The Economic Journal

Savin,, Andrej

doi: 10.1111/j.1468-0297.2006.01069_3.xpmid: N/A

The economic analysis of intellectual property law is a difficult subject. Unlike some other branches of law that show more harmony, intellectual property law consists of a range of areas influenced by statutes and common law and underlined by common principles. By and large, we do not yet properly understand the economic effects of these rules. Also, we do not yet know to what extent the lawmakers are influenced by the economics of intellectual property. Although there has been a notable increase in the amount of relevant academic material in the past two decades, only the unprecedented technological advances of the past years made it one of the most dynamic areas of law that it is today. Blair and Cotter's Intellectual Property – Economic and Legal Dimensions of Rights and Remedies, only the second full‐length book on the subject to date, falls into that still somewhat rare category of works that analyse intellectual property law from an economic perspective. Blair and Cotter's book aims to fill some of the gaps in literature by giving a comprehensive overview of the economics of the enforcement process of intellectual property rights. This perspective distinguishes it from other similar works, notably Landes and Posner's The Economic Structure of Intellectual Property Law (Belknap, Harvard 2003). Whereas the latter attempts to give an overview of the field, the former restrict themselves to rights and remedies. The book follows up on authors’ previous research in this area and is limited almost entirely to US law. In an interconnected world, influenced by international efforts and plagued by global worries, such as patents on drugs or the Internet, this may be seen as a missed opportunity. After a brief introduction into the field of intellectual property law, the authors propose a general theory of damages rules. According to their model, the plaintiff in a patent, copyright, trademark or trade secret dispute should be able to recover ‘the greater of her lost profits attributable to the infringements or the defendant's profits so attributable’ (page 68). They then proceed to examine the departure from this model in each of the areas of intellectual property law they address (chapter 4), moving on to the problem of liability standards for intellectual property rights (chapter 5), and issues concerning proper identification of defendants and plaintiffs (chapters 6 and 7) concluding with the calculation of monetary damages (chapter 8). The importance of Blaire and Cotter's work lies in its context. The law and economics of IP enforcement has not caught proper attention in academic literature and is generally poorly understood. The detailed and competent coverage offered here may serve as a starting point for further research. The model the authors take as a basis for their book can be a convenient tool for analysing the remedies, even if it is at times speculative, lacking the detailed analysis of the evidence. In their model, damages either compensate the victims or deter potential infringers. This model, although appealing by its simplicity, does not reflect the complexity of the American intellectual property system. Finally, the authors focus on damages and their structure and operation. It remains unclear why so little attention has been paid to injunctive relief, which, at least in the United States, has generated a lot of attention in recent years. Although the authors expressly disclaim the analysis of normative, believing in a ‘platform‐independent’ model (page 5), and assuming that the scope and extension of protection are set, doubt remains if remedies can completely be understood from this standpoint. This is mainly because statistical evidence has not been provided and court practice in the United States has been analysed to a limited extent, sometimes putting the authors’ work more on the speculative side. The merit of this, as they point out, is the model's universality. The economic analysis of intellectual property law, as a discipline, eventually attempts to answer the question of the extent of protection accorded to intellectual property. Although the economics of intellectual property law can be analysed even if the crucial question of the right balance between incentive and access is not addressed, in other words, if the political side of the question is left out, this approach leaves some important question unanswered and notably that of the causes of expanding intellectual property protection. In this respect, the peculiar, somewhat theoretical, perspective of the book need not be taken as a weakness, but rather as a measure of the complexity of the questions posed. It is a needed and welcome addition to the literature on the subject that has grown to be so important. © Royal Economic Society 2006.

Backhouse, Roger, E.

doi: 10.1111/j.1468-0297.2006.01069_7.xpmid: N/A

Robert Lowe started his career as an economist as a leader writer for The Times in the year of the Great Exhibition and just a few years after John Stuart Mill's Principles had provided the definitive treatment of the subject. The Corn Laws had been repealed, the Bank Charter Act had been passed, and Britain was experiencing prosperity that was widely attributed to free trade. Confidence in classical political economy was a its height and Lowe was one of its leading representatives: he was one of Marx's ‘vulgar’ political economists, never wavering in his confidence that Adam Smith had laid down the principles of the subject. However, this does not make Lowe uninteresting. He was an economist when classical economics was being challenged by moralist, socialist and historicist critics, and by the beginnings of ‘marginalist’ economics in the 1870s. He was a front‐bench politician – in turn Vice President of the Board of Trade, Vice President of the Board of Education, Chancellor of the Exchequer and Home Secretary. Britain was negotiating commercial treaties (which were agreed) and European monetary union (which was not), embarking on a historic shift from indirect to direct taxation, and moving towards a compulsory education system. This makes Lowe a fascinating subject for study. He was a colourful character. Maloney (page 3) points out that ‘he was the only Chancellor to have his skull fractured at an election meeting, or to be challenged to two duels, or to take refuge from demonstrators underneath the palace of Westminster, or to have to withdraw his entire budget.’ He was not afraid to court controversy (‘a hard, fierce man with a gift for enmity’), and was almost alone in being willing to challenge Mill on economic questions in Parliament. Maloney convincingly makes the case that Lowe was not an unthinking exponent of laisser faire: he was ‘a Benthamite coming to terms with the power of the nineteenth‐century state’ (ibid.). He never abandoned the principles of Adam Smith, but was creative in the way these were understood and applied. Consistency was not one of his virtues. For example, he could both praise the Bank Charter Act, and advocate policies that effectively undermined it. He changed his position on most issues, sometimes for good reasons, sometimes for bad. Though his economics was based on deduction from ‘true’ premises, which did not change, conclusions had to be take account of circumstances, so summarising the way his views developed is very difficult. He believed in unilateral free trade but changed his mind over supporting a commercial treaty that lowered tariffs. Discussions of education had to take account not only of whether parents knew what was best for their children but also of the need to educate those who were receiving the votes in the Reform Acts of 1867 and 1884. Tax policy had not only to avoid introducing distortions but also to balance the interests of different classes. Monetary union (replacing the pound with a new one worth 25 Francs, thereby reducing its gold content by l%) raised the question of what it meant to honour existing debt contracts. There was ample scope for Lowe to swing back and forth between inconsistent positions, and sometimes to hold them simultaneously. Maloney's analysis of Lowe's views proceeds from issue to issue, though in most cases this corresponds to the development of his career. Though he tries to understand the context in which Lowe's ideas developed (without which it would be impossible even to begin to understand them) he is not afraid to comment on how those views seem to a modern observer. There is a very modern tone to many of the debates: perhaps because of the way Maloney tells the story, but perhaps because many of the issues are contemporary ones too (the size of government, the mix of direct and indirect taxation, monetary union, education – even the merits of a national curriculum). Occasionally it was frustrating not finding out what happened after Lowe left a field. For example, as Chancellor, Lowe's caution led to a series of surpluses (greater than anticipated) and repayment of the national debt. But his resignation in 1873, a cyclical peak perhaps comparable to the one that occurred exactly a century later, meant that one never found out what happened afterwards, and whether Lowe and his contemporaries realised the link between the business cycle and the budget. The book is written in a lively, colourful style that fits its subject and, even if he had not stated this, it would be obvious that Maloney very much enjoyed writing the book. It shows that the three decades after 1850 were a far more interesting period in British economics than one would have inferred from most textbooks on the history of economic thought. It is well worth reading. © Royal Economic Society 2006.

ter Weel,, Bas

doi: 10.1111/j.1468-0297.2006.01061.xpmid: N/A

Abstract The emergence and rapid diffusion of information and communication technology is the most radical technological transformation of the last couple of decades and has had an enormous economic impact. This bundle of four articles contains research that discusses major economic implications of the diffusion of IT. They focus on two main implications of IT: (i) networks and industry dynamics, and (ii) IT and workplace organisation. The two sets of two articles contribute both original theoretical insights and new empirical results to these themes. The adoption and rapid diffusion of the bundle of new information and communication technologies has been the most radical technological change of recent decades. Economists have been focussing on the economic consequences for education, wages, workplace organisation, markets and competition, and networks and innovation of this technological revolution for quite some time now. The four contributions cover core issues of networks and industry dynamics, and workplace organisation and the demand for labour in relation to the IT revolution. The IT revolution has caused the costs of many kinds of interactions to drop by making a great many processes operate more efficiently and it has allowed the opportunity to engage in new interactions that have become cost effective. At first, revolutionary technologies are implemented in a rather mechanical way to make it less costly to produce the same level of output. The main reason for Charles Babbage to manufacture and use a ‘computer’ was to deal more accurately and efficiently with information (Babbage, 1832).1 Such ‘number crunchers’ were the main appearance of computers until the late 1960s mainly used to carry out mathematical and statistical procedures, mass‐integrated data processing and simulations of Bayesian decision‐making.2 Over time, a revolutionary technology changes and amends (in an endogenous or co‐invention manner) to do entirely new things, such as to change the way in which markets are functioning and structured, demand and supply are brought together, or restructure the way in which a firm is organised and influence the way in which innovation activities are pursued.3 These more incremental changes or innovations are the result of the opportunities offered by the main breakthrough offered by ‘general purpose technologies’ (e.g., Bresnahan and Trajtenberg, 1995). These latter consequences of new technologies are often hard to predict at first appearance but lie at the heart of the study of the economic consequences of technological change. An example of how difficult it appears to predict technological changes can be found in a study of Klahr and Leavitt (1967). They acknowledged that in the mid‐1960s computer technology used in firms was being complemented by communication technology but they could not forecast the way in which computer technology would change decision making inside firms:4 ‘Presumably the organisational impact of this wave [of computer technology] may be broader and may move into somewhat higher levels of the organisational pyramid. One effect of this new thrust seems to be the binding together of several subparts of the total structure, often severely modifying some in the process. Another is to move the locus of large amounts of information to some central point; and, in some cases, to change the locus of certain decisions from one part of the structure to another. The still speculative part of real‐time information systems forecasts the provision of instantaneous information to top executives at their will – though no one is quite sure what they should will, or how such innovations will affect present structures.’ (p. 108). This view on IT describes the developments until the mid‐1980s relatively well, because mainframe computers were in heavy use and companies such as IBM and Computer Associates were big players in the market for hardware and software. Sweeping changes took place in the early 1990s, when smaller, more mobile and networked computer technology became available. This client‐server type of computing gave companies more flexibility in applying information technologies without losing computing power. In addition, the compatibility of different types of software and hardware increased substantially, making an impact on the way in which agents work together within organisations and ‘meet’ suppliers in consumer markets. Client‐server computing puts the power of the mainframe computer into a server and networks permit a business system to run on less powerful and more mobile clients.5Bresnahan and Greenstein (1996) use Jonathan Swift's Gulliver's Travels to describe the advantage of client‐server computing compared to mainframe computing: ‘Before 1989 workstations and personal computers could no more replace mainframes than could the people of Lilliput wrestle Gulliver to the ground. Yet, like the Lilliputians’ ropes, networking cables created strength from numbers. Workstations, a technology originally intended to serve individuals, were deployed as servers. They did not need as much capacity as hosts, since PCs, deployed as clients, assumed some of the computing tasks (such as effectively interfacing with people). This technical opportunity produced large market and organisational change. IBM, having dominated large‐scale business computing for decades, no longer found itself a leader. The MIS departments of large organisations were excited as well as threatened by the new opportunities. Both IBM and these departments began to appear as the inert Gulliver, roped and staked to the ground.’ (pp. 3–4) It is the economic aspects of this modern form of computer technology this Feature is concerned with. In particular, this Feature consists of a collection of articles intended to provide an overview of and new insights into the industry and labour‐market effects of information and communication technology diffusion over the past two decades. There are two major themes: (i) networks and industry dynamics, and (ii) IT and workplace organisation. Both sets of articles have three things in common. On the empirical side it has been hard to identify network effects and changes in workplace organisation from existing data sources and the contribution of this Feature is to give new estimates by exploring network effects in the global market for workgroup servers and by investigating the consequences of computer technology adoption for workplace organisation and task assignment of different types of workers, using several large establishment/firm level data sets from the Netherlands and France. On the methodological side this Feature offers new insights into the role of IT in improving the matching between economic agents by reducing search frictions and in establishing a more efficient division of labour by looking at productivity increases and falling communication costs. These theoretical contributions add rigour to our understanding of the fundamental changes computer technologies have brought about. Finally, each of the contributions gives a perspective on useful and interesting future research into studying technological changes in networks from a theoretical and empirical point of view, and a research perspective on how to improve the understanding of the interplay between workplace organisation and the demand for labour in several ways. Together these four articles add to the substantial and increasing body of papers in this research area and yield a consistent understanding of how the IT revolution has had an impact and still affects market structures and the need for insight on how IT has generated market power for some parties, and how it has affected workplace organisation and the demand for labour. 1. Networks and Industry Dynamics To maintain the analogy with Gulliver's Travels the first two contributions in this Feature are about the robes, nodes and strength of the ‘network’ with which Gulliver was tied by the people of Lilliput. From a theoretical point of view IT is able to have an influence on all three features of the network. First, it could establish new robes that were previously too expensive to explore, which extends the breath of the network; it could make the nodes denser in the sense that more information can come together in one node and in the sense that agents can meet more often; and, it has the potential to strengthen the network by reducing the number of weak links and by making markets more transparent. This could induce price decreases and quality improvements of the products and services traded in the network but it could also induce market power and lead to lower equilibrium quality and higher prices when one single player supplying a particular product becomes dominant. Gerard van den Berg's theory article ‘Revolutionary Effects of New Information Technologies’ explores the market consequences of IT by focusing on the interaction between IT and differences in the composition of production technologies across agents or firms (van den Berg, 2006). The main aim of his contribution is to analyse the effects on the search process of agents of improvements in the information technology they have at their disposal. In his model the adoption of a novel technology improves the realisation of good matches for single agents, because improvements in IT increase the rate at which agents meet producers. This yields a better overview and a larger choice set on the side of agents. At the market level this may entail the removal of inefficient production technologies, because agents require less effort to find superior production technologies at lower prices. This gain in efficiency leads to upward jumps in the performance of networks. The key feature of the model is that due to information frictions an equilibrium exists in which inferior and superior production technologies coexist, but a decrease in the amount of information frictions results in firms using inferior production technologies disappearing from the market. The specific role of IT is to improve the technology of information processing and reduce the amount of frictions in the market. This in turn may result in the adoption of new production technologies and affect the composition of different technologies in use, which will be more efficient. This simple model of IT reducing search frictions is able to explain a number of recently observed phenomena, such as the success of Internet job search in improving the quality of matches between workers and firms.6 Another key insight of the model is that the effects of improved information technologies spill over to other markets in which the technology is not being used. The mechanism causing this effect is driven by agents’ beliefs. At first the good equilibrium could be obtained but many markets are stuck in bad equilibria. After the revolution and the successful adoption of IT in some markets, the good equilibrium is attained in these markets. Now, if agents are trading simultaneously in different markets using different information technologies or if they are able to apprehend the changes in search behaviour in sectors in which the new technology is adopted, they start acting as if the new technology is implemented in all sectors of the economy. This induces a shift from the inefficient to the efficient equilibrium and drives out bad firms. Finally, the implications of the exit of bad firms and the flourishing of good firms are numerous. Some of the most important implications are the following. First, the growth rate of the economy increases and the agents’ purchasing power rises because of falling prices, which is consistent with low inflation during the 1990s. Second, long‐run growth levels remain similar because the long‐run growth rate is determined by the arrival rate of new technologies, which is assumed constant. Finally, the exit of inferior technologies reduces the prices of commodities and price dispersion, which is consistent with findings suggesting that prices of commodities fall and converge if they become increasingly traded via the Internet. John Van Reenen analyses a specific example of changing markets as a result of IT and his focus in ‘The Growth of Network Computing: Quality‐adjusted Prices for Network Servers’ is especially on quality‐adjusted prices in the network server market (Van Reenen, 2006). Since the late 1980s computing architecture went through a paradigm shift, in which mainframe‐orientated systems were replaced by PC client/server architecture as briefly summarised above. Instead of computer intelligence being centralised and users interacting via terminals, processing power was more decentralised, distributed between PCs with their own operating systems and servers linking these PCs together. This market for servers has become a major antitrust concern in Europe and the US in recent years.7 The present analysis is mainly concerned with (i) analysing the market for network servers and (ii) estimating the quality‐adjusted prices for network servers using hedonic price indices. The server market is analysed from 1996 to 2001 and characterised by a number of large players on both the hardware and the software side. Servers are mainly used to share computer facilities among users. For example, servers are applied to handle security, share files, process large data files, carry out print jobs and send emails. The market for servers is relatively large with companies spending about one‐third of their annual computer budget on servers. To estimate the prices of network servers over time for Europe and the US both the rapid decrease in observed prices and the increase in quality have to be taken into account. In addition, inferior technologies have exited the market, which leads to a selection bias in the estimated price if the prices of these technologies are not taken into account. Furthermore, a hedonic index does have estimation (as well as sampling) variance and hence can be imprecise compared to a matched model of measuring prices. These three reasons form the basis of selecting a ‘hybrid model’ comparable and in line with Pakes (2003), which includes variance reduction features and selection correction properties. Using data from a server tracker database combined with sales data from the major players in the market for all quarters between 1996/1 and 2001/4 the evolution of prices, revenues and number of servers sold can be investigated. The main finding of the article is that quality‐adjusted prices for network servers have fallen dramatically by some 30% per year over the period 1996–2001. The estimates reveal that using the more precise hybrid model yields an estimate almost 15 percentage points higher than when using a matched model only, which does not take exit into account. These price falls have been similar in Europe and the US. The general interest in these results is that the improved quality and increasing presence of network servers has improved information flows between agents and driven out inferior models, which contributes to the switching to an equilibrium in which the average price and price dispersion is lower. This switch to an efficient equilibrium is likely to have contributed significantly to the dramatic fall in server prices since the mid‐1990s, which is consistent with the theoretical predictions of reducing market frictions proposed by van den Berg (2006). 2. IT, Workplace Organisation and The Demand for Labour The tying down of Gulliver required a huge coordination and team effort involving a great many citizens of Lilliput. In the first chapter, right after he landed and recovered his consciousness, Gulliver notes that ‘[t]hese people are most excellent mathematicians’ to be able to get all different tasks fulfilled to tie him down. The division of labour into separate tasks or jobs is determined by the benefits of specialisation and the cost of communication. The IT revolution has improved communication possibilities as well as production opportunities and this has affected the division of labour and the demand for labour in fundamental ways. The two contributions by Lex Borghans and Bas ter Weel, and Patrick Aubert, Eve Caroli and Muriel Roger complement the two studies discussed above by investigating the interplay and complementarity between IT, workplace organisation and the demand for labour within the firm. In ‘The Division of Labour, Worker Organisation, and Technological Change’Borghans and ter Weel (2006) investigate the relationship between the adoption of computer technology and workplace organisation both from a theoretical and empirical point of view. The main aim of the article is to present a simple model of the costs and benefits of specialisation to examine how far one can go toward explaining changes in the division of labour resulting from the adoption of IT. A second aim is to estimate the key predictions of the model using a panel of Dutch establishments surveyed in the 1990s. The model is based on a density function of a continuum of tasks, representing the work that has to be carried out in a firm. The way a firm assigns these tasks to workers describes the organisational structure of a firm.8 Such an organisational structure is determined by the trade‐off between the benefits of specialisation and communication costs where the firm decides upon assigning particular (sets of) tasks to individual workers. The adoption of IT might affect both the time needed for each task (production effect) and the costs of communication with fellow workers (communication effect) in production. The impact on the organisational structure and the division of labour might go two ways, however. The productivity effect will decrease the benefits of specialisation because the time devoted to a task falls and one worker is able to carry out more tasks during working time. This leads to smaller team sizes and to educational segregation within firms. The communication effect, on the other hand, will induce more cooperation among workers and hence increase the level of specialisation within the firm. Assigning workers with different skill levels will become more beneficial because for each task the right person, in terms of skill levels, can be hired. Borghans and ter Weel show that both effects resulting from IT adoption increase the demand for higher educated workers and hence induce a skill bias in the demand for labour. Empirically the article's aim is to distinguish between firms that adopt IT because of production or communication reasons. When the latter is relevant they argue that firms adopt IT at once for the entire (or a large share of the) workforce to reap the fruits from improved communication possibilities, whereas if the former is more relevant individual (or team) decisions with regard to adoption are more relevant. Based on a panel of more than 2,000 Dutch establishments their empirical results suggest that the productivity effect dominates the communication effect with respect to the overall pattern of organisational change in the 1990s, associated with the adoption and diffusion of IT within firms. This result suggests that improvements in productivity associated with the ‘stand‐alone’ character of IT have been more important than the increased possibilities of communication. The empirical evidence also suggests that there exists heterogeneity in the way IT improves productivity. Especially for larger firms, firms competing in the high‐quality segment and exporting firms, improved communication possibilities seem to be the dominant factor of IT adoption, while in firms in which the productivity effect dominates a tendency towards generalisation of work is observed. In ‘New Technologies, Workplace Organisation and Age: Firm‐level Evidence’Aubert et al. (2006) investigate the impact of IT and workplace organisation on the age structure of the workforce, using French firm‐level data. There are two views on the effects of new technologies and innovative workplace organisations on the demand for older workers.9 First, the new technologies and forms of firm organisation complement existing skills, so there will be a positive relationship between more experienced and older workers and the adoption of these innovative modes of production. Second, the argument that skills become obsolete over time implies that older workers’ skills are less adaptable to new technologies compared to younger workers, so their labour‐market fortunes worsen when a new technology or organisational innovation arrives. The empirical strategy in Aubert et al. is to estimate wage‐bill shares to capture the effects of innovations on the demand for older workers, but also to account for the inflow and outflow of workers. Particularly the latter is important (and by and large neglected in previous studies) in determining the possibility of an age bias of technological change because the observed composition of a firm's workforce might be the result of labour‐turnover policies that are not necessarily age neutral. The database used consists of a sample of almost 4,000 French manufacturing firms in the late 1990s and results from combination of three separate data sources that include information about the firms’ workplace organisations, the demographics of the workforces and the firms’ financial situation. The estimation results offer at least three novel insights. First, there is a negative and significant correlation between wage‐bill shares by age (measured in four ten‐year categories from 20 to 59) and the use of computer technology, the Internet and the presence of innovative workplace practices. When these results are decomposed by gender it turns out that the negative relationship between the wage‐bill shares by age and IT are stronger for women and that younger men seem to be benefiting relatively more from innovative workplace practices.10 Decomposing the data into three major occupational categories suggests that, besides a skill bias in labour demand, the negative age bias is present across all occupations. The second contribution of the article is the analysis of labour turnover by explicitly investigating the hiring and firing policies of firms. The estimates suggest that more innovative firms hire relatively younger workers. The results with respect to firing decisions are less clear and seem to point towards a need for more experienced workers to ensure coordination of work. Finally, this contribution extends the work of Bresnahan et al. (2002) and the earlier contribution of Caroli and Van Reenen (2001) on the interplay between organisational change, IT adoption and the demand for labour by explicitly focusing on the age composition of the workforce within a large sample of firms. The evidence in favour of skills obsolescence suggests that innovations in terms of technology adoption and workplace organisation do not only yield a skill bias in the demand for labour but also a sizeable age bias. Footnotes 1 " Babbage created an ‘analytical engine’ around 1845. Babbage's experiments with computing did not find continuation until in the 1930s electrical technology could provide an alternative for the exorbitantly expensive and difficult to handle mechanical systems he had to rely on. 2 " The first more or less useful results were obtained around the beginning of World War II, by Konrad Zuse in Germany and Howard H. Aiken in the US. These computers were referred to as electromechanical computing machines (e.g., Mark I (1939), Z3 (1943) and ENIAC (1945)). Advanced theoretical work on developing computer technology was done in the 1930s by the British mathematicians Church and Turing. See e.g., David (1989) for a more elaborate historical overview and comparison to the electric dynamo. 3 " See for example Simon (1986) and David (1990) who draw analogies with other major invention such as the steam engine (Simon) and the electric dynamo (David). 4 " Another intriguing example of forecasting the way in which computer technology was likely to change the world is Leavitt and Whistler's (1958) essay on how firms would deal with computer technology in the 1980s. Their view was that most of the work would be automated and managers would control firms by pushing buttons. 5 " A more detailed overview of recent developments in computer technology can be found in the volume edited by Soete and ter Weel (2005). See Freeman and Soete (1990, Chapter 3) for a detailed description of the developments in computer technology until the late 1980s. 6 " Bontemps et al. (1999) develop a more complicated model building on similar assumptions with the presence of many different production technologies in which unemployment is falling when information frictions are reduced. 7 " Van Reenen (2004) analyses the market for work‐group servers with a focus on antitrust concerns. The main finding of there is that demand is highly inelastic leading a monopolist to raise prices and reduce welfare. 8 " This way of modelling jobs is consistent with and extends the approaches of Autor et al. (2003) and Borghans and ter Weel (2004; 2005). These approaches do not explicitly take the division of labour within firms into account and neglect the gains from specialisation, which is the core focus of the present contribution. Borghans and ter Weel (2004, 2005) develop a number of examples and speculate how the division of labour is likely to change, but do not formally model the division of labour as in the present contribution. 9 " See also Weinberg (2005) for an empirical analysis of the relationship between experience and technology adoption. His findings, using US data, suggest that IT has complemented the skills of high‐school educated workers and substituted for college graduates’ skills. Hence, skills complementarity seems to hold for lower educated workers, whereas skills obsolescence plays a more important role for relatively more educated workers when adopting IT. 10 " This result is consistent with the findings of Weinberg (2000). In a study using the US CPS data he finds that particularly women have benefited from computer technology adoption because computers have de‐emphasised physical strength and stamina. Given the focus on manufacturing firms in Aubert et al. (2006) this seems to be a particularly relevant observation in light of their estimates. References Aubert , P. , Caroli , E. and Roger , M. ( 2006 ). ‘New technologies, organisation and age: firm‐level evidence’ , Economic Journal , this issue. OpenURL Placeholder Text WorldCat Autor , D.H. , Levy , F. and Murnane , R.J. ( 2003 ). ‘The skill content of recent technological change: an empirical exploration’ , Quarterly Journal of Economics , vol. 118 ( 4 ), pp. 1279 – 334 . Google Scholar Crossref Search ADS WorldCat Babbage , C. ( 1832 ). On the Economy of Machinery and Manufactures , London: Frank Cass . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Bontemps , C. , Robin , J.M. and van den Berg , G. J. ( 1999 ). ‘An empirical equilibrium job search model with search on the job and heterogeneous workers and firms’ . International Economic Review , vol. 40 ( 4 ), pp. 1039 – 74 . Google Scholar Crossref Search ADS WorldCat Borghans , L. and Ter Weel , B. ( 2004 ). ‘What happens when agent T gets a computer? The labor market impact of cost efficient computer adoption’ , Journal of Economic Behavior & Organization , vol. 54 ( 2 ), pp. 137 – 51 . Google Scholar Crossref Search ADS WorldCat Borghans , L. and Ter Weel , B. ( 2005 ). ‘How computerization has changed the labour market: a review of the evidence and a new perspective’, in ( L. Soete and B. Ter Weel, eds.) The Economics of the Digital Society , pp. 219 – 47 , Cheltenham: Edward Elgar . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Borghans , L. and Ter Weel , B. ( 2006 ). ‘The division of labour, worker organisation, and technological change’ , Economic Journal , this issue. OpenURL Placeholder Text WorldCat Bresnahan , T.F. , Brynjolfsson , E. and Hitt , L.M. ( 2002 ). ‘Information technology, workplace organization, and the demand for skilled labor: firm level evidence’ , Quarterly Journal of Economics , vol. 117 ( 1 ), pp. 339 – 76 . Google Scholar Crossref Search ADS WorldCat Bresnahan , T.F. and Greenstein , S. ( 1996 ). ‘Technical progress and co‐invention in computing and in the uses of computers’ , Brookings Papers on Economic Activity: Microeconomics , pp. 1 – 77 . OpenURL Placeholder Text WorldCat Bresnahan , T.F. and Trajtenberg , M. ( 1995 ). ‘General purpose technologies: ‘‘engines of growth’’?’ , Journal of Econometrics , vol. 65 ( 1 ), pp. 83 – 108 . Google Scholar Crossref Search ADS WorldCat Caroli , E. and Van Reenen , J. ( 2001 ). ‘Skill‐biased organizational change? Evidence from a panel of British and French establishments’ , Quarterly Journal of Economics , vol. 116 ( 4 ), pp. 1449 – 92 . Google Scholar Crossref Search ADS WorldCat David , P.A. ( 1989 ). ‘Computer and dynamo: the modern productivity paradox in a not‐too‐distant mirror’ , Center for Economic Policy Research Discussion Paper No. 172, Stanford University , July. David , P.A. ( 1990 ). ‘The dynamo and the computer: an historical perspective on the modern productivity paradox’ , American Economic Review , vol. 80 ( 2 ), pp. 355 – 61 . OpenURL Placeholder Text WorldCat Freeman , C. and Soete , L. ( 1990 ). Work for All or Mass Unemployment: Computerised Technical Change into the 21st Century , London: Pinter . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Klahr , D. and Leavitt , H.J. ( 1967 ). ‘Tasks, organizational structures, computers’ , in ( C.A. Myers, ed.) The Impact of Computers on Management , Cambridge MA: MIT Press . OpenURL Placeholder Text WorldCat Leavitt , H.J. and Whisler , T.L. ( 1958 ). ‘Management in the 1980’s’ , Harvard Business Review , vol. 36 ( 6 ), pp. 41 – 8 . OpenURL Placeholder Text WorldCat Pakes , A. ( 2003 ). ‘A reconsideration of hedonic price indices with an application to PC's’ , American Economic Review , vol. 93 ( 5 ), pp. 1578 – 96 . Google Scholar Crossref Search ADS WorldCat Simon , H.A. ( 1986 ). ‘The steam engine and the computer: what makes technology revolutionary?’ , EDUCOM Bulletin , vol. 22 ( 1 ), pp. 2 – 5 . OpenURL Placeholder Text WorldCat Soete , L. and Ter Weel , B. ( 2005 ). The Economics of the Digital Society , Cheltenham: Edward Elgar . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC van den Berg , G.J. ( 2006 ). ‘Revolutionary effects of new information technologies’ , Economic Journal , this issue. OpenURL Placeholder Text WorldCat Van Reenen , J. ( 2004 ). ‘Is there a market for work group servers? Evaluating market level demand elasticities using micro and macro models’ , CEP Discussion Paper No. 650, London School of Economics , September. Van Reenen , J. ( 2006 ). ‘The growth of network computing: quality‐adjusted price changes for network servers’ , Economic Journal , this issue. OpenURL Placeholder Text WorldCat Weinberg , B.A. ( 2000 ). ‘Computer use and the demand for female workers’ , Industrial and Labor Relations Review , vol. 53 ( 2 ), pp. 290 – 308 . Google Scholar Crossref Search ADS WorldCat Weinberg , B.A. ( 2005 ). ‘Experience and technology adoption’ , Working Paper, Ohio State University , February. Author notes " This Feature is a selection of contributions to the 3–4 June 2004 MERIT Workshop on ‘Information Technology and New Industry and Labour Market Dynamics’ in Maastricht. Financial support from the Netherlands Organisation for Scientific Research (NWO), Priority Research Programme ‘Society and the Electronic Highway’, the Maastricht Economic Research Institute on Innovation and Technology (MERIT), and the Maastricht Research School of Economics of Technology and Organizations (METEOR) is gratefully acknowledged. © Royal Economic Society 2006.

Bandyopadhyay,, Siddhartha;Chatterjee,, Kalyan

doi: 10.1111/j.1468-0297.2006.01068.xpmid: N/A

Abstract We survey some of the major advances in coalition theory in recent years. We then look at the recent applications of the theory in the area of political economy and touch upon some important theoretical and applied open questions in the field. The use of extensive‐form models of bilateral bargaining under complete information has become popular in both the theoretical and applied literature since the work of Rubinstein (1982). Muthoo (1999) contains an excellent exposition of this work. A natural extension of this approach would be to understand the nature of coalition formation in multiperson bargaining through similar models. Coalition formation is important in many real‐world contexts, including the formation of cartels or bidding rings, alliances or trading blocs among nation states, research joint ventures, political parties and groupings of political parties, to name but a few. Just as interesting as the coalitions that do form are the ones that do not, for example, on the issue of climate change or, to choose an older instance, the law of the sea. The question has been studied in co‐operative game theory (in fact, originating in the work of von Neumann and Morgenstern, who considered coalition formation the most unportant issue in multiperson games and who developed a solution concept for such games). The concepts of the core and the Shapley value are the two most important ideas to come out of this literature; see, for example, the book by Rosenmüller (1981). Both are concerned with the payoffs obtained by individual players when the coalition of all the players forms. These are therefore unsuitable for explaining the formation of coalitions smaller than the ‘grand coalition’, the grand name for the coalition of all players.1 The co‐operative approach relies on various informal motivating stories to illustrate features that the model needs to take into account; these stories are converted into axioms and the consequences of these axioms are then derived. The approach is mathematically satisfying. The programme of research we shall be concerned with in this article, however, seeks to model the motivating stories as explicitly as possible in a game tree.2 We lose the neat characterisations of the axiomatic approach; what is gained is an intuition for the strategic elements that are important in determining when there is agreement and what kind of agreement there is. We will, of course, use some basic notions from the co‐operative theory, such as the notion of characteristic function, and we will come back to the core and the Shapley value through a more circuitous route. This article will therefore focus on some models that illustrate how coalition formation can be studied using extensive‐form games. We will simplify as much as possible in developing the theory and focus on a specific area of application, in political economy. The reader might feel that the work of the present authors gets more attention here than it should. We apologise in advance to researchers whose work has not been addressed adequately here but our aim is not to provide an exhaustive discussion of work done in this area (this would indeed need a very large book), only to whet the readers’ appetites for more. We now introduce the basic framework and notation we shall use in the article. Let N be the set of all players in the game. (We use n to denote the number of players, with a representative player i = 1,2,…n.) Let Sj, j = 1,2…k, Sj ⊇ N, ⋃j Sj = N, Sj ∩ Sk = ∅ for any j, k, constitute a partition of the set of players to be denoted as π. The Sj represent various coalitions that form3 during the game and no player can belong to more than one coalition.4 The worth of a coalition Sj, in a coalition structure π, is given by v(Sj | π). If xℓ,ℓ∈Sj is the payoff to a player ℓ in this coalition, we could assume that . For the most part, we shall make this assumption, that of transferable utility. There is a considerable literature in co‐operative game theory on games with non‐transferable utility, where it would be more appropriate to think of 𝒮j–tuples (x1,…,x𝒮j) ∈ V(Sj | π), where now V(·|·) is some set of feasible utility vectors attainable by some coalition Sj. However, transferable utility games will suffice for our purposes for most of this paper. In Section 2, we shall assume that v(Sj | π) is, in fact, independent of π, so long as Sj is an element of the partition. This rules out externalities that formation of a particular coalition might have on other coalitions. (We shall discuss externalities briefly, however, in section 3 as well as in discussing applications in section 4.) If the worth of a coalition is independent of the way in which players outside the coalition organise themselves, the function v(·) is called a characteristic function. The more general function is called a partition function. Even the latter is not the most general way to characterise coalitional payoffs. For example, as in Jackson and Wolinsky (1996), we could define a function based on the graph of connections between the players. In the next Section, we consider some examples of extensive‐form, coalition‐formation games using characteristic functions. Section 2 is a very brief one on games with externalities. Section 3 discusses applications in political economy and Section 4 concludes. 1. Extensive‐Form Models of Coalitional Bargaining 1.1. Overview The questions that we address here are as follows: (i) What coalitions form in equilibrium and (ii) how is the worth of a coalition divided among its members? Various extensive forms have been proposed to answer these questions; we shall primarily deal with a model of sequential offers and discuss some alternative approaches.5 We start with the model and investigate what answers the model gives to the two questions above. There is also a literature that takes the reverse approach; starting with one of the myriad co‐operative game solutions and devising an extensive form whose equilibrium (in some sense) coincides with that of the cooperative solution. We find that the core, the Shapley value and the Nash bargaining solution make appearances as equilibria in our models. Other solution concepts, such as the nucleolus, are implemented by various games, see for example, the work of Serrano (1993).6 The extensive‐form models we consider usually use discounting, in the tradition of Rubinstein, and consider the limiting equilibrium allocation as the discount factor δ → 1. An earlier tradition (Harsanyi, 1974; Selten, 1981) considers models without discounting and we shall mention a couple of these as well. As one would expect, models without discounting have a multiplicity of stationary, subgame perfect equilibria and are useful in circumstances where such multiplicity is desired (as in finding a game that implements all the allocations in the core). 1.2. Sequential Offers Models Selten (1981) proposed a sequential offers model of coalitional bargaining in which there is a fixed order of players and the first person in this order makes a proposal, naming a coalition S (of which the proposer is a member) and a payoff division among the players in S. Each named player responds (in sequence) either accepting or rejecting. All the members of S have to accept before the coalition forms. As soon as one coalition forms, the game ends. Since no discounting is used, the characteristic function satisfies a rule called zero normalisation,7 which is not satisfied in the models with discounting that we shall consider but is commonly used in co‐operative games. If someone in S rejects, that person makes the offer in the next period. Selten had to supplement the non‐cooperative approach with axioms to select among stationary equilibria, since he did not use discounting. A sequential offers model with discounting was proposed by Chatterjee et al. (1993). This model, like Selten's, has a fixed order of players, the first one of whom makes an offer naming a coalition and an allocation of the coalitional worth among the members of S. Other members of S accept or reject. If everyone accepts, S leaves the game and the initiative passes to the specified first player in N∖S, without discounting. (Note that this model does not assume that only one coalition forms.) If some member of S rejects, that person makes the next offer in the next period. All players discount the future with the discount factor being δ. (That is, a payoff of x in period t + 1 is equivalent to a payoff of δx in period t.) If a coalition S forms and obtains a coalitional worth v(S) in period t, player i in S obtains a payoff δt−1xi, where Σi∈Sxi = v(S). In other variants of this model, if an offer is rejected, the rejector does not make the next offer but the next player in the pre‐specified order does (as in the Shaked analysis of the unanimity game where all members of N have to agree to a proposal for it to take effect – see Osborne and Rubinstein (1990)) or the next proposer is chosen randomly, as in Okada (1996). It turns out that these variants differ in one important respect from Chatterjee et al., though most of the results are quite similar. The sequential offers extensive form problematic, as it does not take into account competition among different coalitions for some members who are common to both. We discuss this point in the last Section. However, the experiment in Bolton et al. (2003) suggests that the sequential offers model reflects the interplay between competition and equity that one observes in free‐form bargaining. The solution concept used in all these papers is that of stationary, subgame perfect equilibrium. ‘Stationary’ in this context means that offers made and response strategies (that is, whether or not to accept an offer (S, x) currently on the table) depend only on the set of players in the game and not on the history of past offers and counter‐offers.8 The Chatterjee et al. model has two kinds of negative results and one positive characterisation result. The first negative result is that the grand coalition need not form for a given order of proposers, even with a non‐empty core, and therefore that the equilibrium of the game need not be in the core. To see this, consider the following example of a three‐player game taken from Chatterjee et al. (1993). Example 1.2.1. The characteristic function is given by v({1, 2, 3}) = 1, v({1, 2}) = 0.7, v({1, 3}) = v({2, 3}) = 0.2, v(S) = 0 otherwise. As δ → 1, the limiting stationary subgame perfect equilibrium allocation depends on who proposes. If Player 1 or 2 proposes, each will propose the coalition {1, 2} and the other will accept. Player 3 will get 0 and Players 1 and 2 will each get 0.35 (in the limit). If Player 3 proposes, he proposes the grand coalition and the limiting equilibrium allocation is (0.35,0.35,0.3). Thus, if Player 3 proposes, the equilibrium outcome is in the core; otherwise, it is not even efficient. The same point can be made even more forcefully in the following example, where the equilibrium allocation is inefficient for every order of proposers. Again, we only need a three‐player game for this example. Example 1.2.2. Suppose v({1, 2, 3,}) = 1.2, v({1, 2}) = 1, v({1, 3}) = 0.99, v({2, 3}) = 0.4, v(S) = 0 otherwise. Here the limiting equilibrium allocation will be (0.5,0.5,0) if either Player 1 or Player 2 is the first proposer and (0.5,0,0.49) if Player 3 is the first proposer. Note that the core is non‐empty‐for example (0.8,0.2,0.2) is in the core of the game. The problem that arises in both these examples is that the per capita payoff is greater in the two‐player coalition than in the efficient, three‐player one. In the unanimity game, on the other hand, in which v(N) = 1, v(S) = 0 otherwise, one would expect equal division to be the limiting (stationary) equilibrium payoff. (It is.) A strong condition, called domination by the grand coalition in Chatterjee et al. guarantees efficient grand coalition formation for all proposers. Essentially this condition states that the per capita payoff of the grand coalition must be greater than that of any other coalition, so, in some sense, the structure of sub‐coalitions is rendered superfluous by the possibility of proposing the grand coalition. Another interesting example of inefficiency in the sequential offers model arises because of equilibrium delay, even in stationary equilibrium. The following example is due to Elaine Bennett and Eric van Damme and is reproduced in the Chatterjee et al. paper. Here we need at least four players. Example 1.2.3. Suppose v({1, j}) = 50, j = 2, 3, 4, v({2, 3}) = 101, v({2,4}) = 100, v({3, 4}) = 99, v(S) = 0 otherwise and the fixed order is (1,2,3,4). Here, for high values of δ, Player 1 will never make an acceptable offer, since she would want two of the players to leave the market so that she could combine with the remaining player to get 25 (slightly discounted) rather than something close to 0. The possibility of equilibrium delay and unacceptable offers creates severe difficulties with any characterisation results. Chatterjee et al. show that a sufficient condition for no delay is for the game to exhibit a high degree of increasing returns to coalition size, namely that it be strictly convex.9 The main positive result for the sequential offers model is that for strictly convex games, for all sufficiently high values ofδ, there exists an efficient equilibrium for some order of proposers and the limit of the efficient equilibrium allocation, which depends onδ, asδ → 1, is the allocation that maximises the product of utilities of players among all allocations in the core. Thus, for games showing sufficiently strong increasing returns, we get a unique limiting allocation in the core, and moreover the ‘most equal’ point in the core. (We can also think of this as a modified Nash bargaining solution, where the Nash product is maximised over all allocations in the core. Binmore (1985) comes to a similar conclusion in a different three‐player game.) The Nash bargaining solution and the core, derived on very different grounds, make their reappearance here. As mentioned earlier, the paper of Okada (1996) differs in one important respect from that of Chatterjee et al. If the rejector of an offer is not necessarily the next proposer, who is chosen randomly, there is no equilibrium delay. Okada's result is proved for superadditive games but the intuition can be seen in the non‐superadditive case of the Bennett‐van Damme example. If the order of proposers is as given, then Player 4 does not gain by rejecting Player 1's offer, since if he does, the initiative passes to Player 2 in the following period and Player 2 will make an acceptable offer to Player 3, thus leaving Player 4 with only Player 1 as a partner. Player 4 might just as well accept Player 1's offer and save a period's delay. (In strictly superadditive games, there can be delay in equilibrium but not in stationary equilibrium.) As we mentioned at the beginning of this Section, various authors have sought to obtain all the points in the core rather than a unique point. Perry and Reny (1994) and Moldovanu and Winter (1995) have models without discounting. Perry and Reny use continuous time and also allow coalitions to continue in the game even after they have formed. Moldovanu and Winter confine their interest to equilibria that remain equilibria for all orders of proposer (remember they do not have discounting). Evans (1997) gets to the heart of the motivating assumptions behind the core by considering a game where players compete first for the right to make offers. 1.3. Gul's Bilateral Random Matching Model Gul (1989) proposed a model that gives the Shapley value as the limiting ‘efficient, stationary subgame perfect equilibrium’ allocation. However, as Hart and Levy (1999) pointed out, his result needs acceptance of all offers and this is not implied by efficiency. (Hart and Levy construct an eight‐player example to show that efficiency does not imply that all offers are acceptable.) It turns out (according to Hart and Levy, who attribute this to a response to their paper from Gul), that, again, strictly convex characteristic function games do not have this problem. In describing the Gul model, we shall assume that all offers are accepted on the efficient equilibrium path, so that the problem identified by Hart and Levy does not arise. Each player, i = 1, 2, …n owns an asset. In each period a randomly selected, ordered pair {i, j} is selected. Player i makes an offer to Player j to buy out Player j's asset. If this is accepted, player j leaves the game and Player i gets a flow payoff of v({i, j}) in that period (less the amount she paid j for the asset). The number of players in the game is now reduced by 1 though Player i has two assets. Another randomly ordered pair is selected and so on. If an offer is rejected at any period, a player gets the flow payoff that period corresponding to the assets she already owns. There is a common discount factor δ. For example, in Example 1.2.1, there are six possible ordered pairs that could be chosen in the first period, namely {1, 2},{2, 1},{1, 3},{3, 1},{2, 3}.{3, 2}. Suppose {1, 3} is chosen. Then Player 1 can make an offer p3 to buy out Player 3's asset. If this is accepted, Player 1 gets a payoff in period 1 of 0.2 –p3. (If it is rejected, each player gets a payoff of 0 and they move to the next period when one of the six pairs is again chosen.) If Player 2 leaves the game, the following period will see a match between Players 1 and 2, in which one of them will be randomly chosen to make an offer. Suppose it is Player 1 again. She makes an offer p2 to Player 2, who accepts or rejects. If he accepts, Player 1 buys out his asset and enjoys a payoff of 1 in every future period. If he rejects, Player 1 gets a payoff of 0.2 in that period and the game moves to one of the two possible selections next period. Given the caveats at the beginning of this subsection, Gul shows that the equilibrium allocation as δ → 1 is the Shapley value. This is quite an interesting result since the process by growth through random accretion might describe how a given firm might expand by acquisition, with the randomness being provided by the exogenous appearance of market information about the capabilities of the potential acquisition to the acquirer.10 There are other extensive‐form games that give the Shapley value (see Winter (1994) and Dasgupta and Chiu (1998), for example), but the Gul model seems the most natural. 2. Three Models With Externalities We shall briefly discuss three models here, the work of Bloch (1995) and the follow‐up paper of Na (2005), and Ray and Vohra (1999). The first two models, while dealing with coalition formation, are not really about coalitional bargaining, in the same sense as the characteristic function games we discussed in the last Section. Ray and Vohra permit transfers like the models in the previous section.11 Bloch (1995) considers a two‐stage model in which firms independently make Cournot production quantity choices in the second stage. The model analysed is with linear demand and constant marginal cost, to make explicit calculations possible. In the first stage, players propose coalitions. The greater the number of firms in a coalition, the lower the production cost in the second stage for all the firms in the first‐period coalition. (Note that the coalition does not survive in the second stage, no collusion is permitted in the choice of quantities.) Since all firms are ex ante identical, the only choice for a proposer is how many other firms to include in a coalition. Bloch shows that there will be two coalitions in equilibrium, the first approximately three times as large as the second. Na considers the same model with asymmetric firms. Firms differ in the quality of their technology and the first‐stage game has the proposals made in the order given by the technology ranking. Na first investigates the conjecture that ‘birds of a feather flock together’, namely that there will not be coalitions with a technology gap. (If the technology parameters are θ1 > θ2 > θ3, and if firm 1 chooses firm 3, firm 2 will also be part of the coalition.) Na shows through a counter‐example that this conjecture is false. If there are four firms with θ1 > θ2 > θ3 > θ4, Player 1 might choose 3 and exclude 2, because if Players 1 and 2 get together, so will Players 3 and 4; if Player 1 chooses Player 3, firms 2 and 4 will be sufficiently far apart that they will stay apart. Therefore, Player 1's payoff must be compared in the two coalition structures {1, 2}{3, 4} and {1, 3},{2},{4} and could be higher in the second rather than the first. Na then limits the heterogeneity to two types of firm and shows that here the conjecture on no technology gaps holds. Moreover, if the two types of firm are sufficiently different, there is complete segregation, with all the high‐technology firms forming a coalition together and all the low‐technology firms together. While both Bloch and Na assume that the payoff to a player is fixed (by the second stage) once the coalition structure is determined, Ray and Vohra analyse a model in which payoffs are endogenously determined simultaneously with coalition structure. As in the sequential offers model of the previous Section, there is a fixed order of responses and a protocol specifying who the first proposer is if the game is restricted to the set of players S. The proposer makes an offer to a coalition Sj of which she is a member, specifying a payoff to each member of Sj in each partition of N of which Sj is an element. The first rejector of a proposal obtains the right to make a proposal in the following period and there is discounting between periods with a common discount factor δ. The major general result in Ray and Vohra's paper demonstrates existence of a stationary equilibrium in the general partition function game. There is no such general existence result even for characteristic functions in the Chatterjee et al. paper, though explicit characterisations are given in the case where there is no equilibrium delay (for strictly convex games), so the problem solved by Ray and Vohra is a difficult one. Much of the difficulty arises from the possibility that it might be optimal to make unacceptable offers (hence equilibrium delay) and the existence result involves some possible mixing among offers. The proof uses inductive arguments on subgames consisting of subsets of players. Ray and Vohra also provide a fairly complete characterisation of the symmetric case, using a recursive approach based on a proposer choosing to maximise per capita coalition ‘worth’ conditional on the coalition structure predicted by the equilibrium strategies of players who would remain in the game after the chosen coalition has agreed to the proposal and departed from the game. In particular, they discuss the nature of no‐delay equilibria in these games. The Ray‐Vohra paper shows, as the discussion in the previous section presages, that the occurrence of delay on the equilibrium path even in stationary equilibrium is a fundamental question in coalitional bargaining.12 In bilateral bargaining, the Rubinstein result under complete information has led to delay being regarded either as due to incomplete information or to anomalous preferences or mistaken beliefs. Once we go to four or more players, the possibility of delay is no longer an ‘out‐of‐equilibrium’ phenomenon. This is of interest in applications as delay characterises many ‘real life’ bargaining situations. These include delays in forming a government (these vary across countries, in Denmark and Sweden, for example, coalition negotiations usually conclude in about a week. It takes, on average, more than a month in Austria and Belgium, while in the Netherlands it is even longer; almost three months pass before a new coalition takes office).13 These results show that delays may not happen only because of asymmetric information as is commonly assumed. In particular, externalities are crucial in political coalitions since the payoffs of players outside a coalition typically depend on the identity of players in the coalition. 3. Applications of Coalition Theory 3.1. Introduction Coalition theory finds natural applications in political economy, in particular several papers study some form of coalition formation in legislatures. The simplest such models use majority game characteristic functions, where v(S) = 1 if 𝒮 > n/2, and 0 otherwise. Sometimes if 𝒮 = n/2, some special assumption needs to be made. Here v(S) is interpreted as the perks of office that have to be shared among members of the coalition. Some papers on coalition formation (Baron and Ferejohn, 1989 and Eraslan, 2002) do use such formulations. However, bargaining over policy plays an important role in political decision making. This implies that externalities are involved as the policy chosen by the winning coalition affects non‐members. Hence, members not in the coalition would have preferences over different winning coalitions of which they are not members. Thus, the partition function could be used to characterise payoffs. The game‐theoretic literature we have surveyed provides useful insights as many coalition formation processes can be modelled as multi‐party (and potentially multi‐issue) bargaining games. Further, some form of a sequential offer (made in a fixed or random order) bargaining model is employed in many of the applied papers. Hence, the models we have analysed in the previous two Sections provide valuable insights, which the applied work has used. However, the papers to be discussed here typically analyse more specific contexts than the general coalitional bargaining models. The models are therefore tailored to the problems being studied. In terms of the timing of the coalitions that form, coalition formation takes place between groups both before and after elections; however, the most well‐developed models in the literature are those that look at coalitions after the election. There are several papers in this area but we shall focus on only a few and then briefly discuss some of the open questions in this area.14 In particular, we shall discuss some of the issues in the modelling of pre‐electoral coalitions, which have only recently received attention. It is important to remember that most coalition formation in the political arena assumes that only a majority needs to agree for a proposal to be implemented (unlike the unanimity agreements that were required for some of the bargaining games that we reviewed in the earlier sections). 3.2. Coalitions When Bargaining Over A Fixed Pie Predicting what kind of coalition forms in equilibrium can be traced back to Riker (1962). He predicted that coalitions will be minimal winning. This is certainly a strong prediction but as Riker considered the problem as one of dividing a fixed pie, it is perhaps not so surprising that the smallest number of agents needed to win the right to divide this would form a coalition. In the context of political coalitions, this seems to suffer from two major limitations; first, as long as ideology plays a role in the formation of political coalitions, assuming a constant pie is not always a good approximation. Second, as long as one is allowed to make counter offers, these pure cake division problems need an extensive‐form structure for a solution. For example Baron and Ferejohn (1989) assume a ‘random recognition rule’, like Okada (1996). Someone is recognised to be the proposer probabilistically and makes a proposal of a division of the pie, which needs majority agreement. If the majority do not agree then the game moves to the next round. This (potentially) infinite horizon game continues till majority agreement is reached. Assuming stationary strategies, they show that the subgame perfect equilibrium is unique when the recognition probabilities are the same. Eraslan (2002) extends this to show that even when the recognition probabilities differ all stationary subgame perfect equilibria yield the same payoff. However, these papers do not address the first problem, i.e. they do not deal with bargaining over policy. All of these are variants of the basic coalitional bargaining models discussed earlier. 3.3. Bargaining Over A Non‐Constant Sum Pie: The Austen‐Smith‐Banks Model One of the seminal papers to address the issue of bargaining over policy and perquisites is by Austen‐Smith and Banks (1988, henceforth ASB), who look at coalition formaion in a model of proportional representation. They have three parties who care for perquisites and (instrumentally) for policy.15 Party entry is not endogenous but the parties strategically choose positions on a one dimensional policy space S ? R with |S| < ∞. However, once positions are chosen, any policy that deviates from the ideal point of a party in a post‐election coalition government gives them disutility.16 The voters in the model are strategic; they vote taking into account the coalition formation game. The extensive form game they consider (for the coalition formation) is what is called ‘selection in order’ where the party with the largest seat share first proposes a coalition (which is a policy and a division of perks); if the majority approve the game ends. Otherwise the second largest party proposes and so on. If no party manages to form a government then a pre‐determined status quo policy is implemented, which in their game is one where the policy is an average of all three parties and the perks are equally divided. Bargaining over perks is constrained by (i) a fixed amount of total perks G in their model and (ii) a non‐negativity assumption gk ≥ 0 for all k. This game has a multiplicity of equilibria. The equilibrium ASB choose is one where all voters are pivotal. This gives a unique prediction where the largest and smallest party form a coalition and the second largest party is left out. (This is ‘minimal winning’ unless a single party has a majority, in their ‘voting equilibrium’ no party gets a majority of seats.) In equilibrium the two parties in the coalition get all the perks. Given this, and the anticipated voting behaviour, two of the parties choose positions symmetrically around the median, gathering equal numbers of votes and the third party chooses the median position and gets the lowest number of votes. Hence, the final policy of the governments is always away from the median but the expected policy is the median policy. Further, it is never an equilibrium for all voters to vote sincerely. However, as the authors admit, the equilibrium selction is somewhat ad hoc. It is interesting to note that until recently most models of coalitional bargaining in a legislative environment, which followed the paper by ASB, predicted minimal winning coalitions though not necessarily minimal size (a minimum size coalition is the smallest minimal winning coalition).17 In terms of explanatory power, these were clearly unsatisfactory since data on coalition formation showed that minimal winning coalitions accounted for only a third of all coalitions formed after the second world war in European democracies.18 3.4. Explaining Coalition Diversity: The Efficient Bargaining Approach The recent game‐theoretic models of legislative bargaining have made headway in this direction. Two papers by Diermeier and Merlo (2000) and Baron and Diermeier (2001), (collectively referred to as BDM) employ the efficient bargaining approach to coalition politics to provide an explanation for the size diversity of coalitions. Under this approach, the party in charge of putting together a coalition (called the formateur) can buy the support of other parties by adopting a compromise policy position or by making side payments in return for support. The papers characterise equilibrium coalitions when parties have quasi‐linear preferences over policy and perks.19 Note that they allow for the transfer to be negative as well and also require budget balance i.e. the transfers add up to 0. Given the quasi‐linear function, the ‘efficient’ policy (in this case the unweighted sum of the parties ideal points) can always be reached by compensating parties for making policy transfers.20 The use of efficient bargaining where parties can make transfers (and hence also receive negative transfers) is one important point of difference with the previous literature. Their choice of who proposes to form a government is also different from ASB. They use a ‘random recognition’ protocol, i.e. each party is recognised probabilistically (à la Baron and Ferejohn (1989)) to propose a government. If the government is majority approved (and all members in the government agree to be in the government) then the policy implemented is the one that is efficient from the coalition's viewpoint and transfers are made to compensate others for supporting the government. Otherwise, an exogenously given status quo policy is implemented. Using this approach, they are able to generate equilibrium governments which can be minority, minimum winning, or surplus. The BDM papers are thus able to generate results consistent with the observed diversity of coalitions, using a standard bargaining framework. Moreover, the use of efficient bargaining implies that the policy chosen is independent of whether offers are one shot or counter offers are allowed; only the distribution of benefits changes if the bargaining is multi‐period rather than one‐period. However, these papers are not suited to looking at some interesting comparative static issues, in particular the analysis of how the importance of ideology relative to rents from office affect the formation of coalitions. Nor do these papers shed light on the issue of disconnected coalitions.21 (Brams et al. (2002) is one of the few papers that addresses this issue at a theoretical level.) 3.5. Comparative Statics Some of these comparative static questions have been addressed in empirical papers, notably Indridason (2003, 2005) and Kalandrakis (2002) and at a theoretical level by Bandyopadhyay and Oak (2004 and 2005, referred to collectively as BO). In some sense these papers seem to complement each other, as many of the empirical results in Indridason and Kalandrakis are suggested by the theoretical predictions in Bandyopadhyay and Oak. We briefly discuss BO and compare it with the BDM papers before commenting on Indridason's findings. The main contribution of Bandyopadhyay and Oak (2004) is to provide a comparative static analysis of coalitions as functions of ideology and political rent (called power in their model) as well as in providing a counter example to the Duvergerian prediction of more parties under proportional representation than under plurality voting. They construct a game‐theoretic model of proportional representation in which groups of citizens decide to form parties, voting occurs and once seat shares are realised parliament is constituted and coalition formation occurs. The coalition formation procedure is similar to the BDM papers with the important difference that BO do not allow for commitment to policy or transfers at the coalition formation stage, they do not provide microfoundations for the policy making but cite empirical evidence and assume that the policy is a seat weighted average of the coalition members’ ideal points. Like ASB, the total perks cannot exceed a fixed amount P, with a non‐negativity constraint. The lack of commitment implies that this can be shared only among coalition members. This framework lends itself to doing interesting comparative statics. Like BDM, they generate coalitions of a varied nature consistent with empirical evidence and also provide some general insights: for example they predict that centrist parties are more likely to form minority governments and surplus governments form if a formateur wants to balance the policy by taking people of opposing ideologies. Interestingly, they show that a non‐monotonic relation exists between connectedness of coalitions and political rents. This is both consistent with the empirical findings in Indridason and goes against the ‘folk wisdom’ that disconnected coalitions are more likely when rents are important relative to ideology. They also show that in the limit when P becomes very high, coalitions are never surplus (or supermajority) though not necessarily minimal size. This is in contrast to the BDM papers where increasing the perks from office do not change the coalitions that are formed. Both the BDM papers and BO papers in some sense provide extreme views about commitment and transfers, it would be interesting to see what kind of comparative statics emerge when allowing for partial commitment. The choice of protocol also plays a role in the results and while BDM provide some empirical justification for choosing random recognition over selection in order (as in ASB) they do not address the question of what factors (like centrality of party, ability to gather majority support at least cost etc.) really drive the formateur selection procedure. BO do not allow for any coalition formation in the party formation stage though they mention this among possible extensions. Further, using the coalitional set up in their models and in BDM to look at ex ante coalitions remains an important area of future research. 3.6. Pre‐Electoral Coalitions There are, of course, papers on party formation which allow groups to form coalitions but the post‐election game is simplified to a majority voting game thereby losing much of the action that arises from the well‐developed coalitional models of parliamentary democracy we have discussed earlier. However, these papers do provide interesting insight into what are the motives for party formation and extend the citizen candidate models of Osborne and Slivinkski (1996) and Besley and Coate (1997) to allow for groups of individuals to form parties. The two main motives that are considered are cost sharing and co‐ordination. Some of the main papers in this area are Morelli (2004), Levy (2004), Osborne and Tourky (2002) and Riviere (2000).22 We do not provide a detailed description of these papers since the interested reader will find the main insights discussed in a recent survey by Dhillon (2005). We think that integrating these two areas of research would provide insights into how the nature of electoral systems affects the timing of coalition formation. Formally, this could be modelled as a two‐stage coalition formation process, where, in stage 1, parties have an option of forming an ex ante coalition or going it alone. In either case, vote shares are realised and the game from then on can be modelled along the lines of the existing papers in the literature. Clearly, we need to make assumptions about commitment and the precise role that ex ante coalitions serve. For instance, we have to decide whether after seat shares are realised, we allow the parties to form coalitions that include new partners and (perhaps) exclude some of the partners in the pre‐electoral alliance or whether an ex ante coalition must stick together if it wins. These two would correspond to different outcomes in the post‐election stage and anticipating that would lead to different incentives to form ex ante coalitions. One question to study is whether seat sharing arrangements (a big part of the pre‐election bargaining between parties) serve as an effective way to commit to (ex post) sharing of power and compromising over policy by parties entering into an arrangement. It is also interesting to see how far voters can counteract the strategic choices of parties. To illustrate, suppose there are three parties L, M and R. If L and M form an ex ante alliance (corresponding to a seat sharing agreement), M voters in a district may, for instance vote against an L candidate if she is the alliance's candidate in a district to weaken L's position in the ex post coalition that forms (as long as by doing so they do not cause the alliance as a whole to lose). This may perhaps explain why, in plurality voting, we see smaller parties often getting merged into a big party rather than retaining their identity and forming seat‐sharing arrangements. With mergers, voters can no longer distinguish between the different ideologies within a party so easily and thus cannot resort to this strategic manipulation. Incidentally, this may also be the reason why in the US where party control is weaker, Democratic senators who are more rightist get elected in moderate Republican constituencies and vice versa. In general, so far the theoretical (and generally the empirical) literature has been silent on this issue of the timing of coalition formation, though recently there have been a couple of empirical papers by Golder (2003 a, b) on this. Clearly, several modelling issues need to be thought through. However, such models would allow us to study questions like whether a plurality, first‐past‐the‐post system encourages ex ante coalitions (or mergers), while proportional representation encourages ex post coalitions and contribute to our understanding of how electoral rules affect the nature of political competition. 3.7. Stability of Coalitions – A Model and An Example An important issue is how constitutional design affects the formation and stability of coalitions. There have been important papers in this area and this has been surveyed in Diermeier et al. (2003) and the interested reader will find a detailed account of their papers in this area. Essentially, their approach is based on structural estimation of a multilateral bargaining model. They use that to see how constitutional features have an impact on the formation and duration of governments. An interesting issue that they point out is that bicameral parliamentary systems are not inherently more unstable since formateurs compensate for the fact that they have to satisfy both houses by forming larger coalitions. Hence, on average, such legislatures tend to see surplus coalitions more often. They also analyse how several institutional factors such as fixed inter‐election period vs. no fixed inter election periods affect coalition stability. We conclude this Section by focusing on an alternative model (this is taken from Chatterjee and Dutta (2000) unpublished notes) of coalition formation and durability. As this work is not available elsewhere we present it in some detail. The aim is to construct and analyse a model of government formation and operation that explains why coalition governments fall, and to relate the theory to certain stylised facts. These are: The greater the number of parties in a minimal winning coalition, the more unstable the government. The more heterogeneous the composition of the coalition, the less stable the government. The governing coalition might not be a minimal winning coalition, but a supermajority government, despite the formateur having to give up more of the benefits of office to more parties. (Examples are the government formed by Ehud Barak in Israel and the NDA government in India, neither of which is now in power.) Like Laver and Shepsle (1996), the model considers policy disagreements to be at the core of government instability. Similarly, the authors agree with them that there must be something not completely anticipated in advance for instability to occur; no government forms knowing in advance that it is going to fall in two weeks (though the first Atal Bihari Vajpayee government in India in 1996, which lasted 13 days, probably had a very good idea that its days were numbered). However, the Chatterjee‐Dutta model differs from that of Laver and Shepsle in the nature of the uncertain event that causes the government to fall. We now outline their approach and conclude with an example taken from their notes. The model has two periods, followed by an election. The initial conditions specify a random draw from the election lottery, specifying αi, the proportion of seats won by the ith party. Each period, the members of the cabinet collectively enjoy a surplus of one unit. (We think of these as not just perquisites of office and patronage possibilities, but also agreed policies being implemented on issues that are important at the time of government formation.) ‘Tomorrow’, i.e. the second period, one of several issues j could arise and need a decision. The possible decisions on each issue are dj = 0, which we label the ‘left‐wing’ position and dj = 1, which maps into the right‐wing of the ideology space. At the time of government formation, players are uncertain about how important a particular issue will be to the electorate if it comes up in the following period. (Examples could be gun control, which becomes important after some unforeseeable massacre, or immigration, whose importance might depend on the economic situation and the political need to affix blame on easily available groups.) The importance of the issue is given by Kj, which will be explained shortly. Party i does not know in advance what its position will be on the particular nuances of the issue that arises ‘tomorrow’. However, given its general ideological position, there is a probability pi that it will advocate the right‐wing position on any issue. (Thus positions on issues are correlated, another difference of this analysis from that of Laver and Shepsle.) Such a formulation is a reduced form of the following. Suppose party i holds an ideological position xi ∈ [0, 1]. Its ideological payoff for issue j if dj = 0 is qiLj(1 − xi) and if dj = 1, qiRjxi. Suppose qiRj is known but for each issue Party i receives a signal (independently of other signals) about the value of qiLj, which is ex ante random. Depending on this realised value, Party i has a position in favour of dj = 0 or 1. A party that has a low value x, i.e. is left wing, will be more likely to adopt left‐wing positions. If x = 0, the party concerned will always adopt left‐wing positions on every issue. (On the issue of NAFTA for instance, the generally pro‐union Democratic President adopted a position at odds with the unions, while the congressional Democrats often did not. Of course, this is not an example of two formally different parties, but the party structure in the US is looser than it is in parliamentary democracies.) If a party advocates dj = 0, but is part of a government that decides to implement dj = 1, it suffers a loss of Kj per period from its electoral payoff, which is (per period) given as w. There is a common discount factor of δ; for politicians, this is unlikely to be close to 1! Thus, essentially the government falls if an issue is important enough for a sufficiently large number of parties to prefer leaving the government to having the opposite position implemented on a particular issue. The coalition formation game they consider is ‘selection in order’à la Austen‐Smith and Banks. They assume that the party with the highest αi is asked first to form the government. It makes a proposal (allocating portfolios and proposing positions on issues that are known) to a coalition S ⊇ N, where N is the set of all parties (the grand coalition in the game‐theoretic sense, rather than in the sense it was once used in Germany to describe a coalition between the two largest parties). The coalition S must include the proposer and must have a legislative majority. If everyone in S agrees, the coalition forms. If there is a disagreement, the initiative passes to the next largest party and so on. The bargaining is supposed not to take time, compared to the length of a period. If no party is able to form a government, the legislature is dissolved and elections for a new parliament are held, with party i receiving its per period outside option wi. If the government forms, in the next period, issue j arises, with importance Kj, where Kj has a commonly known probability distribution. The parties in government take positions on the issue according to the pi. The Prime Minister proposes a value for dj; members of the government accept or reject in sequence. (Rejection implies withdrawal from the government.) If a majority withdraws from the government, the government falls and the formation bargaining process starts afresh. If a minority withdraws, its shares are reallocated to the now smaller majority, which continues in power. All players are risk‐neutral. The payoffs are the expected payoffs taking into account the probability that the government will fall as a function of the nature and size of the ruling coalition. The model has all the ingredients necessary to generate the three stylised facts from a theoretical analysis. With otherwise similar parties, a formateur will go about selecting ‘like‐minded’ parties (i.e. parties with values of pi close to the formateur's) first before approaching those that are ideologically distant. The reason a supermajority government might be formed has to do with the tradeoff between increased stability on the one hand, and a smaller share of the benefits of office on the other. The uncertainty about the importance of issues that are likely to arise, embodied in Kj, enables us to calculate the probability that any given coalition will be unstable on the basis of the model. We illustrate these features with the following example. Example: We consider a scenario in which there are 5 parties characterised by the p‐vector (0,0.4,0.4,0.5,1), and the w‐vector (0.05,0.05,0.05,0.05,0.06). We assume that αi = 0.2 for all parties, and that the order in which the parties are asked to form the government is 1,2,3,4,5. Suppose that party i is a member of the ruling coalition, and that issue j arises, and that the Prime Minister proposes dj = 0. Then, if the realised position of party i on issue j is dj = 1, and if Kj is high enough, party i will prefer to withdraw from the government irrespective of the share of the surplus offered to i.23 Assuming that each Kj is high enough, one can calculate the probability of survival of each coalition, contingent on who is the Prime Minister (more accurately, on the party to which the Prime Minister belongs). We give below some of these survival probabilities. Probability of Survival (Party listed first contains Prime Minister) Coalition . Probability . {1,3,4} 0.20 {1,2,3,4} 0.60 {2,1,4,5} 0.50 {2,1,3,4,5} 0.76 {3,1,2,5} 0.50 {3,1,2,4,5} 0.76 {4,1,2,3} 0.50 {4,1,2,3,5} 0.74 {5,2,3,4} 0.40 Coalition . Probability . {1,3,4} 0.20 {1,2,3,4} 0.60 {2,1,4,5} 0.50 {2,1,3,4,5} 0.76 {3,1,2,5} 0.50 {3,1,2,4,5} 0.76 {4,1,2,3} 0.50 {4,1,2,3,5} 0.74 {5,2,3,4} 0.40 Coalition . Probability . {1,3,4} 0.20 {1,2,3,4} 0.60 {2,1,4,5} 0.50 {2,1,3,4,5} 0.76 {3,1,2,5} 0.50 {3,1,2,4,5} 0.76 {4,1,2,3} 0.50 {4,1,2,3,5} 0.74 {5,2,3,4} 0.40 Coalition . Probability . {1,3,4} 0.20 {1,2,3,4} 0.60 {2,1,4,5} 0.50 {2,1,3,4,5} 0.76 {3,1,2,5} 0.50 {3,1,2,4,5} 0.76 {4,1,2,3} 0.50 {4,1,2,3,5} 0.74 {5,2,3,4} 0.40 In order to find out the equilibrium proposal of party i, we have to find derive the equilibrium proposals of party (i + 1) for i = 1, 2, 3, 4. So, we start with the equilibrium proposal of 5. Note that party 5 will not select party 1 as a coalition partner since they will never agree on any issue. Party 5 will offer to form a government with 2,3,4. The survival probability of this coalition is 0.4. Party 5 will offer 0.05 to each of its coalition partners since wi is the outside option of i. So, the equilibrium payoff vector associated with 5's offer is (0,0.05,0.05,0.05,0.25). Then, 4 will propose to (1,2,3), the equilibrium shares being (0,0.05,0.05,0.40,0).24 Similarly, 3 proposes to {2,1,5}, the equilibrium being (0,0.05,0.47,0,0). Continuing in this fashion, 1 will propose to (3,4), the equilibrium being (0.20,0,0,0,0). So, in this example, only a majority government forms. The reason for this is that 5 never adds anything to the survival probability of a coalition in which the Prime Minister is from party 1, while party 2 is too ‘costly’. However, one can obtain a supermajority government by modifying the example so that p5 = 1 − ε. Then, 1 would include 5 in the coalition because 5 would have a positive marginal effect on the survival probability of the coalition {1,3,4}. Since the continuation equilibrium payoff of 5 continues to be 0 (so long as ε is ‘small’), 1 gains by including 5 in the coalition. Also, suppose that . Let S = {1, 3, 4, 5}. Then, S is more heterogeneous under than under p5. Also, the survival probability of S is higher under p5, illustrating that heterogeneity reduces stability. The ‘outside options’ in the example above are not derived endogenously. (This is a possible, non‐trivial, extension.) However, we think this provides an interesting framework to analyse stability of coalitions. 4. Conclusions and Open Questions Coalition theory has received considerable attention among social scientists, particularly economists and political scientists. In this survey, we have focussed on the game theoretic literature that has developed in this area and focussed on applications in the political arena. At the theoretical level, coalition formation when agents bargain over more than one dimension remains a big challenge, particularly in the presence of externalities.25 A much under‐researched area is gradualism in coalition formation though, see Seidmann and Winter (1998). Another challenging area is coalition formation and bargaining under incomplete information. Even in the well‐studied characteristic function games, there are several open issues. Some of these, of relevance to applications in bargaining in legislatures, relate to the interest in ‘non‐contingent proposals’, where a proposal to a given player or group of players is not contingent on acceptance by all the players who have been named in an offer. Huang (2002) and Huang and Sjostrom(2003) have specific models along these lines, but a generalisation to a sequential offers coalitional bargaining game would be useful. Another issue arises with respect to the question of competition for bargaining partners. Evans (1997) has an approach to this question by allowing the right to make offers to be subject to competition.26 It would be interesting to devise ways of incorporating this in natural games without explicitly introducing the right to make offers. Some generalisation of a model like Perry and Reny's (1993) bilateral bargaining without procedures might be a fruitful way to study this question. A third issue in this setting is the analysis of the cost (in delay, perhaps) of forming large coalitions. Just as in cartels, co‐operation in large coalitions might be less stable than in small ones. This would lead to a shift away from the emphasis on the theoretical literature to what happens in the grand coalition. In terms of applications, much work remains to be done as well, in particular determining how electoral rules influence the nature and timing of coalitions remains a big challenge. Understanding that would enrich our understanding of the effects of electoral systems on the ideological composition, size and political platforms of parties. Footnotes 1 " Maskin (2003) argues that this is one of the main reasons why co‐operative game theory has been less influential as insisting that the grand coalition always forms is restrictive realistically as well as from the theoretical standpoint. 2 " This is, of course, in common with all of non‐cooperative game theory. Note that this is different from trying to find a game to implement specific co‐operative solution concepts. 3 " As is usual sj is used to denote the cardinality of the coalition Sj. 4 " In the recent state elections in Jharkhand, in India, the total number of legislators claimed by the various coalitions contending for power was greater than the total number of seats in the house. This is disallowed in our framework. 5 " There has been much recent work on extensive‐form models; somewhat different from the model to be discussed here is the work of Moldovanu and Winter (1995), Morelli and Currarini (2000), Ray and Vohra (1997) and Perry and Reny (1994). Gomes (1999) considers a three‐player game with externalities in which players make forward commitments not to leave even when a coalition smaller than the grand coalition forms‐like Perry and Reny‐and obtains a version of the Nash bargaining solution only for unanimity games and for games where subcoalitions have smaller per capita value than the grand coalition. 6 " We have assumed prior knowledge of the concepts of the core and the Shapley value, the book by Rosenmüller (1981) can be referred to for more detail on these. 7 " This condition basically says that if we subtract v({i}) from the worth of each coalition of which i is a member, the resulting characteristic function is strategically equivalent. This is not true in the Rubinstein game with outside options, for example. A game with a pie of 1 and two players with outside options of 0.6 and 0 is not strategically equivalent to one where a surplus of 0.4 is split among two players. (In the first, the Rubinstein limiting solution gives (0.6,0.4); in the second (0.8,0.2). 8 " This is not always a natural assumption and has been criticised, see Osborne and Rubinstein (1990). A formalisation and justification of stationarity as economising on complexity costs was provided for the unanimity game by Chatterjee and Sabourian (2000). 9 " This means that if S ? T, then v({S ∪ i}) − v(s) < v({T ∪ i}) − v(T), for all i, S, T. 10 " As far as we know, such a model of the firm has not been formulated, but it seems reasonable to try to do so. 11 " Maskin (2003) proposes a co‐operative solution concept for coalition formation with externalities, which reduces to the Shapley value for games without externalities. He also proposes a non‐cooperative implementation in which coalitions bid for additional members. Someone who is bid for can join a coalition bidding for him or start a new coalition and bid for other players. 
Our survey only briefly deals with general models with externalities, though we consider specific political economy applications with externalities; for a different emphasis we recommend Debraj Ray's Lipsey lectures on group formation, see Ray (2004). 12 " Delays in games with negative externalities (though not in the context of coalitions) have been analysed in Jehiel and Moldovanu (1995) for example. They show that negative externalities can lead to long delays. 13 " As reported in Müller and Strom (2000). 14 " The reader who is interested in more details is referred to Austen‐Smith and Banks (2005) and Merlo (2005). Lack of space prohibits us from going into many of the recent results on coalitions relevant in the sphere of political economy. 15 " Like nearly all the papers in this vein, preferences here are additively separable in policy and perquisites, or ‘quasi‐linear’ even though some of the results do not depend on that. 16 " This is justified by saying voters punish them next period for deviating from their ideal point. Such punishment occurs only if a party is in the government. 17 " See also Baron (1989, 1991, 1998), Lupia and Strom (1995) and Roemer (2002) for instance. 18 " See for example Laver and Schofield (1990). 19 " Thus the utility function of a party takes the form Ui = ui(x) + yi where x is the policy implemented and ui(x) is the utility to party i from policy x and yi is the transfer (or perks) received by party i. 20 " They point out that the efficient policy being implemented is irrespective of the details of the bargaining process. 21 " A left–right coalition leaving a centrist out is an example of a disconnected coalition. 22 " In contrast to these models Snyder and Ting (2002) offer an informational rationale for parties. 23 " Note that this also means that the Prime Minister will never propose a position on an issue which is at variance with his or her party's position. 24 " 4 effectively leaves out 5 since the survival probability when 5 joins the coalition increases from 0.50 to 0.74, but 5 will agree only if it is offered 0.25, its share in the continuation equilibrium. 25 " See In and Serrano (2003, 2004) for some progress over multi‐issue bargaining. 26 " See also Chatterjee and Dutta (1998). References Austen‐Smith , D. and Banks , J. ( 1988 ). ‘Elections, coalitions and legislative outcomes’ , The American Political Science Review , vol. 82 ( 2 ), June, pp. 405 – 22 . Google Scholar Crossref Search ADS WorldCat Austen‐Smith , D. and Banks , J. ( 2005 ). Positive Political Theory II: Strategy and Structure , Ann Arbor: The University of Michigan Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Bandyopadhyay , S. and Oak , M. ( 2004 ). ‘Party formation and coalitional bargaining in a model of proportional representation’ , Discussion Paper, University of Birmingham . Bandyopadhyay , S. and Oak , M. ( 2005 ). ‘Coalition governments in a model of parliamentary democracy’ , Discussion Paper, University of Birmingham . Baron , D. and Diermeier , D. ( 2001 ). ‘Elections, governments, and parliaments under proportional representation’ , Quarterly Journal of Economics , vol. 116 ( 3 ), pp. 933 – 67 . Google Scholar Crossref Search ADS WorldCat Baron D . and Ferejohn D. ( 1989 ). ‘Bargaining in legislatures’ , American Political Science Review , vol. 83 , pp. 1181 – 206 . Google Scholar Crossref Search ADS WorldCat Besley , T. and Coate , S. ( 1997 ). ‘An economic model of representative democracy’ , Quarterly Journal of Economics , vol. 112 ( 1 ), pp. 85 – 106 . Google Scholar Crossref Search ADS WorldCat Binmore , K. G. ( 1985 ). ‘Bargaining and coalitions’, in ( A. E. Roth, ed.): Game theoretic models of bargaining , Cambridge University Press , New York. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Bloch , F. ( 1995 ). ‘Endogenous structures of associations in oligopolies , Rand Journal of Economics , vol. 26 , pp. 537 – 56 . Google Scholar Crossref Search ADS WorldCat Bolton , G. , Chatterjee , K. and McGinn , K. ( 2003 ). ‘How communication links influence coalition bargaining: a laboratory investigation’ , Management Science , vol. 49 ( 5 ), pp. 583 – 98 . Google Scholar Crossref Search ADS WorldCat Brams S. J. , Jones , M. A. and Kilgour , D. M. ( 2002 ). ‘Single peakedness, coalition formation and disconnected coalitions’ . Journal of Theoretical Politics , vol. 14 ( 3 ) (July), pp. 359 – 83 . Google Scholar Crossref Search ADS WorldCat Chatterjee , K. , Dutta , B., Ray , D. and Sengupta , K. ( 1993 ). ‘A non‐cooperative theory of coalitional bargaining’ , Review of Economic Studies , vol. 60 , pp. 463 – 47 . Google Scholar Crossref Search ADS WorldCat Chatterjee , K. and Dutta , B. ( 1998 ). ‘Rubinstein auctions: on competition for bargaining partners’ , Games and Economic Behavior , vol. 23 , May, 119 – 45 . Google Scholar Crossref Search ADS WorldCat Chatterjee , K. and Dutta , B. ( 2000 ). ‘On stability of coalitions’ (unpublished notes) Chatterjee , K. and Sabourian , H. ( 2000 ). ‘Multiperson bargaining and strategic complexity’ . Econometrica , vol. 68 ( 6 ), pp. 1491 – 509 . Google Scholar Crossref Search ADS WorldCat Dasgupta , A. and Chiu , S. ( 1998 ). ‘On implementation via demand commitment games’ , International Journal of Game Theory , vol. 27 ( 2 ), pp. 161 – 89 . Google Scholar Crossref Search ADS WorldCat Dhillon , A. ( 2005 ). ‘Political parties and coalition formation’ , mimeo (in G. Demange and M. Wooders, eds.), Group formation in economics: networks, clubs and coalitions, Cambridge: Cambridge University Press . Diermeier , D. and Merlo , A. ( 2000 ). ‘Government turnover in parliamentary democracies’ , Journal of Economic Theory , vol. 94 , pp. 46 – 79 . Google Scholar Crossref Search ADS WorldCat Diermeier , D. , Eraslan , H. and Merlo , A. ( 2003 ). ‘The effects of constitutions on coalition governments in parliamentary democracies’ , Working Paper. Eraslan , H. ( 2002 ). ‘Uniqueness of stationary equilibrium payoffs in the Baron‐Ferejohn model’ , Journal of Economic Theory , vol. 103 ( 1 ), pp. 11 – 30 . Google Scholar Crossref Search ADS WorldCat Evans , R. A. ( 1997 ). ‘Coalitional bargaining with competition to make offers’ , Games and Economic Behavior , vol. 19 , pp. 211 – 20 . Google Scholar Crossref Search ADS WorldCat Golder , S. N. ( 2003a ). ‘Electoral institutions and pre‐electoral coalitions in comparative perspective’ , Ph.D thesis , New York University. Golder , S. N. ( 2003b ). ‘A theoretical model of pre‐electoral coalition formation’ , Ph.D thesis , New York University. Gomes , A. ( 1999 ). ‘A theory of negotiations and formation of coalitions’ , CARESS Working Paper 99–12, University of Pennsylvania, Philadelphia. Gul , F. ( 1989 ). ‘Bargaining foundations of the Shapley value’ , Econometrica , vol. 57 ( 1 ), pp. 81 – 95 . Google Scholar Crossref Search ADS WorldCat Hart , S. and Levy , Z. ( 1999 ). ‘Efficiency does not imply immediate agreement’ , Econometrica , vol. 67 ( 4 ), pp. 909 – 12 . Google Scholar Crossref Search ADS WorldCat Harsanyi , J. C. ( 1974 ). ‘An equilibrium‐point interpretation of stable sets and a proposed alternative definition’ , Management Science , vol. 20 , pp. 1422 – 95 . Google Scholar Crossref Search ADS WorldCat Huang , C. Y. ( 2002 ). ‘Multilateral bargaining: conditional and unconditional offers’ , Economic Theory , vol. 20 ( 2 ), pp. 401 – 12 . Google Scholar Crossref Search ADS WorldCat Huang , C. Y. and Sjostrom , T. ( 2003 ). ‘Consistent solutions for cooperative games with externalities’ , Games and Economic Behavior , vol. 43 , pp. 196 – 213 . Google Scholar Crossref Search ADS WorldCat In , Y. and Serrano , R. ( 2003 ). ‘Agenda restrictions in multi‐issue bargaining (II): unrestricted agendas’ , Economics Letters , vol. 79 , pp. 325 – 31 . Google Scholar Crossref Search ADS WorldCat In , Y. and Serrano , R. ( 2004 ). ‘Agenda restrictions in multi‐issue bargaining’ , Journal of Economic Behavior and Organization vol. 53 , pp. 385 – 99 . Google Scholar Crossref Search ADS WorldCat Indridason , I. H. ( 2005 ). ‘A theory of coalitions and clientelism: coalition politics in Iceland 1945–2000’ , European Journal of Political Research , vol. 44 ( 3 ), pp. 439 – 64 . Google Scholar Crossref Search ADS WorldCat Indridason , I. H. ( 2003 ). ‘Coalitions and clientelism: a comparative study’ , mimeo Jackson , M. and Wolinsky , A. ( 1996 ). ‘A strategic model of social and economic networks’ , Journal of Economic Theory , vol. 71 ( 1 ), pp. 44 – 74 . Google Scholar Crossref Search ADS WorldCat Jehiel , P. and Moldovanu , B. ( 1995 ). ‘Cyclical delay in bargaining with externalities’ , Review of Economic Studies , vol. 62 ( 4 ), pp. 619 – 37 . Google Scholar Crossref Search ADS WorldCat Kalandrakis , T. ( 2002 ). ‘Minority governments: ideology and office’ , mimeo Laver , K. and Shepsle , K. A. ( 1996 ). Making and Breaking Governments: Cabinets and Legislatures in Parliamentary Democracies . New York: Cambridge University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Laver , K. and Schofield , N. J. ( 1990 ). ‘Multiparty governments: the politics of coalitions in europe’ , Oxford: Oxford University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Levy , G. ( 2004 ). ‘A model of political parties’ , Journal of Economic Theory , vol. 115 ( 2 ), pp. 250 – 77 Google Scholar Crossref Search ADS WorldCat Lupia , A. and Strom , K. ( 1995 ). ‘Coalition termination and the strategic timing of parliamentary elections’ , The American Political Science Review , vol. 89 , pp. 648 – 65 . Google Scholar Crossref Search ADS WorldCat Maskin , E. ( 2003 ). ‘Bargaining, coalitions and externalities’ , mimeo. Merlo , A. ( 2005 ). ‘Whither political economy? theories, facts and issues’ , Manuscript prepared for the Econometric Society World Congress, 2005. Moldovanu , B. and Winter , E. ( 1995 ). ‘Order independent equilibria’ , Games and Economic Behavior , vol. 9 ( 1 ), pp. 21 – 34 . Google Scholar Crossref Search ADS WorldCat Morelli , M. and Currarini , S. ( 2000 ). ‘Network formation with sequential demands , Review of Economic Design , vol. 5 ( 3 ), pp. 229 – 49 . Google Scholar Crossref Search ADS WorldCat Morelli , M. ( 2004 ). ‘Party formation and policy outcomes under different electoral systems’ , Review of Economic Studies , vol. 71 ( 3 ), pp. 829 – 53 . Google Scholar Crossref Search ADS WorldCat Müller , W. and Strom , K. (ed.), ( 2000 ). Coalition governments in western europe ; Oxford: Oxford University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Muthoo , A. ( 1999 ). Bargaining Theory , Cambridge and Boston, Cambridge University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Na , S. ( 2005 ). ‘Three essays in game theory’ Ph.D Thesis , Pennsylvania State University. Osborne , M. and Slivinski , A. ( 1996 ). ‘A model of political competition with citizen-candidates’ , Quarterly Journal of Economics, vol. 111 ( 1 ), pp. 65 – 96 . Osborne , M. J. and Rubinstein , A. ( 1990 ). Bargaining and Markets , New York, Academic Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Osborne , M. and Tourky , R. ( 2002 ). ‘Party formation in collective decision‐making’ , Working Paper. Okada , A. ( 1996 ). ‘A non‐cooperative coalitional bargaining game with random proposers’ , Games and Economic Behavior , vol. 16 , pp. 97 – 108 . Google Scholar Crossref Search ADS WorldCat Perry , M. and Reny , P. J. ( 1994 ). ‘A non‐cooperative view of coalition formation and the core’ , Econometrica , vol. 62 , pp. 795 – 817 . Google Scholar Crossref Search ADS WorldCat Perry , M. and Reny , P. ( 1993 ). ‘Non cooperative bargaining without procedures’ , Journal of Economic Theory , vol. 59 ( 1 ), vol. 50 – 77 . Google Scholar Crossref Search ADS WorldCat Ray , D. ( 2004 ). ‘Game theoretic perspectives on group formation’ , Lipsey lectures , University of Essex. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Ray , D. and Vohra R. ( 1997 ). ‘Equilibrium binding agreements’ , Journal of Economic Theory , vol. 73 , pp. 30 – 78 . Google Scholar Crossref Search ADS WorldCat Ray , D. and Vohra , R. ( 1999 ). ‘A theory of endogenous coalition structures’ , Games and Economic Behavior , vol. 26 , pp. 286 – 336 . Google Scholar Crossref Search ADS WorldCat Riker , W. ( 1962 ). The Theory of Political Coalitions , New Haven : Yale University Press. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Riviere , A. ( 2000 ). ‘Citizen Candidacy, Party Formation and Duverger's Law’ . Discussion Paper, Royal Holloway , University of London. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Roemer , J. ( 2002 ). Political Competition: Theory and Applications , Boston: Harvard University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rosenmüller , J. , ( 1981 ). The Theory of Games and Markets , Amsterdam: North‐Holland . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rubinstein , A. ( 1982 ). ‘Perfect equilibrium in a bargaining model’ , Econometrica , vol. 50 , pp. 97 – 109 Google Scholar Crossref Search ADS WorldCat Seidmann , D. J. and Winter , E., ( 1998 ). ‘A theory of gradual coalition formation’ . Review of Economic Studies , vol. 65 , pp. 793 – 815 . Google Scholar Crossref Search ADS WorldCat Serrano , R. ( 1993 ). ‘Non‐cooperative implementation of the nucleolus: the 3‐player case’ . International Journal of Game Theory , vol. 22 , pp. 345 – 57 . Google Scholar Crossref Search ADS WorldCat Selten , R. ( 1981 ). ‘A Non‐Cooperative Model of Characteristic Function Bargaining’, reprinted in Models of Strategic Rationality , Kluwer Academic Publishers, 1989 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Snyder , J. and Ting , M. ( 2002 ). ‘An informational rationale for political parties’ , American Journal of Political Science , vol. 46 ( 1 ), pp. 90 – 110 . Google Scholar Crossref Search ADS WorldCat Winter , E. ( 1994 ). ‘The demand commitment bargaining and snowballing cooperation’ , Economic Theory , vol. 4 , pp. 255 – 73 . Google Scholar Crossref Search ADS WorldCat Author notes " We thank an anonymous referee for comments on an earlier draft. © Royal Economic Society 2006.

Poddar,, Sougata

doi: 10.1111/j.1468-0297.2006.01069_9.xpmid: N/A

This is a comprehensive guide to the movie industry written with great clarity and incisiveness, which delivers plenty of original and interesting insights palatable to both the specialist in the movie business and avid reader interested in the movie industry. The book first brings its readers through a detailed outline of the momentous historical developments in the Hollywood film industry, tracing the transition and impacts of these changes in shaping the industry's current distribution arrangements and contractual regulations of actors and actresses. Coming from a political economy analysis approach, the first chapter in the compilation offers a refreshing perspective inimitable in the existing literature of motion pictures, where concepts of ‘social change’, ‘social totality’, and ‘ownership and control’ are applied to understand the operations of the movie business. Question like, why are Hollywood films popular with audiences all over the world, is raised with great interest. Political economists are interested in how US films came to dominate international markets, what mechanisms are in place to sustain such market dominance and how the state involved in this process. From the political economy of film, the book goes on to examine the management strategies adopted by film executives by using econometric models. The studies gathered concur with the established conclusions in the existing film literature that although big‐budget movies lead to higher revenues, they generally produce lower rates of return. Stars do not help or hurt movies, and a G or PG status increases the return on investment. However, unlike studies in the existing literature of motion pictures, for the first time, this book goes on to uncover the paradox of why so many big‐budget movies and star‐studded cast productions are made, and yet so few family‐oriented movies (G and PG rated) are produced in the industry. Ingeniously, the author explained this paradox as ‘sub‐optimal risk‐management strategies adopted by extremely risk‐averse executives’ that thrive in the film industry. On the thorny and complicated subject of creative book‐keeping in the film industry where costs and revenues flow in at different timings and in different forms, the book does a good job in delineating accounting standards in the film industry, illustrating how they must be tailored to meet the disparate interests of the two positions, namely the corporate and the participants’ point of view. Moving away from the subject of profit distribution in a film production, the next chapter examines the distribution patterns in the film industry, and offers an in‐depth analysis on the impact of the different types of theatrical release patterns on the amount of revenue generated by motion pictures. While most analysts examining the distribution sector failed to capture the subjective nature of pre‐release consumer expectations, the authors overcame this impediment by using production budgets of movies as a descriptive statistic that approximates closely to these actual expectations, and subsequently uses production budgets as a measure of the power and importance of the distributor in the industry to construct the impact of theatrical release patterns on the level of revenue generated. Some critical issues, practice and research in the film exhibition business are taken up next. The author talks about the threats and opportunities with the arrival of digital technology in the movie business. For example, if digital projection devices are adopted they will increase exhibitors’ house allowance in the contract and simultaneously increase distributor's control over movie screening in houses. From the consumer side, it will definitely enhance consumers’ experience to enjoy the movie but possibly with a higher price of the ticket. The final chapter talks about the ancillary markets for movies. Videos, television, consumer products and myriad other tangible or intangible products are linked to a movie these days. In total, these ancillary products account for more revenue than the movie itself. Does it mean that movie should follow the other markets or that the movie should lead the other markets? The book ends with this open question. One criticism about the book would be: it focuses on movies mainly originated from Hollywood, and thus studies mainly the US movie industry. A study on another movie industry from a different country, say, France or UK (or suchlike) could have given us a different perspective. In recent times, the Indian movie industry globally known as Bollywood (because majority of the Indian films are actually made in Bombay hence the name) is making waves and making its place in the global market. The market for Bollywood movies is expanding fast. Bollywood produces the highest number of movies every year (higher than Hollywood). Thus, the time has come where one can hardly ignore to study the evolution of Indian movie industry. These criticisms notwithstanding, the book will certainly be found useful by researchers, students, professionals and non‐specialists. The book has contributed new insights to the existing literature of the economics of the movie industry, which effectively advance our understanding of the movie business. © Royal Economic Society 2006.

Barassi, Marco, R

doi: 10.1111/j.1468-0297.2006.01069_4.xpmid: N/A

The book provides a thorough treatment of microeconometric analysis, involving the analysis of individual level data on the economic behaviour of individual agents or firms. The book is written at the mathematical level of similar graduate textbooks such as Greene, Econometric Analysis, 2003, and can therefore be used as a core text in a graduate or PhD microeconometrics course or as a reference work for graduate students or applied researchers. Although the book is oriented to the practitioner, it is not properly designed to be a cookbook in microeconometrics. It definitely does not start from scratch in that assumes a good command of matrix algebra and linear regression models, and can get fairly advanced in places where it treats cases where two or more complications occur simultaneously such as endogeneity in a logit model with panel data. However, considerable cover of practical data problems is provided, and substantial empirical examples are included in most of the chapters to illustrate the methods covered. The book is organised in six parts. The first part is made up of chapters 1–3 and presents the issues involved in microeconometric analysis describing some of the special features of the microeconometric approach to modelling, and microeconometric data structures within the general context of regression analysis. Core methods are the object of the second part of the book occupying chapters 4–10. Results for the classic linear regression models are briefly presented emphasising those issues and methods that are more relevant for understanding the rest of the book. Estimation theory (applicable to nonlinear models) is then presented and asymptotic theory is used to obtain the distribution of the estimators, with particular importance attached to obtaining robust standard error estimates relying on weaker distributional assumptions. Classical hypothesis testing when estimators are nonlinear is also presented in this core part of the book. Semi‐parametric estimation and computational methods used to compute parameters when the estimator is defined implicitly concludes this part of the book. Part 3 of the book covers simulation‐based estimation methods. Bootstrapping (chapter 11) is the first topic covered highlighting the usefulness of the method in cases where obtaining standard errors may cumbersome due to the complexity of asymptotic theory in circumstances such as two‐step estimators, or in the case of small samples. In chapter 12, it is showed how simulation based estimation can be of great help also when the integral over a probability distribution has no closed form. Maximum simulated likelihood estimation and simulated moment based estimation are analysed in this chapter. Bayesian methods conclude part 3. Part 4 (chapters 14–20) is the heart of the book and presents advanced models for cross‐section data. The topics treated are mainly nonlinear models for limited dependent variables and sample selection. Binary outcome models such as logit and probit are thoroughly covered and an extended (multinomial logit). The case where the variable of interest is incompletely observed or it is observed in a selected sample is the object of chapter 16. The cases of truncated data or censored data are dealt with by means of nonlinear Tobit models. Sample selection problems are also widely covered and analysed. The Heckman's two‐step estimator, and the Roy's model are treated in some detail and other techniques such as semi‐parametric estimators are covered. Survival analysis is also included in a separate chapter (chapter 17). Topics covered are the standard regression models including the exponential, the Weibull and the Cox proportional hazard model. The fourth part is then completed by count data models. Part 5 is dedicated to Panel data models. The basic linear panel data model is presented in chapter 21, with particular attention to fixed and random effects models. Extensions of linear models to allow for lagged dependent variables and endogenous regressors are presented in chapter 22. The estimation procedures treated include GMM, Instrumental variables estimation and Arellano–Bond estimator for dynamic Panels. Nonlinear panel data models are the objects of chapter 23. The chapter extends linear panel estimation methods to the nonlinear estimation models of chapters 14–20. Both static and dynamic nonlinear models for, discrete data, selection models, transition data and count data are covered. Non‐parametric estimation is also surveyed. The last part of the book (chapters 24–27) considers some important topics that are related to the models examined in the parts 4 and 5. These topics include: modelling clustered data, treatment evaluation, measurement error and missing data. The book is written elegantly and treats the material in a comprehensive but still accessible way. Undoubtedly, another of its merits is that the book includes a broad range of topics, some of which are still very much object of research. In summary, Microeconometrics by A. C. Cameron, and P. K. Trivedi is very good textbook for students and a precious resource for practitioners and researchers. © Royal Economic Society 2006.

Garratt,, Anthony;Vahey, Shaun, P

doi: 10.1111/j.1468-0297.2006.01067.xpmid: N/A

Abstract We characterise the relationships between preliminary and subsequent measurements for 16 commonly‐used UK macroeconomic indicators drawn from two existing real‐time data sets and a new nominal variable database. Most preliminary measurements are biased predictors of subsequent measurements, with some revision series affected by multiple structural breaks. To illustrate how these findings facilitate real‐time forecasting, we use a vector autoregresion to generate real‐time one step ahead probability event forecasts for 1990Q1 to 1999Q2. Ignoring the predictability in initial measurements understates considerably the probability of above trend output growth. ‘Early estimates of key economic figures … can only tell an incomplete story. The ending to the story … is commonly not revealed until much later when the flow of relevant information from survey respondents has dried up and the results have been cross‐checked with other information. What first appeared to be an unremarkable movement in GDP for example can grow into something more newsworthy, and of course the opposite can happen too.’ David Rhind, Chairman of the Statistics Commission, in Mitchell (2004, vol. 1, p. 3). To aid timely economic decisions, real‐time forecasters and policy makers utilise preliminary data and make assumptions about the relationships between those and subsequent (more accurate) measurements. In principle, a better understanding of the revisions processes can increase the quality of real‐time macroeconomic analysis. The data characteristics of interest include the typical deviation between preliminary measurements and subsequent measurements, the variability of the deviations and any time variation in the revision processes. The same characteristics are also of great relevance to researchers concerned with ex post policy and forecast evaluations. These studies require further assumptions about how real‐time agents deal with preliminary data. In practice, many studies presume incorrectly that the agents know the measurements available subsequently to researchers and that these measurements will be unrevised in the future. In this article, we characterise the revisions to a variety of commonly‐used UK macroeconomic indicators. We find that the preliminary measurements of most real‐side indicators are downwards biased predictors of subsequent measurements (at the sample means). Using the Bai and Perron (2003a,b) test for multiple structural breaks of unknown timing, we find that breaks affect the relationships between early and later measurements for many variables. These breaks often pre‐date the much‐publicised late 1980s and early 1990s reforms to UK statistical reporting stemming from the Pickford Report (1989) and subsequently two phases of ‘Chancellor's Initiatives’(Wroe, 1993). Previous studies, including (among others) Akritidis (2003), Barklem (2000), Symons (2001), Castle and Ellis (2002) and Mitchell (2004) have noted the predictability property for the expenditure measure of output and its components but none provide a formal analysis of structural breaks. Some of the UK indicators characterised in this study are drawn from two existing real‐time data sets, Castle and Ellis (2002) and Egginton et al. (2002). Only Castle and Ellis (2002) characterise the revisions processes in detail, restricting attention to the expenditure measure of output and its components. In addition, we analyse real‐time quarterly monetary aggregates, nominal GDP and price deflator variables collected specifically for this study – and neglected in earlier UK databases. The preliminary measurements of UK monetary aggregates are largely unbiased. In contrast, initial nominal GDP and GDP price deflator measurements typically understate final measurements. The revisions to these nominal variables rarely exhibit structural breaks. The untypical behaviour of monetary aggregate revisions reflects the very different collection processes for these series. Often macroeconomic models perform better with revised data than with preliminary measurements. Real‐time data sets allow researchers to condition their ex post model analyses on the information set actually available to forecasters and policymakers in real time. But if researchers ignore the predictability in initial measurements, real‐time model performance can be misjudged. To illustrates this, we use a vector autoregression (VAR) in UK real output growth and inflation to forecast the (one‐step ahead) probability of above‐trend growth – sometimes referred to as the likelihood of ‘positive momentum’. Ignoring the predictability in initial measurements understates the event probability considerably for our 1990Q1 – 1999Q2 evaluation period. Our forecasting example indicates the scope for improvement in real‐time model performance but the variation in revisions predictability across variables and through time makes generalisation perilous. The remainder of the article is organised as follows. In Section 1, we discuss the sources of UK real‐time data. We describe our methodology for characterising UK real‐time data in Section 2 and report the main results in Section 3. We analyse our illustrative probability forecasting VAR exercise in Section 4. Section 5 concludes with a discussion about the implications of our study for UK macroeconomic statistical reporting. 1. Data Sources Two on‐line real‐time UK data sources have appeared in the last couple of years: Castle and Ellis (2002) and Egginton et al. (2002).1 Both studies adopt the standard terminology used in the more recent literature to describe the data; see, for example, Diebold and Rudebusch (1991). Typical macro databases store each time series variable as a column (or row) vector. In the real‐time data literature, the remeasurements are recorded as successive column vectors, and the data for each variable are usually stored as a matrix. The ‘vintage date’ refers to the release date of each vector of time series measurements and the ‘vintage’ denotes the column vector of time series data. Real‐time data comprises many vintages; each successive column vector represents a vintage containing the data available at that vintage date. The ‘most recent’, ‘current’ and ‘final’ labels are used interchangeably to denote the column with the latest vintage date. These are not the ‘true’ measurements, however, since these will be revised subsequently too. Some researchers, including Egginton et al. (2002), use successive vintages (columns) reflecting common practice by applied econometricians in real‐time policy and forecasting analyses. Others, including Howrey (1978) and Koenig et al. (2003), use measurements that have been revised the same number of times (from the diagonals of the real‐time data matrix for a particular variable). Castle and Ellis (2002) provide the most comprehensive UK real‐time data set.2 The variables comprise the expenditure components measure of real GDP (known as GDP(E)) in constant prices: private consumption, investment, government consumption, changes in inventories, exports, imports and GDP(E). The quarterly seasonally adjusted variables were published initially by the Office for National Statistics (ONS) in Economic Trends and its Annual Supplement. An MS‐Excel file contains separate sheets for each variable. Following the standard conventions in the literature, the columns reflect the vintages, with time series observations in the rows. The first vintage refers to 1961Q1 and currently the last refers to 2003Q4.3 Since a typical quarter contains multiple vintages, the frequency of the vintage dates exceeds the frequency of the time series observations. Egginton et al. (2002) provide additional real‐time data for: GDP(O) (output measure of real GDP), private consumption, retail sales, government surplus, unemployment (total claimant count), M0, M3, M4, industrial production and average earnings.4 The first two quarterly series and the remaining monthly variables came from the ONS publications Economic Trends and Financial Statistics. Variables are downloadable individually in MS‐Excel and ASCII text format. With the exception of the monetary aggregates, the sequence of vintages starts in January 1980 and end in June 1999. For the monetary variables, M0, M3 and M4, the first vintages are June 1981, January 1980 and June 1987 respectively, reflecting availability in the source publications. All variables are seasonally adjusted except the budget surplus. Unfortunately, the Egginton et al. data set contains no ‘deep history’ information. The published versions of the original sources only show a (moving) window of data at any point in time. Empty cells denote data outside of that window – generally in excess of two years before the vintage date. One concern for researchers interested in UK monetary issues is the absence of quarterly monetary aggregates.5 To address this omission, we collected real‐time data on quarterly seasonally adjusted M0 and M4 from the ONS’Economic Trends for the vintages July 1987 to August 2002. We included (from the same sources) additional real‐time information on nominal GDP(E), GDP price deflator, M0 velocity and M4 velocity. The interest in money velocity stems from its pivotal role in the UK's 1980's monetary targeting experiments.6 For the last four variables, the vintages start in November 1981 and end in 2002. Like the Egginton et al. data set, the absence of deep history results in some empty cells. The Appendix contains more complete data descriptions.7 The causes of the UK revisions apparent in all three data sets are discussed in detail by Castle and Ellis (2002) and Mitchell (2004). In brief, revisions occur when the ONS receives new data, changes its methodology or re‐bases variables. The new data category sometimes involves the substitution of delayed survey information for earlier judgement. The changes in methodology, associated with both the major structural reforms, following the Pickford Report and the Chancellor's Initiatives (Wroe, 1993), and other more minor reforms have unknown implementation dates. In contrast, the re‐basing dates are known and occur approximately every five years. Unlike the other variables in our study, the monetary aggregate data were collected by the Bank of England not the ONS. Topping and Bishop (1989) discuss the definitions, collection of, breaks in and revisions to UK monetary aggregates. 2. Methodology Our basic model for characterising UK remeasurements is: (1) where defines the ‘revisions’, denotes the growth rate of the ‘final’ measurement and denotes the kth measurement of the growth rate of the macro variable, k = 1,…,K where K < F. Notice that the preliminary measurement on the right hand side predates the final measurement used to construct the left hand side variable. The model corresponds to the ‘news’ or ‘rational forecast’ specification analysed by (among others) Mankiw et al. (1984). The null hypothesis of unbiasedness, αk = 0 and βk = 0, indicates unpredictable data revisions. The orthogonality error condition of ordinary least squares ensures that revision errors are uncorrelated with preliminary measurements.8 Since the index k = 1, … ,K denotes the successive measurements for each time series observation, the variable is formed from many ‘vintages’: one data point is taken from each vintage. In the results that follow, we restrict attention to the k =1 case for brevity. Results for the k = 2, … ,K case can be obtained from the authors on request.9 For the vector of ‘final’ data, , we use the vintage available from the ONS’Economic Trends, 6 March 2003 (electronic version). A substantial time interval exits between the respective sample end dates and the final vintage date to allow revisions to occur.10 Our model could be extended to allow other macro indicators from the same information set as as explanatory variables; see, for example, Swanson and Van Dijk (2006). Revisions are ‘efficient’ if, and only if, αk, βk and the coefficients on the additional explanatory variables are zero. Unfortunately, theory provides no guidance on what other variables might be useful for testing efficiency and unrestricted searches for predictability undoubtedly result in a degree of data snooping. In the absence of a theoretical basis for an examination of the predictability arising from other variables, we prefer to test for bias – a sufficient (but not necessary) condition for inefficiency – and test for multiple structural breaks. Given the unknown implementation dates of some wide‐ranging reforms to the UK data reporting processes (Wroe, 1993), we adopt the methodology proposed by Bai and Perron (2003a,b) to search for multiple structural breaks of unknown timing.11 We introduce some additional notations to our basic revisions equation (1): (2) for j = 1, … ,m + 1. The linear regression has m breaks (m + 1 regimes) where the indices (T1, … ,Tm) — the break points — are unknown, with T0 = 0 and Tm+1 =T. So for the one break point case, m = 1 and j = 1,2, and the pair of estimated parameters corresponds to the sample t = 1, … ,T1 and , corresponds to the sample t = T1 + 1, … ,T. We define a break as a change in at least one of the parameters αk and βk. The Bai and Perron (2003a,b) algorithm conducts efficient automated searches for multiple breaks based on a dynamic programming approach. The researcher chooses a maximum number of candidate breaks, N, and a trimming factor, τ. Given these inputs the algorithm splits the sample into feasible sub‐samples. The maximised value of the residual sum of squares identifies the candidate breaks for each number of breaks, n = 1, … ,N. The researcher tests the null hypothesis of no structural change against the alternative of many changes by a Sup Wald test. Having identified at least one change, the number of breaks is identified by specifying the null of n = L (1 ≤ L < N) changes against L + 1 changes and conducting a sequence of SupF(L + 1|L) tests. The Bai‐Perron approach is robust to serial correlation and heteroscedasticity. Minor reforms to statistical reporting procedures could induce the latter and slow adjustment by the agency would cause the former; see Barklem (2000). Our approach tests the stability of bias allowing for badly‐behaved errors. 3. Characterising Uk Revisions 3.1. Data For our characterisations of UK data remeasurements we use sixteen variables in total. The first six variables are from the Castle‐Ellis data set and comprise GDP(E), consumption, investment, government expenditure, exports and imports, all for the period 1961Q3–1999Q2; the second six are from the Garratt‐Vahey data set and are nominal GDP and the GDP price deflator for 1981Q1–1999Q4, M0 and M4 for 1987Q1–1999Q4, M0 velocity for 1987Q1–1999Q4 and M4 velocity for 1986Q4–1998Q4; the final four are from the Egginton et al. data set and are average earnings for 1979M11–1997M1, industrial production for 1979M11–1997M8, claimant count unemployment for 1979M12–1997M10 and retail sales for 1986M2–1997M12. Since the indicators vary by source and time series frequency, the sample size, the trimming factor τ (as a proportion of the sample) and the maximum number of breaks, N, vary.12 We set N = 5 and τ = 0.15 for the 150 plus observations for both the quarterly GDP(E) components in the Castle‐Ellis data set and the monthly indicators from Egginton et al.13 For the Garratt‐Vahey monetary aggregates and velocities, where there are 52 or 53 quarterly observations, we set N = 1 and τ = 0.25. For nominal GDP and the price deflator, there are 76 quarterly observations and we set N = 2 and τ = 0.25. We use quarterly or monthly growth rates as appropriate throughout the empirical section.14 This approach mitigates the level effects that result from base year changes (Patterson and Hervai, 1991). In general, conventional unit root tests indicate that the variables in (2) are stationary, despite the small samples and the likely presence of structural breaks. Table 1 reports the means and standard deviations of revisions, . In general, the mean revisions are positive: preliminary measurements understate final measurements but there is considerable variation across variables. Approximately half of the indicators have statistically significant mean revisions at the 5% level (denoted by * in Table 1). Investment has the largest (quarterly) mean revisions: nearly twice as big as GDP(E).15 The notably small M0 and M4 mean revisions are insignificantly different from zero at the 5% level. The mean absolute error for the monetary aggregates is also notably lower than for the other variables. The preliminary analysis suggest little predictability for monetary aggregate revisions. Table 1 Summary Statistics for Revisions, . Sample . Mean . MAE . SD . SD 1990s . GDP(E): 1961Q3 1999Q2 0.24* 0.88 1.20† 0.31 Consumption: 1961Q3 1999Q2 0.10* 0.72 0.95† 0.48 Investment: 1961Q3 1999Q2 0.49* 1.87 2.40† 1.71 Government expenditure: 1961Q3 1999Q2 −0.07 0.96 1.32 1.07 Exports: 1961Q3 1999Q2 0.3* 1.45 1.80† 1.52 Imports: 1961Q3 1999Q2 0.16 1.44 1.84† 1.33 Nominal GDP: 1981Q1 1999Q4 0.29* 0.56 0.70 0.64 GDP deflator: 1981Q1 1999Q4 0.07 0.62 0.79† 0.64 M0: 1987Q1 1999Q4 0.05 0.36 0.52† 0.29 M4: 1987Q1 1999Q4 −0.01 0.26 0.35 0.31 M0 velocity: 1987Q1 1999Q4 0.25* 0.61 0.71† 0.40 M4 velocity: 1986Q4 1999Q4 0.39* 0.54 0.68 0.67 Average earnings: 1979M11 1997M1 0.03 0.49 0.68 0.77 Industrial production: 1979M11 1997M8 0.05 0.68 0.93† 0.72 Unemployment: 1979M12 1997M10 0.02 0.41 0.61 0.72 Retail sales: 1986M2 1997M12 0.04 0.56 0.73† 0.60 . Sample . Mean . MAE . SD . SD 1990s . GDP(E): 1961Q3 1999Q2 0.24* 0.88 1.20† 0.31 Consumption: 1961Q3 1999Q2 0.10* 0.72 0.95† 0.48 Investment: 1961Q3 1999Q2 0.49* 1.87 2.40† 1.71 Government expenditure: 1961Q3 1999Q2 −0.07 0.96 1.32 1.07 Exports: 1961Q3 1999Q2 0.3* 1.45 1.80† 1.52 Imports: 1961Q3 1999Q2 0.16 1.44 1.84† 1.33 Nominal GDP: 1981Q1 1999Q4 0.29* 0.56 0.70 0.64 GDP deflator: 1981Q1 1999Q4 0.07 0.62 0.79† 0.64 M0: 1987Q1 1999Q4 0.05 0.36 0.52† 0.29 M4: 1987Q1 1999Q4 −0.01 0.26 0.35 0.31 M0 velocity: 1987Q1 1999Q4 0.25* 0.61 0.71† 0.40 M4 velocity: 1986Q4 1999Q4 0.39* 0.54 0.68 0.67 Average earnings: 1979M11 1997M1 0.03 0.49 0.68 0.77 Industrial production: 1979M11 1997M8 0.05 0.68 0.93† 0.72 Unemployment: 1979M12 1997M10 0.02 0.41 0.61 0.72 Retail sales: 1986M2 1997M12 0.04 0.56 0.73† 0.60 Notes. The revisions, , are defined as the final measurement, , minus the first measurement, . Each measurement, Xt, refers to the quarter‐on‐quarter (first 12 variables) or month‐on‐month (last 4 variables) growth rate in %. MAE is the mean absolute error; SD refers to standard deviation and SD 1990s refers to the standard deviation for the 1990s. The symbol * denotes statistical significance at the 5% level using a Newey‐West (1987) corrected t‐statistic based on a regression of the revision on a constant. Significantly lower variance for the 1990s at the 5% level using a variance ratio test is denoted by † (for break dates see main text). Open in new tab Table 1 Summary Statistics for Revisions, . Sample . Mean . MAE . SD . SD 1990s . GDP(E): 1961Q3 1999Q2 0.24* 0.88 1.20† 0.31 Consumption: 1961Q3 1999Q2 0.10* 0.72 0.95† 0.48 Investment: 1961Q3 1999Q2 0.49* 1.87 2.40† 1.71 Government expenditure: 1961Q3 1999Q2 −0.07 0.96 1.32 1.07 Exports: 1961Q3 1999Q2 0.3* 1.45 1.80† 1.52 Imports: 1961Q3 1999Q2 0.16 1.44 1.84† 1.33 Nominal GDP: 1981Q1 1999Q4 0.29* 0.56 0.70 0.64 GDP deflator: 1981Q1 1999Q4 0.07 0.62 0.79† 0.64 M0: 1987Q1 1999Q4 0.05 0.36 0.52† 0.29 M4: 1987Q1 1999Q4 −0.01 0.26 0.35 0.31 M0 velocity: 1987Q1 1999Q4 0.25* 0.61 0.71† 0.40 M4 velocity: 1986Q4 1999Q4 0.39* 0.54 0.68 0.67 Average earnings: 1979M11 1997M1 0.03 0.49 0.68 0.77 Industrial production: 1979M11 1997M8 0.05 0.68 0.93† 0.72 Unemployment: 1979M12 1997M10 0.02 0.41 0.61 0.72 Retail sales: 1986M2 1997M12 0.04 0.56 0.73† 0.60 . Sample . Mean . MAE . SD . SD 1990s . GDP(E): 1961Q3 1999Q2 0.24* 0.88 1.20† 0.31 Consumption: 1961Q3 1999Q2 0.10* 0.72 0.95† 0.48 Investment: 1961Q3 1999Q2 0.49* 1.87 2.40† 1.71 Government expenditure: 1961Q3 1999Q2 −0.07 0.96 1.32 1.07 Exports: 1961Q3 1999Q2 0.3* 1.45 1.80† 1.52 Imports: 1961Q3 1999Q2 0.16 1.44 1.84† 1.33 Nominal GDP: 1981Q1 1999Q4 0.29* 0.56 0.70 0.64 GDP deflator: 1981Q1 1999Q4 0.07 0.62 0.79† 0.64 M0: 1987Q1 1999Q4 0.05 0.36 0.52† 0.29 M4: 1987Q1 1999Q4 −0.01 0.26 0.35 0.31 M0 velocity: 1987Q1 1999Q4 0.25* 0.61 0.71† 0.40 M4 velocity: 1986Q4 1999Q4 0.39* 0.54 0.68 0.67 Average earnings: 1979M11 1997M1 0.03 0.49 0.68 0.77 Industrial production: 1979M11 1997M8 0.05 0.68 0.93† 0.72 Unemployment: 1979M12 1997M10 0.02 0.41 0.61 0.72 Retail sales: 1986M2 1997M12 0.04 0.56 0.73† 0.60 Notes. The revisions, , are defined as the final measurement, , minus the first measurement, . Each measurement, Xt, refers to the quarter‐on‐quarter (first 12 variables) or month‐on‐month (last 4 variables) growth rate in %. MAE is the mean absolute error; SD refers to standard deviation and SD 1990s refers to the standard deviation for the 1990s. The symbol * denotes statistical significance at the 5% level using a Newey‐West (1987) corrected t‐statistic based on a regression of the revision on a constant. Significantly lower variance for the 1990s at the 5% level using a variance ratio test is denoted by † (for break dates see main text). Open in new tab To illustrate the scale of revisions, Figure 1 plots GDP(E) from 1961Q3 to 1999Q2 for the first and final measurements. The deviation between the two shows the k = 1 revision. At times, these are larger in absolute size than the quarterly economic growth rate. Figure 1 also shows that the final measurements are much less volatile post‐1989, reflecting the relative stability of the 1990s boom. Fig. 1. Open in new tabDownload slide GDP(E) Growth, First and Final Measurements Fig. 1. Open in new tabDownload slide GDP(E) Growth, First and Final Measurements To check for structural change in the mean revision of each variable, we estimated a restricted version of equation (2) with . We used the Bai‐Perron methodology to identify structural breaks of unknown timing in the intercept. There are breaks in the means only for exports (1993Q3) and imports (1992Q1).16 (The results reported in the next Section based on unrestricted estimation of (2) suggest that the data reject the restriction and that structural breaks are much more prevalent.) To investigate time variation in the standard deviations for each GDP(E) component, we split the sample into two sub‐samples, corresponding approximately to the 1980s and 1990s.17 The results suggest a fairly consistent pattern: lower standard deviations for the 1990s. For 10 of the 16 variables, the data reject the null hypothesis of equal variances for the two sub‐samples at the 5% level using a variance ratio test (denoted by † in Table 1). We conclude from this preliminary investigation that revisions are often predictable and typically positive, with considerable variation in size across variables and lower 1990s’ revision volatility. 3.2. Testing for Bias Tables 2, 3 and 4 summarise the results from our regressions based on (2) using 16 macro indicators for the first measurements (k = 1).18 In each case, we report the p‐value for the Wald test of the null hypothesis for unbiasedness, α1 = β1 = 0, Newey‐West heteroscedasticity and serial correlation consistent standard errors and an LM‐test statistic for serial correlation.19 The tables show the bias for each parameter‐stable segment; if there are no structural breaks, we report the results for the full sample. The break points are also shown on a time line in Figure 2. Fig. 2. Open in new tabDownload slide Break Points Fig. 2. Open in new tabDownload slide Break Points Castle‐Ellis Variables. Table 2 reports the results for GDP(E) and its components. Most of these variables have breaks that pre‐date structural reforms to ONS practices in the late 1980s and early 1990s. The exports break in 1993Q3 coincides with the rebasing of national accounts. In general, the null hypothesis of α = β =0 can be rejected at the 1% level, with variation in the size of the bias across variables. Initial measurements are unfailingly revised upwards (at the sample means). For example, the estimated α and β values for GDP(E) (investment) are in the region of 0.4 (0.6) and −0.6 (−0.3) respectively. This implies preliminary GDP(E) (investment) measurements around the sample mean (quarterly) output growth of 0.4% (0.5%) would be revised to nearly 0.6% (0.9%). Nearly all variables subject to structural breaks display bias before and after the breaks; the absolute values of the coefficients are sometimes larger post‐break. The null hypothesis of unbiased revisions can only be not rejected in one sub‐sample: for ports 1972Q1–1980Q3. Table 2 Revisions Regressions, Castle‐Ellis . Sample . α . β . . Wald‐test . LM‐test . GDP(E): 1961Q3 1999Q2 0.444 −0.573 0.58 0.00 0.30 (0.065) (0.050) Consumption: 1961Q3 1967Q2 0.478 −0.682 0.64 0.00 0.86 (0.120) (0.087) 1967Q3 1999Q2 0.252 −0.288 0.20 0.00 0.00 (0.072) (0.051) Investment: 1961Q3 1999Q2 0.563 −0.320 0.17 0.00 0.01 (0.147) (0.074) Government expenditure: 1961Q3 1975Q1 0.421 −0.591 0.24 0.01 0.17 (0.185) (0.182) 1975Q2 1999Q2 0.181 −0.816 0.33 0.00 0.01 (0.083) (0.128) Exports: 1961Q3 1971Q4 0.630 −0.237 0.43 0.00 0.13 (0.117) (0.032) 1972Q1 1980Q3 0.219 −0.068 0.01 0.34 0.03 (0.156) (0.077) 1980Q4 1993Q3 0.495 −0.525 0.45 0.00 0.00 (0.139) (0.063) 1993Q4 1999Q2 1.274 −0.458 0.25 0.00 0.43 (0.290) (0.159) Imports: 1961Q3 1985Q4 0.156 −0.120 0.05 0.03 0.00 (0.130) (0.043) 1986Q1 1999Q2 0.920 −0.473 0.42 0.00 0.67 (0.290) (0.082) . Sample . α . β . . Wald‐test . LM‐test . GDP(E): 1961Q3 1999Q2 0.444 −0.573 0.58 0.00 0.30 (0.065) (0.050) Consumption: 1961Q3 1967Q2 0.478 −0.682 0.64 0.00 0.86 (0.120) (0.087) 1967Q3 1999Q2 0.252 −0.288 0.20 0.00 0.00 (0.072) (0.051) Investment: 1961Q3 1999Q2 0.563 −0.320 0.17 0.00 0.01 (0.147) (0.074) Government expenditure: 1961Q3 1975Q1 0.421 −0.591 0.24 0.01 0.17 (0.185) (0.182) 1975Q2 1999Q2 0.181 −0.816 0.33 0.00 0.01 (0.083) (0.128) Exports: 1961Q3 1971Q4 0.630 −0.237 0.43 0.00 0.13 (0.117) (0.032) 1972Q1 1980Q3 0.219 −0.068 0.01 0.34 0.03 (0.156) (0.077) 1980Q4 1993Q3 0.495 −0.525 0.45 0.00 0.00 (0.139) (0.063) 1993Q4 1999Q2 1.274 −0.458 0.25 0.00 0.43 (0.290) (0.159) Imports: 1961Q3 1985Q4 0.156 −0.120 0.05 0.03 0.00 (0.130) (0.043) 1986Q1 1999Q2 0.920 −0.473 0.42 0.00 0.67 (0.290) (0.082) Notes. Revisions regression (2), . Newey‐West (1987) standard errors (truncation factor 4) are in parentheses. We report p‐values of the Wald‐test for α = β = 0 and the LM‐test statistic up to 4th‐order serial correlation. Open in new tab Table 2 Revisions Regressions, Castle‐Ellis . Sample . α . β . . Wald‐test . LM‐test . GDP(E): 1961Q3 1999Q2 0.444 −0.573 0.58 0.00 0.30 (0.065) (0.050) Consumption: 1961Q3 1967Q2 0.478 −0.682 0.64 0.00 0.86 (0.120) (0.087) 1967Q3 1999Q2 0.252 −0.288 0.20 0.00 0.00 (0.072) (0.051) Investment: 1961Q3 1999Q2 0.563 −0.320 0.17 0.00 0.01 (0.147) (0.074) Government expenditure: 1961Q3 1975Q1 0.421 −0.591 0.24 0.01 0.17 (0.185) (0.182) 1975Q2 1999Q2 0.181 −0.816 0.33 0.00 0.01 (0.083) (0.128) Exports: 1961Q3 1971Q4 0.630 −0.237 0.43 0.00 0.13 (0.117) (0.032) 1972Q1 1980Q3 0.219 −0.068 0.01 0.34 0.03 (0.156) (0.077) 1980Q4 1993Q3 0.495 −0.525 0.45 0.00 0.00 (0.139) (0.063) 1993Q4 1999Q2 1.274 −0.458 0.25 0.00 0.43 (0.290) (0.159) Imports: 1961Q3 1985Q4 0.156 −0.120 0.05 0.03 0.00 (0.130) (0.043) 1986Q1 1999Q2 0.920 −0.473 0.42 0.00 0.67 (0.290) (0.082) . Sample . α . β . . Wald‐test . LM‐test . GDP(E): 1961Q3 1999Q2 0.444 −0.573 0.58 0.00 0.30 (0.065) (0.050) Consumption: 1961Q3 1967Q2 0.478 −0.682 0.64 0.00 0.86 (0.120) (0.087) 1967Q3 1999Q2 0.252 −0.288 0.20 0.00 0.00 (0.072) (0.051) Investment: 1961Q3 1999Q2 0.563 −0.320 0.17 0.00 0.01 (0.147) (0.074) Government expenditure: 1961Q3 1975Q1 0.421 −0.591 0.24 0.01 0.17 (0.185) (0.182) 1975Q2 1999Q2 0.181 −0.816 0.33 0.00 0.01 (0.083) (0.128) Exports: 1961Q3 1971Q4 0.630 −0.237 0.43 0.00 0.13 (0.117) (0.032) 1972Q1 1980Q3 0.219 −0.068 0.01 0.34 0.03 (0.156) (0.077) 1980Q4 1993Q3 0.495 −0.525 0.45 0.00 0.00 (0.139) (0.063) 1993Q4 1999Q2 1.274 −0.458 0.25 0.00 0.43 (0.290) (0.159) Imports: 1961Q3 1985Q4 0.156 −0.120 0.05 0.03 0.00 (0.130) (0.043) 1986Q1 1999Q2 0.920 −0.473 0.42 0.00 0.67 (0.290) (0.082) Notes. Revisions regression (2), . Newey‐West (1987) standard errors (truncation factor 4) are in parentheses. We report p‐values of the Wald‐test for α = β = 0 and the LM‐test statistic up to 4th‐order serial correlation. Open in new tab Garratt‐Vahey Variables. Table 3 reports the results for the six Garratt‐Vahey nominal variables. With the exception of the early 1990s’ breaks for M0 and its velocity, these variables show stability over the period. Although nominal GDP revisions and the GDP price deflator both exhibit significant bias at the 1% level, the monetary aggregates do not, with p‐values above 10% and smaller coefficients (in absolute value). The narrower measure, M0, displays bias before the early 1990s’ break. In general, the revisions to the money velocities are biased at the 1% level – reflecting the predictability of nominal GDP revisions – with an early 1990s’ break for the narrower measure. Table 3 Revisions Regressions, Garratt‐Vahey . Sample . α . β . . Wald‐test . LM‐test . Nominal GDP: 1981Q1 1999Q4 0.929 −0.431 0.27 0.00 0.22 (0.141) (0.097) GDP deflator: 1981Q1 1999Q4 0.696 −0.595 0.30 0.00 0.16 (0.135) (0.078) M0: 1987Q1 1993Q2 0.555 −0.494 0.50 0.00 0.09 (0.131) (0.078) 1993Q3 1999Q4 0.114 −0.035 −0.03 0.46 0.08 (0.077) (0.041) M4: 1987Q1 1999Q4 −0.020 0.006 −0.02 0.97 0.16 (0.081) (0.030) M0 velocity: 1987Q1 1992Q1 0.937 −1.040 0.72 0.00 0.97 (0.049) (0.190) 1992Q2 1999Q4 0.025 −0.538 0.48 0.00 0.26 (0.074) (0.114) M4 velocity: 1986Q4 1999Q4 0.254 −0.141 0.04 0.00 0.91 (0.111) (0.80) . Sample . α . β . . Wald‐test . LM‐test . Nominal GDP: 1981Q1 1999Q4 0.929 −0.431 0.27 0.00 0.22 (0.141) (0.097) GDP deflator: 1981Q1 1999Q4 0.696 −0.595 0.30 0.00 0.16 (0.135) (0.078) M0: 1987Q1 1993Q2 0.555 −0.494 0.50 0.00 0.09 (0.131) (0.078) 1993Q3 1999Q4 0.114 −0.035 −0.03 0.46 0.08 (0.077) (0.041) M4: 1987Q1 1999Q4 −0.020 0.006 −0.02 0.97 0.16 (0.081) (0.030) M0 velocity: 1987Q1 1992Q1 0.937 −1.040 0.72 0.00 0.97 (0.049) (0.190) 1992Q2 1999Q4 0.025 −0.538 0.48 0.00 0.26 (0.074) (0.114) M4 velocity: 1986Q4 1999Q4 0.254 −0.141 0.04 0.00 0.91 (0.111) (0.80) Notes. Revisions regression (2), . Newey‐West (1987) standard errors (truncation factor 4) are in parentheses. We report p‐values of the Wald‐test for α = β = 0 and the LM‐test statistic for up to 4th‐order serial correlation. Open in new tab Table 3 Revisions Regressions, Garratt‐Vahey . Sample . α . β . . Wald‐test . LM‐test . Nominal GDP: 1981Q1 1999Q4 0.929 −0.431 0.27 0.00 0.22 (0.141) (0.097) GDP deflator: 1981Q1 1999Q4 0.696 −0.595 0.30 0.00 0.16 (0.135) (0.078) M0: 1987Q1 1993Q2 0.555 −0.494 0.50 0.00 0.09 (0.131) (0.078) 1993Q3 1999Q4 0.114 −0.035 −0.03 0.46 0.08 (0.077) (0.041) M4: 1987Q1 1999Q4 −0.020 0.006 −0.02 0.97 0.16 (0.081) (0.030) M0 velocity: 1987Q1 1992Q1 0.937 −1.040 0.72 0.00 0.97 (0.049) (0.190) 1992Q2 1999Q4 0.025 −0.538 0.48 0.00 0.26 (0.074) (0.114) M4 velocity: 1986Q4 1999Q4 0.254 −0.141 0.04 0.00 0.91 (0.111) (0.80) . Sample . α . β . . Wald‐test . LM‐test . Nominal GDP: 1981Q1 1999Q4 0.929 −0.431 0.27 0.00 0.22 (0.141) (0.097) GDP deflator: 1981Q1 1999Q4 0.696 −0.595 0.30 0.00 0.16 (0.135) (0.078) M0: 1987Q1 1993Q2 0.555 −0.494 0.50 0.00 0.09 (0.131) (0.078) 1993Q3 1999Q4 0.114 −0.035 −0.03 0.46 0.08 (0.077) (0.041) M4: 1987Q1 1999Q4 −0.020 0.006 −0.02 0.97 0.16 (0.081) (0.030) M0 velocity: 1987Q1 1992Q1 0.937 −1.040 0.72 0.00 0.97 (0.049) (0.190) 1992Q2 1999Q4 0.025 −0.538 0.48 0.00 0.26 (0.074) (0.114) M4 velocity: 1986Q4 1999Q4 0.254 −0.141 0.04 0.00 0.91 (0.111) (0.80) Notes. Revisions regression (2), . Newey‐West (1987) standard errors (truncation factor 4) are in parentheses. We report p‐values of the Wald‐test for α = β = 0 and the LM‐test statistic for up to 4th‐order serial correlation. Open in new tab Eggintonet al. Variables. Table 4 reports the results for the remaining four variables, all taken from the Egginton et al. data set. Both unemployment and industrial production have one break (in the early 1990s and mid‐1980s, respectively); average earnings has two breaks (one in the late 1980s, the other in the early 1990s). In contrast, retail sales exhibits no breaks. In general, the preliminary measurements are downwards biased predictors of subsequent measurements at their sample means – matching the pattern observed for real‐side quarterly indicators. The exceptions are unemployment, average earnings and industrial production before their respective first breaks. These sub‐samples display unbiasedness at the 15% level and above. All three indicators exhibit bias at the 1% level for subsequent sub‐samples, consistent with statistical quality degradation. Table 4 Revisions Regressions, Egginton‐Pick‐Vahey . Sample . α . β . . Wald‐test . LM‐test . Average earnings: 1979M11 1987M11 0.108 −0.138 0.05 0.23 0.00 (0.072) (0.079) 1987M12 1992M9 0.464 −0.688 0.45 0.00 0.01 (0.091) (0.129) 1992M10 1997M1 0.257 −0.907 0.87 0.00 0.19 (0.028) (0.045) Industrial production: 1979M11 1986M5 0.050 −0.033 −0.01 0.80 0.06 (0.081) (0.076) 1986M6 1997M8 0.103 −0.508 0.29 0.00 0.00 (0.047) (0.092) Unemployment: 1979M12 1992M10 0.063 −0.026 0.00 0.42 0.00 (0.056) (0.30) 1992M11 1997M10 −0.428 −0.328 0.25 0.01 0.13 (0.154) (0.101) Retail sales: 1986M2 1997M12 0.120 −0.390 0.41 0.00 0.00 (0.029) (0.042) . Sample . α . β . . Wald‐test . LM‐test . Average earnings: 1979M11 1987M11 0.108 −0.138 0.05 0.23 0.00 (0.072) (0.079) 1987M12 1992M9 0.464 −0.688 0.45 0.00 0.01 (0.091) (0.129) 1992M10 1997M1 0.257 −0.907 0.87 0.00 0.19 (0.028) (0.045) Industrial production: 1979M11 1986M5 0.050 −0.033 −0.01 0.80 0.06 (0.081) (0.076) 1986M6 1997M8 0.103 −0.508 0.29 0.00 0.00 (0.047) (0.092) Unemployment: 1979M12 1992M10 0.063 −0.026 0.00 0.42 0.00 (0.056) (0.30) 1992M11 1997M10 −0.428 −0.328 0.25 0.01 0.13 (0.154) (0.101) Retail sales: 1986M2 1997M12 0.120 −0.390 0.41 0.00 0.00 (0.029) (0.042) Notes. Revisions regression (2), . Newey‐West (1987) standard errors (truncation factor 4) are in parentheses. We report p‐values of the Wald‐test for α = β = 0 and the LM‐test statistic for up to 12th‐order serial correlation. Open in new tab Table 4 Revisions Regressions, Egginton‐Pick‐Vahey . Sample . α . β . . Wald‐test . LM‐test . Average earnings: 1979M11 1987M11 0.108 −0.138 0.05 0.23 0.00 (0.072) (0.079) 1987M12 1992M9 0.464 −0.688 0.45 0.00 0.01 (0.091) (0.129) 1992M10 1997M1 0.257 −0.907 0.87 0.00 0.19 (0.028) (0.045) Industrial production: 1979M11 1986M5 0.050 −0.033 −0.01 0.80 0.06 (0.081) (0.076) 1986M6 1997M8 0.103 −0.508 0.29 0.00 0.00 (0.047) (0.092) Unemployment: 1979M12 1992M10 0.063 −0.026 0.00 0.42 0.00 (0.056) (0.30) 1992M11 1997M10 −0.428 −0.328 0.25 0.01 0.13 (0.154) (0.101) Retail sales: 1986M2 1997M12 0.120 −0.390 0.41 0.00 0.00 (0.029) (0.042) . Sample . α . β . . Wald‐test . LM‐test . Average earnings: 1979M11 1987M11 0.108 −0.138 0.05 0.23 0.00 (0.072) (0.079) 1987M12 1992M9 0.464 −0.688 0.45 0.00 0.01 (0.091) (0.129) 1992M10 1997M1 0.257 −0.907 0.87 0.00 0.19 (0.028) (0.045) Industrial production: 1979M11 1986M5 0.050 −0.033 −0.01 0.80 0.06 (0.081) (0.076) 1986M6 1997M8 0.103 −0.508 0.29 0.00 0.00 (0.047) (0.092) Unemployment: 1979M12 1992M10 0.063 −0.026 0.00 0.42 0.00 (0.056) (0.30) 1992M11 1997M10 −0.428 −0.328 0.25 0.01 0.13 (0.154) (0.101) Retail sales: 1986M2 1997M12 0.120 −0.390 0.41 0.00 0.00 (0.029) (0.042) Notes. Revisions regression (2), . Newey‐West (1987) standard errors (truncation factor 4) are in parentheses. We report p‐values of the Wald‐test for α = β = 0 and the LM‐test statistic for up to 12th‐order serial correlation. Open in new tab 3.3. Discussion The predictability of revisions indicates the potential for improvements in UK statistical quality. Although an agency aiming to minimise revisions could exploit revision predictability via filtering, the UK approach apparently involves the less direct route based on the periodic overhauling of survey quality and in‐house estimates. The UK's well‐known statistical reforms, associated with the Pickford Report and the subsequent Chancellor's Initiatives (Wroe, 1993), had minor impacts on predictability. As shown in Figure 2, only five structural breaks occurred in the 1989–95 period. For unemployment, predictability increased post‐break. The monetary aggregates produced by the Bank of England were unaffected by the reforms to ONS procedures. Both exports and average earnings exhibit statistically significant predictability after their early 1990s’ breaks. Our preliminary analysis indicated that there was, however, some evidence that the volatility of revisions fell after the Pickford Report. To check the robustness of this characterisation in the presence of structural breaks, we tested for constant variances across each break identified by the Bai‐Perron approach. Using a variance ratio test, the null of no difference in the variance can be rejected at the 5%, with revisions volatility lower post‐break for most cases. The exceptions are unemployment and average earnings (second break). 4. Forecasting Case Study Strong revisions predictability gives scope for improving real‐time forecast performance. To illustrate this, we consider a probability event forecasting exercise. We compute one step ahead out‐of‐sample forecasts for the evaluation period 1990Q1–1999Q2 using the unrestricted VAR estimated recursively:20 (3) where , s = 1, F and B. The variables y and π denote quarterly output growth and inflation (defined using the GDP price deflator). Where the superscript s = 1, F and B denotes the separate VAR estimated for the first, final and bias‐adjusted measurements respectively. We define the bias‐adjusted measurements, , as: (4) where denotes the first measurement. We assume that the forecaster knows the true values of α and β and that they are equal to the respective sample coefficients from (2).21 To arrive at our preferred specification for the forecasting VAR, we first tested for stationarity and then selected the lag order. We could not reject the null of a unit root in the levels data but could reject the null in first differences at the 5% level using augmented Dickey‐Fuller tests (for both first and final measurement data). We selected the lag order by estimating a sequence of unrestricted VAR(p), p = 0,1,2, … ,6 models. For the first measurement data, s = 1, the optimal Akaike Information Criteria selected lag length was zero; but for final data, s = F, the lag order equalled four. Bearing in mind that unnecessary lags cause inefficiency but not bias in the OLS estimators, we standardised the lag length at four for first, final and the bias‐adjusted data. For model evaluation, we consider an economic agent monitoring business cycle turning points by calculating the probability of above trend output growth. This is sometimes referred to as ‘positive momentum’ or ‘above speed limit’ growth in the monetary policy literature (Walsh, 2003). We take the (final data) average economic growth rate for the evaluation period, 0.52%, as the ‘trend’. The agent calculates the probability , for s = 1, F and B where Ωt−1 denotes the information set dated t − 1. Confidence intervals are of limited help to our agent because the concern with turning points implies little interest in whether any particular forecast confidence interval encompass a specific value for output. Garratt et al. (2003a) and Clements (2004) discuss the appropriateness of probability forecasts and their relationships to standard forecast confidence intervals in detail. We compute the probability forecasts by stochastic simulation by the methods described by Garratt et al. (2003a, appendix).22Figure 3 plots the probabilities of the event for the three data types. For most of the evaluation period, final data results in a higher probability of above mean output growth than with first measurement data. The average difference in probabilities is 11.6 percentage points (with a standard deviation of 22.4%). Using bias‐corrected measurements rather than first‐measurement data considerably reduces the mean (absolute) difference in forecast probabilities to 4.8 percentage points (with a standard deviation of 21.7%), although substantial differences remain at times. Fig. 3. Open in new tabDownload slide One Step Ahead Probability of Above‐trend Output Growth Fig. 3. Open in new tabDownload slide One Step Ahead Probability of Above‐trend Output Growth For more formal forecasts evaluation, Table 5 reports the proportion of correctly forecast events, P, the Kuipers score statistics, KS, and the Pesaran and Timmermann (1992) directional market timing statistics, PT. Table 5 Evaluation of Probability Event Forecasts Measurements . P (%) . KS . PT . First 55.3 0.047 0.270 Final 73.7 0.514 2.865 Bias adjusted 63.2 0.368 1.428 Measurements . P (%) . KS . PT . First 55.3 0.047 0.270 Final 73.7 0.514 2.865 Bias adjusted 63.2 0.368 1.428 Open in new tab Table 5 Evaluation of Probability Event Forecasts Measurements . P (%) . KS . PT . First 55.3 0.047 0.270 Final 73.7 0.514 2.865 Bias adjusted 63.2 0.368 1.428 Measurements . P (%) . KS . PT . First 55.3 0.047 0.270 Final 73.7 0.514 2.865 Bias adjusted 63.2 0.368 1.428 Open in new tab We consider 38 events in total; one event (above trend growth) for each time period in the 1990Q1 – 1999Q2 evaluation period. We assume that an event can be correctly forecast if the associated probability forecast exceeds 50%. Although over 70% of events can be correctly forecast using final data, using first measurements and bias‐adjusted measurements reduces the success rate by approximately 19 and 10 percentage points respectively. The Kuipers scores also suggest that bias adjustment improves forecast performance. This statistic measures the proportion of above mean growth rates that were correctly forecast minus the proportion of below mean growth rates that were incorrectly forecast. The test provides a measure of the accuracy of directional forecasts, with high positive numbers indicating high predictive accuracy. Using first measurements gives a KS of approximately 0.05; bias‐adjustment betters this score by 0.32 – considerably closer to the final data score of 0.51. The PT statistic allows a formal hypothesis of directional forecasting performance. As shown in Granger and Pesaran (2000), this hypothesis test uses the same information as the Kuipers score. Under the null hypothesis that the forecasts and realisations are independently distributed the PT statistic has a standard normal distribution. The first measurement data reject the null of no ability to forecast oberved changes with a probability value of 0.78. Bias‐adjustment reduces the probability value to 0.15 – indicating rejection at the 15% level. Final data give clear rejection at the 1% level. We conclude that bias adjustment improves probability forecasting performance for this particular forecasting example.23 Although this analysis indicates the scope for exploiting revision predictability, we emphasise that the variation in predictability across variables and through time ensures that performance improvement is case specific. Furthermore, the parameters of (2) were assumed to be known by the agent (and identified as the population coefficients). In the presence of structural breaks, parameter learning may limit the scope for increasing forecast accuracy. Modelling the impacts of bounded rationality on real‐time forecast and policy model performance is an interesting area for subsequent research. 5. Conclusions By utilising both existing and new sources of real‐time data, this article has characterised the revision processes for 16 UK macro indicators. The main finding – that the preliminary measurements of UK macro variables are generally biased – confirms a widely‐held suspicion that UK macro measurements are inefficient. Where present, the bias causes preliminary measurements to understate later measurements (at the sample means) and structural breaks result in some time variation in revisions predictability. Monetary aggregates are typically unbiased (at least, post‐break) reflecting the untypical data collection processes for these variables. Using a forecasting probability example, we have demonstrated how real‐time economic agents (both public and private) might exploit revision predictability to enhance forecast ability. However, the variation in real‐time data characteristics across variables and through time makes it difficult to generalise about the extent of the improvement in model performance. Since one long‐run aim for statistics producers is to minimise preliminary data inaccuracies, there is also scope for the ONS to exploit revisions predictability. Filtering initial measurements prior to release is an obvious route to more efficient statistical reporting. But if the filtering process is unknown by the statistics user, transformed preliminary measurements severely complicate inferences about the data generating process, as noted by Sargent (1989). Hence, the production of filtered data can contribute to monetary and fiscal control problems. An alternative strategy for statistics producers focuses on improving transparency about real‐time data characteristics and the causes of revisions. Then the end‐users can filter preliminary measurements appropriately. In this respect, the recent moves towards more comprehensive analyses of revisions by the UK statistical authorities are encouraging; see for example Mitchell (2004). It remains striking, however, that although the volatility of revisions has diminished recently, the preliminary measurements of many UK macroeconomic indicators understate substantially subsequent measurements. We expect that future UK statistical reforms will result in diminished bias. Appendix Appendix: Summary of Garratt‐Vahey Real‐Time Data In this Appendix, we describe the real‐time data collected specifically for this study (referred to as the Garratt‐Vahey data set in the main text). The data consist of monthly vintages of nominal macroeconomic variables. Each variable has many different vintages — reflecting the revisions and updates that occur over time. In the MS‐Excel files, the data are stored as a matrix for each variable. Successive column vectors of the matrix represent different (more recent) vintages of data; each contains the most recent measurements available at that vintage date. The data were collected by examining various issues of Economic Trends, which is published by the ONS (formally the Central Statistical Office). The figures reported were in the public domain at the end of the month in question. For each vintage, the observations are identical to those in the relevant published source. The window length reported by the source publications is affected by page layout considerations — it varies by variable and by vintage date. Missing data are recorded as empty cells. The two excel files containing the data described below, nomY&Pdef.xls and money.xls, are available from the authors on request. In the following Section, the definition, source, code, period and relevant notes are described for each variable. Nominal GDP (Excel file: nomY&Pdef.xls, Spreadsheet: nominal_mktp(sa)). 
Definition: Gross domestic product at market prices, current price £ Million, seasonally adjusted. 
Source: ONS Economic Trends. 
Code: FNAM (from November 1981 to September 1985), CAOB (from October 1985 to September 1998) and YBHA (from October 1998 onwards). Period: Monthly vintages from November 1981 to August 2002, on quarterly observations 1976Q1 to 2002Q1. GDP price deflator (Excel file: nomY&Pdef.xls, Spreadsheet: deflator_mktp). 
Definition: Implied market price deflator (average estimate). 
Source: ONS Economic Trends. 
Code: DJDT (from November 1981 to October 1998) and YBGB (from October 1998 onwards). 
Period: Monthly vintages from November 1981 to August 2002, on quarterly observations 1976Q1 to 2001Q4. M0 money (Excel file: money.xls, spreadsheet: M0_sa). 
Definition: M0, £ Million, Amount outstanding, seasonally adjusted. 
Source: ONS Economic Trends. 
Code: AVAE. 
Period: Monthly vintages from July 1987 to August 2002, on quarterly observations 1983Q1 to 2002Q1. M4 money (Excel file: money.xls, spreadsheet: M4_sa). 
Definition: M4, £ Million, Amount outstanding, seasonally adjusted. 
Source: ONS Economic Trends. 
Code: AUYN. 
Period: Monthly vintages from July 1987 to August 2002, on quarterly observations 1983Q1 to 2002Q1. VM0 money (Excel file: money.xls, spreadsheet: V(M0)). 
Definition: Velocity of circulation. 
Source: ONS Economic Trends. 
Code: AVAM. 
Period: Monthly vintages from July 1987 to August 2002, on quarterly observations 1983Q1 to 2002Q1. VM4 money (Excel file: money.xls, spreadsheet: V(M4)). 
Definition: Velocity of circulation. 
Source: ONS Economic Trends. 
Code: AUYU. 
Period: Monthly vintages from July 1987 to August 2002, on quarterly observations 1983Q1 to 2002Q1. Footnotes 1 " US data can be downloaded from the Philadelphia Federal Reserve Bank http://www.phil.frb.org/econ/forecast. Croushore and Stark (2001) describe data set construction. 2 " Download from http://www.bankofengland.co.uk/statistics/gdpdatabase. 3 " Annual updates occur in the Spring of each year. 4 " Download from http://www.econ.cam.ac.uk/dae/keepitreal. 5 " The monthly seasonally adjusted monetary aggregates contained in Egginton et al. (2002) were seasonally adjusted on a different basis from the quarterly equivalents for some of the period. 6 " See for example Jansen (1998). Although money velocities can be constructed from the component variables, nominal GDP and the relevant monetary aggregates, we report the official measures for completeness. 7 " The data are available in MS‐Excel format on request from [email protected]. 8 " The ‘noise’ model analysed by Mankiw et al. (1984) has the ‘final’ measurements as the explanatory variable. In this case, the unbiased revisions are orthogonal to final measurements — the data collection agency remeasures with errors in variables. 9 " Contact [email protected]. We set K = 8 (16) for the quarterly (monthly) variables. 10 " We repeated our analysis reported below treating as the final measurements. Although this limits the number of revisions allowed in each case the results were qualitatively similar. Again, the tables can be obtained from the authors on request. 11 " Bai‐Perron Gauss code can be downloaded from http://econ.bu.edu/perron/code.html. Swanson and van Dijk (2006) consider structural breaks in US revisions but restrict attention to just one break. 12 " Bai and Perron (2003b) discuss the appropriate parameter values in small samples. 13 " The Castle‐Ellis data set contains (at times) more than one vintage per quarter. We used the vintage available at the start of each quarter and treated the Garratt‐Vahey variables analogously. 14 " The growth rates for Xt were defined as 100( logeXt −  logeXt−1). 15 " The GDP(E) revisions are comparable in size to those documented by Faust et al. (2005). 16 " The pre and post‐break means were 0.23 (Newey‐West (1987) coefficient standard error 0.091) and 0.81 (0.204) for exports and 0.02 (0.176) and 0.73 (0.219) for imports. 17 " The sample mid‐points defined the break dates for the Garratt‐Vahey and Egginton et al. variables. 18 " Tables for subsequent measurements (up to two years after the initial measurement) can be obtained from the authors on request. Except for the monetary aggregates, the data reject the null hypothesis of unbiasedness for all k at the 1% level. However, the degree but not the direction of bias varies considerably with k. 19 " The Newey‐West (1987) truncation factor was 4; and the serial correlation test was for up to 4th (12th) order for the quarterly (monthly) data. 20 " The sample start date reflects the availability of real‐time GDP price deflator data. 21 " Real‐time (s = 1) GDP growth and GDP price deflator inflation exhibit no breaks (see Section 3). The values for and for GDP growth are 0.444 and −0.573 and for GDP price deflator inflation are 0.696 and −0.595 respectively. 22 " To obtain probability forecasts by stochastic simulation we simulate values of where T runs from 1989Q4 to 1999Q1, the parameter estimates vary with each recursion, the superscript ‘(r)’ refers to the rth replication of the simulation algorithm (r = 1,2, … 1000) and the s are drawn using a non‐parametric method with replacement. Garratt et al. (2003b) label this type of uncertainty as the effects of unobserved future shocks. 23 " We also used the ‘probability integral transform’ (PIT) method, due to Rosenblatt (1952) and discussed in detail by Clements (2004). The two events considered were above‐trend output growth and above‐trend inflation, giving 76 probability forecasts and their associated realisations for the 38 quarters from 1990Q1 to 1999Q2. We calculated the probability of observing values no greater than the actual (final data) values. Under the null hypothesis that the set of density forecasts match the actual data generating density, the PITs are uniformly distributed U[0,1]. The Kolmogorov‐Smirnov statistics indicate marginal rejection for final data but clear rejection with first measurements at the 5% significance level. The bias‐adjusted measurements indicated marginal rejection at the same significance level. References Akritidis , L. ( 2003 ). ‘Revisions to quarterly GDP growth and expenditure components’ , Economic Trends , vol. 601 , pp. 69 – 85 . OpenURL Placeholder Text WorldCat Bai , J. and Perron , P. ( 2003a ). ‘Computation and analysis of multiple structural change models’ , Journal of Applied Econometrics , vol. 18 , pp. 1 – 22 . Google Scholar Crossref Search ADS WorldCat Bai , J. and Perron , P. ( 2003b ). ‘Critical values for multiple structural change tests’ , Econometrics Journal , vol. 6 , pp. 72 – 8 . Google Scholar Crossref Search ADS WorldCat Barklem , A. ( 2000 ). ‘Revisions analysis of initial estimates of key economic indicators and GDP components’ , Economic Trends , vol. 556 , pp. 31 – 52 . OpenURL Placeholder Text WorldCat Castle , J. and Ellis , C. ( 2002 ). ‘Building a real‐time database for GDP(E)’ , Bank of England Quarterly Bulletin , February, pp. 42 – 9 . OpenURL Placeholder Text WorldCat Clements , M.P. ( 2004 ). ‘Evaluating the Bank of England density forecasts of inflation’ . Economic Journal, vol. 114 , pp. 844 – 66 . Croushore , D. and Stark , T. ( 2001 ). ‘A real‐time data set for macroeconomists’ , Journal of Econometrics , vol. 105 , pp. 111 – 30 . Google Scholar Crossref Search ADS WorldCat Diebold , F. X. , and Rudebusch , G. D. ( 1991 ). ‘Forecasting output with the composite leading index: a real‐time analysis’ , Journal of the American Statistical Association , vol. 86 , pp. 603 – 10 . Google Scholar Crossref Search ADS WorldCat Egginton , D.M. , Pick , A. and Vahey , S.P. ( 2002 ). ‘‘‘Keep it real’’ A real‐time UK macro data set’ , Economics Letters , vol. 77 , pp. 15 – 20 . Google Scholar Crossref Search ADS WorldCat Faust , J. , Rogers , J.H. and Wright , J.H. ( 2005 ). ‘News and noise in G7 GDP announcements’ , Journal of Money, Credit and Banking , vol. 37 , pp. 403 – 20 . Google Scholar Crossref Search ADS WorldCat Garratt , A. , Lee , K., Pesaran , M.H. and Shin , Y. ( 2003a ). ‘Forecast uncertainties in macroeconometric modelling: an application to the UK economy’ , Cambridge University Discussion Paper, available at http://www.econ.cam.ac.uk/faculty/pesaran. Garratt , A. , Lee , K., Pesaran , M.H. and Shin , Y. ( 2003b ). ‘Forecast uncertainties in macroeconometric modelling: an application to the UK economy’ , Journal of American Statistical Association, Applications and Case Studies , vol. 98 ( 464 ), pp. 829 – 38 . Google Scholar Crossref Search ADS WorldCat Granger , C.W.J. and Pesaran , M.H. ( 2000 ). ‘Economic and statistical measures of forecast accuracy’ , Journal of Forecasting , vol. 19 , pp. 537 – 60 . Google Scholar Crossref Search ADS WorldCat Howrey , E.P. ( 1978 ). ‘The use of preliminary data in econometric forecasting’ , Review of Economics and Statistics , vol. 60 , pp. 193 – 200 . Google Scholar Crossref Search ADS WorldCat Jansen , N. ( 1998 ). ‘The demand for M0 in the United Kingdom reconsidered: some specification issues’ , Bank of England Working Paper 83. Koenig , E. , Dolmas , S. and Piger , J. ( 2003 ). ‘The use and abuse of ‘‘real‐time” data in economic forecasting’ , Review of Economics and Statistics , vol. 85 , pp. 618 – 28 . Google Scholar Crossref Search ADS WorldCat Mankiw , N.G. , Runkle , D.E and Shapiro , M.D. ( 1984 ). ‘Are preliminary announcements of the money stock rational forecasts’ , Journal of Monetary Economics , vol. 14 , pp. 15 – 27 . Google Scholar Crossref Search ADS WorldCat Mitchell , J. ( 2004 ). Revisions to Economic Statistics , Statistics Commission Report no 17, vols. 1,2 and 3, April. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Newey , W.K. and West , K.D. ( 1987 ). ‘A simple positive semidefinite, heteroskedasticity and autocorrelation consistent covariance matrix’ , Econometrica , vol. 55 , pp. 703 – 8 . Google Scholar Crossref Search ADS WorldCat Patterson , K. D. , and Heravi , S. M. ( 1991 ). ‘Data revisions and the expenditure components of GDP’ , Economic Journal , vol. 101 , pp. 887 – 901 . Google Scholar Crossref Search ADS WorldCat Pesaran , M.H. and Timmermann , A. ( 1992 ). ‘A simple nonparametric test of predictive performance’ , Journal of Business and Economic Statistics , vol. 10 , pp. 461 – 5 . OpenURL Placeholder Text WorldCat Pickford S. ( 1989 ). Government Economic Statistics: A Scrutiny Report , Cabinet Office , HMSO. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rosenblatt , M. ( 1952 ). ‘Remarks on a multivariate transformation’ , Annals of Mathematical Statistics , vol. 23 , pp. 470 – 2 . Google Scholar Crossref Search ADS WorldCat Sargent , T. , ( 1989 ). ‘Two models of measurements and the investment accelerator’ , Journal of Political Economy , vol. 97 , pp. 251 – 87 . Google Scholar Crossref Search ADS WorldCat Swanson , N.R. and Van Dijk , D. ( 2006 ). ‘Are statistical reporting agencies getting it right? Data rationality and business cycle asymmetry’ . Journal of Business and Economic Statistics , forthcoming. OpenURL Placeholder Text WorldCat Symons , P. ( 2001 ). ‘Revisions analysis of initial estimates of annual constant price GDP and its components’ , Economic Trends , vol. 568 , pp. 48 – 65 . OpenURL Placeholder Text WorldCat Topping , S.L. and Bishop , S. L. ( 1989 ), ‘Breaks in monetary series’ . Bank of England Discussion Paper, Technical Series, No. 23. Walsh , C. ( 2003 ). ‘Speed limit policies: the output gap and optimal monetary policy’ , Amercian Economic Review , vol. 93 , pp. 265 – 78 . Google Scholar Crossref Search ADS WorldCat Wroe , D. ( 1993 ). ‘Improving macro‐economic statistics’ , Economic Trends , vol. 471 , pp. 191 – 9 . OpenURL Placeholder Text WorldCat Author notes " We thank Alex Brazier, Dean Croushore, Colin Ellis, George Kapetanios, Simon van Norden, Stephen Machin, Peter Tinsley, Tony Yates and two anonymous referees for helpful comments. We are grateful to seminar participants at the University of Cambridge and the Reserve Bank of New Zealand. Kateryna Rakowsky and Mutita Akusuwan provided excellent research assistance. Financial support from the Department of Applied Economics and from the ESRC (Research Grant No. RES‐000‐22‐1342) is acknowledged gratefully. The views expressed in this paper are those of the authors and do not reflect those of the Reserve Bank of New Zealand. © Reserve Bank of New Zealand 2006.

doi: 10.1111/j.1468-0297.2006.01070.xpmid: N/A

The inclusion of a book or publication in this list does not preclude publication in a future issue of the Journal. Pensions at a Glance: Public Policies Across OECD Countries, Paris: OECD, 2005. Pp. 204. £17.00 paperback, $29.00 paperback. ISBN 92 64 01871 9. Adnett (Nick) and Hardy (Stephen). The European Social Model: Modernisation or Evolution. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xix+244. £55.00 hardback. ISBN 1 84376 125 4. Agarwal (Bina) and Vercelli (Alessandro) (Eds). Psychology, Rationality and Economic Behaviour: Challenging Standard Assumptions. Basingstoke and London: Palgrave, 2005. Pp. xiii+270. £60.00 hardback. ISBN 1 4039 4253 6. Aghion (BeatrizArmendariz deAghion) and Morduch (Jonathan). The Economics of Microfinance. Cambridge, Mass, and London: MIT Press, 2005. Pp. xiv+346. £29.95 hardback. ISBN 0 262 01216 2. Aglietta (Michel) and Reberioux (Antoine). Corporate Governance Adrift: A Critique of Shareholder Value. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xiv+306. £69.95 hardback. ISBN 1 84542 137 X, 1 84542 138 8. Atzeme (Oedzge), Rietveld (Piet) and Shefer (Daniel) (Eds). Regions, Land Consumption and Sustainable Growth: Assessing the Impact of the Public and Private Sectors. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. x+213. £49.95 hardback. ISBN 1 84542 124 8. Banerjee (Debdas). Globalisation, Industrial Restructuring and Labour Standards: Where India Meets the Global. New Delhi: Thousand Oaks/London: Sage Publications, 2005. Pp. 321. £37.50 hardback, £14.99 paperback. ISBN 0 7619 3355 7, 0 7619 3356 5. Beal (Tim). North Korea: The Struggle Against American Power. London: Pluto Press, 2005. Pp. ix+342. £18.99 hardback. ISBN 0 7453 2013 9. Bellofiore (Riccardo) and Taylor (Nicola) (Eds). The Constitution of Capital: Essays on Volume 1 of Marx's Capital. Basingstoke and London: Palgrave, 2004. Pp. ix+315. £60.00 hardback. ISBN 1 4039 0798 6. Bergin (James). Microeconomic Theory: A Concise Course. Oxford: Oxford University Press, 2005. Pp. xxi+356. £30.00 hardback. ISBN 0 19 928029 0. Blair (Roger D.) and Cotter (Thomas F.). Intellectual Property: Economic and Legal Dimensions of Rights and Remedies. Cambridge and New York: Cambridge University Press, 2005. Pp. ix+304. £45.00 hardback, $75.00 hardback, £17.99 paperback, $29.99 paperback. ISBN 0521833167, 0521540674. Bliss (Christopher), Cohen (Avi J.) and Harcourt (G.C.) (Eds). Capital Theory: Volume 1 2 3. Aldershot and Lyme, NH: Edward Elgar, 2005. £395.00 hardback. ISBN 1 84064 481 8. Boeri (Tito), Boca (DanielaDel) and Pssarides (Christlpher) (Eds), et al. Women at Work: An Economic Perspective. Oxford: Oxford University Press, 2005. Pp. xiii+278. £55.00 hardback. ISBN 0 19 928187 4. Borjas (George J.) and Crisp (Jeff) (Eds). Poverty, International Migration and Asylum. Basingstoke: Palgrave Macmillan, 2005. Pp. xix+445. £65.00 hardback. ISBN 1 4039 4365 6. Burns (LawtonRobert) (Ed). The Business of Healthcare Innovation. Cambridge and New York: Cambridge University Press, 2005. Pp. xxiii+373. £50.00 hardback, $90.00 hardback, £24.99 paperback, $43.00 paperback. ISBN 0 521 83898 3, 0 521 54768 7. Buti (Marco) and Franco (Daniele). Fiscal Policy in Economic And Monetary Union: Theory, Evidence and Institutions. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xi+308. £69.95 hardback. ISBN 1 84542 017 9. Butler (Richard J.) and Park (Yong‐seung). Safety Practices, Firm Culture, and Workplace Injuries. Kalamazoo, Michigan: W.E. Upjohn Institute, 2005. Pp. ix+105. $40.00 hardback, $15.00 paperback. ISBN 0 88099 277 8, 0 88099 275 1. Cameron (A. Colin) and Trivedi (Pravin K.). Microeconometrics: Methods and Applications. Cambridge and New York: Cambridge University Press, 2005. Pp. xxii+1034. £45.00 hardback, $75.00 hardback. ISBN 0 521 84805 9. Cao (TianYu) (Ed). The Chinese Model of Modern Development. London and New York: Routledge, 2005. Pp. xi+323. £75.00 hardback. ISBN 0 415 34518 9. Carnot (Nicolas), Koen (Vincent) and Tissot (Bruno). Economic Forecasting. Basingstoke and London: Palgrave, 2005. Pp. xix+315, £22.50 paperback. ISBN 1 4039 3654 4. Cencini (Alvaro). Macroeconomic Foundations of Macroeconomics, London and New York: Routledge, 2005. Pp. xvii+358. £85.00 hardback. ISBN 0 415 31265 5. Chakravarti (Ashok). Aid, Institutions and Development: New Approaches to Growth, Governance and Poverty. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. x+190. £49.95 hardback. ISBN 1 84542 190 6. Collins (Michael) and Baker (Mae) (Eds). Commercial Banks Industrial Finance in England and Wales, 1860–1913. Oxford: Oxford University Press, 2003. Pp. x+296. £53.00 hardback. ISBN 0 19924986 5. Cooper (Joseph) (Ed). Global Agricultural Policy Reform Trade: Environmental Gains and Losses. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xvii+202. £55.00 hardback. ISBN 1 84376 887 9. Copeland (Brian R.) and Taylor (M. Scott). Trade and the Environment: Theory and Evidence. Princeton, NJ: Princeton University Press, 2003. Pp. viii+295. £14.95 hardback. ISBN 0 691 12400 0. Cornia (GiovanniAndrea) (Ed). Inequality, Growth, and Poverty in an Era of Liberalization and Globalization. Oxford: Oxford University Press, 2005. Pp. xxii+438. £27.50 hardback. ISBN 0 19 9284 10 5. Cragg (Wesley) (Ed). Ethics Codes, Corporations and the Challenge of Globalization. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xv+396. £75.00 hardback. ISBN 1 84542 102 7. Crespo (MilagrosGarcia) (Ed). Public Expenditure Control in Europe: Coordinating Audit Functions in the European Union. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xvi+256. £59.95 hardback. ISBN 1 84376 837 2. Cuff (Timothy). The Hidden Cost of Economic Development: The Biological Standard of Living in Antebellum Pennsylvania. Aldershot and Brookfield, VT: Ashgate, 2005. Pp. xvii+277. £55.00 hardback. ISBN 0 7546 4119 8. Dasgupta (Biplab). Globalisation: India's Adjustment Experience. New Delhi: Thousand Oaks/London: Sage Publications, 2005. Pp. 283. £37.50 hardback, £14.99 paperback. ISBN 0 7619 3305 0, 0 7619 3306 9. Davis (Timothy). : Money, Trade Cycles, & Growth. Cambridge and New York: Cambridge University Press, 2005. Pp. xii+316. £45.00 hardback, $75.00 hardback. ISBN 0521844746. Enevoldsen (Martin). The Theory of Environmental Agreements and Taxes: CO2 Policy Performance in Comparative Perspective. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. x+294. £69.95 hardback. ISBN 1 84376 880 1. Epstein (Gerald A.) (Ed). Capital Flight and Capital Controls in Developing Countries. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xiii+343. £69.95 hardback. ISBN 1 84376 931 X. Ermisch (John F.) and Wright (Robert E.) (Eds). Changing Scotland: Evidence from the British Household Panel Survey. Bristol: Policy Press, 2005. Pp. xv+310. £65.00 hardback. ISBN 1 86134 593 3. Ezzamel (Mahmoud) (Ed). Governance, Directors and Boards. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xxv+526.£150.00 hardback. ISBN 1 84542 103 5. Falvey (Rod) and Kreickemeier (Udo) (Eds). Resent Developments in International Trade Theory. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xix+655. £165.00 hardback. ISBN 1 84376 882 8. Feinstein (Charles H.). An Economic History of South Africa: Conquest, Discrimination and Development. Cambridge and Nev York: Cambridge University Press, 2005. Pp. xxiii+302. £45.00 hardback, $75.00 hardback, £19.99 paperback, $34.99 paperback. ISBN 0521850916, 0521616417. Filatotchev (Igor) and Wright (Mike) (Eds). The Life Cycle of Corporate Governance. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. viii+293. £89.95 hardback. ISBN 1 84542 214 7. Flath (David). The Japanese Economy: Second Edition. Oxford: Oxford University Press, 2005. Pp. xviii+371. £25.00 hardback. ISBN 0 19 927861 X. Folmer (Henk) and Tietenberg (Tom) (Eds). The International Yearbook of Environmental and Resource Economics 2005/2006: A Survey of Current Issues. Aldershot and Lyme, NH: Edward Elgar 2005. Pp. x+324. £75.00 hardback. ISBN 1 84542 206 6. Fontaine (Philippe) and Leonard (Robert) (Eds). The Experiment in the History of Economics. London and New York: Routledge, Pp. xii+159. £65.00 hardback. ISBN 0 415 34429 8. Foxall (Gordon R.). Understanding Consumer Choice. Basingstoke and London: Palgrave, 2005. Pp. xvi+262. £60.00 hardback. ISBN 1 4039 1492 3. Frank (Robert H.). What Price the Moral High Ground? Ethical Dilemmas in Competitive Environments. Princeton, NJ: Princeton University Press, 2004. Pp. xii+203. £10.95 paperback. ISBN 0 691 12401 9. Giavazzi (Francesco), Goldfajn (Ilan) and Herrera (Santiago) (Eds). Inflation Targeting, Debt, and the Brazilian Experience 1999 to 2003. Cambridge, Mass, and London: MIT Press, 2005. Pp. xxii+303. £29.95 hardback. ISBN 0 262 07259 9. Giddens (Anthony) and Diamond (Patrick) (Eds). The New Egalitarianism. Bristol: Policy Press, 2005. Pp. ix+254. £45.00 hardback, $55.00 hardback, £14.99 paperback, $22.95 paperback ISBN 0 7456 3430 3, 0 7456 3431 1. Gillespie (Alexander). Whaling Diplomacy: Defining Issues in International Environmental Law. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xxii+509. £95.00 hardback. ISBN 1 84542 107 8. Grauwe (PaulDe). Economics of Monetary Union: Sixth Edition. Oxford: Oxford University Press, 2005. Pp. x+282. £29.99 paperback. ISBN 0 19 927700 1. Green (Christopher J.), Kirkpatrick (Colin H.) and Murinde (Victor) (Eds). Finance and Development: Surveys of Theory, Evidence and Policy. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xiv+450. £95.00 hardback. ISBN 1 84376 030 4. Griffin (James M.) and Puller (Steven L.) (Eds). Electricity Deregulation: Choices and Challenges. Chicago and London. University of Chicago Press, 2005. Pp. vi+446. £45.50 hardback, $65.00 hardback. ISBN 0 226 30856 1. Grosse (Robert) (Ed). International Business and Government Relations in the 21st Century. Cambridge and New York: Cambridge University Press, 2005. Pp. xiii+527. £30.00 hardback. ISBN 0 521 85002 9. Gui (Benedetto) and Sugden (Robert) (Eds). Economics and Social Interaction: Accounting for Interpersonal Relations. Cambridge and New York: Cambridge University Press, 2005. Pp. xv+299. £45.00 hardback. ISBN 0 521 84884 9. Haan (JakobDe), Eijffinger (Sylvester C. W.) and Waller (Sandra). The European Central Bank: Credibility, Transparency and Centralization. Cambridge, Mass, and London: MIT Press, Pp. vi+264. £22.95 hardback. ISBN 0 262 04226 6. Hansjurgens (Bernd) (Ed). Emissions Trading for Climate Policy: US and European Perspectives. Cambridge and New York: Cambridge University Press, 2005. Pp. xiii+245. £48.00 hardback, $85.00 hardback. ISBN 0 521 84872 5. Harvie (Charles) and Lee (Boon‐Chye) (Eds). Sustaining Growth and Performance in East Asia: The Role of Small and Medium Sized Enterprises. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xix+372. £85.00 hardback. ISBN 1 84064 808 2. Head (Simon) (Ed). The New Ruthless Economy: Work and Power in the Digital Age. Oxford: Oxford University Press, 2005. Pp. xviii+222. £8.50 paperback. ISBN 0 19 517983 8. Held (David) (Ed). Debating Globalization. Oxford: Polity Press, 2005. Pp. xii+205. £45.00 hardback, $49.95 hardback, £12.99 paperback, $19.95 paperback. ISBN 0 7456 3524 5, 0 7456 3525 3. Henson (Spencer) and Wilson (John S.) (Eds). The WTO and Technical Barriers to Trade. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xxv+531. £135.00 hardback. ISBN 1 84542 049 7. Ito (Takatoshi) and Rose (Andrew K.) (Eds). International. Trade in East Asia: NBER‐East Asia Seminar on Economics, volume 14. Chicago and London: University of Chicago Press, 2005. Pp. ix+419. £56.00 hardback, $80.00 hardback. ISBN 0 226 37896 9. Iversen (Torben). Capitalism, Democracy and Welfare. Cambridge and New York: Cambridge University Press, 2005. Pp. xvii+315. £40.00 hardback. ISBN 0 521 61307 8. Jackson (John E.), Klich (Jacek) and Poznznska (Krystyna). The Political Economy of Poland's Transition: New Firms and Reform Governments. Cambridge and New York: Cambridge University Press, 2005. Pp. xvi+280. £45.00 hardback, $80.00 hardback. ISBN 0 521 83895 9. Jones (Chris). Applied Welfare Economics. Oxford: Oxford University Press, 2005. Pp. xxiv+308. £35.00 hardback. ISBN 0 19 928197 1. Kaplinsky (Raphael). Globalization, Poverty and Inequality: Between a Rock and a Hard Place. Oxford: Polity Press, 2005. Pp. xix+280. £50.00 hardback, $64.95 hardback, £14.99 paperback $26.95 paperback. ISBN 0 7456 3554 7. Kleit (Andrew N.) (Ed). Antitrust and Competition Policy. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xix+641. £160.00 hardback. ISBN 1 84376 319 2. Koenker (Roger). Quantile Regression. Cambridge and New York: Cambridge University Press, 2005. Pp. xv+349. £19.99 paperback $34.99 paperback. ISBN 0 521 60827 9. Lesbirel (S. Hayden) and Shaw (Daigee) (Eds). Managing Conflict in Facility Siting: An International Comparison. Aldershot Lyme, NH: Edward Elgar, 2005. Pp. xii+220. £59.95 hardback. ISBN 1 84376 523 3. Lofgren (Orvar) and Willim (Robert) (Eds). Magic, Culture and the New Economy. New York and Oxford: Berg, 2005. Pp. xi+145. £50.00 hardback, $89.95 hardback. ISBN 184520 090 X. Martin (Lisa L.) (Ed). International Institutions in the New Global Economy. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xv+616. £150.00 hardback. ISBN 1 84376 425 3. Mavroidis (Petros C.) and Sykes (Alan O.) (Eds). The WTO and International Trade Law/Dispute Settlement. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xiv+697. £160.00 hardback. ISBN 1 84376 343 5. Mclean (Lain). The Fiscal Crisis of the United Kingdom. Basingstoke and London: Palgrave, 2005. Pp. xxi+250. £55.00 hardback. ISBN 1 4039 0366 2. Mehrling (Perry). Fisher Black and the Revolutionary Idea of Finance. New York and Chichester: John Wiley & Sons Ltd., 2005. Pp. xv+374. £19.99 hardback. ISBN 0 471 45732 9. Meier (Peter) and Munasinghe (Mohan). Sustainable Energy in Developing Countries: Policy Analysis and Case Studies. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xx+283. £69.95 hardback. ISBN 1 84376 753 8. Milanovic (Branko). Worlds Apart: Measuring International And Global Inequality. Princeton, NJ: Princeton University Press, 2005. Pp. ix+227. £18.95 hardback. ISBN 0 691 12110 9. Millward (Robert). Private and Public Enterprise in Europe: Energy, Telecommunications and Transport, 1830–1990. Cambridge and New York: Cambridge University Press, 2005. Pp. xix+351. £50.00 hardback, $90.00 hardback. Moore (Don A.), Cain (Daylian M.) and Loewenstein (George) (Eds), et al. Conflicts of Interest: Challenges and Solutions in Business, Law, Medicine, and Public Policy. Cambridge and New York: Cambridge University Press, 2005. Pp. xi+300. £40.00 hardback, $65.00 hardback. ISBN 0 521 84439 8. Moosa (Imad A.). Exchange Rate Regimes: Fixed, Flexible Something In Between? Basingstoke and London: Palgrave, 2005. Pp. xvii+267. £60.00 hardback. ISBN 1 4039 3672 2. O’kane (Rosemary H.T.) (Ed). Terrorism: Volume 1 and 2. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. 984. £250.00 hardback. ISBN 1 84376 064 9. Ozawa (Terutomo). Institutions, Industrial Upgrading, and Economic Performance in Japan: The ‘Flying Geese’ Paradigm of Catch‐up Growth. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xvii+234. £59.95 hardback. ISBN 1 84376 959 x. Paula (Silvana De) and Dymski (Gary A.) (Eds). Reimaging Growth: Towards a Renewal of Development Theory. London and New Jersey: Zed Books, 2005. Pp. xii+308. £70.00 hardback, $85.00 hardback, £19.95 paperback, $27.50 paperback. ISBN 1 84277 584 7, 1 84277 585 5. Philp (Bruce) (Ed). Reduction, Rationality and Game Theory in Marxian Economics. London and New York: Routledge, 2005. Pp. xiii+143. £60.00 hardback. ISBN 0 415 28765 0. Rebick (Marcus). The Japanese Employment System: Adapting to a New Economic Environment. Oxford: Oxford University Press, 2005. Pp. xvii+196. £45.00 hardback. ISBN 0 19 924724 2. Redclift (Michael) and Woodgate (Graham) (Eds). New Developments in Environmental Sociology. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xxvi+645. £160.00 hardback. ISBN 1 84376 115 7. Roberts (Karen), JR. (John F. Burton) and Bodah (Matthew) (Eds) Workplace Injuries and Diseases: Prevention and Compensation. Michigan: W.E. Upjohn Institute for Employment Research, 2005. Pp. xv+301. $20.00 paperback. ISBN 0 88099 324 3. Ross (Don). Economic Theory and Cognitive Science: Microexplanation. Cambridge, Mass, and London: MIT Press, 2005, Pp. x+444. £29.95 hardback. ISBN 0 262 18246 7. Salanie (Bernard). The Economics of Contracts: A Primer, Second Edition. Cambridge, Mass. and London: MIT Press, 2005. Pp. viii+244. £22.95 hardback. ISBN 0 262 19525 9. Sampson (Gary) and Whalley (John) (Eds). The WTO, Trade and the Environment. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. XXV+701. £170.00 hardback. ISBN 1 84376 839 9. Sampson (Gary P.) (Ed). The WTO and Sustainable Development. The Netherlands: United Nations University, 2005. Pp. xiii+315. $45.00 paperback. ISBN 92 808 1115 0. Seabright (Paul). The Company of Strangers: A Natural History of Economic Life. Princeton, NJ: Princeton University Press, 2005. Pp. x+304. £19.95 hardback, £12.50 paperback. ISBN 0 691 11821 3, 0 691 12452 3. Shionoya (Yuichi). Economy and Morality: The Philosophy of the Welfare State. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xi+355. £79.95 hardback. ISBN 1 85898 480 7. Sinha (Aseema). The Regional Roots of Developmental Politics in India: A Divided Leviathan. Bloomington and Indianapolis: Indiana University Press, 2005. Pp. xxiii+356. $64.95 hardback, $27.95 paperback. ISBN 0 253 34404 2, 0 253 21681 8. Sullivan (Rory). Rethinking Voluntary Approaches Environmental Policy. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. ix+216. £55.00 hardback. ISBN 1 84542 210 4. Takeyama (Lisa N.), Gordon (Wendy J.) and Towse (Ruth) (Eds). Developments in the Economics of Copyright: Research and Analysis. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xix+197. £55.00 hardback. ISBN 1 84376 930 1. Taylor (Stephen J.). Asset Price Dynamics, Volatility and Prediction. Princeton, NJ: Princeton University Press, 2005. Pp. xv+525. £41.95 hardback. ISBN 0 691 11537 0. Teranishi (Juro). Evolution of the Economic System in Japan. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xiv+343. £69.95 hardback. ISBN 1 84376 163 7. Thaler (Richard H.) (Ed). Advances in Behavioral Finance: Volume II. Princeton: Princeton University Press, 2005. Pp. xxi+712. £48.95 hardback, £29.95 paperback. ISBN 0 691 12174 5, 0 691 12175 3. Thompson (Andrew). The Empire Strikes Back? The Impact of Imperialism on Britain From the Mid‐Nineteenth Century. Essex: Pearson Education Limited, 2005. Pp. xvii+374. £19.99 paperback ISBN 0 582 43829 2. Uzan (Marc) (Ed). The Future of the International Monetary System. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. ix+365. £75.00 hardback. ISBN 1 84376 805 4. Uzawa (Hirofumi). Economic Analysis of Social Common Capital. Cambridge and New York: Cambridge University Press, 2005. Pp. ix+406. £45.00 hardback, $80.00 hardback. ISBN 0 521 84788 5. Vatn (Arild). Institutions and the Environment. Lyme, NH: Edward Elgar, 2005. Pp. xiv+481. £79.95 hardback. ISBN 1 84376 100 9. Vogel (David). The Market for Virtue: The Potential and Limits of Corporate Social Responsibility. Washington, DC: Brookings Institution Press, 2005. Pp. xii+222. £17.99 hardback. ISBN 0 8157 9076 7. DeVries (Manfred F.R. Kets). Lessons on Leadership by Terror: Finding Shaka Zulu in the Attic. Aldershot and Lyme, NH: Edward Elgar, 2004. Pp. xvii+197. £12.95 paperback. ISBN 1 84542 368 2. Weiss (John) (Ed). Poverty Targeting in Asia. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xvii+282. £65.00 hardback. ISBN 1 84542 123 X. Whitelaw (Diana M.) and Visgilio (Gerald R.) (Eds). America's Changing Coasts: Private Rights and Public Trust. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. xviii+248. £59.95 hardback. ISBN 1 84542 080 2. Whitten (Stuart M.) and Bennett (Jeff). Managing Wetlands for Private and Social Good: Theory, Policy and Cases From Australia. Aldershot and Lyme, NH: Edward Elgar, 2005. Pp. x+286. 59.95 hardback. ISBN 1 84064 898 8. Wise (David A.) (Ed). Analyses in the Economics of Aging. Chicago and London: University of Chicago Press, 2005. Pp. ix+424. £59.50 hardback, $85.00 hardback. ISBN 0 226 90286 2. Wood (John H.). A History of Central Banking in Great Britain and the United States. Cambridge and New York: Cambridge University Press, 2005. Pp. xv+439. £50.00 hardback, $90.00 hardback. ISBN 0 521 85013 4. © Royal Economic Society 2006.

Liu,, Xiaoyu

doi: 10.1111/j.1468-0297.2006.01069_5.xpmid: N/A

China has, since 1978, experienced a host of reforms. In 2001, she entered the WTO to begin her comprehensive integration into the global market. In order to understand these reforms and the impact of globalisation theoretically, and evaluate them practically a group of scholars and Chinese officials gathered in Hangzhou, China, for a symposium on ‘The Chinese Model of Modern Development’. This book is the outcome. It explores, in depth, the goals of China's reforms, ways to achieve them, the most urgent issues facing China today, and how these should be addressed. China's economic and cultural fabric suffered considerable destruction between 1957 and 1977. After the ‘Great Leap Forward’ and the ‘Cultural Revolution’, the situation was extreme enough finally for people to be aroused. There was a re‐evaluation of the theory and practice of the Mao Zedong era. Concurrently, the Chinese Communist Party realised the importance of both strengthening democracy within the Party and improving the legal system, and switched focus from ideological and political struggle to the economy. The slogan of China's reform was the ‘four modernisations’, and, along with political reform the Party aimed, to build a highly civilised, highly democratised, modernised and powerful socialist nation. Deng Xiaoping called the reform the self‐improvement, rather than the abandonment, of socialism. He emphasised the need for a market economy and openings to the West, as part of a process of building socialism with Chinese characteristics. While twenty‐three years of economic growth and loosening of political control are indisputable facts, exploitation, social injustice, corruption and inequality have also developed rapidly. Under present difficulties, people are uncertain whether they are closer to or further away from socialism, and there is no consensus about how successful the reforms have been. The future direction of reform and orientation of policy are subject to ongoing debate. The book consists of four parts. Part 1 calls for conceptual and institutional innovations that are relevant to China today. It is claimed that, after twenty years focused on building the economy, technological innovation and accelerated development, it is time to think about the whole system. Perry Anderson's essay constructs four alternative models, namely ‘market socialism’, ‘socialism with Chinese characteristics’, ‘clean‐path capitalism’ and ‘actually existing capitalism’, and analyses them. Part 2 discusses the historical experience of China's reforms. The essay of Yu Guangyuan places China's reforms in an historical and global context, explaining clearly why the reforms were implemented and providing a personal evaluation. The focus is on some major decisions by the Party at the beginning of the reforms. The essay of Wang Hui looks at the economic reforms that took place between 1978 and 1989, and describes the underlying causes of the social upheaval of 1989. Qin Hui's essay is concerned mainly with the ‘reform of state enterprises’, comparing this reform's success and failure with the experience of privatisation in Eastern European countries and Russia. Part 3 examines the theoretical issues crucial to an understanding of the reforms. Topics included are: the relationship between the state and society, liberal socialism and the future of China, and some basic ideas about Marxism. Part 4 is concerned with China's position in global integration. There are some who advocate nationalism and oppose globalisation; others who bow to the United States and incline towards cosmopolitanism. What is sure is globalisation is a reality that must be faced, not a process to be bypassed. In this part, the impact of globalization on China's economic and political development is discussed, and China's comparative advantage analysed. The study of globalisation trends, and the impact of globalisation on other underdeveloped countries, fosters an awareness that China can continue to develop her own productivity, and promote her own interests, while participating in the international division of labour. Reforms, modernisation, globalisation and ideology are hot topics in China. In the book, they are discussed by writers of different backgrounds from their different perspectives. The authors divide into two groups. They are either designers of, or active participants in the reforms, or else western scholars whose ideas have had a large impact on China's intellectuals. The authors discuss the various issues both objectively and fearlessly, and sensitive topics, such as the reform of the political system, are not shirked. In my opinion, this is a book to be read by all those interested in China's reforms. © Royal Economic Society 2006.

Volz,, Ulrich

doi: 10.1111/j.1468-0297.2006.01069_8.xpmid: N/A

This book is the essence of Ronald McKinnon's recent work on the exchange rate policies pursued by the East Asian countries. It contains eight essays, of which three are written with his co‐author, Gunther Schnabl of Tiibingen University. Three other essays are co‐authored with Rishi Goyal of the IMF, Kenichi Ohno of the National Graduate Institute of Policy Studies in Tokyo, and Huw Pill of the Harvard Business School, respectively. McKinnon, the William D. Eberle Professor of International Economics at Stanford University and one of the main contributors to the theory of optimum currency areas, has been a leading figure in international economics for more than forty years. This book is yet another proof of his mastery. McKinnon argues forcefully and convincingly that the conventional wisdom that East Asian countries should adopt flexible exchange rates might be an ill advice. Instead, he makes a case that East Asian countries should co‐ordinate their policies to keep exchange rates stable against the dollar. McKinnon demonstrates that the East Asian dollar pegs are entirely rational from the individual country's perspective. He shows that the motivation for dollar pegging does not primarily arise because of strong trade ties between East Asia and the US. Instead, he highlights three other reasons for exchange rate stability vis‐à‐vis the dollar. Firstly, for East Asia – a natural trading region for which intra‐regional trade accounts for more than fifty per cent of overall trade – exchange rate spillover effects from one country to another are of great importance. This is even more so, because East Asian countries also compete against one another in third markets. Hence, McKinnon writes, mutual exchange rate stability is the ‘quintessential public good’. Because of neighbourhood effects, national decisions to fix or float exchange rates should not be made independently. McKinnon's second reason for maintaining the dollar standard builds on what has become known as ‘original sin’ (Eichengreen and Hausmann 2004). Because countries cannot borrow in their own currencies, be it internationally or domestically, exporters and importers have problems hedging against exchange rate fluctuations. Hence, governments have an incentive to provide an informal forward hedge for short‐term transactions by keeping the exchange rate stable. As most East Asian countries have turned into creditor countries after the Asian crisis, McKinnon's third argument is of particular importance: The excess build‐up of foreign exchange assets has created a problem which McKinnon calls ‘conflicted virtue’. Conflicted virtue describes a country's inability to lend in its own currency, forcing creditor countries to cumulate a currency mismatch. While Japan, Singapore and Taiwan have had current account surpluses for more then two decades, and China has had more modest current account surpluses since 1995, even the five former crisis economies – Indonesia, Korea, Malaysia, Philippines and Thailand, which had large current account deficits before 1997 – have now accumulated large stocks of liquid dollar assets in both private and official portfolios. With mounting dollar claims, domestic holders of dollar assets have to worry more about a run into the domestic currency, which would cause an appreciation and hence a decline of their net wealth. Countries are hence inclined to avoid large‐scale appreciation of their currencies, which might invoke protests from deficit countries about unfair competition through an undervalued currency. Because of financial fragility in debtor economies with original sin and in creditor countries with conflicted virtue, as well as the importance of exchange rate stability within the region, McKinnon views continued dollar pegging as the most sensible policy option for East Asian countries. He points out that using the dollar as the key currency for stabilising relative exchange rates within East Asia is tenable as long as the Federal Reserve maintains the international purchasing power of the dollar. Until the East Asian countries opt for the creation of a common currency (an ‘Asian euro’, which McKinnon thinks will not materialise until ‘the very distant future’), which could float freely against the dollar, the East Asian countries should co‐operate to keep their exchange rates stable against the dollar, and thus stable against one another. While one may not necessarily agree with McKinnon's policy conclusion, i.e. that East Asian countries should focus their co‐operation efforts on maintaining their dollar anchors, one will agree that this book provides an excellent analysis of the dilemmas East Asian central banks face. In times where floating exchange rates are hailed as the panacea to global imbalances, this book offers some stimulating and controversial thoughts. No one studying or working on East Asian exchange rate policies could afford to ignore this book. Reference Eichengreen , B. and Hausmann , R., eds. ( 2004 ). Other People's Money: Debt Denomination and Financial Instability in Emerging Market Economies , Chicago: University of Chicago Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC © Royal Economic Society 2006.