Does it make a difference? Hybrid regimes and their consequences on citizens wellbeing

Transcription

1 Does it make a difference? Hybrid regimes and their consequences on citizens wellbeing Abstract Many processes of regime transition that occurred during the so-called third wave did not result in full democratization. The recent proliferation of hybrid regimes political entities characterized by a mixture of authoritarian and democratic institutions raised several questions about the consequences of this phenomenon. In particular, while a transition to democracy is typically said to bring about also an improvement of citizens quality of life, it is not clear whether we should expect a similar change after the institutionalization of a hybrid regime. A time-series cross-sectional analysis on the consequences of hybrid regimes on several indicators of material well-being reveals that, because of their peculiar institutional nature, hybrid regimes performance differs from both democracies and other non-democratic regimes.

2 Andrea Cassani Ph.D. Candidate Dept. of Social and Political Sciences Università degli Studi di Milano Draft version, comments welcomed. Please do not quote without author permission. Paper prepared for delivery at the panel The economic, political, social consequences of democratization (Chairs Giovanni Carbone, Davide Grassi), XXVII Annual Conference of the Italian Political Science Society (SISP) Florence, September 12 th -14 th This paper is part of a research project on The economic, social and political consequences of democratic reforms. A quantitative and qualitative comparative analysis (COD), funded by a Starting Grant of the European Research Council (Grant Agreement no , Ideas, 7th Framework Programme of the EU).

3 Introduction Between the mid 70s and the first half of the 90s, a sensational number of transitions from authoritarian rule occurred in close sequence all over the globe. Huntington captured the significance of that moment with the idea of a third wave of democratization (1991) that, one after the other, was overwhelming Mediterranean Europe, Latin America, Eastern Europe, Asia, and sub-saharan Africa. Beyond that evocative image, however, the reality was much more varied. Many of these processes of regime change effectively resulted in the introduction of political democracy. For several others, however, the outcome was less certain. Often the crisis of an existing authoritarian regime caused its collapse, started a phase of transition often accompanied by the partial opening of the political system, yet it did not lead to democracy. Not always the call of free elections, not even their institutionalization as the main instrument to gain political power, corresponded to the democratization of a country. The explanation of the non-linear trajectory of political change followed by these regimes might differ from one case to another being it either the absence of the economic, social, cultural prerequisites (for a review see Diamond et al., 1990), political elites merely instrumental commitment to democracy, or the lack of linkages to the West (Levitsky and Way, 2010). In virtually all cases, however, the results was the institutionalization of a hybrid form of political regime, characterized by the co-existence and interaction of formally democratic institutions with persistently authoritarian practices of governance. This led to the formation of a gray zone between democracy and autocracy (Carothers, 2002). Despite initial skepticism about their non-ephemeral nature, hybrid regimes have triggered a lively academic debate. The relative novelty of this political phenomenon challenged most consolidated theories of democratization and raised many new questions. Researchers studied the origins of hybrid regimes (Levitsky and Way, 2002; Schedler, 2002; Ottaway, 2003), theorized about their functioning (Lust-Okar, 2006; Magaloni, 2006; Gandhi and Przeworski, 2006; Brownlee, 2007; Boix and Svolik, 2008), and analyzed their endurance (Epstein et al., 2006; Hadenius and Teorell, 2006; Howard and Roessler, 2006; Brownlee, 2009). These works made a valuable contribution to our understanding of the phenomenon. We learned about the causes of its recent spread in conjuncture with the end of the Cold War. We understood that, rather than a source of fragility deriving from their apparent incompatibility, democratic and autocratic institutions may interact in ways that elude conventional wisdom, and even strengthen the hold on power of the incumbent leaders. To date, however, relatively little attention has been paid to an issue of even more substantive interest: the consequences of this form of political regime on the wellbeing of their citizens.

4 The crisis and collapse of an authoritarian regime are invariably accompanied with hope and demand for a better future. Not surprisingly, researchers found evidence that a democracy s chances to survive are enhanced by socioeconomic development (Przeworski et al., 1996), while the ability of democratic countries to provide it has recently triggered a lively debate. Similarly to many young democracies, contemporary hybrid regimes emerged from the failure of the development goals legitimizing heavily repressive dictatorships. Therefore, it is very likely that also the fate of a newly established hybrid regime will be influenced by its ability to meet citizens expectations. The research agenda on hybrid regimes, therefore, needs to be enriched with a new line of inquiry. The investigation into the consequences of hybrid regimes should complement the study of their functioning and the analysis of their future prospects. Do hybrid regimes promote social welfare? Are they better or worse providers of public goods than their respective plainly authoritarian and fully democratic counterparts? Does the introduction of some formally democratic institutions in a context of persistently authoritarian governance make a difference? Is it enough to determine an improvement in people lives? The goal of this paper is to start addressing these questions, from both a theoretical and empirical point of view. The next section delves into a preliminary but fundamental issue: the conceptualization of the notion of hybrid regime. Following a review of the relevant literature, the third section proposes a few hypotheses about how hybrid regimes should be expected to perform in the field of social welfare. The main argument is that, given their peculiar institutional nature, any superficial comparison with democracy and other forms of authoritarianism would be misleading. Finally, the paper presents the results of an empirical analysis conducted on a sample of 132 developing countries observed from 1980 to 2010 on the consequences of hybrid regimes on several indicators of well-being in the sectors of education and health. Defining hybrid regimes The conceptualization of the object of study, hybrid regimes, is the necessary first step of any research work. Conceptualization is the formulation of a systematized concept through reasoning about the background concept, in the light of the goals of research (Adcock and Collier, 2001: 531). A background concept is the broad constellation of meanings ( ) associated with a given concept (ivi). A systematized concept, in turn, is a specific definition of the concept under examination. The latter should not be thought of as a mere list of the attributes that form the intension of a given concept (Sartori, 1984). It is also the result of an ontological effort to establish how to see, or perceive, the corresponding object. Given the same set of defining attributes, depending on our ontological preferences we will coin different concepts, study a phenomenon

5 from different perspectives, identify different empirical referents, and plausibly get different conclusions. Conceptualizing hybrid regimes, therefore, requires the prior analysis of how the latter have been defined by students of the phenomenon, and the formulation of a specific definition that will orient our research. Preliminary evidence suggests that the debate on hybrid regimes has experienced a proliferation of alternative approaches to the conceptualization of the notion. A survey of the most influential studies on the topic published during the last ten years demonstrates that scholars frequently differ in their opinion concerning the identity of hybrid regimes. Five broad alternative positions can be identified. Scholars defined hybrid regimes as either a diminished type of democracy, a diminished type of authoritarianism, a third intermediate type of regime, an outright instance of authoritarianism, or a specific subtype of non-democracy. The causes of this proliferation should be traced back to the rapid, somewhat chaotic, development of the debate. The consequence is conceptual confusion. In order to formulate a definition of hybrid regime without falling in a conceptual impasse, we should go beyond existing divergences. The way to do so is to move from the macro to the micro level. The macro-level is the regime level which is the main source of confusion about the identity of hybrid regimes. The micro-level, in turn, is represented by those individual institutions notably, the ones regulating the procedure of leadership selection and the exercise of power (Munck, 1996: 5) that, taken together, define a political regime. Once the macro-level obstacles are eliminated, we may note that all conceptualizations of hybrid regime rely on the following core set of microlevel institutional attributes: (1) periodic multiparty elections for the selection of the executive; (2) an elected legislature in which the opposition is represented; (3) few limits to the arbitrary power of the chief executive, and (4) frequent violations of the citizens political and/or civil rights. Starting from this core set of defining attributes, it is also possible to reason about the identity of such a political regime. Given the context of the present research the investigation into the consequences of a specific form of regime conceptualizing hybrid regimes as a subtype of nondemocracy has some merit. First, it represents a fair balance between parsimony and accuracy. It is parsimonious in that it preserves the basic dichotomous division between democracy and nondemocracy. It is accurate because it distinguishes neatly a hybrid institutional arrangement from other non-democratic regimes. This allows a neater and fine-grained comparison that neither diminished types because of their often blurred lines nor third types given their tendency to become residual categories could achieve. Second, from a theoretical point of view, it better highlights the transformative potential that the interaction of democratic and authoritarian

6 institutions has on a political regime s structure, which would be missed by sticking with a simple dichotomy. To conclude, hybrid regimes are defined as a specific subtype of autocracy characterized by the presence of formally democratic institutions. A hybrid regime is authoritarian for two reasons: it does not fulfill all the requirements to be a full democracy; and less than full democracy is just synonym of non-democracy, or autocracy (cf. Sartori, 1987; Alvarez et al., 1994). Contrary to the cited authors, however, in order to capture hybrid regimes peculiar nature, we should go beyond a purely procedural definition of democracy. The democratic component of hybrid regimes refers to the presence of a limited form of political competition, put into effect by the institutionalization of periodic multi-party and multi-candidate elections for both the executive and legislative, in which opposition is allowed to run, gain seats in the legislature and, in principle, to win (Hyde and Marinov, 2011). This is to say that democratic and hybrid regimes formally abide by the same fundamental rules of the game. This procedures matter in that they makes hybrid regimes different from both exclusive political systems such as military and hereditary regimes and from other inclusive yet non-competitive regimes such as single-party regimes. By focusing on this sole element, however, we would overlook what makes hybrid regimes so different from democracies: the frequent/systematic violations of the same rules deriving from persistent restrictions of political and civil rights, and the lack of constraints to the executive power, which substantially spoil opposition s chances of victory. So defined, the label that best captures this form of political regime is probably the one coined by Levitsky and Way (2002). Hybrid regimes are competitive authoritarian regimes. Argument and hypotheses Before delving into theory, it should be clarified that the primary goal of a political regime is to provide an efficient mechanism for the selection of the leadership and, more generally, to define the rules that regulates the politics in a country. Political regimes are not meant to provide for social welfare. Governments and their public policies are. The socioeconomic consequences of a political regime, if any, are a by-product (Carbone, 2009: 127). When analyzing the consequences of hybrid regimes on social welfare, therefore, we are actually investigating whether, how, and to what extent the routine functioning of their institutions the incentive and constraints they engender may (indirectly) influence a government s attitude/commitment to promote it. In the absence of a full-blown theory of the socioeconomic consequences of hybrid regimes, I suggest, we need to unpack the phenomenon, adopt a comparative perspective, and analyze the

7 consequences that three elements namely the authoritarian nature of the regime, the presence of democratic institutions, and the consolidation of such a regime may have on citizens wellbeing. The authoritarian dimension As a starting point we should clarify why the leader of a government should care about citizens wellbeing. The literature on the consequences of political regimes generally agrees on two important points. First, from a political leader s point of view, to meet citizens needs is one viable strategy to gain the support and the cooperation of the society in order to pursue his/her own interests, whatever they are. Second, the likelihood of choosing that option depends on the institutions of a political regime. What, on the contrary, is not always highlighted is that in this context democratic and authoritarian leaders are not really different. Often autocrats are implicitly assumed to prefer violence and repression by default, given the power and the degree of arbitrariness they enjoy. When dealing with authoritarianism, to invest in citizens support and cooperation is too often considered just not an option. With very few exceptions, however, the image of a legibus solutus despot is too simplistic a description of any contemporary autocrat. Olson describes authoritarian leaders as stationary bandits willing to boost the productivity of the society in order to increase their own revenues (1993). Wintrobe depicts them as neurotic dictators fearing for personal safety (1998). In both cases, autocrats do have incentives to improve citizens wellbeing. Starting from this assumption, we should weigh the relative cost of this option for an authoritarian leader compared to a democratic one. In this context, autocracies differ from democracies in two important aspects: the lack of communication between government and society; the absence of an effective mechanism of political competition. Authoritarian regimes are invariably characterized by few limits to the power of the chief executive and by the relatively easy recourse to violence (or realistic threat) in order to repress dissent. Because of the absence of institutionalized channels through which citizens may signal their true preferences, citizens are reluctant to signal their true preferences. Authoritarian governments, as a consequence, suffer from an information deficit that tends to make policy making inefficient. The incentives they have to invest in citizens wellbeing in order to buy off their support and solicit their cooperation are therefore counterbalanced by the difficulties to implement public policies on target. Authoritarian regimes also lack an effective (i.e. functioning) mechanism of political competition. Several theories agree that the institutionalization of a procedure of leadership selection based on the competitive struggle for the people s vote (Schumpeter) may have positive effects on the quality of life of the latter (cf. Meltzer and Richards, 1981; Lake and Baum, 2001). Yet non-

8 democratic regimes are invariably characterized by what Bueno de Mesquita and his colleagues defined a small winning coalition (1999; 2003). Authoritarian and democratic governments differ from each other because they need different numbers of supporters, i.e. of people whose support is to be bought off. In particular, given the small size of his/her winning coalition, autocrats will find more efficient to maximize the welfare of their backers by investing in private (rather than public) goods. As other forms of authoritarianism, hybrid regimes limit the channels through which citizens may communicate their dissatisfaction, preferences, and needs. As in other dictatorships, incumbent leaders in hybrid regimes need the support of a few key actors, rather than the majority of voters. Therefore, the authoritarian dimension of a hybrid regime seems to counterbalance and nullify the existing incentives for a government to invest in the promotion of social welfare. Our first testable hypothesis is: HP1. Hybrid regimes provide social welfare less efficiently than democratic regimes. The democratic dimension Albeit authoritarian, however, hybrid regimes however are also characterized by the presence of formally democratic institutions. So far, the analysis of formally democratic institutions elections, parties, and legislatures in authoritarian contexts has focused on how they have been used and/or manipulated by incumbent dictators in order to pursue strategic goals, all ultimately extending the duration of their rule. To the extent that their presence makes a hybrid regimes different from other subtypes of authoritarianism, however, it should also be expected to shape politics, including the options available to a political leaders and consequently his/her attitudes toward the welfare of the citizens. Following Bueno de Mesquita s argument, the presence of similarly democratic institutions in a non-democratic context triggers a sort of vicious circle that may ultimately result in a further deterioration of citizens quality of life. The author describes the democratic dimension of a hybrid regime as the institutionalization of a large selectorate. Similarly to democracies, by institutionalizing periodic free elections hybrid regimes grant virtually all citizens the right to have a formal say in choosing the government (1999: 148). Contrary to democracies, however, elections in a hybrid regime do not really allow citizens to vote an incumbent out of office. His/her fate rather depends on the support of a still relatively small group of actors. Therefore, in a hybrid regime the enfranchisement of the selectorate corresponds to an enlargement of the pool of potential supporters that may become members of a political leader s small winning coalition. This peculiar institutional environment is particularly fertile for the adoption of predatory attitudes. The more easily

9 replaceable each member of the current winning coalition, the less political leader needs to invest in his/her loyalty, the more state revenues the dictator may save for own interest. Similar conclusions are rigorous from a strictly theoretical point of view. Yet they overlook the transformative potential of democratic institutions in an authoritarian context and their consequences on political leaders options with respect to their priorities of buying off support and soliciting cooperation. Autocrats, it has been said, seek the support of the masses in order to thwart rebellions and coups (Geddes, 2005). They would be willing to meet at least part of their demands, but they lack credible information about citizens preferences, priorities, and needs. Periodic multi-party elections may represent an efficient means to collect information. Voting for the opposition in a hybrid regime can hardly contribute to defeat the ruling party. Yet it is not meaningless. Citizens can use their vote to reveal their preferences at a relatively low cost. Voting for an opposition party that abides by the rules of the game has less consequences than campaigning against the government or taking the streets. The fragmentation that typically affects the opposition fronts in a hybrid regime (Lust-Okar, 2004), moreover, may even improve the quality of the information contained by a vote. Disunited opposition parties are likely to have different programs, and promote different interests. Their supporters have no incentive to strategic voting behaviours. Voting for a specific opposition party, therefore, may signal citizens sensitivity toward a specific issues. Autocrats, moreover, have an interest in soliciting the economic cooperation of the society in order to increase their own returns (Gandhi, 2008). They would be willing to incentivize citizens productivity by investing in their wellbeing, but their commitment is not credible. Multi-party elected legislatures may serve as a forum to reach a compromise between the government and the different sectors of the civil society (Gandhi and Przeworski, 2006). They may represent an arena where cooperation and compliance may be traded for policy concessions. Rather than triggering a vicious circle, therefore, the democratic dimension of hybrid regimes tends to mitigate the distortive effects of their authoritarian nature. Hybrid regimes differ from other forms of authoritarian rule because formally democratic institutions partly compensate the negative consequences deriving from the lack of an effective mechanism of political competition. Rather than nullifying political competition, hybrid regimes limit its consequences on the executive, while taking advantage of its beneficial effects in terms of quality of information about citizens preferences/needs. To conclude, hybrid regimes multi-party elections and legislatures seems to be able to partly fill the gap between authoritarianism and democracy. Our second hypothesis is: HP2. Hybrid regimes provide social welfare in a more efficient way than other nondemocratic regimes.

10 The consolidation of a hybrid regime A third factor that should be taken into account is the consolidation of a hybrid regime and its potential effects on the dynamics described above. Different types of political regimes have different prospects of consolidation (Geddes, 1999). The ability of hybrid regimes to endure, in particular, has been the object of intense scrutiny during the last decade and nowadays scholars generally agree that hybrid regimes are proving more stable than initially expected. Accordingly, hybrid regimes may consolidate and this is likely to have consequences. The analysis of the consequences of regime consolidation on citizens wellbeing has almost exclusively focused on democracy. In very general terms, it has been argued that regime consolidation has a positive effect on the efficiency of policy-making which, in a democratic context, results also in a more efficient provision of public goods. The empirical evidence strongly confirmed this and similar arguments (Muller, 1988; Gerring, 2005). Scholars however seldom investigated the effects of the consolidation of a non-democratic regime. In most cases, the neglect was justified by the concern that, given the arbitrariness typically enjoyed by dictators, authoritarian governance is too unpredictable to consider consolidation a significant intervening variable. The exception is Bell s recent work on the effect of regime consolidation in democratic and non-democratic countries (2011). Hinging on the Selectorate theory, the author argues that regimes characterized by a large winning coalition become better providers of public goods as they consolidate, while in the early phases following their institutionalization they need to appease dissatisfied elites with private goods. As small winning coalition regimes consolidate, on the contrary, governments get insulated from outside pressure and they no longer need to deter popular uprisings with public goods. Similar considerations may hold true also for hybrid regimes. The more consolidated the existing structure of power, the more frustrated opposition ambitions and expectations are likely to be. In the absence of alternation in power, the continuous and repeated practice of a formally democratic routine tend to dissipate its initial benefits. Facing the impossibility of a change of government, opposition leaders are increasingly likely to seek rents. Incumbents governments will no longer deal with an opposition responsive to its constituents, but with leaders willing to get co-opted and. The lack of alternation in power will also increase voters dissatisfaction with politics, and their indifference. The consolidation of a hybrid regimes, in other words, is likely to make the above described mechanisms less efficient. Our third hypothesis is: HP3. Hybrid regimes become worse providers of social welfare as they consolidate. Empirical analysis

11 The three hypotheses discussed in the previous section have been tested by means of a time-series cross-sectional (TSCS) analysis, working on a sample of 132 developing countries observed from 1980 to Countries have been selected on the basis of the following criteria: being a current member of the UN, having more than 500,000 inhabitants, not being a OECD high income (GNI >= $12,616) member. Newly independent countries former colonies, former subunits of a larger state entity have been observed since the year of their international recognition. The dataset consists of 3,554 observations. The remainder of this section is devoted to the presentation of the analysis results. I proceed by illustrating the dependent, the independent, and the control variables, describing how the analysis was performed, and then conclude by commenting the results. Dependent variable As dependent variables, twenty-five indicators of the performance of a country in the sector of education (15) and health (10) have been selected. In so doing it has been tried to give an exhaustive overview of the consequences of hybrid regimes on the two non-economic dimensions of the concept of human development (UNDP) namely the possibility to lead a long and healthy life and the opportunity to expand personal choices. When data were available, special attention has also been paid to the distribution of wellbeing across genders. In order to draw a picture as complete as possible of the effects of hybrid regimes on the sectors of education and health, the selected indicators refer to both inputs (i.e. public spending) and outcomes. A complete list of the indicators used, along with a summary of the results is available in the Appendix at the end of the paper. All indicators of well-being have been collected from the World Bank Development Indicators dataset. Independent variable Regime type is the independent variable of the analysis. A first raw distinction has been made between democratic and non-democratic regimes. Following Przeworski et al. (2000), necessary (non-sufficient) conditions for a country to be democratic are: popular election (either direct or indirect) of the chief executive; an elected legislature; multi-party elections. Although I agree that alternation in power is the ultimate goal of these three conditions, however, I do not consider the observation of an executive turnover to be a sine qua non requirement. Unlike the authors, moreover, additional necessary conditions for a democracy to be in place are: extensive protection of both civil and political rights; substantial limitations on executive authority. Non-democratic regimes have been classified as follows: hybrid regimes, hereditary regimes, military regimes, one-party regime, hegemonic-party regimes, plus a residual category other. Cases of non-independent authority either periods of foreign occupation or failed states have

12 been kept out of the analysis. Without delving into the coding rules of each non-democratic regime type, a hybrid regime has been recorded any time the following conditions jointly hold true: an elected (either directly or indirectly) chief executive and legislature; multi-party elections; a multiparty legislature. The five sub-types of non-democratic regime, the democratic type, plus the residual and the noauthority categories, are mutually exclusive and together they form an exhaustive typology of political regimes. Without compromising these desirable properties, the two -party types could be collapsed in a single-party category, while the four non-democratic/non-hybrid types (plus the residual other ) may form a more general other authoritarian category. By so doing, it is possible to perform the analysis of HP2 at three different levels of generality: (1) hybrid regimes vs. [other authoritarian] regimes; (2) hybrid regimes vs. [military, hereditary, single-party, other] regimes; (3) hybrid regimes vs. [military, hereditary, one-party, hegemonic-party, other] regimes Control variables Starting from a base model common to all the dependent variables analyzed, depending on the nature of the dependent variable either an input or an outcome and the sector either education or health slightly different model specifications have been built. The base model for hypotheses 1 and 2, that were tested jointly, included the main variable of interest, regime type (divided in regime type dummies), a variable for regime consolidation, and a measure of national income. Hypothesis 3 was tested separately by adding to the base model an interaction term between the consolidation variable and the hybrid regime dummy. Regime consolidation was measured as the incremental number of consecutive years a given regime has been in place. To be sure, I consider this factor of primary importance when the definition and implementation of public policies is concerned. Yet, because the analysis focused on both democratic and non-democratic regimes, I do not have specific expectations on the effect (either positive or negative) of that variable. I included a measure of gross domestic product per capita (PPP converted at constant prices, taken from Penn World Table) for three reasons. First, preliminary analyses showed that GDP is correlated both with most dependent variables and with regime types. Second, scholars generally agree that wealthier citizens are typically more concerned about the quality of public services (Lake and Baum, 2001). Third wealthier countries are better equipped to provide them (Brown and Hunter, 2004, Stasavage, 2005), by devoting a larger portion of tax revenues to these sectors without producing acute trade-offs between them and other public sectors (Ghobarah, 2004; Mulligan et al., 2004).

13 The control variables that complete the specification of each model vary depending on the indicator of wellbeing under examination, but they do not change from a hypothesis to another. The specification of each model was driven by two goals. I sought completeness by including at least one variable representing each family of factors economic, demographic, political that, based on literature and personal judgment, may influence the dependent variable under examination. I pursued parsimony by limiting the number of variables referring to the same family, and related risks of redundancy and collinearity. Unless differently specified, all control variables referring to economic and demographic factors come from the World Bank dataset. I consider government spending on education and health to be influenced mainly by economic factors. A country s economic performance represents a significant constraint on government expenditures on these two sectors. Yet, its effect is rather controversial. On the one hand, one might expect that periods of economic growth correspond to higher state revenues, and subsequently to a larger budget to invest (Brown and Hunter, 2004). On the other, it should be noted that social spending tend to be countercyclical: to remain low in time of growth and to expand during recession (Rudra and Haggard, 2005). Following Mulligan et al. (2004), in the model for education spending I also controlled for the fraction of national value added from agriculture because it might signal the relative importance assigned to education in a country. When health spending were analyzed, in turn, I also controlled for size of a country s dependent population (under 14 and over 65), because health services are most often demanded by both the young and the old (Gandhi, 2008). I consider education and health outcomes to be influenced mainly by demographic factors. Both the actual provision of education services (especially when primary and secondary levels are concerned) and of basic health care services are likely to be affected by the geographical distribution of the population, notably the urban-rural divide, although its effects may be hardly predictable. On the one hand, it is relatively easier to provide such services in concentrated urban environments than in more sparsely populated rural communities (Lake and Baum, 2001). Following Keefer (2005), moreover, the percentage of population living in urban areas is also a proxy for how informed citizens are about the quality of public services and consequently for the content of their demands. On the other hand, fast-paced urbanization often leads to the proliferation of peripheral urban slums, where people are exposed to several health risks, where school infrastructures are scarce, and citizens are politically marginalized (Ghobarah, 2004). In the models for education outcomes both enrollment and attainment I also included a measure of the raw size of young population. The higher the number of school-age residents in a country, the more challenging is to ensure education to all. A negative sign of the coefficient would cast doubts on the actual efficiency of the service. In the models for some health outcomes namely infant

14 mortality and immunization I also controlled for fertility rates: the higher the number of new born, the higher the number of infant to be vaccinated, and, in situation of inadequate access to neonatal cares, the more children die (Zweifel and Navia, 2000). In the analysis of infant mortality, a measure of HIV infections prevalence rates was included because, unlike other diseases, AIDS is not a direct consequence of poverty and lack of health care facilities and therefore not directly under the control of a government (Ross, 2005). Finally, when life expectancy was under scrutiny, I also included a measure of demographic pressure, because a rapid population growth is often considered as having a negative effect (Frey, 1999). Additional control variables, whose effects have been tested throughout the different models are: trade openness (Brown and Hunter, 2004; Gandhi, 2008; Rudra and Haggard, 2005), external debt (Rudra and Haggard, 2005), external development assistance (Stasavage, 2005), relative weight of oil rents (Brown, 1999), ethnic fractionalization (Ghobarah, 2004; Stasavage, 2005; Vollmer, 2009), Muslim majority (Brown, 2004), British colonial past (Brown, 2004; Mulligan, 2004); current/former communist country (Mulligan, 2004; Ross, 2005), war fought within the country borders (Ghobarah, 2004; Vollmer, 2009). Analysis TSCS analysis presents several advantages in terms of analysis, robustness of the findings, and interpretation. Yet it also raises severe challenges to some fundamental assumptions of the linear regression model, notably no specification errors, homoskedasticity of the conditional distribution of the errors, and non-serially correlated residuals. A first concern derives from the possibility that when pooling a large number of countries together, we are overlooking important sources of heterogeneity whose omission may bias our results. Given the difficulty of identifying such factors, the inclusion in the model of panel fixed effects (FE) country dummies to control for all those unit-specific unobservable elements that do not vary over time has at least the merit of admitting our ignorance. Each time it has been used, the appropriateness of the FE model has been assessed by a Hausman test and by a Wald test of joint significance of the country dummies. In the presence of heteroskedasticity, a viable strategy is to compute the standard errors through a Huber/White ( sandwich ) estimator of variance that is robust to violation of the assumption of constant variance of the errors. Whenever serial-correlation has been detected, the corresponding dynamic of stickiness has been modeled as part of the specification. Following Beck and Katz (1995), in particular, it has been assumed that the errors follow an autoregressive process, and as many lags of the dependent variable as warranted by a Lagrange Multiplier test have been included in the model specification. Furthermore, in order to limit endogeneity concerns, each regression has been re-run

15 using one-period lagged independent variables, and the results compared on the basis of AIC and BIC scores. As an alternative to the FE model, each dependent variable has also been tested by estimating the coefficients with OLS while computing panel-corrected standard errors (PCSEs). Since the publication of Beck and Katz s seminal article (1995), this practice has become standard in TSCS analysis. According to the authors, PCSEs outperform robust SEs because they allow to deal with both ordinary, panel-level heteroskedasticity, and contemporaneous correlation at a time. The PCSEs model has an additional advantage over the FE one in that it allows to control for specific time-invariant and/or rarely changing variables (such as several from the list of additional control variables above). Finally, as an alternative way to model the dynamics inevitably involved in TSCS analysis, each dependent variable has also been tested using a simplified version of the single-equation error correction model (ECM). This model uses a first-differenced dependent variable, while the righthand side of the regression includes a lagged dependent variable (plus lags of the first-differenced d.v., if necessary) and both lags and first-differences of each regressors. The ECM is traditionally considered necessary when a dependent variable shows signs of non-stationarity (as suggested by the observation of either a lagged dependent variable s coefficient really close to 1, random walk - like individual time-series, or the results of an augmented Dickey-Fuller test) and there s a risk of co-integration with one or more regressors. Recently however it has become a rather diffused practice also in the absence of integrated time series, given its ability to disentangle long and short term effects (De Boef and Keele, 2004). Further robustness checks have been performed by including in the analysis also advanced economies, by dating back the beginning of the time coverage of five years, by omitting the regime consolidation variable. Results The Appendix presents the full outputs of two dependent variables for education and two dependent variables for health, selected on the basis of the representativeness of the conclusions that can be drawn from the overall analysis. In the Appendix a synthesis of the results achieved for each dependent variable is also available. The table below, in turn, summarizes the results of the analysis, by dividing them by sector and topic. Table 1

16 Dependent Variable HP1 HP2 HP3 Education spending Partly accepted Rejected Rejected Education outcomes Rejected Accepted Partly accepted Health spending Partly accepted Partly accepted Rejected Health outcomes Rejected Accepted Partly accepted Starting from the first hypothesis, the empirical evidence suggests the existence of a pattern more complex than the one initially theorized. Expectations are confirmed only in the case of government spending and also in this case the analysis provided only weak support. Hybrid regimes invest significantly less money than democracies in both the education and health sector. When outcomes are concerned, however, it is not possible to observe a similar gap. More often than not hybrid regimes performance does not significantly differ from the performance of democracies. Most (few) exceptions to the latter conclusion do not hold to the robustness check performed, with the exclusion of pre-primary school enrollment. Although evidence is weak, in a few other cases hybrid regimes appear to do even better than democratic ones. Moving to the second hypothesis, the empirical analysis confirmed early expectations concerning the different socioeconomic performance between hybrid regimes and other forms of autocracy. Hybrid regimes outperform other non-democratic regimes in both education and health sectors. They are able to ensure primary and secondary education to a larger number of children in general, and they promote the enrollment of female children in particular, so that the gender gap between female and male pupils decreases. More precisely, strong evidence has been found of a significant difference between hybrid regimes and military regimes, single-party regimes in general, and oneparty regimes in particular. Relatively strong evidence also confirms that hybrid regimes do better than hegemonic-party regimes. The comparison between hybrid and hereditary regimes, on the contrary, provided less clear results. In many cases hypothesis 2 is confirmed, in several other it is not, in a few cases however hereditary regimes appear to do better than hybrid regimes. Hybrid regimes are also able to provide more adequate health care services to their citizens than other non-democratic regimes. This is reflected by the lower rates of child mortality, by the higher life expectancy, and by the larger number of people with access to safe drinking water. With the exception of the latter indicator, here the pattern is even more evident than in the sector of education. There s strong evidence that hybrid regimes do better than military, single-party regimes in general, one-party (especially when controlling also for communist rule) and hegemonic party regimes in particular. Once again, the analysis failed to clarify the comparison between hybrid and

17 hereditary regimes: the former outperform the latter only when life expectancy is concerned. Surprisingly, given the gap in child mortality rates, it is not possible to observe a similar difference in terms of neonatal immunization. As a matter of fact, some evidence suggests that both one-party and hereditary regimes do better than hybrid regimes. Concerning hypothesis 3, the analysis shed a pale light on the existence of a meaningful interaction between the hybrid nature of a political regime and its consolidation. In most cases the regression analysis produced interactions terms characterized by non-significant coefficients. The most relevant exceptions (for a total of 8 out of 25) are school enrollment (both primary and secondary, both total and female), and life expectancy. Yet an in-depth analysis of an apparently negligible piece of evidence may tells us a much more complex story. In the vast majority of cases (21 out of 25, including the 8 significant coefficients) we may note that the sign of interaction term is negative. A marginal effects analysis of being ruled by a hybrid regime as its duration increases demonstrated that this is not accidental. In 15 cases out of 21 we found evidence that the performance of a hybrid regime in the sectors of education and health tends to decrease as the process of regime consolidation advances. In most cases these results are significant up to mean (7) and medium/high (7) levels of consolidation. All the 15 cases refers to outcomes rather than investments. Conclusions The present analysis is one of the first attempts to investigate the consequences of hybrid regimes on citizens wellbeing from a theoretical and from an empirical point of view. By linking the now consolidated branch of literature on the consequences of democratization to the more recent debate on the functioning of hybrid regimes, a few hypotheses have been proposed and tested. The empirical analysis confirmed some expectations but also call for a refinement of the argument theorized. Based on the results presented above, the research may be summarized in the following conclusions. The citizens of a hybrid regime enjoy a higher quality of life than the citizens of any other nondemocratic regime. The only meaningful exception are hereditary regimes. The comparison between these two forms of authoritarianism failed to clarify if a difference exists and its direction. Even more interestingly, the citizens of a hybrid regime do as well as the citizens of a democratic country. Also when advanced economies have been included in the analysis, it is clear that the fact of being ruled by a hybrid rather than a democratic regime does not really make a difference when the sectors of education and health are examined. Another important conclusions concerns public spending. Hybrid regimes invest in education and health as many state revenues as other non-

18 democratic regimes, yet they achieve better results. In the same sectors hybrid regimes invest less money than democratic countries, yet in most cases they achieve very similar standards. Similar considerations, that deserve further examination, have obvious meaningful consequences for our understanding of hybrid regimes, and their future prospects. Furthermore they also contribute to debate on the consequences of democratization and, in particular, to the re-evaluation of the causal mechanism(s) that have been theorized so far.

19 Appendix Summary of the results Dependent Variable Dem Aut Mil Her Sin One Heg Education spending (% gdp) Education spending (% gov budget) ary school completion rate Progression to 2ary school Enrollment pre-1ary Enrollment 1ary Enrollment 2ary ary school completion rate (female) Progression to 2ary school (female) Enrollment 1ary (female) Enrollment 2ary (female) ary school completion rate (gender ratio) / Progression to 2ary school (gender ratio) Enrollment 1ary (gender ratio) Enrollment 2ary (gender ratio) Health spending (% gdp) Health spending (% gov spending) Life expectancy Life expectancy (female) +1/ Child mortality (<5) Child mortality (<1) Child mortality (neo natal) Immunization DPT Immunization measles Safe drinking water (% people with access) +2/ Legend: Positive numbers (ranging from 1 to 3) indicate the number of models (FE, PCSE, ECM) confirming that hybrid regimes have a significantly higher performance than the regime type of the corresponding column. Negative numbers indicate just the opposite. Dem = democracies ; Aut = all other non-democratic (non-hybrid regimes) ; Mil = military regimes ; Her = hereditary regimes ; Sin = single-party regimes ; One = pure one-party regimes ; Heg = hegemonic-like single-party regimes.

20 Education spending % GDP FE(1) FE(2) FE(3) PCSE ECM PCSE(Hp3) DV L1 omitted omitted omitted omitted omitted omitted Hybrid (L1) (0.37) (0.33) (0.15) (0.23) (0.51) (0.74) Hybrid (D) (0.68) Democracy (l1) (1.43) (1.23) (1.31) (0.51) (0.79) (1.00) Democracy (D) (0.68) Hereditary (L1) (5.20)*** (4.95)*** (0.19) (0.20) Hereditary (D) (4.11)*** Single-party (L1) (0.86) (0.44) Single-party (D) (1.69)* Other (L1) (0.61) (0.68) (0.82) (0.72) Other (D) (0.40) One-party (1.38) (0.28) Hegemonic-party (0.63) (0.90) Hybrid*Duration (0.50) Duration (L1) (0.11) (0.16) (0.01) (2.92)*** (2.66)*** (2.74)*** Duration (D) (0.68) Income pc (L1) (0.63) (0.58) (0.56) (0.60) (0.51) (0.73) Income pc (D) (1.18) Agriculture (L1) (3.39)*** (3.35)*** (3.33)*** (1.52) (0.97) (1.56) Agriculture (D) (0.08) Growth (L1) (2.32)** (2.32)** (2.31)** (1.99)** (0.84) (2.03)** Growth (D) (2.35)** _cons (1.76)* (1.76)* (1.73)* (1.29) (0.98) (1.36) R N * p<0.1; ** p<0.05; *** p<0.01; robust s.e. s in parentheses. The significance of the difference between the coefficients of the dummy Hybrid and the other regime dummies has been assessed by means of a Wald test. (L1) refers to the ECM and indicates that the variable is 1 year lagged. (D) refers to the ECM and indicates that the variable is first-differenced.

21 Marginal effect Marginal effect of HR vs Auth as they consolidate Duration

22 Health spending % GDP FE(1) FE(2) FE(3) PCSE ECM PCSE(Hp3) DV L1 omitted omitted omitted omitted omitted omitted Hybrid (L1) (2.05)** (2.09)** (2.09)** (1.01) (1.14) (0.75) Hybrid (D) (1.49) Democracy (l1) (1.50) (1.51) (1.51) (1.79)* (2.31)** (1.80)* Democracy (D) (1.28) Hereditary (L1) (1.14) (1.14) (0.04) (0.17) Hereditary (D) (0.20) Single-party (L1) (0.17) (1.07) Single-party (D) (0.95) Other (L1) (0.03) (0.03) (1.63) (1.35) Other (D) (2.22)** One-party (0.89) Hegemonic-party (0.17) (1.17) Hybrid*Duration (1.11) Duration (0.31) (0.45) (0.45) (1.76)* (1.85)* (1.66)* Duration (D) (0.50) Income pc (l1) (2.13)** (2.17)** (2.17)** (1.12) (1.63) (1.15) Income pc (D) (1.00) Growth (L1) (2.36)** (2.35)** (2.35)** (2.83)*** (0.06) (2.78)*** Growth (D) (1.86)* Dep. pop (L1) (1.20) (1.18) (1.18) (1.09) (1.32) (0.72) Dep. pop (D) (0.91) _cons (1.35) (1.34) (1.34) (1.91)* (2.07)** (2.10)** R N 1,556 1,556 1,556 1,556 1,552 1,556 * p<0.1; ** p<0.05; *** p<0.01; robust s.e. s in parentheses. The significance of the difference between the coefficients of the dummy Hybrid and the other regime dummies has been assessed by means of a Wald test. (L1) refers to the ECM and indicates that the variable is 1 year lagged. (D) refers to the ECM and indicates that the variable is first-differenced.