00 Years in area 1.04** 0.01 1.15** 0.01 0.94** 0.01 Constant 0.39 0.06 0.16 0.02 5.96 0.85 Random effects Year 0.02 0.01 0.01 0.01 0.01 0.01 Years/individuals 11/46,177 11/50,056 11/32,283 . Model 1: NW in area . Model 2: House part of NW . Model 3: Would join NW . OR . SE . OR . SE . OR . SE . Year 1.11** 0.03 1.08** 0.02 1.00 0.02 Year2 0.995** 0.00 0.996** 0.00 1.00 0.00 Accommodation (ref.: detached) Semi-detached 0.81** 0.05 0.79** 0.05 0.92 0.08 Terraced 0.57** 0.04 0.51** 0.04 0.89 0.08 Flat 0.46** 0.04 0.33** 0.04 0.72** 0.07 Other 0.79 0.11 0.60** 0.10 0.77** 0.12 Semi-detached × year 0.99 0.00 0.98** 0.00 0.99 0.01 Terraced × year 1.00 0.00 0.99* 0.01 0.99* 0.01 Flat × year 1.01 0.01 1.01 0.01 0.98* 0.01 Other × year 1.00 0.01 1.00 0.01 0.98 0.01 Income (ref.: under £20,000) £20,000+ 1.53** 0.08 1.65** 0.09 1.41** 0.10 £20,000+ × year 0.99** 0.00 0.98** 0.00 0.99 0.00 Tenure (ref.: owner) Private rented 0.73** 0.07 0.76* 0.09 0.50** 0.05 Social rented 0.55** 0.03 0.51** 0.04 0.63** 0.04 Private rented × year 1.01 0.01 0.99 0.01 1.01 0.01 Social rented × year 1.03** 0.00 1.01* 0.01 0.99 0.01 Feel very safe at home 0.95 0.04 1.02 0.05 0.84** 0.05 Feel very safe at home × year 1.01* 0.00 1.00 0.00 1.00 0.00 Female 1.03 0.02 1.09** 0.03 1.21** 0.03 Age (centred) 1.01** 0.00 1.02** 0.00 0.998** 0.00 Years in area 1.04** 0.01 1.15** 0.01 0.94** 0.01 Constant 0.39 0.06 0.16 0.02 5.96 0.85 Random effects Year 0.02 0.01 0.01 0.01 0.01 0.01 Years/individuals 11/46,177 11/50,056 11/32,283 NW, Neighbourhood Watch; OR, odds ratio. *p < 0.05, **p < 0.01. Open in new tab Table 1 Multilevel repeated cross-sectional analysis of changing participation in Neighbourhood Watch (1988–2010/11) . Model 1: NW in area . Model 2: House part of NW . Model 3: Would join NW . OR . SE . OR . SE . OR . SE . Year 1.11** 0.03 1.08** 0.02 1.00 0.02 Year2 0.995** 0.00 0.996** 0.00 1.00 0.00 Accommodation (ref.: detached) Semi-detached 0.81** 0.05 0.79** 0.05 0.92 0.08 Terraced 0.57** 0.04 0.51** 0.04 0.89 0.08 Flat 0.46** 0.04 0.33** 0.04 0.72** 0.07 Other 0.79 0.11 0.60** 0.10 0.77** 0.12 Semi-detached × year 0.99 0.00 0.98** 0.00 0.99 0.01 Terraced × year 1.00 0.00 0.99* 0.01 0.99* 0.01 Flat × year 1.01 0.01 1.01 0.01 0.98* 0.01 Other × year 1.00 0.01 1.00 0.01 0.98 0.01 Income (ref.: under £20,000) £20,000+ 1.53** 0.08 1.65** 0.09 1.41** 0.10 £20,000+ × year 0.99** 0.00 0.98** 0.00 0.99 0.00 Tenure (ref.: owner) Private rented 0.73** 0.07 0.76* 0.09 0.50** 0.05 Social rented 0.55** 0.03 0.51** 0.04 0.63** 0.04 Private rented × year 1.01 0.01 0.99 0.01 1.01 0.01 Social rented × year 1.03** 0.00 1.01* 0.01 0.99 0.01 Feel very safe at home 0.95 0.04 1.02 0.05 0.84** 0.05 Feel very safe at home × year 1.01* 0.00 1.00 0.00 1.00 0.00 Female 1.03 0.02 1.09** 0.03 1.21** 0.03 Age (centred) 1.01** 0.00 1.02** 0.00 0.998** 0.00 Years in area 1.04** 0.01 1.15** 0.01 0.94** 0.01 Constant 0.39 0.06 0.16 0.02 5.96 0.85 Random effects Year 0.02 0.01 0.01 0.01 0.01 0.01 Years/individuals 11/46,177 11/50,056 11/32,283 . Model 1: NW in area . Model 2: House part of NW . Model 3: Would join NW . OR . SE . OR . SE . OR . SE . Year 1.11** 0.03 1.08** 0.02 1.00 0.02 Year2 0.995** 0.00 0.996** 0.00 1.00 0.00 Accommodation (ref.: detached) Semi-detached 0.81** 0.05 0.79** 0.05 0.92 0.08 Terraced 0.57** 0.04 0.51** 0.04 0.89 0.08 Flat 0.46** 0.04 0.33** 0.04 0.72** 0.07 Other 0.79 0.11 0.60** 0.10 0.77** 0.12 Semi-detached × year 0.99 0.00 0.98** 0.00 0.99 0.01 Terraced × year 1.00 0.00 0.99* 0.01 0.99* 0.01 Flat × year 1.01 0.01 1.01 0.01 0.98* 0.01 Other × year 1.00 0.01 1.00 0.01 0.98 0.01 Income (ref.: under £20,000) £20,000+ 1.53** 0.08 1.65** 0.09 1.41** 0.10 £20,000+ × year 0.99** 0.00 0.98** 0.00 0.99 0.00 Tenure (ref.: owner) Private rented 0.73** 0.07 0.76* 0.09 0.50** 0.05 Social rented 0.55** 0.03 0.51** 0.04 0.63** 0.04 Private rented × year 1.01 0.01 0.99 0.01 1.01 0.01 Social rented × year 1.03** 0.00 1.01* 0.01 0.99 0.01 Feel very safe at home 0.95 0.04 1.02 0.05 0.84** 0.05 Feel very safe at home × year 1.01* 0.00 1.00 0.00 1.00 0.00 Female 1.03 0.02 1.09** 0.03 1.21** 0.03 Age (centred) 1.01** 0.00 1.02** 0.00 0.998** 0.00 Years in area 1.04** 0.01 1.15** 0.01 0.94** 0.01 Constant 0.39 0.06 0.16 0.02 5.96 0.85 Random effects Year 0.02 0.01 0.01 0.01 0.01 0.01 Years/individuals 11/46,177 11/50,056 11/32,283 NW, Neighbourhood Watch; OR, odds ratio. *p < 0.05, **p < 0.01. Open in new tab Explaining membership in Neighbourhood Watch To provide a more detailed assessment of the household and neighbourhood characteristics that remain influential for participation in NW, we estimate multilevel logistic regression models (Goldstein 2011) for each of our three measures of NW participation in 2010–11. Household characteristics To account for differences in involvement in NW between households, we include details of the household reference person’s gender, age, ethnicity and socio-economic status. We also record the number of children in the household, how long they have been resident in the area and whether it is a multiple adult household. Details of the type of accommodation, housing tenure, whether there is a visible burglar alarm, whether the house is left unoccupied during the day and whether the house is surrounded by visible signs of disorder4 are also included. Finally, we also identify those respondents who reported being a victim of crime in the previous 12 months, whether they report worrying about crime in general and whether they believe the local police do a good job (all binary coded).5 Full descriptive details are included in Appendix Table A2. Area characteristics To examine differences in NW participation between different local communities, we link the household-level data from the CSEW to Middle layer Super Output Areas (MSOA)—a census geography with each area comprising approximately 2,000 households that were grouped together based on similarity of tenure and accommodation type. Respondents are clustered in a total of 2,230 MSOA. For each area, we include a number of area measures derived using data from the 2001 census.6 A total of 21 different census variables were combined using a factorial ecology approach (Rees 1971) to produce five distinct indices measuring differences between each area (full factor loadings in Appendix Table A3). These cover: Concentrated disadvantage—with more disadvantaged areas typically having higher numbers of single-parent families, households on income support and unemployed people, but fewer residents identified as in managerial and professional occupations, and with fewer owner-occupiers. Urbanicity—higher scores characterize areas with higher population density and more domestic properties, but relatively little green space. Population mobility—areas score higher if there are higher levels of in- and out-migration, as well as more single-person households. Age profile—higher scores associated with areas that have a younger population. Housing structure—higher scores for areas with more terraced and vacant properties, but fewer flats. We also include a general measure of the level of crime in each area, calculated as the population weighted average score for the crime component of the 2010 Index of Deprivation. This is the most temporally proximate year of the Index of Deprivation that precedes our survey data, and is based on police recorded crime in 2008. In addition to main effects for all neighbourhood characteristics, we also include the interaction between the neighbourhood level of concentrated disadvantage and the crime rate to see whether objective crime risks moderate the link from deprivation to participation in NW (Skogan 1989; Pattavina et al. 2006). Results By 2010–11, NW was operating in one third of areas in England and Wales, although only 40% of households in those areas were active members of the scheme. Table 1 includes results from three multilevel models examining in more detail which areas and households were most likely to be part of NW.7 Consistent with expectations, we find that NW is significantly less likely to be in operation in areas where there are greater levels of concentrated disadvantage, with the odds of a scheme being in operation approximately 24% lower for every 1 SD increase in the level of disadvantage (Model 4). NW schemes are also significantly less likely to be found in more urban areas (with an estimated odds ratio of 0.85) and areas with a younger age profile (0.91). We also find NW is less likely in areas that are characterized by more terraced accommodation and vacant properties (0.87). These areas also tend to have slightly lower than average numbers of flats, suggesting a more complex picture of the role that housing structure may play in the presence of NW schemes (Table 2). Table 2 Multilevel logistic regression models examining participation in Neighbourhood Watch (2010–11) . Model 4: NW in area (excluding DK) . Model 5: House part of NW . Model 6: Would join NW . OR . SE . OR . SE . OR . SE . Household characteristics Female (HRPa) 0.92 0.08 1.11 0.17 1.14 0.12 Age (HRPa) 1.00 0.00 1.03** 0.01 1.00 0.00 Number of children in HH 0.91* 0.04 0.96 0.09 1.01 0.06 Years in area 1.02 0.03 1.15** 0.05 0.98 0.03 Cohabiting HH 1.03 0.09 1.24 0.19 1.51** 0.17 Black, Asian and Minority Ethnic (HRPa) 1.11 0.17 0.89 0.28 1.51* 0.31 Accommodation (ref.: detached) Semi-detached 0.86 0.09 0.43** 0.08 1.05 0.15 Terraced 0.82 0.10 0.41** 0.09 0.83 0.13 Flat/other 0.90 0.14 0.51* 0.14 0.63* 0.12 Socio-economic status (ref.: managerial/professional) Intermediate occupations 1.26 0.17 0.74 0.18 1.05 0.19 Small employers and own account workers 0.91 0.11 0.75 0.16 0.81 0.13 Lower supervisory and technical occupations 1.06 0.13 0.87 0.19 0.92 0.15 Semi-routine and routine occupations 1.03 0.10 0.69* 0.13 0.97 0.12 Never worked and long-term unemployed 1.14 0.26 0.98 0.41 0.60 0.16 Tenure (ref.: owner-occupier) Social rented sector 1.28* 0.15 0.86 0.19 0.59** 0.08 Private rented sector 0.84 0.10 0.51** 0.13 0.69** 0.10 Victim of crime 0.93 0.08 1.31 0.22 1.00 0.11 Worried about crime overall 1.03 0.08 1.12 0.16 1.70** 0.18 Local police do a good job 1.45** 0.11 1.29 0.18 1.25* 0.12 Has a visible burglar alarm 1.20* 0.10 1.18 0.18 1.12 0.13 Home unoccupied during week 0.99 0.12 1.22 0.28 1.20 0.19 Household disorder 0.83** 0.04 0.81* 0.08 0.90* 0.05 Area characteristics Concentrated disadvantage 0.76** 0.05 0.72** 0.09 0.95 0.07 Crime rate 1.29** 0.12 1.45* 0.25 0.91 0.10 Crime rate × disadvantage 1.14* 0.07 1.22 0.14 1.15 0.09 Urbanicity 0.85** 0.04 0.96 0.09 1.05 0.07 Population mobility 0.98 0.05 0.91 0.08 1.02 0.06 Age profile 0.91* 0.04 0.96 0.07 0.99 0.05 Housing structure 0.86** 0.04 0.99 0.09 0.96 0.05 Constant 0.41 0.09 0.37 0.15 2.32 0.65 Random effects Area 0.52 0.12 0.81 0.41 0.32 0.18 Approximate ICC 0.14 0.20 0.09 Areas/individuals 2,230/4,237 1,126/1,550 1,874/2,999 . Model 4: NW in area (excluding DK) . Model 5: House part of NW . Model 6: Would join NW . OR . SE . OR . SE . OR . SE . Household characteristics Female (HRPa) 0.92 0.08 1.11 0.17 1.14 0.12 Age (HRPa) 1.00 0.00 1.03** 0.01 1.00 0.00 Number of children in HH 0.91* 0.04 0.96 0.09 1.01 0.06 Years in area 1.02 0.03 1.15** 0.05 0.98 0.03 Cohabiting HH 1.03 0.09 1.24 0.19 1.51** 0.17 Black, Asian and Minority Ethnic (HRPa) 1.11 0.17 0.89 0.28 1.51* 0.31 Accommodation (ref.: detached) Semi-detached 0.86 0.09 0.43** 0.08 1.05 0.15 Terraced 0.82 0.10 0.41** 0.09 0.83 0.13 Flat/other 0.90 0.14 0.51* 0.14 0.63* 0.12 Socio-economic status (ref.: managerial/professional) Intermediate occupations 1.26 0.17 0.74 0.18 1.05 0.19 Small employers and own account workers 0.91 0.11 0.75 0.16 0.81 0.13 Lower supervisory and technical occupations 1.06 0.13 0.87 0.19 0.92 0.15 Semi-routine and routine occupations 1.03 0.10 0.69* 0.13 0.97 0.12 Never worked and long-term unemployed 1.14 0.26 0.98 0.41 0.60 0.16 Tenure (ref.: owner-occupier) Social rented sector 1.28* 0.15 0.86 0.19 0.59** 0.08 Private rented sector 0.84 0.10 0.51** 0.13 0.69** 0.10 Victim of crime 0.93 0.08 1.31 0.22 1.00 0.11 Worried about crime overall 1.03 0.08 1.12 0.16 1.70** 0.18 Local police do a good job 1.45** 0.11 1.29 0.18 1.25* 0.12 Has a visible burglar alarm 1.20* 0.10 1.18 0.18 1.12 0.13 Home unoccupied during week 0.99 0.12 1.22 0.28 1.20 0.19 Household disorder 0.83** 0.04 0.81* 0.08 0.90* 0.05 Area characteristics Concentrated disadvantage 0.76** 0.05 0.72** 0.09 0.95 0.07 Crime rate 1.29** 0.12 1.45* 0.25 0.91 0.10 Crime rate × disadvantage 1.14* 0.07 1.22 0.14 1.15 0.09 Urbanicity 0.85** 0.04 0.96 0.09 1.05 0.07 Population mobility 0.98 0.05 0.91 0.08 1.02 0.06 Age profile 0.91* 0.04 0.96 0.07 0.99 0.05 Housing structure 0.86** 0.04 0.99 0.09 0.96 0.05 Constant 0.41 0.09 0.37 0.15 2.32 0.65 Random effects Area 0.52 0.12 0.81 0.41 0.32 0.18 Approximate ICC 0.14 0.20 0.09 Areas/individuals 2,230/4,237 1,126/1,550 1,874/2,999 DK, don't knows; HH, household; HRP, Household Reference Person; ICC, intraclass correlation; NW, Neighbourhood Watch; OR, odds ratio. aUses respondent in Model 3 because asking for individual opinion. *p < 0.05, **p < 0.01. Open in new tab Table 2 Multilevel logistic regression models examining participation in Neighbourhood Watch (2010–11) . Model 4: NW in area (excluding DK) . Model 5: House part of NW . Model 6: Would join NW . OR . SE . OR . SE . OR . SE . Household characteristics Female (HRPa) 0.92 0.08 1.11 0.17 1.14 0.12 Age (HRPa) 1.00 0.00 1.03** 0.01 1.00 0.00 Number of children in HH 0.91* 0.04 0.96 0.09 1.01 0.06 Years in area 1.02 0.03 1.15** 0.05 0.98 0.03 Cohabiting HH 1.03 0.09 1.24 0.19 1.51** 0.17 Black, Asian and Minority Ethnic (HRPa) 1.11 0.17 0.89 0.28 1.51* 0.31 Accommodation (ref.: detached) Semi-detached 0.86 0.09 0.43** 0.08 1.05 0.15 Terraced 0.82 0.10 0.41** 0.09 0.83 0.13 Flat/other 0.90 0.14 0.51* 0.14 0.63* 0.12 Socio-economic status (ref.: managerial/professional) Intermediate occupations 1.26 0.17 0.74 0.18 1.05 0.19 Small employers and own account workers 0.91 0.11 0.75 0.16 0.81 0.13 Lower supervisory and technical occupations 1.06 0.13 0.87 0.19 0.92 0.15 Semi-routine and routine occupations 1.03 0.10 0.69* 0.13 0.97 0.12 Never worked and long-term unemployed 1.14 0.26 0.98 0.41 0.60 0.16 Tenure (ref.: owner-occupier) Social rented sector 1.28* 0.15 0.86 0.19 0.59** 0.08 Private rented sector 0.84 0.10 0.51** 0.13 0.69** 0.10 Victim of crime 0.93 0.08 1.31 0.22 1.00 0.11 Worried about crime overall 1.03 0.08 1.12 0.16 1.70** 0.18 Local police do a good job 1.45** 0.11 1.29 0.18 1.25* 0.12 Has a visible burglar alarm 1.20* 0.10 1.18 0.18 1.12 0.13 Home unoccupied during week 0.99 0.12 1.22 0.28 1.20 0.19 Household disorder 0.83** 0.04 0.81* 0.08 0.90* 0.05 Area characteristics Concentrated disadvantage 0.76** 0.05 0.72** 0.09 0.95 0.07 Crime rate 1.29** 0.12 1.45* 0.25 0.91 0.10 Crime rate × disadvantage 1.14* 0.07 1.22 0.14 1.15 0.09 Urbanicity 0.85** 0.04 0.96 0.09 1.05 0.07 Population mobility 0.98 0.05 0.91 0.08 1.02 0.06 Age profile 0.91* 0.04 0.96 0.07 0.99 0.05 Housing structure 0.86** 0.04 0.99 0.09 0.96 0.05 Constant 0.41 0.09 0.37 0.15 2.32 0.65 Random effects Area 0.52 0.12 0.81 0.41 0.32 0.18 Approximate ICC 0.14 0.20 0.09 Areas/individuals 2,230/4,237 1,126/1,550 1,874/2,999 . Model 4: NW in area (excluding DK) . Model 5: House part of NW . Model 6: Would join NW . OR . SE . OR . SE . OR . SE . Household characteristics Female (HRPa) 0.92 0.08 1.11 0.17 1.14 0.12 Age (HRPa) 1.00 0.00 1.03** 0.01 1.00 0.00 Number of children in HH 0.91* 0.04 0.96 0.09 1.01 0.06 Years in area 1.02 0.03 1.15** 0.05 0.98 0.03 Cohabiting HH 1.03 0.09 1.24 0.19 1.51** 0.17 Black, Asian and Minority Ethnic (HRPa) 1.11 0.17 0.89 0.28 1.51* 0.31 Accommodation (ref.: detached) Semi-detached 0.86 0.09 0.43** 0.08 1.05 0.15 Terraced 0.82 0.10 0.41** 0.09 0.83 0.13 Flat/other 0.90 0.14 0.51* 0.14 0.63* 0.12 Socio-economic status (ref.: managerial/professional) Intermediate occupations 1.26 0.17 0.74 0.18 1.05 0.19 Small employers and own account workers 0.91 0.11 0.75 0.16 0.81 0.13 Lower supervisory and technical occupations 1.06 0.13 0.87 0.19 0.92 0.15 Semi-routine and routine occupations 1.03 0.10 0.69* 0.13 0.97 0.12 Never worked and long-term unemployed 1.14 0.26 0.98 0.41 0.60 0.16 Tenure (ref.: owner-occupier) Social rented sector 1.28* 0.15 0.86 0.19 0.59** 0.08 Private rented sector 0.84 0.10 0.51** 0.13 0.69** 0.10 Victim of crime 0.93 0.08 1.31 0.22 1.00 0.11 Worried about crime overall 1.03 0.08 1.12 0.16 1.70** 0.18 Local police do a good job 1.45** 0.11 1.29 0.18 1.25* 0.12 Has a visible burglar alarm 1.20* 0.10 1.18 0.18 1.12 0.13 Home unoccupied during week 0.99 0.12 1.22 0.28 1.20 0.19 Household disorder 0.83** 0.04 0.81* 0.08 0.90* 0.05 Area characteristics Concentrated disadvantage 0.76** 0.05 0.72** 0.09 0.95 0.07 Crime rate 1.29** 0.12 1.45* 0.25 0.91 0.10 Crime rate × disadvantage 1.14* 0.07 1.22 0.14 1.15 0.09 Urbanicity 0.85** 0.04 0.96 0.09 1.05 0.07 Population mobility 0.98 0.05 0.91 0.08 1.02 0.06 Age profile 0.91* 0.04 0.96 0.07 0.99 0.05 Housing structure 0.86** 0.04 0.99 0.09 0.96 0.05 Constant 0.41 0.09 0.37 0.15 2.32 0.65 Random effects Area 0.52 0.12 0.81 0.41 0.32 0.18 Approximate ICC 0.14 0.20 0.09 Areas/individuals 2,230/4,237 1,126/1,550 1,874/2,999 DK, don't knows; HH, household; HRP, Household Reference Person; ICC, intraclass correlation; NW, Neighbourhood Watch; OR, odds ratio. aUses respondent in Model 3 because asking for individual opinion. *p < 0.05, **p < 0.01. Open in new tab Areas with a higher crime rate are significantly more likely to be part of NW, although this effect is dependent on the level of neighbourhood disadvantage. Figure 2 demonstrates this moderating effect for a ‘typical’ area (when all other individual and neighbourhood characteristics are held at their mean). In areas where there is a low level of neighbourhood disadvantage (bottom 25% of the distribution), the probability of an NW scheme being in operation is comparatively high and is not influenced by the surrounding levels of crime. Conversely, in areas with higher levels of disadvantage (top 25% of the distribution), NW is less likely to be in operation and is more closely dependent on the levels of crime. So disadvantaged areas with a higher level of crime are more likely to operate an NW scheme than similar areas with a low level of crime, but both are less likely to have a scheme in operation than areas with a low level of disadvantage. Additional sensitivity checks restricting the analysis to the most disadvantaged neighbourhoods confirmed this non-linear relationship with crime, but that all other effects operated as shown in the main models. Significant differences between neighbourhoods are still evident when these area characteristics have been accounted for, with a residual neighbourhood variance of 0.52. Fig. 2 Open in new tabDownload slide Predicted probability that area is part of a Neighbourhood Watch scheme. Predicted probabilities are calculated for the ‘typical’ area, with all other variables held at their mean value Fig. 2 Open in new tabDownload slide Predicted probability that area is part of a Neighbourhood Watch scheme. Predicted probabilities are calculated for the ‘typical’ area, with all other variables held at their mean value Differences are also evident between particular types of households. Households that hold more positive views of the local police have almost 50% higher odds of living in NW areas. Social housing and households that have a visible burglar alarm are also more likely to be in NW areas. In contrast, households surrounded by more signs of low-level disorder are significantly less likely to be in NW areas (0.83), as are households containing more children (0.91). Restricting the focus to those areas where a scheme is known to be operating (Model 5), households are more likely to belong to NW if they have been resident in the area for longer and when the head of the household is older. People living in terraced housing, semi-detached accommodation and flats are less likely to participate, as are those who rent their property. People living in houses surrounded by greater levels of disorder are also significantly less likely to participate. The included area characteristics are generally not influential in determining whether particular households opt to be part of an NW scheme, although we do find that houses in more disadvantaged areas are less likely to participate while those in high-crime areas are more likely to be involved. Significant differences between neighbourhoods remain, with a residual neighbourhood variance of 0.81 suggesting that other features of the local neighbourhood may be influential in determining whether particular households choose to opt into an NW scheme. Despite the steady fall in the proportion of households opting to be part of NW since 1988, the levels of interest from non-member households have remained relatively stable. Model 6 shows that Black, Asian and Other Minority ethnic households have approximately 50% higher odds of indicating that they would be willing to join a scheme. People who worry about crime and those who have more favourable attitudes towards the police are also more likely to want to be part of NW. In contrast, those living in flats, as well as those in rented accommodation, have significantly lower odds of wanting to participate. We also find that people living in houses that are surrounded by more visible signs of disorder are also less likely to want to be part of NW. However, we find no systematic differences between areas. Conclusion This article has examined patterns of participation in NW in England and Wales since its inception in the early 1980s to the end of the New Labour era (2010–11). This issue is an important one since the co-production of services has been a major theme within the narratives of successive governments in England and Wales, as elsewhere. The influence of co-production is also contested contest as it is based on an assumption that citizens will work with service providers to construct public goods (Mattson 1986; Pestoff 2009). We add to the theoretical and empirical debates regarding who ‘produces’ crime control in two key ways. First, we add to the body of literature that suggests that it should not be assumed that citizens will co-produce (Osborne and Strokosch 2013). We find only moderate uptake of NW in England and Wales and a general decline in individual membership. While the number of areas with NW in England and Wales rose quickly following their initial inception in the early 1980s, they have remained stable, at about 30%, since the early 2000s. And at the same time, household membership has been falling—from 80% in 1988 to about 40% by 2010–11. As a result, only one-in-ten households in England and Wales was part of NW by this time. NW rose to prominence at the height of the neo-liberal governments of the 1980s–1990s. But declined during or shortly after New Labour politics came to the fore. The reasons for this cannot be revealed by the present analysis. However, our findings undermine any assumption that citizens are routinely prepared to accept the burden of crime control that current political discourses promote and economic necessity may demand. They also question the long-term viability of approaches to crime control which seek to co-opt the initiative of citizens (see also Mukherjee and Wilson 1987; Rosenbaum 1988). Second, we add to the body of literature regarding who co-produces, and in particular the vexed question of whether NW has regressive effects (Rosenbaum 1987; Skogan 1988; Laycock and Tilley 1995; Hope 2000; Bullock 2014). NW has often been associated with wealthier, long-term residents who own their own homes and so have attachments to the security of place and property (e.g. Lavrakas and Herz 1982; Henig 1984; Shernock 1986; Hope 1988; McConville and Shepherd 1992; Frank et al. 1996; Pattavina et al. 2006). We make similar observations though we report a small increase in reported NW membership among those in social housing. Indeed, consistent with expectations, we found that NW is significantly less likely to be in operation in areas where there are greater levels of concentrated disadvantage. However, we have also demonstrated that crime rates moderate the effect of disadvantage. And that disadvantaged areas with a higher level of crime are more likely to operate an NW scheme than similar areas with a low level of crime. There are several possible explanations for this. First, in areas of relative affluence, the costs of establishing and maintaining NW, which may not ask much of its members, may be low. Hence, they can be established even where crime risks are low and action again crime not a necessity. The situation is different in less affluent areas. In these cases, there are hurdles—such as fewer opportunities to volunteer, lower levels of trust between citizens and between citizens and public officials—that may need to be overcome to seed and sustain activity are more extensive. As such, reflecting studies that suggest that propensity to co-produce is related to the salience of a social problem (Loeffler and Bovaird 2016; Van Eijk and Steen 2016) efforts may be warranted only where the crime risk is particularly elevated (see also Skogan 1989). Second, it is plausible that practitioners exert greater efforts to establish NW in disadvantaged areas with higher crime risk (Gresham et al. 2004). However, research has failed to demonstrate that increased efforts by practitioners in areas of disadvantage lead to the establishment of more NWs (Rosenbaum 1987). Third, our observations may reflect variation in the characteristics of those individuals who live in high-risk areas, for example better-off individuals in high-risk areas may be driving participation. To assess this possibility, we conducted additional analyses looking specifically at the most disadvantaged neighbourhoods (available on request). However, this did not reveal further area or household influences on the proliferation of NW. Whatever the explanation, our finding that citizens will respond where objective risks are high in areas where conditions are unfavourable is an important one. It has implications for how we think about the concept of co-production and more specifically how we think about citizen participation in crime control at the local level. Our observations call into question the oft-held view that neighbourhood disadvantage undermines co-production. Suggesting instead that citizens will co-produce in unfavourable circumstances—but the conditions have to be right. Footnotes 1 Additional analyses (available on request) suggest a generally stable picture since 2010–11. 2 Despite asking about the presence of an NW scheme in ‘this area’, respondents were not required to make reference to a specific geographical boundary when answering. Respondents may therefore be reporting on different spatial geographies, leading to response variation within each area. As a result, this is treated as a household-level measure in the analysis. 3 The apparent fall in 2004–05 is likely to reflect a change in question wording. Prior to 2004–05, the survey asked whether an NW scheme had ever been in operation in the area. In 2004–05, this was changed to ‘in the last year’. 4 To measure signs of disorder, we used factor analysis to combine the scores from three interviewer ratings of the immediate area around each household covering (factor scores in parentheses): levels of litter/rubbish (0.78), vandalism/graffiti (0.83) and housing condition (0.84). 5 The CSEW selects one individual from each household to complete the survey; therefore, it is not possible to further unpick the relative influence of households and individuals. We therefore assume that these individual measures provide a reasonable approximation for the shared views and experiences of the household. 6 Boundary changes precluded us from incorporating data from the 2011 census. 7 Figure 1 also demonstrated that 15% of people did not know if a scheme was in operation in the area. People were more likely to be uncertain about the existence of NW if they had recently arrived in the area, had not been a victim of crime, did not own their home and lived in a more urban area (see Appendix Table A4). Appendix Table A1 Reasons for not joining Neighbourhood Watch (2012–13) . Frequency . % . No one has asked us to join 425 27 No particular reason 383 24 Too busy/not enough time 213 14 Haven’t got around to it yet/never thought about it 193 12 Another reason 162 10 Don’t know how to join 143 9 Don’t know enough about the scheme 119 8 Not interested 95 6 Not much crime in the local area 70 4 Don’t think they are effective 21 1 Don’t know 5 0 Sample size 1,564 . Frequency . % . No one has asked us to join 425 27 No particular reason 383 24 Too busy/not enough time 213 14 Haven’t got around to it yet/never thought about it 193 12 Another reason 162 10 Don’t know how to join 143 9 Don’t know enough about the scheme 119 8 Not interested 95 6 Not much crime in the local area 70 4 Don’t think they are effective 21 1 Don’t know 5 0 Sample size 1,564 Open in new tab Table A1 Reasons for not joining Neighbourhood Watch (2012–13) . Frequency . % . No one has asked us to join 425 27 No particular reason 383 24 Too busy/not enough time 213 14 Haven’t got around to it yet/never thought about it 193 12 Another reason 162 10 Don’t know how to join 143 9 Don’t know enough about the scheme 119 8 Not interested 95 6 Not much crime in the local area 70 4 Don’t think they are effective 21 1 Don’t know 5 0 Sample size 1,564 . Frequency . % . No one has asked us to join 425 27 No particular reason 383 24 Too busy/not enough time 213 14 Haven’t got around to it yet/never thought about it 193 12 Another reason 162 10 Don’t know how to join 143 9 Don’t know enough about the scheme 119 8 Not interested 95 6 Not much crime in the local area 70 4 Don’t think they are effective 21 1 Don’t know 5 0 Sample size 1,564 Open in new tab Table A2 Descriptive statistics . Mean . SD . Dependent variables Neighbourhood Watch area 0.35 0.48 House in Neighbourhood Watch 0.43 0.50 Would join Neighbourhood Watch 0.75 0.44 Household characteristics Female (HRP) 0.40 0.49 Age (HRP) 0.21 17.25 Number of children in HH 0.46 0.88 Years in area 4.54 1.79 Cohabiting HH 0.50 0.50 Non-white (HRP) 0.08 0.27 Accommodation type (ref.: detached) Semi-detached 0.32 0.47 Terraced 0.27 0.45 Flat/other 0.14 0.34 Socio-economic status of HRP (ref.: managerial/professional) Intermediate occupations 0.09 0.29 Small employers and own account workers 0.11 0.32 Lower supervisory and technical occupations 0.12 0.32 Semi-routine and routine occupations 0.27 0.44 Never worked and long-term unemployed 0.03 0.17 Tenure (ref.: owner-occupier) Social rented sector 0.17 0.37 Private rented sector 0.16 0.37 Victim of crime 0.23 0.42 Worried about crime overall 0.30 0.46 Local police do a good job 0.59 0.49 Has a visible burglar alarm 0.26 0.44 Home unoccupied during week 0.90 0.30 Household disorder 0.00 0.91 Area characteristics Concentrated disadvantage −0.04 0.96 Urbanicity −0.17 0.98 Population mobility −0.03 0.92 Age profile −0.04 1.03 Housing structure 0.04 0.90 Crime rate −0.09 0.71 . Mean . SD . Dependent variables Neighbourhood Watch area 0.35 0.48 House in Neighbourhood Watch 0.43 0.50 Would join Neighbourhood Watch 0.75 0.44 Household characteristics Female (HRP) 0.40 0.49 Age (HRP) 0.21 17.25 Number of children in HH 0.46 0.88 Years in area 4.54 1.79 Cohabiting HH 0.50 0.50 Non-white (HRP) 0.08 0.27 Accommodation type (ref.: detached) Semi-detached 0.32 0.47 Terraced 0.27 0.45 Flat/other 0.14 0.34 Socio-economic status of HRP (ref.: managerial/professional) Intermediate occupations 0.09 0.29 Small employers and own account workers 0.11 0.32 Lower supervisory and technical occupations 0.12 0.32 Semi-routine and routine occupations 0.27 0.44 Never worked and long-term unemployed 0.03 0.17 Tenure (ref.: owner-occupier) Social rented sector 0.17 0.37 Private rented sector 0.16 0.37 Victim of crime 0.23 0.42 Worried about crime overall 0.30 0.46 Local police do a good job 0.59 0.49 Has a visible burglar alarm 0.26 0.44 Home unoccupied during week 0.90 0.30 Household disorder 0.00 0.91 Area characteristics Concentrated disadvantage −0.04 0.96 Urbanicity −0.17 0.98 Population mobility −0.03 0.92 Age profile −0.04 1.03 Housing structure 0.04 0.90 Crime rate −0.09 0.71 Open in new tab Table A2 Descriptive statistics . Mean . SD . Dependent variables Neighbourhood Watch area 0.35 0.48 House in Neighbourhood Watch 0.43 0.50 Would join Neighbourhood Watch 0.75 0.44 Household characteristics Female (HRP) 0.40 0.49 Age (HRP) 0.21 17.25 Number of children in HH 0.46 0.88 Years in area 4.54 1.79 Cohabiting HH 0.50 0.50 Non-white (HRP) 0.08 0.27 Accommodation type (ref.: detached) Semi-detached 0.32 0.47 Terraced 0.27 0.45 Flat/other 0.14 0.34 Socio-economic status of HRP (ref.: managerial/professional) Intermediate occupations 0.09 0.29 Small employers and own account workers 0.11 0.32 Lower supervisory and technical occupations 0.12 0.32 Semi-routine and routine occupations 0.27 0.44 Never worked and long-term unemployed 0.03 0.17 Tenure (ref.: owner-occupier) Social rented sector 0.17 0.37 Private rented sector 0.16 0.37 Victim of crime 0.23 0.42 Worried about crime overall 0.30 0.46 Local police do a good job 0.59 0.49 Has a visible burglar alarm 0.26 0.44 Home unoccupied during week 0.90 0.30 Household disorder 0.00 0.91 Area characteristics Concentrated disadvantage −0.04 0.96 Urbanicity −0.17 0.98 Population mobility −0.03 0.92 Age profile −0.04 1.03 Housing structure 0.04 0.90 Crime rate −0.09 0.71 . Mean . SD . Dependent variables Neighbourhood Watch area 0.35 0.48 House in Neighbourhood Watch 0.43 0.50 Would join Neighbourhood Watch 0.75 0.44 Household characteristics Female (HRP) 0.40 0.49 Age (HRP) 0.21 17.25 Number of children in HH 0.46 0.88 Years in area 4.54 1.79 Cohabiting HH 0.50 0.50 Non-white (HRP) 0.08 0.27 Accommodation type (ref.: detached) Semi-detached 0.32 0.47 Terraced 0.27 0.45 Flat/other 0.14 0.34 Socio-economic status of HRP (ref.: managerial/professional) Intermediate occupations 0.09 0.29 Small employers and own account workers 0.11 0.32 Lower supervisory and technical occupations 0.12 0.32 Semi-routine and routine occupations 0.27 0.44 Never worked and long-term unemployed 0.03 0.17 Tenure (ref.: owner-occupier) Social rented sector 0.17 0.37 Private rented sector 0.16 0.37 Victim of crime 0.23 0.42 Worried about crime overall 0.30 0.46 Local police do a good job 0.59 0.49 Has a visible burglar alarm 0.26 0.44 Home unoccupied during week 0.90 0.30 Household disorder 0.00 0.91 Area characteristics Concentrated disadvantage −0.04 0.96 Urbanicity −0.17 0.98 Population mobility −0.03 0.92 Age profile −0.04 1.03 Housing structure 0.04 0.90 Crime rate −0.09 0.71 Open in new tab Table A3 Rotated component loadings from factorial ecology Neighbourhood measure . Concentrated disadvantage . Urbanicity . Population mobility . Age profile . Housing profile . Working population on income support 0.89 0.25 0.19 0.14 0.09 Lone parent families 0.85 0.22 0.00 0.26 0.15 Local authority housing 0.85 0.06 −0.01 0.15 −0.17 Working population unemployed 0.84 0.29 0.17 0.12 0.13 Non-car owning households 0.80 0.42 0.36 −0.01 0.06 Working in professional/managerial role −0.79 0.00 0.15 0.15 −0.37 Owner occupied housing −0.61 −0.25 −0.35 −0.57 0.05 Domestic property 0.10 0.92 0.17 0.05 0.11 Green space −0.21 −0.90 −0.18 −0.01 −0.04 Population density (per km2) 0.25 0.82 0.26 0.15 −0.14 Working in agriculture −0.13 −0.66 −0.01 −0.18 −0.03 In-migration −0.07 0.10 0.92 0.07 0.07 Out-migration −0.02 0.16 0.90 0.12 0.13 Single-person, non-pensioner households 0.36 0.36 0.74 0.13 −0.09 Commercial property 0.38 0.43 0.53 0.02 −0.09 More than 1.5 people per room 0.43 0.47 0.51 0.20 −0.33 Resident population over 65 −0.05 −0.21 −0.27 −0.89 −0.02 Resident population under 16 0.43 0.04 −0.46 0.64 0.19 Terraced housing 0.32 0.26 0.10 0.27 0.69 Vacant property 0.32 −0.12 0.49 −0.17 0.53 Flats 0.45 0.36 0.49 0.01 −0.52 Eigen value 9.3 3.3 1.9 1.4 1.3 Neighbourhood measure . Concentrated disadvantage . Urbanicity . Population mobility . Age profile . Housing profile . Working population on income support 0.89 0.25 0.19 0.14 0.09 Lone parent families 0.85 0.22 0.00 0.26 0.15 Local authority housing 0.85 0.06 −0.01 0.15 −0.17 Working population unemployed 0.84 0.29 0.17 0.12 0.13 Non-car owning households 0.80 0.42 0.36 −0.01 0.06 Working in professional/managerial role −0.79 0.00 0.15 0.15 −0.37 Owner occupied housing −0.61 −0.25 −0.35 −0.57 0.05 Domestic property 0.10 0.92 0.17 0.05 0.11 Green space −0.21 −0.90 −0.18 −0.01 −0.04 Population density (per km2) 0.25 0.82 0.26 0.15 −0.14 Working in agriculture −0.13 −0.66 −0.01 −0.18 −0.03 In-migration −0.07 0.10 0.92 0.07 0.07 Out-migration −0.02 0.16 0.90 0.12 0.13 Single-person, non-pensioner households 0.36 0.36 0.74 0.13 −0.09 Commercial property 0.38 0.43 0.53 0.02 −0.09 More than 1.5 people per room 0.43 0.47 0.51 0.20 −0.33 Resident population over 65 −0.05 −0.21 −0.27 −0.89 −0.02 Resident population under 16 0.43 0.04 −0.46 0.64 0.19 Terraced housing 0.32 0.26 0.10 0.27 0.69 Vacant property 0.32 −0.12 0.49 −0.17 0.53 Flats 0.45 0.36 0.49 0.01 −0.52 Eigen value 9.3 3.3 1.9 1.4 1.3 Open in new tab Table A3 Rotated component loadings from factorial ecology Neighbourhood measure . Concentrated disadvantage . Urbanicity . Population mobility . Age profile . Housing profile . Working population on income support 0.89 0.25 0.19 0.14 0.09 Lone parent families 0.85 0.22 0.00 0.26 0.15 Local authority housing 0.85 0.06 −0.01 0.15 −0.17 Working population unemployed 0.84 0.29 0.17 0.12 0.13 Non-car owning households 0.80 0.42 0.36 −0.01 0.06 Working in professional/managerial role −0.79 0.00 0.15 0.15 −0.37 Owner occupied housing −0.61 −0.25 −0.35 −0.57 0.05 Domestic property 0.10 0.92 0.17 0.05 0.11 Green space −0.21 −0.90 −0.18 −0.01 −0.04 Population density (per km2) 0.25 0.82 0.26 0.15 −0.14 Working in agriculture −0.13 −0.66 −0.01 −0.18 −0.03 In-migration −0.07 0.10 0.92 0.07 0.07 Out-migration −0.02 0.16 0.90 0.12 0.13 Single-person, non-pensioner households 0.36 0.36 0.74 0.13 −0.09 Commercial property 0.38 0.43 0.53 0.02 −0.09 More than 1.5 people per room 0.43 0.47 0.51 0.20 −0.33 Resident population over 65 −0.05 −0.21 −0.27 −0.89 −0.02 Resident population under 16 0.43 0.04 −0.46 0.64 0.19 Terraced housing 0.32 0.26 0.10 0.27 0.69 Vacant property 0.32 −0.12 0.49 −0.17 0.53 Flats 0.45 0.36 0.49 0.01 −0.52 Eigen value 9.3 3.3 1.9 1.4 1.3 Neighbourhood measure . Concentrated disadvantage . Urbanicity . Population mobility . Age profile . Housing profile . Working population on income support 0.89 0.25 0.19 0.14 0.09 Lone parent families 0.85 0.22 0.00 0.26 0.15 Local authority housing 0.85 0.06 −0.01 0.15 −0.17 Working population unemployed 0.84 0.29 0.17 0.12 0.13 Non-car owning households 0.80 0.42 0.36 −0.01 0.06 Working in professional/managerial role −0.79 0.00 0.15 0.15 −0.37 Owner occupied housing −0.61 −0.25 −0.35 −0.57 0.05 Domestic property 0.10 0.92 0.17 0.05 0.11 Green space −0.21 −0.90 −0.18 −0.01 −0.04 Population density (per km2) 0.25 0.82 0.26 0.15 −0.14 Working in agriculture −0.13 −0.66 −0.01 −0.18 −0.03 In-migration −0.07 0.10 0.92 0.07 0.07 Out-migration −0.02 0.16 0.90 0.12 0.13 Single-person, non-pensioner households 0.36 0.36 0.74 0.13 −0.09 Commercial property 0.38 0.43 0.53 0.02 −0.09 More than 1.5 people per room 0.43 0.47 0.51 0.20 −0.33 Resident population over 65 −0.05 −0.21 −0.27 −0.89 −0.02 Resident population under 16 0.43 0.04 −0.46 0.64 0.19 Terraced housing 0.32 0.26 0.10 0.27 0.69 Vacant property 0.32 −0.12 0.49 −0.17 0.53 Flats 0.45 0.36 0.49 0.01 −0.52 Eigen value 9.3 3.3 1.9 1.4 1.3 Open in new tab Table A4 Do not know whether area is part of Neighbourhood Watch . Odds ratio . SE . Household characteristics Female (HRP) 1.01 0.10 Age (HRP) 0.98** 0.00 Number of children in HH 0.92 0.05 Years in area 0.82** 0.02 Cohabiting HH 0.92 0.10 Non-white (HRP) 1.38* 0.22 Accommodation type (ref.: detached) Semi-detached 1.12 0.15 Terraced 1.15 0.17 Flat/other 1.09 0.20 Socio-economic status of HRP (ref: Managerial/professional) Intermediate occupations 0.89 0.15 Small employers and own account workers 0.90 0.15 Lower supervisory and technical occupations 1.11 0.17 Semi-routine and routine occupations 1.34* 0.16 Never worked and long-term unemployed 0.93 0.25 Tenure (ref.: owner-occupier) Social rented sector 1.28 0.18 Private rented sector 1.60** 0.20 Victim of crime 0.83 0.09 Worried about crime overall 0.89 0.09 Local police do a good job 0.89 0.08 Odds ratio SE Has a visible burglar alarm 1.00 0.11 Home unoccupied during week 0.86 0.13 Household disorder 1.05 0.06 Area characteristics Concentrated disadvantage 1.08 0.08 Crime rate 0.93 0.10 Crime rate × disadvantage 1.02 0.07 Urbanicity 1.08 0.06 Population mobility 1.05 0.06 Age profile 1.04 0.05 Housing structure 0.90* 0.05 Constant 0.34 0.09 Random effects Area 0.39 0.18 Approximate ICC 0.11 Areas/individuals 2,430/4,958 . Odds ratio . SE . Household characteristics Female (HRP) 1.01 0.10 Age (HRP) 0.98** 0.00 Number of children in HH 0.92 0.05 Years in area 0.82** 0.02 Cohabiting HH 0.92 0.10 Non-white (HRP) 1.38* 0.22 Accommodation type (ref.: detached) Semi-detached 1.12 0.15 Terraced 1.15 0.17 Flat/other 1.09 0.20 Socio-economic status of HRP (ref: Managerial/professional) Intermediate occupations 0.89 0.15 Small employers and own account workers 0.90 0.15 Lower supervisory and technical occupations 1.11 0.17 Semi-routine and routine occupations 1.34* 0.16 Never worked and long-term unemployed 0.93 0.25 Tenure (ref.: owner-occupier) Social rented sector 1.28 0.18 Private rented sector 1.60** 0.20 Victim of crime 0.83 0.09 Worried about crime overall 0.89 0.09 Local police do a good job 0.89 0.08 Odds ratio SE Has a visible burglar alarm 1.00 0.11 Home unoccupied during week 0.86 0.13 Household disorder 1.05 0.06 Area characteristics Concentrated disadvantage 1.08 0.08 Crime rate 0.93 0.10 Crime rate × disadvantage 1.02 0.07 Urbanicity 1.08 0.06 Population mobility 1.05 0.06 Age profile 1.04 0.05 Housing structure 0.90* 0.05 Constant 0.34 0.09 Random effects Area 0.39 0.18 Approximate ICC 0.11 Areas/individuals 2,430/4,958 *p < 0.05, **p < 0.01. Open in new tab Table A4 Do not know whether area is part of Neighbourhood Watch . Odds ratio . SE . Household characteristics Female (HRP) 1.01 0.10 Age (HRP) 0.98** 0.00 Number of children in HH 0.92 0.05 Years in area 0.82** 0.02 Cohabiting HH 0.92 0.10 Non-white (HRP) 1.38* 0.22 Accommodation type (ref.: detached) Semi-detached 1.12 0.15 Terraced 1.15 0.17 Flat/other 1.09 0.20 Socio-economic status of HRP (ref: Managerial/professional) Intermediate occupations 0.89 0.15 Small employers and own account workers 0.90 0.15 Lower supervisory and technical occupations 1.11 0.17 Semi-routine and routine occupations 1.34* 0.16 Never worked and long-term unemployed 0.93 0.25 Tenure (ref.: owner-occupier) Social rented sector 1.28 0.18 Private rented sector 1.60** 0.20 Victim of crime 0.83 0.09 Worried about crime overall 0.89 0.09 Local police do a good job 0.89 0.08 Odds ratio SE Has a visible burglar alarm 1.00 0.11 Home unoccupied during week 0.86 0.13 Household disorder 1.05 0.06 Area characteristics Concentrated disadvantage 1.08 0.08 Crime rate 0.93 0.10 Crime rate × disadvantage 1.02 0.07 Urbanicity 1.08 0.06 Population mobility 1.05 0.06 Age profile 1.04 0.05 Housing structure 0.90* 0.05 Constant 0.34 0.09 Random effects Area 0.39 0.18 Approximate ICC 0.11 Areas/individuals 2,430/4,958 . Odds ratio . SE . Household characteristics Female (HRP) 1.01 0.10 Age (HRP) 0.98** 0.00 Number of children in HH 0.92 0.05 Years in area 0.82** 0.02 Cohabiting HH 0.92 0.10 Non-white (HRP) 1.38* 0.22 Accommodation type (ref.: detached) Semi-detached 1.12 0.15 Terraced 1.15 0.17 Flat/other 1.09 0.20 Socio-economic status of HRP (ref: Managerial/professional) Intermediate occupations 0.89 0.15 Small employers and own account workers 0.90 0.15 Lower supervisory and technical occupations 1.11 0.17 Semi-routine and routine occupations 1.34* 0.16 Never worked and long-term unemployed 0.93 0.25 Tenure (ref.: owner-occupier) Social rented sector 1.28 0.18 Private rented sector 1.60** 0.20 Victim of crime 0.83 0.09 Worried about crime overall 0.89 0.09 Local police do a good job 0.89 0.08 Odds ratio SE Has a visible burglar alarm 1.00 0.11 Home unoccupied during week 0.86 0.13 Household disorder 1.05 0.06 Area characteristics Concentrated disadvantage 1.08 0.08 Crime rate 0.93 0.10 Crime rate × disadvantage 1.02 0.07 Urbanicity 1.08 0.06 Population mobility 1.05 0.06 Age profile 1.04 0.05 Housing structure 0.90* 0.05 Constant 0.34 0.09 Random effects Area 0.39 0.18 Approximate ICC 0.11 Areas/individuals 2,430/4,958 *p < 0.05, **p < 0.01. Open in new tab References Alford , J . ( 2009 ), Engaging Public Sector Clients: From Service Delivery to Co-production . Palgrave Macmillan . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Alford , J. and Yates , S. ( 2016 ), ‘ Co-production of Public Services in Australia: The Roles of Government Organisations and Co-producers ’, Australian Journal of Public Administration , 75: 159–75. Google Scholar OpenURL Placeholder Text WorldCat Bennett , T . ( 1989 ), ‘ Factors Related to Participation in Neighbourhood Watch Schemes ’, British Journal of Criminology , 29 : 207 – 18 . Google Scholar Crossref Search ADS WorldCat —— ( 1990 ), Evaluating Neighbourhood Watch . Gower . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Bennett , T. , Holloway , K. and Farrington , D. ( 2008 ), ‘ The Effectiveness of Neighbourhood Watch in Reducing Crime ’, Campbell Systematic Reviews , 18 , available online at http://www.campbellcollaboration.org/. Bovaird , T . ( 2007 ), ‘ Beyond Engagement and Participation – User and Community Co-production of Public Services ’, Public Administration Review , 67 : 846 – 60 . Google Scholar Crossref Search ADS WorldCat Brandsen , T. and Honingh , M. ( 2015 ), ‘ Distinguishing Different Types of Coproduction: A Conceptual Analysis Based on the Classical Definitions ’, Public Administration Review , 76 : 427 – 35 . Google Scholar Crossref Search ADS WorldCat Bullock , K . ( 2014 ), Citizens, Community and Crime Control . Palgrave Macmillan . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Cabinet Office ( 2010 ), The Coalition: Our Programme for Government . Cabinet Office , available online at https://www.gov.uk/government/publications/the-coalition-documentation. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Crawford , A . ( 1999 ), The Local Governance of Crime: Appeals to Community and Partnership . Oxford University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC DuBow , F. and Podolefsky , A. ( 1982 ), ‘ Citizen Participation in Community Crime Prevention ’, Human Organization , 41 : 307 – 14 . Google Scholar Crossref Search ADS WorldCat Frank , J. , Brandl , S., Worden , R. E. and Bynum , T. ( 1996 ), ‘ Citizen Involvement in the Coproduction of Police Outputs ’, Journal of Crime and Justice , 19 : 1 – 30 . Google Scholar Crossref Search ADS WorldCat Gelman , A. and Hill , J. ( 2007 ), Data Analysis Using Regression and Multilevel/Hierarchical Models . Cambridge: Cambridge University Press. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Goldstein , H . ( 2011 ), Multilevel Statistical Models , 4th edn. Wiley . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Gresham , P. , Grainger , E., Stockdale , J., Woodhead , D. C. and Shermer , D. ( 2004 ), ‘Urbecon’ Report to the Home Office, Crime Strategy and Resources Unit, on the Neighbourhood Watch Development Project . Urbecon Ltd , available online at http://www.nwinfo.co.uk/urbecon/urbecon.pdf (accessed 27 August 2012 ). Henig , J . ( 1984 ), Citizens against Crime: An Assessment of the Neighbourhood Watch Programme in Washington, D.C . Occasional Paper No. 2, Center for Washington Area Studies, George Washington University , available online at http://www.popcenter.org/library/scp/pdf/85-Henig.pdf. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC HM Treasury ( 2010 ), Spending Review 2010, CM7942 . HMSO , available online at https://www.gov.uk/government/publications/spending-review-2010 (accessed 7 March 2017 ). Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC HMIC ( 2017 ), PEEL: Police Effectiveness 2016 – A National Overview . HMIC , available online at http://www.justiceinspectorates.gov.uk/hmic/wp-content/uploads/peel-police-effectiveness-2016.pdf (accessed 5 May 2017 ). Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Home Office ( 2004 ), Building Communities, Beating Crime: A Better Police Service for the 21st Century , CM 6360. Home Office , available online at http://www.archive2.official-documents.co.uk/document/cm63/6360/6360.pdf. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC —— ( 2010 ), Policing in the 21st Century: Reconnecting Police and the People , CM 7942. Home Office , available online at: https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/118241/policing-21st-full-pdf.pdf (accessed 7 March 2017 ). Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Hope , T . ( 1988 ), ‘ Support for Neighbourhood Watch: A British Crime Survey Analysis ’, in T. Hope and M. Shaw, eds, Communities and Crime Reduction . H.M. Stationery Office . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC —— ( 1995 ), ‘ Community Crime Prevention ’, in M. Tonry and D. P. Farrington, eds, Building a Safe Society: Strategic Approaches to Crime Prevention , 21 – 90 . University of Chicago Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC —— ( 2000 ), ‘ Inequality and the Clubbing of Private Security ’, in T. Hope and R. Sparks, eds, Crime, Risk and Insecurity , 83 – 106 . Routledge . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Hourihan , K . ( 1987 ), ‘ Local Community Involvement and Participation in Neighbourhood Watch: A Case Study in Cork, Ireland ’, Urban Studies , 24 : 129 – 36 . Google Scholar Crossref Search ADS WorldCat Husain , S . ( 1988 ), Neighbourhood Watch in England and Wales: A Locational Analysis , Crime Prevention Unit Paper 12, Home Office . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Lavrakas , P. and Herz , E. ( 1982 ), ‘ Citizen Participation in Neighborhood Crime Prevention ’, Criminology , 20 : 479 – 98 . Google Scholar Crossref Search ADS WorldCat Laycock , G. and Tilley , N. ( 1995 ), Policing and Neighbourhood Watch: Strategic Issues . Police Research Group Crime Detection and Prevention Series Paper No. 60. Home Office , available online at http://www.popcenter.org/problems/credit_card_fraud/PDFs/Laycock_Tilley_1995.pdf. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Lister , S. and Rowe , M. ( 2015 ), ‘ Electing Police and Crime Commissioners in England and Wales: Prospecting for the Democratisation of Policing ’, Journal Policing and Society , 25 : 358 – 77 . Google Scholar Crossref Search ADS WorldCat Loeffler , E. and Bovaird , T. ( 2016 ), ‘ User and Community Co-production of Public Services: What Does the Evidence Tell Us ?’, International Journal of Public Administration , 39 : 1006 – 19 . Google Scholar OpenURL Placeholder Text WorldCat Lowndes , V. and Pratchett , L. ( 2011 ), ‘ Local Governance under the Coalition Government: Austerity, Localism and the “Big Society” ’, Local Government Studies , 38 : 21 – 40 . Google Scholar Crossref Search ADS WorldCat Mattson , G . ( 1986 ), ‘ The Promise of Citizen Coproduction: Some Persistent Issues ’, Public Productivity Review , 10 : 51 – 6 . Google Scholar Crossref Search ADS WorldCat McConville , M. and Shepherd , D. ( 1992 ), Watching Police, Watching Communities . Routledge . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Mukherjee , S. and Wilson , P. ( 1987 ), ‘ Neighbourhood Watch: Issues and Policy Implications ’, in Trends and Issues in Crime and Criminal Justice , No. 8. Australian Institute of Criminology , available online at http://www.popcenter.org/library/scp/pdf/131-Mukherjee_Wilson.pdf. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Nunnally , J . ( 1978 ), Psychometric Theory . McGraw-Hill . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Office for National Statistics (ONS) ( 2012 ), Crime Survey for England and Wales (CSEW) Dataset User Guide: Adults Aged 16 and Over 2011/12 . Crime Statistics and Analysis . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Osborne , S. P. , Radnor , A. and Strokosch , K. ( 2016 ), ‘ Co-production and the Co-creation of Value in Public Services: A Suitable Case for Treatment ?’, Public Management Review , 18 : 639 – 53 . Google Scholar Crossref Search ADS WorldCat Osborne , S. P. and Strokosch , K. ( 2013 ), ‘ It Takes Two to Tango? Understanding the Co-production of Public Services by Integrating the Services Management and Public Administration Perspectives ’, British Journal of Management , 24 : S31 – 47 . Google Scholar Crossref Search ADS WorldCat Ostrom , E . ( 1990 ), Governing the Commons: The Evolution of Institutions for Collective Action . Cambridge University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Ostrom , R . ( 1996 ), ‘ Crossing the Great Divide: Coproduction, Synergy, and Development ’, World Development , 24 : 1073 – 87 . Google Scholar Crossref Search ADS WorldCat Parrado , S. , Van Ryzin G., Bovaird , T. and Löffler , E. ( 2013 ), ‘ Correlates of Co-production: Evidence from a Five-Nation Survey of Citizens ’, International Public Management Journal , 16 : 85 – 112 . Google Scholar Crossref Search ADS WorldCat Pattavina , A. , Byrne , J. and Garcia , L. ( 2006 ), ‘ An Examination of Citizen Involvement in Crime Prevention in High-Risk versus Low- to Moderate-Risk Neighbourhoods ’, Crime and Delinquency , 52 : 203 – 31 . Google Scholar Crossref Search ADS WorldCat —— ( 2009 ), ‘ Towards a Paradigm of Democratic Participation: Citizen Participation and Co-participation and Co-production of Personal Social Services in Sweden ’, Annals of Public and Cooperative Economics , 80 : 197 – 224 . Crossref Search ADS WorldCat —— ( 2012 ), ‘ Co-production and Third Sector Social Services in Europe—Some Crucial Conceptual Issues ’, in V. Pestoff , T. Brandsen and B. Verschuere, eds, New Public Governance: The Third Sector and Co-production , Chapter 2. Routledge . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Pestoff , V. , Osborne , S. P. and Brandsen , T. ( 2006 ), ‘ Patterns of Co-production in Public Services ’, Public Management Review , 8 : 591 – 5 . Google Scholar Crossref Search ADS WorldCat Putnam , R . ( 2000 ), Bowling alone: The collapse and revival of community in America . New York : Simon & Schuster . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Quinton , P. and Morris , J. ( 2008 ), Neighbourhood Policing: The Impact of Piloting and Early National Implementation . Home Office Online Report 01/08. Home Office , available online at http://rds.homeoffice.gov.uk/rds/pdfs08/rdsolr0108.pdf. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rawlings , P . ( 2002 ), Policing: A Short History . Willan . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rees , P. H . ( 1971 ), ‘ Factorial Ecology: An Extended Definition, Survey and Critique of the Field ’, Economic Geography , 47 : 220 – 33 . Google Scholar Crossref Search ADS WorldCat Reisig , M . ( 2007 ), ‘ Procedural Justice and Community Policing—What Shapes Residents’ Willingness to Participate in Crime Prevention Programs ?’, Policing , 1 : 356 – 69 . Google Scholar Crossref Search ADS WorldCat Rosenbaum , D . ( 1987 ), ‘ The Theory and Research Behind Neighbouhood Watch: Is It a Sound Fear and Crime Reduction Strategy ’, Crime and Delinquency , 35 : 105 – 34 . Google Scholar OpenURL Placeholder Text WorldCat —— ( 1988 ), ‘ Community Crime Prevention: A Review and Synthesis of the Literature ’, Justice Quarterly , 5 : 323 – 95 . Crossref Search ADS WorldCat Scarman , L . ( 1981 ), The Brixton Disorders 10–12 April 1981 , Cmnd 8427. Her Majesty’s Stationary Office . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Shernock , S . ( 1986 ), ‘ A Profile of the Citizen Crime Prevention Activist ’, Journal of Criminal Justice , 14 : 211 – 28 . Google Scholar Crossref Search ADS WorldCat Skogan , W . ( 1988 ), ‘ Community Organisation and Crime ’, Crime and Justice , 10 : 39 – 78 . Google Scholar Crossref Search ADS WorldCat —— ( 1989 ), ‘ Communities, Crime, and Neighbourhood Organisation ’, Crime and Delinquency , 35 : 437 – 57 . Crossref Search ADS WorldCat —— ( 2008 ), ‘ An Overview of Community Policing: Origins, Concepts and Implementation ’, in T. Williamson, ed., The Handbook of Knowledge-Based Policing, 43–50 . Wiley . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Stoker , G . ( 2006 ), Why Politics Matters: Making Democracy Work . Palgrave Macmillan . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Tilley , N . ( 2008 ), ‘ Modern Approaches to Policing: Community, Problem-Oriented and Intelligence-Led Policing ’, in T. Newburn, ed., Handbook of Policing , 2nd edn, 95–116. Willan . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Van Eijk , C. and Steen , T. ( 2016 ), ‘ Why Engage in Co-production of Public Services? Mixing Theory and Empirical Evidence ’, International Review of Administrative Sciences , 82 : 28 – 46 . Google Scholar Crossref Search ADS WorldCat Van Eijk , C. , Steen , T. and Verschere , B. ( 2017 ), ‘ Co-producing Safety in the Local Community: A Q-Methodology Study on the Incentives of Belgian and Dutch Members of Neighbourhood Watch Schemes ’, Local Government Studies , 43 : 323 – 43 . Google Scholar Crossref Search ADS WorldCat Verschuere , B. , Brandsen , T. and Pestoff , V. ( 2012 ), ‘ Co-production: The State of the Art in Research and the Future Agenda ’, Voluntas , 23 : 1083 – 101 . Google Scholar Crossref Search ADS WorldCat Voorberg , W. , Bekkers V. and Tummers , L. ( 2015 ), ‘ A Systematic Review of Co-creation and Co-production: Embarking on the Social Innovation Journey ’, Public Management Review , 17 : 1333 – 57 . Google Scholar Crossref Search ADS WorldCat Webb , J . ( 1994 ), ‘ Neighbourhood Concern Groups: Daring to Care in the Inner City ’, unpublished thesis, The Safer Communities Project . Google Scholar Yarwood , R. and Edwards , B. ( 1995 ), ‘ Voluntary Action in Rural Areas: The Case of Neighbourhood Watch ’, Journal of Rural Studies , 11 : 447 – 59 . Google Scholar Crossref Search ADS WorldCat © The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. © The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD).

Molnar,, Adam;Whelan,, Chad;Boyle, Philip, J

doi: 10.1093/bjc/azy014pmid: N/A

Abstract Extending inquiries into the dynamics underpinning the ‘iterative’ development of security governance at mega-events, this article explores practices of knowledge sharing and policy transfer at major political summits. Through detailed interviews with police involved in the Toronto 2010 G20 and the Brisbane 2014 G20 summits, and through analysing supporting documentation, we examine the ways in which police interpret past events, as either ‘failures’ or ‘successes’, specifically in the context of public order policing. The article extends insights into how such perceptions are facilitated through transnational exchanges, particularly where event-related ‘failures’ might be considered as a benchmark for iterative policy developments. We explain this process as a form of ‘failure-inspired social learning’ that questions the effectiveness, norms and legitimacy of established policies, practices and institutions involved in security governance, which can influence future transformations in global ‘best practices’. Introduction A breadth of research on mega-event security governance has focused on practices of security and surveillance before, during and after mega-events (Fussey and Coaffee 2011; Fussey et al. 2011; Bennett and Haggerty 2011; Klauser 2013). Recognizing mega-events as spectacularly risk-averse projects, researchers have focused on the techniques and technologies used to manage perceived risks and the implications of such initiatives for urban environments and communities (Boyle and Haggerty 2009; Fussey et al. 2011; Giulianotti and Klauser 2011; Kitchen and Rygiel 2014). Within this body of research, however, very few have examined knowledge-sharing policy transfer between hosts of mega-events (Boyle 2011; Fussey et al. 2011; Klauser 2013). Such analyses have largely been framed through critical geographical studies examining how policy knowledges circulate across jurisdictions, within and through fields of power and politics (Stone 2004; Peck and Theodore 2010; McCann 2011). In addition, criminological studies have called attention to loosely coupled knowledge networks for furthering the wider objectives of institutions tasked with various dimensions of mega-event security (Johnston and Shearing 2003; Dupont 2004; Wood and Dupont 2006; Whelan 2012; Whelan and Dupont 2017). For instance, Boyle (2011) has examined the formal networks whereby event organizers—such as the International Olympic Committee (IOC), Commonwealth Games federation (GCF) and Federation Internationale de Football Association (FIFA)—record and facilitate knowledge sharing between host nations. Others have examined how the private sector is involved in globalized assemblages that promotes the private interests of security and technology markets within the domain of mega-events (Samatas 2007; Molnar and Snider 2011). Building on inquiries into the ‘iterative’ development of security governance at mega-events (Boyle et al. 2015) as well as knowledge sharing and policy transfer in the trans-nationalization of criminal justice policy (Jones and Newburn 2006; Jones and Newburn 2007; Vanstone 2008; McGuire and William 2013; Wood 2014), this article explores the practices of knowledge transfer and policy learning between actors responsible for public order policing at major political summits. A prominent theme in the policing literature is tracing ongoing strategic innovations (Gillham 2011; Monaghan and Walby 2012) and tactical developments in public order policing models such as with regard to what have been referred to as ‘strategic incapacitation’ (Gillham 2011; Gillham, Edwards and Noakes 2013) and ‘intelligent control’ (de Lint and Hunt 2009). Speaking generally, such models refer to the blending of ‘hard’ tactics—involving pre-event surveillance and intelligence operations, pre-emptive arrest and broad spatial contaminant strategies—and ‘soft’ tactics such as negotiated management (Whelan and Molnar 2017a). Other policing scholars have begun to sketch out the ways in which public order policing strategies circulate across national borders (King and Waddington 2006; della Porta and Tarrow 2012; Wood 2014), highlighting the importance of inter-organisational relationships in shaping public order policing policies and practices. Situated in the latter area of inquiry, the article draws on extensive interviews with key individuals from the Queensland Police Service (QPS) and the Toronto Police Service (TPS) to show how police responsible for the Brisbane 2014 G20 Leaders’ Summit evaluated the experience of the 2010 Toronto G20 and, according to them, adjusted their approach in light of that knowledge. We focus, in particular, on the role of operational failure, or perceived failures, in prompting reflexive innovations in police organizations. Examining when and how organisations understand failure and the conditions under which organizations learn (or do not) from their own failures or the failures of others has a long history in public policy and administration literature pertaining to topics such as natural disasters, aviation disasters, oil spills, nuclear energy policy and ‘homeland security’ (e.g. Birkland 1997; 2006; Donahue and Tuohy 2006). However, few aside from Jones and Newburn (2007) have extended analyses of organizational learning to the policing literature. Following their suggestion of expanding scholarly analyses of organizational learning and knowledge transfer to wider aspects of criminological inquiry (Jones and Newburn 2007: 163–164), our examination focuses on how the Toronto G20 constituted a negative lesson (Stone 2004) for QPS that stimulated changes in how the organisation planned for Brisbane 2014. We thus eschew imposing our own interpretation of ‘what went wrong’ in Toronto (see for example Beare et al. 2015; Baker et al. 2017) in favour of seeking to understand how QPS identified and reflexively adapted to the perceived failures that underpinned the widespread property damage, public disorder and mass arrests occurring at Toronto 2010. Following what Hall (1993) refers to as a process of ‘social learning’—that is, ‘a deliberate attempt to adjust the goals or techniques of policy in response to past experience and new information’ (Hall 1993: 278)—we suggest that a process of failure-inspired social learning constitutes an important pathway through which policies and practices of public order policing are refined amongst transnational police communities. The article is organised in four key sections. The first advances a conceptual approach to elaborate on how a certain event comes to be interpreted as a ‘focusing event’ in the eyes of policing organizations. We round off this section with a more detailed consideration of the literature on ‘social learning’ (Hall 1993), ‘epistemic communities’ (Haas 1992) and knowledge transfer before looking further at the features of failure-inspired learning. The second section provides an overview of the Toronto 2010 G20, highlighting how that event provided a backdrop for ‘failure-inspired learning’ for the Brisbane 2014 G20. The third outlines our methods and data. The fourth section concentrates on our findings, specifically in terms of the lessons QPS drew from Toronto 2010. We focus, in particular, on the problems encountered in Toronto with command and control and the adaptations made by QPS in an effort to avoid similar problems in Brisbane. The final section concludes the article and raises some supplementary observations regarding failure-inspired social learning at mega-events. Failure-Inspired Social Learning Focusing events can provide an opportunity for social learning. A focusing event refers to how ‘disruptive’ events that occur in a given policy domain come to be interpreted as normative failures after the event has ended (Sabatier and Weible 2007: 202). Such failures are regarded as an opportunity to revise established approaches in order to learn, adapt and avoid similar outcomes for other actors (Alink et al. 2001; Boin et al. 2005; Norhstedt and Weible 2010). Focusing events are thus crucial for shaping future policy directions (Sabatier and Weible 2007: 202), especially as they may lead organisations to modify their policies through a process of social learning. Hall’s (1993) theory of social learning has been found to provide important insights into learning through focusing events (e.g. Birkland 1998). For Hall, social learning can encompass what are referred to as first-, second- and third-order policy change. First-order change emerges from dissatisfaction with past policy decisions or implementations, where revisions in tactics occur but the policy objectives remain the same. Although the instruments and objectives remain consistent in first-order change, incremental adjustments occur in the ‘precise settings’ of those instruments (1993: 278). Second-order policy change includes the recalibration and recombination of existing and new policy instruments with no change to the overarching objectives. This form of change is in part attributable to significant adjustments in organizational norms that affect strategic orientations and policy instruments. Third-order change, on the other hand, consists of changes not only to the goals, instruments and settings of inherited policy approaches but also invokes deep-seated shifts in the entire interpretative Gestalt of a policy framework. Hall likens this process to a Kuhnian paradigm shift in which one policy paradigm displaces another in a more radical ‘disjunctive process associated with periodic discontinuities in policy’ (1993: 279). Though Hall’s (1993) analyses were couched in a single political context, his insights can be extended to examine how police expertise circulates and is refined in response to the experiences of counterparts elsewhere. These processes of social learning are facilitated by loosely structured transnational knowledge networks (Boyle 2011). In its broadest interpretation, the concept of ‘knowledge transfer’ encompasses ‘the processes by which knowledge about policies, administrative arrangements, institutions and ideas in one political system (past and present) is used in the development of policies, administrative arrangements, institutions and ideas in another political system’ (Dolowitz and Marsh 2000: 5). Such processes can range from either voluntary adoption or the coercive imposition of programs or policies that appear to work elsewhere, though most will arise from a mix of the two. All policy processes will also be conditioned by a process of ‘becoming’ (Bennett 1991: 219) that is further attributable to a particular ‘structural and cultural context’ (Jones and Newburn 2007: 27). Haas (1992) identifies this structural and cultural context as ‘epistemic communities’ (Haas 1992: 2), which establish important normative and ideational parameters for ‘weighing and validating’ the merit of potential policy actions and desired outcomes. This is especially the case in moments of uncertainty brought about through focusing events. In other words, epistemic communities enact a sort of ‘bounded rationality’ that can circumscribe the range of policy alternatives under evaluation through knowledge transfer. Although Haas (1992) has pointed to epistemic communities as specialist groups that are ‘external’ to government processes, in the context of public order policing, Wood (2014) regards these processes as being conditioned more narrowly amongst police organizations across national jurisdictions. Wood (2014) considers the degree of integration and the extent that a shared professional identity exists among any particular policing community, suggesting that the greater the shared identity and beliefs, the greater the likelihood of knowledge transfer. The public order policing literature suggests that experiences at previous events and developments in emerging technologies can influence how policing and security is pursued at future events. Wood (2014: 20), again for example, notes how ‘the Seattle protests, the Rodney King riots in Los Angeles, the 9/11 attacks in New York City, and the G20 protests in London and Toronto have each stimulated tactical innovation, diffusion, and shifts in practice’. More broadly, della Porta and Tarrow (2012: 127) have highlighted how changes in public order policing are often augured through a process of inter-agency assessment, where ‘the analysis of information on past events and their definition as successes or failures…leads to adaptation of the innovation to new sites and situations’. Likewise, Waddington’s (2007: 130) comprehensive review of changes in public order policing over the last two decades underscores the importance of lessons learned from ‘previous encounters or the vicarious experience of confrontations involving other forces’. However, much of the public order policing literature remains preoccupied with a technocratic focus on tactics (see Gillham 2011) at the expense of considering the wider organizational practices that inform ‘cultures of control’ (Garland 2001). Although Wood and others have recognized that tactical innovations take place in response to experiences elsewhere, researchers are yet to fully appreciate how police organizations themselves make sense of such failures. The lack of explicit inquiry into the internal characteristics of these processes means researchers are left making deceptive inferences that some mechanism of knowledge transfer is at work based on externally observable traits (Bennett 1991). We suggest that Hall’s theory of policy change through social learning, which is based on an analysis of British macroeconomic policy making in the 1970s and 1980s, is a useful framework for considering how perceived failures can inform future philosophies and practices in public order policing. Our analysis is, thus, motivated by the view that the literature must attend directly to both the conceptual parameters of failure-inspired social learning as well as the empirical aspects of knowledge transfer. The former provides the explanation of how event-related failures influence the types of change that occur (first-, second-, or third-order). The latter provides an account of how relationships amongst law enforcement agencies, or what Haas (1992) would term ‘epistemic communities’, facilitate these changes. We do this by examining the processes of failure-inspired social learning between the TPS and QPS in relation to their role in policing G20 events in Toronto and Brisbane. From Toronto 2010 to Brisbane 2014 The Toronto 2010 G20 was an exceptional event by Canadian standards. It is the most expensive domestic security event ever held, costing approximately 1CAD billion in public funds exclusively for security, trumping the Vancouver 2010 Olympic security operation that took place a mere 5 months earlier (Kitchen and Rygiel 2014). The scale of the security operation involved over 6000 law enforcement personnel across a range of police and security agencies, including the TPS, Royal Canadian Mounted Police (RCMP) and Ontario Provincial Police (OPP), and officers from municipal police agencies across Canada. In a similar way to most mega-events (Fussey et al. 2011; Bennett and Haggerty 2011; King and Waddington 2013; Whelan 2014; Fussey 2015), the scale of G20 security also translated into an extensive convergence of crime control and counter-terrorism strategies (Kitchen and Rygiel 2014). For example, the event involved a significant public order operation that relied heavily on intelligence-gathering and surveillance (Monaghan and Walby 2012), the militarization of crowd control and the associated implementation of strategies and tactics of spatial control, including the controversial practice of ‘kettling’ (Fernandez 2008; Wood 2015). The adoption of these tactics was buoyed by the passage of local ordinances to regulate populations through access control and a lowered threshold for conducting ‘stop-and-search’ practices of citizens (Milberry and Clement 2015; Roach 2015). Ultimately, the event was marked by mass public protests that reportedly involved up to 10,000 protesters and called into action up to 20,000 police and military personnel. Claims of excessive use of force by the TPS and Integrated Security Unit followed, as did several convictions of police in criminal courts (Hasham 2013; CBC News 2015). By the end of the 2-day summit over 1,000 people were arrested and detained for up to 24 hours in a makeshift holding centre built specifically for the G20, leading commentators to describe the Toronto G20 as one of the most significant contraventions of civil liberties in Canadian history (Benzie and Ferguson 2010; OIPRD 2011; Beare et al. 2015). The fallout from Toronto continued for several years after the event, resulting in ongoing claims of excessive use of force and a number of associated government investigations (OIPRD 2011; House of Commons Canada 2011). In total, a class-action lawsuit was launched, over 30 police faced disciplinary hearings for misconduct and three police were convicted of criminal charges involving the misuse of force (Hasham 2013; CBC News 2015). In early 2012, when Australia was requested to host the 2014 G20, the media were replaying events from Toronto so frequently that it seemed many Australian cities did not want the event. In a conversation between then Toronto mayor David Miller and Melbourne’s then Lord Mayor Robert Doyle, Miller reportedly insisted: ‘For God’s sake don’t get this conference;” “this is a conference you do not want’ (Moore 2012). After Melbourne’s withdrawal in July 2012, Brisbane was elected over Sydney to host the event. From that moment, over 2 years prior to the hosting of the event, Brisbane commenced planning the security operation. In the lead up to the event, considerable media attention and public concern was captured in headlines, such as ‘G20 anarchists vow chaos and mayhem for Brisbane’s streets’ (Ironside and Chamberlin 2013) and ‘Welcome to Brisbane, ferals: special forces troops show muscle ahead of the G20 summit’ (Courier Mail 2014). The media displayed footage of police vehicles being set on fire and people being subject to considerable force by Toronto Police, leading many to conclude that a similar outcome would happen in Brisbane. Amidst ongoing media concerns about violent protests in Australia, and as the Toronto Police dealt with significant public scrutiny after the close of the G20 including a class-action lawsuit in response to the use of kittling (Perkel 2016), the QPS actively sought to learn from the Toronto G20 (QPS 2014; 2016). The 2014 G20 was held in inner city Brisbane and event organizers were under pressure to minimize any disruptions to residents and businesses. In line with measures introduced at previous political summits, the Brisbane G20 was precipitated by the introduction of the G20 (Safety and Security) Act (Queensland Government 2013), which was passed in October of that year and created new offences and powers for authorized officers in designated security areas. For example, this legislation included: specifying prohibited items in designated security areas; imposing requirements on those within such areas to undergo various searches and disclose personal details to police if required; restrictions on the movement of people and prohibited persons; and imposing requirements for those intending to organise an assembly or protest to consult with a QPS ‘liaison officer’ to discuss the location, time and date (among other details) of the proposed assembly. The Act was accompanied by the G20 (Safety and Security) Regulation (Queensland Government 2014), which established the designated security areas for each of the venues for the meetings as well as accommodation precincts for leaders’ and staff. There was also a G20 Finance Ministers and Central Bank Governors Meeting in Cairns—a small city in Northern Queensland—held on 20–21 September to which the same legislation applied. The security operation involved approximately 6,650 police, including 4,500 from QPS, 1,500 from other state and territory police organizations and New Zealand, and 650 members of the Australian Federal Police (AFP). It also involved many other agencies, including the Australian Defence Force, the Australian Customs and Border Protection Service (ACBPS)1 and the Australian Security Intelligence Organisation (ASIO) as well as private security services within the secure event zones. The total costs of the security operation were $AU450 million (Crime and Misconduct Commission 2015). Independent legal observers, organized by Brisbane’s Caxton Legal Centre, were appointed to monitor and record on camera any incident between the police and the public during the event. Data and Methods We base our analysis of these events on evidence gathered as part of a larger project following an exploratory case study design (Yin 2014) covering various aspects of the Brisbane 2014 G20. The project is based on detailed, qualitative interviews with key stakeholders involved in the G20 security operation across a number of agencies and the analysis of available security documentation. In the current article, we focus mainly on interviews with members of the G20 Group, a standalone division of QPS established in late 2012 to assume overall responsibility for planning and coordinating the G20 police operation. The G20 Group involved a familiar hierarchical structure, with an Assistant Commissioner as the Program Executive, a Chief Superintendent as Program Director and then three Superintendents responsible for the portfolios of Operational Planning, Program Control and Support Services. Project managers, who were normally at the level of Inspector, were appointed under the respective superintendents. Examples of project managers from the operational planning portfolio include: command and control; dignitary protection; intelligence; investigations and offender processing; public safety and specialist support; traffic; and venue security. The program control and support service portfolios included a further ten project managers. Our focus within these portfolios was limited to the areas of information technology integration, logistics, risk and quality assurance, and training. In the lead-up to Brisbane 2014, members of the G20 Group travelled extensively and hosted international peer-reviewers and observers from North America and the United Kingdom before and during the event. We have interviewed a number of international peer-reviewers, with access largely brokered by members of the QPS. The project involves 15 interviews with members of the QPS G20 Group, ten interviews with members of the G20 Taskforce—the agency established within the Department of Prime Minister and Cabinet to coordinate the event—and three interviews with international stakeholders from Canada, New Zealand and the United Kingdom. Other interviewees include members of the Australian Federal Police. In total, 30 interviews were conducted. Interviews were semi-structured, between 45 and 90 minutes long and were conducted in 2015–16. We draw mostly on interviewees that travelled to Toronto (of which there were three members) as well as a key official from Toronto that was involved in hosting and consulting with the QPS. Our primary focus is the internal workings of the security operation, particularly in relation to public order policing, and as such, we draw heavily on interview data. Although we recognize the limits of relying on interview data, including the potential for there to be a significant difference between what respondents may say in an interview and how they actually think and act, in reality interviews are the only feasible methodology for addressing the internal reflections of policing operations as much of the finer details are not captured in documents such as post-event reviews. The interview schedule open-ended questions, inviting interviewees to share their experiences of being involved in Brisbane 2014. For example, interviewees were asked to tell us about their role, the challenges they faced in carrying out their role and the main lessons they learnt for securing mega-events in the future. Interviews were transcribed and interviewees were given an opportunity to review and amend the interview transcript, principally to ensure that no security-sensitive information was inadvertently released, as a condition of the research agreements. Interview data were coded and analysed using NVivo 11. Coding concentrated on classifying emerging themes from the interviews. Data were cross-referenced across interviewees from each organization as well as against official documents (where possible). It should be noted that only personal views of interviewees are quoted in this article and these views are not necessarily representative of the QPS or any other organization.2 Policing Through Social Learning and Transnational Knowledge Networks QPS documentation (QPS 2014) as well as our interviewees suggest that the QPS placed ‘a heavy reliance on lessons learnt’ (interviewee, QPS G20 Group, 7 May 2015) as a project management philosophy that would help them adjust their own goals and techniques of policy in light of assessments of previous events. The QPS suggest they ‘plucked all the lessons that had come from everything we could find in similar countries’ (interviewee, QPS G20 Group, 25 March 2015), which involved a number of staff travelling to the United Kingdom, Europe and North America seeking to learn any lessons they felt could be useful. According to our interviewees, the principle of building on the experience of others applied broadly within the organization. As suggested by the following interviewee: Everyone who came into the organisation had an induction into what we were going to achieve and how we were going to achieve it. Part of that induction was that they had to do all this research—we actually brought the research in and they had to be across all the research—we put the lessons learnt on paper. … So we put this together and everyone who came into the organisation had to be across that. We also then made a very mindful decision that people were to be sent across the world because reports—reports give you some things but reports don’t give you those gold nuggets from talking to people on the ground as to what actually happened at their events. The organizational emphasis on learning, which appears to portray a deliberate attempt to engage in social learning, was formalized in a number of ways. Foremost among this was a risk registry system that centralized documentation from site visits and consultations with others on various aspects of the Brisbane G20 security operation (QPS 2016). Most of the key planning portfolios—such as dignitary protection, public order policing, offender management, etc.—were peer-reviewed on two separate occasions during the planning cycle, whereas the ‘risk management system’ and ‘lessons learned system … was spoken about every day of the week’ (interviewee, QPS G20 Group, 26 March 2015). Although our interviewees from QPS explained that they cast the net broadly regarding transnational engagement, all emphasized the importance of learning from Toronto. Many factors were put forward to explain this desire to establish an epistemic community with Toronto Police, including the comparable features between Australian and Canadian systems of government, similarities between the cities of Brisbane and Toronto, and a shared language. As one of our interviewees indicated, ‘there hadn’t been a G20 in a downtown city area in a Western democracy since Toronto’ (interviewee, QPS G20 Group, 20 March 2015), making Toronto a primary focal point for QPS when planning for the 2014 G20. A total of three QPS officials travelled to Toronto for a period of 7–10 days to discuss many aspects of mega-event security. Our interview data infers learning emerged through the relationships forged by QPS with other agencies, particularly TPS, particularly when such relationships were perceived to be based on an ‘open and honest exchange’ (interviewee, QPS G20 Group, 25 March 2015). The emphasis on learning was driven by a strong if not overriding desire to avoid repeating the problematic events characterizing the Toronto G20. Although QPS were sympathetic to the challenges faced in Toronto—many interviewees noted that TPS had only 6 months to plan for the G20 whereas QPS had 2 years—much of the analysis of Toronto by QPS had to do with determining ‘what happened [and] why it went so bad’ (interviewee, QPS G20 Group, 20 March 2015). One of our respondents spoke at length about this focus: Our Assistant Commissioner was very firm that a number of staff should go to Toronto and interrogate the Toronto Police about what they did well, what they didn’t do well and what they perhaps didn’t foresee in their planning that caused them problems during the event. We spent a week with the Toronto Police picking apart their G20. They were very generous in their information and advice to us about what didn’t work for them and what did and we were at least then able to combine that information with research we could do from London and Ireland for the G8 and G20 and a couple of other major events around the world. … If we didn’t go on that trip we would have made a lot of the same mistakes Toronto made or we would have made the same mistakes London made. The Toronto site visit provided a forum through which mistakes could be conveyed and social learning could occur. In particular, the TPS noted that, though the physical security of the G20 meetings was never compromised, plans to properly manage events outside the G20 secure zone were considered inadequate with hindsight, because of the limited timeframe that the police had to prepare for the event. For QPS observers, then, one of the lessons drawn from Toronto was not to concentrate on securing the physical locations of the G20 meetings themselves at the expense of any potential public disorder that might occur ‘outside the fence’. Similar to other mega-events, implementing this lesson was further motivated by a desire to safeguard the ‘reputation’ of parties involved in the event (Jennings and Lodge 2011; White 2014). One of our respondents elaborated on this point as follows: We had some very honest feedback from the Toronto G20 organisers, who by their own admission failed in terms of the intelligence effort in that they looked at just securing the event itself and didn’t look at what was going on in the periphery. So, we learned from that and said, we’ve got to look at the periphery and as you know from Toronto, there were a lot of riots in town, a lot of damage and very much reputational harm done to the Toronto Police and the RCMP and government. The Summit itself was safe and secure but that’s not good enough for us because one of our threats was reputational harm to government or the organisation. As this interviewee indicates, preparations towards managing events ‘outside the fence’ were seen as not only matters of intrinsic public good but investments in managing the reputational and financial risks that public disorder poses to the police. As such, instead of invalidating the strategies and tactics of TPS’s public order approach, the recognition of a ‘failure’ to account for a broader set of activities and the potential harm that this could introduce for the QPS reinforced the legitimacy of existing models and tactics, with the caveat that they needed to be executed much more effectively. Although QPS effected a number of additional failure-inspired adaptations in light of the Toronto experience, one of the most prominent was revisions to its command and control system. A widely-held standard for command and control among police services across the global north involves a three-tiered command structure often referred to as Gold, Silver and Bronze (GSB) (VPD 2011: 26). Gold Command is usually a high-ranking official who serves as the main strategic decision maker at a remote location. Silver Command decides on tactical deployment at the event itself, operating as the conduit between strategic decisions made at the Gold level and front-line officers. Silver Command also relays information and intelligence to Gold Command from on the ground. Finally, Bronze Command makes up the operational front-line officers who carry out the tactical plans decided upon by Silver. Recent innovations in police command and control include grafting the GSB model onto police operations centres (POCs) during high-profile but temporary events, which then serve as the physical location for centralized decision-making from Gold Command (Whelan and Molnar 2017b). In keeping with our focus on ‘failure-inspired’ developments in policing, the GSB command structure was developed in response to the perceived failures of command during the Brixton riots in which the excessive force of front-line officers was an exacerbating factor. Although the specific articulation of the GSB model will differ across police agencies, its overriding purpose is to provide front-line officers discretion and flexibility to deal with situations falling within well-defined limits and clear lines of command for situations that exceed those limits (Waddington 2007: 27–28). Since its development in the 1980s the model has been widely adopted and refined through numerous high-profile events, including the 2012 London Olympics (Fussey 2015), to become the benchmark in public order policing. TPS employed a modified version of the GSB command system known as the Incident Command System for the Toronto G20 (TPS 2011; London Emergency Services Liaison Panel 2015, pp. 21–22). As with other articulations of the GSB model, top-level strategic command in Toronto was embedded in TPS’s existing operations centre to create what was known for the duration of the G20 as the Major Incident Command Center (MICC) (RCMP 2011). Our interviews and post-event reviews conducted by police and civilian oversight bodies (RCMP 2011; OIPRD 2011) indicate that much of the turmoil experienced during the event stemmed from dysfunctional communication between the MICC and field officers. This was principally explained to have resulted from a lack of time to train officers seconded from other agencies in the command system (OIPRD 2011: xvi). These communication problems were exacerbated on the afternoon of Saturday, 26 June, by rising frustration within the MICC with the inability of public order units (POUs) to keep pace with fast-moving protest groups being observed on TV and CCTV feeds. These delays led to tactical and operational decisions to be made by strategic leaders within the MICC rather than by field officers (OIPRD 2011). A key TPS official describes this situation: Decisions started to be made from the Command Centre [MICC] level in terms of operational decisions that were ... that needed to be made. And we found that we lost our reliance on the commanders that were on the ground, facing the situation. They tended to step back and hesitate, waiting for instructions from a Command Centre above that was watching the situation occur on a video wall. My personal opinion is that we need to provide more autonomy to the commanders on the ground and let them make the decisions based on what they’re facing at the time, rather than, you know, through a process hesitating and waiting for instructions from above. I found it to be contrary to what we needed to do at the time. We didn’t realise that going in. But it certainly played out that way. As Saturday evening wore on, MICC commanders made the decision to ‘take back the streets’ (Perkel 2016) and issued sweeping instructions to restore order. Mass arrests occurred at multiple locations across the city. Furthermore, the following day, these instructions resulted in the kettling of approximately 400 persons for several hours during a thunderstorm at a well-known downtown intersection that has become an enduring symbol of excessive use of force. Preparations by QPS for the 2014 G20 included visiting a number of national and international police command centres and operations centres of corporate enterprises (QPS 2014: 54), but the command and control problems experienced in Toronto reportedly stood out as a prominent focal point (QPS 2016). Based specifically on lessons drawn from their evaluation of Toronto, QPS refined the precise setting of their GSB model to ensure greater ‘autonomy of command’ to Bronze-level field commanders (interviewee, QPS G20 Group, 25 March 2015). Specifically, QPS adopted a decentralized form of command in which lower commanders were ‘freed them up to take command’ (interviewee, QPS G20 Group, 25 March 2015) unless circumstances on the ground exceeded a specified risk threshold. One officer described these parameters, saying ‘if it was this, this and this, you can make the decision. If it involved this, you come to me … in the POC’ (interviewee, QPS G20 Group, 25 March 2015). In other words, whereas in Toronto decision-making authority tended to be exercised from within the POC, decision-making authority in Brisbane was pushed out to front-line officers unless support from higher levels of command was specifically requested by front-line officers. A QPS officer contrasted the approach taken in Brisbane to that of Toronto in the following way: I think for Toronto that was difficult. They were getting told over the radio to do something or not to do something and they weren’t allowed to make their own decisions. And it was quite apparent at one stage where one of the police vehicles was set on fire and there was quite a bit of physical violence at a police line. There were other police 50 meters away watching it happen but were directed not to intervene because their job was to protect the fence. … We decided that we were never going to let our commanders suffer that same fate. If they were standing outside the Hilton Hotel and something was happening across the road and they were of the view that their staff could assist in resolving that quickly and peacefully then that’s what they should be doing. They shouldn’t be asking me for advice on what to do. I think we freed them up to take command of it and that is probably the big difference. Another QPS officer reflected on how the lessons from Toronto led to changes in how QPS planned for the Brisbane G20: Well the huge difference was we empowered our commanders at a tactical level to make their own decisions. Toronto and what we heard from their debrief, very clearly they were told, they were getting instructions from their operations centre [MICC]; do this, don’t do that. The reality is the only person with the relevant situational awareness of what’s going on is the person there on the ground at the time. The QPS officer suggested this resulted in decisions being made with very little awareness or ‘feel’ about what was occurring, noting that ‘what we learned out of Toronto, some of those key decisions, tactical decisions were waiting for support from the Police Operations Centre’. The same official continued to describe how, at the Brisbane G20, the role of higher levels of command were redefined to provide support for field and front-line commanders rather than instruct their operations: I was [a] coordinator from the command centre during the event and what I would tell my … commanders was you are in charge of that [location], you make decisions that you think are appropriate to respond to whatever is going on. My job in the city here is to provide you with the resources for you to do what you need to do. It’s not to tell you what to do and it is not to tell you how to do it. Obviously, they could ask your advice and some did, but they were well aware that the responsibility … was theirs and theirs alone and what we had to do was provide them with sufficient training, knowledge and staff to generally run … as we had planned. In contrast to Toronto, where a short planning cycle prevented training of officers of their role and responsibilities within the overall command system, the 2-year preparation time for the Brisbane G20 allowed for much more in-depth training, creating what the QPS after-action report called a ‘foundation of proficiency’ throughout all levels of the command and control system (QPS 2016: 86). The expansion of discretionary authority for front-line officers extended to mobile POUs, which was a particularly problematic area for police at the Toronto G20. Mobile POUs are designed to provide a tactical rapid-response to address ‘flashpoints’ in different parts of the city. QPS viewed this as an explicit ‘mobility programme’ that was ‘built-in’ to the operation as a means to introduce greater agility and responsiveness to public order events in ways that would ameliorate gaps that plagued the Toronto G20. These revisions in command and control were accompanied with various subtler strategies and tactics regarding the way in which QPS sought to manage protestors. For example, in an effort to avoid the coercive control (de Lint and Hunt 2009) that was on wide display in Toronto, QPS purposefully avoided provocative displays of police in ‘full kit’ in favour of a ‘reflexive dramatisation’ promoting business-as-usual. In this way, the social learning that took place reflected Hall’s (1993) idea of second-order social learning—the adjustment of norms of public order policing, in particular those concerning the public representation of the policing response insofar as it involved a refinement of policy instruments to align with these norms and expectations. However, at the same time, the coercive edge of ‘hard’ tactics were never far away in Brisbane. As one of our interviewees explained, ‘we didn’t roll out public order from the get-go for this. Toronto did; they had public order police in full kit. We didn’t. All pre-staged, all hidden away. They were there if we needed them but we didn’t’ (interviewee, QPS G20 Group, 25 March 2015). A further difficulty encountered in Toronto was that operational orders issued from the MICC were not updated or revised as commanders rotated through their shifts, resulting in front-line confusion as to whether orders remained in place or, in the case of conflicting orders, which orders were to be followed (OIPRD 2011). In either case, front-line officers in Toronto were unable to adjust their tactics in light of the situation at hand. A QPS officer explained how efforts in Brisbane sought to empower officers working directly at the frontlines of police–protestor engagement, otherwise termed by QPS as being ‘at the coalface’ (QPS 2014: 86): I didn’t have the situational awareness even though I had vision, we had vision everywhere, I still couldn’t sense the mood of the crowd sitting up there so that is why that decision had to come from them. … I know in Toronto there were decisions about releasing protesters out at certain areas—it took hours and there were people on the ground saying we need to let these people go. Well [in Brisbane] that was up to the people on the ground to make that decision. But having said that though, I had to train them, give them the confidence that they were allowed to make that decision and that they would have our support in making that decision. And that directly came out of the research we did out of Toronto. It worked very well and I would never go back to the way we did it before. It just can’t happen. As the interviewee suggests, front-line decision-making with Issue Motivated Groups (IMGs) was supported by training as well as the practice of having senior officers in the field (interviewee, QPS G20 Group, 25 March 2015), thus attempting to make engagement through a more agile and mobile response. As one interviewee suggests, ‘decision-making [was] pushed down to such a level that it was front-line decision making. It didn’t have to come to the POC for the decision; someone was there to do it’ (interviewee, QPS G20 Group, 25 March 2015). In many respects, through enhanced training and an apparent revision in the QPS command and control structure, front-line officers were similarly burdened with managing the reputational risk associated with the QPS public order response (Jennings and Lodge 2011; White 2014). Conclusion Many scholars have noted the apparent diffusion of the strategies and tactics of strategic incapacitation beyond the initial laboratories of the United States and United Kingdom insofar as they reflect a technocratic emergence of policing tactics (Waddington 2007; della Porta and Tarrow 2012; Wood 2014). However, few have specifically analysed the internal form and content of these processes from the perspective of police agencies involved. Based on wide-ranging interviews with both ‘sender’ and ‘recipient’ police agencies (della Porta and Tarrow 2012), this article has directly examined these transnational processes of organizational learning and adaptation through an examination of how QPS sought to evaluate and learn from the Toronto 2010 G20. We have focused, in particular, on the refinements that QPS claim to have made to its command and control architecture in an effort to provide greater autonomy for front-line officers to avoid some of the operational challenges encountered in Toronto. We suggest that this process of failure-inspired learning sits alongside the imagination of potential worst-case scenarios as a significant driver in the scope and scale of mega-event security (Boyle and Haggerty 2009; Aradau and van Munster 2007; Boyle et al. 2015). Although worst-case scenarios promote extensive and expensive security plans, so too does the prospect of repeating the high-profile mistakes of others. But, how ‘failure’ is defined and discerned is a matter of perspective. After all, whether Toronto 2010 was a ‘failure’ for the police is not straightforward, as security for the G20 meeting itself was never in jeopardy. It may be that, similar to the APEC Summit in Vancouver 13 years prior (Ericson and Doyle 1999), Toronto police were willing to ‘die in a ditch’ (King and Waddington 2013: 262) once circumstances began to spiral out of control rather than allow the meetings to be compromised. By that measure the Toronto G20 was a success, however one that came with significant political, reputational and potentially financial risks to TPS and its allied agencies. What matters for our analysis is that QPS defined these outcomes as unacceptable, adamant that they would not let ‘things go wrong like they did at Toronto’ (interviewee, QPS G20 Group, 20 March 2015), and consequently undertook first- and second-order policy changes in light of the political morals of Toronto. These refinements left the underlying paradigm of public order policing at mega-events intact, even if they sought to ameliorate organizational complications that were seen to exacerbate the problems encountered in Toronto. What we wish to emphasize is that while police can, and do, learn from one another in transnational epistemic communities, they do so not in a free market of ideas but in fields already structured by power and politics. As Wood (2014) has already argued, the diffusion of police practices is contingent on preconditions specific to the underlying beliefs, values and attitudes toward organizational learning as well as wider geopolitical factors between nations that will structure transnational relationships between police and security actors (Dupont 2004; Boyle 2011; Whelan 2012; Brewer 2014; Whelan and Dupont 2017). In this case, the presence of a strong professional identity, sense of shared ‘mission’ between our case studies and a political proximity between Canada and Australia facilitated what QPS officers perceived as the exchange of ‘honest views of what went wrong’ (interviewee, QPS G20 Group, 26 March 2015). At the same time, this largely voluntary process of social learning unfolded within a socio-political context that, though instruments were refined, still reaffirmed the dominant global paradigm for public order policing. The social learning activities of QPS were consequently limited to ‘action-oriented conclusions’ (Rose 1991: 7) about how best to optimize the ‘precise settings’ of the strategies and tactics that continue to refine global ‘best-practice’ in public order policing (see Gillham 2011). Furthermore, we are not suggesting, of course, that QPS was the passive recipient of a fully formed model of police practice formulated elsewhere and faithfully recreated in Brisbane. Lesson-drawing, as Rose (1991: 24) articulates it, ‘is not a mechanical set of deterministic procedures leading to unalterable conclusions’. Instead, the lesson-drawing activities we have focused on are better understood as part of a continual and dynamic process of trial-and-error experimentation in which the instruments of a given policy framework are evaluated, refined, discarded or recombined in novel ways based on the experience of counterparts. Although it is useful to understand public order policing as models with identifiable traits (Gillham 2011), it needs to be remembered that such models are ‘mutable mobiles’ (Peck 2005: 767) that are articulated in new ways from event to event and city to city. Failure-inspired social learning within transnational policing communities—as embodied in the relationship between the TPS and QPS—is subject to local variances in how these learnings eventually unfolded in Brisbane. Our analysis of failure-inspired learning in public order policing, as a result, also points to a broader logic of social learning which is intent on eradicating failure from security governance more generally. Leese (2015) highlights the paradox in which the continuing failure of security technologies to deliver on their promised outcomes typically leads to renewed optimism for their eventual effectiveness. This is because technological failure is often promoted from the outset by suppliers and government buyers alike as an opportunity to fine-tune operational effectiveness. Failure in this logic is regarded as non-terminal insofar as it provides provocation for future first- or second-order adjustments (Hall 1993) rather than indicating a lack of efficacy (Leese 2015: 277; see also Malpas and Wickham 1995). Our analysis raises questions about a similarly recursive dynamic at work in transnational security governance. Scholarly and civil rights analyses regularly raise questions about the proportionality and legitimacy of contemporary public order policing practices through a critique of the apparent ‘failures’ at the Toronto G20 or other summits, with the aim of making public order policing more democratically responsive in the future. However, these activities must contend with the operationalization of failures within policing organizations in which failure can serve as an incitement to further refine the same policing practices that civil rights observers find undemocratic in the first instance. Failure, in other words, might be productively absorbed into the next iteration of ‘best practice’ rather than demonstrating a need to develop alternatives. Thus, rather than opening up space for public order policing to articulate in new and more democratically sensitive and accountable ways, the mechanism of failure-inspired social learning we have examined in this article may, in fact, enable the extension of troubling directions in security governance that avoid much public challenge or scholarly attention simply because they are not, in their new and ‘improved’ form, noticed at all. Funding There were no funds to support this research. Acknowledgements The authors would like to acknowledge the essential contribution of all of our interviewees, who willingly made time for this project in their busy schedules and who shared their experiences so openly. We, especially, would like to extend thanks to the Queensland Police Service for responding very positively to our requests to conduct interviews across the planning team, for providing various documents and for often brokering access to other interviewees. We would also like to thank the anonymous reviewers for their helpful comments on an earlier draft of this paper. References Alink , F. , Boin , A. and T’Hart , P . ( 2001 ), ‘ Institutional Crises and Reforms in Policy Sectors: The Case of Asylum Policy in Europe ’, Journal of European Public Policy , 8 : 286 – 306 . Google Scholar Crossref Search ADS Aradau , C. and van Munster , R . ( 2007 ), ‘ Governing Terrorism Through Risk: Taking Precautions, (Un)Knowing the Future ’, European Journal of International Relations , 13 : 89 – 115 . Google Scholar Crossref Search ADS Baker , D. , Bronitt , S. and Stenning , P . ( 2017 ), ‘ Policing Protest, Security and Freedom: The 2014 G20 Experience ’, Policing Practice and Research: An International Journal , 18 : 425 – 48 . Google Scholar Crossref Search ADS Beare , M. , Des Rosiers , N. and Deshman , A. C ., eds. ( 2015 ), Putting the State on Trial: The Policing of Protest During the G20 Summit . UBC Press . Bennett , C. J . ( 1991 ), ‘ What is Policy Convergence and What Causes It ?’, British Journal of Political Science , 21 : 215 – 33 . Google Scholar Crossref Search ADS Bennett , C. J. and Haggerty , K ., eds. ( 2011 ), Security Games: Security and Surveillance at Mega-Events . Routledge . Benzie , R. and Ferguson , R . ( 2010 ), ‘ Ombudsman Charges G20 Secret Law was ‘Illegal’ ’, Toronto Star [online], available online at https://www.thestar.com/news/gta/g20/2010/12/07/ombudsman_charges_g20_secret_law_was_illegal.html (accessed 23 May 2018). Birkland , T. A . ( 1997 ), After Disaster: Agenda Setting, Public Policy and Focusing Events . Cambridge University Press . —— ( 1998 ), ‘ Focusing Events, Mobilization, and Agenda Setting ’, Journal of Public Policy 18 : 53 – 74 . Crossref Search ADS —— ( 2006 ), Lessons of Disaster: Policy Change After Catastrophic Events . Georgetown University Press . Boin , A. , Hart , P. , Stern , E. and Sundelius , B . ( 2005 ), The Politics of Crisis Management: Public Leadership Under Pressure . Cambridge University Press . Boyle , P . ( 2011 ), ‘ Knowledge Networks: Mega-Events and Security Expertise ’, in C. J. Bennett and K. Haggerty , eds., Security Games: Surveillance and Control at Mega-Events , 169 – 84 . Routledge . Boyle , P. , Clément , D. and Haggerty , K. D . ( 2015 ), ‘ Iterations of Olympic Security: Montreal and Vancouver ’, Security Dialogue , 46 : 109 – 25 . Google Scholar Crossref Search ADS Boyle , P. and Haggerty , K . ( 2009 ), ‘ Spectacular Security: Mega-Events and the Security Complex ’, International Political Sociology , 3 : 257 – 74 . Google Scholar Crossref Search ADS Brewer , R . ( 2014 ), Policing the Waterfront: Networks, Partnerships and the Governance of Port Security . Oxford University Press . CBC News . ( 2015 ), ‘Mark Fenton, G20 Police Officer Behind Mass Arrests, Guilty of Discreditable Conduct’ , CBC News, available online at http://www.cbc.ca/news/canada/toronto/mark-fenton-g20-police-officer-behind-mass-arrests-guilty-of-discreditable- conduct-1.3202686. Courier Mail . ( 2014 ), Anarchists’ G20 Corporate Chaos Threat: Swarm and Destroy . Courier Mail . Crime and Misconduct Commission . ( 2015 ), Review of the G20 (Safety and Security) Act 2013: Consultation Paper and Invitation for Public Submissions . Queensland Government . Dolowitz , D. P. and Marsh , D . ( 2000 ), ‘ Learning from Abroad: The Role of Policy Transfer in Contemporary Policy‐making ’, Governance , 13 : 5 – 23 . Google Scholar Crossref Search ADS Donahue , A. and Tuohy , R . ( 2006 ), ‘ Lessons We Don’t Learn: A Study of the Lessons of Disasters, Why We Repeat Them, and How We Can Learn Them ’, Homeland Security Affairs , 2 : 1 – 29 . Dupont , B . ( 2004 ), ‘ Security in the Age of Networks ’, Policing and Society , 14 : 76 – 91 . Google Scholar Crossref Search ADS Ericson , R. and Doyle , A . ( 1999 ), ‘ Globalization and the Policing of Protest: the Case of APEC 1997 ’, British Journal of Sociology , 50 : 589 – 608 . Google Scholar Crossref Search ADS PubMed Fernandez , L . ( 2008 ), Policing Dissent: Social Control and the Anti-Globalization Movement . Rutgers University Press . Fussey , P . ( 2015 ), ‘ Command, Control And Contestation: Negotiating Security at the London 2012 Olympics ’, The Geographical Journal , 181 : 212 – 23 . Google Scholar Crossref Search ADS Fussey , P. and Coaffee , J . ( 2011 ), ‘ Olympic Rings of Steel: Constructing Security for 2012 and Beyond ’, in C. J. Bennett and K. Haggerty , eds., Security Games: Surveillance and Control at Mega-Events . Routledge . Fussey , P. , Coaffee , J. , Armstrong , G. and Hobbs , D . ( 2011 ), Securing and Sustaining the Olympic City: Reconfiguring London for 2012 and Beyond . Ashgate. Garland , D . ( 2001 ), The Culture of Control . Oxford University Press . Gillham , P. F . ( 2011 ), ‘ Securitizing America: Strategic Incapacitation and the Policing of Protest Since the 11 September 2001 Terrorist Attacks ’, Sociology Compass , 5 : 636 – 52 . Google Scholar Crossref Search ADS Gillham , P. F. , Edwards , B. and Noakes , J. A . ( 2013 ), ‘ Strategic Incapacitation and the Policing of Occupy Wall Street Protests in New York City, 2011 ’, Policing and Society , 23 : 81 – 102 . Google Scholar Crossref Search ADS Giulianotti , R. and Klauser , F . ( 2011 ), ‘ Introduction: Security and Surveillance at Sport Mega Events ’, Urban Studies , 48 : 3157 – 68 . Google Scholar Crossref Search ADS Haas , P. M . ( 1992 ), ‘ Introduction: Epistemic Communities and International Policy Coordination ’, International organization , 46 : 1 – 35 . Google Scholar Crossref Search ADS Hall , P. A . ( 1993 ), ‘ Policy Paradigms, Social Learning, and the State: The Case of Economic Policymaking in Britain ’, Comparative politics , 25 : 275 – 96 . Google Scholar Crossref Search ADS Hasham , A . ( 2013 ), ‘ G20 Assault Trial: Guilty Verdict for Officer Who Hit Adam Nobody ’, Toronto Star , available online at https://www.thestar.com/news/gta/2013/09/12/g20_assault_trial_guilty_verdict_for_of cer_who_hit_adam_nobody.html (accessed 23 May 2018 ). House of Commons Canada . ( 2011 ), ‘Issues Surrounding Security at the G8 and G20 Summits: Report of the Standing Committee on Public Safety and National Security’ , available online at http://publications.gc.ca/collections/collection_2011/parl/XC76-403-1-1-04-eng.pdf. Ironside , R. and Chamberlin , T . ( 2013 ), ‘ G20 Anarchists Vow Chaos and Mayhem for Brisbane’s Streets ’, Courier Mail , available online at http://www.couriermail.com.au/news/queensland/g20-anarchists-vow-chaos-and-mayhem-for-brisbanes-streets/story-fnihsrf2-1226778265923 (accessed 23 May 2018). Jennings , W. and Lodge , M . ( 2011 ), ‘ Governing Mega‐Events: Tools of Security Risk Management for the FIFA 2006 World Cup in Germany and London 2012 Olympic Games ’, Government and Opposition , 46 : 192 – 222 . Google Scholar Crossref Search ADS Johnston , L. and Shearing , C . ( 2003 ), Governing Security: Explorations in Policing and Justice . Routledge . Jones , T. and Newburn , T . ( 2006 ), ‘ Three Strikes and You’re Out: Exploring Symbol and Substance in American and British Crime Control Politics ’, British Journal of Criminology , 46 : 781 – 802 . Google Scholar Crossref Search ADS —— ( 2007 ), Policy Transfer and Criminal Justice . McGraw-Hill Education . King , M. and Waddington , D . ( 2006 ), ‘ The Policing of Transnational Protest in Canada ’, in D. Porta , A. Peterson and H. Reiter , eds., The Policing of Transnational Protest , 75 – 96 . —— ( 2013 ), ‘ Riots and Social Protest in an Age of Austerity ’, Policing and Society , 23 : 1 – 5 . Crossref Search ADS Kitchen , V. and Rygiel , K . ( 2014 ), ‘ Privatizing Security, Securitizing Policing: The Case of the G20 in Toronto, Canada ’, International Political Sociology , 8 : 201 – 17 . Google Scholar Crossref Search ADS Klauser , F . ( 2013 ), ‘ Spatialities of Security and Surveillance: Managing Spaces, Separations and Circulations at Sport Mega Events ’, Geoforum , 49 : 289 – 98 . Google Scholar Crossref Search ADS Leese , M . ( 2015 ), ‘ ‘We were Taken by Surprise’: Body Scanners, Technology Adjustment, and the Eradication of Failure ’, Critical Studies on Security , 3 : 269 – 82 . Google Scholar Crossref Search ADS London Emergency Services Liaison Panel . ( 2015 ), Major Incident Procedure Manual , Version 9.4. London : LESLP , available online at https://www.met.police.uk/globalassets/downloads/about-the-met/major-incident-procedure-manual-9th-ed.pdf. de Lint , W. and Hunt , A . ( 2009 ), Intelligence Control: Developments in Public Order Policing in Canada . University of Toronto Press . Malpas , J. and Wickham , G . ( 1995 ), ‘ Governance and Failure: On the Limits of Sociology ’, The Australian and New Zealand Journal of Sociology , 31 : 37 – 50 . Google Scholar Crossref Search ADS McCann , E . ( 2011 ), ‘ Urban Policy Mobilities and Global Circuits of Knowledge: Towards A Research Agenda ’, Annals of the Association of American Geographers , 101 : 101 – 30 . Google Scholar Crossref Search ADS McGuire , E. and William , K . ( 2013 ), ‘ Transferring Criminal Investigation Methods from Developed to Developing Nations ’, Policing and Society , 23 : 346 – 61 . Google Scholar Crossref Search ADS Milberry , K. and Clement , A . ( 2015 ), ‘ Policing as Spectacle and the Politics of Surveillance at the Toronto G20 ’, in M. Beare , N. Des Rosiers and A. C. Deshman , eds., Putting the State on Trial: The Policing of Protest During the G20 Summit , 127 – 47 . UBC Press . Molnar , A. and Snider , L . ( 2011 ), ‘ Mega-Events, Mega-Profits, Unpacking the Mega-Event Security-Development Nexus ’, in C. Bennett and K. Haggarty , eds., Security Games . Taylor & Francis . Monaghan , J. and Walby , K . ( 2012 ), ‘ Making up ‘Terror Identities’: Security Intelligence, Canada’s Integrated Threat Assessment Centre and Social Movement Suppression ’, Policing and Society , 22 : 133 – 51 . Google Scholar Crossref Search ADS Moore , T . ( 2012 ), ‘ Is G20 Worth Brisbane’s While ?’, Brisbane Times , available online at http://www.brisbanetimes.com.au/queensland/is-g20-worth-brisbanes-while-20120711-21wga.html (accessed 23 May 2018). Nohrstedt , D. and Weible , C. M . ( 2010 ), ‘ The Logic of Policy Change After Crisis: Proximity and Subsystem Interaction ’, Risk, Hazards & Crisis in Public Policy , 1 : 1 – 32 . Google Scholar Crossref Search ADS Office Of The Independent Police Review Director . ( 2011 ), ‘Policing the Right to Protest: G20 Systemic Review Report’ , available online at http://www.oiprd.on.ca/EN/PDFs/G20-Systemic-Review-2012_E.pdf. Peck , J . ( 2005 ), ‘ Struggling with the Creative Class ’, International Journal of Urban and Regional Research , 29 : 740 – 770 . Google Scholar Crossref Search ADS Peck , J. and Theodore , N . ( 2010 ), ‘ Mobilizing Policy: Models, Methods, and Mutations ’, Geoforum , 41 : 169 – 74 . Google Scholar Crossref Search ADS Perkel , C . ( 2016 ), ‘ Toronto Cop Who Ordered Mass Arrests at G20 Given Reprimand, Avoids Firing ’, The Canadian Press , available online at https://globalnews.ca/news/2762961/toronto-cop-who-ordered-mass-arrests-at-g20-to-be-sentenced-today/ (accessed 23 May 2018). della Porta , D. and Tarrow , S . ( 2012 ), ‘ Interactive Diffusion: The Coevolution of Police and Protest Behavior with an Application to Transnational Contention ’, Comparative Political Studies , 45 : 119 – 52 . Google Scholar Crossref Search ADS Queensland Government . ( 2013 ), G20 (Safety and Security) Act 2013: Act No. 56 of 2013 , available online at https://www.legislation.qld.gov.au/view/pdf/asmade/act-2013-056 (accessed 23 May 2018 ). —— ( 2014 ), G20 (Safety and Security) Regulation Act 2014: Subordinate Legislation 2014 No. 220 , available online at http://www.parliament.qld.gov.au/Documents/TableOffice/TabledPapers/2014/5414T6411.pdf (accessed 23 May 2018 ). Queensland Police Service . ( 2014 ), G20 Group: Operation Southern Cross: Policy and Procedures Manual . Queensland Police Service . —— ( 2016 ), G20 – Operation Southern Cross: Post Operation Report . Queensland Police Service . RCMP . ( 2011 ), G8 and G20 Summit Integrated Security Unit After Action Report . Royal Canadian Mounted Police . Roach , K . ( 2015 ), ‘ Post-9/11 Policing of Protests: Symbolic but Illusory Law Reform and Real Accountability Gaps ’, in M. Beare , N. Des Rosiers and A. C. Deshman , eds., Putting the State on Trial: The Policing of Protest During the G20 Summit . UBC Press . Rose , R . ( 1991 ), ‘ What is Lesson-Drawing ?’, Journal of Public Policy , 11 : 3 – 30 . Google Scholar Crossref Search ADS Sabatier , P. A. and Weible , C . ( 2007 ), ‘ The Advocacy Coalition: Innovations and Clarifications ’, in P. Sabatier , ed., Theories of the Policy Process , 2 nd edn, 189 – 220 . Westview Press . Samatas , M . ( 2007 ), ‘ Security and Surveillance in the Athens 2004 Olympics: Some Lessons From a Troubled Story ’, International Criminal Justice Review , 17 : 220 – 38 . Google Scholar Crossref Search ADS Stone , D . ( 2004 ), ‘ Transfer Agents and Global Networks in the ‘Transnationalization’ of Policy ’, Journal of European Public Policy , 11 : 545 – 66 . Google Scholar Crossref Search ADS Toronto Police Service . ( 2011 ), ‘G20 Summit: Toronto Police Service After-Action Review’ , available online at http://www.torontopolice.on.ca/publications/files/reports/g20_after_ action_review.pdf. Vanstone , M . ( 2008 ), ‘ The International Origins and Initial Development of Probation: An Early Example of Policy Transfer ’, British Journal of Criminology , 48 : 735 – 55 . Google Scholar Crossref Search ADS VPD Vancouver Police Department . ( 2011 ), Vancouver Police Department 2011 Stanley Cup Riot Review . Vancouver Police Department . Waddington , D. P . ( 2007 ), Policing Public Disorder: Theory and Practice . Willan Publishing . Whelan , C . ( 2012 ), Networks and National Security: Dynamics, Effectiveness and Organisation . Routledge . —— ( 2014 ), ‘ Surveillance, Security and Sports Mega Events: Toward a Research Agenda on the Organisation of Security Networks ’, Surveillance & Society , 11 : 392 – 404 . Whelan , C. and Dupont , B . ( 2017 ), ‘ Taking Stock of Networks Across the Security Field: A Review, Typology and Research Agenda ’, Policing and Society , 27 : 671 – 87 . Google Scholar Crossref Search ADS Whelan , C. and Molnar , A . ( 2017a ), ‘ Policing Political Mega-Events Through ‘Hard’ and ‘Soft’ Tactics: Reflections on Local and Organisational Tensions in Public Order Policing ’, Policing and Society , (online first 23 January 2017). —— ( 2017b ), ‘ Managing Flows During Mega-Events: Taking Account of Internal and External Flows in Public Order Policing Operations ’, Global Crime , 18 : 176 – 97 . Crossref Search ADS White , A . ( 2014 ), ‘ Post-Crisis Policing and Public–Private Partnerships: The Case of Lincolnshire Police and G4S ’, British Journal of Criminology , 54 : 1002 – 22 . Google Scholar Crossref Search ADS Wood , L. J . ( 2014 ), Crisis and Control: The Militarization of Protest Policing . Pluto . Wood , L . ( 2015 ), ‘ Reorganizing Repression: Policing Protest, 1995–2012 ’, in M. Beare , N. Des Rosiers and A. C. Deshman , eds., Putting the State on Trial: The Policing of Protest During the G20 Summit . UBC Press . Wood , J. and Dupont , B . eds. ( 2006 ), Democracy, Society, and the Governance of Security . Cambridge University Press . Yin , R . ( 2014 ), Case Study Research: Design and Methods , 5 th edn. Sage Publications . Footnotes 1 The ACBPS became the Australian Border Force on 1 July 2015. 2 The QPS requires that the following disclaimer be stated: The views expressed in this material are those of the authors and are not those of QPS. Responsibility for any errors of omission or commission remains with the authors. The QPS expressly disclaims any liability for any damage resulting from the use of the material contained in this publication and will not be responsible for any loss, howsoever arising, from the use of or reliance on this material. © The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). All rights reserved. For permissions, please e-mail: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Jackson,, Paul;Kassaye,, Demelash;Shearon,, Edward

doi: 10.1093/bjc/azy030pmid: N/A

Abstract This article examines the introduction of community policing in Dire Dawa, Ethiopia. It shows that the relationship between the security actors within the district is a complex one that neither represents a simple dichotomy between state and non-state, nor an emerging clear and hybrid system. Rather it is a negotiated arrangement between a top-down, statist ideology and local forms of justice process, a balance that has historically characterized Ethiopian internal security for decades. The community police initiative offers a positive way of reducing friction between the different policing providers through acting as interlocutors but also enforcing the state’s legitimacy in others. Local providers can use local actors to enhance their reach and their effectiveness but also extend the reach of the state and the legitimacy of the law at the local level constructing a negotiated ambiguity between central control and local agency in policing. ‘Police in previous times were brutal in their actions. We used to run when we saw them. Now the changes are dramatic and the reason for this comes from two sources: the policy directs all to behave similarly and hence developing a mentality of serving the. In general, the police are now the servants of the community, not superior to the community as in previous times.’1 Introduction This article explores an example of hybrid security provision in Dire Dawa in Ethiopia involving a collaboration between the state and local community and the extensive development of community policing approaches. Within policing, and specifically within the discourse on African community policing, hybridity is used to describe the relationships and structures that arise when different policing providers co-exist and work together to police a particular community. Drawing on the experience of community policing in Dire Dawa, we explore whether this initiative, although successful, has anything to offer beyond a version of legal pluralism or adding additional resources to a stretched police service and, perhaps, to reflect Abrahams ‘…fundamental nature of the problems of relations between ordinary villagers, criminals and government’ (1998:52). We address a fundamental tension within a hybrid approach between community policing as a means of control and as a means of actual policing by the community itself (Ruteere and Pomerolle 2003; Cooper-Knock 2014; Cooper-Knock and Owen 2015). Our argument is that Ethiopia has been subject to a continual balance between a centralized, state system that is seeking hierarchical control and has historically clashed with entrenched local systems of hierarchy, including policing and justice. The introduction of community policing in Dire Dawa represents a new chapter in this continual negotiation between the central and the local and the net effect of the policing approach has been to grant significant roles for the local community in spreading ideas of formal justice as well as recognising the importance of local conflict mediation, for example. In one sense, all parties have an interest in reforming a system that had become rotten, but tensions remain between those seeking more state control and those seeking local power according to tradition. At the heart of the argument is examination of a fundamental tension that remains unresolved in Ethiopia: how does a state-led, top-down, government initiative manage to cede power to local systems of justice and why? Within this contradiction lies a series of issues including the relationship between a previously oppressive police force designed to terrify, the new police force and local resistance from traditional authorities; and the tension between customary law and formal, state law. This paper argues that whilst this tension is unlikely to be resolved any time soon, it forms the basis for a continuing negotiation between the state and traditional authority. Furthermore, whilst there is a recognition that local traditional authorities can play a role in some approaches to crime, conflict and justice, there is also a clear use of community policing to assert the power of the constitution, human rights law and formal systems, and to enhance the role and power of the police at community level. This article begins by outlining the links between community policing and hybrid security. It then briefly sets out the Ethiopian context and discusses the methodology that was used and the limitations of research in relatively unstable environments. It also highlights the lack of statistical or detailed survey information available—or indeed the lack of opportunity to carry out detailed surveys—and the continued reliance on perceptions of security rather than detailed data. The article then sets out the broad situation in Dire Dawa and, drawing on the extensive fieldwork findings, it discusses the tensions between hybrid approaches and assimilation in to the state system. Finally, it makes the point that community policing here has been successful in building considerable trust, but this has resulted in every interviewee reporting reductions in crime, whilst the crime figures have increased. There may be several reasons for that, including under-reporting under the previous, oppressive regime, but this is unclear. Secondly, some specific groups have benefitted from the state-led and sponsored reform, specifically women and girls. Certainly in Dire Dawa, whilst there remains an issue of gender-based violence, female genital mutilation and early marriage, it is clear that the community police have managed to rapidly increase the numbers of females reporting these crimes. Previously, these hidden crimes had been dealt with by local authorities maintaining traditional inequalities and social hierarchies. Finally, the process of community policing has been a delicate balance between recognising the legitimacy of local authorities to tackle certain instances of crime and also educating local people on modern law of the constitution. Ultimately there is a process of incorporation of traditional rule in to legal formality and subordination to the rule of law rather than a new hybrid system. Community Policing and Hybrid Security Arrangements Community policing is a concept that has been adopted across Africa in varying forms, including in Ethiopia (Baker 2013; Di Nunzio 2014), Mozambique (Kyed 2010), Nigeria (Hills 2014), and Sierra Leone (Albrecht et al. 2014). Much of this development has been undertaken by donor-led programmes reinterpreting approaches from the global North. The structures and processes of policing across Africa vary greatly, with institutionalised police structures being patchy and empirical policing characterized by a multiplicity of providers drawing on a wide range of authorities (Bierschenk 2017). As Baker (2009, 2013) points out, the interpretation of community policing in Africa has largely consisted of the creation, recognition and incorporation of a wide range of alternative security providers with a correspondingly wide variation in capability, legitimacy and links to the state. This is a pragmatic recognition of reality, with lack of resources available to the police preventing coverage, particularly in rural areas (Baker and Scheye 2007). It also presents the African state in a permanent state of negotiation, or mediation (Menkhaus 2006) with non-state providers and a hybrid political system (Boege et al. 2008) where the state is only one provider amongst many and where boundaries between state and non-state, international and local, are problematized by complex interactions and interpenetration of providers following different logics and drawing on varying sources of power (MacGinty 2010; Richmond 2011; Luckham and Kirk 2012). The hybridity debate partly addresses this discussion around balances of power between alternative sources of legitimacy, recognising that the extent of hybridity is partly a political choice and partly the result of institutional pressures arising from institutional constraints. The international categories of ‘state’ and ‘non-state’ create a bias that prevents analysis of the real nature of domestic politics, or the ‘politics of the everyday (Schroeder and Chappuis 2014). This raises a number of issues around legitimacy and political power at the local level and part of any analysis of a hybrid system has to include a view of where the power lies within such systems and who exercises power and how. Hybridity is really an attempt to understand the empirical reality of power beyond the simple binaries of analytical categories like international and local. How they interact, with whom and what results from that interaction in both systems is the core concern of hybrid approaches. Other literature examines the nature of these local political interactions and their role in shaping, resisting and constraining international projects. Pouligny (2006) and Autesserre (2010) both focus on viewing peacekeeping from below, whilst there is a growing literature on the role of domestic elites in both taking opportunities to benefit from international intervention, but also in developing international approaches to democratic transition. Richmond (2011) and Mac Ginty’s (2010) work in particular takes this discussion forward by attempting to see the political actions of local actors as being more than mere ‘spoilers’ in a positive process of attaining liberal peace. Hybrid forms of governance are constructed by the complex interaction of multi-layered actors and it is the result of these interactions that can be said to be ‘hybrid’. The particular outcome is unpredictable and variable—not all hybrid orders are the same. State and customary forms of government ‘…do not exist in isolation from each other, but permeate each other and, consequently, give rise to different and genuine political orders that are characterised by closely interwoven texture of their separate sources of origin (Boege et al. 2008)’. In other words, the state becomes just one of many actors within the state building process (Jackson 2011). Hybrid order is difficult to examine empirically and may encompass actors and institutions that rarely measure up to western normative standards of justice, or even gain from political disorder as a tool of political power (Chabal and Deloz 1999). In some places those responsible for delivering justice may actually be those most responsible for insecurity in the first place, or alleged agents of insecurity may be providing a particular form of security. This may explain why the concept of hybridity may actually be more useful as a tool to critique liberal approaches to peacebuilding rather than offering a credible alternative strategy or a practical approach (Luckham and Kirk 2012). This is a feature of what Goodfellow and Lindemann (2013) refer to as substituting ‘hybrid’ for ‘complex’ and of using the term hybrid to describe what is actually institutional multiplicity. Hesselbein et al. (2006) describe institutional multiplicity as a context where multiple rule systems confront individual actors providing distinct and different normative frameworks and incentive structures in which they act. An end-user seeking justice may therefore be faced with a series of institutional choices each with different consequences and different sources of legitimacy, power and agency rather than one coherent hybrid system. Much literature refers to arrangements that are probably best described as institutional multiplicity rather than hybridity as a new form of institutional system (Cross 2016). Multiple institutions co-exist and interpenetrate rather than merge in to completely new systems. International organizations also identify hybridity as one way of boosting the state by adding non-state multipliers to service provision. In community policing, there is logic in seeking greater governance effectiveness by incorporating local providers through, at the very least, improving intelligence gathering and managing local ethnic or family clashes before they get to the state. As Baker (2013) points out in citing Bourdieu, the police may be strong in terms of economic capital (state and donor support), cultural capital (training and experience) and social capital (history), but private firms may be able to compete with the police in terms of economic and cultural capital and so there is a case for hybrid arrangements. Non-state actors are usually strong in terms of symbolic capital (prestige and local legitimacy), particularly where the police are perceived negatively or as agents of foreign powers. The attraction in a hybrid arrangement is that non-state actors and the police can both gain; non-state actors gaining economic, cultural and social capital and the police gaining symbolic capital (Baker 2013). Cooper-Knock and Owen (2015) argue that communities are also implicated in the reproduction of police practice and that police forces can authorize or limit regulatory intervention; exercise a unique bureaucratic power over ‘right’ and ‘wrong’; and, can take activity ‘off the books’ as vigilantism. In Ethiopia, the history of political wrangling between core and periphery means that the development of community policing is the evolution of a political struggle between the central state and the local community. In practice, this is a difficult set of arrangements to develop and critics point to the idea of ‘bargain based governance’ (Meagher 2012: 1078) or even an emphasis on potential coercion, vigilantism and settling local scores rather than service provision (Cooper-Knock 2014; Tapscott 2016). Local institutions may not have a monopoly on local legitimacy and also tend to reflect local social hierarchies. Access to resources and networks with which to engage in hybrid arrangements are not equal at the local level, which creates its own political logic. Actual hybridity may be characterised by a series of exchanges of capital on the boundaries between actors (Baker 2013). Given this, such arrangements will tend to be dynamic and uneven and some will have more options than others. They will also rely on the motivations of the actors involved to frame the nature of those dynamics. The Context of Ethiopian Community Policing Ethiopia has a population of around 102 million people, including more than 80 distinct ethnic groups and is governed by the Ethiopian People’s Revolutionary Democratic Front (EPRDF), which follows a broadly socialist and state-led model of development. Emperor Haile Selassie ruled until 1974 with an elitist approach that created deep societal cleavages and inequality. He was eventually overthrown in 1974 by the Derg military regime of Colonel Mengistu Haile Mariam that espoused socialist intentions but was extremely violent, and left a lasting legacy of police brutality. As Ethiopia became increasingly riven by ethno-nationalist conflict, the Derg regime used the police to maintain control, meaning violence against the local population. The background to any police reform, therefore, remains a history of arbitrary arrests, violence and rape. In 1991, the Derg was overthrown, and replaced by Meles Zenawi, who ruled until his death in 2012. Meles remains an important figure, not least because he managed to use his cross-ethnic power base and credibility as former head of the Tigray People’s Liberation Front (TPLF), to establish a new constitution, including reform of the police. In 1994 he established a transitional police regime comprising of former militia whilst he dismissed or retrained the Derg police force and trained a new force (Denney and Kassaye 2013). A history of violent regime change, oppressive militarized rule and strong separatist movements has produced a history of central-local tension within Ethiopia since the early twentieth century. The police have been at the centre of that. Regime security is regarded as a critical element of state strategy given external threats from Somalia and Eritrea, internal threats from several separatist groups, and a significant threat from domestic and international terrorism2. High rates of violence, particularly against women, mean that personal security is also high on the agenda (Burton et al. 2011). However, as Baker (2013) points out, the state itself is aware of the limited resources available to it and its own inability to extend effective state control in some areas, particularly over the 83% of the population in rural areas. In keeping with much African practice, security is delivered alongside local providers, and traditional authorities retain influence at the local level. These authorities, however, have their own resource constraints and may also espouse values that are incompatible with modern perceptions of human rights or gender issues, for example. With a poor policing infrastructure, users are faced with a choice of inadequate state provision and lengthy delays in investigation or court cases, or facing a justice system that reflects local political structures and social hierarchies. The plural Ethiopian legal system, for example, incorporated strong influences from continental civil law and common law in the 1950s and 60s alongside traditional (or customary) laws and norms. This has resulted in multiple layers of formal, customary and religious laws and processes (Green et al. 2012). The pluralistic legal system is partly mirrored by policing, with a strong influence of traditional forms of policing beyond the formal police structure. The formal structure is divided in to Federal and Regional Police. The Regional Police manage the nine regional states whereas the Federal police are responsible for the federally administered areas of Addis Ababa and Dire Dawa. As the best trained and equipped force, they also provide support to Regional Police, varying from the ‘more developed’ regions of Amhara, Harare, Oromia, Southern nations, Nationalities and People’s Region and Tigray where support is a negotiated activity, through to the ‘less developed’ areas of Afar, Benishangul-Gamuz, Gambella and Somali, where the Federal Police operate more extensively and frequently beyond the remit of the regional government (Green et al. 2012). In reality, policing in many rural areas in particular is far more complex and is characterized by a multiplicity of institutions. These incorporate a variety of conflict resolution mechanisms including sharia courts, women and youth groups, peace committees and particularly traditional authorities and elders who are the most used reconciliation mechanism in Ethiopia (Baker 2013). These leaders are in very respected positions and preside over crimes ranging from minor individual issues through to clan disputes over cattle raiding and even revenge killing, and the police frequently refer cases to them for decisions (Green et al. 2012; Baker 2013). The judgement of the elders is frequently taken as the main form of justice and any leadership exercised in terms of community policing must take these leaders in to consideration. This leadership class can oppose or enact significant elements of security and justice emanating from the central state. There are also militia present at the local level, usually significantly outnumbering the Regional Police. Militia are effectively a politicized paramilitary comprising armed volunteers, frequently from the ethno-nationalist militia that opposed the Derg. They usually mount community patrols and man checkpoints on major roads, but whilst they outnumber the police, they are also seen as a political wing of the government and are therefore viewed with some element of suspicion (Denney and Kassaye 2013). Policing and justice for many Ethiopians is therefore a process of negotiation between a series of multiple agencies and agents. There is also a form of continual negotiation between the central (and centralizing) authority of the state and state policing, and the local agency of the customary and religious authorities. Community policing is seen as a way of bringing some of these diverse elements together and providing a leadership role for the police following perceptions of rising crime during the dramatic 2005 elections (Denney and Kassaye 2013). In 2006, community policing was included as part of a broader police reform programme. In many ways this form of centralized introduction was surprising for a policy that is held to be a bottom up solution. Communities were essentially informed that they would get community policing as the new model and there does not seem to be any evidence indicating broad demand. In Dire Dawa, these reforms were embraced by the police and community policing incorporated in to police training. In addition, local community members were selected (by the community) to go for police training and were then returned to the communities as community police officers. As Denney and Kassaye (2013) point out, the process of community policing was partly perceived at the local level as a means of formalising structures that had existed in some for some time, but it should also be noted that many of these systems have grown up in the absence of effective policing and whilst Ethiopia has a long history of dualism it has not been a hybrid, but has been one of interaction, not integration (Baker 2013). The development of community policing since 2006 has been a fundamental change as the Government of Ethiopia explicitly pursues a policy of hybridity, by which they mean the incorporation of local justice in to the regular policing activity (Green et al. 2012). A key narrative arising from the research in Dire Dawa is the underlying development of a relationship between the police and the community itself. This was routinely stated within interviews, usually using the official phrase of ‘bringing the police and the people near’. Interviewees repeatedly identified this change while communicating support for the reduction in crime and satisfaction with the current practice of policing. Nearly all interviews contained references to practices during the time of Haile Selassie’s time in power and the transition to the Derg regime: ‘During previous times, we were afraid not only them, but their shadows as well. We attempted to escape … when we saw them patrolling ... We saw the police as if they are always there to attack people.’ Many of the police commanders, clan elders, and community advisors lived through previous eras of brutality. These experiences influenced the national desire for police reforms and in their lifetime they experience a transition to a current of reduced crime and increased trust with the state police. In some ways, community policing arrived at a time when everyone had an interest in reforming a rotten system. Baker (2013) outlines a number of reasons for the changes to a community policing approach and, in general, our research supports his assertion that whereas before 2006 and probably up until 2010, the process was really one of recognition and interaction rather than integration. After this, the process intensified and community policing has become about integration in to an approach that incorporates traditional policing. Given the top-down introduction of the reforms, it is perhaps unsurprising that there are several features that point to the formalization and incorporation of traditional mechanisms in to the state system rather than the genuine creation of a new hybrid. The overall philosophy of the state draws on a socialist and developmental ideology that seeks to provide services to all people. However, lack of resources meant that this could never be realized and so the state needed partners. Statist ideology also dictates that non-state actors should not sit outside the centralized system and so the process of incorporation has been largely that of systematizing non-state actors including popular mechanisms of justice provision (Baker 2013; Denney and Kassaye 2013; Di Nunzio 2014). In practice, police officers have more than one option open to them in dealing with any one crime and can refer cases to the traditional authorities instead of waiting for the inefficient court system to deal with the case. Our research also supported the view taken by Denney and Kassaye (2013) that the absence of resources and the ideological drive of the state for control are not the only things that influence community policing. Whilst it may be that this makes community policing attractive for the government, the police themselves and the communities ascribe a different set of reasons for the introduction and subsequent success of community policing. The intention behind community policing is to improve outcomes, reduce crime rates and improve relationships with the community. Firstly, the police themselves very much saw community policing as a means of reducing crime. This was routinely mentioned in interviews, and the police are very keen to present community policing as being effective in giving people responsibility to tackle crime and support the police. Police officers brought up community policing as being effective in gaining intelligence on criminal activity, but also in empowering citizens to talk to the community police officers about crimes.3 Research approach The research was undertaken by local and international researchers in Ethiopia and the UK between February and June 2017, including fieldwork in Dire Dawa in March 2017. The research focused on the district (woreda) and village (kebele) levels where most disputes are mediated. Seven kebeles were visited, four urban and three rural, and at each location police commanders, community police officers, clan leaders, and members of advisory councils were interviewed. When possible, a mixture of male and female police officers were interviewed. Advisory councils consisted of a mixture of male and female. Clan leaders are male, but in rural areas women’s associations exist, and members of those groups were interviewed. Non-police interviews were conducted in private rooms without police presence. In total, 15 in-depth interviews were conducted with police commanders in both urban and rural locations and with 15 NGO representatives. There were 12 focus group discussions with clan leaders, police officers, local associations, and community advisory council members. These were led by both an international and local researcher from the University of Addis Ababa, who also acted as translator. In 2016, Ethiopia experienced political protests resulting in a declaration of emergency, the suspension of some internet communication services and restricted movement in the country; while in Dire Dawa, the researchers were escorted from site to site by assistants of the area police commander. At the same time, civilians involved with the community policing structure were selected for interview by the researchers and were interviewed and participated in the focus groups without a police presence. External organizations were also accessed by the research team to improve the cross-referencing of the responses. Clearly respondents were those who had some role within the policing and justice area either through membership of a committee or a local advocacy NGO. The security situation and the relationship with the police prevented a wider survey approach. This assessment is also cross-referenced with a wide range of published and unpublished studies in Ethiopia, all of which experience similar issues. The research team strove to make the research as objective as possible given the need for reliance on the police for access. However, as Green et al. (2012) notes, it is important to recognize that systematic government or other statistical or survey data on community-level insecurity, crime and justice problems in Ethiopia is scarce. All of the issues relating to the collection of data across much of Africa are present, but are also exacerbated by a combination of sensitivity about crime and security and weak, patchy and unreliable collection and reporting of data. There are very few large scale surveys with relevant data, and police and court data on reported crimes is not publicly available. Police statistics are made available to Ministers in a summary format in annual reports and is generally difficult to interpret, whilst court data may be more available in urban areas but reliability is unclear (Greene et al. 2009). It is also clear that the majority of crimes and disputes are handled outside this system where there are no records. Perception surveys therefore remain key sources of general crime trends in the absence of reliable crime statistics, even though they may not correspond to actual rates of criminality. Perceptions of criminality, however, do matter because even if they are not accurate representations of empirical reality of crime, they directly affect peoples’ behavior. In Ethiopia, perceptions of risks of violence may affect willingness to travel to markets, schools, health care, etc. What Does Policing Look Like in Dire Dawa? Policing structures in Dire Dawa represent something of a spectrum. There are five main policing departments: Criminal Investigation, Police Academy, Human Resources, Finance/Logistics and Community Policing. The Criminal Investigation unit involves armed, rapid response units as well as units who handle more serious crimes. The Human Resources and Finance/Logistics departments provide administrative support for the police. The Community Policing department oversees the management of all the community policing officers in the region. Community police officers are nominated by the community to serve in that community, which also vets them. These officers are then centrally trained in policing and then posted back in to those communities as police officers. The community is additionally responsible for funding the construction of the community police stations in their immediate area. These stations are typically one room buildings open 24 hours a day with living areas attached to the station. Communities open bank accounts and citizens contribute money until the goal is met, then the community also participates in the construction of these structures. The Community Policing structure in Dire Dawa consists of 11 urban zones and 4 rural zones. Each zone is then broken down into different subdivisions: kebeles, katana, blocks and ‘family police’. An important, but potentially confusing conflation of the word ‘policing’ emerges at this level. There are community police officers who are paid by the state who perform official roles, but equally important to the strategy of community policing is the assistance of the community in the process, and these members are also referred to as ‘police’ or community advisors. In one zone, there were 115 advisory council members, all elected by members of that zone. This is the same for village ‘police’, block ‘police’, campus ‘police’ and family ‘police’. At the village and block level, the ‘police’ are members of the community, often elders, priests, or other individuals of influence. At campus level, the ‘police’ are children whose responsibilities include making sure peers do not bring cell phones to school and making sure students do not skip classes. The concept of ‘police’ is deeply entrenched within communities. One of the most Orwellian sounding aspects of this approach is the idea of ‘family police’. These are typically fathers or older sons within large family units at the local level, mirroring the social hierarchies of traditional structures. Family police essentially control small scale activity like petty theft, truancy and family disputes, but may also get involved in broader crimes related to marriage of underage girls or genital mutilation where the family police are expected to liaise with the community police. Once it goes beyond the family’s ability to resolve the issue this will then be reported to block police and the community police will become involved. In this way, local disputes can be managed by customary or family structures that are incorporated in to a bigger policing structure, but which don’t necessarily involve the police themselves. Crime in Dire Dawa is dominated by ‘assault’. Between 2012 and 2016 there were 6903 assaults, 3011 thefts, 2883 incidences of anti-social behaviour or minor fights and 2067 minor disputes.4,5 Homicide is low with only 84 murders and 61 attempted murders recorded. Anti-social behaviour is acting outside of the desired cultural norms and ‘minor fighting’ was defined as a small fight involving punching or the ‘breaking of teeth’, but different from assault as there was no intent to cause great harm or potential death. All of these figures fluctuated significantly over the period, but there has been a relative increase. For example, theft increased from 301 in 2012 to 856 in 2016, which is clearly at odds with the overwhelming sense of reduced crime amongst local people. These are official figures, and so do not record instances that happened and were dealt with in the informal system. It may also reflect a willingness to report crime. This is certainly claimed by local activists advocating against gender-based violence and several members of the community and police reported an increased rate of individuals feeling empowered to report crimes, specifically domestic violence.6 A women’s group interviewed stated that one of the key factors limiting women from reporting domestic violence is an economic dependence upon males. Prior to the community policing reforms, if a wife’s complaint about domestic abuse sent her husband to court, it might be months before the case was heard, and the accused might be held in prison the whole time, negatively impacting the family’s financial situation. Thus, the decision to seek justice for abuse victims was both a question of justice and economics. Being able to address some instances of domestic violence at home, rather than involve the court, is reportedly encouraging women to increase their use of community policing/local mediation to address this. As one member of a women’s organization stated: ‘Female applicants are far more willing to come to us for justice than in earlier days. They do not remain in the home when they may be beaten or raped by males’7 Whilst the interpretation of crime figures is difficult, the perception of our interviewees was that community policing has certainly increased trust, which has in turn resulted in increased reporting. Additionally, interviews with police and other leaders identified interconnected changes in and around Dire Dawa. Both police and local leaders identified reduced caseloads in the court, reduced incarceration, lessening fear of police, and improved coordination with community leaders as direct results from the implementation of community policing. Previously courts were swamped and the accused forced to wait in cells for months. The extension of community policing and the incorporation of customary approaches has meant that petty cases can be dealt with outside the court system. Similar reductions in caseloads were also reported by the police as community leaders not only handled some small cases, but provided conflict mediation. Mediation is important in a context where conflicts between pastoralists and farmers frequently lead to land disputes resulting in violence.8 Police and clan leaders worked with people to reduce the long-term impact of conflicts, through discouraging fighting with farm tools, for example. Clan leaders also created a rotating system of access to communal water to reduce tensions between farmers and pastoralists.9 Land is one issue that traditional leaders identified as being continually negotiated and where there is some resistance to police intervention. Disputes over land are extremely common and traditional leaders frequently play an important role in managing land for the community and presiding over disputes. One meeting with elders and clan leaders specifically centred around land disputes10 and revealed a dual responsibility: a government mandate for the police, but a need for community reconciliation and patience for long-term peace. Beyond land, the most difficult issues related to the experience of women and girls. Female genital mutilation (FGM), child marriage and child abduction were three issues mentioned in multiple interviews.11 Despite the ongoing difficulty of under-reporting, several individuals commented that the rate of reporting instances of domestic violence to police or local mediators had increased.12 Certainly, local advocacy NGOs and women interviewed all recognized that improvements had taken place, although, despite these gains, interviewees identified gender issues as a major continuing issue because it was so deep-rooted. Impetus for these police reforms came from the highest levels of the Ethiopian government and resistance has been relatively limited in Dire Dawa. Clan leaders do occasionally attempt to settle violent crimes without notifying the police, but those interviewed insisted such behaviour was becoming increasingly rare and typically limited to rural areas.13 The way community policing had been induced appeared to have placated local chiefs in terms of losing influence. The idea that they were still responsible for minor offences and took part in detailed consultations with the police, reassures them that they were still part of the security infrastructure. Where they seemed to have given ground was certainly in their management of social hierarchies, particular gender relations and this has gone from a system where almost nothing was reported, to one where the police deal directly with certain traditional practices. With FGM, for example, chiefs have largely stepped back from involvement, recognizing that resistance is futile. Where there have been disputes over this, the police are also quick to invoke the constitution and this seems to have an overarching positive effect on the incorporation of local communities. What has certainly happened is a shift in perception of the police by local communities, summarized by one member of an Advisory Council: ‘Police in the previous days were distant from people. Contemporary policing … cannot discharge all its responsibilities unless it obtains support from the public. Both parties in community policing attempt to build trust to collaboratively work against crime and antisocial behaviours.’ This would suggest that there has been buy-in from the local community in to the idea of improved information flows between the community and the police in tackling crime. The broad approach to policing has meant that community members can use their own accepted means of mediation and conflict resolution if they wish. It is also possible for community police officers to refer individuals to non-state conflict resolution groups. This is not really seen as a means of avoiding involvement of the police, but as the police being respectful towards the role of traditional institutions. It is also notable how the top-down nature of the relationship works here. Whilst there may be traditional rules for specific offences, the constitution overrules most of them: ‘…neighbourhood communities cooperate in passing information to police when they observe criminals in their surroundings. However, harmful practices, such as those oppressing women that are now against the constitution, still exist in those communities.’14 In other words, the Ethiopian Constitution has banned such practices so the only remaining task is to educate those failing to comply. As a rural chief told us, ‘…we developed the program to educate people to enhance their level of consciousness about the constitution of the country’. The overriding understanding of community policing both publicly at high levels and also within the community at kebele level is the ideology of state development and of community policing as an integral part of that strategy. This includes the practice of further spreading state influence in to areas where it has previously been patchy. As one community police officer stated: ‘Before the Police Commission opened an office at this rural village, the community was unaware of what “police” meant. They do not recognize that the police have legitimate power to arrest criminal offenders. The community was resistant [to an external imposition of police]. However, the community did not take long time to understand the value of the police and their presence in creating a safe and secure environment.’ The ideology of community and police being part of an overarching development strategy is present in several of the interviews. These issues were raised in conjunction with ‘taking the constitution to the people’ and overriding local practices. As one police chief stated: ‘The main problem here is … [that some decisions of the elders] … are compliant with the constitution of the country, while some others are designed to regulate the social order of the area. For example, when a homicide happens, elders attempt to settle the cases through traditional means, whereas homicide should only be subject to the formal, state judicial system. These are cases where the elders do not have a mandate and need to be placed under the court procedures. To avoid these misunderstandings, we developed a program to educate people to enhance their level of consciousness about the constitution of the country.’ The police, therefore, at least partly see community policing as spreading a judicial system in to community areas and using the constitution as the overarching vehicle for doing so. The power of the constitution is such that this was frequently cited by traditional leaders as a reason why they had changed their behaviour. The law seems to be winning. Hybridity or Assimilation? The top-down approach to community policing is an attempt to overcome some of the legacies of brutal regimes that used the police as a tool of oppression. Whilst labelling the approach as ‘top down’ implies a number of characteristics, it does represent a radically different approach to the past, not least in using persuasion and partnership rather than force. It also involves a wide network of actors collaborating to make it work. The initial decision to adopt community policing came from the national Ethiopian government. In each district, the police commander and their subordinates are responsible for implementing the reforms. With regional commanders onboard, it then became a process of convincing clan leaders, and creating community advisory councils in urban areas to assist in ushering in the desired change. Communities are now required to nominate individuals to be officers in their area which places the vetting demands on the community before an individual can even begin training at the police academy. This networked approach also requires a shared philosophy and set of aims that we explored during the research. For the police, the main driver is one of using decentralization as a means of reducing local crime and increasing control. When asked about their aim for their community, officers repeatedly returned to the idea of a ‘space without crime’, or, as one police inspector stated: ‘to see an area where crime is less and [make] the neighborhood community free from the fear of crime’.15 Similarly, one community advisory board member said ‘our vision is to see an area not affected by crime and suitable to the [movement of people] without fear of crime’.16 A separate community advisory board member articulated a very comprehensive answer: ‘We need to see our country at peace. Our vision is to create a community sensible to peace and enjoying security of living environment. We need to ensure peace and reducing the congestions of the justice system is one of our major aims. This will be sustained by creating a number of platforms where people can freely discuss issues related to creating crime free areas and also social and economic development.’17 A youth committee interviewed echoed these notions. Taken together, a particular picture of policing in Ethiopia begins to emerge. The top-down reforms were adopted quickly, and though the intention of decentralizing the structure of police to better reflect local needs, both local police and community advisory members mentioned that their actions were because it was written ‘in the constitution’.18 Clan leaders and community members regarded changes to gender-based issues as issues of the past because the constitution resolved them. The view of police reform, emerging from both police and members of the community was that policing focusses not only on crime, but on constructing the good relationships necessary for a peaceful community. This was not always straightforward and met increased resistance in rural areas where clan leaders were more resistant to changes in their own power. Even here clan leaders work with the police if a crime occurs. If people are ‘…aggressive, their case will be directed to the police after leaders of the clan have exhausted all means of traditional problem solving approaches’.19 One clan leader said: ‘I can say that the work of the clan leaders is almost the work of the government. Community policing in this village is organized by Clan leaders, the police and the community. The police teach people not to commit crime and inform the police when they observe suspects.’20 A significant issue facing the police reforms is the general view of women in society. This is a challenge because the Ethiopian constitution aims to have women filling 50% of government jobs, including the police, and generally improving their everyday lives.21 One proverb dominated the all discussions around gender during the research: “women are for the kitchen.” A local women’s association interviewed was both realistic about the struggle, but encouraged by the progress. They said: ‘…problems arising from culture still exist. For example, FGM was dates from Ferron22 times and is religiously accepted. We struggled with feminist thinking and have tried to mainstream ideas of social and gender equality across the community.’23 This women’s association found encouragement through a general reduction in early marriage, married women continuing their education and increasing women’s literacy. Any progress made has been achieved through dialogue with others, often in the context of traditional coffee ceremonies. This is a space where community police officers are particularly useful because they are embedded in the community and see gender equality and awareness as an important part of their job.24 Gender-related crimes in particular were described as being ‘un-seen’ in previous regimes but reported now, which may indicate an increased confidence in the system and the police more generally, although this may be difficult to interpret. Certainly, women’s groups recognized that gender-related crime remained a problem, but were also reduced and increasingly regarded as unacceptable. It is interesting to note that the networked and broad leadership outlined above contains both enablers and obstacles to reform. The network at a local level is described by a sergeant in Village Six as follows: ‘Traditional leadership has its own advantages and disadvantages. Ethnic conflict is common in Dire Dawa and the police are in charge of stopping the conflict. The main method the police have been using is through traditional conflict resolution by elders, working with the police in mediation. Participation of social institutions such as churches, mosques and other associations formed to extend help to each other during funerals and marriages may also be involved in the mediation process. Conflicts between the dominant ethnic groups – the Oromo and Somali - have been solved using all these approaches.’25 Such an approach requires four main pillars. Firstly, both sets of leaders within the police and community have to recognize that they have important roles to play in this. Such disputes cannot be resolved by either leadership set alone. This leads on to collaborative leadership based on trust. Secondly, this mutual partnership relies heavily on trust, both between the parties but also between the leaders and the community. Thirdly, the alliance between the police and community leaders is one of establishing legitimacy in resolving the dispute as well as a recognition that these conflicts relay on underlying issues that cannot be resolved by solving a case in a formal police enquiry or a court. Finally, what it also illustrates is some form of shared vision of why policing is important. What our fieldwork shows clearly is that shared vision comes partly from the legitimacy of the state as embodied in the constitution, and strongly identifying reduced crime with ‘development’, but also from the shared belief that reducing crime is a public good. This raises interesting issues around the process itself and whether this is actually community policing, or an attempt by the state to use community systems as a multiplier to reduce crime. Certainly it is clear that community policing in Ethiopia more generally cannot be divorced from the political-economic context within which it takes place (Denney and Kassaye 2013). There is clearly an element of that in the approach of the Ethiopian state, although that would be reducing the initiative in too reductionist a way. What does not seem to be emerging is a separate and completely new hybrid system, rather the development of a partnership approach whereby the role of the police is enhanced at community level through increased collaboration and intelligence and also the use of customary methods to deal with low level crime. The end result supports Baker’s (2013) proposal that the community policing model is really about incorporation in to the state itself rather than the creation of a new system. The pattern of police engagement is also enlightening in this respect and implies that people don’t follow the constitution because they know no better and therefore need to be educated. It also implies that if local authorities do not obey the law, then they will be educated: the law wins. Conclusion This paper has argued that community policing in Dire Dawa is certainly a success in terms of perceptions of criminality. Its origins in a top-down approach have produced side effects that mean the system is not really a fully separate and new hybrid, but more of an extension of police power through operating alongside existing customary systems managed by traditional authorities. At the same time, the system has also managed to expand police power and influence through cooperation rather than through force, which in the context of Ethiopia is something of a novelty in itself. We partly agree with Baker (2013) who argues that the Ethiopian case is not simply a community policing model, but a more comprehensive system where there is not only an attempt to manage non-state systems, but a recognition and transfer of jurisdiction to them. In Dire Dawa, the police refer cases through the local court systems and police practice has altered as a result. Much of this is done through cooperation. However, it is arguable that the Dire Dawa model is not necessarily a hybrid system but has some similarities with other donor-led police reforms where community policing is about gathering intelligence to support the police, or the development of complex systems designed to extend central control as seen in Kenya (Ruteere and Pomerolle 2003) and in Rwanda (Purdekova 2011). The Ethiopian approach can be said to be distinct in terms of accepting some principles of local justice, the approach to partnership, and that the community police are from and live within the communities themselves—it is an embedded approach (Baker 2013). The research also implies that the system relies on a complex web of leadership and agents who are constantly negotiating, developing and resisting. This requires mediation between the formal state policing backed by the developmental ideology of the government and the local communities that have been locked in a localised struggle for authority over decades. Community policing in this sense is a means of reducing friction between these two distinct spheres of juridical activity, using the community police themselves as a bridge between the two spheres. Moreover, the use of the constitution as an overarching framework is used by the police to assert the authority of the state over local justice systems where those systems are said to have encroached on to the legal ground occupied by formal institutions. This is primarily about serious cases of murder, rape or equivalent where the state will assert authority over local institutions even when those institutions feel that they should deal with those cases. Our research implies that there has been some conflict with regard to some of these cases with some local officials feeling that they should be responsible for all justice, but where the state has then stepped in. One of the biggest areas of change has been in gender-based crime, where the evidence shows a clear increase in reporting of crime to formal authorities. Respondents indicate that sexual and gender-based crime was previously hidden, which would support the findings of earlier surveys that suspected under-reporting within a generally conservative attitude to women where beating was commonly accepted (Burton et al. 2011).26 The same survey also reported that more than half of the women surveyed felt that there was no-one they could talk to about domestic violence. The development of a highly publicized gender-related legal framework and training within the police appears to have emboldened women in terms of going to the authorities to report violence against them. Evidence from local women’s groups as well as individual interviewees suggests that the community police are able to mediate and approach this issue far more than before and, at the same time, given that the local authorities rarely intervened in a case of domestic violence—reflecting the attitude that the male knows best within the household—this seems to be an area where local authority is happy to accept the community police as legitimate under the umbrella of the constitution. The centrally led community policing programme is at least partly an exercise in recognising the benefits of working together rather than attempting a wholesale replacement of a local legal system perceived as legitimate by local people. The interviews show that, even though reported crimes have increased, the increase itself is a result of fewer crimes and increased reporting partly due to improved relations between the formal and informal agents. However, whilst there is a recognition of the benefit on behalf of the police, there is also a hierarchy involved here. A very common theme within police interviews at the local level is ‘they do not understand therefore we had to educate them’, reflecting a modernising philosophy deriving from the overarching developmental approach of the Ethiopian state. The extensive use of the constitution as a framework has not only been successful but also indicates a process whereby local legal systems may in fact be incorporated in to formal, state approaches rather than forming a new, hybrid system of policing: there is a clear sense of direction towards state law here. Ethiopian community policing at least partly reflects the desire for the state to spread, but also recognition of the limitations of the state. This system of working together has had clear benefits for the population in terms of crime, or at least the perception of crime. Acknowledgements The authors would like to thank the two anonymous reviewers for their very constructive criticism of the paper. Footnotes 1 Interview with Tadele and Motuma, aged 22 and 18, Village 5, 12/3/2017. 2 The internal threats include active separatist movements for Afar, Gambela, Ogaden, Oromia and Sidama, or five out of the nine internal states of Ethiopia, whereas terrorist threats are frequently related to either Ethiopia’s Somali Region on the East and South, but also from al-Shabaab (Green et al. 2012). 3 Interview with Dire Dawa Police Commissioner, Dire Dawa, 15/3/2017. 4 We have already outlined the serious lack of reliable data in Ethiopia. This crime data was compiled and provided to the researchers by the Dire Dawa police commissioner. This information does not seem to be centralized in any meaningful way. 5 Out of a total population of 341,834 and total reported crimes of 16,978. 6 Interview with a Deputy Inspector in the Department of Women and Children Services, Gendekore, 13/3/2017 7 Interview with a Deputy Inspector in the Department of Women and Children Services, Gendekore, 13/3/2017. 8 Interview with Police Chief, Wahibia, 13/3/2017. 9 Interview with Clan Leader, Wahibia, 13/3/2017. 10 Interestingly there was also discussion of FGM, which these leaders all said had been ‘un-seen before the police came along, but was being dealt with in a positive fashion. There was recognition that women could play an important part in the community. 11 Interview with representatives from Dire Dawa Women’s Association Interview, Dire Dawa, 14/3/2017. 12 Interview with Officers in Women and Children’s Department, Dire Dawa, 12/3/2017. 13 Interview with Dire Dawa Police Commissioner, Dire Dawa, 15/3/2017. 14 Interview with Police Chief, Police Land Station, 9/3/2017. 15 Interview with Police Officer, Village Six, 13/3/2017. 16 Interview with Community Advisory Council Member, Amese, 11/3/2017. 17 Interview with an Orthodox Priest and Community Advisory Council Member, Village Six, 13/3/2017 18 Interview with Dire Dawa Police Commissioner, Dire Dawa, 15/3/2017. 19 Interview with Chief of Police, Wahibia, 11/3/2017. 20 Interview with Clan Leader, Wahibia, 11/3/2017. 21 VAW Report 2013: 6, Interview with Youth Association leaders, Dire Dawa, 13/3/2017. 22 This word caused significant debate since none of us knew its meaning. In the end we discovered that it is a reference to ‘Pharaonic’ meaning that this was an ancient practice. 23 Interview with Police Commissioner, Dire Dawa, 15/3/2017. 24 Interview with Police Commissioner, Dire Dawa, 15/3/2017. 25 Interview with Police Sergeant, Village Six, Dire Dawa, 16/3/2017 26 This report also contains significant baseline data from the Ethiopia Gender Survey that compared under-reporting of gender violence, and to a lesser extent early marriage, with a comprehensive attitudinal survey. References Abrahams , R . ( 1998 ), Vigilant Citizens: Vigilantism and the State . Polity Press . Albrecht , P. , Olushegu , G. , Gibson , A. and Thomas , S . ( 2014 ), ‘ Community Policing in Sierra Leone: Local Police Partnership Boards ’. DIIS Report 16. Danish Institute for International Studies . Autesserre , S . ( 2010 ), The Trouble with the Congo: Local Violence and the Failure of International Peacekeeping . Cambridge University Press . Baker , B . ( 2009 ), ‘ A Policing Partnership for Post-War Africa? Lessons from Liberia and South Sudan ’, Policing and Society , 19 : 372 – 89 . Google Scholar Crossref Search ADS —— ( 2013 ), ‘ Hybridity in Policing: the Case of Ethiopia ’, The Journal of Legal Pluralism and Unofficial Law , 45 : 296 – 313 . Crossref Search ADS Baker , B. and Scheye , E . ( 2007 ), ‘ Multi-Layered Justice and Security Delivery in Post-Conflict and Fragile States ’, Conflict, Security and Development , 7 : 503 – 28 . Google Scholar Crossref Search ADS Bierschenk , T . ( 2017 ), ‘ Who are the Police in Africa ?’, in J. Beek , M. Göpfert , O. Owen and J. Steinberg , eds., Police in Africa: The Street Level View . Hurst and Co . Boege , V. , Brown , A. , Clements , K. and Nolan , A . ( 2008 ), On Hybrid Political Orders and Emerging States: State Formation in the Context of ‘Fragility’ . Berghof Center . Burton , P. , Artz , L. , Teferra , S. and Upton , J . ( 2011 ), ‘Scoping Study: Violence Against Women and Girls in Ethiopia, Centre for Justice and Crime Prevention’ , Unpublished report prepared for DFID Ethiopia (confidential). Chabal , P. and Deloz , J . ( 1999 ), Africa Works: Disorder as Political Instrument . James Currey . Cooper-Knock , S.-J . ( 2014 ), ‘ Policing in Intimate Crowds: Moving Beyond “the Mob” in South Africa ’, African Affairs , 113/453 : 563 – 82 . Google Scholar Crossref Search ADS Cooper-Knock , S.-J. and Owen , O . ( 2015 ), ‘ Between Vigilantism and Bureaucracy: Improving our Understanding of Police Work in Nigeria and South Africa ’, Theoretical Criminology , 19 : 355 – 75 . Google Scholar Crossref Search ADS Cross , C . ( 2016 ), ‘ Ulinzi Shirikishi: Popular Experiences of Hybrid Security Governance in Tanzania ’ Development and Change , 47 : 1102 – 24 . Google Scholar Crossref Search ADS Denney , L. and Kassaye , D . ( 2013 ), Securing Communities for Development: Community Policing in Ethiopia’s Amhara National Regional State . Overseas Development Institute . Di Nunzio , M . ( 2014 ), ‘ Thugs, Spies and Vigilantes: Community Policing and Street Politics in Inner City Addis Ababa ’, Africa , 84 : 444 – 65 . Google Scholar Crossref Search ADS Goodfellow , T. and Lindemann , S . ( 2013 ), ‘ The Clash of Institutions: Traditional Authority, Conflict and the Failure of “Hybridity” in Buganda ’, Commonwealth and Comparative Politics , 51 : 3 – 26 . Google Scholar Crossref Search ADS Green , O. , Baker , B. , Horn , A. , Kebede , E. , Jennings , M. and Othieno , T . ( 2012 ), ‘Security and Justice Provision in Ethiopia’ , Report commissioned for DFID , 7 May (Unpublished and confidential). Greene , O. , Sansculotte-Greenidge , K. , Murithi , T. and Dagne , K . ( 2009 ), ‘Study on the Security, Justice and Dispute Resolution Needs of the Poor in Ethiopia’ , Draft Final Report 6 March 2009. University of Bradford . Hesselbein , G. , Golooba-Mutebi , F. and Putzel , J . ( 2006 ), Economic and Political Foundations of State Making in Africa: Understanding State Reconstruction . LSE Crisis States Research Centre , Crisis States Working Article Series, 3.2. Hills , A . ( 2014 ), ‘ Partnership Policing: Is it Relevant in Kano, Nigeria ?’ Criminology and Criminal Justice , 14 : 8 – 24 . Google Scholar Crossref Search ADS Jackson , P . ( 2011 ), ‘ Decentralised Authorities, Decentralised Power and Who gets Access to Justice in Sierra Leone ’, Journal of Legal Pluralism , 63 : 207 – 31 . Google Scholar Crossref Search ADS Kyed , H . ( 2010 ), The Contested Role of Community Policing: “New” Non-State Actors in the Plural Landscape of Mozambique , DIIS Working Paper 26. Danish Institute for International Studies . Luckham , R. and Kirk , T . ( 2012 ), Security in Hybrid Political Contexts: An End-User Approach . The Justice and Security Research Programme, LSE . MacGinty , R . ( 2010 ), ‘ Hybrid Peace: The Interaction Between Top-Down and Bottom-Up Peace ’, Security Dialogue , 41 : 391 – 412 . Google Scholar Crossref Search ADS Meagher , K . ( 2012 ), ‘ The Strength of Weak States? Non-State Security Forces and Hybrid Governance in Africa ’, Development and Change , 43 : 1073 – 101 . Google Scholar Crossref Search ADS Menkhaus , K . ( 2006 ), ‘ The Rise of a Mediated State in Northern Kenya: The Wajir Story and its Implications for State-Building ’, Afrika Focus , 21 : 23 – 38 . Pouligny , B . ( 2006 ), Peace Operations Seen from Below: UN Missions and Local People . Kumarian Press . Purdekova , A . ( 2011 ), ‘ “Even if I am not Here, There are so Many Eyes”: Surveillance and State Reach in Rwanda ’, Journal of Modern African Studies , 49 : 475 – 97 . Google Scholar Crossref Search ADS Richmond , O . ( 2011 ), A Post-Liberal Peace . Routledge . Ruteere , M. and Pomerolle , M.-E . ( 2003 ), ‘ Democratizing Security or Decentralizing Repression? The Ambiguities of Community Policing in Kenya ’, African Affairs , 102 : 587 – 604 . Google Scholar Crossref Search ADS Schroeder , U. and Chappuis , F . ( 2014 ), ‘ New Perspectives on Security Sector Reform: The Role of Local Agency and Domestic Politics ’, International Peacekeeping , 21 : 133 – 48 . Google Scholar Crossref Search ADS Tapscott , R . ( 2016 ), ‘ Where the Wild Things are Not: Crime Preventers and the 2016 Ugandan Elections ’, Journal of Eastern African Studies 10 : 693 – 712 . Google Scholar Crossref Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). All rights reserved. For permissions, please e-mail: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Steffensmeier,, Darrell;Lu,, Yunmei;Kumar,, Sumit

doi: 10.1093/bjc/azy011pmid: N/A

Abstract In this study, we used age–crime statistics from India to investigate the two core tenets of Hirschi and Gottfredson’s (HG) invariance thesis that the age distribution of crime is always and everywhere adolescent spiked and thereafter declines continuously into elder ages. Besides comparisons to the inverted J-shaped distribution projected by HG, we further compared Indian age–crime patterns with those in United States (Western nation; main evidential source for HG invariance projection) and Taiwan (like India, a non-Western collectivist society). Findings suggest considerable divergence in India’s age–crime patterns compared with HG invariance norm and US age–crime distributions, but overall similarity with Taiwan’s age–crime schedules. Implications for research and theory on the age–crime relation more broadly are discussed. Introduction In this research, we used age–crime data from India to investigate the influential thesis of Hirschi and Gottfredson (HG) that always and everywhere crime emerges in adolescence, increases sharply to a peak in mid-to-late adolescence and then declines rapidly or slowly for the rest of the life course.1 Acceptance of the invariance position has led many criminologists and social-behavioural scientists to suppose that a large portion of adolescent crime is a naturally evolved consequence of teen biology or highly symptomatic of members of that age group who are unavoidably rebellious or reckless. Some recent biosocial critiques of sociological explanations of crime in relation to ageing and the life course take as the starting point a ubiquitous age–crime curve (Barnes et al. 2015). Surely, contends biosocial criminologist Anthony Walsh (2009: 156), if social influences shorn of evolved developmental patterns were what really mattered, ‘we would have accounts of some cultures and sometimes when the age–crime curve did not exist, or was even reversed, but we do not’. Rather than social influences, a better explanation for the ostensibly universal and much higher teen rates of crime is to be sought in underlying hormonal and neurological changes that accompany adolescence. Walsh writes Age is an index of a series of developmental stages from prenatal to senescence that we all go through if we live long enough with each having its own characteristic physiology. There must be something special requiring its own explanations going on during adolescence that dramatically, albeit temporarily, greatly increases the probability of antisocial behavior (2009:156, emphasis ours). Embracing this representation and the HG invariance thesis, a new genre of biological and neuro-developmental theoretical approaches have appeared that attribute the projected adolescent spike in crime to pre-programmed hormonal or cerebral changes over the adolescent to young adulthood period. Kanazawa and Still (2000), for example, posited an evolutionary explanation, whereas Ellis et al. (2012) proposed a hormonal theory. Many scholars today have argued for a strong cerebral or neural basis, suggesting that adolescents, as compared with adults, have different brain maturation timetables or circuitry so that adolescents are considerably more programmed for sensation-seeking and involvement in risky behaviours, such as crime. Of particular prominence is Steinberg’s ‘dual systems’ model of adolescent neural development that characterizes the adolescent spike in crime as a normative by-product of an imbalance in incentive (sensation seeking, reward) and inhibitory (cognitive control) circuitries that sets the stage for suboptimal decision-making and risk-taking, including crime (Steinberg 2008). The imbalance or temporal gap in maturation timetables of two different brain systems is posited as the root of the ubiquitous peaking of crime at around ages 15–17 (projected peak is age 17). Other analysts, however, consider the matter of an invariant age–crime relation to be unsettled (e.g. Epstein 2007; Greenberg 2008; Steffensmeier et al. 2017). First, these analysts acknowledge that crime tends to decline with age, but contend that important variation exists between countries or historical periods, and/or between crime types, in the parameters of the age–crime curve (e.g. peak age, rate of decline from peak ages) and in the proportionate amount of total crime that youth account for (Steffensmeier et al. 1989; Tittle and Grasmick 1997; Greenberg 2008). Second, they observe that current empirical and theoretical understanding of the age–crime relation and the extent to which it is adolescent driven is based almost entirely on the United States or a few similar Western societies, where age-specific crime data are more readily available and where age–crime schedules do typically rise to a peak in adolescence and then decline rapidly or slowly. Notably, HG’s support for invariance in the age–crime curve is based largely on age–crime plots derived from: (1) 19th century England and Wales magistrate-court data and (2) 1970s US arrest data. Relying on data from western societies has largely neglected both the theoretical perspectives and the empirical research from non-Western societies and it is problematic on several grounds (Aas 2012; Connell 2006). Our study of the relation between age and crime in India, therefore, is important and distinctive for addressing what is most missing from the scholarship evaluating the premise of a ubiquitous teen-driven age–crime schedule—the scarcity of international or cross-cultural evidence across diverse cultural context. India is a uniquely suited site for evaluating the universality assumption because their age–crime statistics are among the best available among non-Western nations and because India represents a ‘collectivist-hierarchical’ Asian cultural template as compared to an ‘individualistic-egalitarian’ Western culture, such as characterizes the United States. The considerable impact of a collectivist versus individualist cultural syndrome on a variety of societal phenomena (e.g. business practices, child rearing, policing, schooling [for reviews, see: Bond 2010; Hofstede et al. 2010]) suggests that notable differences also might be found in India’s age–crime schedules as compared to the invariance norm. Prominent also is a recent study showing divergence in age–crime patterns in the collectivist nation of Taiwan as compared to the reverted J-shaped distribution projected by HG (Steffensmeier et al. 2017), suggesting that the age–crime association may differ in collectivist as compared to individualistic societies. Along with the issue of ubiquity, application of cross-cultural or international data to the debate about the age–crime invariance issue is especially crucial because there is no reason to believe that biological factors implicated in determining age–crime patterns will vary across countries (Mannheim 1965). The degree to which age differences in crime result from biological or fixed developmental factors implies that there should be no international variation in these differences. It seems particularly important for researchers to include non-Western societies in assessments of whether there is cross-national variation in the age–crime relation and extent of youth crime. Cross-national data and analyses also provide a basis for investigating societal factors that might produce variation in age–crime patterns. Prominent criminologist Braithwaite (2015: 183; see also Aas 2012; Stamatel 2014) has warned that cross-national criminological studies are needed to ‘provide lenses through which we can see a need to revise explanatory theories of crime from those dished up by western criminal law and criminology’. Besides comparisons to the HG invariance norm, we also compare India age–crime patterns to those in the United States and Taiwan. As noted previously, US age–crime patterns have been the prominent basis for the projection of an invariant adolescent-spiked age–crime relation and the findings from collectivist non-Western Taiwan seem strongly at odds with the invariance projection. We proceed, first, by outlining how social and cultural properties of nations or collectivities might create both constancy and divergence in the age distribution of crime and, second, by considering how central structural and cultural elements of the United States and Western societies seem to contribute to relatively high rates of youth crime, and then reviewing prior research that proffers some supporting evidence for country-level variation in the age–crime relation. Next, we discuss India as a strategic research site and why we might find differences in age–crime schedules between India and the HG invariance standard. We then discuss the data and methods and present the results of our analysis. Sociocultural production of constancy and variability in age–crime relation There are obviously biological dimensions to the age–crime relationship, since ageing itself is a biological, neurological, psychological and social process. Physical and neuropsychological development and ageing (especially in the early life course) set the parameters of possibility and limitation for behaviour, including criminal behaviour. The physical foundation of ageing has long been seen as relevant to the age–crime connection by criminologists (Lemert 1951; Steffensmeier and Allan 2000). However, the notions that the age–crime curve may vary across cultures and time, and that social contexts matter in shaping the age–crime relationship, draws from prominent strands of sociological and criminological theorizing and research. One consistent theme that runs across these writings and research is that ‘older’ age–crime distributions (less in way of spiked adolescent rates) are likely to be associated with societies, cultures or historical periods in which youth have greater access to conventional opportunities and integration into adult society; and/or population groups for whom generational isolation and normative leeway do not markedly change during the youth-to-adulthood transition. First, some consistency in age differences in crime across space and time could instead indicate that age socialization and age-graded norms are markedly constant across times and settings for reasons that are socially practical and only indirectly biological. For any society to thrive, elder carriers of institutions must socialize youth to become productive members of the groups that fill structural roles, and they must also ensure an adequate level of conformity among post-youth age cohorts (Parsons 1951). One would expect that such socialization and pressures for conformity inevitably would be problematic and incomplete for some youth, but such pressures would increase with age. Although aggregate rates of crime may tend to peak in youth and then decline with age, collective variation in the parameters of age–crime distributions and even the contours of its shape also is expected (e.g. peak age, median age, rate of decline from peak age) due to age-graded processes of socialization and social control that differ over time and across cultures or places (Mannheim 1965; Greenberg 1985). Second, as based on information about the United States and other western societies, important sociocultural and temporal variations in the age–crime curve are projected by major sociological and criminological theories of crime. Key tenets of such theories emphasize the likelihood that cultural or structural properties of collectivities (e.g. nations, time periods) may differ in terms of: (1) integration and bonding of youth into adult society (e.g. Schlegel and Barry 1991); (2) normative leeway and exposure to youth peer subcultures (social learning theory via; Akers 2009); (3) status frustration and stress in youth lives, including economic or monetary pressures to purchase consumer goods (anomie and strain theories via Agnew 2006); or (4) the connection of youth to non-economic, normative institutions (Lemert 1951; Clinard and Abbott 1973). These factors all, in turn, are theorized to contribute to variability in youthful offending. In short, the biological facts of development and ageing are obviously important, but they occur within social structures, cultures and time periods, which shape and channel the life course (especially youth) in different ways and give different forms to the patterning of criminal behaviour across age groups. Third, drawing on these theoretical themes, along with recognizing the potential influence of biological factors (e.g. changes in physical capability; sexual interest), sociologists historically have attributed the spiking of rates of adolescent crime in the United States and Western nations to their isolation from adult society, socialization for individuation and autonomy, heavy reliance on their peers for approval, ambivalent tolerance and cultural support for teenage nonconformity, and lack of funds to support leisure-time activities (Lemert 1951; Clinard and Abbott 1973; Steffensmeier and Allan 2000; see also Moffitt 1993). In turn, the rapid decline in crime in the late teens has been attributed to rapidly occurring changes in how the ‘maturing’ adolescent is treated by adult society and what they expect of themselves. In the abrupt transition to young adulthood upon graduating from high school, they typically leave old crime-reinforcing peer groups behind, quickly acquire adult privileges and greater financial resources, and make crucial life-course decisions and forge mainstream institutional ties by leaving for college, joining the military or landing a job and perhaps forming serious romantic relationships, marrying, having children and so on (Steffensmeier et al. 1989; Warr 1993). Along with self-other expectations of it is time to ‘grow up’ and ‘settle down’ (Steffensmeier et al. 2000), these role transitions increase the stakes for conformity and orientation into conventional society at the same time that legal sanctions increase substantially. Social processes also might account for continual declines. Offenders (or would-be offenders) gradually may learn that gains from crime are not worth the risk or effort and that with age they have more to lose (Steffensmeier and Allan 2000). As they age, moreover, offenders may lose suitable co-offenders as partners are incarcerated, become physically incapacitated or die. All of these factors, obviously, are likely to be accompanied by biological and psychological processes of ageing such as physical deterioration or the realization that time is a diminishing and increasingly valuable resource. As well, an important (often unrecognized) contributor to the post-youth decline in crime is a continuous attrition in the pool of high-risk offenders who face high odds of injury, death or incarceration. Fourth, there are important slices of comparative-historical evidence that do show variation in the parameters of the age–crime curve depending on social context. For example, studies of US age–crime patterns covering a longer time period (e.g. pre- and post-World II period from1940 to 1980) show that compared with earlier eras, the modal or peak crime age of youths in the United States is younger in later time periods, the rate of decline from the mode is faster and there is considerable variation in age–crime curves by the type of crime. Violent crimes seem to peak at later ages and decline more slowly than in the case for property crimes (Steffensmeier et al. 1989). Similar age–crime trends over time in the direction of earlier peaks and more youthful distributions were observed for England (Farrington 1986). Several age-period-cohort studies using data covering the post-World II period also documented that more recent birth cohorts are more delinquent in the teenage years than previous ones (O’Brien et al. 1999). Some research, moreover, shows within-country variation in the age–crime association depending on racial or ethnic groups (Jennissen 2014). In addition, research on pre-industrial societies (based on Human Relations Files) shows historical variation in the age distribution for crime across societies, and even shows that adult involvement in antisocial behaviour (e.g. violence) actually is higher than among adolescents in a sizable number of societies (Schlegel and Barry 1991). Also, a recent study of the relationship between age and crime in Taiwan found considerable differences not only in age–crime parameters (e.g. peak age), but also in the shape of the curve (e.g. skewness), as compared to the adolescent spike and thereafter continuous decline observed in US age–crime schedules (Steffensmeier et al. 2017); notably, a much lower concentration of offending was found among the young in modern times in Taiwan. India as strategic research site India is a strategic international site for testing the age–crime invariance thesis because its hierarchical-interdependence cultural orientation may contribute to lower levels of crime both in general and among youth in particular, as compared to a highly individualistic-egalitarian Western culture like the United States or England that are the main basis for the invariance premise. Our treatment here focuses on differences in India and the United States or Western cultures and institutions in how youth experience adolescence and the transition to adulthood, which is at the centre of contemporary scholarly attention on the age–crime invariance issue. We draw, in particular, from the observations of prominent Indian social scientists who have written about crime patterns in India and about youth crime or delinquency in particular.2 Although some disagreement exists as to whether youth crime is rising in recent years (some say yes, others say no), there is broad consensus that delinquency levels—both currently and historically—are low, and low in particular when compared with Western nations like the United States. In light of greater familiarity with the United States (or Western) context and the novelty of India, we focus on the latter’s context.3 Located in southern Asia with a history dating as far back as 2,500 BCE, India came into existence as a nation state in 1947 following its independence from two centuries of British rule. The nation that emerged was composed from numerous provinces, kingdoms and independent principalities. Today, India is the world’s most populous and ethnically diverse democracy with more than a billion residents speaking some 22 different officially recognized languages. Though housing some of the world’s largest cities, over 800 million (70%) of India’s 1.2 billion people live in rural areas. Primarily a rural/agrarian society, the nation is rapidly industrializing, modernizing and Westernizing (Menon 2017). Following market-based economic reforms in 1991, India is now the world’s fifth largest economy with prominent geopolitical clout. Religion plays a vital role in the Indian way of life. Although over 80% of Indian population is Hindu, India is constitutionally a secular state. A melting pot of religions and faiths over centuries has resulted into a unique syncretic, pluralistic culture composed of now over 800 million Hindus, 140 million Muslims, 28 million Christians and significant numbers of Sikhs, Buddhist and Jains and other indigenous faiths (Census of India 2011). In a country as diverse as India, generalizations can be tricky but there are certain cultural values that are mostly uniform across populations, often in contrast with the western cultures. First, a deeply ingrained gestalt of family obligation and interdependence-dependence has always been the foundation of Indian society. Children and adolescents have always occupied a special place in Indian family system (Saraswathi 1999; Larson et al. 2003). Second, unless necessary (e.g. homelessness), financial or personal autonomy in adolescence is neither cherished nor encouraged and it is not uncommon for adolescents and those in their early adulthood to live with their parents or relatives for a considerably longer time than their counterparts in the western societies. Instead, interdependency, support and nurturance across the generations are unique values in Indian family systems. In general, control and supervision of children and adolescents is pervasive and deeply entrenched in the Indian culture. Although some large and pervasive changes in Indian culture and modes of organization have taken place in recent decades, traditional Indian practices and culture largely have prevailed (Saraswathi 1999; Menon 2017). Ancient customs exist side by side with the latest advances of civilization and science. Indians are immersed in a network of role relationships that involve a variety of obligations towards kin, religion, caste, village or community (Mandelbaum 1970; Kakar and Kakar 2007). The social system is organized around these ties and the personal loyalties and interdependency they demand. Compromise rather than conflict in all relationships is highly valued. When in conflict, communal rights is given precedence over individual rights. Preserving harmony, hierarchy and authority involves avoiding confrontation and in-group aggression. According to prominent Indian historian Kakar (1979: 104): ‘socialization practices are deeply embedded in social relations, particularly as is reflected in deep (lifelong) loyalty to family/kin and to select groups and individuals . . . as a link in the chain of generations and in the history of the race’. Research on adolescence in India has been growing in recent years. Key features of family life and Indian culture make the experience of adolescence quite distinctive from adolescence in many other parts of the world, particularly when compared with their American and European counterparts. Although the Indian social world is characterized by multiple religions and ethnic groups with their particular practices, there are many similarities with regard to the socialization of children and youth. The similarities include the following: There is less separation between adult responsibility and adolescence in India when compared with the United States and Western countries. The basic function of that stage in India is on learning, preparing and assuming of adult roles (Chaudhary and Sharma 2011). Parents are strongly involved in everyday lives of youth and for the rest of their lives. This involvement involves advice and assistance, as well as intense supervision and interference in children or youth’s decision-making (Sandhu 1987). Social climate is one of the close relationships between adolescents and parents (Hartjen 1982; Larson et al. 2003). There is little in way of generational isolation or segregation of youth from adult society, but instead children and adolescents form an integrated part of the collective family, religious and community activities (Kakar 1979; Kakar and Kakar 2007). Belief in lifelong commitment to the family in Indian culture fosters family cohesion and highly interdependent family–kinship relationships. The single most important socialization tenet that characterizes Indian adolescents is their constant pull towards the family ethos that encourages them to place individual needs secondary to family needs and subjugate their decisions to those made by the family to maintain cohesiveness (Chaudhary and Sharma 2011). There is less in the way of time spent with peers and less of a peer culture in India when compared with the United States (Verna et al. 2002). Indian adolescents spend much more time with their families, evidence much less in way of conflict with parents, and truly believe that their parents have their (the young persons’) best interest in mind (Larson et al. 2003). Self-reliance and engagement with age-mates, which more so is the central part of US teenagers’ life, is not a critical expectation on part of adolescents (Chaudhary and Sharma 2011). Hypotheses Although it is difficult to know exactly how these differences in cultural and institutional features operate, they do provide a basis for expecting some to moderately large differences in the age distribution of crime in India when compared with the HG premise of a sharply peaked adolescent-driven distribution. For Indian youth, in contrast to their Western counterparts, there is much less generational separation but instead greater integration with adult society; much less in way of a prominent peer youth culture, and much less normative leeway or ambivalent tolerance of deviant behaviour. Combined with our earlier review of cross-national empirical evidence on the age–crime relation showing some evidence contrary to the HG thesis, the pronounced differences we observed in the Indian cultural template imply that we might find older peak ages and a more evenly distributed age schedule for crime in India when compared with the United States, especially for violent crime. Instead of finding a peak and then sharp drop-off in crime in late teens, we may find that post-adolescent crime levels remain relatively stable or even increase. Data and method The main source of data for our analysis is age-specific arrest statistics from India that are used for two sets of comparison: (1) main analysis involving the comparison of Indian age–crime distributions with the inverted J-shaped invariance norm; (2) supplemental analysis involving comparisons of Indian age–crime distributions with the age–crime distributions for the United States and Taiwan. The data and methods applied for the main and the supplemental analyses overlap, but also differ somewhat in data sources and analysis. We highlight here the Indian and US age–crime data and review the Taiwan age–crime data below. Compiled annually by National Crime Records Bureau of India’s central government in Crime in India (CII thereafter), Indian arrest statistics are broken out by age groupings and provided for various offense types and for total arrests. Similar to the format of US Uniform Crime Reports (UCR hereafter), CII presents the data in tables showing the number of arrests in a given year by offense type and age category of arrestees. Age is coded in the CII data in grouped ages of: 15–17, 18–29, 30–44, 45–59, and ages 60 and above. In the UCR, age is coded as: individual ages 15–24, and in 5-year groupings for the remaining ages. We include the following crime categories available in both CII and UCR: homicide, assault, theft, drug violations and gambling, as well as total arrests (summary offense category, which includes all arrests for any crime). We also calculated a summary measure for property crime (sum of theft, burglary and robbery) and violent crime (sum of homicide, assault and rape). We used a combination of statistical procedures to study India’s age–crime patterns when compared with the invariance norm. First, we calculated age-specific rates using the appropriate population estimates and averaged them for the years 2000–2002 (Indian Statistical Yearbook; U.S. Census Bureau). The rates were calculated for multiple offense types for India and for the 1980 US burglary that is the basis for the HG invariance projection. Second, we applied interpolation to the five-year grouped rates for the 1980 UCR burglary (25–29 . . . 55–59) and for the age-grouped rates in India (15–17, 18–29, 30–44, 45–59) to create age categories to represent single years (see Steffensmeier et al. 2017).4 The interpolated rates are useful for establishing equivalency in age categories and applying added statistical tests (Shyrock and Siegel 1976). Third, we used the age-specific rates to calculate the percentage age involvement (PAI) of offending that measures the timing of offending across the life span. Representing the percentage of total arrests (e.g. for a particular offense) accounted for by each age group after adjusting for the group’s relative population size, PAI is an easily interpreted and demographically sensitive measure that provides a robust comparison of crime levels across age groups (e.g. between youth and adult rates) and easy identification of differences in age–crime distributions between countries and across offenses (Britt 1992). The PAI also was used to generate age–crime plots and calculate measures of the age–crime distribution (e.g. median age, peak age, skewness) to evaluate the invariance. The formula for calculating the PAI is PAIij= rij∑rij×100 where r is age-specific arrest rate, i age category and j is offense category. These multiple measures and tests help us to take into account some imprecision in HG in defining invariance in the age–crime distribution (e.g. whether peak is mid- or late teen years and whether decline from peak is rapid or gradual). Drawing from Britt’s influential treatment (1992), we consider two kinds of invariance—parameter invariance and shape or form invariance—that basically constitute ‘stringent’ versus ‘lenient’ interpretations of uniformity or divergence. Parameter invariance (‘stringent’ test) would require the same (or nearly so) mean peak or right-side skewness in offending across all population groups, whereas shape invariance (‘lenient’ test) would require only a general similarity in the age–crime curve across all populations. The general view in the scholarly writings is that invariance in underlying shape (general similarity) covers any age–crime distribution characterized by a peaking of crime in adolescence followed by a post-adolescent decline, but steepness of the up-slope or the down-slope can vary in small to higher degree (Steffensmeier et al. 1989; Tittle and Grasmick 1997). The databases have some limitations that involve potential concerns about country-level differences in the classification or definition of offenses and possible variability in the extent of crime underreporting, as arrests are not a direct indicator of crime (Young 2004). The first concern is less of an issue in this study, as we include indices of crime that represent the sum of offenses across a number of offense categories (e.g. total, violent, property). Notably, to minimize idiosyncratic age patterns for this or that individual offense, HG (1983) argue for the use of broad crime indexes, and specifically total or overall violent crime counts. And, we include homicide, a crime that is widely projected to be reliably reported. The matter of underreporting of crime, moreover, is less of an issue because the focus in this study is not a direct comparison of the magnitude or absolute level of age–crime rates, but rather a comparison of relative age differences in crime rates to the HG invariance norm. Arrest data are problematic for estimating country-level differences in absolute levels of crime, such as the magnitude of youth rates or adult rates, but provide reasonably robust indicators of between-country relative differences in levels of crime across age groupings. Scholars generally advocate for the suitability of international crime data for disaggregated analysis, such as age (Lynch and Pridemore 2011). Notably also, HG (1983) and others (e.g. Walsh 2009; Sweeten et al. 2013) have relied on arrest data in making the age–crime claim, and they also have strongly asserted that an invariant age–crime curve is remarkably resilient regardless of data source. Logically, tests of that claim also should rely on arrest data. Results ‘For all offenses, at all times, and in all places and for all races and both sexes, involvement in crime reaches its peak in the middle to late teens and then declines rapidly throughout life’ (Gottfredson and Hirschi 1990: 219). Our tasks were to investigate the HG position of a ubiquitous and remarkably robust age–crime relation in the context of India, and then to broaden the analysis with added comparisons of age–crime patterns observed in India with those in United States and Taiwan. We are concerned with the two main projections of the invariance thesis: (1) the spiking of crime in mid-to-late adolescence and (2) the continuous downward post-adolescence slope in crime following the peak. Since much of the discussion surrounding the invariance issue revolves around age–crime plots, we began by presenting the relevant plots as a starting point and then proceeded to examine statistics that further compare India age–crime distributions to the invariance norm which projects a rapid increase in criminal offending in adolescence is followed by a rapid decrease. Because we found considerable similarity in Indian age distributions for each of the violent crimes (homicide, rape, assault) and each of the property crimes (robbery, burglary, theft), for parsimony we rely mainly on the summary indexes (total, violent, property) for presentation of results. Figure 1 shows graphically the relation between age and arrests in India for three summary indices of crime (violent crime, property crime and total); for homicide, assault, theft, drug-law violations and gambling; and for the invariance norm. Recall that, we used UCR 1980 burglary to replicate the reverted J-shaped age–crime schedule prominently displayed in HG writings (1983: 558) and used it as a standard and compared its age patterning with the age distributions across India. This standard (early peak and then decline) characterizes what HG single out as representative of the age–crime relation more generally. There is strong consensus that homicide arrest information is reliably reported and HG advocate for the use of summary indices of age-specific offending, particularly total counts. As compared with the invariance norm, the plots for India suggest: that (relative to older age groupings) teen levels of crime in India are diminished for homicide, for the other offense types (e.g. assault, drug violations) and for all three indices; that crime rates are typically the highest in mid-to-late thirties and that fairly high crime rates in proportionate terms prevail well into the forties and fifties. The timing of violent crime as represented in the plots for India appears robustly at odds with the invariance hypothesis.5 Fig. 1 View largeDownload slide Age–arrest distributions by offense type for India vs. invariance norm. Note—Norm: age schedule that matches closely the reverted J-shaped age–crime distribution projected by Hirschi and Gottfredson (1983) as invariant across social and cultural conditions. PAI is defined as percentage of arrests represented by each age after adjusting for age composition of the population, as based on 15–59 age range. India violent index includes: homicide, rape and assault; property index includes: robbery, burglary and theft. Fig. 1 View largeDownload slide Age–arrest distributions by offense type for India vs. invariance norm. Note—Norm: age schedule that matches closely the reverted J-shaped age–crime distribution projected by Hirschi and Gottfredson (1983) as invariant across social and cultural conditions. PAI is defined as percentage of arrests represented by each age after adjusting for age composition of the population, as based on 15–59 age range. India violent index includes: homicide, rape and assault; property index includes: robbery, burglary and theft. To establish further whether the relation between age and crime is invariant, we next compare the age–crime schedules on a variety of measures. Table 1 shows summary measures (e.g. peak age, median age, skewness) derived from the age distributions shown in Figure 1. Besides the three indices and homicide, the measures display the timing of crime across age groups in India for a range of offense types (e.g. drug violations, gambling, theft and assault). Similar statistics are provided for the invariance norm (inverted J-curve pattern) where the peak age is at age 16 to 17, the median age is located at age 19 and the skewness measure shows the age curve as sharply peaked and strongly skewed to the right. In contrast, the peak age for India always is beyond the teen-age range, with crime typically continuing to rise well into the thirties. Notably, the skewness values are mostly negative or approach zero, strongly confirming that arrests are much less concentrated among adolescents (also, young adulthood) in India but instead display considerable spread and are more heavily concentrated in older age groups. Table 1 Timing of crime by type of offense in India vs. invariance norm Peak Median Skewnessb Index of dissimilarity (D)c (norm vs. India) HG Norma 16 19 1.82 N/A Homicide 37 37 −1.16 58.6 Assault 37 36 −1.05 57.2 Theft 23 33 −0.25 51.3 Drug 37 36 −1.01 57.0 Gambling 37 37 −1.42 58.8 Violent Index 37 36 −1.03 57.1 Property Index 23 33 −0.26 51.2 Total 37 37 −0.96 60.0 Peak Median Skewnessb Index of dissimilarity (D)c (norm vs. India) HG Norma 16 19 1.82 N/A Homicide 37 37 −1.16 58.6 Assault 37 36 −1.05 57.2 Theft 23 33 −0.25 51.3 Drug 37 36 −1.01 57.0 Gambling 37 37 −1.42 58.8 Violent Index 37 36 −1.03 57.1 Property Index 23 33 −0.26 51.2 Total 37 37 −0.96 60.0 aNorm: Age schedule that matches closely with the reverted J-shaped age–crime distribution projected by Hirschi and Gottfredson (1983) as invariant across social and cultural conditions. bSkewness value of zero (or near zero) indicates a symmetrical distribution. Positive value indicates the distribution is skewed to the right (i.e. ‘younger’ age curve), while a negative value indicates the distribution is skewed to the left (i.e. ‘older’ age curve). cD is interpreted as the percentage of arrests that would have to be redistributed among age groups to achieve perfect congruence between two age curves. The formula for D is: D=(1/2)∑|oij−bi| ⁠, where oij is the percentage of arrests for offence j in age category i of India, and bi is the percentage of arrests for HG norm arrests falling in age category i. The form or shape of age–crime distributions is defined as significantly different from each other if the D value is 15 or higher. View Large Table 1 Timing of crime by type of offense in India vs. invariance norm Peak Median Skewnessb Index of dissimilarity (D)c (norm vs. India) HG Norma 16 19 1.82 N/A Homicide 37 37 −1.16 58.6 Assault 37 36 −1.05 57.2 Theft 23 33 −0.25 51.3 Drug 37 36 −1.01 57.0 Gambling 37 37 −1.42 58.8 Violent Index 37 36 −1.03 57.1 Property Index 23 33 −0.26 51.2 Total 37 37 −0.96 60.0 Peak Median Skewnessb Index of dissimilarity (D)c (norm vs. India) HG Norma 16 19 1.82 N/A Homicide 37 37 −1.16 58.6 Assault 37 36 −1.05 57.2 Theft 23 33 −0.25 51.3 Drug 37 36 −1.01 57.0 Gambling 37 37 −1.42 58.8 Violent Index 37 36 −1.03 57.1 Property Index 23 33 −0.26 51.2 Total 37 37 −0.96 60.0 aNorm: Age schedule that matches closely with the reverted J-shaped age–crime distribution projected by Hirschi and Gottfredson (1983) as invariant across social and cultural conditions. bSkewness value of zero (or near zero) indicates a symmetrical distribution. Positive value indicates the distribution is skewed to the right (i.e. ‘younger’ age curve), while a negative value indicates the distribution is skewed to the left (i.e. ‘older’ age curve). cD is interpreted as the percentage of arrests that would have to be redistributed among age groups to achieve perfect congruence between two age curves. The formula for D is: D=(1/2)∑|oij−bi| ⁠, where oij is the percentage of arrests for offence j in age category i of India, and bi is the percentage of arrests for HG norm arrests falling in age category i. The form or shape of age–crime distributions is defined as significantly different from each other if the D value is 15 or higher. View Large Comparison of Indian age–crime patterns with those of United States and Taiwan Our findings so far suggest considerable divergence in India’s age–crime patterns as compared with the HG projection of an inverted J-shaped distribution. To better situate these findings in broader international context, we extended the analysis to provide comparisons between India’s age–arrest patterns with those in contemporary United States and Taiwan. HG relied heavily on US age–crime distributions as evidence in support for their invariance thesis, and the prominence more generally of US databases on crime in the criminological literature is widely recognized. Taiwan is a strategic comparison because it represents a collectivist non-Western society, as is India, and also because a recent analysis of Taiwan’s age–crime patterns found considerable divergence as compared to the invariance norm (Steffensmeier et al. 2017). Information on the age distribution of arrests in Taiwan is drawn from the Taiwan Criminal Investigation Bureau of Police Administration Agency (TCIB) with age coded as individual ages for 15–39 and ten-year groupings for the remaining ages (for details, see Steffensmeier et al. 2017). As each of the three countries has its age categorization schemes, we first standardized the age classifications by collapsing the detailed age-specific arrest counts in the United States and Taiwan databases to align with the broad age categories in India’s crime database (15–17, 18–29, 30–44, 45–59). Using the arrest counts in combination with the population estimates for each the four age categories (Indian Statistical Yearbook; Taiwan Statistical Yearbook, U.S. Census Bureau), we calculated age-grouped rates for each offense (or index) in each country. Considering the differences in the absolute crime level among these three countries, we standardized the age-grouped crime rates through a procedure proposed by Blumstein and Wallman (2006).6 Using the US rates as the referent, we first calculate the standardized ratio with the following formula: Ratiojk=Rj (k=US)Rjk where R is total arrest rates (age 15–59), j offense category, k is country (Taiwan or India). Next, we calculated the standardized age-specific rates by multiplying the original age-specific rates by the ratio. Standardized Rateijk=rijk*Ratiojk where r is an age-specific arrest rate, i age category, j offense category, k is country (Taiwan or India). Application of these standardization procedures adjusts for the absolute differences in age-specific rates between the three countries and establishes equivalence for comparing their age–crime distributions. For parsimony and space constraints, we focus on the summary indices of total arrests, violent-crime arrests (homicide, rape, assault) and property-crime arrests (theft, robbery, burglary). Figure 2 presents histograms that compare the three countries as based on the four age–crime groupings: 15–17, 18–29, 30–44 and 45–59. The findings are straightforward in showing—first, that India and Taiwan age–crime patterns are older and flatter than US patterns. Second, on a continuum moving from older to younger distributions, Indian age–crime distributions are the oldest, whereas US age–crime curves stand out as being most strongly youth peaked. Taiwan’s age-curves track fairly closely with those for India, except for the proportionately lower involvement in crime on the part of Indian juveniles (ages 15–17) across all three crime measures. Third, although US age–crime distributions are right-skewed, they nonetheless are older and less sharply skewed as compared to the HG invariance norm. This US divergence from the invariance norm is particularly the case for the total measure and the violent-crime index. Fig. 2 View largeDownload slide Age–arrest distributions by offense index (total, violent, property) for the United States, Taiwan and India, with Spearman Rho when paired with invariance norm. Note—Rho is the Spearman rank-order correlation comparing the rank ordering of arrest rates for the four age categories as based on the invariance norm with the rank ordering as observed for the offense-index measure for each of the three countries (United States, Taiwan, India). The formula of Spearman’s rank-order correlation is: 1[6∑di2/n(n2−1)] ⁠, where di is the difference in paired ranks and n is the number of cases. The ranking for the invariance norm is: 1 = 15–17 (highest rate), 2 = 18–29, 3 = 30–44 and 4 = 45–59 (smallest rate). Rho value ranges from −1 to 1. A Rho of 1.0 means that the two age–arrest schedules have the same ranking (e.g. as is observed for the United States when paired with invariance norm), whereas a Rho of −1.0 means the two schedules have an opposite ranking (e.g. 1, 2, 3, 4 vs. 4, 3, 2, 1). Fig. 2 View largeDownload slide Age–arrest distributions by offense index (total, violent, property) for the United States, Taiwan and India, with Spearman Rho when paired with invariance norm. Note—Rho is the Spearman rank-order correlation comparing the rank ordering of arrest rates for the four age categories as based on the invariance norm with the rank ordering as observed for the offense-index measure for each of the three countries (United States, Taiwan, India). The formula of Spearman’s rank-order correlation is: 1[6∑di2/n(n2−1)] ⁠, where di is the difference in paired ranks and n is the number of cases. The ranking for the invariance norm is: 1 = 15–17 (highest rate), 2 = 18–29, 3 = 30–44 and 4 = 45–59 (smallest rate). Rho value ranges from −1 to 1. A Rho of 1.0 means that the two age–arrest schedules have the same ranking (e.g. as is observed for the United States when paired with invariance norm), whereas a Rho of −1.0 means the two schedules have an opposite ranking (e.g. 1, 2, 3, 4 vs. 4, 3, 2, 1). As also is shown in Figure 2, to further clarify the cross-country age–crime patterns with the invariance norm, we applied Spearman’s Rho to measure the correlations between the rank ordering of the four age categories of crime rates for each country to the rank order of age–crime rates for the invariance norm where 1 is the highest rank and 4 is the lowest rank. When applied, therefore, the Spearman’s rho coefficient ranges from −1 to 1 in which the value of 1.0 indicates the same rank order (or uniformity) in age–crime rates of a country (e.g. India or United States) as compared to the HG invariance norm, whereas a Rho value at or −1.0 indicates the two schedules have an opposite ranking.7 Key findings are as follows. First, we find similarity or uniformity between the rank ordering for the invariance norm as compared to US age–crime schedules where the youngest age group is ranked as having the highest arrest rate (a score of ‘1’) and there is a strict age-driven monotonic decline thereafter in rank-order scores (i.e. Rho is 1.0 across all comparisons). Second, in sharp contrast are the rank orderings for India and also for Taiwan. In India, the standardized age rates of the youngest age group have the lowest rank ordering across all three measures—i.e. opposite to what is found for the United States and invariance norm. Also very prominent for India is the ordering for the mid-age group (ages 30–44) as yielding either the highest ranking for two measures (total, violent index) or second highest for the other measure (property index). Third, generally similar rank orderings are found between India and Taiwan, with exception in Taiwan of a comparatively higher ranking for the youngest age group for the property-crime measure. Particularly noteworthy here are the Rho calculations for India and Taiwan (as compared to the invariance norm) that yield coefficients that fall well below 0.50 and in some instances are negative. Taken together, the rank orderings and the Rho calculations further confirm the robust differences in age–crime schedules for India (and Taiwan) as compared to the invariance norm and US age–crime distributions. Post-adolescent comparisons in age–crime schedules Last, we present in Figure 3 histograms that only consider the age scheduling of crime across three post-adolescent age groupings—18–29, 30–44, 45–59—in order to address more fully the second core projection of the HG invariance thesis of a continuous decline in the age–crime distribution either immediately or shortly after adolescence or young adulthood. The histograms also allow us to appraise further the possibility that India’s older and more spread out age–crime patterns might be an artifact of reporting practices. Post-adolescent comparisons also are presented for the United States and Taiwan, though our focus is on Indian post-adolescent age–crime patterns. Fig. 3 View largeDownload slide Post-adolescence age–arrest patterns by offense type (total, violent, property) for the United States, Taiwan, and India (ages 18–59). Fig. 3 View largeDownload slide Post-adolescence age–arrest patterns by offense type (total, violent, property) for the United States, Taiwan, and India (ages 18–59). Turning first to the post-peak decline projection, recall that an evaluation of the HG invariance thesis involves two distinct predictions: one is the universal peaking of crime in adolescence, and the other is the universal decline in crime from the peak (i.e. regardless of whether the peak is in adolescence or, as HG sometimes qualify, in late teens or early twenties). At issue is the direction or shape of age–crime schedules in India after young adulthood (ages 18–29). Do Indian age–crime rates decline (as HG would predict) or do they plateau or possibly even rise? Clearly, any instance of crime peaking near to or beyond age 30 is at odds with the projected ubiquity of a reverted J-shaped age–crime relation. Figure 3 is strikingly informative for clarifying just how robustly different the age–crime distributions in India are as compared to the HG invariance projection. Rather than declining after young adulthood, age–crime rates in India remain high well into the late thirties and forties. Notably, even after the twenties, there is a rise in crime rates in the majority of comparisons or measures (i.e. age rates are actually ticking upward and thus in reverse direction to the HG projection). For example, Indian rates for the violent index and total index are actually higher in the 30–44 age group than in the young-adult group (18–29). Of note also is the overall similarity in India’ post-young adulthood age–crime patterns to those observed for Taiwan—where there is a small rise in rates for the 30–44 age group as compared to the 18–29 group. In contrast to India (also Taiwan), moreover, the age–crime patterns in the United States display the downward decline projected by the invariance thesis. The findings shown in Figure 3 also are useful for considering further the possibility that observed differences in age curves for India may reflect variable law enforcement rather than differences in criminal behaviour. One robust way to address this possibility is to evaluate whether the large contrasts between the invariance norm and Indian age–crime schedules persist even when arrests are removed for juveniles (i.e. ages 15–17), when differential enforcement by age might affect the age patterns observed. Our reading of policing practices in India is that, whatever the presence of enforcement discretion towards teen law violators, it ends at around age 18 or shortly after (Sandhu 1987; Choudary and Sharma 2011). We know of no evidence to the contrary.8 A finding that India’s age–crime distributions become more right-skewed when using an age range in which teens are dropped from analysis would suggest differential enforcement. This of course does not happen, as we have seen. Instead, India age–crime distributions remain skewed to the left, and its age-specific crime rates continue to rise or hold steady well past the teen and young-adult ages when some reporting or enforcement discretion might be in play. Added evidence that the Indian findings are not an artifact of differential enforcement comes from the older age-schedule observed for homicide (reliably reported crime) and by the overall consistency of left-skewed age distributions across a variety of offense types. Instead, the Indian age–crime schedules (see also Figure 1) exhibit a much higher and longer right-hand tail than is observed for US age–crime distributions and for the J-shaped inverted curve that characterizes the HG invariance norm.9 Discussion and conclusion Current empirical and theoretical understanding of the relation between age and crime is based almost entirely on data from the United States and a few prototypical Western societies for which age-specific crime information across offense types is available. By using Western databases, Hirschi and Gottfredson (1983) projected that the age distribution of crime is always and everywhere robustly right-skewed with (1) a sharp spiked adolescent peak and (2) a continuous decline thereafter. HG’s position is contested but overall is widely accepted in criminology and social science writings. In the study described here, we investigated these two core postulates of the age–crime invariance thesis by comparing age–crime patterns in India (a collectivist non-Western Hindu society) with the inverted J-shaped invariance proposed by HG and also by comparing the Indian patterns to age–crime schedules in Taiwan and the United States. Our findings indicate considerable divergence in India’s age–crime patterns, compared with the age schedule projected by Hirschi and Gottfredson. Rather than peaking early and then declining, Indian age–crime distributions have later or ambiguous peaks and more spread-out distributions. Rather than being right-skewed, they are left-skewed or nearly symmetrical. Regardless of whether parameter or shape assessments of invariance are considered, the evidence of age–crime patterns as based on this Hindu-collectivist society is strongly at odds with the invariance thesis. India’s age–crime distributions do not display an early peak, nor is there a continuous decline in crime levels past the adolescent or young-adult years. A comparison of Indian age–crime distributions with those observed in the United States and Taiwan are also illuminating. The findings are straightforward in showing—first, that India and Taiwan age–crime patterns are older and flatter than US patterns. Second, on a continuum moving from older to younger distributions, Indian age–crime distributions are the oldest whereas US age–crime curves stand out as being most strongly youth peaked. Taiwan’s age-curves track fairly closely with those for India, except for the proportionately-lower involvement in crime on the part of Indian juveniles (ages 15–17). Third, although US age–crime distributions are right-skewed, they nonetheless are older and less sharply skewed as compared to the HG invariance norm (see Steffensmeier et al. 2017). Our results have a number of implications for research and theory on the aggregate age–crime relationship and on adolescent or youth offending in particular. First, they suggest skepticism for strong claims of invariance in the age–crime association in which there is a sharp peaking of crime in adolescence and a continuous decline in crime following young adulthood. Indian age–violence schedules were found to differ considerably from the invariance norm both in their parameters and underlying shape. In generally similar fashion, Taiwan’s age–crime schedules also seem strongly at odds with the invariance thesis. Moreover, though US age–crime curves display overall consistency with the HG invariance norm, they do show some variation over time and across offense types. Second, among its theoretical implications, therefore, the findings about India’s age–crime patterns can be seen as supportive of sociological life-course approaches and their emphasis on the importance of social context (Elder and Shanahan 2006). Adding to this support is the evidence reviewed earlier that also confirms the importance of social context—notably, international and historical studies documenting variation in the age–crime relation across space and time. As noted previously, finding consistent age differences around the world would indicate either that differences have a biological basis or that age socialization is remarkably constant; whereas a finding of international differences indicates that whatever biological factors may be implicated, age differences in crime are also strongly shaped by historical and country-specific social and cultural properties. Besides space and time variation in the age distribution of crime in the aggregate, findings based on individual-level data question the HG invariance position and proffer support for expecting variability in the age–crime relation depending on social context. First, significant variability between individuals in the same age groups and multiple age–crime offending trajectories among individuals has been found in longitudinal studies (e.g. Ezell and Cohen 2005; Piquero et al. 2007). Second, longitudinal studies also show that standard sociological variables (e.g. peer influence) explain fairly well the post-adolescence desistance in crime in some western countries (see review in Warr 1993; Sweeten et al. 2013). Third, plentiful research on desistance more generally further confirms the robust influence of social factors on ‘ageing out’ of crime (Uggen 2000; Farrall et al. 2009; Savolainen 2009; Skardhamar et al. 2014). Normative biological processes might play an important role in accounting for the age–crime relation more generally, but the evidence also suggests considerable plasticity in neurobiological and human development and the prominent role of social factors. A third implication is the need for more non-Western and international research on the age–crime relation. One priority is to conduct in-depth case studies of Indian culture and institutions to better understand how key explanatory factors (among others) raised in our speculative conceptual treatment might actually contribute to the age–crime contrasts observed in our analysis. Empirical and theoretical writings on Indian (and Taiwan) culture suggest a relatively high commitment to societal institutions such as family and school among teenagers as compared to their western counterparts, projecting a strong association between legitimacy of traditional institutions and low levels of youth crime in these two countries. Future studies to further assess this projection seem particularly useful for the development of the ‘new institutionalist’ criminology, with its emphasis on linking cross-national variation in crime levels (or criminal punishment) to citizenry trust, commitment and normative compliance to social institutions (see review in Karstedt 2010). The cases of India and Taiwan may suggest, also, that revitalization of traditional institutions in the lives of teenagers and young adults may help to attenuate relatively high level of youth crime in some countries. A second priority is the implementation of self-report studies of offending over the adolescent–adulthood transition period in India in order to better assess the possible effects of policing on age-specific arrest profiles and evaluate the application extant Western-based theories of crime. Ongoing monitoring of India’s age–crime patterns is also needed in light of its dynamic transition towards greater access to schooling and economic prosperity that may propel the presence of adolescence as a more distinct life-course phase. In a complex society like India, it also might be expected that alternative adolescent-developmental profiles may prevail across class groups (e.g. where the upper-class youth already experience a prolonged adolescence involving advanced schooling and delayed marriage). A third priority is to study age–crime patterns in other non-Western societies and collectivist cultures in particular (e.g. China, Pakistan, South Korea). Cross-national inquiries are indispensable for establishing the generality of theories of youth crime and elucidating the range of possibilities for minimizing levels of youth crime and violence (Aas 2012; Stamatel 2014). International inquiry also enables an assessment of the role of macrosocial and societal factors in the explanation of youth crime and the age–crime relation. Ideally, information on enough countries could be collected to enable modelling of international (also, historical) variations in the age–crime schedule with country-level predictors. Besides the collectivism-individualism dimension, one approach might involve applying added cultural maps of societies that may contribute to differences in the age–crime relationship—e.g. Ingelhart and Welzel’s (2005) ‘traditional/rational’ cultural dimension or Hofstede et al.’s (2010) six-fold dimensions of national cultures (e.g. short-term vs. long-term orientation, indulgence-restraint). Establishing cross-national differences in the age–crime relationship might also be an important departure for more generally studying factors effecting crime at the social system or group level. Acknowledgments Along with the editor and reviewers, we would like to thank S. George Vincentnathan for their helpful comments on prior drafts of the paper. References Aas , K . ( 2012 ), ‘ The Earth Is One but World Is Not: Criminological Theory and its Geopolitical Divisions ’, Theoretical Criminology , 16 : 5 – 20 . Google Scholar Crossref Search ADS Agnew , R . ( 2006 ), Pressured into Crime: An Overview of General Strain Theory . Roxbury Publishers . Akers , R . ( 2009 ), Social Learning and Social Structure: A General Theory of Crime and Deviance , 1st edn. Transaction Publishers . Barnes , J. , Boutwell , B. and Beaver , K . ( 2015 ), ‘ Contemporary Biosocial Criminology: A Systematic Review of the Literature, 2000–2012 ’, in A. Piquero , ed., The Handbook Criminological Theory , 5 – 99 . Wiley Blackwell . Bennett , R . ( 1991 ), ‘ Development and Crime: A Cross-National, Time-Series Analysis of Competing Models ’, Sociological Quarterly , 32 : 343 – 63 . Google Scholar Crossref Search ADS Blumstein , A. and Wallman , J . ( 2006 ), The Crime Drop in America , 2nd edn. Cambridge University Press . Bond , M . ( 2010 ), The Oxford Handbook of Chinese Psychology , 1st edn. Oxford University Press . Braithwaite , J . ( 2015 ), ‘ Rethinking Criminology Through Radical Diversity in Asian Reconciliation ’, Asian Journal of Criminology , 10 : 183 – 91 . Google Scholar Crossref Search ADS Britt , C . ( 1992 ), ‘ Constancy and Change in the U.S. Age Distribution of Crime: A Test of the “Invariance Hypothesis” .’ Journal of Quantitative Criminology , 8 : 175 – 87 . Google Scholar Crossref Search ADS Chaudhary , N. and Sharma , N . ( 2011 ), ‘ Adolescence in India ’, in J. Arnett , ed., Adolescent Psychology Around the World , 103 – 18 . Taylor and Francis . Clinard , M. and Abbott , D . ( 1973 ), Crime in Developing Countries: A Comparative Perspective . Wiley Interscience . Connell , R . ( 2006 ), ‘ Northern Theory: The Political Geography of General Strain Theory ’, Theoretical Sociology , 35 : 237 – 64 . Google Scholar Crossref Search ADS Elder , G. and Shanahan , M . ( 2006 ), ‘ The Life Course and Human Development ’, in R. Lerner , ed., Handbook of Child Psychology: Theoretical Models of Human Development , 665 – 715 . Wiley . Ellis , B. , Del Giudice , M. , Dishion , T. , Fuqueredo , A. , Gray , P. , Griskevicius , V. and Wilson , D . ( 2012 ), ‘ The Evolutionary Basis of Risky Adolescent Behavior: Implications for Science, Policy, and Practice ’, Developmental Psychology , 48 : 598 – 623 . Google Scholar Crossref Search ADS PubMed Epstein , R . ( 2007 ). ‘ The Myth of the Teen Brain ’, Scientific American Mind , 18 : 56 – 63 . Google Scholar Crossref Search ADS Ezell , M. and Cohen , L . ( 2005 ), Desisting from Crime: Continuity and Change in Long-term Crime Patterns of Serious Chronic Offenders . Oxford University Press . Farrall , S. , Godfrey , B. and Cox , D . ( 2009 ), ‘ The Role of Historically-Embedded Structures in Processes of Criminal Reform: A Structural Criminology of Desistance ’, Theoretical Criminology , 13 : 79 – 104 . Google Scholar Crossref Search ADS Farrington , D . ( 1986 ), ‘ Age and Crime ’, Crime and Justice , 7 : 189 – 250 . Google Scholar Crossref Search ADS Gottfredson , M. and Hirschi , T . ( 1990 ), A General Theory of Crime . Stanford University Press . Greenberg , D . ( 1985 ), ‘ Age, Crime, and Social Explanation ’, American Journal of Sociology , 91 : 1 – 21 . Google Scholar Crossref Search ADS —— ( 2008 ), ‘ Age, Sex, and Racial Distribution of Crime ’, in E. Goode , ed., Out of Control: Assessing the General Theory of Crime , 38 – 48 . Stanford Social Sciences . Hakeem , F. and Lukash , A . ( 2017 ), ‘ Self-Reported Delinquency in India: Findings from the Third International Survey ’, Paper Presented at the Annual Meeting of the American Society of Criminology , Philadelphia, PA . Hartjen , C . ( 1982 ), ‘ Delinquency, Development, and Social Interaction in India ’, Social Problems , 29 : 465 – 73 . Google Scholar Crossref Search ADS Hirschi , T. and Gottfredson , M . ( 1983 ), ‘ Age and the Explanation of Crime ’, American Journal of Sociology , 89 : 552 – 84 . Google Scholar Crossref Search ADS Hofstede , G. , Hofstede , G. and Minkov M . ( 2010 ), Cultures and Organizations: Software of the Mind , 3rd edn. McGraw-Hill Education . Ingelhart , R. and Welzel , C . ( 2005 ), Modernization, Cultural Change and Democracy . Cambridge University Press . Jennissen , R . ( 2014 ), ‘ On the Deviant Age-Crime Curve of Afro-Caribbean Populations: The Case of Antilleans Living in the Netherlands ’, American Journal of Criminal Justice , 39 : 571 – 94 . Google Scholar Crossref Search ADS Kakar , S . ( 1979 ), Indian Childhood: Cultural Ideals and Social Reality . Oxford University Press . Kakar , S. and Kakar , K . ( 2007 ), The Indians: Portrait of a People . Viking Penguin . Kanazawa , S. and Still , M . ( 2000 ), ‘ Why Men Commit Crimes (and Why They Desist) ’, Sociological Theory , 18 : 434 – 47 . Google Scholar Crossref Search ADS Karstedt , S . ( 2010 ), ‘ New Institutionalism in Criminology: Approaches, Theories, and Themes ’, in E. McLaughlin and T. Newburn , ed., The Sage Handbook of Criminology , 337 – 59 . Sage . Larson , R. , Verma , S. and Dworkin , J . ( 2003 ), ‘ Adolescence Without Family Disengagement: The Daily Family Lives of Middle-Class Teenagers in India ’, in T. S. Saraswathi , ed., Cross-Cultural Perspectives in Human Development , 258 – 86 . Sage . Lemert , E . ( 1951 ), Social Pathology: A Systematic Approach to the Theory of Sociopathic Behavior . McGraw-Hill . Lynch , J. and Pridemore , W . ( 2011 ), ‘ Crime in International Perspective ’, in J. Wilson and J. Petersilia , ed., Crime and Public Policy , 5 – 52 . Oxford University Press . Mandelbaum , D . ( 1970 ), Society in India: Continuity and Change , Vol. 1. University of California Press . Mannheim , H . ( 1965 ), Comparative Criminology: A Textbook , Vol.1, 1st edn. Boston Houghton Mifflin . Menon , D . ( 2017 ), Cultural History of Modern India . Orient Blackswan . Moffitt , T . ( 1993 ), ‘ Adolescence-limited and Life-Course Persistent Antisocial Behavior: A Developmental Taxonomy ’, Psychological Review , 100 : 674 – 701 . Google Scholar Crossref Search ADS PubMed Nickels , E. and Verma , A . ( 2008 ), ‘ Dimensions of Police Culture: A Study in Canada, India, and Japan ’, Policing , 31 : 61 – 78 . O’Brien , R. M. , Stockard , J. and Isaacson , L . ( 1999 ), ‘ The Enduring Effects of Cohort Characteristics on Age-Specific Homicide Rates, 1960–1995 ’, American Journal of Sociology , 104 : 1061 – 95 . Google Scholar Crossref Search ADS Parsons , T . ( 1951 ), The Social System , 1st edn. Free Press . Piquero , A. , Farrington , D. and Blumstein , A . ( 2007 ), Key Issues in Criminal Career Research: New Analyses of the Cambridge Study in Delinquent Development . Cambridge University Press . Sandhu , H . ( 1987 ), ‘ Low Rates of Delinquency and Crime in India: A Case for Strong Social Controls ’, Indian Journal of Criminology , 15 : 9 – 16 . Saraswathi , T . ( 1999 ), ‘ Adult-Child Continuity in India: Is Adolescence a Myth or an Emerging Reality ?’ in T. Saraswathi , ed., Culture, Socialization, and Human Development , 213 – 32 . Sage . Savolainen , J . ( 2009 ), ‘ Work, Family and Criminal Desistance: Adult Social Bonds in a Nordic Welfare State ’, The British Journal of Criminology , 49 : 285 – 304 . Schlegel , A. and Barry , H . ( 1991 ), Adolescence: An Anthropological Inquiry . Free Press . Shyrock , H. and Siegel , J . ( 1976 ), The Methods and Materials of Demography . Academic Press . Singh , U. and Nayak , A . ( 2017 ), Crime in India . Global Open University . Skardhamar , T. , Monsbakken , C. W. and Lyngstad , T. H . ( 2014 ), ‘ Crime and the Transition to Marriage: The Role of the Spouse’s Criminal Involvement ’, British Journal of Criminology , 54 : 411 – 27 . Google Scholar Crossref Search ADS Stamatel , J . ( 2014 ), ‘ Revisiting Nations Not Obsessed with Crime ’, Crime, Law and Social Change , 94 : 113 – 29 . Google Scholar Crossref Search ADS Steffensmeier , D. and Allan , E . ( 2000 ), ‘ Looking for Patterns: Gender, Age, and Crime ’, in J. Sheley , ed., Criminology: A Contemporary Handbook , 85 – 128 . Wadsworth . Steffensmeier , D. , Allan , E. , Harer , M. and Streifel , C . ( 1989 ), ‘ Age and the Distribution of Crime ’, American Journal of Sociology , 94 : 803 – 31 . Google Scholar Crossref Search ADS Steffensmeier , D. , Streifel , C. and Shihadeh , E . ( 1992 ). ‘ Cohort Size and Arrest Rates over the Life Course: The Easterlin Hypothesis Reconsidered ’, American Sociological Review , 57 : 306 – 14 . Google Scholar Crossref Search ADS Steffensmeier , D. , Zhong , H. and Lu , Y . ( 2017 ), ‘ Age and its Relation to Crime in Taiwan and the United States: Invariant, or Does Cultural Context Matter ?’, Criminology , 55 : 377 – 404 . Google Scholar Crossref Search ADS Steinberg , L . ( 2008 ), ‘ A Social Neuroscience Perspective on Adolescent Risk-Taking ’, Developmental Review , 28 : 78 – 106 . Google Scholar Crossref Search ADS PubMed Sweeten , G. , Piquero , A. and Steinberg , L . ( 2013 ), ‘ Age and the Explanation of Crime, Revisited ’, Journal of Youth and Adolescence , 42 : 921 – 38 . Google Scholar Crossref Search ADS PubMed Tittle , C. and Grasmick , H . ( 1997 ), ‘ Criminal Behavior and Age: A Test of Three Provocative Hypotheses ’, Journal of Criminal Law and Criminology , 88 : 309 – 42 . Google Scholar Crossref Search ADS Uggen , C . ( 2000 ), ‘ Work as a Turning Point in the Life Course of Criminals: A Duration Model of Age, Employment, and Recidivism ’, American Sociological Review , 65 : 529 – 46 . Google Scholar Crossref Search ADS Verna , S. , Sharma , D. and Larson , R . ( 2002 ), ‘ School Stress in India: Effects on Time and Daily Emotions ’, International Journal of Behavioral Development , 26 : 506 – 8 . Walsh , A . ( 2009 ), ‘ Crazy by Design: A Biosocial Approach to the Age-Crime Curve ’, in A. Walsh and K. Beaver , ed., Biosocial Criminology: New Directions in Theory and Research , 154 – 75 . Routledge . Warr , M . ( 1993 ), ‘ Age, Peers, and Delinquency ’, Criminology , 31 : 17 – 40 . Google Scholar Crossref Search ADS Young , J . ( 2004 ). ‘ Voodoo Criminology and the Numbers Game ’, in J. Ferrell , K. Hayward , W. Morrison and M. Presdee , ed., Cultural Criminology Unleashed , 13 – 27 . Glasshouse Press . Footnotes 1 For parsimony, we hereafter use the term ‘age–crime’ when referring to the age distribution of arrests (i.e. measure of ‘crime’ used by Hirschi and Gottfredson and in our analysis). However, we recognize the limitations of using arrest data as a proxy of age-specific criminal behaviours (see discussion in Data and Method). 2 We also have benefitted from qualitative information from ‘indigenous experts’ about Indian social and cultural practices through interviews with Indians (government officials, law enforcement, businessmen and university faculty) who have also lived in the United States for extended periods of time. Of note: (1) there is sizable Indian community in the city where our university is located; (2) one of the co-authors of this study is a native of India who has working experience with various State Police organizations in India and with juvenile and adult prison systems in the states of Maharashtra, Delhi and Chhattisgarh. 3 Owing to both page and citation limitations, not all the literature we reviewed could be included in this article. Thus, we selected works or pieces of scholarship that seemed particularly relevant to the age–crime issue in the context of our focus on India. 4 We applied the approx() function in R statistics package for linear interpolation. This statistical function linearly interpolates given data points (i.e. the age-grouped rates of the United States and India) and returns a list of points (i.e. individual age–arrest rates). The interpolation is basically a smoothing technique that does not change the overall relationship between arrests and age or produce a different total arrest rate (Steffensmeier et al. 1992, 2017; see also: Shyrock and Siegel 1976). We also tested the results with other smoothing functions available in R (e.g. spline() and smooth()), they demonstrate consistent patterns. Moreover, added analysis revealed similarity in findings with or without interpolation; results available upon request. 5 The somewhat higher youth involvement in property crime may reflect the sizable numbers of adolescents and youth in India who have lost family relationships and were thus severed from economic support (see Hartjen 1982; Singh and Nayak 2017). 6 Our approach here follows that of Blumstein and Wallman (2006) who applied this standardization method to establish equivalence between crime trends of different offense types in which the size of the rates varied considerably (e.g. where the homicide rate is much smaller than assault rate). 7 The formula for Spearman’s rank-order correlation is: 1[6∑di2/n(n2−1)] where di difference in paired ranks and n = number of cases. 8 Of note, also, is the conclusion of some prominent cross-national scholars that the reasons for reporting and/or not reporting crime are roughly similar in developing (like India) as compared to developed countries (like the United States [Bennett 1991; Nickels and Verma 2008]). 9 Added evidence for the conclusion of lower rates of delinquency in India, including in comparison to older age groups, can be drawn (1) from observations of prominent Indian social scientists who also observe low levels of youth crime in India (e.g. see Chaudhary and Sharma 2011; Sandhu 1987); (2) self-report delinquency research in India showing very low levels self-reported crime by Indian adolescents (Hartjen 1982; Hakeem and Lukash 2017); and (3) qualitative information from interviews of ‘indigenous experts’ (see endnote #1) who uniformly were in agreement of (1) low levels of delinquency in India today as compared with the United States and (2) adolescence being a much less demarcated stage of development in India (see also Chaudhary and Sharma 2011). © The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). All rights reserved. For permissions, please e-mail: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Leerkes,, Arjen;Martinez,, Ramiro;Groeneveld,, Pim

doi: 10.1093/bjc/azy021pmid: N/A

Abstract Immigrants and their native-born children tend to be overrepresented among crime suspects in Europe. Using a representative Dutch survey, we examine whether inhabitants of Turkish and Moroccan origin also self-report more crimes than the native Dutch. In addition, we test various explanations for ethnic differences in crime, partly using variables that are unavailable in administrative data (socio-economic status [SES], perceived discrimination, neighbourhood disadvantage and control, family bonds, religiousness). We discover two ‘minority paradoxes’. Firstly, contrary to analyses using administrative data, both minorities have similar to lower self-reported crime rates compared to the majority group when age, sex, urbanization, SES and social desirability are controlled. Secondly, first-generation immigrants report fewer crimes than expected given their social disadvantage, thus indicating a notable ‘righteous migrant effect’. Introduction Immigrants and their native-born children are generally found to be overrepresented among crime suspects and convicted offenders in Europe (De Haen-Marshall 1997; Tonry 1997; Bucerius and Tonry 2014). A recent Dutch study, e.g., reported that the country’s four largest ethnic minorities, which originate from Turkey, Morocco, Surinam and the Dutch Antilles, had between 2.2 (first-generation Turkish immigrants) and 8.1 (native-born with two Moroccan-born parents) higher odds of being a crime suspect than the native Dutch (Blom and Jennissen 2014). Variables such as age, sex, socio-economic status (SES), and degree of residential urbanization only partially explained the higher odds. Although there is considerable doubt about whether official crime data indicate ethnic differences in criminal behaviour (Tonry 1997; Van der Leun and Van der Woude 2011), European scholars tend to see the overrepresentation of ethnic minorities in official crime figures as a behavioural consequence of a hampered social integration of migrants and their children in the host society (Tonry 1997; Killias 2009; Koopmans 2010; Engbersen et al., 2014). Interestingly, in the traditional immigration countries outside Europe (United States, Canada and Australia), first- and second-generation immigrants are typically not, or less, overrepresented in crime statistics (Lynch and Simon 1999; Wortley 2009; Bersani 2014; Bucerius and Tonry 2014). American researchers actually speculate about a protective effect of immigration, which is assumed to dissipate over time. In that connection, Martinez (2002) and Sampson and Bean (2006) introduced the notion of a ‘Latino paradox’ in American criminology, which holds that foreign-born Latinos in particular do better on various social indicators, including violence, than would be expected given their social disadvantage. Here, crime among immigrants and their offspring is actually seen as a—paradoxical—consequence of increased integration into (segments of) US society (Portes and Zhou 1993; Sampson et al. 2005; Rumbaut and Ewing 2007; Sampson 2008). The available quantitative studies on migration, ethnicity and crime are almost entirely based on criminal justice and other administrative data, and two limitations result from that focus. Firstly, it is unclear to what extent findings regarding the overrepresentation of ethnic minorities among crime suspects in Europe indicate differences in criminal behaviour or mostly point at ethnic differences in the probability of criminal behaviour being detected, reported and prosecuted. Official crime data are likely to be biased against ethnic minorities, especially those with lower SES. A considerable body of research points at the importance of stereotyping and prejudice among officials and the wider public to the disadvantage of individuals of low SES and/or ethnic minority background (Weenink 2009; Van der Leun and Van der Woude 2011). Additional biases can be expected from Black’s (1970) social distance theory, which argues that the probability of formal law being applied increases with the social distance between offender and victim (also see Rojek et al. 2010). Because of family ties and network homophily, criminal victimization tends to be intra-ethnic, especially for crimes, such as intimate partner violence, that typically occur within the offender’s primary group (O’Brien 1987). However, individuals from a small group are more likely to meet individuals from a large group than vice versa. Given that ethnic minority populations are smaller than the ethnic majority, offenders with minority status will therefore, for statistical reasons, be relatively likely to victimize somebody outside their ethnic group—thus putting themselves at a higher risk of formal sanctioning. Such biases may be less influential in the United States and Canada than in Western European countries, which are relatively reluctant to receiving immigrants (cf. Cornelius et al. 2004; Esipova et al. 2015) and generally have lower percentages of foreign-born. By implication, transatlantic differences in the association between migration and crime may be smaller than official crime figures indicate. Although most Dutch criminologists contend that selective law enforcement only partially explains the overrepresentation of ethnic minorities in officially recorded crime (for a discussion see Engbersen et al. 2014), the extent to which such figures do indicate ethnic differences in criminal behaviour is unclear. Secondly, quantitative analyses of the causes of ethnic differences in criminal behaviour have largely been limited to variables that researchers could measure with administrative data—using variables such as age, sex, ‘generation’ and, in more advanced analyses, using administrative microdata (Blom and Jennissen 2014; Bovenkerk and Fokkema 2016), socio-economic position and demographic household characteristics. Other relevant factors—both factors that put ethnic minorities at risk of offending (e.g. perceived ethnic discrimination) and factors that may be protective for them (e.g. family bonds and religiosity)—have rarely been considered in studies of this type. Thus, to better understand whether and how ethnicity is related to criminal behaviour, we have to strive for methodological triangulation and also examine data that (1) are not directly influenced by selective law enforcement and (2) include information that is unavailable in administrative data. This article aims to fill both gaps with data from the Netherlands Longitudinal Lifecourse Study (NELLS). Conducted by a team of sociologists, the NELLS is a large survey (N = 5,312) on the living situation and attitudes of the Dutch population in 2010 (De Graaf et al. 2010).1 The survey included several items on offending and is representative for the native Dutch population aged between 15 and 45 years and for ethnic Turks and Moroccans—the country’s two largest ethnic minorities—in that age span. Hereafter, these two groups will be called the ‘Turkish Dutch’ and ‘Moroccan Dutch’ to acknowledge their connection to the Netherlands (most persons in these groups also have Dutch nationality). The ‘native Dutch’ are defined as people with two Netherlands-born parents. The following research questions guided the analysis: (1) Do first- and second-generation Turkish and Moroccan Dutch report more crimes than native Dutch respondents? (2) How can ethnic differences in self-reported offending be explained? (3) How do the patterns of self-reported offending compare to police data? Using the NELLS data, we consider five factors that may promote ethnic differences in criminal behaviour: (1) SES, (2) perceived ethnic discrimination, (3) neighbourhood disadvantage and control, (4) family bonds and (5) religiousness. We juxtapose levels of self-reported crime by ethnic group and generation to police data, and compare our multivariate results to similar analyses regarding crime suspects (Blom and Jennissen 2014). We are aware of four (mostly older) studies that have previously compared rates of self-reported offending among different ethnic groups. These studies have yielded differing results: two found higher rates of self-reported offending among ethnic minorities (but not of the magnitudes that are regularly observed in official data), one found lower rates and one study found no significant differences. Killias (2009) conducted a survey in 2006 among 2,067 native Swiss youth and 772 migrant youths, and, for most crimes, found two to three times higher rates for minority youth from ‘Balkan countries’, and smaller differences for other immigrant youth. Wittebrood (2003) analysed data from a Dutch school survey that included 420 youngsters of Moroccan or Turkish origin and found that they were somewhat more likely to report crimes than native Dutch youth (16% against 12%). In 1993 and 1994, Murad et al. (2003) conducted a survey among 363 youth of Turkish origin and 1,098 native Dutch adolescents: 31% of the native Dutch males scored above the 75th percentile on a delinquency scale against 21% of the Turkish Dutch males. Junger and Polder (1992) compared self-reported crime and police data on Moroccan (N = 182), Turkish (N = 196), Surinamese (N = 206) and native Dutch (N = 204) youth living in the same neighbourhood, all with comparable SES, and found no significant ethnic differences in self-reported offending. Junger and Polder stressed that their survey results should be distrusted as respondents of immigrant origin allegedly underreported criminal behaviour more than native Dutch youth. The recent methodological literature indeed gives some credibility to the latter claim (cf. Kim and Kim 2015). First-generation immigrants in particular—especially those from relatively ‘collectivistic’ societies—typically show higher magnitudes of social desirability bias than the majority group. Among the advantages of the NELLS data compared to the existing studies is that the data include various items from the Marlowe–Crowne Social Desirability Scale, which allowed us to assess and control ethnic and generational differences in social desirability (cf. Saunders 1991). The theoretical background of our hypotheses is discussed in the Hypotheses, Risk and Protective Factors section. As we will see, each hypothesis involves two elements: it is argued why a factor explains criminal behaviour and why that factor is expected to promote higher or lower crime rates among the Turkish and/or Moroccan Dutch relative to the native Dutch. As the second element requires some knowledge of the characteristics of the ethnic groups included, the next section first gives some concise information about immigration in the Netherlands. The Netherlands as a Reluctant Country of Immigration In 2010, when the NELLS was conducted, 10.3% of the Dutch population (N = 16,574,989) was foreign-born, and 10.0% was native-born with one or two foreign-born parents. With 2.3% and 2.1% respectively, both the Turkish and Moroccan Dutch represented about one-tenth of those of foreign (parental) origin. Other immigrant groups mostly originate from former colonies (e.g. Surinam, the Dutch Antilles), from source countries of asylum migration (e.g. Afghanistan, Somalia) and from European Union member states (e.g. Germany, Poland). The Netherlands has been characterized as a ‘reluctant country of immigration’ (Cornelius et al. 2004). To the extent that the Dutch government actively sought to attract migrants, it mainly recruited ‘guest workers’ from Mediterranean countries in the 1960s and early 1970s. Like other Western European countries, it first attracted workers from Southern Europe, before recruiting Turks and Moroccans. The workers were expected to return to their countries of origin, but a significant part of the Turkish and Moroccan workers in particular reunified with their families and settled permanently. Afterwards, a considerable number of youth with Turkish or Moroccan parents married a spouse from their parents’ country of origin (Leerkes and Kulu-Glasgow 2011). Both Turkey and Morocco—especially the rural areas where most migrants originated from—are relatively poor and ‘traditional’ societies with high levels of communal social control, as indicated by a patriarchal family structure and high rates of religiousness (Lucassen and Penninx 1997; Pels 2000). In spite of gradual upwards social mobility (Heath et al. 2008) and stricter immigration policies (Leerkes and Kulu-Glasgow 2011), the Turkish and Moroccan Dutch still have a relatively low SES. In 2010, e.g., the percentage of households with a disposable annual income below €10,000 was 10% for the ‘Turks’ and 9.2% for the ‘Moroccans’, against 4.5% for the native Dutch.2 Both minorities are also overrepresented in relatively disadvantaged neighbourhoods (cf. Leerkes and Bernasco 2010). The Turkish and Moroccan Dutch tend to have relatively tight familial bonds. Both groups—especially the Turkish Dutch—typically marry and procreate at a relatively early age, and they are relatively unlikely to live on their own as a student (Leerkes and Kulu-Glasgow 2011). In 2015, 40% and 50% of the Turkish and Moroccan Dutch reported having daily contacts with their family, against 30% among the native Dutch (Statistics Netherlands 2016). Among the native Dutch in particular, there has been a notable increase in ‘liquid love’ (cf. Bauman 2003): as lifelong relationships became less common, a substantial number of people live in single-person households during periods of their lives. In 2010, 39% of all households among persons in the 15–45 age span were single-person households.3 Almost all Turkish Dutch (94%) and Moroccan Dutch (97%) are Muslims, and a majority (56% and 84%) are practicing Muslims, meaning that they pray daily, follow a halal diet and participate in Ramadan. A part of them also regularly visit mosques (70% among practicing ‘Turks’ and 51% among practicing ‘Moroccans’) (Maliepaard and Gijsbers 2012). Among the general population, 49% self-identifies as religious, making the Netherlands one of the world’s most atheist countries (De Hart and Dekker 2013; Röder 2015). In 2013, the majority of the Turkish (73%) and Moroccan Dutch (76%) claimed to have experienced some form of discrimination that year (Andriessen et al. 2014). Among those experiencing discrimination, 54% felt discriminated because of ethnicity, and 51% (Moroccans) and 39% (Turks) felt discriminated because of religion. Among the native Dutch, less than 1% felt discriminated on ethnic or religious grounds. Field experiments show the existence of actual ethnic discrimination on the labour market (Andriessen et al. 2012). Van der Leun and Van der Woude (2011) hypothesized that the increasingly ‘actuarial’ Dutch criminal justice system—which has led to proactive stop and search powers being expanded—entails a risk of ethnic profiling. Two recent ethnographies indeed highlight widespread practices of ethnic and racial profiling among Dutch police, which, according to the researchers, are partly due to racism among police (Mutsaers 2014; Çankaya 2015). Older studies found little evidence for such racism and argued that eventual ethnic selectively mostly arose from a desire among police to optimize ‘organizational output’ under conditions of limited resources: at most, police tended to monitor groups that they had come to suspect of crime (see e.g. Junger 1990). In the latter view—which is still dominant in Dutch criminology—selective law enforcement possibly inflates but does not ultimately cause the overrepresentation of ethnic minorities in official crime figures (for a fuller discussion, see Engbersen et al. 2014). It is pointed out that various ‘white’ minorities (Yugoslavs, Russians, Poles) are also prominent in crime statistics, although they are less likely to experience ethnic profiling. Some ‘visible’ minorities, such as the Chinese Dutch, are actually underrepresented among crime suspects. In 2011, 71% of the Turkish Dutch and 67% of the Moroccan Dutch had ‘sufficient’ trust in the police, against 78% of the native Dutch (Huijnk and Andriessen 2016). Hypotheses, Risk and Protective Factors We started with the hypothesis that first- and second-generation Turks and Moroccans report higher levels of criminal behaviour than respondents of native Dutch origin (H1). The hypothesis is in line with all Dutch studies using police suspect data (e.g. Blokland et al. 2010; Blom and Jennissen 2014; Engbersen et al. 2014), the dominant view among Dutch criminologists that selective law enforcement only partially explains ethnic differences in suspect rates and Wittebrood’s (2003) findings regarding self-reported delinquency. In what follows, we discuss five additional hypotheses (H2–H6) that pertain to the second research question. We first discuss factors that are expected to put ethnic minorities such as the Turkish and Moroccan Dutch at risk of criminal offending (SES, perceived ethnic discrimination, neighbourhood deprivation and control), followed by a discussion of two factors that may be protective for these groups compared to the native Dutch (family bonds and religiousness). Ellis and McDonald (2001) compared 273 studies into the relation between SES and criminal behaviour and concluded that SES was a significant predictor of criminal behaviour in many, though not all, studies. According to general strain theory, which builds on Merton’s (1967) anomie theory, a low SES may put pressure on people to engage in criminal behaviour (cf. Agnew 2001). Moreover, social bonds are assumed to be relatively weak among the lowest social strata in particular. Blom and Jennissen (2014) indeed found that ethnic differences in SES explain part of the overrepresentation of the Turkish and Moroccan Dutch among crime suspects. For these reasons, we hypothesize (H2): The effect of being a first- or second-generation Turk or Moroccan on self-reported offending diminishes after controlling SES. Immigrants are often labelled with an ethnic ‘master status’ (cf. Hughes 1945), i.e. a status overshadowing other traits. Such forms of labelling risk being coupled with ethnic discrimination, both actual and perceived, which promote criminal behaviour for two reasons. Firstly, discrimination is likely to be felt as an additional strain on top of low SES (Simons et al. 2003). Secondly, it is an indicator of a social environment seeing certain groups as ‘deviant’, possibly promoting ‘secondary deviance’ via the labelling mechanism (cf. Grattet 2011). We therefore hypothesize (H3): The effect of being a first- or second-generation Turk or Moroccan on self-reported crime diminishes after controlling perceived ethnic discrimination. Neighbourhood social control reduces neighbourhood deviance and crime, including crimes committed by residents (cf. Sampson et al. 1997). Poorer immigrants are likely to start their housing careers in relatively ‘disorganized’ and socially deprived neighbourhoods. Over time, they obtain the resources to relocate to more prosperous and ‘organized’ neighbourhoods, or may reinvigorate their neighbourhoods of settlement (Sampson 2008). Given that the Turkish and Moroccan Dutch are still overrepresented in disadvantaged and ethnically diverse neighbourhoods, we hypothesize (H4): The effect of being a first- or second-generation Turk or Moroccan on self-reported crime diminishes after controlling differences in neighbourhood disadvantage and control. Family bonds tend to reduce criminality as they typically foster informal social control. In his classical formulation of social bond theory, Hirschi (1969: 16) explicitly mentions family relations: ‘Elements of social bonding include attachment to families, commitment to social norms and institutions (school, employment), involvement in activities, and the belief that these things are important’. Before we started this study, it was somewhat unclear whether familial bonds are stronger or weaker among Turkish and/or Moroccan Dutch than among the native Dutch. On the one hand, there was evidence—which was discussed in the previous section—that familial bonds are relatively strong among the minorities, leading us to hypothesize (H5a): The effect of being a first- or second-generation Turk or Moroccan on self-reported crime increases after controlling family bonds. Junger and Polder (1992), by contrast, found that ‘Moroccan’ and ‘Turkish’ youth had higher arrest rates than the native Dutch of comparable SES living in the same neighbourhood, which they implicitly attribute to differential family bonds, including lower parental supervision of ethnic minority youth, especially among ‘Moroccans’. Although the NELLS data mostly pertain to young adults (rather than youth), we therefore formulated the following counterhypothesis: The effect on self-reported crime of being a first- or second-generation Moroccan in particular diminishes after controlling familial bonds (H5b). Religiosity is generally found to be negatively associated with crime (Johnson et al. 2000). In the European context in particular, first- and (less so) second-generation immigrants are more religious than the majority population, which is especially true for migrants originating from rural areas in ‘Islamic countries’ (Röder 2015). As a rule, religiosity can therefore be expected to protect Dutch Moroccans and Turks from crime relative to the native Dutch, leading us to hypothesize (H6): The effect of being a first- or second-generation Turk or Moroccan on self-reported crime increases after controlling religiousness. Methodological Approach Data and missing values The NELLS is a large survey for sociological research on the living situation and opinions of residents of the Netherlands (for a more elaborate methodological discussion, see De Graaf et al. 2010). It pertains to the 15–45 years age span and was carried out in the 4 biggest cities (Amsterdam, Rotterdam, The Hague and Utrecht) and 31 randomly selected municipalities. Random samples were taken from three types of inhabitants. Firstly, persons from Turkish origin (born in Turkey or with at least one parent born in Turkey); secondly, persons from Moroccan origin (born in Morocco or at least one Moroccan-born parent) and thirdly, the remainder of the population, mostly native Dutch (two Netherlands-born parents), but also inhabitants with other parental origins (the latter were excluded from the analyses). The ‘Turks’ and ‘Moroccans’ were oversampled so as to obtain a sample that would also be representative for the two minority groups. We used the first wave, which was carried out in 2010 (N = 5,312). It consisted of two parts, a face-to-face interview and a questionnaire, which included the items measuring criminal behaviour. Operationalization Five items were combined to measure criminal behaviour. Respondents were asked if, in the last 12 months, they had been involved in the following activities: ‘stolen something from a person or a shop’, ‘damaged or demolished property of others’, ‘carried a weapon (knife, gun)’, ‘threatened someone’, ‘kicked or punched someone or participated in a fight’. For each item, respondents could answer on a four-point scale (‘never’, ‘once’, ‘two–three times’ and ‘four times or more’). We calculated the sum of reported crimes for each respondent with the category ‘two–three times’ being counted as 2.5, and ‘four times or more’ as 4. We also created a dichotomous variable measuring whether or not the respondent reported at least one crime. Two variables measured SES: income and educational level. Income is the net monthly income of the respondent and eventual partner. Incomes were classified into low (net monthly income of up to €1,499), middle (€1,500–€2,499) and high (€2,500 or more). Educational attainment was similarly classified in three categories: ‘no formal education or primary education’, ‘secondary education’ (high school, lower professional education) and ‘tertiary education’ (higher professional education, university degrees). When respondents were still in school, the unfinished school level was used. Respondents who obtained a diploma outside of the Netherlands were asked to indicate whether the diploma corresponded to a low, middle or high education level. Perceived ethnic discrimination was measured by averaging the following items: ‘Did you in any of the following situations experience that you were discriminated on the basis of your ethnic background (1) during a job application, (2) at work, (3) at school, (4) on the street, in stores, or in public transport, (5) in associations or sports clubs, (6) during nightlife or discotheques’. Respondents could answer 1: ‘no, never’; 2: ‘yes, once in a while’ or 3: ‘yes, quite often’. We used the average value of the scores on these six variables and deducted 1, thereby obtaining a scale from 0 (no perceived discrimination) to 2 (high perceived discrimination). Cronbach’s alpha was 0.81. Perceived ethnic discrimination was set at zero for all native Dutch respondents. The NELLS did not ask the native Dutch about perceived ethnic discrimination, probably because studies show that they hardly experience it (cf. Andriessen et al. 2014). As was reported in the previous section, the latter assumption is more or less supported by other survey data. Three community measures were included. Firstly, neighbourhood disadvantage, a neighbourhood-level factor score based on four indicators by Statistics Netherlands: (1) the average real estate value in the neighbourhood (factor loading: −0.53), (2) the percentage of rental housing (0.47), (3) the average income per income recipient (−0.46) and (4) the percentage of residents receiving welfare benefits (0.53). The factor had an eigenvalue of 2.9 and explained 74% of the variation in the four indicators mentioned. In addition, we calculated neighbourhood collective efficacy and neighbourhood contacts, by aggregating to the neighbourhood level the mean values on selected survey items for all respondents living in the same neighbourhood, regardless of ethnic origin. Neighbourhood collective efficacy is the neighbourhood average of six 5-point items, each ranging from 0 to 4, namely people in this neighbourhood: (1) ‘greet each other’, (2) ‘trust each other’, (3) ‘go along well’, (4) ‘know each other’, (5) ‘like to help each other’ and (6) ‘would say something against youth causing nuisance’ (Cronbach’s alpha = 0.85). Neighbourhood contacts is the neighbourhood average of the seven-point item ‘How often do you have personal contact with a person in your neighbourhood’, which was recoded from low (0) to high (6). Three variables measured family bonds. Firstly, whether or not the respondent lived with a child and/or partner. Secondly, whether or not the respondent lived with at least one parent.4 Thirdly, parental relationship quality, based on six items measuring contact and level of satisfaction with each parent.5 The first four items were as follows: ‘How much contact do you have with (1) your mother face to face, (2) your father face to face, (3) with your mother via calling, email, text or chat and (4) with your father via calling, email, text or chat?’ All items had the same seven answer categories ranging from ‘(almost) every day’ to ‘never’. The fifth and sixth variable ‘How satisfied or unsatisfied are you at the moment with the relation that you have with your mother?’ and ‘How satisfied or unsatisfied are you at the moment with the relation that you have with your father?’ had five-point response options ranging from ‘very satisfied’ to ‘very unsatisfied’. The first four items were recoded into a five-point scale before being combined with the latter two items into one scale measuring parental relationship quality, ranging from ‘very unsatisfied’ (0) to ‘very satisfied’ (4) (Cronbachs’s alpha = 0.74). Personal religiousness is a five-point item (‘How important is religion for you personally’), ranging from not important at all (0) to very important (4). Finally, the following control variables were used: age, sex, degree of urbanization of the municipality and social desirability. Urbanization follows the official classification by Statistics Netherlands. The options were ‘small village’, ‘big village’, ‘small city’ and ‘big city’. The NELLS dataset includes 5 three-point items from Crowne and Marlowe’s (1960) Social Desirability Scale: (1) ‘There have been occasions when I have taken advantage of someone’, (2) ‘I’m always willing to admit it when I make a mistake’, (3) ‘I sometimes feel resentful when I don’t get my way’, (4) ‘I am always courteous, even to people who are disagreeable’ and (5) ‘No matter who I’m talking to, I’m always a good listener’. We averaged these items into a scale ranging from 0 to 2. The reliability of the scale was relatively low (alpha = 0.49) and could not be improved by dropping items. Analytical method We compared self-reported offending rates from the NELLS to population suspect rates by Statistics Netherlands, both with and without controls for social desirability. Subsequently, all hypotheses were tested in Stata 13 (StataCorp LLC, Texas, USA) using negative binomial regression with robust standard errors allowing for clustering of observations at the neighbourhood level, using the preexisting NELLS population weights. Robust regression was preferred to multilevel regression, because it was impossible to conduct weighted multilevel analyses (the NELLS data only include individual-level weights). Clustering of observations was limited at both the neighbourhood (intraclass correlation [ICC] = 0.077) and the municipality level (ICC = 0.025) and was more of a ‘nuisance’ than a focal point of interest (cf. Cameron and Miller 2015). Negative binomial regression was used because the dependent variable—the total amount of self-reported crimes committed in the last 12 months—was skewed and overdispersed (cf. Land et al. 1996): most respondents did not report crimes, a significant minority reported one or a few crimes and a small minority reported many crimes. In all models, we controlled age, sex and degree of urbanization of the municipality, thus excluding basic demographic explanations for ethnic differences in criminal behaviour. We first estimated a model with four dummies indicating the respondent’s ethnic group by generation (being of native Dutch origin was a reference category), and with sex, age and urbanization as controls. We added relevant variables block by block, while examining how their addition affected the coefficients of the dummies indicating ethnic minority membership by generation. Using seemingly unrelated estimation, it was tested whether the changes in coefficients from one model to the next were statistically significant. We estimated all models with and without the social desirability control. For each model, we report coefficients of the dummies indicating ethnic origin with and without the social desirability control; the reported coefficients of all other coefficients were obtained while including the social desirability measure. Prior to the analysis, all missing values were deleted, so as to keep the number of observations constant between different models. After selecting ‘Turkish’, ‘Moroccan’ and ‘native Dutch’ respondents under 45 who filled in the questionnaire, which included the questions on criminal behaviour, 4,294 respondents remained. The eventual regression analyses pertain to 4,074 respondents (95%), living in 261 different neighbourhoods across 35 municipalities. A small number of cases (~1%) could not be used because the respondent filled in less than two items measuring delinquency. The remaining missing values are mostly due to missing information on parental relationship quality (about 4.5%) and, less so, on perceived ethnic discrimination and personal religiousness variable (less than 1%). The respondents refusing to state their income were classified as missing (about 10%) and kept in the analysis by creating a dummy ‘income missing’. Validity Self-reported offending is assumed to be a reasonably valid measure of ‘less serious’ criminal behaviour (cf. Jolliffe et al. 2003), and we are fairly confident that the NELLS data measure ethnic differences in relatively minor forms of crime reasonably well and are an important addition to criminal justice data. That position can be explained by discussing two limitations of survey data at some length. Firstly, criminal behaviour, which is a sensitive subject for most people, is likely to be underreported, possibly leading to social desirability respondent bias (SDRB). Various techniques have been developed to control SDRB, and we followed Saunders’ (1991) recommendation to introduce the Marlowe–Crowne scale as a control variable in multiple regression analyses (also see Sutton and Farrall 2005). It was checked whether there were interaction effects between ethnic origin and social desirability on self-reported offending, which was not the case, suggesting that the items measuring criminal behaviour were not differentially sensitive for minority respondents with an inclination to social desirability than for such respondents among the native Dutch. It should be mentioned, however, that there is still the possibility that the items measuring criminal behaviour were relatively sensitive for all respondents of immigrant origin regardless of their score on the social desirability scale, e.g., because they feared that reporting criminal behaviour would harm the image of the immigrant group. The Marlow–Crowne scale has been criticized for being confounded with genuine individual differences in religiosity (Watson et al. 1986) and prosocial behaviour (Mills and Kroner 2005). Introducing it as a control variable may therefore represent a kind of ‘overcontrol’; it could lead us, e.g., to underestimate the protective effects of ethnic minority status on crime via religiousness and (other) variables that are mediated by prosocial attitudes, such as familial bonds. We therefore estimated separate models with and without the social desirability control. The results without the social desirability control are likely to represent an underestimation of the crime involvement of respondents with minority status relative to ethnic majority respondents. The results with the social desirability control are expected to represent an overestimation of criminal behaviour among ethnic minorities relative to the majority group, given that the former are expected to score relatively high on religiousness and familial bonds. SDRB was also reduced in other ways: (1) the questionnaire only included questions about relatively ‘minor’ crimes, (2) respondents were asked to answer a broad gamut of sociological questions, not just the sensitive items on crime and (3) the delinquency items were included in the written questionnaire, so as to prevent interviewer effects. A second methodological concern is that high-rate offenders are unlikely to participate in surveys (Junger-Tas and Marshall 1999). Such a bias would be especially problematic should there be ethnic differences in the sampling and response rates of such offenders. For example, if those committing crimes with ethnic minority status had elevated chances of being imprisoned, they would have a lower chance to be sampled, potentially leading to an underestimation of criminal behaviour among the ethnic minorities. Fortunately, Dutch imprisonment rates are too low to create a meaningful non-response. Even among ‘Moroccans’, the ethnic group with the highest imprisonment rate, no more than 3 per 1,000 persons were imprisoned on a given day in 2011 (no data for 2010).6 Ethnic differences in response rates were limited: the overall response rates among the Turkish and Moroccans Dutch and other respondents (mostly native Dutch) were 46%, 50% and 56% respectively. The influence of undersampling of potential offenders, such as males, was reduced by using the NELLS population weights. Although all this gives some confidence in the NELLS, it should be stressed that it is possible to remain somewhat agnostic about its ability to measure the ‘true’ rates of relatively minor criminal behaviour in different ethnic groups, while still identifying central risk and protective factors for ethnic minorities and the majority group, which was the second motivation underlying our study. The coefficients of the dummies indicating the effects of ethnicity on crime may be too low, or too high, but the changes in these coefficients from model to model still provide important evidence about risk and protective factors. A dummy going down from one model to the next indicates that the added variables tend to put the minority group at risk of criminal behaviour compared to the ethnic majority group; a dummy going up from one model to the next indicates that the added variables are protective for the minorities. Results Self-reported offending versus police suspect data Table 1 compares the NELLS self-reported offending rates by ethnic group and generation to population suspect rates by Statistics Netherlands for a similar age span. We also report figures on self-reported offending that are adjusted for social desirability (see note for details).7 Table 1 Crime suspects (12–45 years) and self-reported criminal behaviour (15–45 years) by ethnic group and generation (in 2010) Turkey first generation Turkey second generation Morocco first generation Morocco second generation Native Dutch r (correlation with suspect rates) Population  Persons 114,648 118,168 99,845 97,035 5,302,320  % Suspects 3.9 5.8 5.4 11.1 2.0 NELLS 590 352 607 363 2,340  % Reported a crime 5.7 12.5 7.1 9.7 8.5 0.35  % Reported a crime (low social desirability) 9.7 18.2 12.8 14.3 11.1 0.46  Average no. of crimes per respondent 0.14 0.23 0.23 0.20 0.15 0.49  Average no. of crimes per respondent (low social desirability) 0.22 0.39 0.36 0.32 0.21 0.52 Turkey first generation Turkey second generation Morocco first generation Morocco second generation Native Dutch r (correlation with suspect rates) Population  Persons 114,648 118,168 99,845 97,035 5,302,320  % Suspects 3.9 5.8 5.4 11.1 2.0 NELLS 590 352 607 363 2,340  % Reported a crime 5.7 12.5 7.1 9.7 8.5 0.35  % Reported a crime (low social desirability) 9.7 18.2 12.8 14.3 11.1 0.46  Average no. of crimes per respondent 0.14 0.23 0.23 0.20 0.15 0.49  Average no. of crimes per respondent (low social desirability) 0.22 0.39 0.36 0.32 0.21 0.52 View Large Table 1 Crime suspects (12–45 years) and self-reported criminal behaviour (15–45 years) by ethnic group and generation (in 2010) Turkey first generation Turkey second generation Morocco first generation Morocco second generation Native Dutch r (correlation with suspect rates) Population  Persons 114,648 118,168 99,845 97,035 5,302,320  % Suspects 3.9 5.8 5.4 11.1 2.0 NELLS 590 352 607 363 2,340  % Reported a crime 5.7 12.5 7.1 9.7 8.5 0.35  % Reported a crime (low social desirability) 9.7 18.2 12.8 14.3 11.1 0.46  Average no. of crimes per respondent 0.14 0.23 0.23 0.20 0.15 0.49  Average no. of crimes per respondent (low social desirability) 0.22 0.39 0.36 0.32 0.21 0.52 Turkey first generation Turkey second generation Morocco first generation Morocco second generation Native Dutch r (correlation with suspect rates) Population  Persons 114,648 118,168 99,845 97,035 5,302,320  % Suspects 3.9 5.8 5.4 11.1 2.0 NELLS 590 352 607 363 2,340  % Reported a crime 5.7 12.5 7.1 9.7 8.5 0.35  % Reported a crime (low social desirability) 9.7 18.2 12.8 14.3 11.1 0.46  Average no. of crimes per respondent 0.14 0.23 0.23 0.20 0.15 0.49  Average no. of crimes per respondent (low social desirability) 0.22 0.39 0.36 0.32 0.21 0.52 View Large Both minorities turn out to have much higher suspect rates than expected given their self-reported offending rates. The official suspect rate among the two minority groups varies between 3.9% (first-generation Turks) and 11.1% (second-generation Moroccans), which is between 2.0 and 5.6 times higher than for the native Dutch (2.0) in this age span. When social desirability is not controlled, the self-reported crime rates are actually lower among first-generation immigrants than among native Dutch, and they are only 1.1 (second-generation Moroccans) and 1.5 (second-generation Turks) times higher among the second generation. Adjusting the rates for social desirability leads to higher self-reported offending scores for all groups but changes the rates of first-generation immigrants the most: their rates become comparable to the self-reported rates among the native Dutch. The elevated second-generation self-reported crime rates are associated to their relatively young age within the 15–45 age span (see Model 1, next section). The two final rows in Table 1 show the ‘incidence density’, i.e. the average number of crimes per respondent specified by ethnic group and generation (the criminal justice equivalent for this figure is unavailable). The self-reported incidence density after adjustment for social desirability shows the highest correlation with the percentage of crime suspects (r = 0.52, see the last column in Table 1). If we, therefore, assume that the self-reported incidence density under low social desirability represents the best measurement of the true incidence of criminal behaviour, the percentage of crime suspects among the minorities is still between 1.6 (first-generation Moroccans and second-generation Turks), 1.9 (first-generation Turks) and 3.6 (second-generation Moroccans) higher than the NELLS data would suggest.8 Descriptives The remainder of the Results section pertains to the NELLS data. Table 2 shows means of the dependent and independent variables by ethnic group and generation. All figures pertain to respondents without missing values. For that reason, the average number of reported crimes by respondent and ethnic group differs somewhat from Table 1. Table 2 Weighted means of the dependent and independent variables by ethnic group and generation Turkey first generation (N = 539) Turkey second generation (N = 347) Morocco first generation (N = 563) Morocco second generation (N = 348) Native Dutch (N = 2,277) Dependent variable  Self-reported crimes (av.) 0.13 0.23 0.22 0.19 0.15 Independent variables  Men (%) 50.7 48.2 49.8 46.4 50.5  Age (av.) 33.1 23.8 32.8 22.4 29.3  Large city (%) 65.8 54.8 57.2 50.8 21.1  City (%) 20.5 27.1 30.6 33.6 24.3  Town (%) 13.1 17.1 12.0 15.3 20.0  Rural municipality (%) 0.6 1.0 0.3 0.3 34.6  Income low (%) 39.7 56.0 43.0 65.0 36.4  Income middle (%) 30.1 17.4 32.2 13.2 19.6  Income high (%) 21.0 13.0 14.0 10.8 35.6  Income unknown (%) 9.2 13.6 10.8 11.0 8.4  Primary education (%) 50.6 32.9 52.6 27.7 21.1  Secondary education (%) 27.8 44.2 31.4 50.1 39.1  Tertiary education (%) 21.6 22.9 16.0 22.2 39.8  Living with partner and/or child (%) 72.7 33.9 73.1 21.7 44.1  Living with parent(s) (%) 14.4 49.6 11.9 63.1 28.3  Living without partner, children or parents (%) 12.9 16.5 15.0 15.2 27.6  Parental relationship quality (%) 2.7 3.1 2.7 3.2 2.9  Neighbourhood disadvantage (av.) 1.1 0.8 1.0 0.8 −0.6  Collective efficacy (av.) 2.2 2.2 2.2 2.2 2.4  Neighbourhood contacts (av.) 3.8 3.9 3.9 4.0 3.8  Perceived discrimination (av.) 0.3 0.3 0.3 0.4 0.0  Religiousness (av.) 3.4 3.3 3.7 3.6 1.6  Social desirability (av.) 1.5 1.3 1.5 1.4 1.2 Turkey first generation (N = 539) Turkey second generation (N = 347) Morocco first generation (N = 563) Morocco second generation (N = 348) Native Dutch (N = 2,277) Dependent variable  Self-reported crimes (av.) 0.13 0.23 0.22 0.19 0.15 Independent variables  Men (%) 50.7 48.2 49.8 46.4 50.5  Age (av.) 33.1 23.8 32.8 22.4 29.3  Large city (%) 65.8 54.8 57.2 50.8 21.1  City (%) 20.5 27.1 30.6 33.6 24.3  Town (%) 13.1 17.1 12.0 15.3 20.0  Rural municipality (%) 0.6 1.0 0.3 0.3 34.6  Income low (%) 39.7 56.0 43.0 65.0 36.4  Income middle (%) 30.1 17.4 32.2 13.2 19.6  Income high (%) 21.0 13.0 14.0 10.8 35.6  Income unknown (%) 9.2 13.6 10.8 11.0 8.4  Primary education (%) 50.6 32.9 52.6 27.7 21.1  Secondary education (%) 27.8 44.2 31.4 50.1 39.1  Tertiary education (%) 21.6 22.9 16.0 22.2 39.8  Living with partner and/or child (%) 72.7 33.9 73.1 21.7 44.1  Living with parent(s) (%) 14.4 49.6 11.9 63.1 28.3  Living without partner, children or parents (%) 12.9 16.5 15.0 15.2 27.6  Parental relationship quality (%) 2.7 3.1 2.7 3.2 2.9  Neighbourhood disadvantage (av.) 1.1 0.8 1.0 0.8 −0.6  Collective efficacy (av.) 2.2 2.2 2.2 2.2 2.4  Neighbourhood contacts (av.) 3.8 3.9 3.9 4.0 3.8  Perceived discrimination (av.) 0.3 0.3 0.3 0.4 0.0  Religiousness (av.) 3.4 3.3 3.7 3.6 1.6  Social desirability (av.) 1.5 1.3 1.5 1.4 1.2 View Large Table 2 Weighted means of the dependent and independent variables by ethnic group and generation Turkey first generation (N = 539) Turkey second generation (N = 347) Morocco first generation (N = 563) Morocco second generation (N = 348) Native Dutch (N = 2,277) Dependent variable  Self-reported crimes (av.) 0.13 0.23 0.22 0.19 0.15 Independent variables  Men (%) 50.7 48.2 49.8 46.4 50.5  Age (av.) 33.1 23.8 32.8 22.4 29.3  Large city (%) 65.8 54.8 57.2 50.8 21.1  City (%) 20.5 27.1 30.6 33.6 24.3  Town (%) 13.1 17.1 12.0 15.3 20.0  Rural municipality (%) 0.6 1.0 0.3 0.3 34.6  Income low (%) 39.7 56.0 43.0 65.0 36.4  Income middle (%) 30.1 17.4 32.2 13.2 19.6  Income high (%) 21.0 13.0 14.0 10.8 35.6  Income unknown (%) 9.2 13.6 10.8 11.0 8.4  Primary education (%) 50.6 32.9 52.6 27.7 21.1  Secondary education (%) 27.8 44.2 31.4 50.1 39.1  Tertiary education (%) 21.6 22.9 16.0 22.2 39.8  Living with partner and/or child (%) 72.7 33.9 73.1 21.7 44.1  Living with parent(s) (%) 14.4 49.6 11.9 63.1 28.3  Living without partner, children or parents (%) 12.9 16.5 15.0 15.2 27.6  Parental relationship quality (%) 2.7 3.1 2.7 3.2 2.9  Neighbourhood disadvantage (av.) 1.1 0.8 1.0 0.8 −0.6  Collective efficacy (av.) 2.2 2.2 2.2 2.2 2.4  Neighbourhood contacts (av.) 3.8 3.9 3.9 4.0 3.8  Perceived discrimination (av.) 0.3 0.3 0.3 0.4 0.0  Religiousness (av.) 3.4 3.3 3.7 3.6 1.6  Social desirability (av.) 1.5 1.3 1.5 1.4 1.2 Turkey first generation (N = 539) Turkey second generation (N = 347) Morocco first generation (N = 563) Morocco second generation (N = 348) Native Dutch (N = 2,277) Dependent variable  Self-reported crimes (av.) 0.13 0.23 0.22 0.19 0.15 Independent variables  Men (%) 50.7 48.2 49.8 46.4 50.5  Age (av.) 33.1 23.8 32.8 22.4 29.3  Large city (%) 65.8 54.8 57.2 50.8 21.1  City (%) 20.5 27.1 30.6 33.6 24.3  Town (%) 13.1 17.1 12.0 15.3 20.0  Rural municipality (%) 0.6 1.0 0.3 0.3 34.6  Income low (%) 39.7 56.0 43.0 65.0 36.4  Income middle (%) 30.1 17.4 32.2 13.2 19.6  Income high (%) 21.0 13.0 14.0 10.8 35.6  Income unknown (%) 9.2 13.6 10.8 11.0 8.4  Primary education (%) 50.6 32.9 52.6 27.7 21.1  Secondary education (%) 27.8 44.2 31.4 50.1 39.1  Tertiary education (%) 21.6 22.9 16.0 22.2 39.8  Living with partner and/or child (%) 72.7 33.9 73.1 21.7 44.1  Living with parent(s) (%) 14.4 49.6 11.9 63.1 28.3  Living without partner, children or parents (%) 12.9 16.5 15.0 15.2 27.6  Parental relationship quality (%) 2.7 3.1 2.7 3.2 2.9  Neighbourhood disadvantage (av.) 1.1 0.8 1.0 0.8 −0.6  Collective efficacy (av.) 2.2 2.2 2.2 2.2 2.4  Neighbourhood contacts (av.) 3.8 3.9 3.9 4.0 3.8  Perceived discrimination (av.) 0.3 0.3 0.3 0.4 0.0  Religiousness (av.) 3.4 3.3 3.7 3.6 1.6  Social desirability (av.) 1.5 1.3 1.5 1.4 1.2 View Large We find that first-generation immigrants were relatively old, with an average age of 33 years, whereas second-generation immigrants were five to seven years younger than the native Dutch. The majority of the ‘Turks’ and ‘Moroccans’ lived in large cities, against one-quarter of the native Dutch. Both minorities have a considerably lower mean SES than the majority population. The native Dutch are almost twice as likely to live without a partner, child and/or (grand) parent. Parental relationship quality is somewhat higher among the second-generation than the native Dutch, but lower among the first-generation. Ethnic minority status is associated with living in poorer neighbourhoods, with second-generation immigrants being slightly better off than first-generation immigrants. The native Dutch reside in neighbourhoods with somewhat higher average levels of collective efficacy, but there are no clear ethnic differences in neighbourhood contacts. Among ‘Turks’ and ‘Moroccans’, the average level of perceived discrimination is about 0.3 (0.4 for second-generation Moroccans). Both minority groups attach more importance to religion than the native Dutch. The mean value for the native Dutch is 1.6 (between ‘not important’ and ‘important/not important’), whereas the minorities score in the 3.3–3.7 range (‘important’ to ‘very important’). First-generation immigrants score considerably higher on social desirability than the native Dutch; second-generation immigrants are halfway between first-generation immigrants and the native Dutch. Multivariate results Table 3 presents six negative binomial regression models. Model 1 only includes the four dummies indicating ethnic minority membership by generation, and controls age, sex and urbanization of the municipality and social desirability (thus excluding basic demographic explanations for ethnic differences in criminal behaviour). Between parentheses, we also report the coefficients of the dummies indicating ethnic origin without the social desirability control. Table 3 Determinants of the number of self-reported crimes Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Native Dutch (ref.)  Turkey first generation 0.00 (−0.22) −0.54** (−0.81***) −0.77** (−1.08***) −0.84** (−1.17***) −0.70** (−0.96***) −0.46 (−0.65*)  Turkey second generation 0.39* (0.20) 0.02 (−0.17) −0.15 (−0.36*) −0.17 (−0.38) −0.07 (−0.25) 0.18 (0.07)  Morocco first generation 0.68** (0.42) 0.05 (−0.27) −0.21 (−0.54*) −0.29 (−0.64**) −0.18 (−0.48*) 0.10 (−0.12)  Morocco second generation 0.40 (0.21) −0.01 (−0.21) −0.27 (−0.51) −0.24 (−0.46) −0.16 (−0.38) 0.11 (−0.03) Female (ref.)  Male 1.26*** 1.25*** 1.20*** 1.15*** 1.09*** 1.05***  Age −0.34*** −0.22*** −0.23*** −0.23*** −0.24*** −0.25***  Age squared/100 0.52*** 0.35*** 0.38*** 0.37*** 0.38*** 0.39*** Large city (ref.)  City −0.47** −0.53*** −0.54** −0.34** −0.36** −0.37**  Small town 0.17 −0.01 −0.05 0.17 0.19 0.19  Rural municipality −0.44* −0.76*** −0.78*** −0.45 −0.42 −0.42 Income high (ref.)  Income low 0.87*** 0.83*** 0.78*** 0.64** 0.67**  Income middle 0.55** 0.56** 0.47* 0.45* 0.50**  Income missing 0.78** 0.74** 0.73** 0.71* 0.75** Tertiary education (ref.)  Primary education 1.10*** 1.11*** 1.15*** 1.17*** 1.16***  Secondary education 1.15*** 1.16*** 1.21*** 1.25*** 1.26***  Perceived ethnic discrimination 0.53** 0.57** 0.54** 0.57**  Neighbourhood disadvantage 0.06 0.04 0.04  Collective efficacy −0.17 −0.10 −0.16  Neighbourhood contacts −0.36** −0.31* −0.28* Lives without parent, partner, child (ref.)  Lives with partner and/or child −0.50*** −0.45**  Lives with parent(s) −0.18 −0.22  Parental relationship quality −0.19** −0.16*  Religiousness −0.15**  Social desirability −1.07*** −1.09*** −1.08*** −1.02*** −0.97*** −0.95***  Constant 3.59*** 0.55 0.83*** 2.45* 2.94** 3.23*** Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Native Dutch (ref.)  Turkey first generation 0.00 (−0.22) −0.54** (−0.81***) −0.77** (−1.08***) −0.84** (−1.17***) −0.70** (−0.96***) −0.46 (−0.65*)  Turkey second generation 0.39* (0.20) 0.02 (−0.17) −0.15 (−0.36*) −0.17 (−0.38) −0.07 (−0.25) 0.18 (0.07)  Morocco first generation 0.68** (0.42) 0.05 (−0.27) −0.21 (−0.54*) −0.29 (−0.64**) −0.18 (−0.48*) 0.10 (−0.12)  Morocco second generation 0.40 (0.21) −0.01 (−0.21) −0.27 (−0.51) −0.24 (−0.46) −0.16 (−0.38) 0.11 (−0.03) Female (ref.)  Male 1.26*** 1.25*** 1.20*** 1.15*** 1.09*** 1.05***  Age −0.34*** −0.22*** −0.23*** −0.23*** −0.24*** −0.25***  Age squared/100 0.52*** 0.35*** 0.38*** 0.37*** 0.38*** 0.39*** Large city (ref.)  City −0.47** −0.53*** −0.54** −0.34** −0.36** −0.37**  Small town 0.17 −0.01 −0.05 0.17 0.19 0.19  Rural municipality −0.44* −0.76*** −0.78*** −0.45 −0.42 −0.42 Income high (ref.)  Income low 0.87*** 0.83*** 0.78*** 0.64** 0.67**  Income middle 0.55** 0.56** 0.47* 0.45* 0.50**  Income missing 0.78** 0.74** 0.73** 0.71* 0.75** Tertiary education (ref.)  Primary education 1.10*** 1.11*** 1.15*** 1.17*** 1.16***  Secondary education 1.15*** 1.16*** 1.21*** 1.25*** 1.26***  Perceived ethnic discrimination 0.53** 0.57** 0.54** 0.57**  Neighbourhood disadvantage 0.06 0.04 0.04  Collective efficacy −0.17 −0.10 −0.16  Neighbourhood contacts −0.36** −0.31* −0.28* Lives without parent, partner, child (ref.)  Lives with partner and/or child −0.50*** −0.45**  Lives with parent(s) −0.18 −0.22  Parental relationship quality −0.19** −0.16*  Religiousness −0.15**  Social desirability −1.07*** −1.09*** −1.08*** −1.02*** −0.97*** −0.95***  Constant 3.59*** 0.55 0.83*** 2.45* 2.94** 3.23*** ***p < 0.01; **p < 0.05; *p < 0.10. View Large Table 3 Determinants of the number of self-reported crimes Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Native Dutch (ref.)  Turkey first generation 0.00 (−0.22) −0.54** (−0.81***) −0.77** (−1.08***) −0.84** (−1.17***) −0.70** (−0.96***) −0.46 (−0.65*)  Turkey second generation 0.39* (0.20) 0.02 (−0.17) −0.15 (−0.36*) −0.17 (−0.38) −0.07 (−0.25) 0.18 (0.07)  Morocco first generation 0.68** (0.42) 0.05 (−0.27) −0.21 (−0.54*) −0.29 (−0.64**) −0.18 (−0.48*) 0.10 (−0.12)  Morocco second generation 0.40 (0.21) −0.01 (−0.21) −0.27 (−0.51) −0.24 (−0.46) −0.16 (−0.38) 0.11 (−0.03) Female (ref.)  Male 1.26*** 1.25*** 1.20*** 1.15*** 1.09*** 1.05***  Age −0.34*** −0.22*** −0.23*** −0.23*** −0.24*** −0.25***  Age squared/100 0.52*** 0.35*** 0.38*** 0.37*** 0.38*** 0.39*** Large city (ref.)  City −0.47** −0.53*** −0.54** −0.34** −0.36** −0.37**  Small town 0.17 −0.01 −0.05 0.17 0.19 0.19  Rural municipality −0.44* −0.76*** −0.78*** −0.45 −0.42 −0.42 Income high (ref.)  Income low 0.87*** 0.83*** 0.78*** 0.64** 0.67**  Income middle 0.55** 0.56** 0.47* 0.45* 0.50**  Income missing 0.78** 0.74** 0.73** 0.71* 0.75** Tertiary education (ref.)  Primary education 1.10*** 1.11*** 1.15*** 1.17*** 1.16***  Secondary education 1.15*** 1.16*** 1.21*** 1.25*** 1.26***  Perceived ethnic discrimination 0.53** 0.57** 0.54** 0.57**  Neighbourhood disadvantage 0.06 0.04 0.04  Collective efficacy −0.17 −0.10 −0.16  Neighbourhood contacts −0.36** −0.31* −0.28* Lives without parent, partner, child (ref.)  Lives with partner and/or child −0.50*** −0.45**  Lives with parent(s) −0.18 −0.22  Parental relationship quality −0.19** −0.16*  Religiousness −0.15**  Social desirability −1.07*** −1.09*** −1.08*** −1.02*** −0.97*** −0.95***  Constant 3.59*** 0.55 0.83*** 2.45* 2.94** 3.23*** Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Native Dutch (ref.)  Turkey first generation 0.00 (−0.22) −0.54** (−0.81***) −0.77** (−1.08***) −0.84** (−1.17***) −0.70** (−0.96***) −0.46 (−0.65*)  Turkey second generation 0.39* (0.20) 0.02 (−0.17) −0.15 (−0.36*) −0.17 (−0.38) −0.07 (−0.25) 0.18 (0.07)  Morocco first generation 0.68** (0.42) 0.05 (−0.27) −0.21 (−0.54*) −0.29 (−0.64**) −0.18 (−0.48*) 0.10 (−0.12)  Morocco second generation 0.40 (0.21) −0.01 (−0.21) −0.27 (−0.51) −0.24 (−0.46) −0.16 (−0.38) 0.11 (−0.03) Female (ref.)  Male 1.26*** 1.25*** 1.20*** 1.15*** 1.09*** 1.05***  Age −0.34*** −0.22*** −0.23*** −0.23*** −0.24*** −0.25***  Age squared/100 0.52*** 0.35*** 0.38*** 0.37*** 0.38*** 0.39*** Large city (ref.)  City −0.47** −0.53*** −0.54** −0.34** −0.36** −0.37**  Small town 0.17 −0.01 −0.05 0.17 0.19 0.19  Rural municipality −0.44* −0.76*** −0.78*** −0.45 −0.42 −0.42 Income high (ref.)  Income low 0.87*** 0.83*** 0.78*** 0.64** 0.67**  Income middle 0.55** 0.56** 0.47* 0.45* 0.50**  Income missing 0.78** 0.74** 0.73** 0.71* 0.75** Tertiary education (ref.)  Primary education 1.10*** 1.11*** 1.15*** 1.17*** 1.16***  Secondary education 1.15*** 1.16*** 1.21*** 1.25*** 1.26***  Perceived ethnic discrimination 0.53** 0.57** 0.54** 0.57**  Neighbourhood disadvantage 0.06 0.04 0.04  Collective efficacy −0.17 −0.10 −0.16  Neighbourhood contacts −0.36** −0.31* −0.28* Lives without parent, partner, child (ref.)  Lives with partner and/or child −0.50*** −0.45**  Lives with parent(s) −0.18 −0.22  Parental relationship quality −0.19** −0.16*  Religiousness −0.15**  Social desirability −1.07*** −1.09*** −1.08*** −1.02*** −0.97*** −0.95***  Constant 3.59*** 0.55 0.83*** 2.45* 2.94** 3.23*** ***p < 0.01; **p < 0.05; *p < 0.10. View Large On average, first-generation Moroccans reported more crimes than the native Dutch when age, sex, municipality and social desirability are held constant. The effects of the other dummies indicating ethnic minority status are positive, too, but are not significant. In this model, being a first-generation Moroccan is associated with a 1.97 higher incidence rate ratio (not shown), meaning that the number of crimes per respondent is approximately twice the rate for the ethnic majority when age, sex, municipality and social desirability are held constant. As a result, we only partially accept hypothesis 1 (First- and second-generation Turks and Moroccans report higher levels of criminal behaviour than respondents of native Dutch origin). Model 2 adds income and education, which show the expected negative relationship with crime. Due to their inclusion, the effects of the dummies indicating ethnicity go down significantly, confirming hypothesis 2 (The effect of being a first- or second-generation Turk or Moroccan on self-reported offending diminishes after controlling SES). Being a first-generation Moroccan no longer has an independent positive effect on crime, suggesting that the effect in Model 1 is related to the lower SES of this group. Adding SES leads to a significant negative effect of being first-generation Turk: Turkish immigrants report fewer crimes than would be expected on the basis of their SES. By contrast, Blom and Jennissen (2014), who conducted logistic regression analyses using police data, found that ethnic minority status, including being a first-generation Turk, was associated with a considerably higher odds of being a crime suspect after controlling SES: the ‘residual’ odds ratio of ethnic minority status on being a crime suspect, after controlling age, sex and urbanization of the municipality, household structure and SES, varied between 1.6 (first-generation Turks) and 4.0 (second-generation Moroccans with two foreign-born parents). Model 3 adds perceived ethnic discrimination, which is associated with a higher number of reported crimes (b = 0.52; p < 0.05). Like in Model 2, the effects of the dummies indicating ethnic minority status go down significantly, indicated by the coefficients getting closer to zero. In other words, Moroccan and Turkish Dutch experiencing discrimination reported more crimes than Moroccans and Turks (and native Dutch) not experiencing it, thereby confirming hypothesis 5 (The effect of being a first- or second-generation Turk or Moroccan on self-reported crime diminishes after controlling perceived ethnic discrimination). Model 4 adds three neighbourhood measures: neighbourhood disadvantage, collective efficacy and contacts. The coefficients of these measures are in the expected direction, but only neighbourhood contacts reach conventional levels of statistical significance. Their addition does lead to slightly lower negative coefficients for first-generation immigrants in particular, but the changes are not significant, leading us to reject hypothesis 4 (The effect of being a first or second-generation Moroccan or Turk on self-reported crime diminishes after controlling differences in neighbourhood disadvantage and control). Apparently, these minorities have become relatively established in terms of neighbourhood ties. The negative effect of being a first-generation Moroccan is significant at the p = 0.05 level in this model, when the social desirability measure is not included. We now turn to two factors possibly explaining why first-generation immigrants reported fewer crimes than would be expected on the basis of their social disadvantage. Model 5 adds family bonds. Living with a partner or child (b = −0.50; p < 0.01) and good parental relationships (b = −0.19, p < 0.05) indeed have a negative effect on crime. Due to their inclusion, the effects of the dummies indicating ethnic minority status go up significantly, indicating that stronger family bonds among the minorities counterbalance the potentially criminogenic effects of their social disadvantage, leading us to accept hypothesis 4a (The effect of being a first- or second-generation Turk or Moroccan on self-reported crime increases after controlling family bonds) and to reject hypothesis 4b. Finally, Model 6 adds religiosity, which shows the expected negative relationship with self-reported crime (b = −0.15, p < 0.05), and also significantly increases the coefficients of the dummies indicating ethnic origin, leading us to accept hypothesis 6 (The effect of being a first- or second-generation Turk or Moroccan on self-reported crime increases after controlling religiousness). Religiousness, too, tends to counterbalance the criminogenic effects of minority disadvantage. Conclusion and Discussion One of the puzzles in the field of migration, ethnicity and crime is that first- and second-generation immigrants are generally overrepresented in criminal justice data in Europe, but not, or not as much, in traditional immigration countries, including the United States. We reasoned that official crime figures are likely to be biased against newcomers and their children, and that such biases may explain part of the transatlantic differences in the relationship between immigration and crime, as indicated by official data. Consequently, we analysed the NELLS data so as to examine whether the two largest ethnic minorities in the Netherlands—the Turkish and Moroccan Dutch—also self-report more criminal offending than the native Dutch. An additional reason for analysing the NELLS data was to examine underresearched mechanisms that are difficult or impossible to consider with administrative data, including perceived ethnic discrimination, family bonds and religiousness. The NELLS also allowed us to assess and control ethnic and generational differences in social desirability. We find two ‘minority paradoxes’, both of which suggest that transatlantic differences in the relationship between migration and crime may well be smaller than official crime data indicate. The first paradox is that findings regarding ethnic differences in criminal behaviour turn out to be highly sensitive to the data source used. Analyses using police data have found rather large positive effects of being Turkish or Moroccan Dutch on the chances of being a crime suspect, even when sex, age, urbanization and SES are controlled (e.g. Blom and Jennissen 2014). In the NELLS, however, only first-generation Moroccan Dutch reported significantly more crimes than the native Dutch when age, sex and urbanization of the municipality (and social desirability) were controlled. Both minority groups reported similar to lower rates than the native Dutch when group differences in SES were controlled. Although first-generation immigrants in particular scored relatively high on social desirability, ethnic and generational differences in social desirability were found to only partially account for the apparent contradictions between self-reported crime rates and police figures. When the number of crimes is adjusted for group differences in social desirability, the percentage of crime suspects among the minorities is still between 1.6 and 3.6 times higher than expected on the basis of the NELLS data. These disparities between survey and official crime data warrant more research along three lines. Firstly, there is the possibility that official data are—seriously—biased against ethnic minorities, warranting more research in the spirit of Weenink (2009) and Rojek et al. (2010), especially for the early stages of policing, including reactive policing. How does the ethnic and class origin of offenders, victims, bystanders and police—and their complex interactions—structure the selection of cases that end up being formally sanctioned? Secondly, the survey data may still understate criminal behaviour among ethnic minorities relative to the majority group in spite of the techniques that were used to promote construct validity and representativeness. Perhaps high-rate offenders among those with minority status were less likely to cooperate with the survey than majority ‘criminals’ with a similar high crime incidence; perhaps we did not eliminate all ethnic and generational differences in social desirability bias. Given the latter possibility, it is advisable to also use other methods that have been developed to deal with social desirability, such as list experiments (cf. Kim and Kim 2015). Thirdly, there is the possibility that both administrative and self-report data measure criminal behaviour reasonably well, but measure different types of crime. Perhaps ethnic differences in the European context mostly pertain to relatively serious crimes that are registered by the police, whereas surveys, which focus more on minor forms of crime, predominantly show ethnic similarities. Better longitudinal studies are therefore in order that examine whether practices of selective law enforcement lead to an escalation of deviance among immigrants and their descendants, especially in contexts that are relatively hostile to migration and ethnic diversity (cf. Marx 1981; Grattet 2011). In the absence of additional research, we should not take ethnic differences in official crime figures as definitive proof for large ethnic differences in criminal behaviour. We should be especially cautious in assuming that immigrants and their descendants commit more crimes than can be accounted for by general demographic and socio-economic factors. The second paradox is that first-generation immigrants reported fewer crimes than would be expected given their disadvantage, which resembles findings for the United States based on official crime data (cf. Sampson and Bean 2006, Bersani 2014). On the basis of the present analysis, we propose to label that protective effect the ‘righteous migrant effect’, which contends that the criminogenic effects of ethnic minorities’ social disadvantage are partially or fully annulled by their stronger family bonds and religiousness under the influence of cultural traditionalism in migrants’ countries of origin. By coining that term, we mean to broaden and ‘decontextualize’ the notion of a Latino or Hispanic paradox (Martinez 2002; Sampson and Bean 2006). In addition, we intend to create an analogy to the so-called ‘healthy migrant effect’ in medical sociology, which argues that migrants are healthier than would be expected given their SES (Razum et al. 1998). Like the healthy migrant effect, the righteous migrant effect seems to be related to selective migration, albeit in different ways. Although the healthy migrant effect is related to positive selection of (labour) migrants at the individual level, the righteous migrant effect seems to be primarily associated to unintended group-level selection: most Turkish and Moroccan immigrants, including those who were not recruited as labour migrants, originate from rural societies with relatively high levels of informal social control. The persistence of the righteous migrant effect over the life course may well be driven by (psycho)cultural processes that also explain the relative durability of the healthy migrant effect; Durkheim (1915) already pointed to the similarities in the social causation of health and crime, and highlighted the importance of family bonds and religiousness. Our findings also suggest a complex relationship between immigrant integration and crime. In general sociology, the term ‘integration’ is generally used in a Durkheimean sense: it refers to the different social ties of individuals and groups that enable social order. However, in the sociology of migration—especially in the United States—the term is often used as synonym for assimilation, as the gradual diminishing of the socio-economic and cultural differences between immigrant groups and (segments of) the receiving society (cf. Portes and Zhou 1993; Alba and Nee 1997). So understood, American scholars are certainly right in arguing that a relative lack of assimilation may, in some ways, protect migrants’ and their offspring against criminal behaviour. In the present analysis, minorities’ relatively traditional family structure and religiousness are found to be at least partially responsible for that effect. It is nonetheless misleading to suggest that assimilation only has the effect of fostering criminal behaviour. Structural assimilation to the mainstream—as indicated by diminishing socio-economic disparities between the ethnic minorities and the ethnic majority and diminished levels of interethnic discrimination—certainly reduces crime among poorer migrants and their descendants. If there is a ‘paradox of assimilation’ (Rumbaut and Ewing 2007), it is not that crime increases with assimilation; rather, different forms of assimilation produce opposite effects. Acknowledgements An earlier version of this paper was presented at the annual meeting of the American Sociological Association. We would like to thank two anonymous reviewers for their valuable comments. References Alba , R. and Nee , V . ( 1997 ), ‘ Rethinking Assimilation Theory for a New Era of Immigration ’, International Migration Review , 31 : 826 – 74 . Google Scholar Crossref Search ADS PubMed Agnew , R . ( 2001 ), ‘ Building on the Foundation of General Strain Theory: Specifying the Types of Strain Most Likely to Lead to Crime and Delinquency ’, Journal of Research in Crime and Delinquency , 38 : 319 – 61 . Google Scholar Crossref Search ADS Andriessen , I. , Nievers , E. , Dagevos , J. and Faulk , L . ( 2012 ), ‘ Ethnic Discrimination in the Dutch Labor Market Its Relationship With Job Characteristics and Multiple Group Membership ’, Work and Occupations , 39 : 237 – 69 . Google Scholar Crossref Search ADS Andriessen , I. , Fernee , H. and Wittebrood , K . ( 2014 ), ‘ Ervaren Discriminatie in Nederland’ . SCP . Bauman , Z . ( 2003 ), Liquid Love: On the Frailty of Human Bonds . Polity Press . Bersani , B . ( 2014 ), ‘ An Examination of First and Second Generation Immigrant Offending Trajectories’ , Justice Quarterly , 31 : 315 – 43 . Google Scholar Crossref Search ADS Black , D . ( 1970 ), ‘ Production of Crime Rates ’, American Sociological Review , 35 : 733 – 48 . Google Scholar Crossref Search ADS Blokland , A. , Nieuwbeerta , P. and Bernasco , W . ( 2010 ), ‘ Criminaliteit en Etniciteit. Criminele Carrières van Autochtone en Allochtone Jongeren uit het Geboortecohort 1984 ’, Tijdschrift voor Criminologie , 52 : 122 – 152 . Blom , M. and Jennissen , R . ( 2014 ), ‘ The Involvement of Different Ethnic Groups in Various Types of Crime in the Netherlands ’, European Journal on Criminal Policy and Research , 20 : 51 – 72 . Google Scholar Crossref Search ADS Bovenkerk , F. and Fokkema , T . ( 2016 ), ‘ Crime Among Young Moroccan Men in the Netherlands: Does Their Regional Origin Matter ?’, European Journal of Criminology , 13 : 352 – 71 . Google Scholar Crossref Search ADS Bucerius , S. and Tonry , M. , eds ( 2014 ), The Oxford Handbook of Ethnicity, Crime and Immigration . Oxford University Press . Cameron , A. and Miller , D . ( 2015 ), ‘ A Practitioner’s Guide to Cluster Robust Inference ’, Journal of Human Resources , 50 : 317 – 72 . Google Scholar Crossref Search ADS Çankaya , S . ( 2015 ), ‘ De Politiële Surveillance van Ras en Etniciteit ’, in L. Moor, J. Janssen, M. Easton and A. Verhage , eds, Cahiers Politiestudies: Etnisch Profileren en Interne Diversiteit bij de Politie , 13 – 33 . Maklu . Cornelius , W. , Tsuda , T. , Martin , P. And Hollifield , J ., eds ( 2004 ), Controlling Immigration: A Global Perspective . Stanford University Press . Crowne , D. and Marlowe , D . ( 1960 ), ‘ A New Scale of Social Desirability Independent of Psychopathology ’, Journal of Consulting Psychology , 24 : 349 – 54 . Google Scholar Crossref Search ADS PubMed De Graaf , P. , Kalmijn , M. , Kraaykamp , G. and Monden , C . ( 2010 ), The Netherlands Longitudinal Lifecourse Study (NELLS Wave 1). Dataset . Tilburg University and Radboud University Nijmegen . De Haen-Marshall , I ., ed. ( 1997 ), Minorities, Migrants and Crime: Diversity and Similarity across Europe and the United States . Sage . De Hart , J. and Dekker , P . ( 2013 ), ‘ Religion, Spirituality and Civic Participation’ , in J. De Hart , P. Dekker and L. Halman , eds, Religion and Civil Society in Europe , 169 – 88 . Springer . Durkheim , E . ( 1915 ), The Elementary Forms of Religious Life . Free Press . Engbersen , G. , Leerkes , A. and Snel , E . ( 2014 ), ‘ Ethnicity, Crime and Immigration in the Netherlands’ , in S. Bucerius and M. Tonry , eds, The Oxford Handbook of Ethnicity, Crime, and Immigration , 766 – 90 . Oxford University Press . Ellis , L. and McDonald , J . ( 2001 ), ‘ Crime, Delinquency, and Social Status ’, Journal of Offender Rehabilitation , 32 : 23 – 52 . Google Scholar Crossref Search ADS Esipova , N. , Ray , J. , Pugliese , A. and Tsabutashvili , D . ( 2015 ), How the World Views Immigration , IOM . Grattet , R . ( 2011 ), ‘ Societal Reactions to Deviance ’, Annual Review of Sociology , 37 : 185 – 204 . Google Scholar Crossref Search ADS Heath , A. , Rothon , C. and Kilpi , E . ( 2008 ), ‘ The Second Generation in Western Europe: Education, Unemployment, and Occupational Attainment ’, Annual Review of Sociology , 34 : 211 – 35 . Google Scholar Crossref Search ADS Henneken-Hordijk , I. and Van Gemmert , A . ( 2012 ), Gedetineerd in Nederland 2011 . DJI . Hirschi , T . ( 1969 ), Causes of Delinquency . University of California Press . Hughes , E . ( 1945 ), ‘ Dilemmas and Contradictions of Status ’, American Journal of Sociology , 50 : 353 – 59 . Google Scholar Crossref Search ADS Huijnk , W. and Andriessen , I . ( 2016 ), Integratie in zicht? , SCP . Johnson , B. , Larson , D. , Li , S. and Jang , S . ( 2000 ), ‘ Escaping From the Crime of Inner-Cities: Church Attendance and Religious Salience Among Disadvantaged Youth ’, Justice Quarterly , 17 : 377 – 92 . Google Scholar Crossref Search ADS Jolliffe , D. , Farrington , D. P. , Hawkins , J. D. , Catalano , R. F. , Hill , K. G. and Kosterman , R . ( 2003 ), ‘ Predictive, Concurrent, Prospective and Retrospective Validity of Self-reported Delinquency ’, Criminal Behavior and Mental Health , 13 : 179 – 97 . Google Scholar Crossref Search ADS Junger , M . ( 1990 ), Delinquency and Ethnicity: An Investigation on Social Factors Relating to Delinquency Among Moroccan, Turkish, Surinamese and Dutch Boys . Kluwer . Junger , M. and Polder , W . ( 1992 ), ‘ Some Explanations of Crime Among Four Ethnic Groups in the Netherlands ’, Journal of Quantitative Criminology , 8 : 51 – 78 . Google Scholar Crossref Search ADS Junger-Tas , J. and Marshall , I . ( 1999 ), ‘ The Self-report Methodology in Crime Research ’, Crime and Justice , 25 : 291 – 367 . Google Scholar Crossref Search ADS Killias , M . ( 2009 ), ‘ Paradise Lost? New Trends in Crime and Migration in Switzerland ’, Sociology of Crime, Law and Deviance , 13 : 33 – 45 . Google Scholar Crossref Search ADS Kim , S. H. and Kim S . ( 2015 ), ‘ Ethnic Differences in Social Desirability Bias: Effects on the Analysis of Public Service Motivation ’, Review of Public Personnel Administration , 48 : 444 – 76 . Koopmans , R . ( 2010 ), ‘ Trade-Offs Between Equality and Difference: Immigrant Integration, Multiculturalism and the Welfare State in Cross-National Perspective ’, Journal of Ethnic and Migration Studies , 36 : 1 – 26 . Google Scholar Crossref Search ADS Land , K. , McCall , P. and Nagin , D . ( 1996 ), ‘ A Comparison of Poisson, Negative Binomial, and Semiparametric Mixed Poisson Regression Models ’, Sociological Methods and Research , 24 : 387 – 42 . Google Scholar Crossref Search ADS Leerkes , A. and Bernasco , W . ( 2010 ), ‘ The Spatial Concentration of Illegal Residence and Neighborhood Safety’ , Journal of Urban Affairs , 32 : 367 – 92 . Google Scholar Crossref Search ADS Leerkes , A. and Kulu-Glasgow , I . ( 2011 ), ‘ Playing Hard(er) to Get: The State, International Couples, and the Income Requirement’ , European Journal of Migration and Law , 13 : 95 – 121 . Google Scholar Crossref Search ADS Lucassen , J. and Penninx , R . ( 1997 ), Newcomers: Immigrants and Their Descendants in the Netherlands 1550–1995 . Spinhuis . Lynch , P. and Simon , R . ( 1999 ), ‘ A Comparative Assessment of Criminal Involvement Among Immigrants and Natives Across Seven Nations ’, International Criminal Justice Review , 9 : 1 – 17 . Google Scholar Crossref Search ADS Maliepaard , M. and Gijsbers , M . ( 2012 ), Moslim in Nederland 2012 . SCP . Martinez , R . ( 2002 ), Latino Homicide: Immigration, Violence and Community . Routledge . Marx , G . ( 1981 ), ‘ Ironies of Social Control: Authorities as Contributors to Deviance Through Escalation, Nonenforcement and Covert Facilitation ’, Social Problems , 28 : 221 – 46 . Google Scholar Crossref Search ADS Merton , R . ( 1967 ), Social Theory and Social Structure . Collier-Macmillan . Mills , J. and Kroner , D . ( 2005 ), ‘ An Investigation Into the Relationship Between Socially Desirable Responding and Offender Self-Report’ , Psychological Services , 2 : 70 – 80 . Google Scholar Crossref Search ADS Murad , S. D. , Joung , I. , Lenthe , F. J. , Bengi‐Arslan , L. and Crijnen , A. A . ( 2003 ), ‘ Predictors of Self-reported Problem Behaviours in Turkish Immigrant and Dutch Adolescents in the Netherlands ’, Journal of Child Psychology and Psychiatry , 44 : 412 – 23 . Google Scholar Crossref Search ADS PubMed Mutsaers , P . ( 2014 ), ‘ An Ethnographic Study of the Policing of Internal Borders in the Netherlands: Synergies Between Criminology and Anthropology’ , British Journal of Criminology , 54 : 831 – 48 . Google Scholar Crossref Search ADS O’Brien , R . ( 1987 ), ‘ The Interracial Nature of Violent Crime: A Reexamination’ . American Journal of Sociology , 92 : 817 – 35 Google Scholar Crossref Search ADS Pels , T . ( 2000 ), ‘ Muslim Families from Morocco in the Netherlands: Gender Dynamics and Fathers’ Roles in a Context of Change ’, Current Sociology , 48 : 75 – 93 . Google Scholar Crossref Search ADS Portes , A. and Zhou , M . ( 1993 ), ‘ The New Second-generation: Segmented Assimilation and Its Variants Among Post-1965 Immigrant Youth ’, American Academy of Political and Social Science , 530 : 74 – 96 . Google Scholar Crossref Search ADS Razum , O. , Zeeb , H. , Akgün , H. S. and Yilmaz , S . ( 1998 ), ‘ Low Overall Mortality of Turkish Residents in Germany Persists and Extends Into a Second Generation: Merely a Healthy Migrant Effect ?’, Tropical Medicine & International Health , 3 : 297 – 303 . Google Scholar Crossref Search ADS Röder , A . ( 2015 ), ‘ Immigrants’ Attitudes Toward Homosexuality: Socialization, Religion, and Acculturation in European Host Societies ’, International Migration Review , 49 : 1042 – 70 . Google Scholar Crossref Search ADS Rojek , J. , Rosenfeld , R. and Decker , S . ( 2010 ), ‘ Policing Race: The Racial Stratification of Searches in Police Traffic Stops ’, Criminology , 50 : 993 – 1024 . Google Scholar Crossref Search ADS Rumbaut , R. and Ewing , W . ( 2007 ), The Myth of Immigrant Criminality and the Paradox of Assimilation: Incarceration Rates Among Native and Foreign-Born Men . Immigration Policy Center . Sampson , R . ( 2008 ), ‘ Rethinking Crime and Immigration ’, Contexts , 7 : 28 – 33 . Google Scholar Crossref Search ADS Sampson , R. and Bean , L . ( 2006 ), ‘ Cultural Mechanisms and Killing fields: A Revised Theory of Community-Level Racial Inequality ’, in R. Peterson , L. Krivo and J. Hagan , eds, The Many Colors of Crime: Inequalities of Race, Ethnicity and Crime in America . New York University Press . Sampson , R. Morenoff , J. and Raudenbush , S . ( 2005 ), ‘ Social Anatomy of Racial and Ethnic Disparities in Violence ’, American Journal of Public Health , 95 : 224 – 32 . Google Scholar Crossref Search ADS PubMed Sampson , R. , Raudenbush , F. and Earls , F . ( 1997 ), ‘ Neighborhoods and Violent Crime: A Multilevel Study of Collective Efficacy ’, Science , 277 : 918 – 24 . Google Scholar Crossref Search ADS PubMed Saunders , D . ( 1991 ), ‘ Procedures for Adjusting Self-reports of Violence for Social Desirability Bias’ , Journal of Interpersonal Violence , 6 : 336 – 44 . Google Scholar Crossref Search ADS Simons , R. L. , Chen , Y. F. , Stewart , E. A. and Brody , G. H . ( 2003 ), ‘ Incidents of Discrimination and Risk for Delinquency: A Longitudinal Test of Strain Theory With an African American sample ’, Justice Quarterly , 20 : 827 – 54 . Google Scholar Crossref Search ADS Statistics Netherlands (CBS) ( 2016 ), Jaarrapport Integratie 2016 . CBS . Sutton , R. and Farrall , S . ( 2005 ), ‘ Gender, Socially Desirable Responding and the Fear of Crime. Are Women Really More Anxious About Crime’ . British Journal of Criminology , 45 : 212 – 24 . Google Scholar Crossref Search ADS Tonry , M. , ed. ( 1997 ), Ethnicity, Crime and Immigration: Comparative and Cross-National Perspectives . University of Chicago Press . Van der Leun , J. and Van der Woude , M . ( 2011 ), ‘ Ethnic Profiling in the Netherlands? A Reflection on Expanding Preventive Powers, Ethnic Profiling and a Changing Social and Political Context ’, Policing and Society , 21 : 444 – 55 . Google Scholar Crossref Search ADS Watson , P. J. , Morris , R. J. , Foster , J. E. and Hood , Jr., R. W . ( 1986 ), ‘ Religiosity and Social Desirability ’, Journal for the Scientific Study of Religion , 25 : 215 – 32 . Google Scholar Crossref Search ADS Weenink , D . ( 2009 ), ‘ Explaining Ethnic Inequality in the Juvenile Justice System. An Analysis of the Outcomes of Dutch Prosecutorial Decision Making ’. British Journal of Criminology , 49 : 220 – 42 . Google Scholar Crossref Search ADS Wittebrood , K . ( 2003 ), ‘ Juvenile Crime and Sanctions in the Netherlands ’, Journal of Contemporary Criminal Justice , 19 : 435 – 53 . Google Scholar Crossref Search ADS Wortley , S . ( 2009 ), ‘ Introduction. The Immigration-Crime Connection: Competing Theoretical Perspectives ’, Journal of International Migration and Integration , 10 : 349 – 58 . Google Scholar Crossref Search ADS Footnotes 1 The data are deposited at the DANS data archive: https://dans.knaw.nl/en/front-page?set_language=en. 2 Statistics Netherlands, http://statline.cbs.nl, visited August 2017. 3 Statistics Netherlands, http://statline.cbs.nl, visited August 2017. 4 We also ran models with separate dummies for living with one or two parents. The coefficient of the dummy living with two parents was the most negative of the two, but did not reach conventional levels of statistical significance. 5 If information on the father was missing, we used information on the mother and vice versa. 6 On 1 September 2011, 548 of 10,975 prisoners were Moroccan-born, and about 500 prisoners were second-generation Moroccans (Henneken-Hordijk and Van Gemmert 2012). 7 The group figures under low social desirability were calculated by estimating a model with five IVs: the four dummies indicating ethnic origin by generation and social desirability. We then calculated the percentage for each group based on the constant and the coefficient for the dummy indicating ethnic group and generation. 8 For example (11.1/2.0)/(0.32/0.21) = 3.6. © The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). All rights reserved. For permissions, please e-mail: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Ryan,, Nicole;Ackerman,, Jeff;Bond,, Christine;Ready,, Justin;Kinner, Stuart, A

doi: 10.1093/bjc/azy027pmid: N/A

Abstract This study explores whether differences in Indigenous and non-Indigenous peoples’ risk of reincarceration can partially be explained by their social experiences prior-to-prison and their prison-life experiences. Using administrative and self-report data from 1,238 incarcerated Indigenous (n =303) and non-Indigenous (n = 935) people in Australia, we conducted a series of Cox proportional hazards regressions. We found that Indigenous people had a significantly increased risk of reincarceration compared to non-Indigenous people, and that this can partially be explained by social experiences prior-to-prison. However, after conducting multivariate analyses, the association between prison life-experiences and reincarceration was attenuated to the null. The implications for policy and theory are discussed. Introduction Indigenous people across post-colonial societies are over-represented in prisons. Despite the introduction of diversionary measures, targeted interventions, specialized Indigenous courts and changes in legislation, this over-representation continues (United States Sentencing Commission 2013; Australian Bureau of Statistics (ABS) 2016b; The Sentencing Project 2016; NZStats 2017; Statistics Canada 2018). Although Indigenous prison populations have increased by 8 per cent in New Zealand between 1999 and 2014, to 392 per cent in Australia since 1991 (ABS 2016b; NZStats 2017), the number of Indigenous people released from prison has also increased. One possible explanation for this is that Indigenous people are more likely than non-Indigenous people to be reincarcerated after release from prison (see Broadhurst et al. 1988; Weatherburn et al. 2006; Alaska Judicial Council 2007; Willis and Moore 2008; Giles and Whale 2016). Despite extensive literature examining prisoner re-entry, comparatively few studies have examined reincarceration among Indigenous people. Therefore, it remains unclear why Indigenous people are more likely than non-Indigenous people in post-colonial societies to be reincarcerated. Unfortunately, what is known about Indigenous peoples’ risk of reincarceration is piecemeal at best. Indigeneity—Limitations of prior research and what we know Most prisoner re-entry scholarship has not compared race as a primary point of interest. Besides this limitation, there are notable methodological/design limitations in previous Indigenous re-entry research that is arguably problematic for policymakers. For example, results from prior studies cannot be generalized to all Indigenous people as the samples used are often restricted by type of offence and/or gender. This was a limitation of Wundersitz’s (2010) study that employed a sample of violent male offenders only and Weatherburn’s (2010) research that focused on convicted burglars. By concentrating on type of offence, and/or gender, large proportions of Indigenous ex-prisoners are excluded. As such, these findings are not reflective of the Indigenous offender population, as group differences between types of Indigenous offenders may exist. Furthermore, most previous studies comparing Indigenous and non-Indigenous ex-prisoners’ re-entry patterns have only identified whether Indigenous people are more likely to return, and return at a faster rate—however, the reasons why, have for the most part, been left unanswered (Willis and Moore 2008). Studies attempting to answer ‘why’, examined a limited number of mediating risk factors associated with reincarceration (Giles and Whale 2016), largely ignoring the range of empirically based risk factors that may be associated with Indigenous and non-Indigenous reincarceration. Comparative studies that consider the possibility of multiple mediating effects when examining risk of reincarceration are outdated (Broadhurst et al. 1988). This is problematic as many risks can change overtime due to individual and structural changes (i.e. homelessness, unemployment, substance use, etc). Opinions regarding why Indigenous people are over-represented in prisons are somewhat polarized. One argument focuses on the cumulative effect of colonization. It is suggested that Indigenous people experience a level of disadvantage above and beyond that experienced by non-Indigenous people, even when both have the same level of socio-economic status, due to ongoing effects of dispossession of land, culture and the forcible removal of children (Johnston 1991; Cunneen 2001; 2006). Alternatively, it is argued that over-representation of Indigenous people in prisons can be explained more parsimoniously by differences in traditional criminogenic risk factors that are more prevalent among Indigenous people, such as alcohol dependence, illicit drug use, mental health problems, welfare dependency, chronic socio-economic disadvantage, over-policing, poor education and extensive criminal histories (Bursill 1999; Weatherburn et al. 2003; Blagg 2008; Stubbs 2011). It is well established that Indigenous people are more likely than non-Indigenous people to reside in communities characterized by poverty, substance abuse, violence, overcrowding, teenage pregnancies, high crime rates, high mortality rates, as well as low levels of nutrition, parental supervision, social capital and social support (Weatherburn and Lind 2001). Indigenous people are also on average sentenced to prison longer than non-Indigenous people (Jeffries and Bond 2009). However, it is poorly understood how, or if, these variables affect Indigenous peoples’ risk of reincarceration. Furthermore, we do not know whether Indigenous peoples’ social experiences prior-to-prison and/or their prison-life experiences help to explain why they are more likely to be reincarcerated than non-Indigenous people. Prior social experiences Homelessness is a strong predictor for the onset and persistence of offending behaviour (Makarios et al. 2010; Huebner and Berg 2011). Not only are homeless people at an increased risk of victimization and criminalization, homelessness may also serve as a barrier to employment, subsequently increasing ex-prisoners’ risk of reoffending, as they cannot afford the basic costs of living (Agnew 2006; Makarios et al. 2010). Moreover, research shows that homelessness prior-to-prison is a strong predictor of homelessness post-prison (Baldry et al. 2006). Compared to non-Indigenous people, Australia’s Indigenous people are 14 times more likely to be homeless (Australian Institute of Health and Wellbeing (AIHW) 2014), Canada’s Indigenous people are 8 times more likely (Patrick 2014) and New Zealand’s Maori people are 5 times more likely to be homeless (Amore 2016). In addition, it is estimated that more than 50 per cent of Native Americans living in tribal areas are considered homeless (Pindus et al. 2017). Alcohol and illicit substance use are also well documented predictors of criminality and reincarceration (Visher et al. 2003; Kinner et al. 2009). Indigenous people in Australia, Canada, the United States and New Zealand are more likely to abstain from alcohol than members of the general population, yet besides Native Americans, Indigenous people are also significantly more likely to engage in high-risk alcohol use, than non-Indigenous people (Ministry of Health 2016; Pauly et al. 2016; AIHW 2017; Pindus et al. 2017). However, it must be noted that information regarding Native Americans’ alcohol use is sparse and may not accurately reflect their patterns of alcohol use. Furthermore, Indigenous people in Canada, the United States and New Zealand engage in illicit drug use at disproportionate rates compared to non-Indigenous people (Ministry of Health 2016; Pauly et al. 2016; Pindus et al. 2017). However, evidence of disproportionate illicit substance use between Australian Indigenous and non-Indigenous people is inconclusive. Although the AIHW (2017) reports Indigenous people are 6.8 times more likely to use cannabis, 6 times more likely to use amphetamines and 4.8 times more likely to use heroin, Australian Indigenous people only constituted 1.9 per cent of their sample, and Indigenous people living in very remote areas were under-represented, whereas those living in major cities were over-represented. As such, to generalize results to all Indigenous people is inappropriate. These statistics are remarkably different to those in the AIHW Prisoner Health Report (2015) that suggests a higher prevalence of illicit drug use among non-Indigenous than Indigenous Australians. Prior re-entry literature has mainly focused on examining ‘any’ illicit drug use, not type or frequency of illicit drug use that is problematic because there may be group differences between type, and frequency of drug use, as well as, between Indigenous and non-Indigenous people. Reporting practices that only report ‘any use’ may be masking not only the true patterns of substance use but also the effects of type of substance use on risk of reincarceration. Of concern is the high proportion of ex-prisoners with unresolved substance-use problems at release, as their risk of reincarceration is significantly greater than ex-prisoners without a substance problem (Thomas et al. 2013). A systematic review of 13 studies with a total of 7,563 prisoners found 18–30 per cent of male prisoners and 10–24 per cent of female prisoners reported alcohol abuse and/or dependency (Fazel et al. 2006). Drug dependency is considered a motivating factor for property and non-violent crimes (Kopak and Hoffmann 2014). Moreover, risky substance use can have adverse effects on an individual’s physical and mental health, social-support networks and employment (Fergusson et al. 2013), thus potentially having a multiplicative effect on the risk of reincarceration. Furthermore, Indigenous compared to non-Indigenous people across post-colonial societies are more likely to be unemployed. In Australia, the unemployment rate for Indigenous people is approximately 20 per cent (ABS 2016a), compared to the national unemployment rate of 5.6 per cent (ABS 2018). Similar patterns can be seen in unemployment rates in Canada (16.7 per cent vs. 6.3 per cent) (Statistics Canada 2018), the United States (12 per cent vs. 5.2 per cent) (US Census Bureau 2015) and in New Zealand (11.4 per cent vs. 5.1 per cent) (NZStats 2017). Employment is one of the greatest barriers to reintegration (Travis et al. 2001). The likelihood of obtaining and maintaining employment post-prison is further decreased if a person has mental health problems, which Indigenous people are more likely to have than non-Indigenous people (ABS 2016a). Ex-prisoners rely heavily on family and friends to obtain employment post-release (Visher et al. 2003; Duwe and Clark 2013). However, given Indigenous people compared to non-Indigenous people are more likely to be imprisoned, homeless, suffering mental health problems and/or have substance abuse problems, they may have less social networks to provide such employment opportunities. Prison-life experiences Prisoner visitation is considered a protective factor for reincarceration (Bales and Mears 2008; Derkzen et al. 2009; Duwe and Clark 2013). Prisoners who are visited close to release (Bales and Mears 2008) receive private family visits (Derkzen et al. 2009), and/or receive visits from prosocial people such as a chaplain or mentor, (Duwe and Clark 2013) are less likely to reoffend post-release. Prisoner visitation is believed to provide prisoners with an invaluable source of social capital, resources and informal social control, which reduces their risk of reincarceration. Maintaining prosocial bonds and remaining connected to family and friends may reduce feelings of strain associated with incarceration and promote prosocial behaviour while in prison and post-release (Duwe and Clark 2013). Furthermore, visitation helps foster optimism, which may contribute to a prisoner’s successful reintegration (Burnett and Maruna 2004; Visher and O’Connell 2012). Participation in prison-based programming is considered a protective factor against reincarceration (Pearson et al. 2002; Prendergast et al. 2002). However, due to the lack of perceived cultural relevance of prison-based programs by Indigenous people, program participation may not reduce their risk of reoffending (Willis 2008). Most prison-based programs for Indigenous people stem from westernized programs promoting individual accountability and self-disclosure, and are adapted to include Indigenous cultural aspects (Willis 2008). As such, the reintegration of Indigenous ex-prisoners may be hampered by limited Indigenous-specific programs provided in prisons. Furthermore, due to the multiple types of disadvantage Indigenous people may experience, Gilbert and Wilson (2009) recommended their non-criminogenic needs such as grief for loss of culture, country and spirit be addressed before addressing their criminogenic needs. Many Indigenous scholars and activists have argued that the needs of Indigenous people ‘cannot be accommodated within an adaptation of a programme from another Indigenous culture, or programmes from another country’ (Yavu-Kama-Harathunian 2002). Finally, prison-to-community transitions can be emotionally and mentally challenging. During this time, prisoners are at an increased risk of mental health problems, such as depression, anxiety and psychiatric co-morbidity. Indigenous people in post-colonial societies are more likely to suffer with mental health problems such as depression and anxiety than non-Indigenous people (AIHW 2015; Ministry of Health 2016). Despite this, our knowledge regarding the mental health of Indigenous people while in prison is limited, and findings are mixed. An Australian study found most Indigenous people (72.8 per cent of women and 86.1 per cent of men) in prison suffered from at least one mental disorder, most commonly being post-traumatic stress disorder, and major depression (Heffernan et al. 2012). However, the AIHW (2015) found the number of Indigenous (43 per cent) and non-Indigenous people (45 per cent) that were informed about having mental health problems by a health professional were very similar. Whereas, Korff (2016) found as high as 93 per cent of Indigenous prisoners showed signs of mental health problems. Ex-prisoners with mental health problems face greater re-entry challenges and an increased risk of reincarceration (Cutcher et al. 2014). Besides immediate challenges of housing, employment and re-establishing support networks, ex-prisoners with mental health problems need access to ongoing medical treatment and a plausible mental healthcare plan to reduce their risk of reincarceration (Mallik-Kane and Visher 2008). Jordan (2010) suggested ‘the prison milieu is not always conducive to good mental health and is not often a useful catalyst for mental healthcare’ (26). This may explain why only 11.8 per cent of 1,318 Australian ex-prisoners accessed mental healthcare services in the 12 months post-release (Thomas et al. 2016), and only 34 per cent of 138 Australian Indigenous prisoners could correctly identify their medication (Carroll et al. 2014). Without knowing why Indigenous people are more likely to return to prison, introducing Indigenous-specific policy and practices in the absence of strong empirical evidence is unlikely to be effective. With governments focusing on best practices, cultural competence and evidenced-based policy and programming, the practice of generalization of evidence across cultures is no longer acceptable. Although it is important to know if Indigenous compared to non-Indigenous people are more likely to be reincarcerated, this provides little empirical evidence that can inform targeted interventions and rehabilitation programs. We propose to contribute to the limited knowledge in this area by examining (1) whether there are group differences in Indigenous and non-Indigenous ex-prisoners’ health, social and criminal justice characteristics; (2) whether Indigenous compared to non-Indigenous people are more likely to be reincarcerated after release from prison; and (3) identifying to what extent can Indigenous and non-Indigenous peoples’ social experiences prior-to-prison, and prison-life experiences, explain any differences in their risk of reincarceration. Methodology Data source Data were drawn from the ‘Passports study’, which consists of information collected from Queensland prisoners during interviews conducted within six weeks of their anticipated release (baseline), and at one, three and six months post-prison. Potential participants were identified via prison records. Prisoners on remand were excluded due to uncertainty surrounding their release date (Kinner et al. 2013). Interviews were structured and took 60–90 minutes to complete. Data on participants’ demographic characteristics, criminal history, prior juvenile detention, pre-prison social circumstances, physical and mental health, substance use, social support and in-prison experiences were collected in baseline interviews. Self-report data were linked prospectively with administrative data from Queensland Corrective Services (QCS)1. The QCS data consisted of dates of participants reincarceration for a new offence or parole breach, until 31 December 2013, previous adult incarcerations dating back to 2006, vocational program participation and transition (re-entry) program participation (Kinner et al. 2013). Data were linked probabilistically based on participants’ identifying information, including all known aliases, given evidence that this increases sensitivity of linkage, without adversely affecting specificity (Larney and Burns 2011). In total 1,325 Queensland prisoners completed the baseline survey for the ‘Passports study’. Reincarceration Time to reincarceration as a dependent variable is suggested to be a good proxy measure for serious offending over time (Mears et al. 2012). However, we argue that reincarceration is an important measure in its own right, and should not be a proxy measure for serious offending, as people can be reincarcerated for a parole breach and not a serious offence. Also, considering Indigenous people are 13 times more likely to be incarcerated (ABS 2016a), the measurement of reincarceration may mask structural and systemic bias in policing, parole and court practices. Prior scholarly work on prisoner re-entry has employed reincarceration as the outcome measure to avoid overestimates of reoffending, thus decreasing the probability of finding a type one error (Ulmer 2001). Reincarceration provides a conservative measure of reoffending, and is commonly used in scholarly work examining prisoner re-entry (Payne 2007; Stahler et al. 2013). As such, we measured whether any participant is reincarcerated for a parole breach, or a new offence, before the completion of the follow-up period. Independent variables All independent variables included have empirical or theoretical importance. Indigenous status (Aboriginal, Torres Strait Islander and/or South Sea Islander, vs. not) was collected by self-report during the baseline interview, as well as from QCS data. In accordance with best practices for recording Indigenous status (Dugbaza et al. 2012), participants who identified as Indigenous at least once, were recorded as Indigenous. Measures of prison-life mechanisms included prison employment, visited in prison, no reintegration plan2, prison program participation, low perception of social support and high psychological distress. Other variables used included demographics, criminal history and social experiences prior-to-prison. Demographic factors included gender, education, forcible removal from family as a child and age at release. Measures of prior criminal history included prior adult prison sentences, type of offence most recently imprisoned for (violent, property, drug or traffic/public disorder offence) and number of prior juvenile detentions. Social experiences prior-to-prison included unemployment, homelessness and high-risk substance use (alcohol, marijuana, methamphetamine and heroin). Table 1 provides full descriptions and percentage of missing data for each variable. Table 1 Description of variables and number of participants with missing data Variable Description (number with missing data out of full sample) Demographic  Indigenous Identified as Aboriginal, Torres Strait Islander or South Sea Islander (0). No = 0; Yes = 1  Gender Identified as male (0). No = 0; Yes = 1  Removed from family Was removed from their family as a child (2). No = 0; Yes = 1  Education Achieved year 9 or less in education (4). No = 0; Yes = 1  Age Age at time of release. QCS data (0) Criminal history  Adult imprisonment Number of prior adult prison sentences categorized as first time prisoner (no prior adult prison sentence) = 0; one to four prior adult prison sentences = 1; 5 or more prior adult prison sentences = 2 (QCS data)  Violent offence Most recently imprisoned for a violent offence (0). No = 0; Yes = 1. (QCS data)  Property offence Most recently imprisoned for a property offence (0). No = 0; Yes = 1. (QCS data)  Drug offence Most recently imprisoned for a drug offence (0). No = 0; Yes = 1. (QCS data)  Traffic/public order offence Most recently imprisoned for a traffic public order offence (0). No = 0; Yes = 1. (QCS data)  Juvenile detention Number of prior juvenile detentions (7) categorized as no prior juvenile detentions = 0; 1–4 juvenile detentions = 1; 5 or more juvenile detentions = 2. (QCS data)  Days served in prison Number of days most recently served in prison (0). (QCS data) Prior prison social experiences  Unstable housing PP No permanent address in the 6 months PP (0). No = 0; Yes = 1  Unemployment PP Mainly unemployed for the 6 months PP (0). No = 0; Yes = 1  HR alcohol use PP Score of ≥16 on the Alcohol Use Disorder Identification Test (Bador et al 2001) indicates high-risk drinker to possible alcohol dependence in the 12 months PP (2)  HR marijuana use PP Score of ≥4 for marijuana on the ASSIST (Humeniuk et al. 2010) indicates moderate to HR marijuana use in the 3 months PP (1)  HR methamphetamine use PP Scored ≥4 for methamphetamine on the ASSIST. Indicates moderate to HR methamphetamine use in the 3 months PP (2)  HR heroin use PP Scored ≥4 for heroin on the ASSIST. Indicates moderate to HR heroin use in the 3 months PP (3) Prison-life experiences  Prison work Employed in a prison work program while most recently incarcerated (0), Yes = 0; No = 1. (QCS data)  No. of visits Number of visits received in the month prior to release (0) (QCS data)  No reintegration plan Did not have a reintegration plan when released (2), Yes = 0; No = 1 (QCS data)  Incomplete transition program Did not complete a transition program prior to release from prison (1), Yes = 0; No = 1 (QCS data)  Low perceived social support Scored ≥18 on the five item ENRICHD Social Support Inventory (Mitchell et al. 2003) (4), No = 0; Yes = 1  High psychological distress Scored ≥22 on the Kessler Psychological Distress Scale (Kessler et al. 2002) (4). Indicates high to very high psychological distress in the month prior to release from prison, No = 0; 1 = Yes Variable Description (number with missing data out of full sample) Demographic  Indigenous Identified as Aboriginal, Torres Strait Islander or South Sea Islander (0). No = 0; Yes = 1  Gender Identified as male (0). No = 0; Yes = 1  Removed from family Was removed from their family as a child (2). No = 0; Yes = 1  Education Achieved year 9 or less in education (4). No = 0; Yes = 1  Age Age at time of release. QCS data (0) Criminal history  Adult imprisonment Number of prior adult prison sentences categorized as first time prisoner (no prior adult prison sentence) = 0; one to four prior adult prison sentences = 1; 5 or more prior adult prison sentences = 2 (QCS data)  Violent offence Most recently imprisoned for a violent offence (0). No = 0; Yes = 1. (QCS data)  Property offence Most recently imprisoned for a property offence (0). No = 0; Yes = 1. (QCS data)  Drug offence Most recently imprisoned for a drug offence (0). No = 0; Yes = 1. (QCS data)  Traffic/public order offence Most recently imprisoned for a traffic public order offence (0). No = 0; Yes = 1. (QCS data)  Juvenile detention Number of prior juvenile detentions (7) categorized as no prior juvenile detentions = 0; 1–4 juvenile detentions = 1; 5 or more juvenile detentions = 2. (QCS data)  Days served in prison Number of days most recently served in prison (0). (QCS data) Prior prison social experiences  Unstable housing PP No permanent address in the 6 months PP (0). No = 0; Yes = 1  Unemployment PP Mainly unemployed for the 6 months PP (0). No = 0; Yes = 1  HR alcohol use PP Score of ≥16 on the Alcohol Use Disorder Identification Test (Bador et al 2001) indicates high-risk drinker to possible alcohol dependence in the 12 months PP (2)  HR marijuana use PP Score of ≥4 for marijuana on the ASSIST (Humeniuk et al. 2010) indicates moderate to HR marijuana use in the 3 months PP (1)  HR methamphetamine use PP Scored ≥4 for methamphetamine on the ASSIST. Indicates moderate to HR methamphetamine use in the 3 months PP (2)  HR heroin use PP Scored ≥4 for heroin on the ASSIST. Indicates moderate to HR heroin use in the 3 months PP (3) Prison-life experiences  Prison work Employed in a prison work program while most recently incarcerated (0), Yes = 0; No = 1. (QCS data)  No. of visits Number of visits received in the month prior to release (0) (QCS data)  No reintegration plan Did not have a reintegration plan when released (2), Yes = 0; No = 1 (QCS data)  Incomplete transition program Did not complete a transition program prior to release from prison (1), Yes = 0; No = 1 (QCS data)  Low perceived social support Scored ≥18 on the five item ENRICHD Social Support Inventory (Mitchell et al. 2003) (4), No = 0; Yes = 1  High psychological distress Scored ≥22 on the Kessler Psychological Distress Scale (Kessler et al. 2002) (4). Indicates high to very high psychological distress in the month prior to release from prison, No = 0; 1 = Yes Note: Measurements for variables are from self-report data unless specified otherwise. ASSIST = Alcohol, Smoking and Substance Involvement Screening Test; HR = high risk; PP = prior-to-prison. View Large Table 1 Description of variables and number of participants with missing data Variable Description (number with missing data out of full sample) Demographic  Indigenous Identified as Aboriginal, Torres Strait Islander or South Sea Islander (0). No = 0; Yes = 1  Gender Identified as male (0). No = 0; Yes = 1  Removed from family Was removed from their family as a child (2). No = 0; Yes = 1  Education Achieved year 9 or less in education (4). No = 0; Yes = 1  Age Age at time of release. QCS data (0) Criminal history  Adult imprisonment Number of prior adult prison sentences categorized as first time prisoner (no prior adult prison sentence) = 0; one to four prior adult prison sentences = 1; 5 or more prior adult prison sentences = 2 (QCS data)  Violent offence Most recently imprisoned for a violent offence (0). No = 0; Yes = 1. (QCS data)  Property offence Most recently imprisoned for a property offence (0). No = 0; Yes = 1. (QCS data)  Drug offence Most recently imprisoned for a drug offence (0). No = 0; Yes = 1. (QCS data)  Traffic/public order offence Most recently imprisoned for a traffic public order offence (0). No = 0; Yes = 1. (QCS data)  Juvenile detention Number of prior juvenile detentions (7) categorized as no prior juvenile detentions = 0; 1–4 juvenile detentions = 1; 5 or more juvenile detentions = 2. (QCS data)  Days served in prison Number of days most recently served in prison (0). (QCS data) Prior prison social experiences  Unstable housing PP No permanent address in the 6 months PP (0). No = 0; Yes = 1  Unemployment PP Mainly unemployed for the 6 months PP (0). No = 0; Yes = 1  HR alcohol use PP Score of ≥16 on the Alcohol Use Disorder Identification Test (Bador et al 2001) indicates high-risk drinker to possible alcohol dependence in the 12 months PP (2)  HR marijuana use PP Score of ≥4 for marijuana on the ASSIST (Humeniuk et al. 2010) indicates moderate to HR marijuana use in the 3 months PP (1)  HR methamphetamine use PP Scored ≥4 for methamphetamine on the ASSIST. Indicates moderate to HR methamphetamine use in the 3 months PP (2)  HR heroin use PP Scored ≥4 for heroin on the ASSIST. Indicates moderate to HR heroin use in the 3 months PP (3) Prison-life experiences  Prison work Employed in a prison work program while most recently incarcerated (0), Yes = 0; No = 1. (QCS data)  No. of visits Number of visits received in the month prior to release (0) (QCS data)  No reintegration plan Did not have a reintegration plan when released (2), Yes = 0; No = 1 (QCS data)  Incomplete transition program Did not complete a transition program prior to release from prison (1), Yes = 0; No = 1 (QCS data)  Low perceived social support Scored ≥18 on the five item ENRICHD Social Support Inventory (Mitchell et al. 2003) (4), No = 0; Yes = 1  High psychological distress Scored ≥22 on the Kessler Psychological Distress Scale (Kessler et al. 2002) (4). Indicates high to very high psychological distress in the month prior to release from prison, No = 0; 1 = Yes Variable Description (number with missing data out of full sample) Demographic  Indigenous Identified as Aboriginal, Torres Strait Islander or South Sea Islander (0). No = 0; Yes = 1  Gender Identified as male (0). No = 0; Yes = 1  Removed from family Was removed from their family as a child (2). No = 0; Yes = 1  Education Achieved year 9 or less in education (4). No = 0; Yes = 1  Age Age at time of release. QCS data (0) Criminal history  Adult imprisonment Number of prior adult prison sentences categorized as first time prisoner (no prior adult prison sentence) = 0; one to four prior adult prison sentences = 1; 5 or more prior adult prison sentences = 2 (QCS data)  Violent offence Most recently imprisoned for a violent offence (0). No = 0; Yes = 1. (QCS data)  Property offence Most recently imprisoned for a property offence (0). No = 0; Yes = 1. (QCS data)  Drug offence Most recently imprisoned for a drug offence (0). No = 0; Yes = 1. (QCS data)  Traffic/public order offence Most recently imprisoned for a traffic public order offence (0). No = 0; Yes = 1. (QCS data)  Juvenile detention Number of prior juvenile detentions (7) categorized as no prior juvenile detentions = 0; 1–4 juvenile detentions = 1; 5 or more juvenile detentions = 2. (QCS data)  Days served in prison Number of days most recently served in prison (0). (QCS data) Prior prison social experiences  Unstable housing PP No permanent address in the 6 months PP (0). No = 0; Yes = 1  Unemployment PP Mainly unemployed for the 6 months PP (0). No = 0; Yes = 1  HR alcohol use PP Score of ≥16 on the Alcohol Use Disorder Identification Test (Bador et al 2001) indicates high-risk drinker to possible alcohol dependence in the 12 months PP (2)  HR marijuana use PP Score of ≥4 for marijuana on the ASSIST (Humeniuk et al. 2010) indicates moderate to HR marijuana use in the 3 months PP (1)  HR methamphetamine use PP Scored ≥4 for methamphetamine on the ASSIST. Indicates moderate to HR methamphetamine use in the 3 months PP (2)  HR heroin use PP Scored ≥4 for heroin on the ASSIST. Indicates moderate to HR heroin use in the 3 months PP (3) Prison-life experiences  Prison work Employed in a prison work program while most recently incarcerated (0), Yes = 0; No = 1. (QCS data)  No. of visits Number of visits received in the month prior to release (0) (QCS data)  No reintegration plan Did not have a reintegration plan when released (2), Yes = 0; No = 1 (QCS data)  Incomplete transition program Did not complete a transition program prior to release from prison (1), Yes = 0; No = 1 (QCS data)  Low perceived social support Scored ≥18 on the five item ENRICHD Social Support Inventory (Mitchell et al. 2003) (4), No = 0; Yes = 1  High psychological distress Scored ≥22 on the Kessler Psychological Distress Scale (Kessler et al. 2002) (4). Indicates high to very high psychological distress in the month prior to release from prison, No = 0; 1 = Yes Note: Measurements for variables are from self-report data unless specified otherwise. ASSIST = Alcohol, Smoking and Substance Involvement Screening Test; HR = high risk; PP = prior-to-prison. View Large High-risk substance use was measured using the Alcohol Smoking and Substance Involvement Screening Test (ASSIST) (Humeniuk et al. 2010). The ASSIST consists of an eight-item questionnaire per substance that was interviewer administered and enabled the identification of high-risk substance use (WHO 2002). Also, the ENRICHD Social Support Inventory was used, which is a seven-item questionnaire that assesses three distinct types of support (perceived support, enacted support and social integration), to measure an individual’s perception of social support during their last month in prison (Mitchell et al. 2003). The Kessler Psychological Distress Scale (K10), a ten-item self-report measure about anxiety and depressive symptoms, was also used to measure a person’s psychological distress in the month prior to baseline (Kessler et al. 2002). Statistical analyses To assess participants’ risk of reincarceration over time, and identify the risk/protective factors for reincarceration, we conducted a series of Cox proportional hazard regression analyses (Cox and Oakes 1984). Censoring events included in the analyses were participant’s death, which was obtained from the National Death Index, the ‘Passports study’s’ right censoring date (31 December 2013) and participants’ reincarceration for parole breach or new offence. First, we conducted a series of bivariate Cox proportional hazard regressions with each covariate to establish an association with reincarceration. Second, we examined the attenuating effect that each variable had on the ‘Indigenous–Reincarceration’ relationship. Third, we constructed a multivariate model that examined the different effects the covariates (demographics, criminal history variables and prior prison social experience variables) had on the ‘Indigenous–Reincarceration’ relationship. Results To begin, we conducted descriptive analyses stratified by Indigenous status on all variables included in the study. For a full description of the characteristics of Indigenous and non-Indigenous ex-prisoners interviewed at baseline, see Table 2. Of the 1,238 ex-prisoners included in the study, 651 (52.6 per cent) were reincarcerated during a maximum 5-year follow-up (Mdn follow-up = 2.94 years). This included 71.3 per cent (n = 216) Indigenous participants and 48.1 per cent (n = 450) non-Indigenous participants. To establish a statistical difference in risk of reincarceration for Indigenous and non-Indigenous participants, we conducted a Kaplan–Meier survival analysis that indicated Indigenous people had a significantly higher risk of reincarceration than non-Indigenous people (p ≤ 0.001) (see Figure 1). Next, we compared the median number of days until reincarceration for Indigenous (Mdn = 263 days) and non-Indigenous participants (Mdn = 244 days). The difference in number of days to reincarceration between the two groups was not statistically significant (p = 0.307). Table 2 Indigenous and non-Indigenous participants’ characteristics at baseline Non-Indigenous (n = 935) Indigenous (n = 303) Covariates % % χ2 φc Gender (male) 81.7 71.6 14.17*** 0.11 Removed from family 15.9 31.7 35.53*** 0.17 Education (≤9 years) 37.9 57.1 33.98*** 0.17 Prior adult imprisonment (first time) 39.8 16.8 59.83*** 0.22  1–4 prior imprisonments 35.3 41.3  5 or more prior imprisonments 24.8 41.6 Violent offence 47.2 61.4 18.51*** 0.12 Theft offence 23.3 22.8 0.04 0.01 Drug offence 5.8 2 7.15** 0.08 Public order offence 9 4.3 6.98** 0.08 Prior juvenile detention (none) 79.4 50.8 95.98*** 0.28  1–4 prior juvenile detentions 15.6 30.7  5 or more juvenile detentions 4.8 16.8 High-risk marijuana use PP 42.8 56.4 17.60*** 0.12 High-risk methamphetamine use PP 40.6 29 12.59*** 0.10 High-risk heroin use PP 19.6 10.2 13.73*** 0.11 High-risk alcohol use PP 36.3 61.1 59.23*** 0.22 Unemployed 6 month PP 46.3 71.9 60.32*** 0.22 Unstable housing PP 18.7 26.1 7.59** 0.08 No reintegration plan 79.9 79.9 0.036 0.01 Participation in transition program 85.1 79.5 4.79* 0.06 Not employed in prison 52.8 55.8 0.8 0.03 Low perception of social support 16 16.2 0.01 0.01 High level of psychological distress 24.9 28.7 1.83 0.04 Covariates M (SD) M(SD) t Cohen’s d  Age at release in years 33.89 (11.8) 29.98 (8.15) 5.36*** 0.04 Covariates Mdn (SD) Mdn (SD) t Cohen’s d  No. of time visited 1.76 (2.64) 0.82 (1.71) 5.85*** 0.42  Days served in prison 176 (527.34) 120 (476.04) 1.2 0.08 Non-Indigenous (n = 935) Indigenous (n = 303) Covariates % % χ2 φc Gender (male) 81.7 71.6 14.17*** 0.11 Removed from family 15.9 31.7 35.53*** 0.17 Education (≤9 years) 37.9 57.1 33.98*** 0.17 Prior adult imprisonment (first time) 39.8 16.8 59.83*** 0.22  1–4 prior imprisonments 35.3 41.3  5 or more prior imprisonments 24.8 41.6 Violent offence 47.2 61.4 18.51*** 0.12 Theft offence 23.3 22.8 0.04 0.01 Drug offence 5.8 2 7.15** 0.08 Public order offence 9 4.3 6.98** 0.08 Prior juvenile detention (none) 79.4 50.8 95.98*** 0.28  1–4 prior juvenile detentions 15.6 30.7  5 or more juvenile detentions 4.8 16.8 High-risk marijuana use PP 42.8 56.4 17.60*** 0.12 High-risk methamphetamine use PP 40.6 29 12.59*** 0.10 High-risk heroin use PP 19.6 10.2 13.73*** 0.11 High-risk alcohol use PP 36.3 61.1 59.23*** 0.22 Unemployed 6 month PP 46.3 71.9 60.32*** 0.22 Unstable housing PP 18.7 26.1 7.59** 0.08 No reintegration plan 79.9 79.9 0.036 0.01 Participation in transition program 85.1 79.5 4.79* 0.06 Not employed in prison 52.8 55.8 0.8 0.03 Low perception of social support 16 16.2 0.01 0.01 High level of psychological distress 24.9 28.7 1.83 0.04 Covariates M (SD) M(SD) t Cohen’s d  Age at release in years 33.89 (11.8) 29.98 (8.15) 5.36*** 0.04 Covariates Mdn (SD) Mdn (SD) t Cohen’s d  No. of time visited 1.76 (2.64) 0.82 (1.71) 5.85*** 0.42  Days served in prison 176 (527.34) 120 (476.04) 1.2 0.08 Note: PP = prior-to-prison. ***p < 0.001 or less; **p < 0.010 or less; *p < 0.050 or less. *Measurement is only for those who have been reincarcerated post-release. View Large Table 2 Indigenous and non-Indigenous participants’ characteristics at baseline Non-Indigenous (n = 935) Indigenous (n = 303) Covariates % % χ2 φc Gender (male) 81.7 71.6 14.17*** 0.11 Removed from family 15.9 31.7 35.53*** 0.17 Education (≤9 years) 37.9 57.1 33.98*** 0.17 Prior adult imprisonment (first time) 39.8 16.8 59.83*** 0.22  1–4 prior imprisonments 35.3 41.3  5 or more prior imprisonments 24.8 41.6 Violent offence 47.2 61.4 18.51*** 0.12 Theft offence 23.3 22.8 0.04 0.01 Drug offence 5.8 2 7.15** 0.08 Public order offence 9 4.3 6.98** 0.08 Prior juvenile detention (none) 79.4 50.8 95.98*** 0.28  1–4 prior juvenile detentions 15.6 30.7  5 or more juvenile detentions 4.8 16.8 High-risk marijuana use PP 42.8 56.4 17.60*** 0.12 High-risk methamphetamine use PP 40.6 29 12.59*** 0.10 High-risk heroin use PP 19.6 10.2 13.73*** 0.11 High-risk alcohol use PP 36.3 61.1 59.23*** 0.22 Unemployed 6 month PP 46.3 71.9 60.32*** 0.22 Unstable housing PP 18.7 26.1 7.59** 0.08 No reintegration plan 79.9 79.9 0.036 0.01 Participation in transition program 85.1 79.5 4.79* 0.06 Not employed in prison 52.8 55.8 0.8 0.03 Low perception of social support 16 16.2 0.01 0.01 High level of psychological distress 24.9 28.7 1.83 0.04 Covariates M (SD) M(SD) t Cohen’s d  Age at release in years 33.89 (11.8) 29.98 (8.15) 5.36*** 0.04 Covariates Mdn (SD) Mdn (SD) t Cohen’s d  No. of time visited 1.76 (2.64) 0.82 (1.71) 5.85*** 0.42  Days served in prison 176 (527.34) 120 (476.04) 1.2 0.08 Non-Indigenous (n = 935) Indigenous (n = 303) Covariates % % χ2 φc Gender (male) 81.7 71.6 14.17*** 0.11 Removed from family 15.9 31.7 35.53*** 0.17 Education (≤9 years) 37.9 57.1 33.98*** 0.17 Prior adult imprisonment (first time) 39.8 16.8 59.83*** 0.22  1–4 prior imprisonments 35.3 41.3  5 or more prior imprisonments 24.8 41.6 Violent offence 47.2 61.4 18.51*** 0.12 Theft offence 23.3 22.8 0.04 0.01 Drug offence 5.8 2 7.15** 0.08 Public order offence 9 4.3 6.98** 0.08 Prior juvenile detention (none) 79.4 50.8 95.98*** 0.28  1–4 prior juvenile detentions 15.6 30.7  5 or more juvenile detentions 4.8 16.8 High-risk marijuana use PP 42.8 56.4 17.60*** 0.12 High-risk methamphetamine use PP 40.6 29 12.59*** 0.10 High-risk heroin use PP 19.6 10.2 13.73*** 0.11 High-risk alcohol use PP 36.3 61.1 59.23*** 0.22 Unemployed 6 month PP 46.3 71.9 60.32*** 0.22 Unstable housing PP 18.7 26.1 7.59** 0.08 No reintegration plan 79.9 79.9 0.036 0.01 Participation in transition program 85.1 79.5 4.79* 0.06 Not employed in prison 52.8 55.8 0.8 0.03 Low perception of social support 16 16.2 0.01 0.01 High level of psychological distress 24.9 28.7 1.83 0.04 Covariates M (SD) M(SD) t Cohen’s d  Age at release in years 33.89 (11.8) 29.98 (8.15) 5.36*** 0.04 Covariates Mdn (SD) Mdn (SD) t Cohen’s d  No. of time visited 1.76 (2.64) 0.82 (1.71) 5.85*** 0.42  Days served in prison 176 (527.34) 120 (476.04) 1.2 0.08 Note: PP = prior-to-prison. ***p < 0.001 or less; **p < 0.010 or less; *p < 0.050 or less. *Measurement is only for those who have been reincarcerated post-release. View Large Fig. 1 View largeDownload slide Indigenous and non-Indigenous peoples’ hazard risk for reincarceration post-release Fig. 1 View largeDownload slide Indigenous and non-Indigenous peoples’ hazard risk for reincarceration post-release Next, to understand why Indigenous people are more at risk of reincarceration than non-Indigenous people, we explored whether participants differed by Indigenous status on each covariate. This was important to establish as differences between groups may help to explain differences in Indigenous and non-Indigenous peoples’ risk of reincarceration. To do this, we conducted a series of chi-square analyses and independent t-tests (see Table 2). There were statistically significant group differences at p < 0.05, in 18 of the 24 covariates in the model. These included gender; removal from family when a child; level of education; prior adult imprisonment; previously imprisoned for either a violent, drug, or public order offence; prior juvenile detention; engaging in high-risk substance use prior-to-prison (marijuana, methamphetamine, heroin and/or alcohol); being unemployed prior-to-prison; having unstable housing prior-to-prison, and not having a re-entry plan; age at release from prison; and the number of times visited in prison in the month prior to baseline. Our results clearly showed a high level of heterogeneity among Indigenous and non-Indigenous ex-prisoners, partially assisting in explaining why Indigenous compared to non-Indigenous people are more likely to be reincarcerated post-release. We next conducted a series of bivariate Cox proportional hazard regressions with all covariates, including Indigenous status, to establish baseline associations with reincarceration (see Table 3), before conducting a multivariate model. Results identified 13 potential risk factors and 3 protective factors for reincarceration. Indigenous status; being male; low education (≤9 years); forcible removal from family as a child; prior adult imprisonments; being imprisoned for a property offence; prior juvenile detentions; serving more days in prison; unstable housing prior-to-prison; unemployment prior-to-prison; and high-risk alcohol, marijuana, methamphetamine and heroin use prior-to-prison were significantly associated with an increased hazard of reincarceration. In contrast, being older at release, currently imprisoned for a drug offence and receiving more visitations in prison the month prior to baseline were all found to be significantly associated with a decreased hazard of reincarceration (see Table 3). Table 3 Reincarceration hazard (n = 1238) Unadjusted Adjusted Covariates HR LL UL p HR LL UL p Indigenous 1.78 1.51 2.09 ≤0.001 1.33 1.10 1.62 0.004 Gender (male) 1.24 1.02 1.51 0.030 1.39 1.11 1.73 0.004 Education achieved (≤9 years) 1.52 1.30 1.76 ≤0.001 1.00 0.84 1.18 0.952 Forcibly removed 1.35 1.12 1.62 0.001 0.85 0.69 1.05 0.122 Age at release 0.96 0.96 0.97 ≤0.001 0.97 0.96 0.98 ≤0.001 Prior adult prison sentences (first-time) ≤0.001 ≤0.001  1–4 prior imprisonments 2.87 2.31 3.56 ≤0.001 2.00 1.59 2.53 ≤0.001  5 or more prior imprisonments 4.20 3.37 5.22 ≤0.001 2.78 2.15 3.61 ≤0.001 Violent offence 1.07 0.92 1.25 0.367 1.20 0.91 1.57 0.191 Property offence 1.46 1.23 1.73 ≤0.001 1.01 0.76 1.36 0.938 Drug offence 0.54 0.36 0.83 0.005 0.85 0.52 1.39 0.520 Public order offence 0.92 0.69 1.23 0.579 1.22 0.84 1.77 0.302 Prior juvenile detentions (none) ≤0.001 0.005  1–4 prior juvenile detentions 1.92 1.60 2.30 ≤0.001 1.23 1.00 1.50 0.050  5 or more prior juvenile detentions 2.72 2.14 3.46 ≤0.001 1.57 1.18 2.09 0.002 Days in prison 1.00 1.00 1.00 0.014 1.00 1.00 1.00 0.048 Unstable housing PP 1.31 1.09 1.56 0.004 1.10 0.91 1.34 0.336 Unemployed 6 months PP 1.61 1.38 1.88 ≤0.001 1.11 0.93 1.32 0.236 HR alcohol use PP 1.18 1.02 1.38 0.032 0.896 0.76 1.06 0.202 HR marijuana use PP 2.07 1.78 2.42 ≤0.001 1.32 1.11 1.56 0.002 HR methamphetamine use PP 2.14 1.84 2.50 ≤0.001 1.40 1.18 1.67 ≤0.001 HR heroin use PP 2.11 1.77 2.52 ≤0.001 1.50 1.23 1.83 ≤0.001 Not employed in prison 1.11 0.95 1.29 0.201 1.09 0.93 1.28 0.290 No. of visits 0.90 0.86 0.94 ≤0.001 0.97 0.93 1.01 0.085 No reintegration plan 0.97 0.80 1.17 0.718 0.88 0.72 1.07 0.195 Participation in transition program 0.93 0.77 1.11 0.403 0.85 0.69 1.05 0.123 Low perception of social support 1.16 0.95 1.41 0.159 0.95 0.76 1.18 0.616 High level of psychological distress 1.01 0.85 1.20 0.909 0.94 0.78 1.131 0.513 Unadjusted Adjusted Covariates HR LL UL p HR LL UL p Indigenous 1.78 1.51 2.09 ≤0.001 1.33 1.10 1.62 0.004 Gender (male) 1.24 1.02 1.51 0.030 1.39 1.11 1.73 0.004 Education achieved (≤9 years) 1.52 1.30 1.76 ≤0.001 1.00 0.84 1.18 0.952 Forcibly removed 1.35 1.12 1.62 0.001 0.85 0.69 1.05 0.122 Age at release 0.96 0.96 0.97 ≤0.001 0.97 0.96 0.98 ≤0.001 Prior adult prison sentences (first-time) ≤0.001 ≤0.001  1–4 prior imprisonments 2.87 2.31 3.56 ≤0.001 2.00 1.59 2.53 ≤0.001  5 or more prior imprisonments 4.20 3.37 5.22 ≤0.001 2.78 2.15 3.61 ≤0.001 Violent offence 1.07 0.92 1.25 0.367 1.20 0.91 1.57 0.191 Property offence 1.46 1.23 1.73 ≤0.001 1.01 0.76 1.36 0.938 Drug offence 0.54 0.36 0.83 0.005 0.85 0.52 1.39 0.520 Public order offence 0.92 0.69 1.23 0.579 1.22 0.84 1.77 0.302 Prior juvenile detentions (none) ≤0.001 0.005  1–4 prior juvenile detentions 1.92 1.60 2.30 ≤0.001 1.23 1.00 1.50 0.050  5 or more prior juvenile detentions 2.72 2.14 3.46 ≤0.001 1.57 1.18 2.09 0.002 Days in prison 1.00 1.00 1.00 0.014 1.00 1.00 1.00 0.048 Unstable housing PP 1.31 1.09 1.56 0.004 1.10 0.91 1.34 0.336 Unemployed 6 months PP 1.61 1.38 1.88 ≤0.001 1.11 0.93 1.32 0.236 HR alcohol use PP 1.18 1.02 1.38 0.032 0.896 0.76 1.06 0.202 HR marijuana use PP 2.07 1.78 2.42 ≤0.001 1.32 1.11 1.56 0.002 HR methamphetamine use PP 2.14 1.84 2.50 ≤0.001 1.40 1.18 1.67 ≤0.001 HR heroin use PP 2.11 1.77 2.52 ≤0.001 1.50 1.23 1.83 ≤0.001 Not employed in prison 1.11 0.95 1.29 0.201 1.09 0.93 1.28 0.290 No. of visits 0.90 0.86 0.94 ≤0.001 0.97 0.93 1.01 0.085 No reintegration plan 0.97 0.80 1.17 0.718 0.88 0.72 1.07 0.195 Participation in transition program 0.93 0.77 1.11 0.403 0.85 0.69 1.05 0.123 Low perception of social support 1.16 0.95 1.41 0.159 0.95 0.76 1.18 0.616 High level of psychological distress 1.01 0.85 1.20 0.909 0.94 0.78 1.131 0.513 Note: HR = high risk; LL = lower level 95% confidence interval; PP = prior-to-prison; UL = upper level 95% confidence interval. View Large Table 3 Reincarceration hazard (n = 1238) Unadjusted Adjusted Covariates HR LL UL p HR LL UL p Indigenous 1.78 1.51 2.09 ≤0.001 1.33 1.10 1.62 0.004 Gender (male) 1.24 1.02 1.51 0.030 1.39 1.11 1.73 0.004 Education achieved (≤9 years) 1.52 1.30 1.76 ≤0.001 1.00 0.84 1.18 0.952 Forcibly removed 1.35 1.12 1.62 0.001 0.85 0.69 1.05 0.122 Age at release 0.96 0.96 0.97 ≤0.001 0.97 0.96 0.98 ≤0.001 Prior adult prison sentences (first-time) ≤0.001 ≤0.001  1–4 prior imprisonments 2.87 2.31 3.56 ≤0.001 2.00 1.59 2.53 ≤0.001  5 or more prior imprisonments 4.20 3.37 5.22 ≤0.001 2.78 2.15 3.61 ≤0.001 Violent offence 1.07 0.92 1.25 0.367 1.20 0.91 1.57 0.191 Property offence 1.46 1.23 1.73 ≤0.001 1.01 0.76 1.36 0.938 Drug offence 0.54 0.36 0.83 0.005 0.85 0.52 1.39 0.520 Public order offence 0.92 0.69 1.23 0.579 1.22 0.84 1.77 0.302 Prior juvenile detentions (none) ≤0.001 0.005  1–4 prior juvenile detentions 1.92 1.60 2.30 ≤0.001 1.23 1.00 1.50 0.050  5 or more prior juvenile detentions 2.72 2.14 3.46 ≤0.001 1.57 1.18 2.09 0.002 Days in prison 1.00 1.00 1.00 0.014 1.00 1.00 1.00 0.048 Unstable housing PP 1.31 1.09 1.56 0.004 1.10 0.91 1.34 0.336 Unemployed 6 months PP 1.61 1.38 1.88 ≤0.001 1.11 0.93 1.32 0.236 HR alcohol use PP 1.18 1.02 1.38 0.032 0.896 0.76 1.06 0.202 HR marijuana use PP 2.07 1.78 2.42 ≤0.001 1.32 1.11 1.56 0.002 HR methamphetamine use PP 2.14 1.84 2.50 ≤0.001 1.40 1.18 1.67 ≤0.001 HR heroin use PP 2.11 1.77 2.52 ≤0.001 1.50 1.23 1.83 ≤0.001 Not employed in prison 1.11 0.95 1.29 0.201 1.09 0.93 1.28 0.290 No. of visits 0.90 0.86 0.94 ≤0.001 0.97 0.93 1.01 0.085 No reintegration plan 0.97 0.80 1.17 0.718 0.88 0.72 1.07 0.195 Participation in transition program 0.93 0.77 1.11 0.403 0.85 0.69 1.05 0.123 Low perception of social support 1.16 0.95 1.41 0.159 0.95 0.76 1.18 0.616 High level of psychological distress 1.01 0.85 1.20 0.909 0.94 0.78 1.131 0.513 Unadjusted Adjusted Covariates HR LL UL p HR LL UL p Indigenous 1.78 1.51 2.09 ≤0.001 1.33 1.10 1.62 0.004 Gender (male) 1.24 1.02 1.51 0.030 1.39 1.11 1.73 0.004 Education achieved (≤9 years) 1.52 1.30 1.76 ≤0.001 1.00 0.84 1.18 0.952 Forcibly removed 1.35 1.12 1.62 0.001 0.85 0.69 1.05 0.122 Age at release 0.96 0.96 0.97 ≤0.001 0.97 0.96 0.98 ≤0.001 Prior adult prison sentences (first-time) ≤0.001 ≤0.001  1–4 prior imprisonments 2.87 2.31 3.56 ≤0.001 2.00 1.59 2.53 ≤0.001  5 or more prior imprisonments 4.20 3.37 5.22 ≤0.001 2.78 2.15 3.61 ≤0.001 Violent offence 1.07 0.92 1.25 0.367 1.20 0.91 1.57 0.191 Property offence 1.46 1.23 1.73 ≤0.001 1.01 0.76 1.36 0.938 Drug offence 0.54 0.36 0.83 0.005 0.85 0.52 1.39 0.520 Public order offence 0.92 0.69 1.23 0.579 1.22 0.84 1.77 0.302 Prior juvenile detentions (none) ≤0.001 0.005  1–4 prior juvenile detentions 1.92 1.60 2.30 ≤0.001 1.23 1.00 1.50 0.050  5 or more prior juvenile detentions 2.72 2.14 3.46 ≤0.001 1.57 1.18 2.09 0.002 Days in prison 1.00 1.00 1.00 0.014 1.00 1.00 1.00 0.048 Unstable housing PP 1.31 1.09 1.56 0.004 1.10 0.91 1.34 0.336 Unemployed 6 months PP 1.61 1.38 1.88 ≤0.001 1.11 0.93 1.32 0.236 HR alcohol use PP 1.18 1.02 1.38 0.032 0.896 0.76 1.06 0.202 HR marijuana use PP 2.07 1.78 2.42 ≤0.001 1.32 1.11 1.56 0.002 HR methamphetamine use PP 2.14 1.84 2.50 ≤0.001 1.40 1.18 1.67 ≤0.001 HR heroin use PP 2.11 1.77 2.52 ≤0.001 1.50 1.23 1.83 ≤0.001 Not employed in prison 1.11 0.95 1.29 0.201 1.09 0.93 1.28 0.290 No. of visits 0.90 0.86 0.94 ≤0.001 0.97 0.93 1.01 0.085 No reintegration plan 0.97 0.80 1.17 0.718 0.88 0.72 1.07 0.195 Participation in transition program 0.93 0.77 1.11 0.403 0.85 0.69 1.05 0.123 Low perception of social support 1.16 0.95 1.41 0.159 0.95 0.76 1.18 0.616 High level of psychological distress 1.01 0.85 1.20 0.909 0.94 0.78 1.131 0.513 Note: HR = high risk; LL = lower level 95% confidence interval; PP = prior-to-prison; UL = upper level 95% confidence interval. View Large To provide a comprehensive answer for why Indigenous people have an increased risk of reincarceration compared to non-Indigenous people, and to address the limitations of previous research, we conducted a multivariate Cox proportional hazard regression with all covariates in the study. In total, 11 covariates were found to be significantly associated with risk of reincarceration. Of the demographic and prior criminal history covariates included in the full multivariate model, gender (male); one to four prior adult imprisonments, and five or more prior adult imprisonments; one to four prior juvenile detentions, and five or more prior juvenile detentions and the number of days last served in prison were all associated with an increased hazard ratio for reincarceration (see Table 3). However, age at release was associated with a reduced hazard ratio for reincarceration, thus indicating that increased age at the time of release may be a protective factor for reincarceration. Of the prior-prison social experiences, high-risk marijuana use prior-to-prison, high-risk methamphetamine use prior-to-prison and high-risk heroin use prior-to-prison were all associated with an increased hazard for reincarceration. After examining the hazard ratios for the prison-life covariates in the multivariate model, results showed that prison-life experiences do not help explain why Indigenous compared to non-Indigenous people have an increased risk of reincarceration as visitation was attenuated to the null. Finally, we examined the Indigenous–reincarceration relationship after controlling for demographics, prior criminal history, prior-prison social experiences and prison-life experiences. Results showed a large attenuating effect on the hazard ratio for Indigenous peoples’ risk of reincarceration. As such, our results indicate that without controlling for the aforementioned covariates, Indigenous people compared to non-Indigenous people are 78 per cent more likely to be reincarcerated post-release. However, using a more comprehensive model to better understand the Indigenous–reincarceration relationship and control for the aforementioned covariates, Indigenous peoples’ risk of reincarceration is reduced to 33 per cent more than non-Indigenous people. Although we have explained 45 per cent of Indigenous peoples’ increased risk of reincarceration, a considerable portion (33 per cent) remains unexplained. Discussion The aim of this study was to better understand why Indigenous compared to non-Indigenous people in post-colonial societies are more at risk of reincarceration post-prison. We focused on addressing limitations from previous research by identifying group differences between Indigenous and non-Indigenous people, and examining if these differences were associated with risk of reincarceration. Furthermore, because of the extensiveness of our dataset, we were able to examine whether social experiences prior-to-prison and a person’s prison-life experiences could help explain why Indigenous people are more at risk of reincarceration than non-Indigenous people. Our findings suggest that just like the front end of the criminal justice system, Indigenous people are also over-represented at the back end of the criminal justice system, with nearly three in every four Indigenous ex-prisoners being reincarcerated within five years of being released from prison, compared to approximately one in two non-Indigenous people. This is consistent with previous literature (Broadhurst et al. 1988; Jones et al. 2006; Baldry 2009) Once we started to explore why Indigenous people were more likely to return to prison post-release, it was clear group differences in demographic, prior criminal history and social experiences prior-to-prison between Indigenous and non-Indigenous people may be contributing to different re-entry outcomes. Group differences between Indigenous and non-Indigenous people in social experiences prior-to-prison were identified, highlighting that even before participants were sentenced to prison there were differences in patterned behaviour between Indigenous and non-Indigenous people. In addition, results showed high-risk marijuana use, high-risk methamphetamine use and high-risk heroin use were significant risk factors for risk of reincarceration post-release. These results highlight two important things for consideration. First, risk of reincarceration may be predicted by patterned behaviour and experiences prior to being labelled a prisoner. Second, risk of reincarceration varies depending on type and prevalence of illicit drug use. This finding is important, for traditionally all substance use has been measured as ‘any’ substance use, treating all substances and their effects as homogenous, when in fact, they are not. Furthermore, previous research demonstrates that high-risk substance use prior-to-prison significantly predicts substance use post-prison (Kinner 2006). However, we advise caution when interpreting these results due to the small proportion of Indigenous people who self-identified to be high-risk heroin users prior-to-prison. Generalizing these findings to all Indigenous ex-prisoners without further research into type and prevalence of substance use would be detrimental to the development and implementation of treatment programs that address type and prevalence of substance use for Indigenous people. Our results suggest that prison-life experiences may not be important for explaining why Indigenous than non-Indigenous people are more at risk of reincarceration, despite there being between group differences in the number of times people were visited in the month prior baseline and transition program participation. Although a significant relationship between visitation and reincarceration was found, this relationship did not persist in the multivariate analysis. This contradicts previous work that have consistently shown visitation as a protective factor for reincarceration/reoffending (Wolff and Draine 2004; Bales and Mears 2008; Derkzen et al. 2009; Duwe and Clark 2013). When interpreting these findings, consideration should be given to the low number of visits participants received in the month prior to baseline. It stands to reason, that for visitation to be a protective factor for reincarceration, people must first be visited while in prison. It may be possible that visitation was not found to be a significant protective factor due to the limited number of participants being visited in the month prior to baseline. Of all the risk factors examined, prior adult imprisonment was found to have the strongest relationship with risk of reincarceration. The number of prior adult imprisonments for Indigenous and non-Indigenous people represents one of the largest differences between the two groups. Our research highlights prior adult imprisonment can significantly increase risk of reincarceration by up to 2.7 times that of a first-time prisoner. This is consistent with previous literature examining Indigenous reoffending (Weatherburn et al. 2003; Snowball and Weatherburn 2006; Willis 2008; Gilbert and Wilson 2009). Considering over 80 per cent of Indigenous people and 60 per cent of non-Indigenous people have previously been incarcerated, there is a need for greater emphasis to be placed on repeat prisoners’ needs, so policymakers and program administrators can better reintegrate prisoners into the community. However, results such as these are often misconstrued and used by politicians and media to highlight the need for increased supervision (over-policing) of Indigenous people, as ‘we know they are going to reoffend’; after all, the best predictor of future behaviour is past behaviour—thus identifying Indigeneity itself as a significant risk factor for crime. This practice not only promotes an individualistic approach to offender management where sole responsibility and accountability is placed on the individual, it dismisses the greater underlying social, environmental and/or psychological problems that can lead to offending behaviour, thus framing criminal behaviour as an individual problem, rather than a social health problem. People sentenced to prison not only lose their freedom, most lose their employment, homes, possessions, friends and family, and many incur a debt that cannot be repaid (Baldry et al. 2006; Makarios et al. 2010). In addition, employment skills become outdated, further increasing the difficulty of gaining employment post-prison (Bowman and Travis 2012). Stigmatization often results in a level of distrust for ex-prisoners where people are less likely to lease a residential premise, or employ them (Pager 2003; Petersilia 2003; 2005). Stigmatization may be worse for Indigenous than non-Indigenous people due to the effect of othering, as seen in US research where employers were more likely to employ a white person with a criminal record over a black person without a criminal record, let alone a black person with a criminal record (Pager 2003; 2007). As such, when a person is imprisoned, they may accumulate a higher level of disadvantage, thus increasing their risk of reincarceration. Therefore, it is important for future research to consider when examining risk of reincarceration that risk factors may be multidirectional and potentially affect risk factors at different levels of society. Implications for theory This is one of the most comprehensive studies to examine individual-level risk factors for reincarceration, yet considering risk of reincarceration still remains significantly higher for Indigenous compared to non-Indigenous people, there is still a considerable amount we cannot explain. Given Indigenous and non-Indigenous people are reincarcerated due to a multitude of variables operating at different levels of the environment in which they reside and interact, it may be time to look beyond individual-level risk factors. We argue the need for a more integrated theoretical approach to understand Indigenous and non-Indigenous peoples’ risk of reincarceration. One theory that may be useful in explaining the influence of a multitude of variables is ecological systems theory (Brofenbrenner 1979). Ecological systems theory views a person’s environment to consist of five different levels, or systems, which the individual resides in and interacts (Bronfenbrenner 1979; Zubrick and Robinson 2003; Wundersitz 2010). These systems include (1) the microsystem that consists of the individual themselves and their immediate family; (2) the mesosystem that extends beyond the individual and includes aspects such as the person’s school and work environment; (3) the exosystem that includes the individual’s community where their social support network can be found; (4) the macrosystem that through society’s institutions, cultural values, norms and ideologies influences the structure of a person’s exosystem and (5) the chronosystem that encompasses broad historical forces still impacting society today, such as colonization (Bronfenbrenner 1979; Zubrick and Robinson 2003; Wundersitz 2010). Adopting an ecological-systems approach shifts focus from an individualistic perspective where blame is firmly on the individual, to a social health perspective that considers not only individual-level variables, but historical, social, environmental and/or psychological problems that may affect an individual’s risk of reincarceration. Furthermore, not only does an ecological-systems approach allow us to focus on risk over time instead of causation, thus acknowledging not all people who exhibit risk factors for reincarceration will be reincarcerated, it also highlights risk factors for reincarceration are interconnected, multidirectional and may differ in affects at different levels of an individual’s environment. By adopting an ecological-systems perspective, we may achieve a better understanding of how experiences can affect a person’s risk of reincarceration. Implication for policy Our findings provide further evidence for why treatment programs should be individually tailored, instead of a one-size-fits-all approach, regardless of psychological, sociological or environmental differences. Considering that high-risk marijuana, methamphetamine and heroin use prior-to-prison were the only risk factors found to be significantly associated with reincarceration besides Indigenous status, gender, age and prior criminal history we support previous calls for improved accessibility to substance treatment programs in and post-prison (Kinner 2006; Kinner et al. 2009). In addition, we also reiterate calls for the incorporation of screening and targeted interventions that is complimented by tailored post-release support services based on individuals’ needs and risks given our results highlight the importance of the type, and prevalence, of substance use (Kinner 2006; Kinner et al. 2009). Limitations We did not differentiate between reincarceration for a new offence, or a parole breach. Therefore, caution should be used when interpreting our results as they may differ had we controlled for the competing risk that was present. Considering a person can breach parole and be reincarcerated for not reporting to their parole officer, which is contextually very different from being reincarcerated for a new offence, it is important for policy and practice to identify whether the risk factors for reincarceration differ for a parole breach and new offence. Given the high number of people who are returned to prison for a parole breach, this is a consideration that future research may wish to examine. Also, by relying on administrative data only for our measurement of reincarceration, it may be possible our findings reflect system practices (i.e. bias in decision-making) and not recidivism. Indigenous people have a long history of abnormally high levels of police surveillance (over-policing) (Cunneen 2001; Weatherburn et al. 2003; Blagg and Valuri 2004; Weatherburn et al. 2006). Having more system contact may affect risk of reincarceration. It may be possible that racial differences in arrest rates (Weatherburn et al. 2006) and sentencing outcomes (Jeffries and Bond 2009; Bond and Jeffries 2012) also permeate the decision-making of parole officers when deciding whether to breach Indigenous compared to non-Indigenous people. This may be a by-product of the ‘rule of colonial difference’, where the labelling of the racialized other justifies maintaining ‘…a level of surveillance, intervention and control outside of, an in contradistinction to the “universal” rights of liberal subjects’ (Baldry and Cunneen 2014: 282). Unfortunately, little research examining decision-making processes and outcomes of parole officers when managing Indigenous people exists. To understand if decision-making processes and outcomes of parole officers are differentially affecting Indigenous and non-Indigenous peoples’ risk of reincarceration more research is required. Despite the limitations, there are considerable strengths in our research. The study is one of few that employ a large representative sample of ex-prisoners, thus allowing results to be generalizable to the wider population of ex-prisoners. This is unique in re-entry research as samples in other studies contain either people sentenced to community-based corrections, and/or do not include females, or have enough females and minority participants to enable any meaningful statistical analyses to be conducted. We also had very rich baseline survey data that allowed us to examine the ‘Indigenous–reincarceration’ relationship more extensively than achieved in prior research, thus contributing valuable knowledge to a considerably under-researched area. Conclusion Reducing the over-representation of Indigenous people in prison has been an unachievable goal of governments and criminal justice systems in post-colonial countries for some time. As policymakers and correctional officers do their best to successfully reintegrate Indigenous people, there is a need for Indigenous-specific research to help understand why Indigenous people compared to non-Indigenous people are more likely to be reincarcerated. Most importantly, we need to identify what the re-entry needs are for Indigenous people, so policymakers can create informed, culturally appropriate policy that enables relevant services to facilitate positive change for Indigenous people, their family and community. Our study extends on previous prisoner re-entry literature by focusing on the risk of reincarceration for Australia’s Indigenous people, and the impact that social experiences prior-to-prison and prison-life experiences have on one’s risk of reincarceration. Furthermore, no prior re-entry study has examined the effect of type and prevalence of drug use on risk of reincarceration. Like previous literature, our study indicates that once Indigenous people experience prison life, they become embedded in the system accumulating a higher level of disadvantage, thus supporting calls for targeted intervention programs to prevent Indigenous people from being imprisoned, and the development of culturally appropriate through-care programs to successfully reintegrate Indigenous people back into the community Funding Professor Stuart Kinner receives salary support from NHMRC Senior Research Fellowship APP1078168. The Passports study was funded by NHMRC ACPP409966 and APP1002463. Acknowledgements The authors wish to thank Queensland Corrective Services for assistance with data collection, and Passports study participants for sharing their stories. The views expressed herein are solely those of the authors, and in no way reflect the views or policies of Queensland Corrective Services. References ABS . ( 2010 ), The Health and Welfare of Australia’s Aboriginal and Torres Strait Islander Peoples . Cat. No. 4704.0. Australian Bureau of Statistics . —— ( 2016a ), National Aboriginal & Torres Strait Islander Social Survey, 2014–15 . Cat.No. 4714.0. Australian Bureau of Statistics . —— ( 2016b ), Prisoners in Australia, 2016 . Cat. No. 4517.0. Australian Bureau of Statistics . —— ( 2018 ), Labour Force, Australia , May 2018. Cat. No. 6202.0. Australian Bureau of Statistics . Agnew , R . ( 2006 ), Pressured into Crime: An Overview of General Strain Theory . Roxbury . AIHW . ( 2014 ), Homelessness among Indigenous Australians . Cat. No. IHW 133. AIHW . —— ( 2015 ), The Health and Welfare of Australia’s Aboriginal and Torres Strait Islander peoples . Cat. No. IHW 147. AIHW . —— ( 2017 ), Specific Population Groups.http://www.aihw.gov.au/alcohol-and-other-drugs/data-sources/ndshs-2013/ch8/#t8_3 Alaska Judicial Council . ( 2007 ), Criminal Recidivism in Alaska . http://www.ajc.state.ak.us/reports/1-07CriminalRecidivism.pdf. Last date accessed 3 March 2017 . Amore , K . ( 2016 ), ‘ Maori Homelessness: Basic Statistics ’, Parity , 29 : 7 . Bador , T. F. , Higgins-Biddle , J. C. , Saunders , J. B. and Monterio , M. G . ( 2001 ), AUDIT. The Alcohol Use Disorders Identification Test: Guidelines for Use in Primary Care . 2 nd edition. Geneva : World Health Organisation . Retrieved 16 February 2017. http://apps.who.int/iris/bitstream/10665/67205/1/WHO_MSD_MSB_01.6a.pdf Baldry , E . ( 2009 ), ‘ Home Safely: Aboriginal Women Post-Prison and their Children’ . Indigenous Law Bulletin , 7 : 14 – 7 . Baldry , E. and Cunneen , C . ( 2014 ), ‘ Imprisoned Indigenous Women and the Shadow of Colonial Patriarchy’ , Australian & New Zealand Journal of Criminology , 47 : 276 – 98 . Google Scholar Crossref Search ADS Baldry , E. , McDonnell , D. , Maplestone , P. and Peeters , M . ( 2006 ), ‘ Ex-Prisoners Accommodation and the State: Post-Release in Australia ’, Australia and New Zealand Journal of Criminology , 39 : 20 – 33 . Google Scholar Crossref Search ADS Bales , W. D. and Mears , D. P . ( 2008 ), ‘ Inmate Social Ties and the Transition to Society: Does Visitation Reduce Recidivism?’ Journal of Research in Crime and Delinquency , 45 : 287 – 321 . Google Scholar Crossref Search ADS —— ( 2008 ), Crime, Aboriginality and the Decolonisation of Justice . Hawkins Press . Blagg , H. and Valuri , G . ( 2004 ), ‘ Aboriginal Community Patrols in Australia: Self-Policing, Self-Determination and Security’ , Policing & Society , 14 : 313 – 28 . Google Scholar Crossref Search ADS Bond , C. E. and Jeffries , S . ( 2012 ), ‘ Indigeneity and the Likelihood of Imprisonment in Queensland’s Adult and Children’s Courts’ , Psychiatry, Psychology and Law , 19 : 169 – 83 . Google Scholar Crossref Search ADS Bowman , S. W. and Travis , R . ( 2012 ), ‘ Prisoner Reentry and Recidivism According to the Formerly Incarcerated and Reentry Service Providers: A Verbal Behavior Approach’ , The Behavior Analyst Today , 13 : 9 – 19 . Google Scholar Crossref Search ADS Broadhurst , R. G. , Maller , R. A. , Maller , M. G. and Duffecy , J . ( 1988 ), ‘ Aboriginal and Non-Aboriginal Recidivism in Western Australia: A Failure Rate Analysis’ , Journal of Research in Crime and Delinquency , 25 : 83 – 108 . Google Scholar Crossref Search ADS Bronfenbrenner , U . ( 1979 ), The Ecology of Human Development: Experiments by Nature and Design . Harvard University Press . Burnett , R. and Maruna , S . ( 2004 ), ‘ So “Prison Works”, Does it? The Criminal Careers of 130 Men Released from Prison Under Home Secretary, Michael Howard’ , Howard Journal of Criminal Justice , 43 : 390 – 404 . Google Scholar Crossref Search ADS Bursill , L . ( 1999 ), Alcohol and Violence’ , in Paper presented to the Australian Institute of Criminology Conference on Best Practices Interventions in Corrections for Indigenous People . Adelaide, Australia Christensen, Erik. “Multivariate Survival Analysis using Cox’s Regression Model.” Hepatology 7.6 (1987): 1346 – 1358 . doi: 10.1002/hep.1840070628. Carroll , M. , Kinner , S. A. and Heffernan , E. B . ( 2014 ), ‘ Medication Use and Knowledge in a Sample of Indigenous and Non-Indigenous Prisoners’ , Australian and New Zealand Journal of Public Health , 38 : 142 – 6 . Google Scholar Crossref Search ADS PubMed Cox , D. R. and Oakes , D . ( 1984 ), Analysis of Survival Data , Vol. 21 . CRC Press . Cunneen , C . ( 2001 ), Conflict, Politics and Crime: Aboriginal Communities and the Police . Crows Nest, NSW : Allen & Unwin . —— ( 2006 ), ‘ Racism, Discrimination and the Over-Representation of Indigenous People in the Criminal Justice System: Some Conceptual and Explanatory issues’ , Current Issues in Criminal Justice , 17 : 329 – 46 . Crossref Search ADS Cutcher , Z. , Degenhardt , L. , Alati , R. and Kinner , S. A . ( 2014 ), ‘ Poor Health and Social Outcomes for Ex-Prisoners with a History of Mental Disorder: A Longitudinal Study’ , Australian and New Zealand Journal of Public Health , 38 : 424 – 9 . doi: https://doi.org/10.1111/1753–6405.12207 . Google Scholar Crossref Search ADS PubMed Derkzen , D. , Gobeil , R. and Gileno , J . ( 2009 ), Visitation and Post-Release Outcomes Among Federally-Sentenced Offenders. Research Report . Correctional Service of Canada . Dugbaza , T. , Scott , B. and McKeawn , S . ( 2012 ), National Best Practice Guidelines for Indigenous Data Linkage Activities Relating to Aboriginal and Torres Strait Islander People . AIHW . Duwe , G. and Clark , V . ( 2013 ), ‘ Blessed Be the Social Tie That Binds: The Effects of Prison Visitation on Offender Recidivism’ , Criminal Justice Policy Review , 24 : 271 – 96 . doi: https://doi.org/10.1177/0887403411429724 . Google Scholar Crossref Search ADS Fazel , S. , Bains , P. and Doll , H . ( 2006 ), ‘ Substance Abuse and Dependence in Prisoners: A Systematic Review’ , Addiction , 101 : 181 – 91 . Google Scholar Crossref Search ADS PubMed Fergusson , D. M. , Boden , J. M. and Horwood , L. J . ( 2013 ), ‘ Alcohol Misuse and Psychosocial Outcomes in Young Adulthood: Results from a Longitudinal Birth Cohort Studied to age 30’ , Drug and Alcohol Dependence , 133 : 513 – 9 . Google Scholar Crossref Search ADS PubMed Gilbert , R. and Wilson , A . ( 2009 ), Staying Strong on the Outside: Improving the Post-Release Experience of Indigenous Young Adults . Indigenous Justice Clearinghouse . Giles , M. and Whale , J . ( 2016 ), Models of the Impacts of Prison Education and Training on Welfare Dependence and Recidivism: Western Australia. Report to Australian Institute of Criminology and Phase 3 Report . Centre for Innovative Practice, Edith Cowan University . Heffernan , E. B. , Anderson , K. C. , Dev , A. and Kinner , S . ( 2012 ), ‘ Prevalence of Mental Illness Among Aboriginal and Torres Strait Islander People in Queensland prisons’ , Medical Journal of Australia , 197 : 37 . Google Scholar Crossref Search ADS PubMed Huebner , M. B. and Berg , M. T . ( 2011 ), ‘ Examining the Sources of Variation in Risk for Recidivism’ , Justice Quarterly , 28 : 146 – 73 . Google Scholar Crossref Search ADS Humeniuk , R. , Henry-Edwards , S. , Ali , R. , Poznyak , V. and Monterio , M. G . ( 2010 ), The Alcohol, Smoking and Substance Involvement Screening Test (ASSIST): Manual for use in Primary Care . World Health Organisation . Jeffries , S. and Bond , C . ( 2009 ), ‘ Does Indigeneity Matter? Sentencing Indigenous Offenders in South Australia’s Higher Courts’ , Australian & New Zealand Journal of Criminology , 42 : 47 – 71 . Google Scholar Crossref Search ADS Johnston , E . ( 1991 ), Royal Commission in to Aboriginal Deaths in Custody . National Report, 5 Volumes, AGPS, Canberra. Jones , C. , Hua , J. , Donnelly , N. , McHutchinson , J. and Heggie , K . ( 2006 ), ‘ Risk of Re-Offending Among Parolees’ , Contemporary Issues in Crime and Justice , no. 91. NSW Bureau of Crime Statistics and Research . Jordan , M . ( 2010 ), ‘ Embracing the Notion that Context is Crucial in Prison Mental Health Care’ , British Journal of Forensic Practice , 12 : 26 – 35 . Google Scholar Crossref Search ADS Kessler , R. C. , Andrews , G. , Colpe , L. J. , Hiripi , E. , Mroczek , D. K. , Normand , S. L. , Walters , E. E. and Zaslavsky , A. M . ( 2002 ), ‘ Short Screening Scales to Monitor Population Prevalences and Trends in Non-Specific Psychological Distress’ , Psychological Medicine , 32 : 959 – 76 . doi: https://doi.org/10.1017/S0033291702006074 . Google Scholar Crossref Search ADS PubMed Kinner , S. A . ( 2006 ), ‘ Continuity of Health Impairment and Substance Misuse Among Adult Prisoners in Queensland, Australia’ , International Journal of Prisoner Health , 2 : 101 – 13 . doi: https://doi.org/10/1080/17449200600935711 . Google Scholar Crossref Search ADS Kinner , S. A. , George , J. , Campbell , G. and Degenhardt , L . ( 2009 ), ‘ Crime, Drugs and Distress: Patterns of Drug Use and Harm Among Criminally Involved Injecting Drug Users in Australia’ , Australian and New Zealand Journal of Public Health , 33 : 223 – 7 . Google Scholar Crossref Search ADS PubMed Kinner , S. A. , Lennox , N. , Williams , G. W. , Carrol , M. , Quinn , B. , Boyle , F. , and Alati , R . ( 2013 ), ‘ Randomised Controlled Trial of a Service Brokerage Intervention for Ex-Prisoners in Australia’ , Contemporary Clinical Trials , 36 : 198 – 206 . doi: https://doi.org/10.1016/jcct.2013.07.001 Google Scholar Crossref Search ADS PubMed Kopak , A. M. and Hoffmann , N. G . ( 2014 ), ‘ Pathways Between Substance Use, Dependence, Offense Type, and Offense Severity’ , Criminal Justice Policy Review , 25 : 743 – 60 . Google Scholar Crossref Search ADS Korff , J . ( 2016 ), Mental Health at its Worst in Prison . Retrieved 13 July 2017. https://www.creativespirits.info/aboriginalculture/law/mental-health-at-its-worst-in-prison Larney , S. and Burns , L . ( 2011 ), ‘ Evaluating Health Outcomes of Criminal Justice Populations Using Record Linkage: The Importance of Aliases’ , Evaluation Review , 35 : 118 – 28 . doi: https://doi.org/10.1177/0193841x11401695 . Google Scholar Crossref Search ADS PubMed Makarios , M. , Steiner , B. and Travis , L. F . ( 2010 ), ‘ Examining the Predictors of Recidivism Among Men and Women Released from Prison in Ohio’ , Criminal Justice and Behavior , 37 : 1377 – 91 . doi: https://doi.org/10.1177/0093854810382876 . Google Scholar Crossref Search ADS Mallik-Kane , K. and Visher , C. A . ( 2008 ), Health and Prisoner Reentry: How Physical, Mental, and Substance Abuse Conditions Shape the Process of Reintegration . Urban Institute . Mears , D. P. , Cochran , J. C. , Siennick , S. E. and Bales , W. D . ( 2012 ), ‘ Prison Visitation and Recidivism’ , Justice Quarterly , 29 : 888 – 918 . Google Scholar Crossref Search ADS Ministry of Health . ( 2016 ), Annual Update of Key Results 2015/16: New Zealand Health Survey . Retrieved 13 February 2017. https://www.health.govt.nz/publication/annual-update-key-results-2015-16-new-zealand-health-survey Mitchell , P. H. , Powell , L. , Blumenthal , J. , Norten , J. , Ironson , G. , Pitula , C. R. , Froelicher , E. S. , Czajkowski , S. , Youngblood , M. , Huber , M. and Berkman , L. F . ( 2003 ), ‘ A Short Social Support Measure for Patients Recovering from Myocardial Infarction: the ENRICHD Social Support Inventory’ , Journal of Cardiopulmonary Rehabilitation and Prevention , 23 : 398 – 403 . Google Scholar Crossref Search ADS NZStats . ( 2017 ), Annual Sentenced Prisoner Population for the latest Fiscal Years (ANZSOC) . Retrieved 12 July 2018. http://nzdotstat.stats.govt.nz/wbos/Index.aspx?DataSetCode=TABLECODE7322&_ga=2.151028542.642037166.1531319055-963756003.1531319055 Pager , D . ( 2003 ), The Mark of a Criminal Record . American Journal of Sociology , 108 : 937 – 975 Google Scholar Crossref Search ADS —— ( 2007 ), Two Strikes and You’re Out: The Intensification of Racial and Criminal Stigma , in S. Bushway , M. Stoll , and D. Weiman eds., Barriers to Reentry? The Labor Market for Released Prisoners in Post-Industrial America , 151 – 73 . Russell Sage Foundation . Patrick , C . ( 2014 ), Aboriginal Homelessness in Canada: A Literature Review . Canadian Homelessness Research Network Press . Pauly , B. B. , Gray , E. , Perkin , K. , Chow , C. , Vallance , K. , Krysowaty , B. and Stockwell , T . ( 2016 ), ‘ Finding Safety: A Pilot Study of Managed Alcohol Program Participants’ Perceptions of Housing and Quality of Life’ , Harm Reduction Journal , 13 : 15 . Google Scholar Crossref Search ADS PubMed Payne , J . ( 2007 ), Recidivism in Australia: Findings and Future Research (No. 80) . Australian Institute of Criminology . Pearson , F. S. , Lipton , D. S. , Cleland , C. M. and Yee , D. S . ( 2002 ), ‘ The Effects of Behavioral/Cognitive-Behavioral Programs on Recidivism’ , Crime & Delinquency , 48 : 476 – 96 . Google Scholar Crossref Search ADS Petersilia , J . ( 2003 ), When Prisoners Return to Communities: Political, Economic and Social Consequences . Oxford University Press . —— ( 2005 ), ‘ Hard Time: Ex-Offenders Returning Home After Prison’ , Corrections Today , 67 : 66 – 72 . Pindus , N. , Kingsley , G. T. , Biess , J. , Levy , D. , Simington , J. , Hayes and Hayes , C . ( 2017 ), Housing Needs of American Indians and Alaska Natives in Tribal Areas: A Report From the Assessment of American Indian, Alaska Native, and Native Hawaiian Housing Needs, Executive Summary . Urban Institute: US Department of Housing and Urban Development Office of Policy Development and Research . Prendergast , M. , Hall , E. and Wellisch , J . ( 2002 ), An Outcome Evaluation of the Forever Free Substance Abuse Treatment Program: One-Year Post-Release Outcomes . Drug Abuse Research Center . Queensland Government . ( 2018 ). Leaving Prison , https://www.qld.gov.au/law/sentencing-prisons-and-probation/prisons-and-detention-centres/leaving-prison. Last accessed 15 June 2018. Snowball , L. and Weatherburn , D . ( 2006 ), ‘ Indigenous Over-Representation in Prison: The role of Offender Characteristics’ , BOCSAR NSW Crime and Justice Bulletins , 20 : 1 – 20 . Stahler , G. J. , Mennis , J. , Belenko , S. , Welsh , W. N. , Hiller , M. L. and Zajac , G . ( 2013 ), ‘ Predicting Recidivism for Released State Prison Offenders: Examining the Influence of Individual and Neighbourhood Characteristics and Spatial Contagion on the Likelihood of Reincarceration’ , Criminal Justice and Behavior , 40 : 690 – 711 . doi: https://doi.org/10.1177/0093854812469609 . Google Scholar Crossref Search ADS PubMed Statistics Canada . ( 2018 ), Adult Custody Admissions to Correctional Services by Aboriginal Identity . Retrieved 12 July 2018. https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=3510001601 Stubbs , J . ( 2011 ), ‘ Indigenous Women in Australia Criminal Justice: Over-Represented But Rarely Acknowledged’ . Australian Indigenous Law Review , 15 : 47 – 63 . The Sentencing Project . ( 2016 ), Race & Justice News: Native Americans in the Justice System . Retrieved 3 May 2018. https://www.sentencingproject.org/news/race-justice-news-native-americans-in-the-justice-system/ Thomas , E. , Degenhardt , L. , Alati , R. and Kinner , S. A . ( 2013 ), ‘ Predictive Validity of the AUDIT for Hazardous Alcohol Consumption in Recently Released Prisoners’ , Drug and Alcohol Dependence , 134 : 322 – 9 . Google Scholar Crossref Search ADS PubMed Thomas , E. G. , Spittal , M. J. , Heffernan , E. B. , Taxman , F. S. , Alati , R. and Kinner , S. A . ( 2016 ), ‘ Trajectories of Psychological Distress After Prison Release: Implications for Mental Health Service Need in Ex-Prisoners’ , Psychological Medicine , 46 : 611 – 21 . doi: https://doi.org/10.1017/S0033291715002123 . Google Scholar Crossref Search ADS PubMed Travis , J. , Solomon , A. L. and Waul , M . ( 2001 ), From Prison to Home: The Dimensions and Consequences of Prisoner Reentry . Urban Institute . Ulmer , J. T . ( 2001 ), ‘ Intermediate Sanctions: A Comparative Analysis of the Probability and Severity of Recidivism’ , Sociological Inquiry , 71 : 164 – 93 . Google Scholar Crossref Search ADS United States Sentencing Commission . ( 2013 ), Quick Facts-Native Americans in the Federal Offender Population . Retrieved 3 May 2017. https://www.ussc.gov/sites/default/files/pdf/research-and-publications/quick-facts/Quick_Facts_Native_American_Offenders.pdf US Census Bureau . ( 2015 ), 2015 American Community Survey 1-Year Estimates . Bureau of Labor Statistics, United States Department of Commerce . Visher , C. A. , La Vigne , N. G. and Farrell , J . ( 2003 ), Illinois Prisoners’ Reflections on Returning Home . The Urban Institute . Visher , C. A. and O’Connell , D. J . ( 2012 ), ‘ Incarceration and Inmates’ Self-Perceptions about Returning Home’ , Journal of Criminal Justice , 40 : 386 – 93 . Google Scholar Crossref Search ADS Weatherburn , D . ( 2010 ), ‘ The Effect of Prison on Adult Re-Offending’ , Contemporary Issues in Crime and Justice . No. 143. NSW Bureau of Crime Statistics and Research . Weatherburn , D. , Fitzgerald , J. and Hua , J . ( 2003 ), ‘ Reducing Aboriginal Over-Representation in Prison’ , Australian Journal of Public Administration , 62 : 65 – 73 . Google Scholar Crossref Search ADS Weatherburn , D. and Lind , B . ( 2001 ), Delinquent Prone Communities . Cambridge University Press . Weatherburn , D. , Snowball , L. and Hunter , B . ( 2006 ), ‘ The Economic and Social Factors Underpinning Indigenous Contact With the Justice System: Results From the 2002 NATSISS Survey’ , Crime and Justice Bulletin , No. 104. NSW Bureau of Crime Statistics and Research . WHO . ( 2002 ), ‘ The Alcohol, Smoking and Substance Involvement Screening Test (ASSIST): Development, Reliability and Feasibility’ , Addiction , 97 : 1183 – 94 . Crossref Search ADS PubMed Willis , M . ( 2008 ), ‘ Reintegration of Indigenous Prisoners: Key Findings’ , Trends & Issues in Crime & Criminal Justice , 364 : 1 – 6 . Willis , M. and Moore , J . ( 2008 ), Reintegration of Indigenous Prisoners. Research and Public Policy Series . No. 90. Australian Institute of Criminology . Wolff , N. and Draine , J . ( 2004 ), ‘ Dynamics of Social Capital of Prisoners and Community Reentry: Ties that Bind ?’ Journal of Correctional Health Care , 10 : 457 – 90 . Google Scholar Crossref Search ADS Wundersitz , J . ( 2010 ), Indigenous Perpetrators of Violence: Prevalence and Risk Factors for Offending . Research and Public Policy Series. No. 105. Australian Institute of Criminology . Yavu-Kama-Harathunian , C . ( 2002 ), The Therapeutic Needs of Indigenous Violent and Sex Offender Males: How can They be Addressed. Paper presented at Probation and Community Corrections: Making the Community Safer Conference, AIC, Perth, 23–24 September 2002. Zubrick , S. R. and Robson , A . ( 2003 ), Resilience to Offending in High-Risk Groups: Focus on Aboriginal Youth: Research Agenda . Criminology Research Council . https://www.ncjrs.gov/App/abstractdb/AbstractDBDetails.aspx?id=206191 Footnotes 1 Prison populations in Australia are not differentiated by federal and state offences. Although jurisdictional differences in legislation between state and federal governments exist, regardless of whether an offender is sentenced to prison for a state or federal crime, the responsibility of housing and rehabilitating all prisoners is overseen by state-run correctional facilities. It is possible that offenders sentenced for minor infractions (i.e., traffic offence) could be housed with offenders sentenced for murder. 2 Soon-to-be-released prisoners are provided support to create a re-entry plan that identifies their risks and needs and relevant available support services in the community that they can access to help adjust to life post-prison (Queensland Government 2018). However, it must be noted that not all prisoners in Queensland are eligible for this support, and the specific design and support provided in the plan can change depending on the prison that the prisoner is being released from, and the prisoners gender (Queensland Government 2018). Although these are important to consider, accounting for the differences is beyond the scope of this article. © The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). All rights reserved. For permissions, please e-mail: [email protected] This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)