Wayne Sandholtz and William Koetzle University of California, Irvine Though corruption poses fundamental challenges to both democratic governance and market economies, political science research has only recently begun to address corruption in a comparative context. In this article we explain variation in the perceived level of corruption ~defined as the misuse of public office for private gain! across fifty countries. We propose a set of hypotheses that explain variation in corruption levels in terms of domestic political-economic structure, democratic norms, inte- gration into the international economy, and Protestant religious affili- ation. Levels of corruption, we propose, are higher: ~1! the lower the average income level, ~2! the greater the extent of state control of the economy, ~3! the weaker are democratic norms and institutions, ~4! the lower the degree of integration in the world economy, and ~5! the lower the share of the population with Protestant religious affiliation. The data analysis broadly confirms our predictions: in the multivariate regres- sion, each of the independent variables is significant in the direction we expect. Corruption, like the poor, will probably always be with us. We can read about it in Cicero, Augustine, and almost daily in the news. In just the past two years a striking number and variety of countries have been shaken by major corruption scandals: France, Italy, Japan, Kenya, Mexico, South Korea, and the United States. The problem of corruption has entered the agendas of major international bodies, including the International Monetary Fund, the World Trade Organiza- tion, and the Organization for Economic Cooperation and Development. 1 Though corruption creates problems for any political system, it directly challenges some of the fundamental principles of democratic governance. Corrupt behaviors take place in secret and provide privileged access to government for some parties ~for example, those who bribe!. Corruption thus violates norms of openness and equality that would seem to be central to democracy. It also contravenes com- mon notions of modern bureaucracythat public agencies should operate not in a personalistic but in an impartial, rule-based fashion. Finally, the phenom- enon of corruption leads into an enduring puzzle in political studies, namely, Authors note: We are grateful to Global Peace and Conf lict Studies at the University of California, Irvine for a grant that supported this research. For their helpful comments on earlier drafts, we thank Russell Dalton, Bernie Grofman, Rein Taagepera, Daniel Treisman, and Sharon Welden. We are also indebted to the late Harry Eckstein, our friend, colleague, and mentor, who with his keen insight and generous spirit encouraged us in this and many other intellectual enterprises. 1 See Washington Post , August 8, 1997, p. A25; Wall Street Journal , September 13, 1996, p. A5; Wall Street Journal , December 18, 1997, p. A16. International Studies Quarterly ~2000! 44, 3150. 2000 International Studies Association. Published by Blackwell Publishers, 350 Main Street, Malden, MA 02148, USA, and 108 Cowley Road, Oxford OX4 1JF, UK. the ubiquitous gap between formal institutions and informal cultures. Cultures tolerate corruption in varying degrees despite nearly universal laws designed to eliminate corrupt practices. Political science research on corruption has f lourished in recent years. The steady procession of major scandals around the globe seems to have provoked a renewal of scholarly interest, much of it focused on corruption problems surfac- ing in one or a few countries. Looking at research on political corruption in the last few decades, we notice that it tends to address a relatively small number of themes: how a political system has dealt with corruption; whether corruption helps or hinders economic development, 2 or how to design public agencies so as to minimize corrupt practices. 3 But we would find few answers to broad com- parative questions, even fundamental ones like: Does the extent of corruption vary across countries? Why does it vary? What is the relationship, if any, between democracy and corruption levels? In this article we offer some first steps toward the comparative analysis of corruption, defined as the misuse of public office for private gain. 4 We propose a set of hypotheses that explain variation in corruption levels in terms of domes- tic economic structures, democracy, involvement in international trade, and the extent of Protestant religious orientation. The study tests these hypotheses using recently compiled cross-national indicators of perceived levels of corruption. In general, the data confirm our propositions. Levels of corruption are higher: ~1! the lower the average income, ~2! the greater the extent of state control of the economy, ~3! the weaker are democratic norms and practices, ~4! the lower the degree of integration in the world economy, and ~5! the smaller the share of the population with a Protestant affiliation. The first section of the paper brief ly reviews existing research on corruption. The second section offers a theoretical framework and derives hypotheses. The third section defines the variables, explains our construction of indicators, and provides descriptive data. The fourth section reports the data analysis. In the conclusion we explore some of the implications for further scholarship. What We Know Though recent years have generated a great deal of political science research on corruption, there is also a substantial body of studies produced mainly in the 1960s and 1970s. Much of the existing research, old and new, tackles issues that remain central to any analysis of corruption. We brief ly discuss two themes that 2 The initial presumption was that corruption inevitably retarded economic and political development. In response, some scholars argued that corruption could actually promote development. Later analysts attempted to specify both the costs and benefits of corruption ~see Wraith and Simpkins, 1963; Huntington, 1968:68; Myrdal, 1968:93758; Scott, 1972; Rose-Ackerman, 1978; Klitgaard, 1987; Bayley, 1989; Johnston, 1989; Leff, 1989; Leys, 1989; Nye, 1989; Theobald, 1990!. 3 Analysts in the political-economic ~sometimes also called public choice or rational choice! mode have in particular concentrated on designing anti-corruption measures. Rose-Ackermans Corruption: A Study in Political Economy ~1978! was an impressively systematic and comprehensive initial elaboration of this approach. Klitgaard ~1987! similarly uses a political-economy model to suggest ways of designing and carrying out successful anti- corruption policies. Other studies, not necessarily in the political-economy vein, explore general principles for reducing corruption ~Gardiner and Lyman, 1989! or assess empirical instances of anti-corruption reforms ~Palmier, 1985; Quah, 1989; von Aleman, 1989; Gillespie and Okruhlik, 1991!. 4 As is apparent from this definition, we are concerned with corruption in the public sphere. Private sector corruption clearly exists and is worthy of study, but it is a different phenomenon. Private corruption presumably occurs when people misuse their organizational position ~in a firm, for example! for personal gain; that is, they violate the norms of the private organization in which they operate. Public corruption, in contrast, involves misuse of public position and raises a distinct set of issues, namely, the more directly political ones of governance, political authority, legitimacy, statesociety relations, and, as we argue, democracy. It is thus reasonable, and common practice in the literature, to address public corruption as a distinct problem, leaving private corruption as a separate area of inquiry ~see, e.g., DeLeon, 1993:6!. 32 Accounting for Corruption have direct import for the present study: the problem of comparative data, and the problem of defining corruption. Case Studies and the Data Problem With a few notable exceptions, the majority of empirical research on corruption has consisted of single-country case studies ~as Manzetti and Blake also note @1996:663# !. For instance, the Heidenheimer, Johnston, and LeVine collection includes twenty-five case studies analyzing a single country or locality, and the bibliography lists dozens more. The volume includes three papers that address two or three cases, and one that analyzes a larger number of Latin American presidents ~Heidenheimer, Johnston, and LeVine, 1989b!. Other volumes simi- larly offer a variety of single-country cases ~Markovits and Silverstein, 1988; Levi and Nelken, 1996; Little and Posada-Carb, 1996; Heywood, 1997!. There are excellent book-length treatments of corruption in France ~Mny, 1992!, Ghana ~LeVine, 1975!, and Mexico ~Morris, 1991!, as well as in larger regions such as Asia ~Palmier, 1985! or Africa ~Williams, 1987!. But few studies are both compar- ative and empirical. Manzetti and Blake ~1996! assess the effects of market reforms on corruption in three Latin American countries. Gillespie and Okruhlik ~1991! propose an analytical framework for studying the politics of corruption cleanups, and provide an analysis of twenty-five Middle Eastern and North African coun- tries. As early as 1972 Scott explicitly confronted the challenges involved in the comparative analysis of corruption, and proposed a number of reasonable solu- tions ~Scott, 1972!. Some recent studies have begun to address those challenges in a systematic way, notably those by Treisman ~1998! and Lipset and Lenz ~1999!. A daunting obstacle to comparative studies of corruption has been the lack of usable data on corrupt activities. Williams even suggests that the lack of hard numbers regarding corruption may explain the relative neglect of the subject by American political scientists. But he also points to the problem of data for comparative analysis: It seems, therefore, almost inevitable that the evidence used by students of corruption is bound to be fragmentary, biased, anecdotal, potentially misleading, impressionistic and inadequate, and thus incapable of sustaining general ~comparative! theory ~Williams, 1987:2728!. The reasons for the lack of numerical data on corrupt practices are not hard to imagine: corrupt actions take place in secret and are generally meant to remain secret; even the victims of corruption frequently are unaware that they have been victimized; those reporting or alleging instances of corruption can be political opponents of the accused with motives to discredit them; critics of corrupt practices often have a separate agenda to extol or denigrate specific groups; and governments may not want researchers probing such sensitive areas ~Noonan, 1984:xii; Klitgaard, 1987:29; Williams, 1987:7, 9!. As a consequence of the lack of cross-national data on corruption, scholars have been unable to say much about relative levels of corruption across coun- tries, or to explain levels of corruption in terms of broad political or economic differences. Along with other recent efforts, our study begins to fill a sizable gap in the research. Defining Corruption Virtually every published work on corruption, from the 1960s to the present, wrestles with the problem of defining it. The ability to articulate broad ~if not universal! propositions that can be checked against empirical evidence requires a definition of corruption that is portable across cultures. But the brute fact is that practices that one society condemns as corrupt are considered harmless or Wayne Sandholtz and William Koetzle 33 even appropriate in another cultural context. The effort to establish general principles thus confronts the inevitable particularity of corruption norms. Corrupt acts are, in every definition, improper or illegitimate. The problem for social science consists of giving the term improper specific content. By what standard would the researcher know if a particular interaction is corrupt? The political science literature offers three main approaches, based, respectively, on the public interest, public opinion, and legal norms ~see Scott, 1972:3!. The first approach identifies improper behavior on the part of political or adminis- trative officials as that which goes against the public interest ~Rogow and Lass- well, 1963:132; Friedrich, 1966:74! That is, public officials sacrifice the general interest in order to favor specific groups, in return for private rewards. But the public-interest definition suffers from irremediable vagueness, as there is no way to identify an objective public interest ~Scott, 1972:3; Theobald, 1990:5!. Almost anything a government does is contrary to somebodys definition of the public interest. Public opinionbased approaches argue that since corruption standards vary, corruption is what the public thinks it is ~Gibbons, 1989:169!. But is the relevant public the political elite, the politically mobilized citizenry, or the whole popu- lation ~Scott, 1972:4; Theobald, 1990:6!? Furthermore, opinion regarding proper conduct in public office varies not just across groups, but also across geographic regions within a country and across time ~Williams, 1987:19!. 5 The formal-legal approach identifies corrupt acts as those that violate specific rules governing the way public duties should be performed ~Williams, 1987:15!, as illegal exchanges of political goods for private rewards ~Manzetti and Blake, 1996:665; see also Williams, 1987:1516; Heidenheimer, Johnston, and LeVine, 1989a:89!. The appeal of such a definition is that it offers, for any given society, a fairly clear standard ~despite the inevitable ambiguities of the law; see Lowen- stein, 1989!. But formal-legal approaches suffer from some of the same limita- tions as the other two, for the laws vary from state to state and also change over time, frequently in response to the manipulations of powerful political actors ~Williams, 1987:18!. Legal definitions also exclude acts that may not be illegal but are widely considered improper ~Moodie, 1989:876; Theobald, 1990:17!. Bribery, for example, is an inherently legal concept, but the formal law can differ from the law as enforced, which can in turn diverge from the moral sense of the community ~Noonan, 1984:xii!. The disputes and difficulties involved in defining corruption sometimes obscure the reasonably stable, general definition of corruption that in fact exists. The real problem is how to operationalize it across different societies. At the heart of this core definition of corruption are three elements. First is a distinction between the public sphere and the private sphere ~see LeVine, 1975:2; Palmier, 1985:1; Theobald, 1990:2!. Nyes widely cited definition highlights the public-private divide: corruption is behavior which deviates from the formal duties of a public role because of private-regarding ~personal, close family, private clique! pecuni- ary or status gains ~Nye, 1989:966!. Corruption, then, involves a breakdown, or blurring, in the public0private distinction. Second is a recognition that corrupt acts involve an exchange, in which one party offers inducements ~frequently but not necessarily monetary! to a public 5 Heidenheimer acknowledged the inevitable variation in beliefs within a society and proposed a spectrum of corruption based on the extent of social consensus regarding the acts in question. Black corruption refers to behaviors that virtually all members of society would condemn; white corruption designates practices that most of society would consider acceptable; and gray corruption is the category for actions about which public opinion is divided ~Heidenheimer, 1989:16063!. Unfortunately, the typology is more a re-labeling of the problem of fragmented opinion than a solution. 34 Accounting for Corruption official in return for special policy or administrative advantages, or political goods ~LeVine, 1975:6; see also Manzetti and Blake, 1996:665!. Some analysts have made the exchange dimension the core of a market-centered ~Heiden- heimer, Johnston, and LeVine, 1989a:9; van Klaveren, 1989:26! or political eco- nomic approach ~Rose-Ackerman, 1978!. The third element in the core definition of corruption is the sense that such exchanges are improper, that is, they violate established norms. Thus corruption is behavior by public officials that deviates from the norm actually prevalent or believed to prevail ~Friedrich, 1989:15! or from accepted norms ~Huntington, 1968:59!; or it is political conduct contrary to political norms ~Morris, 1991:2!. With these three elements, then, we have a usable general definition of corruption: the improper use of public office in exchange for private gain. A Comparative Approach to Corruption Much scholarship on corruption has been bedeviled by the apparent dilemma that general definitions founder on the cultural specificity of norms of corrup- tion. That is, researchers seeking to observe or measure levels of corruption quickly discover that what counts as improper in one society does not in another. The only response that permits comparative research is to specify in advance whose definition of improper the analysis will employ. The research would then take a set of practices defined by a specified social group as corrupt and assess comparatively their incidence in diverse countries. We adopt such a strategy in this article. We do not, therefore, pretend to measure corruption in each society in its own terms. Rather, we offer a definition of corruption and employ data that assess various societies in terms of that definition. This is a common social science device: corruption in this study is an analytical category whose utility is that it permits comparative analysis. An earlier generation of researchers who wrestled with the comparative analy- sis of corruption came to the same conclusion. Bayley, for example, argues that the best solution is to employ broadly shared Western standards regarding spe- cific practices, like bribery, nepotism, and misappropriation. Otherwise the West- ern researcher who chooses the culturally relevant definition. . . will either end by abandoning the term altogether or will find it necessary to define it pecu- liarly, perhaps differently, for every non-Western country studied. Recognizing the diversity of social contexts, the analyst can then also note that some practices classified as corrupt for research purposes do not, in some societies, elicit social or legal condemnation ~Bayley, 1989:938!. Nye notes that this practice has the merit of denoting specific behavior generally called corrupt by Western stan- dards @including bribery, nepotism, and misappropriation#, thus permitting com- parisons ~Nye, 1989:96667!. The distortion inherent in applying Western standards of corrupt behavior to non-Western societies is probably in fact minimal. As Bayley notes, the intelli- gentsia and higher-level civil service in most developing countries are familiar with the Western label corruption, and they apply it to their own countries ~1989:93839!. Nye similarly writes that Western standards are at least partly relevant in most developing countries ~1989:967; see also Klitgaard, 1987:3!. And Scott argues that inasmuch as the legal standards of public conduct in less developed countries are substantially the same as those in use in the westa colonial heritagethe terms of comparison are roughly similar ~Scott, 1972:8!. Indeed, in studies of anti-corruption campaigns in non-Western societies, the practices targeted for elimination are precisely those proscribed by Western norms ~bribery, extortion, misappropriation; see, for example, Palmier, 1985; Quah, 1989!. Wayne Sandholtz and William Koetzle 35 To summarize, for purposes of empirical analysis, corruption in this study refers to practices that are considered improper under Western norms. We focus on certain kinds of administrative corruption, not on campaigns and elections. The dependent variable in this study is perceived levels of administrative cor- ruption, where the perceptions are those of transnational businesspersons and consultants as well as local actors. We describe the data sources more fully in a subsequent section. Predicting Corruption We posit two key dimensions in explaining the level of corrupt behaviors. The first is the structure of opportunities and incentives. This dimension concerns the patterns of potential inducements or sanctions that people confront in a given polity. Though many analysts recognize the importance of the structure of opportunities and incentives, scholars in the political-economy tradition have been most explicitly concerned with assessing the patterns of rewards and pen- alties that legal and bureaucratic systems configure for occupants of various roles ~Rose-Ackerman, 1978; Klitgaard, 1987!. The second dimension is political culture, understood as a repertoire of cognitions, feelings, and schemes of evaluation that process experience into action. These repertoires constitute orientations to action: general disposi- tions to act in certain ways in sets of situations. Actors learn these orientations through a process of cultural socialization ~Eckstein, 1988:790, 791!. A number of researchers have recognized the structural and normative dimensions to the problem of corruption, without necessarily making explicit their interaction ~see Rose-Ackerman, 1978; Morris, 1991; Manzetti and Blake, 1996!. The two dimensions are analytically independent. This implies that two bureau- crats, in two countries, might face similar legal-bureaucratic opportunities, say, to extort kickbacks ~low salaries, sole control over the licenses in question, and lax supervision!. But their behaviors could be quite different if the cultures into which they have been socialized provide different orientations to action; in one culture, administrative extortion might be expected, in the other it might be unthinkable. By the same token, two administrators sharing the same culture would act quite differently if placed in different bureaucratic structures. The repertoires of orientations might be the same, but divergent opportunities and incentives will produce divergent behaviors. Or, put differently, even where the political-economic structure of opportunities would seem to invite corrupt behav- ior, prevailing cultural orientations can nevertheless inhibit people from exploit- ing the opportunities. Conversely, where cultural orientations permit practices that could be deemed corrupt, formal laws and organizations can inhibit them. The next step is to theorize the relationships between political-economic struc- ture and corruption, and between political culture and corruption. To generate hypotheses on political-economic structure we focus on the aver- age income and the degree of individual economic freedom ~the inverse of state control of the economy!. In countries where incomes are relatively low, the economy generates minimal wealth for the average citizen, and almost certainly for the vast majority of citizens. Low average incomes create certain structural incentives for corrupt behaviors. Where incomes are low, economic insecurity, if not outright poverty, means that marginal additions to income can have a large impact on a familys living conditions. Another way of putting this is that in poor countries the marginal value of money is higher than it is in wealthy countries. The high utility attached even to small income supplements affects both givers and takers of bribes: paying a bribe can be worth the expense and risk if it is likely to result in greater gains in income; receiving a bribe produces a direct 36 Accounting for Corruption boost in income. Indeed, a government job that offers the possibility of extra- legal income supplements would be especially attractive. Thus government posi- tions themselves can become the objects of corrupt activities ~nepotism, bribes!. The problem of corruption does not appear, however, to be closely related to low relative salaries for government employees. A plausible argument in the corruption literature has been that underpaid bureaucrats are more likely than well-paid ones to enrich themselves via corrupt transactions. However, La Porta et al. find exactly the opposite, namely, a positive correlation between govern- ment wages ~as a share of per capita GDP! and levels of corruption ~1998:23!. We thus expect a broader logic to be at work, by which relatively low average incomes increase the propensity both to offer and to accept corrupt payments. 6 This argument leads us to our first hypothesis. Hypothesis 1: Average income will correlate inversely with the perceived level of corruption. A second argument relates to the extent of individual economic freedom. When the state and its administrative apparatus exercise relatively greater con- trol over the economy, public officials make decisions that determine who will enjoy access to economic resources and opportunities. Under these conditions, economic success depends less on market activities and more on the ability to inf luence the relevant officials. Thus bribery, extortion, payoffs, and kickbacks become viable means of inf luencing the distribution of wealth. Or, as Scott puts it, the larger is the relative size and scope of the public sector, the greater will be the proportion of certain acts that will meet our criteria of corruption ~1972:9; see also Tanzi, 1994:iii!. Conversely, where economic outcomes are largely the product of private decisions ~outside of state control!, the state will not be seen as the crucial dispenser of economic resources. 7 Private economic activity is more likely than political0bureaucratic inf luence to lead to wealth. A high level of personal economic freedom thus implies reduced political control over eco- nomic opportunities, and fewer incentives to engage in corruption. Thus our second hypothesis. Hypothesis 2: The degree of state control of the economy should correlate positively with corruption. Or, since personal economic freedom is the inverse of state control, the extent of individual economic freedom should correlate negatively with corruption. The next hypothesis concerns the relationship between democracy and cor- ruption. Democracy includes institutional and cultural, or normative, compo- 6 One could also put forward an argument that in poorer countries the pools of money available for offering bribes would also be fewer and smaller, such that in low-income countries both the demand and the supply of bribes would be lower, with no net effect on the level of corruption. We find the reasoning unconvincing. For a supply constraint to exist, everyone would have to be more or less equally poorthere could not be pockets of wealth. Yet we know that even in the poorest countries ~perhaps especially there! the distribution of income is skewed, such that there exist concentrations of wealth in rich individuals and families, not to mention large companies ~domestic as well as foreign!. In other words, we take it as a plausible assumption that in every society there are concentrations of wealth available for offering bribeslow average incomes do not eliminate the supply of bribes. 7 On the issue of state intervention in the economy, the literature on rent-seeking overlaps the literature on political corruption. Seminal works in the rent-seeking literature include Krueger, 1974, and Buchanan, Tollison, and Tullock, 1980. The rent-seeking literature, however, classifies corruption with other, quite distinct activities, like lobbying or even voting. Furthermore, the analysis of political corruption touches on payoffs that are not strictly financial, but can be ideological or political ~see DeLeon, 1993:25!. For a more extended argument regarding the differences between the rent-seeking literature and the corruption literature, see Hutchcroft, 1997:22429. Wayne Sandholtz and William Koetzle 37 nents. Both the institutional and the cultural dimensions of democracy tend to suppress corruption. Though we can distinguish theoretically between institu- tional and normative effects of democracy, we can only begin to separate those effects empirically. The institutional argument is straightforward. In democracies, government officials derive their authority from and serve the interests of the public, at least in principle. Furthermore, the citizenry possesses more, and more effective, means of detecting and punishing corruption in a democracy than it does in more autocratic forms of government. Basic democratic freedoms, like those of assembly, speech, and press, allow people and groups to uncover information, ask questions, demand inquiries, and publicize their discoveries. Corrupt acts are by their nature secret, and secrets are harder to keep in an open society. Once the public is aroused over a corruption scandal, for example, the organs of government have powerful incentives to prosecute and punish the miscreants ~or at least to be seen to do so!. In some countries, citizens can file complaints directly with ombudspersons, who have investigative powers. Elected officials, who are sensitive to public opinion, will be motivated to respond to revelations of corruption by holding hearings, launching legal proceedings, and passing laws. The more extensive are democratic freedoms and the more effective are democratic institutions, the greater will be the deterrent to corruption. The effectiveness of democratic institutions in curbing corruption, however, depends on the presence of a set of democratic norms. The public will not care about detecting, publicizing, and punishing corrupt acts unless broadly shared norms treat corruption as antagonistic to basic democratic values. Thus the institutional argument presumes a normative argument. Political culture consists of orientations that link behavior to sets of situations. Rose-Ackerman, who wrote the initial comprehensive statement of the political-economic approach, argued that both structure and culture ~values! were crucial to explaining corrupt behav- ior. She chose to focus her book on structural incentives, at the same time recognizing that an economic approach to politics. . . cannot explain the orig- ination and transmission of the democratic and personal ideals required to preserve a functioning mixed economy ~1978:56!. In fact, Rose-Ackerman explic- itly argues that if one wants to understand the functioning of democracy, it will not be possible to follow the conventional economists inclination to ignore moral constraints upon self-seeking behavior, and that certain kinds of political virtue and moral beliefs are necessary in both publics and politicians ~1978:216, 23334!. In an empirical study Della Porta and Pizzorno found that variations in the density of corruption depend more on variations in the willingness of people to be corrupted, and therefore on what we would prefer to call the moral cost ~as seen by them! of participation, than on the structure of opportunities ~Della Porta and Pizzorno, 1996:87!. We pursue these suggestions, arguing that demo- cratic norms and values delegitimize corrupt practices. As we noted earlier, definitions of corruption involve a distinction between private and public spheres. This distinction is a modern one, emerging in the eighteenth century but not fully elaborated and diffused until the nineteenth or even the twentieth century. The pre-modern state was an extension of the royal household, and public offices were the personal property of the monarch, to distribute ~or sell! as she or he desired ~Scott, 1972:ch. 3!. Gradually, the right to rule came to be seen not as inhering in the person of the king but as deriving from the will of the people ~Bendix, 1978!. Later, rational-legal bureaucracy ~to borrow Webers term! replaced patrimonial administration. Modern bureau- cracy, like modern political authority, required a clear divide between incum- bent and office, between the private and the public spheres ~Theobald, 1990:21!. Democratic norms embody these ideals; in a democracy, all acts of government 38 Accounting for Corruption derive their legitimacy, in principle, from the will of the citizenry ~or a majority of it!. Public office is therefore a trust exercised not for the personal benefit of the office-holder but on behalf of the people. Corruption perverts two central norms of democracy: equality ~corruption entails special access and inf luence! and openness ~Della Porta and Pizzorno, 1996:74!. It follows, we argue, that democratic norms and values ~cultural orientations! will produce lower levels of corruption for any given structure of opportunities. This is another way of stating the point we made earlier, that the structure of opportunities interacts with the normative context, and together they determine the level of corruption. We assume that in established democracies the following normative orientations are widely shared: all citizens should enjoy equality of opportunity before the state; public office should not be a vehicle for private enrichment; and public office entails a duty to the collective will. Where such norms are strong, people will regard corrupt practices as improper and illegiti- mate. More important, as we argued above, the institutional effects attributed to democracy cannot work to reduce corruption unless the normative dimension is potent. Put another way, openness and political competition will not reduce corruption unless existing social norms condemn it. The challenge for empirical research is to find a way to distinguish the insti- tutional from the normative constraints on corruption. We propose to access the normative dimension of democracy through the notion of socialization. That is, we expect to find a socialization effect, through which democratic norms are transmitted from one generation to the next ~Eckstein, 1988!. The longer a country has experienced democratic rule, the more deeply rooted the associated norms and values should be. A longer history of democracy implies that demo- cratic norms have diffused both more extensively and more intensively. Over time, with continuous democratic governance, norms spread to ever broader segments of the population and become more deeply rooted in individual values and action orientations. More years of uninterrupted democracy should tend to reduce corruption. From the institutional and normative aspects of democracy we derive two additional propositions. Hypothesis 3a: The strength of democratic institutions should correlate nega- tively with the level of corruption. Hypothesis 3b: Longer experience with democratic rule should correlate neg- atively with the level of corruption. In addition, we expect that the level of integration into the world economy, through trade in particular, should have an impact on both business and admin- istrative practices in a given country. Greater involvement in trade could affect both the political-economic structure of opportunities and the cultural norms of a country. On the structural side, freer trade would remove from the hands of officials certain administrative goods ~licenses, permits, waivers, and the like! that they could otherwise exchange for private rewards ~bribes, kickbacks!. Another structural argument would depict bribes as a tax on the producer or importer. In a closed market, the importer sets the price of imported goods above the inter- national price, and the bribe-taking official collects part of the monopoly profits. In a market open to trade, the bribery tax forces returns below the level pre- vailing in the market, and the producers so taxed will drop out. Thus the competition created by free trade penalizes bribery. The cultural argument has to do with socialization. International commerce and finance are dominated by firms from the OECD countries. These corporate Wayne Sandholtz and William Koetzle 39 actors share a transnational business culture based on Western norms ~see LeVine, 1989:689!. This transnational business culture includes relatively super- ficial dimensions ~what suits to wear, how to entertain visitors! but also includes the standards of business practice that prevail in the West. These rules prohibit a broad variety of corrupt practicesbribery, extortion, nepotism, kickbacks. As a countrys businesspeople and officials become increasingly connected to the international economy, they become immersed in the transnational business culture and advocate stricter prohibitions against corrupt practices, producing a convergence toward transnational anti-corruption norms. Thus, for example, Maclean argues that as France opened itself to the world economy in the 1980s, French business culture adapted itself to transnational norms, including greater transparency, which partially explains the rash of political-business scandals in the 1980s and 1990s ~Maclean, 1993:16768!. Hypothesis 4: The degree of integration into the international economy should correlate negatively with the level of perceived corruption. 8 Finally, we want to include controls for other aspects of culture, to ensure that our results are not biased by the omission of additional, potentially significant explanatory variables. Though we emphasize the roles of economic structure, democracy, and trade in explaining levels of perceived corruption, other kinds of cultural orientations could inf luence a societys propensity to engage in or tolerate corruption. Lipset and Lenz, for example, argue that the Protestant religious ethos is more conducive to norm adhering behavior, largely because of its emphasis on individual responsibility for moral conduct and its generally less tolerant attitude toward human failing. Their empirical analysis reveals a strongly negative correlation between Protestantism and perceived corruption ~Lipset and Lenz, 1999:1112!. In addition, Treismans analysis reveals that countries that have in the past been British colonies tend to have lower levels of perceived corruption. He argues that the British passed on to their colonial possessions a powerful norm that emphasizes compliance with established procedures as opposed to personal authority ~Treisman, 1998:24!. In light of these results, we include Protestantism and British colonial experience in our analysis, via the following propositions: Hypothesis 5: The share of the population having Protestant religious affiliation should correlate negatively with the level of perceived corruption. Hypothesis 6: Experience as a British colony should correlate negatively with the level of perceived corruption. Data As stated previously, major obstacles to the comparative study of corruption have been the absence of a general standard of corruption and the lack of cross- national data on corrupt practices. The dependent variable in this study, there- fore, is not an objective measure of corruption ~since none exists! but a subjective measure based upon the perceptions of various relevant groups. Our data come from Transparency Internationals Corruption Perceptions Index, which assigned scores to fifty-four nations in 1996 ~Transparency International, 1996!. The Index 8 The effects of trade on levels of corruption could work through either or both of the two channels we mentioned, that is, structure and culture. Our data do not permit us to determine which effect is stronger, but that might be a worthwhile issue for future research. 40 Accounting for Corruption is actually a poll of polls, combining the results of ten different surveys. The component polls include samples of transnational businesspersons and consul- tants, as well as chambers of commerce and local populations. 9 The central advantages of this Index as a measure of corruption are that it allows for cross- national comparison and it employs the definition of corruption ~the misuse of public power for private gain! we arrived at above. We operationalize the political-economic structure with two variables. For average income we use per capita Gross Domestic Product ~GDP0cap!. To mea- sure state control of economic activities we employ the Freedom House Indica- tors of Economic Freedom ~Messick,1996!. The Economic Freedom index includes six primary factors related to individual economic rights: freedom to hold prop- erty, freedom to earn a living, freedom to operate a business, freedom to invest ones earnings, freedom to trade internationally, and freedom to participate in a market economy. The 1996 Economic Freedom measure provides data for eighty- two countries, including fifty of the fifty-four for which we have a corruption score. 10 This measure is particularly useful for our purposes because it directly accesses the degree to which individuals can participate openly in, and derive the benefits from, a variety of market activities. In order to capture the impact of democracy upon perceived levels of corrup- tion, we employ two indicators. One provides a measure of the extent of dem- ocratic norms and practices, the other addresses the depth of socialization in those norms. The Freedom House indexes of political rights and civil liberties capture the range of democratic practices in each country. Political rights include: free and fair elections of political leaders who possess real policy-making powers; the right to form political parties; freedom from domination by the military or other oligarchic groups; the existence of genuine political opposition; and pro- tection of minority rights. Civil liberties encompass the right to free speech and media, freedom of assembly, the right to organize, and equality under the law ~Karatnycky,1997!. By combining these indexes for 199596 we obtain a variable, Democracy, measuring the breadth of democratic norms and practices for 192 countries ~including 53 of the 54 countries for which we have corruption scores!. 11 9 The index is constructed by surveying the results of ten different polls, each conducted over different time periods and focusing on different sets of nations. Three of these surveys are from the World Competitiveness Report ~Institute for Management Development, Lausanne! conducted in 1993, 1994, and 1995, respectively; they asked top executives about improper practices ~i.e., bribing! in the public sphere. Three of the surveys are from The Political and Economic Risk Consultancy ~Hong Kong! undertaken in 1993, 1994, and 1995 which polled American, European, and Australian Managers about corruption in 11 Asian countries. Other surveys included in this survey are those undertaken by Impulse ~Peter Neumann, 1994! which polled Embassies and Chambers of Commerce about the spread of corruption in 103 countries; the poll conducted by DRI0McGraw-Hill Global Risk Service ~1995! of staff assessments of losses due to corruption; a survey by Political Risk Services ~East Syracuse, New York, 19931995! which polled staff about the likeliness of being asked to provide special and illegal payments as the price of doing business and included 148 countries; and the Internet Corruption Ranking ~Gottingen Univer- sity, 199596! which asked employees of multinational firms about corruption in 58 countries. For a country to be included in the Transparency Index, it had to be covered in at least four surveys, one of which had to include a sample population within the country. Thus 54 countries receive a corruption score. The logic for this is that inaccuracies in a single survey are compensated for by the inclusion of additional polls. In addition, the accuracy of the overall Index is indicated by the fact that most surveys are highly correlated with each other ~Transparency International, 1996:2!. Since for this Index the higher the score the lower the corruption, in the analysis that follows, we use the Index multiplied by 1 such that the higher the score the greater the corruption. This transformation eases the interpretation of the relationships between this Index and the independent variables used to explain variation in these scores. 10 A low score on this measure indicates a high level of state control over the economy; a high score indicates extensive individual economic freedoms. Cameroon, Uganda, Ecuador, and Hong Kong are not scored by the Freedom House in the Economic Freedom Index. 11 A low score indicates minimal individual political and civil liberties; a high score indicates the opposite. Because of this large sample, every country with a corruption ranking is also ranked by Freedom House except Hong Kong. Wayne Sandholtz and William Koetzle 41 Earlier we advanced two arguments, one institutional and one normative, as to why democracy should reduce corruption. Under functioning democratic insti- tutions, government officials are more accountable to the public, and the public has means of scrutinizing the conduct of public officials and punishing corrupt behaviors. On the normative side, corruption violates fundamental democratic values like openness and equal treatment. In a society where democratic values are deeply ingrained, corruption will be seen, by both citizens and officials, as abhorrent and unacceptable. Our Democracy variable does not allow us to dis- tinguish between institutional and normative constraints on corruption. In fact, the institutional argument implicates the normative one as well, for democratic institutions will function to detect and punish corruption only if there are in place social norms that condemn corrupt practices. Put differently, citizens will detect and condemn corrupt acts only if they believe corruption to be wrong. We thus expect the Democracy variable to be negatively related to corruption, but such a finding would confirm both the institutional and the normative argu- ments concerning democracy. In order to test independently the normative argument, we devise a second variable intended to measure the depth of acceptance of democratic norms. Because norms are transmitted through processes of socialization, we should be able to notice differences among countries depending on how long they have had continuous democratic government. The longer democratic norms and val- ues have informed politics in a country, the more ingrained and potent those norms will be. We generate a variable, Democratic Years, that counts the number of years since 1948 that countries have had democratic rule ~Lijphart, 1984:37 38!. The data come from Political Handbook of the World 1997 ~Banks et al., 1997!. Democratic Years should therefore show a negative relationship with perceived levels of corruption. We employ trade indicators to operationalize the degree of integration into the world economy. Our indicator is the sum of exports and imports ~Trade! as a share of Gross Domestic Product, with trade and GDP data taken from the World Tables 1995 ~World Bank, 1995!. 12 Since small countries tend to have a higher ratio of trade to GDP, we also need to control for economic size. In one model we thus include the natural logarithm of Total GDP as control and assess its effects on the Trade variable. Finally, we measure the extent of Protestantism in the sample countries with the percentage of the population claiming affiliation with a Protestant church ~La Porta et al., 1998:Appendix B!. British colonial experience is a dummy vari- able ~1 for countries that are former British colonies; 0 otherwise!, based on information in the Encyclopedia Britannica. The United Kingdom receives a score of 1 on this variable since it was the putative origin of the orientation toward compliance with established procedures. Data Analysis To begin our analysis of the factors that underlie varying levels of corruption, we first turn to an examination of perceived levels of corruption in the countries in our sample. Table 1 presents the Transparency International Perceived Corrup- tion Index for fifty-four nations from all parts of the globe. The table lists countries from least to most corrupt, as perceived by respondents in the polls included in the Transparency scores. A cursory examination of the rankings in the table generally supports our theoretical expectations expressed above: almost 12 The trade data for Taiwan, however, come from The Handbook of International Trade and Development Statistics ~United Nations, 1995!. 42 Accounting for Corruption without exception, those countries seen as least corrupt are those nations that are known to be democratic, have relatively high levels of political and economic freedom, and are highly integrated into the world economy. Nations that fall on this end of the corruption spectrum include New Zealand, Denmark, Sweden, Finland, and Canada. In contrast, the countries that are viewed as most corrupt are those traditionally seen as authoritarian, those that allow little entrepreneur- ship in either the political or economic sphere, and are less integrated into the world economy. Nations seen as most corrupt include Nigeria, Pakistan, Kenya, and Bangladesh ~on Kenya and Nigeria, see Williams, 1987:8086, 9599!. However, given the hypotheses stated above, there are some surprises. Notice, for example, that Singapore is viewed by those doing business there as having very low levels of corruptionlower than Australia, the Netherlands, and the United States, for examplein spite of the fact that residents of Singapore do not have the same level of political or economic freedoms found in many coun- tries perceived as more corrupt. At the other end of the spectrum, India, which has been a democracy for fifty years, is seen as quite corruptin fact only eight other nations in this sample were rated as more corrupt. Finally, countries that fall in the middle of the corruption scale represent a mix of nations with rela- tively high levels of economic integration but lower levels of political and eco- nomic freedom ~e.g., South Africa, South Korea, and Taiwan! along with nations with longer histories of democratic practices ~e.g., Italy and Spain!. Thus, while Table 1. 1996 Transparency International Perceived Corruption Index, by Country Country Score Country Score 1. New Zealand ~NZL! 9.43 28. Greece ~GRC! 5.01 2. Denmark ~DNK! 9.33 29. Taiwan ~TAI! 4.98 3. Sweden ~SWE! 9.08 30. Jordan ~ JOR! 4.89 4. Finland ~FIN! 9.05 31. Hungary ~HUN! 4.86 5. Canada ~CAN! 8.96 32. Spain ~ESP! 4.31 6. Norway ~NOR! 8.87 33. Turkey ~TUR! 3.54 7. Singapore ~SGP! 8.80 34. Italy ~ITA! 3.42 8. Switzerland ~CHE! 8.76 35. Argentina ~ARG! 3.41 9. Netherlands ~NLD! 8.71 36. Bolivia ~BOL! 3.40 10. Australia ~AUS! 8.60 37. Thailand ~THA! 3.33 11. Ireland ~IRL! 8.45 38. Mexico ~MEX! 3.30 12. United Kingdom ~GBR! 8.44 39. Ecuador ~ECU! 3.19 13. Germany ~DEU! 8.27 40. Brazil ~BRA! 2.96 14. Israel ~ISR! 7.71 41. Egypt ~EGY! 2.84 15. United States ~USA! 7.66 42. Colombia ~COL! 2.73 16. Austria ~AUT! 7.59 43. Uganda ~UGA! 2.71 17. Japan ~ JPN! 7.05 44. Philippines ~PHL! 2.69 18. Hong Kong ~HKS! 7.01 45. Indonesia ~IDN! 2.65 19. France ~FRA! 6.96 46. India ~IND! 2.63 20. Belgium ~BEL! 6.84 47. Russia ~RUS! 2.58 21. Chile ~CHL! 6.80 48. Venezuela ~VEN! 2.50 22. Portugal ~PRT! 6.53 49. Cameroon ~CMR! 2.46 23. South Africa ~ZAF! 5.68 50. China ~CHN! 2.43 24. Poland ~POL! 5.57 51. Bangladesh ~BGD! 2.29 25. Czech Republic ~CZE! 5.37 52. Kenya ~KEN! 2.21 26. Malaysia ~MYS! 5.32 53. Pakistan ~PAK! 1.00 27. South Korea ~ROK! 5.02 54. Nigeria ~NGA! .69 Source: Transparency International ~1996!. Note: A score of 0 represents perceived to be totally corrupt; a score of 10 represents perceived to be totally clean. Wayne Sandholtz and William Koetzle 43 the table generally conforms to our theoretical expectations, we need systematic analysis to reach any general conclusions about variation in corruption across countries. In order to test our hypotheses we analyze a set of multivariate models with Perceived Corruption as the dependent variable. Table 2 reports the results. 13 Model 1 includes our core variables and strongly confirms our initial expecta- tions. Average income, economic freedom, democracy, democratic years, and trade integration are all negatively related to perceived corruption and statisti- cally significant, and the model explains just over 80 percent of the variance in levels of perceived corruption for the countries in our sample. 14 The variables related to democratic institutions and norms offer the least predictive leverage in this model, with lower standardized coefficients and lower significance levels than the other variables. In the second model we add Total GDP ~logged! as a control for size, espe- cially with respect to the trade variable ~according to the standard notion that economically smaller states trade a disproportionately large share of their GDP!. With the Total GDP control, all of the other variables retain their significance ~Democratic Years actually moves up to the .05 significance level!. Trade is still highly significant, though with a slightly reduced coefficient ~the standardized coefficient drops from .24 to .21!. The Total GDP variable does not achieve statistical significance and the adjusted R 2 rises by just .01. We therefore discard the Total GDP variable in subsequent models. When we add the British Colony variable in model 3, we observe some minor changes in the democracy-related variables. The coefficient for Democracy rises to .18, but the coefficient for Democratic Years loses statistical significance. In any case, the British Colony variable fails to achieve statistical significance and adds nothing to the total explained variance. British Colony drops out of the remaining model. The percentage of the population with Protestant religious affiliation, in con- trast, shows a strong correlation with perceived corruption in model 4, and is significant at the .01 level. The other variables retain their significance, though with some shifts in strength and significance. The GDP per capita variable drops to the .05 significance level and no longer shows the highest standardized coef- ficient. Protestantism apparently has some relationship with wealth; La Porta et al. find that Protestant countries are in fact richer ~1998:25!. For our purposes, though, the important point is that Protestantism clearly has an independent effect. The Percent Protestant variable itself is highly significant, appearing to confirm the hypothesis that the Protestant emphasis on individual responsibility and rectitude carries over into a reduced toleration for corruption. In sum, model 4, with our economic structure, democracy, trade, and Protes- tantism variables, explains the largest share of variance in perceived corruption, with an adjusted R 2 of .84. We can now examine the standardized slope coeffi- cients in order to draw comparisons regarding the relative explanatory weight of the variables. The most powerful predictors in our final model are Economic Freedom ~ b .26! and Trade ~ b .26!. Greater opportunity for economic gain through the market reduces the incentive to resort to corrupt activities. Increased involvement in international trade can dampen corruption by inten- sifying economic competition, reducing the opportunities for corruption ~state- controlled import and export permits, for example!, and socializing actors into 13 Cameroon, Uganda, Ecuador, and Hong Kong are not included in this analysis because of missing data for some of the independent variables. 14 Though we are unable, given our data, to test for reverse causation ~the possibility that corruption leads to trade closure!, our results are congruent with studies that do control for reverse causation and do find that higher trade is associated with lower corruption ~see Ades and Di Tella, 1997:9697!. 44 Accounting for Corruption Table 2. Multivariate Regression Analysis Explaining Levels of Perceived Corruption Across Countries Dependent Variable: Corruption Perceptions Index Independent Variables Model 1 Model 2 Model 3 Model 4 GDP0capita .00008*** ~.00002! .0001*** ~.00003! .00009*** ~0.0002! .00006** ~.00003! b .31 b .37 b .34 b .23 Economic freedom .21** ~.09! .19** ~.09! .20** ~.09! .19** ~.08! b .29 b .27 b .27 b .26 Democracy .16* ~.09! .16* ~.09! .18* ~.09! .17** ~.08! b .22 b .21 b .25 b .23 Democratic years .02* ~.01! .03** ~.01! .02 ~.01! .02* ~.01! b .18 b .20 b .14 b .15 Trade 1.46*** ~.42! 1.24*** ~.46! 1.35*** ~.43! 1.58*** ~.38! b .24 b .21 b .22 b .26 Total GDP ~log! .186 ~.15! b .11 British colony .44 ~.39! b .08 Percent Protestant .02*** ~.007! b .21 Constant .70 ~.64! 1.59 ~1.97! .79 ~.64! .57 ~.59! Adjusted R 2 .81 .82 .81 .84 N 50 49 50 50 Sources: Data are from sources cited in the Data section. The table reports unstandardized coefficients, with standard errors in parentheses and standardized coefficients ~ b! below. *p , .10; **p , .05; ***p , .01 the predominantly Western norms of the international economy. Higher GDP0 capita ~ b .23! works to reduce corruption by diluting the marginal value of money. Relative poverty, in contrast, can motivate both officials and private citizens to enhance their incomes through corruption. Democratic institutions ~ b .23! provide the citizenry, and elected leaders, with the means to monitor, expose, and penalize corrupt acts. But the Democratic Years variable ~ b .15! also provides support for our proposition that there are socialization effects at work, and that with longer experience of democratic governance, democratic norms that are hostile to corruption ~equality, openness! do become more deeply rooted. Protestant values have an independent effect in the same range as the democracy variables ~ b .21!. Another way to picture the strength of the relationship between our model and levels of perceived corruption is to graph actual levels of corruption against the predictions derived from the multivariate regression in model 4. Figure 1 plots the predicted level of corruption against the actual level of corruption for the countries in our sample. Not surprisingly, given the explanatory power of our model, most of the countries lie on the diagonal between the two axes: countries generally exhibit the level of corruption our model predicts they will have. However, several countries prove to be interesting outliers. First, China appears to be less corrupt than our model suggests it should be. Even so, China did score fairly high on corruption. There is probably a functional limit to perceived levels of corruption that our model cannot capture. Second, there is a set of nations that are per- ceived to be more corrupt than our model predicts they ought to be, notably Italy and Belgium. Italy ~which has the largest residual! and Belgium are inter- esting because both enjoy relatively high levels of economic and political free- dom, experience with democracy, and levels of international trade. Our model suggests that they should have less corruption than they actually experience. Thus, while our model tells us quite a lot about comparative levels of corruption, clearly there is more to the story. RUS BGD VEN PHL COL BOL THA MYS TAI CZE FRA JPN SWE FIN SGP CHE NLD IRL DEU DNK NOR NZL CAN USA BEL AUS GBR AUT ITA ESP HUN PRT ISR CHL POL GRC KOR JOR ARG IND MEX TUR BRA EGY KEN PAK NGA IDN CHN ZAF Fig. 1. Predicted Versus Actual Levels of Corruption 46 Accounting for Corruption Discussion Corruption erodes some of the core values of a democracy, namely, that collec- tive decisions should emerge from public processes guided by known rules, and that all citizens should have equal access to those processes. For political scien- tists interested in questions of democratic governance and representation, cor- ruption should constitute a central normative problem. Yet corruption persists, despite nearly universal laws and agencies designed to eliminate it, even in the most established democracies. Comparative research on corruption has in the past been stymied by the lack of a general definition of corruption and by the absence of data. The defini- tional dilemma arises because standards of corruption are culturally specific, often differing from country to country. One response to this problem is to embrace the cultural specificity of corruption norms and study the meaning, mechanisms, and effects of corruption in a given culture. Most empirical research on corruption thus focuses on one, or a few countries. Consequently, though this scholarship has produced plausible interpretations of the role corruption plays in specific societies, the conclusions apply with any degree of confidence only to the countries studied. But if we want to compare countries, and explain differ- ences in the incidence of corrupt practices, culturally specific definitions will not suffice. The alternative we have pursued here is to choose one perspective and define corruption from that vantage point. For purposes of comparative analysis, some groups definition of corruption has to serve as a yardstick. In this study, we have employed a definition of corruption ~the misuse of public office for private gain! for which a plausible intellectual rationale exists in the literature, and which corresponds to the definition of corruption informing the Transparency International Index. Of course, we recognize that some practices counted as corrupt by those surveyed for the polls that go into the Index may not be considered improper in the society in which they occur ~though that would have to be established and not merely asserted!. But this approach is the only one that permits general, cross-national conclusions about differences in the extent of corrupt activities and factors that might explain these variations. We offer a theoretical framework in which the level of corruption is a product of two primary factors: the political-economic structure of incentives and oppor- tunities, and the cultural norms ~or orientations! that shape peoples percep- tions, evaluations, and choices. Because the corruption data include a strong component of international commerce, we also theorize the impact of trade integration on levels of corruption. From the theory we derive five hypotheses. We predict that a countrys level of corruption will generally be lower: ~1! the higher the average income; ~2! the greater the extent of individual economic freedom and opportunity ~the less the state controls the economy!; ~3! the greater the degree of integration in the international economy; ~4! the stronger is democracy; and ~5! the more prevalent are Protestant values. The data analysis presented above closely conforms to our expectations in each case. In the multi- variate model, each explanatory variable is significant in the direction we expect. Though the model performs well, it also suggests areas for additional inquiry. For instance, Belgium and Italy both have significantly higher corruption scores than our model would predict. Both countries score high on most of the factors that generally correspond with low levels of corruption, the only exception being Percent Protestant. We speculate that Italy and Belgium have something else in common, namely, political systems that fragment the distribution of political and economic goods in ways that are susceptible to patronage and corruption. In Belgium, a substantial share of government programs and economic resources are divided between the Flemish and Walloon communities. An elite within each ethno-linguistic grouping exercises substantial control over the allocation of Wayne Sandholtz and William Koetzle 47 these goods. Because of the extensive devolution of governance to the regions, those elites have substantial leeway to distribute goods so as to encourage loyalty within the community. Likewise in Italy, the distribution of political goods has been fragmented, not along ethnic lines but along party lines. The parties that are victorious in municipal, regional, and national elections win the right to distribute a plethora of administrative and quasi-administrative posts among their supporters. This system lends itself to personalism and patronage, as party workers seek the rewards of public employment and party leaders reward loyalty. Detailed empirical work could begin to test the plausibility of these speculations. Our results suggest some interesting links to other research programs. For instance, David Lake has argued that democracies are more likely than autocra- cies to prevail in war because they extract fewer rents from society. Corrupt payments are one type of rent that state officials in some societies routinely siphon off from the economy. High levels of corruption mean greater overall rents extracted by the state, which constrains the governments ability to commit resources to the military ~Lake, 1992!. Our study suggests that democratic states may fight better because they are less corrupt. In a different vein, Joel Hellman has recently shown that the most democratic post-communist states have had the most successful economic reforms ~Hellman, 1998!. The logic is that govern- ments that are sheltered from electoral pressures face a variety of obstacles to economic reform from actors who benefit from prior, privileged access to state economic assets. We could reformulate this finding in terms of our analysis to suggest that corrupt relationships are an important source of resistance to eco- nomic liberalization; participants in corruption benefit precisely because they have rigged the system in their favor. Democracies are best at post-communist reform because they minimize those corrupt relationships. Finally, our model implies that contemporary trends in world politics, including the third wave of democratization and increasing global economic integration, may over time lead to lower levels of corruption. References Ades, A., and R. Di Tella ~1997! The New Economics of Corruption: A Survey and Some New Results. In Political Corruption, edited by Paul Heywood. Oxford: Blackwell. Banks, A. S., A. J. Day, and T. C. Muller, eds. ~1997! Political Handbook of the World 1997. Binghamton, NY: CSA Publications. Bayley, D. H. @1966# ~1989! The Effects of Corruption in a Developing Nation. In Political Corrup- tion: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Bendix, R. ~1978! Kings or People: Power and the Mandate to Rule. Berkeley and Los Angeles: University of California Press. Buchanan, J. M., R. D. Tollison, and G. Tullock, eds. ~1980! Toward a Theory of the Rent-Seeking Society. College Station: Texas A&M Press. DeLeon, P. ~1993! Thinking About Political Corruption. Armonk, NY: Sharpe. Della Porta, D., and A. Pizzorno ~1996! The Business Politicians: Ref lections from a Study of Political Corruption. In The Corruption of Politics and the Politics of Corruption, edited by M. Levi and D. Nelken. Oxford: Blackwell. Eckstein, H. ~1988! A Culturalist Theory of Political Change. American Political Science Review 82:789804. Friedrich, C. J. ~1966! Political Pathology. Political Quarterly 37:7085. Friedrich, C. J. @1972# ~1989! Corruption Concepts in Historical Perspective. In Political Corruption: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Gardiner, J. A., and T. R. Lyman @1978# ~1989! The Logic of Corruption Control. In Political Corruption: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. 48 Accounting for Corruption Gibbons, K. M. ~1989! Toward an Attitudinal Definition of Corruption. In Political Corruption: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Gillespie, K., and G. Okruhlik ~1991! The Political Dimensions of Corruption Cleanups: A Frame- work for Analysis. Comparative Politics 24:7795. Heidenheimer, A. J. @1970# ~1989! Perspectives on the Perception of Corruption. In Political Cor- ruption: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Bruns- wick, NJ: Transaction. Heidenheimer, A. J., M. Johnston, and V. T. LeVine ~1989a! Introduction. In Political Corruption: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Heidenheimer, A. J., M. Johnston, and V. T. LeVine, eds. ~1989b! Political Corruption: A Handbook. New Brunswick, NJ: Transaction. Hellman, J. ~1998! Winners Take All: The Politics of Partial Reform in Postcommunist Transitions. World Politics 50:203235. Heywood, P., ed. ~1997! Political Corruption. Oxford: Blackwell. Huntington, S. P. ~1968! Political Order in Changing Societies. New Haven, CT: Yale University Press. Hutchcroft, P. ~1997! The Politics of Privilege: Assessing the Impact of Rents, Corruption, and Clientelism on Third World Development. In Political Corruption, edited by P. Heywood. Oxford: Blackwell. Johnston, M. @1986# ~1989! The Political Consequences of Corruption: A Reassessment. In Political Corruption: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Karatnycky, A. ~1997! Freedom in the World: The Annual Survey of Political Rights and Civil Liberties 19961997. New York: Freedom House. Klitgaard, R. ~1987! Controlling Corruption. Berkeley and Los Angeles: University of California Press. Krueger, A. O. ~1974! The Political Economy of the Rent-Seeking Society. American Economic Review 64:291303. Lake, D. A. ~1992! Powerful Pacifists: Democratic States and War. American Political Science Review 86:2437. La Porta, R., et al. ~1998! The Quality of Government. Working paper no. 6727. National Bureau of Economic Research, Cambridge, Mass. ~http:00www.nber.org0papers0w6727!. Leff, N. H. @1964# ~1989! Economic Development Through Bureaucratic Corruption. In Political Corruption: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Levi, M., and D. Nelken, eds. ~1996! The Corruption of Politics and the Politics of Corruption. Oxford: Blackwell. LeVine, V. T. ~1975! Political Corruption: The Ghana Case. Stanford, CA: Hoover Institution Press. LeVine, V. T. ~1989! Transnational Aspects of Political Corruption. In Political Corruption: A Hand- book, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Leys, C. ~1989! What Is the Problem About Corruption? In Political Corruption: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Lijphart, A. ~1984! Democracies: Patterns of Majoritarian and Consensus Government in Twenty-One Coun- tries. New Haven, CT: Yale University Press. Lipset, S. M., and G. S. Lenz ~1999! Corruption, Culture, and Markets. Typescript. George Mason University, Arlington, Va. Little, W., and E. Posada-Carb, eds. ~1996! Political Corruption in Europe and Latin America. Lon- don: Macmillan. Lowenstein, D. H. @1985# ~1989! Legal Efforts to Define Political Bribery. In Political Corruption: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Maclean, M. ~1993! Dirty Dealing: Business and Scandal in Contemporary France. Modern and Contemporary France 1:161170. Manzetti, L., and C. H. Blake ~1996! Market Reforms and Corruption in Latin America: New Means for Old Ways. Review of International Political Economy 3:662697. Markovits, A. S., and M. Silverstein, eds. ~1988! The Politics of Scandal: Power and Process in Liberal Democracies. New York: Holmes and Meier. Mny, Y. ~1992! La corruption de la Rpublique. Paris: Fayard. Messick, R. E., ed. ~1996! World Survey of Economic Freedom 19951996. New York: Freedom House. Wayne Sandholtz and William Koetzle 49 Moodie, G. C. @1980# ~1989! On Political Scandals and Corruption. In Political Corruption: A Hand- book, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Morris, S. D. ~1991! Corruption and Politics in Mexico. Tuscaloosa: University of Alabama Press. Myrdal, G. ~1968! Asian Drama: An Enquiry into the Poverty of Nations, vol. 2. New York: The Twentieth Century Fund. Noonan, J. T. ~1984! Bribes. New York: Macmillan. Nye, J. S. ~1989! Corruption and Political Development: A Cost-Benefit Analysis. In Political Corrup- tion: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Palmier, L. ~1985! The Control of Bureaucratic Corruption: Case Studies in Asia. New Delhi: Allied Publishers. Quah, J. S. T. ~1989! Singapores Experience in Curbing Corruption. In Political Corruption: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Rogow, A. A., and H. J. Lasswell ~1963! Power, Corruption and Rectitude. Englewood Cliffs, NJ: Prentice-Hall. Rose-Ackerman, S. ~1978! Corruption: A Study in Political Economy. New York: Academic Press. Scott, J. C. ~1972! Comparative Political Corruption. Englewood Cliffs, NJ: Prentice-Hall. Tanzi, V. ~1994! Corruption, Government Activities, and Markets. Working paper. International Monetary Fund, Washington, D.C. Theobald, R. ~1990! Corruption, Development and Underdevelopment. Durham, NC: Duke University Press. Transparency International ~1996! Ranking 1996. Internet Center for Corruption Research, Goet- tingen University. http:00www.gwdg.de0;uwvw. Treisman, D. ~1998! The Causes of Corruption: A Cross-National Study. Typescript. University of California, Los Angeles. United Nations Conference on Trade and Development ~1995! Handbook of International Trade and Development Statistics 1994. New York: United Nations. Van Klaveren, J. @1957# ~1989! Corruption as a Historical Phenomenon. In Political Corruption: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Von Aleman, U. ~1989! Bureaucratic and Political Corruption Controls: Reassessing the German Record. In Political Corruption: A Handbook, edited by A. J. Heidenheimer, M. Johnston, and V. T. LeVine. New Brunswick, NJ: Transaction. Williams, R. ~1987! Political Corruption in Africa. Aldershot, England: Gower. World Bank ~1995! World Tables 1995. Baltimore, MD: Johns Hopkins University Press. Wraith, R., and E. Simpkins ~1963! Corruption in Developing Countries. London: Allen and Unwin. 50 Accounting for Corruption