Annu. Rev. Polit. Sei. 2004. 7:273-96 doi: 10.1146/annurev.polisci.7.012003.104815 Copyright (c) 2004 by Annual Reviews. All rights reserved .SB O 3 0- « Political Representation in Comparative Politics G. Bingham Powell, Jr. Department of Political Science, University of Rochester, Rochester, New York 14627; email: gb.powell@rochester.edu Key Words election laws, proportional representation, vote-seat correspondence, -3 § issue congruence § I I ^ I Abstract Two large research programs have analyzed election-based connections 112 between citizens and policy makers in different democracies. Studies of vote-seat g « representation in the tradition of Rae (1967) begin with citizens' party votes and have _S § made substantial progress in elucidating the impact of election laws, geographic vote "g o distributions, and the number of parties and their interactions on the proportionality 1 0 of party representation. Studies of substantive representation in the tradition of Miller I > & Stokes (1963) begin with citizen issue preferences and link these to the positions of q _2 their representatives. Most studies outside the United States, confronting multimember g= g districts and the cohesion of party representatives, have focused on voter-party dyads 1 rather than geographic constituencies, and confirmed the importance of issues linked ?3 > to a common electoral discourse and the greater structure of legislator issue positions. 5 j Recently, a number of explicitly comparative analyses have begun to analyze collective ° -;3 correspondence and confront other limitations of the literature. INTRODUCTION Robert Dahl observed that in the nineteenth century the idea of representation "transformed democracy from a doctrine suitable only for small and rapidly vanishing city-states to one applicable to the large nation-states of the modern age" (Dahl 1989, p. 29). "Yet," he added, "the change in democracy resulting from its union with representation created its own problems. An entirely new and highly complex constellation of political institutions, which we are only beginning to understand, superseded the sovereign assembly that was central to the ancient conception of democracy" (Dahl 1989, p. 30). Democratic representation means that the actions of these policy makers are supposed to be responsive to the wishes of the people. Moreover, simple correspondence between what citizens want and what policy makers do is not enough. A benevolent dictatorship is not a representative democracy. The latter depends not only on correspondence or responsiveness but also on institutionalized 1094-2939/04/0615-0273$14.00 273 274 powell arrangements that reliably create such connections. The most essential and irreplaceable of these institutions is the free and competitive national election in which all citizens can participate equally (Pitkin 1967, pp. 232-34). The language of principals and their agents is sometimes useful. In a representative relationship, we can conceive of the citizens as principals represented by agents to whom the citizens temporarily delegate the power to make public policies. In a democracy, the citizens should be equal to each other in this relationship. Various normative standards can be applied to specify the desired relations between the principals and their agents in detail. Empirical theories explain what shapes these relations in different countries and elections. Virtually all research on citizens, elections, and policy making in contemporary democracies is relevant to democratic representation broadly conceived. Rather than attempting a grand synthesis, this essay reports on two large bodies of research in comparative politics, emergent research programs that have explicitly addressed initial representative connections between the citizens and their elected policy makers. The first body of research, "procedural" representation, begins with citizens' votes for parties in elections. Party voting is then linked to party representation in the legislature through aggregation of party votes into victories. Democratic representation means that votes for parties should correspond to the seats those parties win in the legislature. The second body of research, "substantive" representation, begins with citizens' preferences rather than with their votes. Voter choices under electoral competition link citizens' preferences to the preferences and behavior of legislative representatives. Parties and candidates take positions on issues, and these electoral commitments shape their policy making after the election. Democratic representation means that citizens' issue preferences should correspond to the positions or behavior of their representatives. Of course, these procedural and substantive connections between citizens and policy makers do not exhaust the representational linkages, which must also take account of the other forces that shape the making and implementation of public policies. Nonetheless, the two bodies of research considered here elucidate critical linkages in the representative connection. CITIZENS' VOTES AS THE STARTING POINT: THE VOTE-SEAT PARADIGM IN COMPARATIVE POLITICS Let us begin with what seems a highly simplified formulation of the problem of democratic representation: the relationship between the votes for political parties that citizens cast in elections and the partisan composition of legislatures that emerge from those elections to represent them. This relatively coherent body of theory and research has made substantial progress in understanding one element of democratic representation. Typical of scientific progress, it has made headway in part by simplifying a broad normative problem into a limited empirical one. It deals with (or evades) perhaps the hardest empirical and normative problem in political representation 275 representation analysis, citizen preferences, simply by assuming that all we can or need to know about those preferences is the partisan votes that citizens cast in competitive elections. Election Laws The study of votes, rales, and their electoral consequences has a very long history. John Stuart Mill was well aware that rules with single-member or winner-take-all districts tended to advantage the largest parties (or perhaps "a few large sectional minorities in particular places" (Mill 1958 [1861], p. 111). The family of election rales called "proportional representation" (PR) was invented to create greater proportionality in vote-seat correspondence. In his classic work Political Parties, Duverger (1954, p. 373) discussed the putative advantages of PR rales for vote-seat proportionality and also noted their variations and limitations in practice. But vote-seat studies in the past 35 years have been dominated by Rae's (1967, 1971) wonderful empirical study, The Political Consequences of Electoral Laws. This elegant little book systematically distinguished a variety of types of electoral laws, identified some of their important properties, introduced systematic measures of vote-seat disproportionality and the creation of legislative majorities, as well as the fractionalization of party systems, and performed other essential services on the way to analysis of the empirical consequences of election laws in 115 elections. Rae demonstrated that the critical feature of election rales that shaped vote-seat translation in these elections was their "district magnitude" (the average number of representatives per district), which dwarfed the still significant effects of differences in computation rules and other relevant features (Rae 1967, pp. 138—40). The widely used election rule with the greatest tendency to disproportionality is the single-member district (SMD) plurality rale, also known as first-past-the-post (hereafter, FPTP), used in such countries as the United States, Britain, Canada and, until recently, New Zealand. Rae's landmark empirical contribution has been elaborated and developed in a variety of different kinds of studies, empirical and methodological (e.g., Benoit 2001; Blais & Carty 1987; Cox 1991; Cox & Shugart 1991; Gallagher 1991, 1992; Grofman 1983; Grofman & Lijphart 1986; Gugkin & Taylor 1979; Lijphart 1985,1986,1990; Lijphart&Gibberd 1977;Loosemore&Hanby 1971; Taagepera 1986; Taagepera & Shugart 1989). Such studies, and others to be found in Electoral Studies and many other political science journals, have greatly extended Rae's account of the variations in election rules and their consequences. One line of analysis has also integrated Rae's work with the long-suggested idea of a "cube law" in vote-seat connections in SMD systems (e.g., Johnston et al. 1994, Tufte 1973). The concept of proportionality itself contains alternative normative versions, reflected in part in different PR counting rales, as pointed out by Gallagher (1991). Specific rales are also adapted in various ways to achieve different practical purposes, including political stability and the partisan goals of the rale writers, as various studies of particular countries have elaborated. In 1994, Lijphart and his colleagues replicated and expanded Rae's original work using 276 powell additional data, variables, and measures of disproportionality. Lijphart's (1994) conceptualization and measurement of an "effective threshold" provides a single measure that takes account of a number of specific features of election rules, including both district magnitude and formal minimum thresholds. It proves to be the most powerful predictor of vote-seat disproportionality, although the PR calculation formula and the size of the assembly remain significant (see Katz 1997, Ch. 9, for a reanalysis on a still larger data base). Geographic Distributions The analysis of election laws has dominated work on vote-seat correspondence, but two other important variables have also emerged in comparative research. One is the geographic distribution of the votes, which becomes important for proportionality when the election rules have low district magnitudes, above all in SMD systems. Naturally enough, American, British, and New Zealand scholars whose work has been primarily within such systems have long been sensitive to the role of geography, which is often ignored by scholars working in systems with large-magnitude PR rules. In 1942, Schattschneider provided a particularly clear account of the impact of geographic distribution on outcomes in SMD systems, concluding, "In an extreme case the party in question might win all of the seats or it might win none at all merely by virtue of the fact that it had received 25% of the total vote" (Schattschneider 1942, p. 70). Some insight into the source of variation is provided by the work of political geographers, who visualize the division of the country into districts as a map superimposed over a map of the distribution of preferences (Gudgin & Taylor 1979, Johnston et al. 1994, Taylor & Johnston 1979, Taylor et al. 1986). In a two-party situation, the resulting intersection of geographic party preferences and district boundaries can be described by a curve graphing the distribution, which will usually take a normal bell shape. Lopsided districts are found in the tails; more even distributions are more frequent and are in the center of the distribution. The standard deviation is a measure of dispersion, with large standard deviations reflecting relatively fewer evenly balanced ("marginal") districts and more lopsided ones. Taylor et al. (1986) show how the standard deviation is shaped by areal clusters of opinion in relation to the district sizes and boundaries. As smaller standard deviations mean more evenly balanced districts, this kind of map will produce more "wasted votes," coming from the losers in these districts (and will also produce larger seat swings from marginal vote swings, a point to which we return in a moment). The addition of more parties can produce even more wasted votes if it means that the winners are carrying districts with less than 50% of the votes, or if it means that there are fewer lopsided victories of any kind. Geographically concentrated parties may not create that effect if they simply mean uneven local contests between different parties. Of course, as Schattschneider (1942), and no doubt others before him, had been well aware, the problem with achieving nationally proportional outcomes in political representation 277 systems based on geographic districts is what happens to the votes for the losers in each district. In a SMD system in which the districts are reasonably competitively contested and districts are of equal size, nearly half the votes will be cast by losers in their district. If we counted these losing voters as unrepresented, than we would see a very disproportionate vote-seat relationship, a matter that greatly concerned early PR proponents (e.g., Mill 1958 [1861], p. 111). From this point of view, the amount of disproportionality is measured by the aggregation of "wasted votes"—that is, votes for candidates or parties who lost the district election. Following Rae (1967), the standard cross-national studies of disproportionality do not simply aggregate the losing voters from all the districts. Rather, they usually assume that party is a nationally meaningful concept for voters, and national vote-party relations define the appropriate representative linkage. Thus it may be that one party will lose in some districts and the other party will lose in others. If one party's losers are canceled by the other party's equivalent losers, a SMD-based system could yield a highly proportional outcome, even if it has many closely contested districts. In geographic terms, this can happen if the distribution curve is highly symmetrical. However, a skewing of the distribution, such that one party loses a lot of districts by a small margin, will swiftly create much higher disproportionality. Thus, greater proportionality can emerge from SMD systems either from many highly unbalanced districts or from a very symmetrical distribution of partisan losses. Additional parties within districts will usually reduce proportionality by creating more losers. This possibility of disproportionate results in one district being canceled by opposite disproportionality in another will, of course, be present in any system with districts—and most real-world PR systems have some kind of geographic districts. But PR systems with large district magnitudes will have fewer unrepresented losers in each district (subject to the number of parties competing, as explained below), and thus will be less dependent on canceling of disproportionality in different districts. Thus, in larger-magnitude districts, the interaction between geography and vote preferences is less important. In small-magnitude district systems it is critical because it will affect the canceling probabilities as well as the proportion of losers in the average district. (One way of understanding gerrymandering is as a practice of drawing the boundaries to preclude an even balance of winners and losers on both sides.) On the geographic distribution effects, see Taylor et al. (1986), Powell & Vanberg (2000), and Taagepera & Shugart (1989). Lack of equality in the relationship between numbers of voters and numbers of representatives from each district may also affect disproportionality. This lack of equality may stem from the rales of representation themselves (as in the underrep-resentation of urban districts in Spain), which is usually called malapportionment. It may also be caused by differential rates of turnout in different districts, as when a labor party gains proportionately greater representation because lower turnout in urban areas means its victories are based on fewer votes. Political geographers' analysis of bias effects in two-party situations suggests that they can be disaggregated into four elements (see especially Johnston et al. 1999 and references 278 powell therein): turnout, malapportionment, third party, and distributional (intentional or unintentional gerrymanders). Although this work has thus far been dominated by scholars studying Britain and New Zealand, this may be changing, as reported in Snyder & Samuels's (2001) review of malapportionment in Latin America. Too Many Parties Observers have also long been aware, from various perspectives, that "too many" parties is a problem for representation. Schattschneider (1942, 75ff.) suggested that in SMD plurality systems, the first party was advantaged; the second party was slightly disadvantaged; but additional parties had little chance, and their voters were likely to be un(der)represented, unless they were "sectional" parties. Duverger (1954, p. 374) also noted that effects of FPTP election rales's "exaggeration" of swings to the winner may become "much more capricious" under multiparty competition. The effect of this generalization on aggregate dispropor-tionality depends, of course, on the "additional" parties being weak everywhere; if they have local areas of enough strength to win some districts, this may make up for their underrepresentation in others. Taagepera & Shugart (1989) show that the effect of too many parties on proportionality is not limited to FPTP systems. Rather, disproportionality tends systematically to increase with more competing political parties. They also observe that as district magnitude increases, so does number of parties, as we might expect from Duverger's Law [which states that FPTP leads to two-party systems (Duverger 1954, p. 217)] and from Rae's findings about election rales and fractionalization of the party system. The effect of district magnitude on the number of parties creates a somewhat off-setting effect on the decline in misrepresentation that greater magnitudes create. Taagepera & Shugart (1989, p. 123) refer to this as a "law of conservation of D (disproportionality)." Cox (1997) provides an appealing theoretical framework into which to place the work on election rales, number of parties, and disproportionality. Cox draws on a large, purely theoretical literature on strategic voting under different voting rales, as well as on the empirical studies, to construct a model of the "microfoundations" of Duverger's Law. The work of Gibbard (1973) and Satterthwaite (1975) had demonstrated theoretically, as Leys (1959) and Sartori (1968) had suggested from empirical observation, that "strategic voting" (voting for a less preferred party or candidate because it has a better chance of winning) can be rational under any kind of voting system. One way of understanding Duverger's Law (that we expect only two parties under SMD) is as successful coordination to reduce the number of parties to match a reasonable probability of winning the only seat available in a district. Such coordination involves the strategies of party leaders, the election rales, and expectations about voters. In SMD, when more than two parties compete for votes in a district, some kind of coordination failure has occurred, unless balance is very even. Different voting rales create incentives to reduce the number of parties to varying levels. Explicitly, Cox (1997, pp. 31-32, 99ff.) suggests that the number of parties should be reduced to the district magnitude plus one. political representation 279 When coordination fails, and "too many" parties compete relative to the threshold, we see increased levels of misrepresentation (assuming equivalent cross-district canceling effects). Although any voting system will demand some coordination, the coordination task is greater and the costs of coordination failure are higher in low-district magnitude systems. FPTP is the extreme case, in which a single party may receive substantial votes in many districts and yet fail to carry any of them—thus depriving many voters of representation. Cox's formulation explicitly provides a theoretical explanation for Taagepera & Shugart's empirically derived "law of conservation of D (disproportionality)" and the general meaning and consequence of "too many" parties. Perhaps more importantly, it suggests theoretically the conditions (of party objectives, information, expectations, and rales) under which coordination failures should be expected—quite apart from the geographic issues. The striking failures of large-district PR rules to generate proportional outcomes in a number of the new democracies of Eastern Europe, especially when too many parties compete relative to a fairly low threshold (typically 5%) underline the importance of specifying more exactly how information limitations, party goals, and the lack of expectation constraints operate in new systems. (The failures also emphasize that these are genuine empirical theories not, as sometimes asserted, mere tautological implications of PR rales.) Standard Theory of Vote-Seat Representation It now is possible to speak of a standard (though far from complete) theory of vote-seat representation that draws on these empirical and theoretical works. In this theory, the citizens are principals whose preferences are expressed by their (first-preference) vote for political parties. The agents are the collective party representatives in the legislature, regardless of which district chooses them. The comparison is between all the citizen votes and all the legislative seats with various, similar measures of proportionality as the standard of desirable representation. Representation is shaped by the interaction of competing parties, citizen's choices, and election rales. Even in this standard form, the vote-seat paradigm is far from exhausted. Despite some excellent work by geographers, the problem of the fit between party support and geographic boundaries seems incomplete. These features are so important in low-magnitude systems that it is difficult systematically to make progress on other effects until they are in hand. The number of parties, which in turn affects disproportionality, is doubtless shaped by internal party organization features and histories (e.g., Kitschelt 1994) and by the opinion and organizational diversity of the society (Cox 1997, Ch. 11; Ordeshook & Shvetsova 1994; GB Powell 1982), as well as by election rales and coordination problems. The inventions of new types of rules, especially those that combine SMD and PR in various ways, create need for theories of party leaders' incentives that take account of both of these (Moser 1999, Shugart & Wattenberg 2001). Finally, despite some promising leads 280 powell (Lijphart 1994, Ch. 6; Shugart 1995; Shugart & Carey 1992), the analysis of interaction between party and voter behavior in the context of multiple levels of elections, especially presidential and legislative elections, has far to go. Variations on the Basic Vote-Seat Paradigm Some interesting variations are emerging, which partially violate some (in part normative) assumptions of the standard exemplars but which potentially enrich our understanding of representation of votes into politics. One important variation involves explicitly linking votes to executives or policy makers rather than only to legislatures. If we assume that the legislature is the decisive stage in policy making, and that all legislators have equivalent influence, then it may be satisfactory to compare citizen vote distributions with legislative seat distributions. But if we accept that political and institutional features of the policy-making system imply that at least some legislatures play a relatively small role in policy making, which is rather dominated by the executive, then perhaps citizen vote distributions should be compared with the way those votes are represented among those who effectively make the policies. Pinto-Duschinsky (1998, pp. 123-24) has emphasized the overrepresentation of small parties in some parliamentary governments in PR systems; others have drawn attention to the virtually permanent role in government of large Catholic parties in Italy and the Netherlands from the end of World War II until 1994; still others have noted the near permanent exclusion from government of "extremist" parties of both right and left. Taylor & Lijphart (1985) and Powell (2000) have begun to link party votes to representation of parties in cabinets or policy-making groups. Some recent work has also challenged the standard assumption in the comparative literature that proportionality is the appropriate way to represent votes in legislatures. Interestingly enough, the bulk of recent American politics literature has assumed that a different relationship is normatively preferable. What Gelman & King (1994) call "responsiveness" of seat changes to vote changes emphasizes amplification of the impact of party vote shifts (around the party average) on party seat shifts in the legislature. This follows in a tradition in American and British political analysis that emphasizes the values of competitiveness and swing ratios (see also Rae 1967, pp. 26-27, 145). Gelman & King (1994, pp. 544^5) explicitly contrast the American concept of responsiveness in amplifying vote shifts to the lesser responsiveness of proportionality as a value (see also Rae 1967, pp. 100-1, on SMD rales and "magnification" of vote shifts). In one of the few studies explicitly to examine vote-seat connections from both views, Katz (1997) considers proportionality and responsiveness as competing democratic virtues. His comparative analysis finds that they "are influenced by much the same factors, but are inversely related" to each other (Katz 1997, pp. 138-42). Powell (2000, Ch. 6) follows a similar line of thought and suggests that "proportional" and "majoritar-ian" visions of democracy imply different relationships between votes and their legislative (or policy-making) representation (see also Powell & Powell 1978.) political representation 281 Empirically, he finds that proportional and majoritarian (FPTP and other SMD) systems each perform fairly well by "their own" standards, and quite badly by the standards implied by the opposite vision. This is consistent with Katz's finding and with the idea of a trade-off in the consequences of election laws. However, good responsiveness in the majoritarian systems usually requires counting the party that wins the plurality of votes as entitled to unshared governing power, since very few parties win vote majorities. A Fundamental Problem: "Party" The most fundamental empirical problem with using vote-based approaches to democratic representation lies in the assumptions about the nature of the "political party" itself. The aggregate comparison of citizen vote distributions and legislative (or executive or policy-maker) representative distributions assumes that the same party means the same thing to voters in different districts within a country. This assumption is absolutely essential to a procedure that allows the unrepresented party losers in one district to be compensated by overrepresented party winners in another district. If the same party label means something different in the two districts, then the results of "canceling" across districts will be misleading at best. Parties that are merely coalitions of diverse local notables or regional patrons, or that contain contentious policy factions, may fit the fundamental assumption of party homogeneity across districts rather badly. If the party representatives fail to coordinate in their legislative activity, the problem is compounded. The paradigmatic vote-seat studies, and even their variants, provide few clues as to how one might address this problem, or even what features of the vote or party system one might investigate to see whether the assumption is warranted. A related problem is that the vote-based approaches do not discriminate between parties on any other basis than votes received. Misrepresentation measures indicate only disproportionality in vote-seat ratios, not whether parties favoring some kinds of policies are advantaged. If a plurality winner (or even second-place finisher) gains a legislative majority because two other parties have split the vote, we have no idea whether the winner is substantively far from positions favored by the other parties. Yet, the attraction of vote-seat representation research is not merely the sometimes misleading convenience of its available data. The citizens' votes do have a special status. The election is the formal authorization step in the principal-agent relationship. Citizens who are markedly underrepresented relative to their votes may well feel that basic standards of democratic equality have been violated. Moreover, the vote does require citizens themselves to aggregate the various concerns they may bring to the election into a single act. It does not require the researcher to pull together these concerns into some semblance of estimated coherence or to assume that one dimension or substantive area, such as the economy, prevails. In this sense, the election reveals the preferences of the citizens with unique authority. Finally, even researchers who wish to begin with citizens' preferences (see the next section), or to look at governments as well as legislatures, need a theory that 282 powell takes account of the way party votes are aggregated under different election rales. For these reasons, at least, the party relationship between votes and seats remains an important feature of democratic representation. CITIZENS' PREFERENCES AS THE STARTING POINT: THE ISSUE-CONGRUENCE PARADIGM IN COMPARATIVE POLITICS Studies of substantive representation tend to assume that a serious claim of contemporary democracies is that policy makers should and do take account of what the citizens want. From this point of view, vote-based studies of representation capture, at best, only part of the complex processes that link citizens' preferences and the policies of their elected policy makers. We need a theory of voting behavior and empirical research to show the conditions under which policy preferences shape vote choice. Furthermore, even if votes are shaped by substantive policy preferences rather than candidate personalities or individual pay-offs, they are also constrained by the available choices offered by candidates and parties. If some electoral rales and party systems offer more complete sets of choices, the votes may be a better guide to underlying preferences in these systems than in others (Lijphart 1994, p. 97). Thus, we also need a theory of party policy promises in the election campaign and empirical research on party strategies. Miller-Stokes as Exemplar in Substantive Representation Research Miller & Stokes (1963) used a public opinion survey to ascertain the issue positions of citizens in different U.S. congressional districts and linked these to the preferences, perceptions, and behavior of the representatives of those districts. Their seminal article, "Constituency Influence in Congress," offered several alternative models of the empirical requirements "to ensure constituency control," addressed directly and thoughtfully the difficulties created by low levels of citizen information, considered relationships across several dimensions of public opinion, and discussed the role of political parties in the linkage process, other factors shaping the behavior of congressmen, and so forth. Miller & Stokes saw the election as the causal connection that could require representatives to be influenced by citizen opinion. Their empirical analysis indicated, however, that the nature of this electoral connection was different in different policy domains. "The representative relation conforms most closely to the responsible-party model in the domain for social welfare," where the parties usually recruited candidates who differed systematically in their policy stances (Miller & Stokes 1963, p. 371). In this domain, they noted the much greater correlations between district majorities and winners, and negative correlations with losing candidates (Miller & Stokes 1963, pp. 359-60). In the civil rights domain, on the other hand, their primary theoretical solution was to emphasize Congressmen's political representation 283 and challengers' perception and mutual anticipation of the position held by the majority of citizens in the district. A decade later, Achen (1977, 1978) demonstrated that correlation coefficients could be misleading estimates of representational connections. Variations within and across districts could constrain the correlations; high correlations could be generated even though the positions of the representatives and citizens were substantively far apart. Achen (1978, pp. 481-94) called for multiple measures that would measure both absolute distances between the positions of constituencies and representatives ("proximity" and "centrality") and the relative degree to which knowing the citizens' positions allowed prediction of their representatives' positions. The absolute distance measures required, however, that citizen and representative opinions be measured on the same scale, which Miller & Stokes (1963) had not done, and which seemed to preclude comparing absolute district preferences with roll-call behavior (but see LW Powell 1982 for an ingenious solution using perceptions of activists). Achen's point seems undeniable, but its substantive implementation remains controversial. It is often interpreted as the need to use the same question wording in interviewing both citizens and representatives. Although failure to do this certainly raises cautionary flags about representation inferences (as in Miller et al. 1999 and Kitschelt et al. 1999, discussed below), it is by no means certain that the average citizen and her legislator will interpret an identically worded question in the same fashion. Moreover, when citizen responses are aggregated to the level of district, party, or country, they are much more likely to include guesses and error than are elite responses. This difference creates further difficulties in interpreting levels of correspondence. Following a different line of concern, Weissberg (1978) showed that Miller & Stokes's measures of "dyadic" representation, which linked district preferences and district representatives, could greatly misstate "collective" representation of the citizenry as a whole within the legislature as a whole. For example, losing citizens' preferences could be represented by legislators from other districts, as in the assumptions of the standard vote-seat studies. By the late 1970s, the Miller & Stokes article still reigned supreme as the research exemplar for the empirical study of preference representation, but analysis and data collection requirements to extend their findings (longitudinally or comparatively) were severe. No single, simple measure of representation or misrepresentation seemed adequate to describe issue connections. Despite the difficulties, the first comparative fruits of Miller & Stokes's work (often encouraged by them as well as their Michigan colleagues and foreign visitors) began to appear at about the same time as the publication of the methodological strictures. Applications of the Miller-Stokes Exemplar Outside the United States It was soon apparent that most representation research in other mature democracies would break fundamentally with one basic element in the Miller-Stokes exemplar: 284 powell the selection of the district as the element that identifies which citizens should be linked to which legislators in dyadic representation. Instead, the comparative research focused almost exclusively on political party representatives, rather than the district representative, as the appropriately conceptualized agent. This break appears consciously and explicitly in Barnes's Representation in Italy (1977). It was forced in part by the election rules and associated election process in many European countries, in part by the high levels of legislative party discipline in most parliamentary systems. In large, multimember districts, such as those Barnes was describing in Italy, the district's voters are not collectively choosing a single representative but a set of representatives (about 20 per district in Italy). Voters supporting relatively small parties gain (proportional) legislative representation. Empirically and conceptually, the average citizen preference in the district as a whole was not strongly related to the average preferences of its representatives (Barnes 1977, pp. 121-22). Moreover, as Barnes found when looking at the elite questionnaires, in most cases all the representatives of a particular party gave similar opinions, regardless of geographic district (Barnes 1977, p. 119). This combination of multimember districts and relatively homogeneous party behavior led Barnes and most subsequent analysts to focus on the collective legislative party as the appropriate agent of party voters, who are its principals. In a country such as the Netherlands, in which the whole country is a single district, there was little alternative (Irwin & Thomassen 1975). [Converse & Pierce's (1986) prize-winning work does keep the constituency-individual representative dyad, but they recognize the importance of party connections as critical linkages, documenting the highly, although not perfectly, cohesive party voting of the French Assembly members.] Following the suggestion in Miller & Stokes (1963) about American representation in the social welfare domain, as well as the apparently natural logic of multiparty competition in Europe, the comparative representation work has been theoretically inspired primarily by some version of the "responsible party government" model of representation (see, e.g., Dalton 1985, pp. 268-71; Converse & Pierce 1986, pp. 698 ff.; Thomassen 1994, 1999). As Dalton points out clearly, voters choosing between parties offering alternative policy packages create a theoretical model of citizen influence to underpin collective voter-representative correspondence (Dalton 1985, p. 278.) Dalton (1985) studied the degree to which candidates for the European Parliament were representative of their party voters on a variety of substantive issues. This study is an excellent example of providing multiple measures of empirical issue representation, emphasizing closeness of voters and representatives (centrism) as well as correlations and regression coefficients, and collective correspondence. Dalton also found PR election rules and party diversity related to centrism. By the late 1990s, issue correspondence data in the party dyadic version of the Miller-Stokes tradition had been collected for a growing number of national legislatures. Most of the studies presented analysis within single countries: the Netherlands (Irwin & Thomassen 1975), Italy (Barnes 1977), Germany (Farah political representation 285 1980, Porter 1995), France (Converse & Pierce 1986), Australia (McAllister 1991), Britain (Norris 1995), Sweden (Holmberg 1989, Esaiasson & Holmberg 1996), New Zealand (Vowles et al. 1995, 1998), and Norway (Matthews & Valen 1999). [Also see Manion's (1996) remarkable study of semicompetitive elections in China.] They tend to share at least two virtues with their Miller-Stokes exemplar. One is the serious consideration of multiple political issues. They report measurable, but uneven, correspondence between voters and their party representatives across different issues. A second common virtue is concern with the problems of citizen opinion formation, low levels of citizen knowledge and constraint, and how to contrast these with invariably much more sophisticated and structured elite opinions. They offer great insight into the citizen context in each country and sometimes into comparative differences (often contrasting their results with American analyses). On the negative side, however, the touchstone "responsible party government" model has been offered primarily as a normative model and has usually lacked development of what must be an essential part of any empirical model of party linkage: theory and analysis of party electoral strategies. This lack has probably encouraged failure to consider the possible role of convergent party offerings [along the lines of Downs (1957) or Miller-Stokes's civil rights domain]. A second problem that follows directly from the adoption of the Miller-Stokes dyadic analysis, largely ignoring Weissberg's argument, has been a lack of systematic consideration of how appropriately to aggregate the constituency (individual party) connections into comparison of citizenry and legislature as a whole. It is particularly striking that, in contrast to the vote-seat literature, these studies have seldom worried about the implications of disproportionate seat representation of different parties in the legislature for issue representation. Moreover, the issue representation studies (like the vote-seat literature) have focused almost exclusively on parties in the legislature. Miller & Stokes's (1963) original article mentioned presidential influence in the foreign policy domain but gave no consideration to the president as an elected representative of the people. Converse & Pierce's (1986) impressive analysis of representation in the French legislature entirely ignores the democratically elected and influential French president. The studies in parliamentary systems have similarly ignored the distinctive influence of the government. The Explosion of Comparative Substantive Representation Studies After a long period in which single-country studies were the dominant mode of analysis, it is striking that 1999-2000 witnessed the publication of explicitly cross-national comparative studies involving at least five different research programs in substantive representation. [Also see Miller et al. (2000), a study including citizen-representative issue comparisons in Russia, Ukraine, and Lithuania in 1995.] These cross-national studies have been forced to confront the great methodological difficulties in comparative measurement and specification and have tried to analyze the 286 powell effects of system-level features, such as party systems, election rales, and historical context, on issue representation. Most obviously in the Miller-Stokes tradition is the ambitious collection of original analytic essays edited by Miller himself, Policy Representation in Western Democracies, published shortly after his death (Miller et al. 1999). The authors attempt to take advantage of the availability of the emergent group of studies of citizen-legislator representation within five countries explicitly to test cross-national hypotheses derived largely from the "responsible party government" model. However, they conceptualize and measure their representative relationships differently and reach diverse conclusions about, for example, the relative success of more or less structured party systems in creating correspondence on the average issue (see, e.g., Holmberg 1999, p. 94; Pierce 1999, p. 31; Thomassen 1999, pp. 45-51; Wessels 1999, pp. 148-51). It is difficult to tell whether the divergent findings reflect alternative approaches to analysis or the varying, flawed data sets. For example, a number of these studies still use the 1958 Miller-Stokes data set for analyses involving absolute issue distances between citizens and representatives, despite the dissimilarity in the questions asked of citizens and representatives. In another recent work that builds from the Miller-Stokes tradition, Schmitt & Thomassen (1999) provide a normative model of "responsible party government" that creates issue linkages through parties offering coherent policy choices. This model plays a large role in their conception of the development of "European level" representation. Their surveys of citizens and candidates in the 1994 European parliamentary elections asked questions about self-placement on the left-right scale and three issues of European community politics (employment, open borders, and common currency). Their chapter on "issue congruence" (Schmitt & Thomassen 1999, pp. 186-208), which should be read in conjunction with their earlier article (Thomassen & Schmitt 1997), examines the linkages between party voters and candidates of the same parties on these issues. These two presentations show effectively a very strong relative-responsiveness connection between voters and representatives on the left-right scale, created apparently by voter choices and party alternatives (both national- and European-level parties), consistent with the "responsible party model." The candidate positions on the three European issues are also strongly structured by their left-right positions; the voter issue positions are far less so. There is some relative congruence between voters and candidates on the substantive issues (as shown by correlation coefficients, etas, and scatter-grams), shaped both by party and (more strongly) by country of origin (Schmitt & Thomassen 1999, pp. 200-5; Thomassen & Schmitt 1997, p. 175). But the absolute positions of voters and parties are far apart on the European issues, with most of the candidates far more pro-European on borders and common currency than their voters. These results are, of course, consistent with referenda and opinion polling on these issues. Holmberg's (1999) chapter "Wishful Thinking," describing the (mis)perceptions of their voters by Members of the European Parliament, also contains interesting material. A third recent cross-national analysis in the Miller-Stokes tradition appeared in Beyond Westminster and Congress: The Nordic Experience (Esaiasson & Ffeidar political representation 287 2000). Holmberg's (2000, pp. 155-180) chapter on "Issue Agreement" utilizes at least eight issue questions (five-point scale questions asked identically of citizens and legislators) to compare positions of party voters and their Members of Parliament (MPs) in Denmark, Iceland, Norway, and Sweden. Holmberg also presents dichotomized results showing agreement/disagreement of the majorities of party voters and their MPs. Refreshingly, he reports when citizens and legislatures collectively (as opposed to voter-party dyads) have agreeing/disagreeing majorities. Holmberg focuses primarily on differences across the political issues, a familiar theme in this literature. He expects better correspondence between voters and party representatives on "salient and politicized issues at the center of political discourse," and he generally finds it on four issues associated with the "left-right" dimension (also see Thomassen 1994, 1999). There seem to be both good absolute and good relative issue agreement. Still, even on these issues, the majorities of citizens and their party representatives in Denmark, Norway, and Sweden did not correspond in 15 of 72 pairs (21%). What is especially striking in Holmberg's data is the near unanimity among each party's MPs. Virtually all the Left and Social Democratic MPs took leftist positions; virtually all the Conservative and Progress MPs took rightist positions. Although there is a general increase in the average support for conservative positions among citizens as we move from left to right across the parties, the citizens are generally more divided than their representatives. For many parties, this implies that there is a substantial minority of voters for each party who favor positions represented in the legislature exclusively by representatives of other parties. (Many of the individual country studies have also shown this pattern.) The most striking failure of correspondence observed by Holmberg was on immigration policy. Voters of most parties (except for Progress and the Swedish Conservatives) were in sharp disagreement with their MPs. [These results are reminiscent of the European policy results reported by Thomassen and Schmitt (Schmitt & Thomassen 1999, Thomassen & Schmitt 1997).] As the issue has been highly politicized, with several parties taking sharp positions, the discrepancy cannot be attributed to voter ignorance of it. Kitschelt et al. (1999) discuss the emergence of partisan politics in the new democracies of Bulgaria, the Czech Republic, Hungary, and Poland in the mid 1990s. Although substantive representation is only one theme in their larger analysis (Kitschelt et al. 1999, primarily pp. 309^4), the theoretical formulation and methodological implementation is impressive. (However, their data contain some serious limitations, especially because the issue-distance measures use standardization of different questions asked of citizens and elites.) Kitschelt et al. take to heart Achen's (1978) argument that both absolute (distance) and relative (responsiveness) properties of representation should be considered. In the comparative tradition, they apply these to party voter-representative dyads rather than to geographic districts, and they calculate centrism distances, intercepts, and slopes, as well as correlations as a goodness-of-fit measure. They do not forget to take account of party size in using the statistics as aggregate properties of representation for each issue (and the left-right scale) in each country. 288 powell Theoretically, Kitschelt et al. (1999) argue that any significant, positive regression slope (relative responsiveness) constitutes democratic representation. Low absolute distances between average party voter and representative, slopes close to 1.0, and small intercepts (which mean close proportional correspondence on issue positions between party voters and party representatives) imply perfect "mandate representation." They also suggest two other patterns, "polarized trusteeship" and "moderating trusteeship," that can be important alternative forms of democratic representation. In polarized trusteeship, the party representatives are more extreme than their respective voters; in moderating trusteeship, the representatives are more centrist than their supporters. Rather than assuming any of these as the correct normative baseline, they suggest the consequences of each for such aspects of democratic performance as citizen mobilization, effective policy leadership, and intensification or diffusion of political conflict (Kitschelt et al. 1999, pp. 80-88, 340). Kitschelt et al. analyze the origins of observed representative patterns on different issues and countries in terms of the differing legacies of Communist rule found in the different countries. Beyond the cross-national differences associated with these legacies, in general they find strong patterns of party linkage, with polarizing trusteeship especially notable on salient political issues. Moreover, they go on to analyze the distances between governing parties and the mean and median voter, contrasting "universalistic" correspondences with larger distances associated with a polarized trusteeship form of "responsible party government" (Kitschelt et al. 1999, pp. 329-36). Finally, Powell (2000) examines both procedural and substantive representation in some 20 democracies over the past 25 years, focusing on images and practices in "majoritarian" and "proportional" constitutional designs. The last third of the book analyzes substantive representation at the whole-system (collective) level, using only the left-right scale, which is assumed to be a summary of various specific issues that are salient in the discourse of competition of each country. (Also see associated work in Huber & Powell 1994, Powell & Vanberg 2000.) The work explicitly builds from theories of strategic party competition and government formation in majoritarian and proportional systems. The empirical analysis tries to show where expectations about party connections are realized and where they break down. Powell & Vanberg (2000) show significantly better correspondence of legislative and citizen medians in low-threshold PR than in higher-threshold PR or, especially, SMD election rales. (They also found this PR advantage using party manifestos to estimate party left-right positions.) The poorer correspondence in the SMD systems apparently resulted from the frequent failure of one or both large parties to converge to the median. Because the number of parties seldom reduced to only two (contra Duverger's Law), this outcome was consistent with Cox's prediction of greater coordination problems with these rules (Cox 1997, Ch. 12). However, it is also implied by the "Mays' Law" line of research and the analysis of internal party incentives that argues patterns of recruitment will generally produce political representation 289 party representatives who are more extreme than party voters (e.g., Kitschelt 1989, May 1973, Norris 1995). When the disproportional election outcome creates an absolute legislative majority for one of these large parties, the result is a legislative median well off the citizen median. Powell (2000) goes beyond the legislature to show significantly better correspondence between the citizen median and the cabinet government in the proportional systems (as do Huber & Powell 1994). Both problems and opportunities abound if the new cross-national data sets are to be fully exploited and the conflicting approaches and findings of the 1999/2000 studies are to be unified and reconciled with each other and previous work. Wessels, Kitschelt et al., and Powell claim that we can theoretically explain differences in the causes and consequences of substantive representation. It remains to be seen whether their optimism will be justified. The other authors are much less sanguine. Structuring Citizen Preferences Public opinion studies in many democracies have shown that even in educated societies with developed partisan discourse, citizens tend to have only weakly structured preferences. Our ability to predict a citizen's opinion on one issue by knowing his/her opinions on other issues is fairly limited, and the connection between the issue preferences and party preferences is typically not very strong, although it is shaped by partisan competition itself (Granberg & Holmberg 1988). Nonetheless, through aggregation in electoral competition, moderately one-dimensional citizen opinion structures can be linked to partisan representation. 1. The dyadic party-issue congruence studies suggest that relative issue representation (responsiveness) is greater on the issues more strongly linked to the general dimension of party competition. Indeed, this has become virtually a stylized fact (e.g., Holmberg 2000, Kitschelt et al. 1999, McAllister 1991). 2. However, the comparative literature has paid little attention to issue positions that are eschewed by all the parties because they are unacceptable to voters generally, as Miller & Stokes (1963) suggested was the case with the civil rights domain. Under these circumstances, we might find little relative responsiveness on the issue, as the parties do not offer contrasting choices, yet close proximity of voter and representative positions. Influence would emerge from party anticipation of voter response rather than as outcomes from voter choices. Downsian models would expect this in two-party systems, but it could also emerge in multiparty systems if few voters favor a position. [Empirically, the pornography issue in Holmberg's (2000) Nordic systems seems to look something like this.] 3. The studies of issue preferences and partisanship repeatedly suggest that the preferences and positions of party representatives are usually much more structured (i.e., predictable from their partisanship across a wider range of issues) than the preferences of average citizens. In consequence, representatives' opinions on some issues are strongly linked to their partisanship 290 powell whereas citizens' opinions are not, a discrepancy that may create serious misrepresentation on these issues. Voters for one party may be better represented by MPs of another party. The long-run electoral dynamics of this situation are unclear (but see Holmberg 1997 for an exploration in Sweden). Moreover, because of their greater consistency, the opinions of representatives will often be, or appear, much more extreme than those of their voters. There may be high levels of relative "responsiveness," such that party representatives are differentiated from each other in the same way their supporters are, but the representatives are much more consistently "left" or "right" than their respective followers. This pattern, which is what Kitschelt et al. (1999) call a "polarized trusteeship" connection, appears in many of the empirical studies (e.g., Holmberg 2000). Consistency across related issues, intensity of preferences, and substantive distance from the "center" on single issues can all contribute to this pattern, but they have different implications for advocacy, conflict resolution, and policy satisfaction. Thus, whether low levels of specific citizen information are a problem for representation through elections seems to depend on whether citizens can participate in, or relate in a considered way to, a general partisan, programmatic discourse. Emergence and penetration of such a discourse probably depends on both elites and citizens (Zaller 1992). As Kitschelt (2000) argues, it is up to the political parties to bundle together specific issues so as to make general packages of policy alternatives available to citizens (Kitschelt 2000; similarly, see Esaiasson & Holmberg 1996 on "representation from above"). Aldrich (1995) suggests that this is one of the two fundamental functions of political parties. Reflecting on early experiences of democracy, Bryce (1921, 1:119) observed that no large democracy had been able to do without political parties as the vehicle for organizing and structuring elections. However, parties' incentives to do this are strengthened or weakened by the presence or absence of personalistic elements in the election rules, nationalization of communication patterns and information, and power relationships in the society (Kitschelt 2000). Research on the emergence of a national partisan discourse must be an item on the agenda of substantive political representation studies. A Fundamental Problem: Multidimensional Preference Representation Perhaps the most serious lacuna in comparative studies of preference representation is the absence of both theory and empirical analysis in multidimensional context. In terms of policy advocacy, one would expect that PR election rules with large magnitudes and multiple political parties should be able to represent a multidimensional citizen preference configuration in the legislature. Thomassen's (1999, p. 46) analysis of the self-placement of Dutch voters and MPs on economic policy and abortion issues suggests just this kind of multidimensional correspondence. In a more limited way, we see some specific parties apparently reflecting such political representation 291 multidimensionality on the immigration (Left and Progress parties) and pornography (Christian parties) issues in the Nordic countries (Holmberg 2000). It would seem to be more difficult for a system with small magnitudes and few parties to do this (see McAllister 1991 on the Australian configuration). However, we lack systematic comparative empirical studies. When we move beyond policy advocacy to policy choice, the situation is even murkier. Theoretically, social choice theory seems to show that under majority voting rales a multidimensional preference configuration can yield almost any configuration of outcomes. This situation seems to imply not only instability in the processes but lack of a normatively preferred outcome against which to assess empirical results. Additional Representative Connections Limited space precludes discussion of several other, less extensive research programs that have investigated other elements that shape citizen-policy maker connections. One of these is analysis of "social" or "symbolic" representation, which focuses on demographic correspondences, such as gender, race, and class, between citizens and their representatives (e.g., Norris 1996, Vowles et al. 1998). Another is the analysis of "accountability," emphasizing conditions under which voters penalize policy makers for poor performance (variously measured) in office (e.g., Cheibub & Przeworski 1999; Powell & Whitten 1993; Przeworski et al. 1999). Moreover, a full analysis of democratic representation must incorporate the other forces that shape the making and implementation of public policies, as well as citizens' responses to these. Research on the role of campaign commitments, such as candidate promises and party manifestos, in government policy making (Klingemann et al. 1994, Royed 1996, Stokes 1999, Thomson 2001) has begun to elaborate these stages of representation. Studies of the impact of political parties on public policies in democracies also constitute a large research program that merits extensive consideration in any complete treatment of democratic representation (see Imbeau et al. 2001 for the references and the meta-analysis of the findings of such studies). CONCLUDING COMMENTS The electoral system studies and the policy representation studies have each made "progress" in the way that we might expect from the paradigm/research program view of scientific progress (Kuhn 1970, Lakatos 1978). That is, they have made progress by converting a general problem to very specific, self-defined, self-limited research programs. The vote-seat studies have not worried about how to estimate voter preferences and the intricacies of greater structure in elite preferences. They have made progress on their own terms by assuming that all we want to know about citizens is how they voted, and examining the consequences of those votes interacting with rules, geography, and partisan competition. Issue correspondence studies, on the other hand, have seldom worried about election rales and collective 292 powell preference aggregation; the simplifications of dyadic correspondence have allowed them to concentrate on the problems of preference comparisons. Substantively, our understanding of the larger problem of political representation, even constrained to the empirical connections between voters and legislators, eventually needs to incorporate results from both lines of research. We can see that vote-seat analyses that take no account of the differing meanings of "party" to different voters, or that do not differentiate between centrist and extremist parties, are neglecting some important questions and contexts. We can see that dyadic issue comparisons that ignore artificial and distorted legislative majorities are missing important parts of the larger picture. Examination of the Australia, Britain, and New Zealand studies suggests not only that the respective Labour, Conservative, and National MPs are more extreme than their own voters on many issues, especially those linked to the primary "left-right" dimension of competition, but also that vote-seat distortion gave their parties (disciplined) majorities even further from the median citizen (see tables in McAllister 1991, Norris 1995, Vowles et al. 1995). The shift in New Zealand to PR in 1996 did not greatly change the relationship between party voters and their representatives, but it greatly lessened the distortion of party representation and apparently improved the fit between the legislative median as a whole and the average voter on issues associated with left-right competition (see Vowles et al. 1998, pp. 2,143-50). Overviews of political representation and even, eventually, the two research programs themselves will need to incorporate elements of both programs. Moreover, each will need fully developed theories of party competition and of voter choice, including understanding the conditions for integrating multiple issues into a unified political discourse. Similarly, thus far, comparative studies have offered surprisingly little additional articulation of the normative models that created the interest in democratic representation in the first place. However, it is unclear at what point this merging and increased normative range will be helpful for the dynamics of research. Progress on vote-seat representation has probably benefited from the simple assumption that proportional correspondence between votes and seats is a desirable standard. The vote-seat paradigm overlooks normative alternatives, such as majori-tarian responsiveness, and ignores the fact that most legislatures make policy (and choose governments) by using a variety of complex voting procedures that cover the range from proportionality to majoritarianism. Complex voting rules, party competition strategies, and geographic configurations have created a sufficiently hard problem when linked even to simple dependent variables. Is it yet time to move beyond them? The thread in the issue correspondence studies that considers "responsible party government" as a normative ideal, rather than as one form of partisan competition that can create correspondence, seems thus far to have been more of a distraction than a help. Research progress will be determined by the hard work of the scholars "on the ground," not by reviewers. It is the working scholars who will choose where to go next. It is sufficient here to report the continuing progress and vigorous possibilities in these two research programs contributing to the study of political representation. POLITICAL REPRESENTATION 293 The Annual Review of Political Science is online at http://polisci.annualreviews.org LITERATURE CITED Achen C. 1977. Measuring representation: perils of the correlation coefficient. Am. J. Polit. Sei. 21:805-15 Achen C. 1978. Measuring representation. Am. J. Polit. Sei. 22:477-510 Aldrich J. 1995. Why Parties? Chicago: Univ. Chicago Press Barnes SH. 1977 .Representation in Italy: Institutionalized Tradition and Electoral Choice. Chicago: Univ. Chicago Press Blais A, Carty RK. 1987. The impact of electoral formulae on the creation of majority governments. Elect. Stud. 6:209-18 Benoit K. 2001. District magnitude, electoral formula and the number of parties. Eur. J. Polit. Res. 39:203-24 Bryce J. 1921. Modern Democracies. 2 vols. New York: Macmillan Cheibub IA, Przeworski A. 1999. Democracy, elections and accountability for election outcomes. In Democracy, Accountability, and Representation, ed. A Przeworski, S Stokes, B Manin, pp. 222-49. New York: Cambridge Univ. Press Converse PE, Pierce R. 1986. Political Representation in France. Cambridge, MA: Harvard Univ. Press Cox GW. 1991. SNTV and d'Hondt are "equivalent." Elect. Stud. 10:118-32 Cox GW. 1997. Making Votes Count: Strategic Coordination in the World's Electoral Systems. Cambridge, UK: Cambridge Univ. Press Cox GW, Shugart MS. 1991. Comment on Gallagher's "Proportionality, disproportionality and electoral systems." Elect. Stud. 10:348-52 Dahl RA. 1989. Democracy and Its Critics. New Haven, CT: Yale Univ. Press Dalton R. 1985. Political parties and political representation: party supporters and party elites in nine nations. Comp. Polit. Stud. 18:267-99 Downs A. 1957. An Economic Theory of Democracy. New York: Harper and Row Duverger M. 1954. Political Parties: Their Organization and Activity in the Modern State. Transl. B North, R North. New York: Wiley Esaiasson P, Heidar K. 2000. Beyond Westminster and Congress: The Nordic Experience. Columbus: Ohio State Univ. Press Esaisson P, Holmberg S. 1996. Representation from Above: Members of Parliament and Representative Democracy in Sweden. Aldershot: Dartmouth Publ. Farah BG. 1980. Political representation in West Germany: the institution and maintenance of mass-elite linkages. PhD thesis, Univ. Mich., Ann Arbor Gallagher M. 1991. Proportionality, disproportionality, and electoral systems. Elect. Stud. 10:33-51 Gallagher M. 1992. Comparing proportional representational electoral systems: quotas, thresholds, paradoxes and majorities. Br. J. Polit. Sei. 22:469-96 Gelman A, King G. 1994. Enhancing democracy through legislative redistricting. Am. Polit. Sei. Rev. 88:541-59 Gibbard A. 1973. Manipulation of voting schemes: a general result. Econometrica 41:587-601 Granberg D, Holmberg S. 1988. The Political System Matters: Social Psychology and Voting Behavior in Sweden and the United States. New York: Cambridge Univ. Press Grofman B. 1983. The measures of bias and proportionality in seats-votes relationships. Polit. Methodol. 9:295-327 Grofman B, Lijphart A, eds. 1986. Election Laws and Their Political Consequences. New York: Agathon Gudgin G, Taylor PI. 1979. Seats, Votes and the Spatial Organization of Elections. London: Pion Holmberg S. 1999. Collective policy 294 POWELL congruence compared. See Miller et al. 1999, pp.87-109 Holmberg S. 1997. Dynamic opinion representation. Scand. Polit. Stud. 20:265-83 Holmberg S. 2000. Issue agreement. See Esa-iasson & Heidar 2000, pp. 155-80 Holmberg S. 1989. Political representation in Sweden. Scand. Polit. Stud. 12:1-36 Holmberg S. 1999. Wishful thinking among European parliamentarians. See Schmitt & Thomassen 1999, pp. 235-54 Huber JD, Powell GB. 1994. Congruence between citizens and policymakers in two visions of liberal democracy. World Polit. 46:291-326 Imbeau L, Petry F, Laman M. 2001. Left-right party ideology and government policies: a meta analysis. Eur. J. Polit. Res. 40:1-29 Irwin GA, Thomassen J. 1975. Issue-consensus in a multi-party system: voters and leaders in the Netherlands. Acta Polit. 10:389-420 Johnston RJ, Pattie C, Fieldhouse E. 1994. The geography of voting: regions and the declining importance of the cube law. In Labour's Last Chance, ed. A Heath, R Jowell, J Curtis, pp. 255-74. Aldershot: Dartmour Johnston RJ, Rossiter D, Pattie C. 1999. Integrating and decomposing the sources of partisan bias: Brookes' method and the impact of redistricting in Great Britain. Elect. Stud. 18:367-78 Katz R. 1997. Democracy and Elections. New York: Oxford Univ. Press Kitschelt H. 1989. The internal politics of parties. Polit. Stud. 37:400-21 Kitschelt H. 1994. The Transformation of European Social Democracy. New York: Cambridge Univ. Press Kitschelt H. 2000. Linkages between citizens and politicians in democratic polities. Comp. Polit. Stud. 33:845-79 Kitschelt H, Mansfeldova Z, Markowski R, Toka G. 1999. Post-Communist Party Systems: Competition, Representation and Inter-Party Cooperation. New York: Cambridge Univ. Press Klingemann H, Hofferbert RI, Budge I. 1994. Parties, Policies and Democracy. Boulder, CO: Westview Kuhn TW. 1970. The Structure of Scientific Revolutions. Chicago: Univ. Chicago Press Lakatos I. 1978. The Methodology of Scientific Research Programmes. Cambridge, UK: Cambridge Univ. Press Leys C. 1959. Models, theories and the theory of political parties. Polit. Stud. 7:127-46 Lijphart A. 1985. The field of electoral systems research: a critical survey. Elect. Stud. 4:3-14 Lijphart A. 1986. Degrees of proportionality of proportional representation formulae. See Grofman & Lijphart 1986, pp. 170-79 Lijphart A. 1990. The political consequences of electoral laws, 1945-85. Am. Polit. Sci. Rev. 84:481-96 Lijphart A. 1994. Electoral Systems and Party Systems: A Study of Twenty-Seven Democracies, 1945-1990. New York: Oxford Univ. Press Lijphart A, Gibberd RW. 1977. Thresholds and payoffs in list systems of proportional representation. Eur. J. Polit. Res. 5:219-30 Loosemore J, Hanby V. 1971. The theoretical limits of maximum distortion. Br. J. Polit. Sci. 1:467-77 Manion M. 1996. The electoral connection in the Chinese countryside. Am. Polit. Sci. Rev. 90:736^8 Matthews DR, Valen H. 1999. Parliamentary Representation: The Case of the Norwegian Storting. Columbus: Ohio State Univ. Press May JD. 1973. Opinion structure of political parties: the special law of curvilinear disparity. Pott. Stud. 21:136-51 McAllister I. 1991. Party elites, voters and political attitudes: testing three explanations of mass-elite differences. Can. J. Polit. Sci. 24:237-68 Mill JS. 1958 (1861). Considerations on Representative Government, ed. CV Shields. Indianapolis: Bobbs-Merrill Miller AH, Erb G, Reisinger W, Hesli VL. 2000. Emerging party systems in post-Soviet societies: fact or fiction? /. Polit. 62:455-90 Miller WE, Pierce R, Thomassen J, Herrera R, POLITICAL REPRESENTATION 295 Holmberg S, et al. 1999. Policy Representation in Western Democracies. Oxford, UK: Oxford Univ. Press Miller WE, Stokes D. 1963. Constituency influence in Congress. Am. Polit. Sci. Rev. 57:165-77 Moser RG. 1999. Electoral systems and the number of parties in postcommunist states. WorldPolit. 51:359-84 Norris P. 1995. May's law of curvilinear disparity revisited . Party Polit. 1:29-41 Norris P. 1996. Legislative recruitment. lnCom-paring Democracies: Elections and Voting in Global Perspective, ed. L LeDuc, R Niemi, P Norris, pp. 184-215. London: Sage OrdeshookP, ShvetsovaO. 1994. Ethnic heterogeneity, district magnitude, and the number of parties. Am. J. Polit. Sci. 38:100-23 Pierce R. 1999. Mass-elite issue linkages and the responsible party model of representation. In Policy Representation in Western Democracies, ed. WE Miller, pp. 9-32. Oxford, UK: Oxford Univ. Press Pinto-Duschinsky M. 1998. Send the rascals packing: defects of proportional representation and the virtues of the Westminster model. Times Lit. Suppl., Sep. 25, pp. 10-12 Pitkin HF. 1967. The Concept of Representation. Berkeley: Univ. Calif. Press Porter SR. 1995. Political representation in Germany: effects of candidate selection committees. PhD thesis. Univ. Rochester Powell GB. 1982. Contemporary Democracies: Participation, Stability and Violence. Cambridge, MA: Harvard Univ. Press Powell GB. 2000. Elections as Instruments of Democracy: Majoritarian and Proportional Visions. New Haven, CT: Yale Univ. Press Powell GB with Powell LW. 1978. The analysis of citizen-elite linkages: representation by Austrian local elites. In The Citizen and Politics: A Comparative Perspective, ed. S Verba, LW Pye, pp. 197-218. Stamford, CT: Greylock Powell GB, Vanberg G. 2000. Election laws, disproportionality and the left-right dimension. Br. J. Polit. Sci. 30:383^111 Powell GB, Whitten GD. 1993. A cross-national analysis of economic voting: taking account of political context. Am. J. Polit. Sci. 37:391^14 Powell LW. 1982. Issue representation in Congress. /. Polit. 44:658-78 Przeworski A, Stokes SC, Manin B. 1999. Democracy, Accountability, and Representation. New York: Cambridge Univ. Press RaeD. 1967,1971. The Political Consequences of Electoral Laws. New Haven, CT: Yale Univ. Press Royed TJ. 1996. Testing the mandate model in Britain and the United States. Br. J. Polit. Sci. 26:45-80 Sartori G. 1968. Political development and political engineering. In Public Policy, ed. JD Montgomery, AO Hirschman. Cambridge, UK: Cambridge Univ. Press Satterthwaite MA. 1975. Strategy-proofhess and Arrow's conditions. /. Econ. Theory 10:1-7 Schattschneider EE. 1942. Party Government. New York: Holt, Rinehart and Winston Schmitt H, Thomassen J. 1999. Political Representation and Legitimacy in the European Union. Oxford, UK: Oxford Univ. Press Shugart MS. 1995. The electoral cycle and institutional sources of divided presidential government. Am. Polit. Sci. Rev. 89:327-43 Shugart MS, Carey JM. 1992. Presidents and Assemblies: Constitutional Design and Electoral Dynamics. New York: Cambridge Univ. Press Shugart MS, Wattenberg MP. 2001. Mixed-Member Electoral Systems: The Best of Both Worlds? New York: Oxford Univ. Press Snyder R, Samuels D. 2001. Devaluing the vote. J.Democr. 12:146-59 Stokes SC. 1999. What do policy switches tell us about democracy. See Przeworski et al. 1999, pp. 98-130 Taagepera R. 1986. Reformulating the cube law for proportional representation elections. Am. Polit. Sci. Rev. 80:489-504 Taagepera R, Shugart MS. 1989. Seats and Votes: The Effects and Determinants of 296 POWELL Electoral Systems. New Haven, CT: Yale Univ. Press Taylor PJ, Gudgin G, Johnston RJ. 1986. The geography of representation. See Grofman & Lijphart 1986, pp. 183-92 Taylor PJ, Johnston RJ. 1979. The Geography of Elections. New York: Holmes and Meier Taylor PJ, Lijphart A. 1985. Proportional tenure vs. proportional representation. Eur. J. Polit. Res. 13:387-99 Thomassen J. 1994. Empirical research into political representation. In Elections at Home and Abroad, ed. MK Jennings, TE Mann, pp. 237-64. Ann Arbor: Univ. Mich. Press Thomassen J. 1999. Political communication between political elites and mass publics: the role of belief systems. See Miller et al. 1999, pp.33-58 Thomassen J, Schmitt H. 1997. Policy representation. Eur. J. Polit. Res. 32:165-84 Thomson R. 2001. The programme to policy linkage: the fulfillment of election pledges on socio-economic policy in the Netherlands, 1986-1998. Eur. J. Polit. Res. 40:171-97 Tufte ER. 1973. The relationship between seats and votes in two-party systems. Am. Polit. Sei. Rev. 67:540^7 Vowles J, Aimer P, Banducci S, Karp J. 1998. Voters' Victory? New Zealand's First Election Under Proportional Representation. Auckland: Univ. Auckland Press Vowles J, Aimer P, Catt H, Lamare J, Miller R. 1995. Towards Consensus? The 1993 Election in New Zealand and the Transition to Proportional Representation. Auckland: Univ. Auckland Press Weissberg R. 1978. Collective versus dyadic representation in Congress. Am. Polit. Sei. Rev. 72:535^7 Wessels B. 1999. System characteristics matter: empirical evidence from ten representation studies. See Miller et al. 1999, pp. 1137-61 Zaller J. 1992. The Nature and Origin of Mass Opinion. New York: Cambridge Univ. Press \JX\Annual Review of Political Science Volume 7, 2004 Contents Not Your Parents' Political Socialization: Introduction for a New Generation, Virginia Sapiro l Downs and Two-Party Convergence, Bernard Grofman 25 - 3 "Business" Is Not an Interest Group: On the Study of 11 Companies in American National Politics, David M. Hart 47 cd o 11 Citizenship and Education, Eamonn Caiian 71 11 Latino Politics, Rodoifo O. de la Garza 91 g 2 Global Media and Politics: Transnational Communication S 3 Regimes and Civic Cultures, w. Lance Bennett 125 -o I Actors, Norms, and Impact: Recent International I § Cooperation Theory and the Influence of the II Agent-Structure Debate, Kate O'Neill, Jorg Baisiger, id3 and Stacy D. VanDeveer 149 ON ^ s ■ States as Laboratories: A Reprise, Sarah M. Morehouse ^ > and Malcolm E. Jewell 111 § d Measuring the Effects of Televised Political Advertising ^ I in the united states, Kenneth Goldstein and Travis N. Ridout 205 Oakeshott and Political Science, Kenneth Minogue 221 What Does Political Economy Tell Us About Economic Development—and Vice Versa?, Philip Keefer 247 Political Representation in Comparative Politics, G. Bingham Powell, Jr. 273 A Critical Guide to Bush v. Gore Scholarship, Richard L. Hasen 291 Public Deliberation, Discursive Participation, and Citizen Engagement: A Review of the Empirical Literature, Michael X. Delli Carpini, Fay Lomax Cook, and Lawrence R. Jacobs 315 Advocacy and Scholarship in the Study of International War Crime Tribunals and Transitional Justice, Leslie Vinjamuri and Jack Snyder 345 o 3 0- « 0Ü -§ ix x contents Martin Shapiro and the Movement from "Old" to "New" institutionalist studies in public law scholarship, Howard Gillman 363 The Centrality of Race in American Politics, Vincent L. Hutchings and Nicholas A. Valentino 383 Models of Vetoes and Veto Bargaining, Charles Cameron and Nolan McCarty 409 Democratic Politics in Latin America: New Debates and Research Frontiers, Gerardo L. Munck 437 go^ Direct Democracy: New Approaches to Old Questions, ° § Arthur Lupia and John G. Matsusaka 463 11 Bayesian Analysis for Political Research, Simon Jackman 483 Indexes Subject Index Cumulative Index of Contributing Authors, Volumes 1-7 Cumulative Index of Chapter Titles, Volumes 1-7 Errata An online log of corrections to Annual Review of Political Science chapters may be found at http://polisci.annualreviews.org/ 507 523 525