Introduction
Between the mid-1970s and late 1990s, a wave of democratization overwhelmed the developing world (Huntington, 1991). Rather than eradicating authoritarianism, the wave led to its transformation. In several cases, citizens’ aspirations for political freedom were soon frustrated. The collapse of an existing dictatorship prompted the introduction of nominally democratic institutions, yet failed to prevent old and new elites from practicing persistently authoritarian methods of governance. Following the end of the Cold War, electoral authoritarianism spread throughout Africa, part of Asia and the post-communist region and soon proved more resilient than initially expected (Brownlee, Reference Brownlee2009; cf. Knutsen and Nygard, Reference Knutsen and Nygard.2015).
Scholars’ commitment to unmask the actual nature of these pseudo-democratic regimes and of their rulers has overshadowed as much important dynamics that go beyond the strictly political sphere. Whether and how people living conditions have changed as a consequence of recent political transformations, for instance, are issues deserving consideration. The non-democratic universe displays a great deal of variation in terms of socioeconomic performance (Besley and Kudamatsu, Reference Besley and Kudamatsu.2007). Can the institutionalization of limited competition explain part of this variance? Outside the West, moreover, social welfare programmes have mostly originated under authoritarianism (Mares and Carnes, Reference Mares and Carnes.2009), even if subsequent democratization often proved key to their effectiveness (Carbone, Reference Carbone2009). But what if political change stops short of democratization? Are nominally democratic institutions just new dresses for old settings, or do they mark substantial changes in politics, power relationships, and leaders’ survival strategies, so as to call attention towards the needs of the populace? More generally, was transition from closed to electoral authoritarianism worth the trouble for ordinary citizens?
Contrary to the pessimistic scenarios that can be inferred from the literature on the socioeconomic consequences of democratization and the role of nominally democratic institutions in authoritarian settings, recent research suggests that bad things do not necessarily go together. Limited political competition can have mutual returns. Even when they fail to empower citizens, multi-party elections and legislatures trigger informational mechanisms that solicit rulers’ attention towards the needs of society and improve their ability to intervene in the public sector. Following a review of this literature, the paper formulates and tests new comparative hypotheses concerning government spending and living conditions under electoral authoritarianism. My research confirms that, especially in the medium/long-run, electoral autocracies outperform their full dictatorial counterparts, first and foremost military and single-party regimes. The analysis, however, also highlights a few important caveats, concerning performance differences with hereditary and democratic regimes, and asymmetries of findings in terms of investment and policy outcomes.
Nominally democratic institutions, incumbent advantages, and mutual returns
For decades authoritarianism had been considered as a better antidote to bring post-colonial new states out of underdevelopment (Huntington, Reference Huntington1968; O’Donnell, Reference O’Donnell1973; Evans, Reference Evans1989). Recently an opposite trend has emerged that sees democracy as compatible and even conducive to socioeconomic progress. Yet, while empirical research confirms that full democratization has positive implications for citizens’ material well-being (Brown and Hunter, Reference Brown and Hunter.1999; Zweifel and Navia, Reference Zweifel and Navia.2000; Gerring et al., Reference Gerring, Thacker and Alfaro.2012), it is not clear whether we should expect similar results as a consequence of pseudo-democratization – that is, in the case of electoral authoritarianism.
During the third wave of democratization, especially following the end of the Cold War, electoral authoritarianism proliferated throughout the developing world. From a procedural point of view, these regimes differ from full or closed forms of dictatorial rule. Electoral autocracies are broadly inclusive, in that regular elections for the executive and legislative offices are held under universal suffrage (Schedler, Reference Schedler2006). They are also minimally competitive, as ‘opposition is allowed, multiple parties are legal, and more than one candidate is allowed on the ballot’ (Hyde and Marinov, Reference Hyde and Marinov.2011: 195). While meaningful, elections do not meet minimum democratic standards of freedom and fairness. A systematic disjuncture between formal rules and actual behaviours makes the playing field ‘heavily skewed in favour of incumbents’, and competition ‘real but unfair’ (Levitsky and Way, Reference Levitsky and Lucan.2010: 5). Even when opposition parties are able to contest vigorously for power, their only realistic ambition is a share of parliamentary seats and some bargaining power. These regimes thus qualify as a distinct but plainly non-democratic regime type (Diamond, Reference Diamond2002).Footnote 1
Based on existing theories of the socioeconomic consequences of democratization, the electoral authoritarian institutional apparatus discourages easy generalizations. When explaining democratic governments’ higher responsiveness to the needs of society, scholars typically focus on two key explanatory factors. The first is electoral competition, an incentive for political leaders to meet middle and lower class voters’ demands for redistribution (Meltzer and Richards, Reference Meltzer and Richard.1981). Alternatively, they stress vertical accountability, a constraint to incumbents’ propensity to engage in cleptocratic behaviours (Lake and Baum, Reference Lake and Baum.2001). Theory thus points to causal mechanisms that can hardly be expected to work under electoral authoritarianism, where nominally democratic institutions fail to give citizens more decisional power. Quite the contrary, in dictators’ hands these institutions can become outright instruments of authoritarian consolidation against the threats coming from both the international and the domestic arena. Concerning the latter, autocrats have learned how to use democratic procedures to influence the behaviour of the other members of the ruling elite, opposition forces, and the masses, as an alternative and complementary strategy to repression. Authoritarian elections and legislatures regularize intra-elite interactions, deterring the risk of defections (Brownlee, Reference Brownlee2007; Magaloni, Reference Magaloni2008; Boix and Svolik, Reference Boix and Svolik.2013). By coopting selectively opposition parties into legislatures through minimally competitive elections, autocrats raise the cost of being an outsider and fragment anti-government coalitions (Magaloni, Reference Magaloni2006; Gandhi, Reference Gandhi2008). Most importantly, these institutions can be used to manipulate citizens’ choices. Authoritarian elections often translate in mechanisms of patronage distribution (Lust-Okar, Reference Lust-Okar2006; Magaloni, Reference Magaloni2006). To buy off the loyalty of the relatively few key actors that keep them in power, incumbents steal resources from the public sector and prioritize the production of private goods.
Electoral autocracies are thus often portrayed as environments ‘quite beneficial for leaders’ that help them ‘keep their jobs and their access to private benefits, even when they perform poorly in the policy arena’ (Bueno de Mesquita et al., Reference Bueno de Mesquita, Morrow, Siverson and Smith.1999: 154). Yet, similar conclusions underestimate the transformative power of nominally democratic procedures (Schedler, Reference Schedler2013). Besides working as constraints on elite members, oppositions and citizens, these institutions also shape rulers’ opportunities and choices. Recently, for instance, it has been posited that political change short of democratization can have mutual returns. Nominally democratic institutions engender new incentives and capabilities for rulers to improve citizens’ living conditions.
Considering the relationship between rulers and ruled, the idea that leaders in electoral autocracies govern by repression and manipulation alone is overly simplistic. As Svolik – and Machiavelli centuries before – puts it, ‘the lack of popular consent (…) is the original sin of dictatorships’ (Reference Svolik2012: 10). Support and cooperation from society is especially important to civilian leaders, who cannot rely on a royal court or military junta (Olson, Reference Olson1993; Gandhi and Przeworski, Reference Gandhi and Przeworski.2006, Reference Gandhi and Przeworski.2007). As a strategy to expand the social basis of the regime, sedate social unrest, improve human capital and the productivity of the country, investments in citizen well-being should thus appeal civilian autocrats, including ‘elected’ ones. But the point still stands. Apart from incentives, can electoral authoritarian institutions enhance rulers’ ability to deliver public goods – for example, health and education services? In other words, can the provision of well-being become a viable plan of action for an autocrat seeking support and cooperation from society?
Under electoral authoritarianism, it could. Two major obstacles that autocrats typically face in policy-making are poor knowledge of society, a consequence of the systematic use of repression (Wintrobe, Reference Wintrobe1998), and low state capacity, resulting from an underdeveloped institutional apparatus. This is why dictators often find it more convenient to defend the privileges of the few, rather that promoting the interests of the majority (Bueno de Mesquita et al., Reference Bueno de Mesquita, Morrow, Siverson and Smith.1999). In this respect, electoral authoritarian institutions ease the task of governing (Charron and Lapuente, Reference Charron and Lapuente.2011). Elections and legislatures, in particular, can compensate for the structural information deficit traditionally vexing non-democratic rule. Miller (Reference Miller2015a) explains that authoritarian multi-party elections represent an effective instrument to gather information about citizens’ preferences and needs. While they can hardly be defined decisive voters, under electoral authoritarianism citizens are free to express their preference for any party that run elections according to the rules of the game; and voice at a relatively low cost their dissatisfaction with the current state of affairs. Opposition parties, in turn, represent a valid alternative for voters wishing to signal dissatisfaction. Despite several restrictions, under electoral authoritarianism these parties are allowed to operate above ground, develop alternative political platforms, compete with each other and run with the realistic aspiration to gain seats and bargaining power. According to Gandhi (Reference Gandhi2008), moreover, opening parliamentary assemblies to opposition groups favours communication with the different sectors of society.
A comparative perspective: state of the art evidence and new hypotheses
As civilian rulers, electoral autocrats seek support and cooperation from society. Improving people living conditions can help political leaders achieve these goals, although the information and communication deficit typically affecting authoritarianism may hamper the feasibility (and thus the convenience) of a similar strategy. Semi-competitive elections and multi-party legislatures compensate for the lack of knowledge of society’s needs and preferences. Autocrats learn from elections how satisfied people are and, based on the relative strength of various opposition parties, what are the most urgent issues to be addressed. Government–opposition interactions within multi-party legislatures, in turn, favour policy compromise, whereby public goods are exchanged with compliance with the status quo. Electoral authoritarian leaders, in other words, do not suffer – not with the same intensity – from the information deficit vexing their full authoritarian counterparts. Nominally democratic institutions can make the promotion of citizen well-being an attractive strategy, an effective lever to sedate social unrest and solicit cooperation from society. Under electoral authoritarianism, therefore, rulers have both the incentives and the capabilities to improve citizens’ living conditions. To date, however, empirical research has not investigated the issue in a fully satisfactory way.
Case studies, especially from African countries such as Ghana (Carbone, Reference Carbone2011) and Tanzania (Kjaer and Therkildsen, Reference Kjaer and Therkildsen.2012), demonstrate that electoral outcomes and dialogue with opposition forces fine-tune policy-making, even when competition is flawed. Working on a sample of authoritarian multi-party elections, Miller confirms that ‘falling vote totals for ruling parties predict policy concessions on government spending’ (Reference Miller2015a: 715) in the years immediately following an election. From a comparative perspective, however, we confront with mixed and incomplete empirical findings. Relative to closed dictatorship, are electoral authoritarian governments better able to improve citizens’ living conditions? Focussing on policy outputs, Gandhi (Reference Gandhi2008) finds a weak negative impact of nominally democratic institutions on military spending, but no significant positive effect on social expenditure. Looking at policy outcomes, Miller (Reference Miler2015b) finds that electoral autocracies achieve better results than closed autocracies. Taken together, these results suggest that, for similar levels of social investment, citizens’ well-being is higher under electoral authoritarianism, thus highlighting a capability (more than a commitment) gap. While intriguing, these conclusions need further assessment, to dispel uncertainties associated with the different research designs from which they are extrapolated. Hence, the first hypothesis that will be (re-)tested is twofold, and compares citizen well-being under electoral and full autocracies looking at both policy outputs – that is, how much governments invest in the social policy sectors – and outcomes – that is, the material results they achieve in terms of living standards:
Hypothesis 1 Citizens enjoy significantly better living conditions under electoral authoritarianism than under closed dictatorship.
Several equally urgent questions are still waiting for examination, though. One refers to the actual strength of the causal mechanisms triggered by pseudo-democratization vis-à-vis the consequences of full democratization. Do democracies deliver more public goods than electoral autocracies? Despite its peculiarities – nominally democratic institutions that are not just window dressing, but meaningful instruments of government – electoral authoritarianism is a non-democracy, ruled by as much non-democratic leaders. Like in other forms of dictatorships, the distribution of power remains profoundly skewed in favour of the ruling elite. The most evident sign of this asymmetry is the fact that, with few exceptions, electoral autocracies are systems in which ‘opposition parties lose elections’ (Schedler, Reference Schedler2002: 47; cf. Przeworski, Reference Przeworski1991). Most importantly, the discussion has highlighted something that is more of an enabling than a constraining effect. Electoral authoritarian institutions trigger informational mechanisms that expand a ruler’s menu of available strategies to pursue his/her own interests. These mechanisms can have mutual returns and benefit citizens, too. Yet, the incentives faced by electoral authoritarian governments are different and arguably less compelling, compared with the mechanisms of accountability and responsiveness working under democracy (Meltzer and Richards, Reference Meltzer and Richard.1981; Lake and Baum, Reference Lake and Baum.2001). Autocrats under semi-competitive regimes, in which executive turnover represents a remote possibility, are not forced to meet citizens’ demands to hold office. Keeping a twofold focus on outputs (i.e. investment) and outcomes (i.e. impact), the analysis will also test the following hypothesis:
Hypothesis 2 Citizens enjoy significantly better living conditions under democracy than under electoral authoritarianism.
To set more definite borders to the conclusions that will be drawn from the analysis of the first hypothesis, further issues should be examined deriving from the within-regime heterogeneity that characterizes the full authoritarian category. Electoral autocracies should be compared with more specific subtypes of dictatorship. In this respect, a major divide seems to exist between single-party and military regimes on one side, and hereditary regimes on the other. For different reasons, single-party and military regimes suffer from the problems typically associated with closed authoritarianism. Military rule, in which a junta controls the government in either a direct or indirect way, corresponds almost by definition to the denial of political pluralism (Brooker, Reference Brooker2000: 44–52). Military regimes lack institutionalized channels to communicate with society and gather information about citizens’ preferences. As non-civilian governments typically conceiving themselves as temporary solutions, moreover, military juntas do not face strong incentives to seek popular support. Rather than policy concessions, these regimes are more likely to use the heavy hand to pursue their short-term goals. One-party systems have a structure similar to electoral autocracies. Sometimes they hold elections, at least for the legislative, and often arrange a network of organizations to infiltrate civil society. Even when multipartitism is tolerated, however, the political arena remains monopolized by a single party (Brooker, Reference Brooker2000: 37–44), any form of political competition being de facto nullified. As long as they either accept a status of satellites of the ruling party or a condition of marginalized outsiders, opposition parties cannot represent a valid alternative for citizens, by means of which to signal dissatisfaction in elections and promote their interests in parliaments.
Because they are ruled by non-civilian elites, in principle hereditary autocracies should be less worried about soliciting cooperation from society and less incentivized to improve citizens’ living conditions. Dynastic succession, however, ‘can be socially desirable (…) because it may give monarchs more concern for the long run and the productivity of their societies’ (Olson, Reference Olson1993: 572). Monarchs may thus have an interest in taking care of the human capital of their countries, by investing in social sensitive sectors such as education and health. During the past decades, moreover, hereditary regimes have experienced deep transformations and currently present several institutional affinities with electoral autocracies. Some of them have maintained the absolutist-like asset. Yet, many others have introduced minimally competitive elections and/or limited forms of power-sharing at the legislative level, although the power of the king remains unchallenged (Herb, Reference Herb2004).
In conclusion, neither military nor single-party regimes appear to have the incentives and the institutional capacity to deliver public goods efficiently, whereas liberalized contemporary monarchies do. Although we cannot extend the argument to all hereditary regimes, it is unlikely that the analysis will produce univocal results in the latter case. With reference to the first (twofold) hypothesis, therefore, more precise expectations should be formulated:
Hypothesis 3 Citizens enjoy significantly better living conditions under electoral authoritarianism than under military rule.
Hypothesis 4 Citizens enjoy significantly better living conditions under electoral authoritarianism than under single-party rule.
Hypothesis 5 Electoral autocracies and hereditary regimes do not display systematic differences in terms of citizens’ living conditions.
Empirical analysis
Time-series cross-section (TSCS) analysis is used to test the hypotheses. The sample covers 141 developing countries observed from 1980 to 2010.Footnote 2 The section proceeds by illustrating the independent, dependent, and control variables, and how the analysis is performed. Findings are discussed in the next section.
Regime types
The main independent variable is categorical and classifies countries according to regime type: democracy (DEM), electoral authoritarianism (EA), full authoritarianism (FA) – including one-party (ONE), military (MIL), and hereditary (MON) regimes – plus a residual category. The three main categories have been frequently thought of as lying along a continuum of political freedom or democraticness. To preserve concept-measurement consistency, and avoid problems associated with existing indices of democracy (Munck and Verkuilen, Reference Munck and Verkuilen.2002) and the choice of arbitrary thresholds (Bogaards, Reference Bogaards2010), the present analysis relies on a different measurement strategy.
Sticking to a procedural and dichotomous definition of political competition, I identify regimes – either democratic or electoral authoritarian – that fulfil the institutional requirements set by Cheibub et al. (Reference Cheibub, Gandhi and Vreeland.2010; cf. Przeworski et al., Reference Przeworski, Michael, Cheibub and Limongi.2000), with the exception of the alternation rule: multi-party elections for both the executive and legislative office, a multi-party legislature in which opposition is represented.Footnote 3 Within this group of minimally competitive regimes, democracy is distinguished from electoral authoritarianism according to the freedom and fairness of the electoral process (Diamond, Reference Diamond2002: 28). Following a recently suggested practice (Bogaards, Reference Bogaards2012), I use Polity IV disaggregated data. Contrary to the composite index, Polity IV concept variables have an important advantage: values represent qualitatively distinct categories and not just differences in degrees. Accordingly, I code as democratic – that is, regimes in which electoral competition is free and fair – only observations scoring 8 in the Polity IV ‘Executive Recruitment’ indicator, meaning that ‘the electoral process is transparent and its outcomes are institutionally uncertain’ (Marshall et al., Reference Marshall, Jaggers and Gurr.2013: 60), that is: outcomes are neither significantly influenced by the incumbent or non-elected officials, nor the result of predetermined agreements; major opposition parties participate vigorously in the electoral process; elections take place in an environment free from systematic repression (Marshall et al., Reference Marshall, Jaggers and Gurr.2013).Footnote 4 By the same token, the electoral authoritarian category, encompasses all minimally competitive regimes without free and fair elections.
The three subtypes of full authoritarianism, to which Hypotheses 3–5 refer, are identified following Wahman et al. (Reference Wahman, Teorell and Hadenius.2013). To be sure, the authors acknowledge several mixed subtypes, which my analysis does not consider. Specifically, any time the executive is either directly or indirectly ruled by a junta, the regime is classified as military. Likewise, any regime in which the position of effective chief executive is inherited in accordance with an accepted practice is a monarchy. The single-party category refers either to pure forms of one-party rule, or systems in which minor alternative parties exist but cannot compete in antagonistic terms. Observations that cannot be classified according to the above rules enter a residual category encompassing transitional phases, cases of non-independent authority and/or failed states, as signalled by Polity scores −66, −77, −88. A complete list of regime spells can be found in the Appendix.
Citizen well-being
The dependent variable is citizen well-being. Following a consolidated practice, the analysis focusses on the non-strictly monetary dimension of the notion. In this respect, a major source of concern is the adaptive nature of individual preferences. If citizens ‘tend to adjust their aspirations to their possibilities’ (Elster, Reference Elster1982: 219), survey data on perceived well-being are not valid when comparing regimes characterized by different levels of freedom. Accordingly, objective measures of policy outputs and outcomes have been preferred to subjective indicators.Footnote 5 The analysis focusses on the sectors of education and health, in particular, using data from the World Bank’s World Development Indicators. Especially for developing countries, improvements in these sectors are widely viewed as reliable indicators of governments’ commitment to improve citizens’ material well-being and of their ability to do it. Data on spending refer to public health expenditure, measured either as a percentage of gross domestic product (GDP) or as a share of the public budget. Observations start from 1995. Data on education expenditure are not available on a sufficient number of countries and years to allow sound large-n comparison. To measure policy outcomes, infant mortality rate and secondary school enrolment ratio have been selected. Infant mortality is one of the most frequently used indicators in this field of study. Secondary school enrolment represents a better source of variance than primary school enrolment.
Control variables
The base specification of the regression model includes regime type, broken in categorical dummies, and control variables referring to the economic, demographic, and political factors that may influence the relationship under scrutiny. Unless differently specified, these data are gathered from the World Bank. By controlling for these factors, I address one major source of concern: endogeneity deriving from the omission of one or more variables affecting both the dependent and the main independent variables, thus leading to biased estimates and incorrect inferences (e.g. spuriousness).
Among the factors that may directly affect citizens’ living conditions but, at the same time, can also predict an autocrat’s decision to adopt nominally democratic institutions, economic ones are of primary importance. Economic growth has redistributive effects that may directly affect living standards, especially in the long run. Recent research, however, has demonstrated that, to attract private investments and foster growth, autocrats adopt nominally democratic procedures because they reduce uncertainty (Wright, Reference Wright2008; Kenyon and Naoi, Reference Kenyon and Naoi.2010; Jensen et al., Reference Jensen, Malesky and Weymouth.2014). Growth rates, moreover, account for another important source of variance across both regimes and regions, given the remarkable economic performance achieved in the past by some military regimes, such as Chile, and by several non-democratic Asian countries. On the contrary, dictators that can count on natural resources as a direct fount of revenue, alternative to taxation, are less interested in investing in institutions favouring communication with citizens and in promoting well-being to solicit cooperation from society (Gandhi and Przeworski, Reference Gandhi and Przeworski.2006, Reference Gandhi and Przeworski.2007; see also Ross, Reference Ross2012). As many contemporary monarchies are oil- and gas-rich countries, moreover, controlling for this factor should eliminate some noise from the comparison between electoral authoritarian and hereditary regimes. Foreign development assistance, in turn, has a direct effect on the provision of schooling and health services but can also influence the likelihood that a non-democratic regime introduces nominally democratic institutions, given conditionality. Among non-economic factors, I control for regime duration. Measured as the age of a regime since the last transition, duration is expected to ease the provision of public services, as all regimes are subject to the ‘liability of newness’. Yet, regime duration is also a proxy of the time horizon faced by the ruler (Keefer and Khemani, Reference Keefer and Khemani.2005) and may thus influence incentives to invest in institutions that improve governance but can be costly in the short run (Olson, Reference Olson1993; Wright, Reference Wright2008).
The regression analysis also accounts for the impact of a few other potential explanatory factors of the dependent variable. A control variable for economic prosperity – measured as per capita GDP – is included because wealthier societies tend to be more attentive to the quality of public services, and their governments better equipped to provide them (Ghobarah et al., Reference Ghobarah, Huth and Russett.2004; Mulligan et al., Reference Mulligan, Gil and Sala-i-Martin.2004; cf. Przeworski et al., Reference Przeworski, Michael, Cheibub and Limongi.2000 on the relationship between regime type and wealth). Demographic and political factors are likely to matter, too. Although most of our dependent variables are measured as rates, the raw number of recipients influences governments’ ability to provide education and health services. The share of urban population is considered, as it should be relatively easier for governments to provide basic services in circumscribed and highly populated urban areas, than in sparse and small rural communities. At the same time, however, fast-paced urbanization often leads to the proliferation of slums, where people are exposed to several health risks, school infrastructures are scarce, and citizens are politically marginalized. Ethnic fractionalization, in turn, may result in the political exclusion of minorities and an unequal access to public services. The variable is time-invariant and is borrowed from Alesina et al. (Reference Alesina, Devleeschauwer, Easterly, Kurlat and Wacziarg.2003). Finally, an indicator of the presence of a communist-led executive (Cheibub et al., Reference Cheibub, Gandhi and Vreeland.2010) is included in the regression model, given the ideological commitment of these elites to the redistribution of national wealth. The variable is expected to refine the comparison with single-party regimes, in particular.
Model
TSCS analysis uses ordinary least squares estimation on a simplified version of the error correction model (ECM). ECM regresses the change in the dependent variable (ΔY) on its lagged levels and/or changes of all other regressors. The choice of ECM has been dictated both by practical and theoretical reasons. First, ΔY tackles the (near)non-stationarity that characterizes most of our dependent variables.Footnote 6 Second, regressors in lagged level form break the symmetry of the regression model, which reduces the risk of reverse causality. Specifically, in this paper lags and changes are measured based on a 3-year interval. Third, ECM disentangles short- and long-run effects, allowing for a more sensitive treatment of temporal dynamics (De Boef and Keele, Reference De Boef and Keele.2008; Beck and Katz, Reference Beck and Katz.2011). In this regard, while most control variables are likely to influence the selected dependent variables mainly in the medium/long term, the same does not apply to our key explanatory factor. The regime effect might be either exerted immediately after a transition, disclose over time, or represent a sum of both.Footnote 7
To ease the interpretation of the main results, Hypotheses 1 and 2 are tested separately, switching FA and DEM as reference regime categories omitted from the model. For each dependent variable, therefore, the regression coefficient of the regime dummy EA estimates the difference in the performance of electoral autocracies vis-à-vis full dictatorships (Hypothesis 1) and democracies (Hypothesis 2), respectively. In the analysis of Hypotheses 3–5, the reference category is MIL (Hypothesis 3). The regression model includes dummies EA, ONE, and MON. Wald tests is used to assess the significance of the performance differences across the regime categories included in the model – EA vs. ONE (Hypothesis 4) and EA vs. MON (Hypothesis 5), respectively.
Diagnostic analyses show that the lagged dependent variable eliminates most of the existing serial correlation. Heteroskedasticity-robust standard errors clustered by panel have been preferred to Beck and Katz’s (Reference Beck and Katz.1995) panel-corrected standard errors because N is much larger than T. Observations classified in the residual category are dropped from the analysis. Theocratic Iran, Libya under Gaddafi, and apartheid South Africa – authoritarian regimes that hardly fit into one specific subtype – are classified as full autocracies when this represents the reference category (Hypothesis 1), and as (dropped) residuals when the analysis focusses on authoritarian subtypes (Hypotheses 3–5).
Robustness checks
To test the robustness of findings, a few alternative models are run. A first source of concern refers to the hybrid nature of electoral semi-competitive autocracies, and the lack of consensus on their measurement (Cassani, Reference Cassani2014). Democratization studies, moreover, have often proved to be particularly sensitive to researchers’ decisions concerning the operationalization of regime types. The analysis is thus replicated measuring regimes in different ways. Following Brownlee (Reference Brownlee2009), I use Freedom House’s ‘electoral democracy’ indicator (updated for the 1980–88 period) to distinguish DEM from EA, and the Index of Electoral Competitiveness from the Database of Political Institutions (Beck et al., Reference Beck, Clarke, Groff, Keefer and Walsh.2001) to separate EA from FA (threshold at score 6). Alternatively, following Diamond (Reference Diamond2002) and Lindberg (Reference Lindberg2009) as reported in Bogaards (Reference Bogaards2010), I use Freedom House’s Index of Freedom, setting the threshold at score 3.5 (DEM vs. EA). To further show that the conclusions that will be drawn are not driven by arbitrary choices, I also test other indicators of material well-being as dependent variables – namely life expectancy and primary school enrolment ratio – and calculate lags and differences of dependent and independent variables based on a shorter interval (1 year).
Other issues have to do with the cross-sectional and time-series structure of the sample and its consequences in terms of endogeneity. When large number of countries are pooled together, we risk to miss important sources of heterogeneity, referring to unobserved unit-specific characteristics. The analysis, moreover, extends through all the 1980s, a decade particularly unfruitful for the socioeconomic emancipation of developing countries, many of which were full dictatorships. After the end of the Cold War both these trends reversed. A statistical correlation between electoral authoritarianism and better living standards, it could be objected, might simply mirror this simultaneous change. Accordingly, the analysis is replicated on a country-fixed effects model (FE) that includes year dummies.Footnote 8 Country-fixed effects allow to work only with longitudinal variation, which drastically reduces the number of available degrees of freedom, given the relatively small T of the sample. Hence, control variables are dropped from the FE model.
Data availability on the dependent variable is another potential source of endogeneity, as missingness can be correlated with both regime type and levels of development. To show that findings are not driven by some trend hidden by missing data, the analysis is replicated on a data set whose empty cells are filled in using multiple imputation.Footnote 9 Using country-years as units in the study of the consequences of political regimes on health and education services is another point that may raise some criticism. Accordingly, the data set is collapsed in regime-spells and cross-sectional analysis is run using only regimes lasted 5 years or more. Finally, because the main model includes a fairly extensive set of covariates, I also replicate the analysis using a thinner specification, dropping those controls that may downward bias regime type coefficients – namely regime duration, growth, development assistance, and natural resource rents.
Citizen well-being under electoral authoritarianism: evidence and discussion
Table 1 summarizes the findings of the TSCS analysis, along with more detailed information concerning the strength and dynamic nature of the regime effect. Tables 2–4 present the regression outputs of the main analysis. Regression outputs of the robustness checks performed are reported in the Appendix. Overall, most hypotheses are accepted, with a few notable exceptions. The analysis confirms that electoral authoritarian regimes outperform their full authoritarian counterparts, and refines this conclusion shedding light on several previously overlooked details, concerning different subtypes of closed dictatorship. It also emerges that differences across regimes can be better seized looking at policy outcomes, rather than investment. More unexpectedly, democracies do not display a significantly superior performance. Another important conclusion refers to the dynamic nature of the relationship under examination. Even if in several cases a transition effect has been detected, the impact associated with the institutionalization of limited competition tends to disclose slowly and its actual magnitude can be observed only in the medium/long run.
Table 1 Summary of findings
For each hypothesis, the table reports detailed information concerning the results of the time-series cross-section analysis, the dynamic nature – either short run, long run, or both – of the estimated regime effects, and the robustness of these findings to the counter-tests performed. Specifically, ‘Alter X’ tests hypotheses using alternative measures of the independent variable (Freedom House and Database of Political Institutions); ‘Alter Y’ uses alternative dependent variables (life expectancy and primary school enrolment); ‘Alter L./D.’ uses a 1-year interval to compute lags and changes (instead of 3 year); ‘Pars’ uses a thinner model specification; ‘FE’ is a country- and year-fixed effect model without control variables; ‘CS’ is a cross-sectional analysis focussing on regimes lasted 5 years or more; ‘MI’ replicates the analysis using multiple imputation. Shaded cells indicate that the hypothesis is rejected. Asterisks confirm the robustness. Concerning Hypothesis 5, asterisks confirm the absence of significant differences.
GDP=gross domestic product; EA=electoral authoritarianism; FA=full authoritarianism; DEM=democracy; MIL=military regimes; ONE=one-party regimes; MON=monarchies.
Table 2 Electoral authoritarianism vs. full authoritarianism
Robust standard errors clustered by panel in parentheses. Estimates are from an ordinary least squares regression, using an error correction model in a time-series cross-sectional analysis. In both cases, the interval is 3 years. Regression outputs refer to Hypothesis 1.
GDP=gross domestic product; L=lagged; D=differenced; pc=per capita; ODA=foreign development assistance.
*P<0.1; **P<0.05; ***P<0.01.
Table 3 Electoral authoritarianism vs. democracy
Robust standard errors clustered by panel in parentheses. Estimates are from an ordinary least squares regression, using an error correction model in a time-series cross-sectional analysis. In both cases, the interval is 3 years. Regression outputs refer to Hypothesis 2.
GDP=gross domestic product; L=lagged; D=differenced; pc=per capita; ODA=foreign development assistance.
*P<0.1; **P<0.05; ***P<0.01.
Table 4 Electoral authoritarianism vs. full authoritarian subtypes
Robust standard errors clustered by panel in parentheses. Estimates are from an ordinary least squares regression, using an error correction model in a time-series cross-sectional analysis. In both cases, the interval is 3 years. Significant differences between coefficients associated with the regime dummies included in the specification model are tested using Wald test. Regression outputs refer to Hypotheses 3–5.
GDP=gross domestic product; L=lagged; D=differenced; pc=per capita; ODA=foreign development assistance.
*P<0.1; **P<0.05; ***P<0.01.
Differences in spending decisions across regime types are mainly evident when attention is focussed on how rulers allocate the public budget. Levels of investment in terms of GDP do not differ significantly, with few exceptions. Electoral autocrats invest in the health sector more than other dictators, but less than democratically elected governments. After 5 years of rule, the estimated cumulative spending gaps are +1.5 and −1.1%, respectively. Relative to spending, however, Hypothesis 1 can be only partially accepted. A thorough comparison between electoral and subtypes of full autocracy reveals that these results are mainly driven by military juntas. That is to say that, in the analysis of Hypotheses 3–5, only the comparison with military regime produces statistically significant results. Finally, the counter-analyses performed – notably the cross-sectional analysis and the use of a shorter interval – cast several doubts on the robustness of these findings.
Spending decisions tell only part of the story, and not necessarily the most salient one. The analysis of health and education outcomes produces clearer and much more robust evidence, even if not always in the anticipated direction. Electoral authoritarian governments provide more efficient and inclusive schooling systems. In 5 years, following a transition from full to electoral authoritarianism, secondary school enrolment ratios are estimated to be 5.2% higher, and the predicted gap amounts to >8.5% after 10 years. Likewise, citizens under electoral authoritarianism enjoy better healthcare services, as reflected by lower infant mortality rates. The comparison with military and single-party subtypes confirms that for ordinary citizens it is better to live under electoral authoritarianism, when the provision of education and health services is concerned. The same cannot be said in the case of monarchies. As expected, there is no systematic difference in the performance of electoral and hereditary governments, signalling similar capabilities to provide citizens with adequate living conditions. In an admittedly more surprising way, the analysis failed to detect systematic differences between electoral authoritarian and democratic regimes. Relative to policy outcomes, Hypothesis 2 cannot be accepted.
Concerning control variables, regime duration, national wealth, economic growth, development assistance, and higher shares of urban population are positively associated with better policy outcomes. The impact of ethnic fractionalization is invariably negative. Rents from natural resources reduce rulers’ propensity to invest in the health sector, although rentier states display comparatively lower infant mortality rates on average. Governments in more populated countries do not necessarily prioritize health in their budgetary decisions, but demographic size does not necessarily hinder the provision of far-reaching health and education services. The robustness of the presented findings has already been discussed. It is worth stressing that no single replication analysis produced results that systematically contradict the main conclusions. It should also be noted that, according to some analyses using infant mortality rates as dependent variables, electoral autocracies outperform democracies, and are outperformed by hereditary regimes. In both cases, however, evidence is mixed and rather inconclusive.
Electoral autocracies outperform full authoritarian regimes. Differences are magnified when the comparison focusses on military and single-party regimes, whereas no systematic gap exists between electoral and hereditary regimes. These are the main conclusions that can be drawn from the analysis, confirming, strengthening, and refining Miller’s (Reference Miler2015b) recent findings. Contrary to anticipations, however, it is not clear if it makes a difference in terms of material well-being to live under democracy rather than under electoral authoritarianism. Another unexpected result is the asymmetry emerging from the comparison of findings relative to spending and actual achievements. Provision of better education and health services does not necessarily correspond to higher investment, and vice versa. While it is hazardous to go too far commenting statistical non-significance, these results suggest a few considerations.
First, nominally democratic institutions do not act as constraints. Among non-democratic regimes, fairly robust difference in spending has been recorded only comparing electoral autocracies and military regimes, a typically short-lived and increasingly rare form of dictatorial rule. Nominally democratic institutions, in other words, can hardly compensate alone for the distortive effect associated with authoritarianism, and re-balance power relationships between rulers and ruled. Electoral authoritarian leaders are not – not necessarily, at least – weaker dictators, forced to invest in citizen well-being to hold office. A clarification that might lead to a partial revision, or delimitation, of Miller’s (Reference Miller2015a) conclusions on electoral responsiveness in autocratic regimes. Yet, nominally democratic institutions enable. The same amount of resources can be used in more or less efficient ways. In this regard, multi-party elections and opposition inclusion in the political arena appear sufficient to improve the ability of rulers to interpret and address citizens’ preferences and needs. Reconciling the findings of Gandhi (Reference Gandhi2008) on spending and of Miller (Reference Miler2015b) on outcomes, therefore, the present analysis portrays electoral autocrats as well equipped rulers, rather than tied-handed dictators.
Second, it seems that an ‘electoral authoritarianism advantage’, rather than a ‘democracy advantage’ (Halperin et al., Reference Halperin, Siegle and Weinstein.2005) exists, at least from rulers’ viewpoint. The analysis of budgetary decisions seems to confirm that democratically elected governments are more committed to improve citizens’ living conditions. Consistently with the literature (Meltzer and Richard, Reference Meltzer and Richard.1981; Bueno de Mesquita et al., Reference Bueno de Mesquita, Morrow, Siverson and Smith.1999; Lake and Baum, Reference Lake and Baum.2001), this is likely to be a consequence of the higher uncertainty that incumbents face. Yet, democrats do not deliver significantly more public goods than regimes in which political change stopped short of democratization. Put another way, electoral autocrats invest less than democratic governments, but obtain about the same results in terms of policy outcomes. This is something that could derive from the inexperience of young democracies with the management of a fully liberalized political arena. If this were the case, supposedly outdated arguments seeing authoritarianism as a better therapy to underdevelopment (Huntington, Reference Huntington1968; O’Donnell, Reference O’Donnell1973; Evans, Reference Evans1989) would re-gain momentum. In less provoking terms, these findings call for a reappraisal of the theory of the socioeconomic consequences of democratization. How much do the incentives deriving from the actual risk of being voted out of office – something that only democratically elected leaders face – matter vis-à-vis the informational mechanisms triggered by elections and legislatures even in the absence of full democratization? The effect associated with the introduction of democratic procedures, arguably an early step of democratization, seems to be larger in magnitude than the impact that could be ascribed to the actual functioning of these procedures – including the institutionalization of electoral uncertainty – which is something typically achieved at more advanced stages of democratization.
Conclusions
In the wake of the third wave of democratization, electoral authoritarianism spread throughout the developing world. Old and new elites have learned how to use formally democratic procedures to strengthen their grip on power. But, besides autocrats’ advantage, does political change short of democratization matter for the lives of ordinary citizens? Drawing on recent research revealing the incentives and the capabilities of electoral authoritarian governments to promote citizen well-being, this paper has deepened the analysis of this question.
The answer is yes. Nominally democratic institutions shape the political arena in ways that advance also citizens’ material interests. Especially in the medium/long-run, electoral autocracies are found to outperform closed dictatorships, notably single-party and military regimes, in the sectors of education and health. While political freedom remains a chimera, recent transitions from full to electoral authoritarianism have brought substantial pay-offs for citizens of developing countries. The analysis, however, also produced a few unexpected results concerning the comparison between electoral authoritarian and democratic regimes and the differences in terms of spending across regime types. A reminder that, as an object of research, the socioeconomic consequences of political change short of democratization are far from being exhausted. For instance, it is yet to be evaluated whether promoting citizen well-being fosters authoritarian consolidation or propels future democratization, as modernization theory suggests.
Acknowledgements
The author is grateful to Giovanni Carbone for advice, Matthijs Bogaards, Luigi Curini, Vincenzo Memoli, Francesca Pasquali, Alessandro Pellegata for helpful comments, and V. for all the rest.
Financial Support
This article 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. 262873, ‘Ideas’, 7th Framework Programme of the EU).
Data
The replication data set is available at http://thedata.harvard.edu/dvn/dv/ipsr-risp
Supplementary Material
To view supplementary material for this article, please visit http://dx.doi.org/doi:10.1017/ipo.2015.27
