Chapter Two: Distributive Conflict Transitions

Transcription

1 Chapter Two: Distributive Conflict Transitions (from Stephan Haggard and Robert R. Kaufman, Inequality, Distributive Conflict and Regime Change, February 22, 2014) As noted in the preceding chapter, distributive conflict models of regime change have generally sidestepped issues of collective action. Acemoglu and Robinson devote only ten pages of their classic book to this issue (2009, pp ) and generally model citizens as a unified rational actor. Boix (2003) does not focus on the issue at all. In this chapter, we focus on the distributive conflict transitions in our sample: those in which citizens overcome barriers to collective action to successfully challenge incumbent authoritarian regimes. In discussing elite-mass dynamics in these transitions, two core causal factors come into play. The first is the ability of publics to overcome collective action problems. What are the factors structural, organizational, cultural that enable publics to mobilize and raise the cost of repression for authoritarian elites? The second issue is the role of political counterelites who often enter the political scene in the wake of spaces initially opened through mass mobilization. Who are the leaders empowered through mass protests? What is their connection to underlying civil society groups, and how do those connections affect negotiations over the new rules of the political game? As we saw in the preceding chapter, inequality and the gains from equalizing reforms under democratic rule do not appear to be either a necessary nor sufficient causal factor in explaining collective action. The weak effect of inequality on transitions does not mean grievances are unimportant; we simply must look for them elsewhere. We present both statistical and qualitative evidence that aggregate economic performance is a more significant source of grievance and mass mobilization than inequality. In general, however, we follow the literature on civil wars and contentious politics in noting that grievances are a necessary but by no means sufficient basis for explaining collective action (for example, Collier and Hoeffler 2004). We start with a more traditional sociological literature that emphasizes the importance of social structure, class organizations including both unions and left parties--and repertoires of collective action (Moore 1966; Rueschemeyer, Stephens and Stephens 1992). Our statistical work shows that both social structural factors, such as the share of manufacturing and industry in GDP, as well as organizational factors such as union density affect the likelihood of distributive conflict transitions, but not their non-distributive counterparts. When we turn to our qualitative analysis, it becomes clear that transitions driven by these sorts of class organizations often with the participation of other civil society groups constitute only one social foundation for mass mobilization. The other path derives from ethno-nationalist mobilization. However, we show that ethno-nationalist mobilization benefits not only from the focal point of identity but from a variety of organizational factors as well. The second major question is how mass mobilization affects elites hold on power. In the statistical analysis, we take a first cut at this question by identifying circumstances in which incumbent political elites are less capable of exercising coercive options. Poor 1

2 economic performance, which is associated with distributive transitions in some of the models, can be not only a source of grievance but also of a declining capacity to maintain military loyalty and the capacity to repress. Institutional factors may also affect the ability of authoritarian incumbents to coopt oppositions and avoid pressures from below (Ghandi and Przeworksi 2007; Svolik 2013). We find evidence that authoritarian regimes with multiparty legislatures are less prone to distributive conflict transitions, and that military regimes are more so. The depth and durability of anti-regime protests themselves, however, clearly have a major effect on incumbents calculations about their ability to hold on to power. In some cases, mass mobilization appears to emerge almost spontaneously, the result of triggers and cascades that can erupt with surprising suddenness (Kuran 1989; Lohmann 1994). Yet we argue that the bargaining games between authoritarian incumbents and oppositions are rarely, if ever, the result of a single, one-shot uprising. Rather, distributive conflict transitions are most likely to occur where mobilization is backed by durable organizations capable of sustaining pressure on authoritarian incumbents. We begin this chapter with a more extended elaboration of the theoretical perspectives that guide the analysis. We then turn to some statistical models of the determinants of distributive and non-distributive conflict transitions. An important finding is that these two classes of cases are driven by different configurations of causal factors and that both structural and organizational measures of the propensity for collective action appear as significant determinants of distributive conflict cases. In the third section, we probe the distributive conflict transitions in more detail. We begin with the organizational resources in play, drawing a distinction between those transitions in which unions and other civil society groups are the dominant actors nearly two-thirds of the cases and those in which ethno-nationalist political forces play a dominant role. In both types, however, we see a similar sequence: mass mobilization depends on organizational resources, typically in civil society, which can be deployed to challenge state authority. We close by showing how political counter-elites and in some cases insiders--gain strength by drawing on the power of social organizations. We find that parties play a direct role in mass mobilization in only a relatively small number of cases. Nonetheless, alliances between political entrepreneurs, emergent and re-emergent parties and organized social forces provide the political and social foundation for the crafting of new political institutions and ultimately democratic governments. Theory: The Social Foundations and Political Consequences of Mass Mobilization Distributive conflict transitions are defined by the presence of mass mobilization. We pose two clusters of questions to these transitions. First, why would ordinary citizens risk repression to oppose authoritarian rule? How is such action coordinated? Who provides leadership? The second set of questions centers on the elite and counter-elite. How is the success of political entrepreneurs and new parties favoring democracy related to mass mobilization? 2

3 As noted above, distributive conflict models have paid surprisingly little attention to many of these crucial questions. In their abbreviated discussion of the issue, for example, Acemoglu and Robinson build on Mancur Olson to argue that anti-authoritarian movements solve collective action problems with selective incentives such as land for peasants in rebel territory or expectations of booty. But these models rely on the literature on organized insurgencies which rarely lead directly to democratic transitions as we will show and ignore other types of collective action in which these types of selective incentives seem much less plausible. If we look empirically at actual cases of distributive transitions, the popular mobilizations that have pushed authoritarian elites from power have unfolded in a variety of ways. In some cases, protests did appear to emerge almost spontaneously in uncoordinated prairie fires (Kuran 1989; Lohmann 1994). Contingent factors such as the death of a leader or a particularly egregious abuse of power the killing of a protestor or a corruption case-- spawn outrage and protest. Once triggered, information about citizen preferences cascades as those opposed to the regime realize that they are not alone. The Tunisian uprising of 2011 triggered by outrage over a confrontation between a low-level security official and a street vendor provides an out-of-sample example. By contrast, other examples of mass mobilization involve sustained and coordinated campaigns of opposition extending over a period of years, led by class-based or ethnonationalist organizations. The Solidarity movement in Poland or the sustained anti-military campaign of Argentine unions provide examples of the former; the mass uprisings in the Baltics in the last year of the Soviet Union and the emergence of indigenous social movements in Ecuador provide examples of the latter. We argue that the difference between these two apparently different routes may be less sharp than is thought. Even if protests are triggered by contingent factors and appear spontaneous, they are rarely sustained in the absence of civil society groups that can play a leadership and coordinating role: unions, NGOs, churches, ethno-nationalist organizations. Indeed, Acemoglu and Robinson themselves make an important theoretical argument about why this is likely to be the case. If mass mobilization is transient then leaderships are unlikely to yield to it; it makes more sense to coerce, make reversible concessions and wait out the storm. The mass protests that topple regimes, by contrast, typically involve an iterative process in which elites first attempt to contain the protest through combinations of repression and partial concessions. These strategic interactions can go on for months or even years. Incumbents are typically forced to yield when protest persists, backed by resources, organizations and networks of communication that can be activated by movement entrepreneurs and be sustained over time. It is these organizational resources that ultimately back the claims of political counter-elites. What gives rise to such organized pressures in the first place? Both the structural sociological literature on democratization and the resource/opportunity focus of the literature on contentious politics provide fruitful starting points for answering the question. Classic sociological approaches pay greater attention to the underlying social and economic structure. Barrington Moore (1966) emphasized the role of class structure in the democratization process but focused on the bourgeoisie; the class conflicts that gave rise to democracy in his chosen cases pitted emerging middle classes against the aristocracy; Ansell 3

4 and Samuels (2011, 2013) provide a more recent example of this approach. The basic line of reasoning was extended by Reuschemeyer, Stephens and Stephens (1992) and Przeworski (1985, 2008) to the role of working-class pressure in the democratic transitions in Europe in the late-19 th and early 20 th centuries. The core proposition was that industrialization was accompanied by the emergence of working classes that were large enough to have an interest in mass democracy and sufficiently concentrated in the workplace to overcome barriers to collective action. Geographic or workplace concentration made it easier to form organizations, to communicate and coordinate, to socialize people into collective norms, and to monitor individual behavior. We take structural approaches as a starting point, and test for their effect statistically. We also show both statistically and in our qualitative analysis that unions matter, playing a role in a substantial number of the distributive conflict cases. But in developing countries during the third wave, the organizational space is by no means bounded by emergent working class movements. First, we document how unions typically acted in coalition with a plethora of other social forces that did not play the same mobilizational role in earlier transitions: professional associations, churches and particularly NGOs. Second, we find that a number of transitions in the sample are driven not by class grievances, but by the politicization of ethno-nationalist cleavages. We categorize these as distributive conflict transitions because they involve material conflict between contending ethnic groups but also because they similarly rest on organized opposition to incumbent authoritarian regimes. It is important to underscore that these organizations do not typically arise de novo; history matters. In some cases, unions, NGOs and whole social movements survive the constraints imposed by them by authoritarian or competitive authoritarian regimes; this latent organization has been identified in the sociological literature as an abeyance structure (Taylor 1989; Taylor and Crossley 2013). Authorities may seek to jail, exile, or even murder the leaders of opposition groups and disrupt the organizations they led, for example by seizing bank accounts, shutting down headquarters and offices, and disrupting cells. During the 1970s, all of these tactics were deployed by military dictatorships in Argentina, Uruguay, and Chile against industrial unions, NGOs and small revolutionary organizations. But while the activity of such groups might be repressed and driven underground, a social movement in abeyance can provide the foundation for new rounds of mobilization through the persistence of activist networks, repertoires of goals and tactics, and the maintenance of collective identities that constitute a symbolic resource for subsequent mobilization (Taylor and Crossley 2013). What is it that these organizations do? First and most obviously they can turn out their members. To borrow a felicitous phrase from Acemoglu and Robinson (2009), the political challenge to authoritarian regimes ultimately comes from the exercise of de facto power on the part of publics that are very much larger in number than the elites they seek to displace. Organizations can turn out members and direct them to actions that impose costs on the regime. These may range from the proverbial storming of the palace to general strikes--paralyzing mass actions that effect economic activity--to more limited strikes and blockades of select industries, the capital city or key roads. Organizations play the role of vertically coordinating collective action by their members. 4

5 Second, these organizations coordinate horizontally. In the more dense organizational space that characterizes the late-20 th century developing world, civil societies are characterized by a plethora of organized social actors and non-governmental organizations raising a multiplicity of demands and deploying a variety of communication strategies to reach their members. But mass action is also coordinated across groups, and organizations and their leaders play a central role in this process. Coalitions of organized civil society actors can serve as force multipliers that increase the de facto power brought to bear on authoritarian incumbents. Yet there is a third role that organizations play and it can best be captured by drawing on the distinction made by Linz and Stepan (1996) between political and civil society. The social organizations that play a role in mass mobilization are of necessity focused initially on the interests of their members, even as their demands broaden. Yet these organizations also provide resources to political counter-elites and ultimately parties that aggregate grievances into systemic political demands on the incumbent authoritarian leadership. We know that these individuals are significant because they are typically the first targets of crackdowns. Their hand depends critically on their ability to both articulate demands and draw on the de facto political power provided by organizations and mass mobilization. In sum, mobilization is not simply the mobilization of people into the streets, but the mobilization of grievance backed by collective action. The second piece of the analytic puzzle is how mass mobilizations contribute to democratic transitions. It is important to note that mass mobilization can impel authoritarian elites to relinquish power in a several ways. In some cases, the magnitude of mass mobilization relative to existing state capacity makes a coercive response physically or politically impossible, and incumbent rulers are forced to resign and even flee--in favor of interim governments, national constitutional conferences, or oppositions. The ancien regime is overwhelmed. In other cases, incumbents face pressures from below but retain adequate capabilities or support to negotiate their exit, and even to impose conditions on new democratic governments. Distinctions between different modes of exit have been extensively analyzed in the literature on transitions (Rustow 1970; Linz 1978; O Donnell and Schmitter 1986; Share 1987; Karl 1990, Huntington 1991, Collier 1999) and have been traced not only on the strength of mass mobilization, but to a variety of other domestic and international factors that affect the calculus of incumbent elites and their political opponents. In keeping with our broad distinction between distributive and non-distributive transitions, however, our emphasis is on the fact that in all cases of the former, the exercise of de facto power is crucial in challenging the capacity to govern and repress, thus contributing to political liberalization and democratic rule. We thus focus in the last part of the chapter on the relationship between mass mobilization and the political entrepreneurs, coalitions, and parties that ultimately take power. Mobilization is not simply about the exercise of de facto power; it is the exercise of de facto power around some articulated purpose and aim. Political entrepreneurs frame not only social but political demands: liberalization and the relaxation of controls, political and electoral reform, and ultimately the withdrawal of authoritarian elites and the negotiation of democratic institutions that will replace them. 5

6 An important question is whether parties themselves should be conceived of as social organizations akin to unions and NGOs that play a mobilizational role. We treat this ultimately as an empirical question, but in general we expect that parties are likely to be consequential fairly late in the sequence of moves between incumbents and oppositions. First, in many single-party and military authoritarian regimes parties are banned outright, and thus appear on the political scene only relatively late in the protest cycle. In competitive authoritarian regimes, party elites can play a role in mobilizing anti-regime votes or protest against electoral fraud; Lindberg (2009) notes that elections in competitive authoritarian regimes may constitute a moment of vulnerability. We find, however, that even where parties have some scope for political maneuver under authoritarian rule, their power typically lies in the resources of allied unions, civil society organizations, or religious organizations that can sustain mobilization beyond the election period itself. Mass mobilization can depose governments, but how does it form new ones? The answers can be found by looking more closely at how mass mobilization affects the bargaining power of emergent political entrepreneurs and political parties. In some cases, insiders are able to capture mass mobilization to head parties that grow out of the existing elite; this happened in seven of the 52 transitions identified in the CGV and Polity datasets. More typically, however, it is mass mobilization that strengthens the hand of the parties that had been tolerated in semi-competitive regimes, established parties that had been driven underground, and the new parties that emerge in the process of constitutional negotiations and the transition itself. We conclude by noting, however, that the interests of political entrepreneurs, parties and the social organizations that support mass mobilization tend to diverge over time as new governments face the challenge of responding to pent-up social demands. We start by exploring these ideas with some simple statistical analysis of the factors that might be associated with different transition paths: those that involve mass mobilization and those that do not. We then turn to a deeper exploration of the cases, following a broadly temporal sequence: the origins and nature of the union, civil society and ethno-nationalist organizations that are active in distributive conflict transitions; how, precisely, they generate regime change; and the role of mass mobilization in the emergence of the counter-elites and parties that ultimately come to power following the transition. Statistical Evidence The regressions that frame our subsequent discussion are based on a panel of data for between 90 and 95 developing countries between 1980 and 2008 with roughly 2000 country-year observations depending on the model. Since we are interested in the propensity to transition through different routes distributive and non-distributive we run separate regressions with each type of transition as the dependent variable. Tables show the results, first for CGV distributive and non-distributive transitions, then for those coded by Polity. Our main variables of interest are those designed to capture the propensity for collective action, drawing from the sociological and contentious politics literature outlined above. We consider two structural measures that proxy for the size of the working class: value added in manufacturing as a percentage of GDP and industry as a share of GDP 6

7 (World Development Indicators 2011). We assume that a relatively large manufacturing or industrial sector will imply a larger and more concentrated workforce, with attendant opportunities for communication, coordination, and organization. As we demonstrate in more detail below, countries that rank high on the manufacturing and industry variables include ones with a history of labor militancy such as Argentina, Brazil, Thailand, and South Korea. 1 To get a more direct measure of organizational capacity, we use the per capita membership in unions affiliated with the International Trade Union Confederation (ITUC), a merger of International Confederation of Free Trade Unions (ICFTU) and the World Confederation of Labor (WCL). Not all unions have established this international tie, so many countries are given scores of zero; moreover, scores for other countries may also underrepresent the extent of unionization. Nevertheless, the history of the ITUC as the successor to the ICFTU and WCL has certain advantages. The WCL was a Christian Democratic labor confederation formed in 1920; the ICFTU broke off from the World Federation of Trade Unions in 1949 to establish itself as an explicitly anti-communist organization. As a result, with just a few exceptions 2 (ones that weight findings against the hypothesis), almost all of the countries with scores of zero are ones in which unions are in effect appendages of the state and/or where unionization rates are in fact low. These include, for example, China, Vietnam, as well as virtually all of the countries of the Middle East and North Africa. The lowest unionization scores, predictably, are found in poorer countries of sub-saharan Africa or Asia, whereas the highest are in relatively industrialized Latin American and former Communist countries. We were also interested in finding a proxy for historical repertoires of collective action; traditions of collective action that might influence the current likelihood of mass mobilization. We considered three separate measures: anti-government demonstrations, riots and strikes (Banks 2011). In the year prior to the transition, each of these variables proved highly significant in models of distributive conflict transitions against a battery of controls, but had no effect in non-distributive conflict models. We interpret this as strong validation of our distributive conflict transition construct since it indicates that such transitions are in fact correlated with mass mobilization. But while these variables were significant in the short-run they showed no effect when averaged over longer time periods and we omit them from the reported models. The models include a measure of inequality (the share of capital in manufacturing income) the driving causal mechanism of distributive conflict theories as well as ethnolinguistic fragmentation. We presume that the latter might capture the presence of other social cleavages that did not break along class lines; and although this variable did not prove 1. We also considered the effects of sheer population density: the number of people per square kilometer of land area (World Development Indicators 2011). This variable has been used in the civil war literature as a proxy for the ease of mass mobilization and collective action (Hauge and Ellinsen 1998, De Soysa 2002, Urdal 2005, Raleigh and Urdal 2007, Theisen 2008, Raleigh and Hegre 2009; see Dixon 2009 for a review), although there is ongoing controversy over whether it is a valid proxy. Nonetheless, the natural log of this variable was consistently significant in a number of our models. 2 Notably, Uruguay and Bolivia 7

8 significant in our regressions, we will see below that ethno-nationalist cleavages did in fact prove an alternative source of mass mobilization in a number of cases. Although our primary interest is in distributive conflict transitions, finally, it is important to control for several factors that are likely to affect transitions of both sorts. First among these is economic performance, measured as GDP growth in the year prior to the transition. This measure as noted above can be serve as an indicator of both grievance and of the capacity of the government to respond to challenges. Deteriorating economic conditions may generate disaffection and spur mass protest. But it also affects incumbents access to the resources needed to maintain the loyalty of elites particularly the military and police--and wider bases of social support within parties, the state (for example the civil service) or controlled social groups, including labor. We are also interested in controlling for the role that international factors might play in democratic transitions. As an indicator of vulnerability to pressure from external donors, we use the ratio of official aid to GDP, measured in constant U.S. dollars (Dreher 2006). To proxy for neighborhood effects and cross border influences, we include the percentage of neighboring countries that are democratic (from Svolik 2008). This variable is widely assumed to reflect an array of international linkages from diffusion effects, economic and diplomatic incentives, and the political engagement of regional actors. We expect high vulnerability to external leverage measured by aid and neighborhood effects to have more consistent effects in non-distributive transitions where, by definition, domestic pressures from below are weaker. A third major cluster of controls is institutional. There is significant empirical work on how the institutions of authoritarian rule may influence the propensity for collective action. Regimes that rule primarily through the military have been theorized to be more reliant on repression and thus more brittle; by contrast, those that can coopt potential challengers through limited legislative and electoral channels may be able to avoid mass mobilization. To capture these differences, we use two dummies, both from Svolik (2012): the first is for military governments that rely primarily on the military hierarchy for political and social control; the second is a dummy for regimes that permit multiparty legislatures. We expect exclusive military rule to increase the vulnerability of elites to distributive conflict transitions and semi-competitive legislatures to reduce it. Neither factor should be relevant in non-distributive transitions, since elites by definition do not face challenges from below. We included two final controls that we expected to have a positive effect on the likelihood of a transition of both sorts. GDP/per capita controls for the long-standing expectation that democratization is a function of the level of development. We also included a dummy for whether the country had undergone a democratic transition in the past, on the expectation that such a history might increase the likelihood of transitioning again. The results on GDP per capita were mixed, but where significant were negatively signed, reflecting a distinctive feature of the Third Wave: the spread of democratization to lower income countries. Prior transitions were both statistically and substantively significant, suggesting that the transition process frequently involves multiple tries. The specification is a time-series cross-sectional (TSCS) design with country-year as the unit of analysis. We lag all explanatory and control variables by one year and include a 8

9 cubic polynomial to model the temporal dependence of our binary outcome (Carter and Signorino 2010). Democratic transitions occur only rarely; the number of nonevents is far greater than the number of events. In our sample of 95 developing (and non-democratic) countries from 1980 to 2008, there are 42 distributive conflict transitions and 37 nondistributive conflict transitions out of 2337 country years and thus possible transitions; distributive conflict transitions occur in just over one percent of all country-years and nondistributive conflict transitions, in just under one percent a rare event indeed. We therefore use King and Zeng s (2001) rare events logit (ReLogit in R) procedure with countryclustered robust standard errors, a procedure that minimizes loss of efficiency and biased estimates but risks possible underestimation of the event of interest. In sum, we estimate the following model: Pr(transition i,t = 1) = α + βx i,t + ε i,t where X is a vector of explanatory and control variables for country i at time t. We organize the presentation on the basis of the dataset used. Tables 2.1 and 2.2 report the results for CGV distributive and non-distributive conflict transitions respectively. Tables 2.3 and 2.4 repeat the exercise for the Polity transitions. In each table, models 1 through 6 show results for each of our proxies for the propensity for collective action; all other independent variables are shown in each of the models. In general, notwithstanding some differences noted below, results for CGV and Polity are similar. 9

10 Table 2.1. CGV Distributive Conflict Transitions, (Rare Event Logit Likelihood Estimates) Manufacturing (% of GDP) 0.10*** 0.04 Industry (% of GDP) 0.04* 0.02 Union Membership 6.05** 3.05 Population Density 0.00*** 0.00 Aid (% of GDP) Multiparty Legislature -1.72*** -1.39** -1.25** -1.42*** Capital Share Military Dictatorship Dummy 0.89* ** * 0.51 GDP per capita -0.55* Growth -0.10** -0.09*** -0.06** -0.07*** Prior Transition Dummy 2.52*** 2.32*** 2.06** 2.07** Proportion of Democracies Ethnolinguistic Fractionalization N Groups

11 Table 2.2. CGV Non-Distributive Conflict Transitions, (Rare Event Logit Likelihood Estimates) Manufacturing (% of GDP) Industry (% of GDP) Union Membership Population Density Aid (% of GDP) 5.55** 4.53** 5.96* 6.16*** Multiparty Legislature * Capital Share Military Dictatorship Dummy GDP per capita Growth Prior Transition Dummy 1.82** 1.92** 2.63** 2.15** Proportion of Democracies 1.46*** 1.27** 1.63** Ethnolinguistic Fractionalization N Groups

12 Table 2.3 Polity Distributive Conflict Transitions, (Rare Event Logit Likelihood Estimates) Manufacturing (% of GDP) 0.15*** 0.03 Industry (% of GDP) 0.06*** 0.02 Union Membership 5.37** 2.52 Population Density 0.00*** 0.00 Aid (% of GDP) Multiparty Legislature -1.33** -1.15* * Capital Share Military Dictatorship Dummy 1.29** 1.47** 1.11* 1.09* GDP per capita -1.02*** -0.78*** Growth Prior Transition Dummy 2.68*** 2.76*** 2.52*** 2.53*** Proportion of Democracies Ethnolinguistic Fractionalization -1.05* ** N Groups

13 Table 2.4. Polity Non-Distributive Conflict Transitions, (Rare Event Logit Likelihood Estimates) Manufacturing (% of GDP) Industry (% of GDP) -0.06** 0.03 Union Membership Population Density Aid (% of GDP) Multiparty Legislature Capital Share -3.14* * -2.97** Military Dictatorship Dummy ** GDP per capita * -0.46* Growth Prior Transition Dummy 1.37** 1.17* 1.44** 1.52** Proportion of Democracies 1.61** 1.36* 1.53* 1.36* Ethnolinguistic Fractionalization N Groups First, the models provide consistent support for the importance of our proxies for collective action capacity in determining the likelihood that a distributive conflict transition. Controlling for other possible determinants, the share of manufacturing and industry in GDP, population density and union density all have significant effects in both the CGV and Polity samples. Capacities for collective action whether measured structurally or proxied by union density are significant correlates of distributive conflict transitions. By contrast, none of the collective action factors have significant effects in explaining non-distributive conflict transitions in the CGV or Polity samples. 3 3 Although not shown, riots, strike events, and anti-government demonstrations have identical effects: they are significant in distributive transitions, but not in non-distributive ones. 13

14 The regressions also confirm what we found in the last chapter; that inequality is neither the source of grievances nor of the class dynamics postulated by the distributive conflict models. We did expect economic circumstances to operate as a source of grievances in the distributive conflict cases and as a constraint on the capacity of the state to respond in both distributive and non-distributive conflict cases, and the effect of this variable was negative and significant in all of the CGV distributive conflict models. The worse the economy the greater the likelihood of a CGV distributive conflict transition. Table 2.5 GDP Growth Rates in the Year Prior to POLITY Distributive Transition Negative or higher Argentina 1983 Bolivia 1982 Philippines 1987 Bangladesh 1991 Bulgaria 1990 Nepal 2006 Brazil 1985 Estonia 1991 Burundi 2005 Indonesia 1999 Dom Rep 1996 Latvia 1991 Ecuador 2002 Madagascar 1992 Georgia 2004 Mongolia 1992 Guatemala 1996 South Africa 1992 Lesotho 1993 Sudan 1986 Malawi 1994 Ukraine 1991 Peru 1980 Ukraine 1994 S. Korea 1988 Uruguay 1985 Thailand 1992 Zambia Missing: Armenia 1991, Haiti 1990, Lithuania 1991, Poland 1991 [4] Source: World Development Indicators (2011) These findings did not extend to the Polity distributive conflict transitions, but we should approach this null finding with some caution. The debt crisis of the 1980s and the deep transitional recessions that followed the decline and breakup of the Soviet Union clearly had consequences for incumbent authoritarian regimes that persisted well after the worst shocks that passed. However, these effects may not have the same lag structure across countries for a variety of idiosyncratic political reasons, making them difficult to capture in a standard panel design even with the introduction of multiple lags. Table 2.5 shows that half of the distributive Polity transitions (14 of 28) did experience negative or flat growth rates in the year preceding the regime change. Transitions in Argentina, Uruguay and Bolivia, for example, occurred at or near the nadir of the debt crisis in the early 1980s and economic grievances were indeed an important component of the mass mobilizations in all three countries. Indonesia provides another example from the 14

15 second generation of financial crises of the 1990s. In almost all of the post-soviet cases, 4 economic decline during the 1980s (and outright crises in Poland and Romania) had jeopardized the capacity of the ruling party to maintain the loyalty of the underlying population, and dual transitions to democracy and the market began in the midst of deep transitional recessions. The table also shows that 12 countries experienced growth in excess of four percent in t-1, more than enough to assure increases in per capita income. But the cases of Peru and the Philippines nominally high-growth countries in the year prior to their transitions show clearly how a focus in the year prior to the transition can be misleading as the political consequences of prior shocks remained highly consequential. Peru s first economic crisis came in the mid-1970s as a consequence of populist policies pursued under the military government. When the government undertook a delayed stabilization, it faced fierce opposition from corporatist labor unions that it had counted on to support the regime. After a massive general strike in 1977 protesting austerity policies, the military rulers yielded to demands for a Constituent Assembly, leading to the Polity transition coded in The Philippines case is quite similar. The country transitioned to democratic rule early in 1987 as the economy was turning up. But the mass mobilization of the people power movement built on the back of a sustained critique of the economic failure of the Marcos regime that was articulated by the opposition from 1985 onward and that generated protests in the years prior to the transition. If anything, the African countries that transited in the early 1990s had suffered even more severe shocks. Along with the low-growth cases (Madagascar, Zambia, Sudan and South Africa), those experiencing rebounds in the year prior to transition (Burundi, Lesotho, Malawi) had also gone through sustained periods under authoritarian rule of economic crisis, prolonged and ultimately failed structural adjustment efforts and mass protest and political destabilization. The third wave transitions occurred in an international context that was favorable to democratic change. The late- and post-cold War period witnessed a sharp increase in the involvement of international and regional institutions in brokering post-conflict settlements and encouraging both through carrots and sticks political liberalization and democratization. Neighborhood effects and to a lesser extent aid dependence did have a significant strong impact on non-distributive transitions, and we explore these relationships in more detail in Chapter Three. Neither proxy was statistically significant in the distributive transitions although as we will suggest below, external actors did play an important role in a substantial number of these cases. Finally, we considered whether the type of authoritarian regime had any effect on the propensity to transit through either or both of the two stipulated routes. As hypothesized, regimes with multiparty legislatures were consistently less likely to experience distributive conflict. In all of the CGV distributive conflict models and in all but one of the Polity distributive conflict models, legislatures appeared to constitute an effective mechanism of cooptation that deterred mass mobilization. Exclusionary rule by military dictatorships, conversely, were more vulnerable to mass protest. The regressions, moreover, show the 4 With the exception of Georgia which became democratic much later in

16 differential effect of authoritarian institutions on the two types of transition. Neither multiparty legislatures nor military dictatorship had statistically significant effects on elite or externally-driven non-distributive transitions. To summarize, three findings from the statistical exercise are of central theoretical interest. First, different causal dynamics do appear to be at work in distributive conflict and non-distributive conflict transitions. Collective action capacity is associated with a greater likelihood of distributive conflict transitions, but does not appear to affect the propensity to transition via the non-distributive conflict route. Distributive conflict transitions arise not because of objective inequalities, but from the capacity of political leaders and social actors to effectively mobilize collective action and articulate distributive grievances into the political arena. A second finding is that institutions have some effect on distributive conflict transitions but not on their non-distributive conflict counterparts: competitive authoritarian regimes appear less vulnerable to distributive conflict transitions and military regimes more so. Finally, the effects of economic factors are mixed but appear to matter more in the distributive conflict cases. This finding would comport with a model in which mass protest is motivated at least in part by economic grievances. Section IV: Qualitative Evidence and Causal Mechanisms In this section, we go beyond the statistical association between capacity for collective action and distributive conflict transitions considering how anti-regime protests were organized, and the type of support they provided to democratic oppositions. As we discussed in Chapter One, the appropriate empirical strategy for addressing this issue is to select on the dependent variable in order to examine the causal mechanisms and dynamics of these transitions in greater detail. We start by identifying two types of organizations through which collective action manifested itself: those in which unions played a central role, even if in conjunction with other civil society groups; and those in which ethno-nationalist political organizations dominated. In both cases, however, we show that mass mobilization was not simply spontaneous but drew on organizational resources, even if initially in abeyance. We show how these mass mobilizations arose and confronted incumbent authoritarian regimes in the period prior to the transitions. We conclude by considering how these social organizations provide resources to the political entrepreneurs and parties that emerge in the late authoritarian period and thus increase their bargaining power vis-à-vis incumbents. Organizational Foundations of Mass Mobilization Although inequality per se was not found to be a significant source of regime change, distributive conflict transitions do have roots in material grievances between elites and excluded groups. How and by whom are these grievances mobilized? We do find instances of prairie fire protest set off by specific triggers that spreads rapidly; and in a number of cases, authoritarian elites were shaken by uncontrolled, decentralized, and often very violent expressions of popular discontent. But a closer examination of cases reveals that organized social forces typically play a more central role than this spontaneous model of protest might lead us to believe. 16

17 What sorts of organization were important? As our regressions indicate, unions (typically allied with civil society groups) were often significant actors in anti-government protests, strikes and other forms of collective action. In a second class of cases, transitions followed in the wake of mass mobilization around nationalist and ethnic cleavages, typically with strong spatial or regional dimensions, but in these cases as well, established organizations played a crucial role. In Table 2.6, we code these two sources of mass mobilization as mutually exclusive, focusing on what appear to be the most significant sources of mass mobilization in the case. Class and ethnic stratification, to be sure, overlap in complex ways, and we found some instances in which differences both between and within ethnic groups were mobilized. Nonetheless, the coding of the most significant social sources of mass mobilization was generally straightforward. Distributive conflict transitions were coded as union or civil society cases in one of two circumstances: when unions and/or civil society groups were leaders or core members of anti-regime coalitions; or when strikes or walkouts by these organizations added significantly to the overall pressure on the regime even if not closely coordinated with other anti-regime forces. Similarly, distributive conflict transitions were coded as ethno-nationalist when such groups were leaders or core members of anti-regime coalitions; or when protests by these organizations added significantly to the overall pressure on the regime. The largest share of distributive conflict transitions about two-thirds fell into the union-civil society category. Democratization occurred in the context of ethno-nationalist challenges in the remainder. These included five cases in which the incumbent authoritarian regime was the Soviet Union and the demands for regime change were in effect demands for independence: Armenia, the Ukraine, and the three Baltic countries, Estonia, Latvia and Lithuania. But the ethno-nationalist cases also included the transition from apartheid in South Africa, civil conflicts in Latin America in which indigenous peoples played a crucial role (Ecuador, Guatemala [1986 and 1996]), and transitions in societies with polarized ethnic-cum-regional cleavages (Burundi, Fiji, Kyrgyzstan, Sri Lanka and the Sudan). 17

18 Country Table 2.6 The Social Foundations of Distributive Conflict Transitions Democracy Coding Union/civil society transitions Ethnonationalist transitions Revolutionary armies and militias Riots (distinct from protests) Albania 1991 CGV only X X Argentina 1983 CGV/Polity X Armenia 1991 CGV/Polity X X X Bangladesh 1991 Polity only X Benin 1991 CGV/Polity X X Bolivia 1982 CGV/Polity X Brazil 1985 CGV/Polity X Bulgaria 1990 CGV/Polity X Burundi 1993 CGV Only X X X Burundi 2005 Polity Only X X X Republic of the Congo 1992 CGV Only X Dominican Republic 1996 Polity Only X Ecuador 2002 CGV Only X El Salvador 1984 CGV/Polity X X El Salvador 1990 CGV/Polity X X Estonia 1991 CGV Only X Fiji 1992 CGV/Polity X Georgia 2004 CGV Only X X Guatemala 1986 Polity Only X X Guatemala 1996 Polity Only X X Haiti 1990 Polity Only X X Indonesia 1999 CGV/Polity X Kenya 1998 CGV Only X Kyrgyzstan 2005 CGV Only X Latvia 1991 CGV/Polity X Lesotho 1993 Polity Only X Lithuania 1991 CGV/Polity X X X Madagascar CGV/Polity X Malawi 1994 CGV/Polity X X Maldives 2008 CGV Only X Mali 1992 CGV/Polity X X Mongolia 1992 CGV/Polity X Nepal 1990 CGV Only X Nepal 2006/08 CGV/Polity X Niger CGV/Polity X Niger 2000 CGV Only X X Nigeria 1999 CGV Only X X Pakistan 2008 CGV Only X Peru 1980 CGV/Polity X Philippines 1986 CGV/Polity X Poland CGV/Polity X X Romania 1990 CGV Only X X South Africa 1992 Polity Only X X South Korea 1988 CGV/Polity X X 18

19 Sri Lanka 1989 CGV Only X X Sudan 1986 CGV/Polity X X Suriname 1988 CGV Only X Thailand 1992 CGV/Polity X Ukraine 1991 CGV Only X Ukraine 1994 Polity Only X Uruguay 1985 CGV Only X X Zambia 1991 Polity Only X Total cases (52) CGV cases (42) Polity cases (34) Given that mass mobilization constitutes what Acemoglu and Robinson call de facto power, it is useful to consider the role that spontaneous or organized violence plays in these distributive conflict transitions. The table codes for the presence of riots spontaneous violence that went beyond mass protest and the presence of armed insurgent groups and militias. How decisive was such violence? Riots were indeed significant in our models of distributive conflict transitions. Perhaps the most clear-cut illustration is in Haiti , where widespread rioting helped to drive Colonel Prosper Avril from office (at the urging of the United States government), and then helped to derail a brief military uprising aimed at preventing the popularly-elected Jean-Claude Aristide from taking office. In approximately 10 other cases, rioting played at least some role. Yet the Haitian case is the only one in which rioting in support of an opposition alternative (Aristide) really provide decisive in the transition. Authoritarian elites in other settings were sometimes shaken by uncontrolled, decentralized, and often very violent expressions of popular discontent. But rioting was typically accompanied by much more durable and centralized forms of class or ethnic organization that ultimately convinced elites that increased repression was not a sustainable strategy for staying in power. In 17 of the 52 distributive conflict cases, incumbent authoritarian regimes faced armed insurgencies, with half of those ethno-nationalist in nature. El Salvador and Guatemala constitute classic civil war cases; in the other cases, the scope of insurgent violence ranged from major ethnic conflicts as in Sri Lanka to the operation of terrorist networks, as in Pakistan, to more episodic and scattered insurgent efforts. Yet in none of these cases did democratization occur as a result of the outright victory of these forces over the government. As with riots, organized violence at best operated in conjunction with enduring forms of political organization, typically in the cities, that were capable of mobilizing citizens into the streets and thus providing leverage for emergent political entrepreneurs. The Union-Civil Society Cases How widespread and significant was the union role in the so-called union-civil society cases? In several Asian cases, where unions were historically weak, they played a distinctly secondary role to NGOs and other civil society groups. In the Philippines, for 19