Inequality, Collective Action, and Democratization Philip Keefer, World Bank ............................................................................................................................................................................................................................................... E pic redistributive struggles between the rich and poor lie at the heart of prominent theories of eco- nomic development and the emergence of democ- racy (e.g., Boix 2003; Acemoglu and Robinson 2006). The poor pursue democracy to secure cred- ible redistribution away from wealthy elites; elites, fearing redistribution, but also the costs of revolution, decide whether to repress these efforts or to surrender to them. These theo- ries, and the historical examples of working classes exacting redistributive or political concessions from elites, have been interpreted as suggesting that inequality and redistributive struggles should be central features of development and democratization. Where inequality is high, democracy should be unlikely to emerge, or to emerge and be unstable. Because elites in unequal societies are unwilling to adopt institutions that encourage growth and investment (such as institutions that protect non-elites from predation by elites), incomes should be lower as well. More recent country experiences do not seem to follow this pattern, however. In contrast to historical experience, many contemporary democracies seem to have emerged without a contest between the organized rich and poor. Correspond- ingly, redistributive contests between the rich and poor are not particularly notable in many new or poor democracies. Finally, poor democracies, which are also the least stable, do not exhibit particularly high income inequality, relative either to stable, rich democracies or to stable, poor autocracies. This article expands on these puzzles and suggests that a promis- ing strategy to explore them involves examining variations across countries in the degree to which citizens can mobilize for collective action. The ability to act collectively is an explicit assumption of the analyses in Boix (2003), Acemoglu and Robinson (2006), and others. They are clear that their predictions about income distribution and democratization are contingent on elites and non-elites each having the capacity to undertake collective action. Most investigations of these predictions maintain, and do not further examine, this assumption. However, the most striking characteristic of poorer, less stable democracies may not be their greater inequality relative to other countries, as theory suggests would be the case were non-elites and elites capable of collective action. Instead, it may be precisely that large social groups in these countries confront larger obsta- cles to collective action. One condition that favors collective action is the ability of group leaders to mobilize group mem- bers with promises that members will benefit if the mobiliza- tion is successful. Evidence presented below on political parties in poorer and richer democracies suggests that this ability is particularly lacking in the former. A focus on political organization and collective action, while entirely consistent with past research on democratization and development, opens up new lines of inquiry. What drives democratization and instability if large social groups cannot act collectively? Is it inequality, or something else, that explains why collective action is easier in some settings than in others? Or, on the contrary, do obstacles to collective action promote inequality (e.g., by preventing the emergence of institutions that would open economic opportunities to all )? This article offers evidence that these are likely to be important issues in future research on development and democratization. PUZZLES IN INEQUALITY AND DEMOCRATIZATION The argument that democratization is a function of inequal- ity and fears of redistribution, conditional on elites and non- elites being able to act collectively, confronts three puzzles. The first is that evidence on recent experiences with regime change, out of or into democracy, shows little evidence of it. One might expect, for example, that if redistributive conflict underlies democratization, then new democracies should be disproportionately governed by left-wing governments, which are most likely to represent the collective interests of the poor, and new non-democracies by right-wing governments. Table 1, from Keefer (2007a), indicates that this is not typically the case. Of the 36 instances over the period 1975–2004 in which a country went from being governed by competitively elected leaders to non-competitively elected leaders, more of the tran- sitions were to left-wing governments (12 of 32) than to right- wing governments (four of 32). In 91 transitions to democracy, more of the newly democratic governments were right wing (31) than left (25). The second puzzle is that redistributive conflict between elites and non-elites does not seem to be the main issue driv- ing political mobilization. Kaufman (2009) surveys evidence from Latin America indicating, despite high inequality, a sur- prisingly low correlation between economic class and parti- san preference. In some cases, middle-class voters are more likely to vote for left-wing parties than are poor voters; in other cases, apparently left-wing parties woo the poor with cli- entelist rather than broadly redistributive appeals. In any case, competitive elections have not ushered in the massive redistribution from rich to poor, the threat of which is thought to prevent democratization in high-inequality coun- tries: taxes, particularly progressive income taxes, are low in Latin America. Moreover, recent efforts in some countries of ............................................................................................................................................................................................................................................... SYMPOSIUM ............................................................................................................................................................................................................................................... doi:10.1017/S1049096509990035 PS • October 2009 661