Subversive Effects of Elections in Competitive Authoritarian Regimes: Repression and social spending 1

Transcription

1 The XXII World Congress of Political Science (IPSA Annual Conference) July 8-12, 2012 Madrid Conference paper Subversive Effects of Elections in Competitive Authoritarian Regimes: Repression and social spending 1 Margarita Zavadskaya PhD Researcher European University Institute (EUI), Florence, Italy European University at Saint Petersburg, Russia Margarita.Zavadskaya@EUI.EU mzavadskaya@eu.spb.ru 1 DRAFT: Please do not cite or circulate without author s permission

2 Introduction One of the widely shared features of modern autocracies is the presence of democratically designed institutions. Elections, referendums, legislatures and parties are the essential institutions by default. In classical political theory these institutions are seen as a key element of democracy and are designed in order to channel political participation and to allow for the articulation of public demands (Dahl 1971). However, modern non-democratic incumbents have managed to transform and adapt these institutions according to their interests in political survival. Parliaments help sustain autocracy by providing an arena of negotiation and bargaining for those who may credibly threaten the current regime (Lust 2009; Gandhi and Przeworski 2007; Wright and Escribà-Folch 2010; Bueno de Mesquita et al. 2003). Political parties in authoritarian regimes facilitate cadre rotation and cooptation as well as the redistribution of spoils, rents and other benefits (Greene 2009; Magaloni 2006). Elections and referendums, in turn, facilitate mobilization of citizens and control of loyalty among the coalition members. Despite the overwhelming control maintained over representative institutions in nondemocratic political regimes, elections and referendums seem to have gotten out of hand. The wave of color revolutions (Serbia 2000, Ukraine 2004, Georgia 2003, Kyrgyzstan 2005) and the referendum failures (Chile 1988, Zimbabwe 2000) are striking instances of the unanticipated anti-hegemonic consequences of authoritarian institutions. Andreas Schedler raised the question of the conditions under which authoritarian elections fulfill a stabilizing role, and those in which they act as subversive forces (Schedler 2006)? I reframe the question in narrower terms: when do wise and finely tuned pre-electoral strategies help incumbents sustain their power even given unfavorable conditions? And vice versa: why even given satisfactory structural socioeconomic environment incumbents do make fatal mistakes? It is evident that even extremely skillful incumbents are doomed to failure due to the adverse external conditions. It is also expected that even clumsy or incompetent incumbents win reelections in thriving economies. Yet, given the same structural socio-economic conditions, some authoritarian incumbents use electoral manipulations aptly and consequently win whereas others, for some reason, lose. Obviously it is not feasible to respond to this challenging question in this paper in an exhaustive manner. That is why I will focus on theoretical aspects, conceptualization of the dependent variable, and preliminary analysis where I try to establish which variables are of more relevance for further research. In order to test the correlates of unintended consequences I distinguish between structural long-term factors and more contingent actor-based strategies. I must mention that I do not want to refute the existing explanations; rather I am more interested in how elections operate as an arena of potential political risks for authoritarian elites and can trigger uncontrolled processes. The most obvious instances of undesired and subversive consequences are the electoral failure and the occurrence of post-electoral protests. By controlling for socio-economic conditions I test whether the redistributive and repressive policies are associated with unanticipated consequences for the incumbent. 2

3 I do not treat elections as causal variables per se, but rather as arenas of political contestation or focal points in Thomas Schelling s terms (Schelling 1960; Brownlee 2007; Ostrom 1998) 2. Thus, elections can be viewed through the lens of eventful political science or exceptional events which are distinct from politics as usual (Bunce and Wolchik 2011:26-27). Put otherwise, elections are not good or bad as such; rather they activate accumulated disaffection and get out the vote of those who have been silent for a long time (Lohmann 1994, p.44-45). Finally, I must specify the limits of my analysis. First of all, I focus exclusively on two types of actors in the electoral arena: the voters and the incumbent. By doing this I reduce the model to two actors which is definitely a simplification of the reality. Secondly, I analyze only the incumbent s strategies in the electoral arena without accounting for the counteractions from the opposition and voters. Although elections are an interactive game and actors adjust their strategies according to the moves of their opponents or allies. In the first part of the paper I discuss the phenomenon of competitive or electoral authoritarianism as modernized incarnation of the old good dictatorships and the role of elections under non-democratic conditions. The second part is devoted to the conceptualization of the dependent variable which is unintended or unanticipated consequences for the incumbent. Then I proceed with the preliminary empirical test of my argument where I strive to find the correlates of electoral failure. The latter is the ultimate form of unanticipated outcome since there is hardly an incumbent who would have convened elections simply to lose them. My intuition is that the authoritarian electoral the authoritarian electoral success shaped not only by not only structural socio-economic conditions but also by more contingent moves that could tip the balance in favor of the incumbent or vice versa. 2 Some scholars label elections as an opening up of political opportunity structure (Gel'man 2012)(see Gel man 2012 on the interpretation of parliamentary and presidential elections in Russia and the protest wave in the aftermath). For instance, according to Sidney Tarrow the opening up of POS occurs as a result some exogenous event (Tarrow 1989). However, elections cannot be treated as completely exogenous to the political regime. It seems to me that it is rather an endogenous potential threat. 3

4 1. Elections in Authoritarian Regimes One of the main debates unfolds around the question how to orient in the grey zone of the political spectrum which has considerably broadened as a result of the third wave of democratization and inhabited by the so-called hybrid or illiberal democracies (Carothers 2002; Diamond 2002; Zakaria 2007), electoral or competitive authoritarianism (Schedler 2006; Levitsky and Way 2002). In this type of regimes democratic institutions exist but actual practices are far from democratic. Adam Przeworski (Przeworski 2000) emphasizes the importance of the democratic institutions in the stabilization of the political order, put otherwise, elections have become normatively significant element of democracy and not merely a technical procedure of elite selection (Schumpeter 1994). Obviously, the ascension of election into the pedestal of democratization has entailed the harsh critique which treat elections as a necessary but not sufficient condition and by attributing curative properties to elections scholars may be trapped into electoral fallacy (Karl and Schmitter 1991). This debate has been followed by the revision of the established political regimes typologies (Chehabi and Linz 1998; Linz 1985; Dahl 1971), particularly that of nondemocratic spectrum (Brooker 2000; Ezrow and Frantz 2011). Classical scheme contestation-participation has been disaggregated into different, more detailed accounts. Barbara Geddes (Geddes 1999) uses the nature of rulers as a departure point for her distinction between personalist, military, one-party, and hybrid regimes. Juan Linz highlights the crucial role of the domestic economy and the way a ruler uses public and private resources (Chehabi and Linz 1998). The analysis of the political transitions in Africa has reinvigorated Weberian concept of neopatrimonialism (Bratton and Van de Walle 1994). Given that the role of elections varies across different authoritarian regimes, I would not exaggerate the role of elections in the regime breakdown or survival. Moreover, the variety of political institutions under autocratic rule is considerably richer when compared with democracies; every regime suggests its own combination of political recruitment mechanisms. That is why there is a need to homogenize the countries where elections are meaningful even if fraudulent, non-free and unfair. In order to select only relevant political regimes I rely on the concepts of electoral and competitive authoritarian regimes developed by Andreas Schedler, Lucan Way and Steven Levitsky (Levitsky and Way 2010; Schedler 2006), which include both the electoral dimension and political competition aspect. Electoral authoritarian regimes are those regimes holding elections regardless of their quality and the level of competition. Philipp Roessler and Marc Howard, in turn, suggest a typology that allows for even more precise distinction between closed, hegemonic, and competitive authoritarian regimes (Howard and Roessler 2006, p ). I focus only on competitive authoritarian regimes where electoral victory by the opposition is at least technically possible, i.e. opposition parties are formally allowed and legitimate, and there is more than one candidate on the ballot (Hyde and Marinov 2009). Competitive authoritarianism thus qualifies as such a regime, where formal democratic institutions are widely viewed as the principal means of obtaining and exercising political authority. Incumbents violate those rules so often and to such an extent, however, that the regime fails to meet conventional minimum standards for democracy (Levitsky and Way 2002). There is a whole strand of literature scrutinizing the interactions between elections and political regime change. While some scholars explain why a number of regimes resort to electoral means of maintaining power (Geddes 1999; Gandhi and Lust-Okar 2009; Escriba Folch 2003), others employ elections as an explanandum in the analysis. Some 4

5 scholars connect elections and their outcomes with further democratization (Lindberg 2009; Bunce and Wolchik 2010): the longer electoral experience is, the more likely the peaceful democratic transition. Empirical evidence from African countries and the study of electoral revolutions in the Eastern Europe and post-soviet space have provided some grounds for the claims of the existence of democratization by elections or peculiar electoral path to democracy (Lindberg 2009; Bunce and Wolchik 2009). On the other hand, study of the Middle East and East Asia countries elections do not undermine the regime at all; quite the opposite, they do uphold it by reproducing patronage or clientelistic practices, vote buying and/or controlling (Hermet, Rose, and Rouquié 1978; Kitschelt and Wilkinson 2007; Lust 2009). Elections help incumbents maintain political coalitions and redistribution policies (Bueno de Mesquita et al. 2003), facilitate monitoring compliance and punishment noncompliance (Weingast 1996), and maintain political stability by sending signals to the international community that the regime is a reliable and credible partner (Hyde 2011). To conclude, I m convinced that elections are normatively neutral towards democratization or authoritarian strengthening and, therefore, they do not seem to be conducive per se to any unexpected outcomes for the incumbent. Elections may serve as a vehicle for the political alternation under certain circumstances and, thus, in this paper I would like to approach to the question is there any systematic variation in conditions and types of strategies leading to the maintenance of the status quo or unintended consequences. 2. Unintended Outcomes of Elections under Authoritarianism as a Phenomenon to be Explained As Andreas Schedler argues, the nested game of authoritarian elections may facilitate gradual processes of democratization by elections, as in Senegal or Mexico. It may lead to democracy through the sudden collapse of authoritarianism, as in Peru and Serbia in It may provoke an authoritarian regression, with a breakdown of the electoral cycle through military intervention, as in Azerbaijan in 1993 or Côte d Ivoire in It may also lead to extended periods of static warfare in which authoritarian incumbents prevail over opposition parties (Schedler 2006, p. 15). Elections may not only produce color revolutions or ritually reinforce the regime, but rather have a wide range of possible outcomes. Unfortunately, there is no systematized concept of the outcomes induced by the institutions of mass political mobilization in authoritarian polities. Scholars opt either for the straightforward measure (e.g. margin of victory) (Greene 2009; Schedler 2008) disregarding conceptualization, or constrain the possible outcomes to the regime change on the basis of fluctuations in scores of democracy indices (Howard and Roessler 2006; Kalandadze and Orenstein 2009). I suggest a conceptualization of outcomes of participatory institutions that is capable of grasping both short- and longterm effects. Unintended anti-hegemonic outcomes are those that severely diverge from the incumbent s expectations, i.e. goals are not achieved. Direct effects are immediate consequences brought about by elections, such as subsequent mass demonstrations or protests. Indirect long- or medium-term effects for the regime persistence include gradual eroding of the political regime and other types of cumulative effects. 5

6 Table 1. Possible Outcomes of Participatory Institutions PRO-HEGEMONIC ANTI-HEGEMONIC Direct Indirect Elections and referendums sustain an authoritarian regime Managing authoritarian arrangements; learning and adjustment Protest actions in the aftermath, immediate subversive effects, destabilization Gradual eroding, cumulative subversive effects The theoretically possible direct and indirect institutional outcomes mentioned in Table 1 are not mutually exclusive. Direct or short-term effects can be a constituent part of more distant and less visible effects. At the same time, indirect outcomes seem the most methodologically vulnerable categories since they are rarely observable; thereby, I run the risk of using them as residual cells for the a posteriori unexplained variance. In order to avoid this pitfall, it might be helpful to omit temporal dimension for the empirical analysis by limiting the types of post-electoral effects to those in the direct category. It would seem fruitful to extend Gordon Smith s model of two faces of referendums - supportive and catalytic - to elections. The first, pro-hegemonic versus anti-hegemonic dimension indicates the observable outcome which encompasses the immediate electoral outcomes such as a margin of victory and whether an incumbent or ruling party won or lost (Smith 1976, p. 5-7, 19-20). The second dimension indicates the presence of mass and/or opposition s coordinated pressure from below aiming to overthrow the incumbent. I employ mass and opposition protest actions as a single dimension, because only rarely does any bottom-up protest action take place without the active role of representatives from more organized opposition groups and/or as a consequence of intra-elite splits. 6

7 Mass and Opposition Protests NO YES Table 2. Types of observable outcomes PRO-HEGEMONIC OUTCOME (MARGIN OF VICTORY, WIN/LOSE) YES Belarus 2010 NO Ukraine 2004, Zimbabwe 2000 (referendum), Russia 2011, Chile 1988 (referendum) Russia 2008, Singapore (1990s present) Venezuela 2011 (referendum) Taiwan 2000, Mexico 2000, Uruguay 1980 (referendum) At the operational level, there are two ways of converting the constructed four types of outcomes into a measurable variable. Outlined scenarios may be treated either as categorical variables or as an ordinal scale of subversiveness. Anti-hegemonic effect is observed when a frank electoral failure takes place or the margin of victory is considerably lower compared with the previous electoral cycle. This measure is more sensitive to the cases with unsecure electoral victory than a simple win/lose indicator. The margin of victory measure is not to be used across cases (Schedler 2008; Greene 2009); rather the temporal comparison is required here, i.e. a relative loss or gain to the previous tenure. Chart 2. Differentiation between pro- and anti-hegemonic effects Did an incumbent or party officially lose? YES Radical antihegemonic NO Is the margin of victory considerably lower relative to previous elections? YES Subtle antihegemonic 7 NO Pro-hegemonic

8 What should be noted is that these four outcomes do not reflect any further change in political regime since I measure only immediately observable outcomes or a postelectoral situation with unexpected or undesired results. However, these four types of outcomes can be used as a minimal unit of analysis to trace the medium- and long-term effects on a political regime trajectory i.e. what I imply by indirect or cumulative institutional effects. Although in this paper I analyze exclusively radical anti-hegemonic outcomes: the incumbent s or dominant party s electoral failure which is operationalized through the simple binomial indicator where 1 stands for electoral victory and 0 for unambiguously lost elections. 3. Sources of unexpected or anti-hegemonic effects There is an impressive amount of work on authoritarian regime survival and breakdown. And as I have already mentioned elections per se are not the only mechanism which is responsible for the regime breakdown. What is more, the sources of persistence and breakdown mechanisms, though overlap, but are not exactly the same. Given that I do not explain the regime or existing political order breakdown, rather my task at hand is much less ambitious to explain why does incumbent lose or win elections. In other words, the flow of events after the official results having been officially declared are not included in my dependent variable. As to the sources of the political regimes breakdown, political economists suppose that social inequality and poor economic performance are the basic reasons causing the fall of dictatorships (Lipset 1959; Downs 1957; Gurr 1970). Another explanation has been advanced by Levitsky and Way, who stated that the impact of the international environment (linkage and leverage), especially from powerful countries or organizations that donate subsidies or invest in a country s economy or, finally, possess some specific political interests in a region, may speed up authoritarian breakdown or promote steady democratic transformation (Levitsky and Way 2010). Exogenous shocks such as financial crises can also contribute to the attenuation of a regime. For instance, reshuffles within the winning coalition and alliances with a wrong political partner may turn out to be fatal for the regime survival (Gasiorowski 1995). In the situation of the twin crises in Malaysia and Indonesia, the regime of the UMNO has managed to secure its positions due to the co-option of the holders of fixed capital and trade unions, whereas regime in Indonesia placed his bets on the Chinese owners of mobile capital that resulted in the collapse of the New Order and Suharto s resignation (Pepinsky 2009). Some authors address the sources of regime strength and persistence. The most popular explanation is economic prosperity and stable economic growth (Downs 1957; Przeworski 2000). Especially, under authoritarianism extensive state control over the economy, natural resources and redistributive politics may be particularly important in allowing the government to exercise control over elites and voters (Wintrobe 1998; Bueno de Mesquita et al. 2003; Ross 2001). I m interested in explaining whether there is an optimal strategy of repression and rewards holding basic structural factors relatively constant. Otherwise, can the gross blunders bring to life dormant features of the existing institutions, which were abused previously or, the other way round, extend the incumbent s life? Simply put, I hypothesize that even competitive elections with a wise adjustment of repression and rewards can prolong the incumbent s political life and, consequently, a political regime s existence, despite the common wisdom that political competition, as a harbinger of democracy, unavoidably leads to liberalization (Brownlee 2009). At the same time one 8

9 has to bear in mind that the choice of strategy is always limited by the more long-term conditions which define the available room for maneuver. Elections in authoritarian regimes also serve as a tool of the incumbent s support monitoring. At the same time, the incumbent s task is to monitor it without putting him/herself 3 at risk. Thus, the incumbent finds himself in a vulnerable position given that he/she does not know the level of his/her support in advance, and therefore cannot optimally plan redistributive and repressive strategies. This problem is known as the dictator s dilemma, i.e. he/she cannot know the exact level of support due to the information asymmetry which is inherent in an autocratic context (Wintrobe 1998; Rozenas 2009). The absence of reliable surveys, free mass media and trustworthy feedback from officials creates a distorting effect for an incumbent. This dilemma or Kafka s Castle syndrome seems to induce lapses and blatant errors, which disrupt the routine way of interaction between the power holder, coalition members, regional administrations, and voters. When an electoral campaign is carried out clumsily or in a ham-fisted fashion, it may open up scope for unintended anti-hegemonic consequences. The deficit of respect for the rules can be plugged by performance legitimacy, since electoral authoritarian regimes turn crucially on the level of skill by which their controls are recalibrated by rulers (Case 2006, p ). Unfortunately, this explanation runs the risk of being circular or endogenous as it relies on a posteriori estimates. In order to mitigate this problem I intend to control for the temporal sequence of events by using most of the indicators for the year preceding each election, including all instances where subversive outcomes may theoretically occur, and controlling for the conditions prior to and after elections. As Andreas Schedler has put it, authoritarian incumbents choose strategies, not outcomes, since they cannot fully control the effectiveness of their strategic moves (2008, p. 6). 4. Actors and their strategies on the electoral arena: operationalization At the micro-level, elections are reproduced by the strategic interactions between citizens, the opposition, and ruling parties or incumbents. Valerie Bunce and Sharon Wolchik add to this set of actors the role of the international environment, specifically local NGOs sponsored by the Western promoting democracy organizations. The quality of the opposition campaign does make a difference: the opposition unity, voters' registration and turnout drives, coordinated pressures on electoral commissions, collaboration between youth movements, civil society and the opposition. Public opinion polls and parallel vote tabulation allowed the political outsiders to put the incumbent and his government in a shaky position (Bunce and Wolchik 2010). However in this paper I would refrain of plunging into the problem of collective action literature and the drives of the mass protest actions. Admitting the merits of this approach it underestimates the incumbent s and political elite s strategies. Thus, I would rather shift the focus toward the incumbent s electoral strategies, especially taking into consideration that he/she has significantly more options for manipulation vis-à-vis the disgraced opposition and voters. For instance, Arturas Rozenas highlights the role of repression on the eve of elections and assumes the existence of the U-shaped relationship between the probability of electoral success and the level of repression. In other words both insufficient and redundant repression is equally dangerous for the incumbent leader. Under repression 3 To my best knowledge there is no female incumbent in the modern frankly authoritarian regimes, however, in order not to deprive women of possible political carrier of dictator I will use both personal pronouns. 9

10 he implies a wide range of the illegal actions from technical fraud and ballot stuffing to extra-judicial killings and tortures (Rozenas 2009). The use of ad hoc oppression is a sort of inherent feature of authoritarian regimes. However as many scholars recognize repression is hardly an explanation of the incumbent and more generally regime persistence. As Bruce Bueno de Mesquita and colleagues argue, the source of invincibility is not concealed in oppressive measures, but rather in the wise redistribution among relatively few members of the winning coalition and large number of the enfranchised, who remained beyond the circle of elite (Bueno de Mesquita et al. 2003). Both views, however, demonstrate only one side of the coin while obscuring the other. For the sake of clarity I have divided all possible moves into repressive and redistributive strategies towards the main domestic actors at the electoral arena. As I concede that this scheme oversimplifies the real picture, it still serves as useful analytical tool for systematizing complex interactions during elections. Table 3 depicts types of repressive and redistributive incumbent strategies relative to the political groups and elites, opposition and voters. Repression towards voters is actions involving direct violent actions by the police forces, secret agencies or paramilitary groupings, when buying strategies, target social programs, cyclical growth of pensions, social benefits or public sector wages (Treisman and Gimpelson 2001; Blaydes 2011). Elite groups such as business companies, the military, in some cases ethnic or religious authorities can be periodically purged or quite the opposite granted fat jobs, or opportunities for corruptive practices. Basically the same strategies are implemented to the opposition groups: they can be co-opted or severely repressed. For instance, dominant parties such as KANU in Kenya, NDP in Egypt, KMT in Taiwan or PRI in Mexico have been extremely successful in recruiting and promoting new strong and popular candidates in order to maintain necessary cadre rotation, improving the public image of a party and ultimately legitimizing the authoritarian rule (Hermet, Rose, and Rouquié 1978; Langston 2006). Table 3. Types of Incumbent s Strategies Voters Opposition Elite (coalition) Repressive Oppression or controlled vote oppression, exclusion Purges Redistributive Welfare, benefits, social spending; vote buying Co-option Corruption; Pork barrel politics 10

11 Repression can be detected either post factum from the international organizations expert indices and reports (Amnesty International, Electoral Watch, CIRI) or spending of the central government on public order and internal security. The first indicator reflects the actual repressive performance of the regime, when the spending merely points out to the incumbent s or his/her government policy choice. The watershed between distribution and repression from this perspective seems vague because this sort of expenditures is also exposed to the budgetary electoral cyclical manipulations as well as social benefits, pensions, subsidies and transfers. For instance, before the last electoral cycle in Russia the military expenditures have dramatically augmented as well as the wages of police officers their wages have doubled ("How a mysterious change to voting tallies boosted Putin at St Petersburg polling station: a citizen observer reports" ). Subsequently at the empirical level I consider spending on public order and security as investment in repression. In my paper I use both measures Cingranelli and Richards physical integrity index which measures the level of direct oppression and violent actions towards populations in the election year. Expenditures on public order and security as a share of all central government expenditures is a proxy for the regime reliance on repressive strategies. These data are available from 1975 to 2010 for 40 countries (Government Finance Statistics, IMF). By employing both measures I tentatively merge them into the category of repressive strategies, but the most curious results come from the separate analysis. For this reason sum of percentages on social spending is not presented in the paper. As to redistributive strategies, in many developing and non-democratic countries scholars revealed the cyclical changes in budgetary spending on the eve of elections. For instance, Lisa Blaydes provides the evidence of the existence of electoral business cycles in Egypt under Sadat and Mubarak rule where the main target constituencies were public sector employees, the urban poor and farmers (Blaydes 2011). Daniel Treisman and Vladimir Gimpelson (2001) drawing upon the analysis of the early Russian elections argue that the analytical concentration on the only one fiscal or monetary indicator appears to be a non-productive effort since it neglects the possibility that rational incumbents will choose a different portfolio of manipulations in different elections. In fact, incumbents in any given election have a variety of ways to affect voters' economic essessments. This portfolio includes rising minimum wages and pensions, monetary policy, different types of transfers, public spending, or tax cuts. If strategies require increased spending, this can be financed by increasing taxes, borrowing, or the money supply. These different elements of strategy will have different relative costs and benefits for particular incumbents in particular elections. Put otherwise, if there is no clear cyclical pattern in the variation of one indicator it does not mean the sheer absence of electoral business cycles and one should consider all available manipulative options at the incumbent s disposal. Unfortunately the cross-national data available from the IMF Government Finance Statistics do not allow testing hypotheses of the budgetary electoral business cycles, since for the majority of cases the annual indicators simply do not exist. Therefore, it is impossible to trace the annual changes. For this reason I have to delimit my task to the test of the basic associations between the probability of electoral victory and government spending patterns. On the plus side I can test different types of social 11

12 spending separately such as central government outlays on education, healthcare, and social security 4. To continue Treisman and Gimpelson s logic, I should remind that the very choice of strategies varies across countries and elections and it is heavily dependent on the constraints such as the resources available to the incumbent. Structural constraints include the impact of the international environment and state capacity of the regime itself, that is, along the line of Levitsky and Way s argument the international linkage may be transformed into effective leverage only in case of a weak state (Levitsky and Way 2010). The strength of a state, in turn, may be grasped through three dimensions extractive, coercive and administrative capacities. The first refers mainly to the ability to extract revenues and taxes, the second dimension involved the Weberian monopoly over the legitimate violence, that is, the centralized control over the military and police forces; finally, the third dimension refers to policy implementation and routine bureaucratic activities and their efficiency. I address only extractive and coercive aspects of state capacity 5 : military expenditures as a share of GDP 6, level of economic development as GDP per capita in constant US dollars prices 7, tax revenues as GDP share 8. Put bluntly, costs of repression and redistribution should not exceed the incumbent s objective capacities imposed by domestic socio-economic conditions and international environment. In the next part I try to test the objective constraints versus contingent strategies in electoral arena as associates of the probability either of victory or failure. The ambitious task of defining when incumbents are doomed to fail whatever strategies would have been chosen or, when structural factors are so favorable that even obvious mishits are blurred and invisible and, finally, when there is a space for contingency, does not seem feasible at the moment. Instead, I pursue a more modest goal of running associational tests between variables of interest. 5. Preliminary Analysis In this part I briefly describe my sample, case selection procedure and operationalization of variables, then, I move to the simple bivariate tests and multivariate regression analysis of the electoral victory probability, at the end I discuss preliminary results and possible interpretations. I argue that there is a room for further analysis of the stabilizing versus subversive conditions because there is still considerable variation in the outcomes among the regimes that are institutionally able to have electoral competition, even while they are unfree and are systematically violated Chart 1 represents the variation of the rate of 4 Government finance statistics yearbook / International Monetary Fund. Vol The very concept of state capacity and its measuring is a topic that deserves at least a separate essay. For an overview see (Evans, Rueschemeyer, and Skocpol 1985; Migdal 1988; Mann 1984); extensive theoretical overview see in (Taylor 2011) Taylor (2011); on the concept measurement issues see for instance (Hanson and Sigman 2011)Hanson and Sigman (2011). 6 Data source is the Stockholm International Peace Research Institute (SIPRI) // 7 Penn World Tables v. 6.3 (Heston, Summers, and Aten 2010) 8 World Bank Development Indicators (WDI) Annual Report. The World Band Data. 12

13 electoral failure in 349 competitive legislative and executive elections across 78 nondemocratic regimes. Out of 205 legislative elections, 32 were unambiguously lost by the incumbent s party; the rate of electoral failure among the executive elections is almost the same 23 out of 121. Excluding closed authoritarian regimes, hybrid and electoral democracies, there is still a significant variation in electoral outcomes. This variation cannot be attributed solely to the type of political regime or the level of competition or the type of political institutions; rather, other factors must be at play. Chart 1. Electoral Failure across Competitive Authoritarian Elections 9 ( ) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% parliamentary presidential lost won Source: NELDA; author s dataset; N of elections = 349; N of countries = 78 In this analysis I take into account only strategies towards voters and I deliberately neglect the importance of intra-elite redistributive practices. Moreover there hardly exists a relatively reliable indicator of spoils politics in cross-nationally comparable perspective. For instance, Jennifer Gandhi employs as a crude proxy the share of military expenditures to capture this phenomenon (Gandhi 2008). I have also included this indicator as a control variable which partly grasps the state capacity dimension and probably to some extent intra-elite redistribution as Jennifer Gandhi claims. As many scholars stipulate the central division between democracies and nondemocracies is a rule of the few at the expense of the many. Laymen are usually excluded from the decision-making process and demobilized (Przeworski 2000; Boix 2003). Redistributive practices towards population or simply provision of public goods such as healthcare, pensions, social benefits, education et cetera may seem redundant, especially in the case of absence of threatening protest mobilization. As the winningcoalition/selectorate theory predicts the incumbent would get more payoffs if sharing with his or her cronies but not with the population (Bueno de Mesquita et al. 2003). Nevertheless, authoritarian countries, particularly those holding regular elections with 9 Founding elections, elections under transitional government and formally noncompetitive elections are excluded. Elections in the failed states are excluded (e.g. Rwanda, Somali etc.). Elections, where it is impossible to define an incumbent or successor, are also excluded (e.g. Iran 2004, Morocco, Jordan, and Bhutan). I included elections only if Political Rights (Freedom House)t-1 >2 OR Polity IVt-1 < 6, where t stands for the year of observation, t-1 means year before elections. 13

14 higher levels of institutionalization, are supposed to provide some basic public goods and welfare. In a situation when voters are already more or less loyal, or at least neutral towards the regime, there is no need to resort to the costly repression. As Beatriz Magaloni has stated, any political regime, even frank autocracy, needs minimally sufficient level of popular support (Magaloni 2006). Autocrats need support in order to reduce the costs of maintaining power, frustrate opposition, and prevent mass mobilization. For these motives of legitimation 10 those regimes hold elections. Provision of public goods and the probability of electoral victory should be positively associated. Moreover, I expect social expenditures to be more significant than direct oppression since repression may be read as a sign of weakness. Repressive actions such as tortures, ambiguous imprisonments, violent suppression of mass protests, in particular, as instances of instrumental power, when effective redistribution by central government is an essential part of infrastructural power and, therefore, is to be more preferable option. For the empirical analysis I have collected all elections that took place from 1990 to 2011 in competitive authoritarian countries. I follow the strategy suggested by Mahoney and Goertz and pick all theoretically possible cases even if they are negative ones in order to avoid any selection biases (Mahoney and Goertz 2004). The lower threshold includes all positive answers to following questions: 1) are there elections at all? 2) is opposition allowed? 3) is there more than one party legal? 4) is there a choice of candidates on the ballot? 11. If one answer is negative, an election was excluded. The upper threshold has been drawn to distinguish between electoral authoritarian and electoral democracies. If the following logical expression is false: Political Rights (t-1) (Freedom House) <=2 OR Polity IV (t-1) >=6, where t means election year, then cases were classified as democracies or hybrid regimes and excluded from the sample. As a result the dataset contains 349 elections in 78 countries which can be qualified as competitive authoritarian, 55 out of them have been indisputably lost by the incumbent (Howard and Roessler 2006; Hyde and Marinov 2009). However the data on social spending are available only for 100 elections in 40 countries. This is why one should not overestimate the reliability of preliminary results since the sample can be potentially biased. I use statistical analysis rather as a descriptive tool than instrument for any predictions. In democracies it is relatively easy to define when does incumbent or his/her party lose elections or considerable share of the public support. Under authoritarian conditions electoral outcomes such as margin of victory are not such reliable measure due to the electoral fraud, uneven electoral field, and other sorts of abuse. To tackle this issue I plan to use alternative set of indicators which would tell us about the anti-incumbent effect. I use simple win/lose dichotomous variable which captures only radical antihegemonic effects where electoral failure is unambiguous loss of the incumbent in the presidential elections or his/her party in the legislative. 10 There is also external legitimation when the fact of having held free and fair elections serves as a signal of reliability. Specifically when a regime is a recipient of international aid. 11 Data are taken from the project Elections across Democracies and Autocracies (NELDA) // see 14

15 Table 4. Descriptive Statistics Variable Observations Mean Std. Dev. Min Max Polity IV GDP ppp pc t (in constant USD prices) GDP ppp pc t-1 (in constant USD prices) ΔGDP ppp pc Tax revenues t CIRI 12 Physical Integrity Index t Ethnic Fractionalization Win Military expenditures, % of GDP Expenditures on education (% of total central government spending) Expenditures on healthcare (% of total central government spending) Expenditures on social protection (% of total central government spending) Expenditures on Political Order and Security (% of total central government spending) Log (GDP ppp pc t) Log (GDP ppp pc t-1) Sources: Penn World Tables, WDI, IMF Government Finance Statistics, SIPRI, CIRI It is clear that fully-blown electoral failure is fairly rare and unlikely event, so in the available for the analysis sample only 17 failures out of 100 cases have taken place. In order to describe the probability of electoral failure and to test whether there is any sort of systematic variation I plug all variables in the logit regression models. I expect structural variables GDP, generous military expenditures and larger share of tax revenues to be positively associated with incumbent persistence. Social spending in general is to be conducive to the incumbent s electoral victory, when the effect of frank repression may be not so evident. For instance, too much repression could result in the regime de-legitimation and mass protests. Ethnic fractionalization aims at grasping the effect of the ascriptive cleavages within a country which may serve as an additional source of de-stabilization and higher rates are expected to increase the probability of adverse outcome for the incumbent. Finally, even having said that the sample contains exclusively the authoritarian regimes with some space for electoral competition there is 12 This is an additive index constructed from the Torture, Extrajudicial Killing, Political Imprisonment, and Disappearance indicators. It ranges from 0 (no government respect for these four rights) to 8 (full government respect for these four rights). Details on its construction and use can be found in: David L. Cingranelli and David L. Richards "Measuring the Level, Pattern, and Sequence of Government Respect for Physical Integrity Rights." International Studies Quarterly, Vol 43.2: / 15

16 still a variation on the political regime type: some regimes are more repressive, others are less. To tackle this issue of regime heterogeneity I include Polity IV index which mainly reflects the extent of openness and fairness of existing institutional rules. By employing this index against the Freedom House measure the issue of potential reliance on post hoc estimation is partly disentangled, otherwise stated, Polity IV estimates the institutional expression of authoritarianism rather than performance. Table 5 shows the b-coefficients, average marginal effects (AME) and robust standard errors of every predictor 13 introduced earlier. However, the reader should bear in mind that b-coefficients are not technically comparable across the models and they express mainly intensity and direction of causal effect within a given model (for more details see Bartus 2005). Instead of b-coefficients I use average marginal effect to compare models with each other. Average marginal effects express the average of the variation induced in the probability of interest by a marginal change in an independent variable for each individual in the sample. Specifications 1 and 2 model the effects of the long-durée structural factors on the probability of electoral victory. Models 3 and 4 demonstrate the impact of more contingent variables such as repressive and redistributive strategies. The last model demonstrates detected significant interaction model between health expenditures and level of economic development. The cases have been chosen basically according to the availability of the data for the independent variables, especially social spending. This means that I do not deal with a proper random sample. Therefore the indicators of statistical significance are of almost no relevance since I cannot make reliable generalizations at the moment. Nevertheless if we imagine that the sample is random the best model is the redistribution model because all predictors are statistically significant. One percentage point increase in spending on education relative to other budgetary spending items accounts for 3% increase in the probability of the incumbent electoral victory. Expenditures on social security and welfare reveal less clear pattern (significant only at 90% confidence interval) and induces only 1% increase in the probability of incumbent success. The most striking finding is the negative association between the share of health expenditures and likelihood of electoral success. Moreover this estimate is robust in the bivariate tests and it does not correlate with other types of expenditures. It might be the case that the regimes with privatized healthcare systems are more stable than with those where the healthcare system is regulated by the state. In order to shed more light on this counterintuitive linkage I run an additional model with only health spending variable controlling for the level of economic development. I found out that there is a strong significant but negative relationship between GDP per capita and health spending (see the model 5 with the interaction term is presented in Table 5). The higher level of GDP and larger share of expenditures on health produce stronger antihegemonic effect. Lower levels of economic development and tight-fisted politics in healthcare are also more dangerous for incumbent s electoral position. However a combination of higher per capita income and lower share of healthcare spending and, vice versa, low levels of GDP per capita and generous health expenditures would attenuate this association. An interpretation might look as follows: the voters in the regimes with higher levels of economic development per capita cannot be deluded by 13 As far as there are groups in the sample, e.g. more than one election in a given country I have tested the same models with country-adjusted robust standard errors. The estimates turn out to be pretty mush close to those ones presented in Table 5. 16

17 the higher share of health expenditure, but poor voters are more sensitive to the health spending priorities relative to other types of social policies. Central government expenditures on education are positively associated with electoral victory. If an incumbent decides to allocate 1% more budgetary resources at the expense of other items it can increase the chances of victory by 3% on average. At the first glance it does not make too much sense because most of people who are involved in education have not reached the age threshold to vote yet. Is it a wise strategy of resource allocation in favor of the potential non-voters? Although taking into account that education comprises all stages of education it involves students getting the tertiary education and, what is more relevant, schools and university staff. The evidence from the Mexican and Russian elections that school teachers were actively involved as principal contractors in fraudulent practices at the precinct election commissions (Plechov 2011; "How a mysterious change to voting tallies boosted Putin at St Petersburg polling station: a citizen observer reports" ). It is not a secret that in most countries polling stations are located in administrative and also school buildings. The socio-economic profile of teachers is female, low- or middle-income and highly dependent on wages and their workplace. For instance, Mexico s National Union of Education Workers (SNTE) was an extremely powerful association that strongly pushed through its corporate interests and was one of the important allies of the regime under the PRI rule. Lastly, university students are also an important constituency which is more exposed to any sort of protest and anti-regime actions. Perhaps the provision of relatively high standard education plays a role of safety valve and channel potentially subversive activities into more controllable outlets. If to calculate the sum of all social expenditures the model ceases to be different from the empty model (without predictors). It means that one should differentiate between various patterns of government expenditures; they may bring about oppositely directed effects conditional to structural factors. 17

18 Table 5. The associates of the electoral victory probability: Logit regression estimates. Dependent variable win or lose. Variable Structural Factors Models Electoral Strategies Models Model with Interaction M1 M2 M3 (Repression) M4 (Redistribution) M5 GDP*Health *** (0.00) GDP ppp pc t-1 (mean centered) *** (0.00) Polity IV (0.09) (0.09) Log (GDP ppp pc t-1) (0.31) (0.39) Military expenses t-1 (% in GDP) (0.13) (0.15) Tax revenues t-1 in constant USD prices (0.08) (0.09) Ethnic Fractionalization (1.88) CIRI Physical Integrity Index Expenditures on political order and security (0.16) * (0.06) Expenditures on education *** (0.05) Expenditures on health ** (0.07) -0.20** (0.08) Expenditures on social security 0.01 and welfare 0.05* (0.03) Constant (2.57) (4.46) 0.72 (0.91) (0.91) 3.18*** (0.76) 18