Revisiting democratic civil peace

Abstract

The debate on democratic civil peace has centered on three general claims: democracies have a low risk of civil conflict, autocracies have the same low risk of conflict as democracies, and hybrid regimes have the highest conflict risk. We re-evaluate these claims, emphasizing that previous studies have focused on the aggregate categories of regimes, neglecting the role of particular institutional features. We propose focusing on the electoral qualities of regimes, which constitute the core of democracy, and argue that constraints on electoral contestation generate incentives for opposition to use force. Building on this framework, we distinguish between five regime types according to their electoral features – non-electoral autocracies, single-party autocracies, multi-party autocracies, minimalist democracies, and polyarchies – and specify hypotheses regarding the probability of conflict onset in each. In a global statistical analysis spanning 1817–2006 and employing the new Lexical Index of Electoral Democracy (LIED), we find that polyarchies, characterized by unconstrained contestation, have a lower probability of conflict than any other regime type (although minimalist democracies are only slightly more prone to conflict). Subsequently, we find that single- and multi-party autocracies, characterized by non-competitive elections, are more peaceful than non-electoral autocracies. Our analysis also reveals two factors that are particularly associated with civil peace: the presence of (any form of) elections and minimal electoral competition. Overall, our study underscores the importance of focusing on the central attributes of democracy and sheds new light on the relationship between particular regime features (or types) and civil conflict, thereby contributing to the growing efforts in conflict research to disaggregate political regimes.

Keywords

armed conflict autocracy civil war democratic civil peace elections political regime

Introduction

The proposition that democracy leads to civil peace has generated extensive debate in conflict research (Hegre, 2014). This debate has centered on three general claims. First, democracies have a low risk of conflict, as democratic regimes create few political grievances and have institutional means to accommodate contenders nonviolently. Second, autocracies have as low a risk of conflict as democracies, as autocratic regimes have the means to repress contenders violently. Finally, hybrid regimes (‘anocracies’) have the highest conflict risk, as they lack both the established institutions to accommodate contenders and effective means to repress them. Initially, empirical work provided considerable evidence to support these claims, suggesting that the democracy–conflict relationship follows an inverted U (Auvinen, 1997; Fearon & Laitin, 2003; Hegre et al., 2001; Henderson & Singer, 2000; Muller & Weede, 1990; Reynal-Querol, 2002). However, subsequent studies have challenged these claims, showing that anocracies do not in fact have a higher risk of conflict than other regimes (Treier & Jackman, 2008; Vreeland, 2008).

In view of this, we revisit the debate on democratic civil peace. We note that previous research has underexplored particular institutional features (see also Fjelde, 2010; Gleditsch & Ruggeri, 2010; Goldstone et al., 2010). Specifically, previous studies have used indicators representing the overall level of democracy that do not account for specific qualities of regimes theorized to influence peace – for example, institutions regulating access to power. We suggest that research on democratic civil peace should center on the electoral features of regimes, such as the presence and competitiveness of national elections, which form the ‘inescapably sine qua non’ of modern democracy (Huntington, 1991: 9).¹ Hence, we follow the research tradition adopting a ‘realistic’ understanding of democracy, according to which democracy is primarily a method that determines access to government power. This understanding of democracy follows Schumpeter, who defined it as ‘that institutional arrangement for arriving at political decisions in which individuals acquire the power to decide by means of a competitive struggle for the people’s vote’ (1976: 269).

We suggest that the electoral qualities of regimes are central not only to the concept of democracy but also to democratic civil peace. To support this claim, we present a theoretical discussion highlighting the role of electoral contestation (Coppedge, Alvarez & Maldonado, 2008; Dahl, 1971). It synthesizes four mechanisms through which electoral contestation may influence the incentives for the opposition to use force. Specifically, constrained electoral contestation can (i) motivate the opposition to substitute electoral competition with violence, (ii) legitimize the use of anti-government violence, (iii) ‘self-select’ the opposition recruits into violence, and (iv) hinder the incumbent’s strength.

Building on this framework and extant regime typologies (Diamond, 2002; Howard & Roessler, 2006; Møller & Skaaning, 2013), we categorize polities into five ‘electoral regime’ types (political regimes defined by their electoral characteristics): non-electoral autocracies, single-party autocracies, multi-party autocracies, minimalist democracies, and polyarchies. Subsequently, we specify hypotheses on the probability of conflict onset in each.

To shed new light on democratic civil peace, we employ the new Lexical Index of Electoral Democracy (LIED) (Skaaning, Gerring & Bartusevičius, 2015). This index simultaneously performs a classificatory function, each level identifying a unique regime type, and a discriminating function, distinguishing between different levels of democracy. LIED therefore allows not only assessing whether democracies are generally more peaceful than autocracies but also examining conflict risks in concrete regime types. Furthermore, spanning 1800–2013 and including 221 independent polities, LIED has the most comprehensive coverage among extant indices. This enables us to expand the conventional post-1945 time span threefold, to 1817–2006 (delimited by data on other covariates), substantially widening the temporal scope of our inferences.

Our analysis demonstrates that polyarchies, characterized by unconstrained electoral contestation, outperform all other regime types on civil peace. Moreover, we find that hybrid regimes characterized by at least nominal electoral competition are more peaceful than autocracies without elections. These findings suggest that democracy – understood in the electoral sense – relates to civil conflict in a negative monotonic way. Our disaggregated analysis also reveals two particular factors that are most strongly associated with civil peace: the presence of (any form of) elections and minimal electoral competition.

Altogether, this study contributes to the literature on political regimes and conflict in the following ways. Conceptually, it underscores that research on democratic civil peace must focus on the central attributes of democracy – not correlates of it (e.g. state capacity). Theoretically, it sheds new light on the mechanisms through which particular regime characteristics, as contrasted to aggregate categories, can influence conflict. Empirically, it assesses conflict risks in regime types that represent qualitatively meaningful and ordered categories, generating substantive insights into the relationships between particular electoral characteristics and conflict.

More generally, our study contributes to an emerging effort to disaggregate political regimes, arguably a prerequisite for progress in research on democratic civil peace. In a recent literature review, Ari et al. (2016: 4) noted that

we know less about the causal mechanisms tying specific aspects of institutions to conflict outcomes. Disentangling the highly aggregated categories of democracy and democratization […] seems a promising avenue for progress […] One potentially fruitful strategy is to unpack the ‘black box’ of aggregated democracy measures and investigate which regime-type dimensions are conductive to peace.

While similar calls to disaggregate political regimes appeared in previous research (e.g. Gleditsch & Ward, 1997), few studies have attempted to address these concerns. One exception is Gleditsch & Ruggeri (2010).² Criticizing the use of democracy indices as proxies of state capacity, they propose an alternative measure and show how it affects civil conflict. Further, controlling for state capacity, they identify a negative monotonic relationship between their democracy measure (adjusted Polity Index) and conflict. While the Gleditsch and Ruggeri study convincingly isolates the effects of state capacity from democracy, it sheds little light on the actual features of democracy that influence peace. Here, we explicitly pursue this aim: holding state capacity constant, we investigate the disaggregated features of democracy that account for democratic civil peace.

Research on political regimes and civil conflict

The democratic civil peace hypothesis has inspired numerous studies on political regimes and civil conflict (Auvinen, 1997; Dunning, 2011; Fearon & Laitin, 2003; Fjelde, 2010; Gleditsch & Ruggeri, 2010; Goldstone et al., 2010; Hegre et al., 2001; Henderson & Singer, 2000; Muller & Weede, 1990; Reynal-Querol, 2002). Many have stressed that democracies experience fewer political grievances than other regimes, as they are less repressive, more politically inclusive, and tolerant. While this does not imply the absence of societal discontent (as grievances may arise from other sources, e.g. the economy), fewer political grievances lead to fewer reasons for politically motivated, state-targeted action (Buhaug, 2006: 696).

Even if grievances arise, democracies can address them nonviolently. Democratic institutions (e.g. elections and legislatures) provide opportunities for contenders to pursue their interests and voice discontent. While such institutions foster political competition, they offer a major – arguably more attractive – ‘substitute’ for political violence (Dunning, 2011: 329–332; Gleditsch & Ruggeri, 2010: 301); the more competitive the democratic institutions, the less likely that the opposition pursue their interests violently.

Some scholars have claimed, however, that full autocracies may be just as (or even more) peaceful than full democracies (Buhaug, 2006; Goldstone et al., 2010). Due to constrained political competition, corruption, and violations of human rights, autocracies generate more grievances than democracies, but they can effectively subdue the opposition with force. While democracies can also use force against their citizens (e.g. to subdue illegal opposition), autocrats can do so more extensively because they rely less on popular support to remain in power (Hegre, 2014: 163).

While full democracies have the means to accommodate contenders and full autocracies possess the means to repress them, hybrid regimes often lack both, leading to the highest risk of conflict. The inconsistent nature of such regimes is the culprit. By mixing a level of repression that is not comprehensive enough to quell the opposition with some degree of political competition that is not sufficient to fully accommodate the opposition, anocracies both motivate violence and fail to counter it (Henderson & Singer, 2000: 279–280; Muller & Weede, 1990: 627; see also Hegre et al., 2001: 33).

Empirical research has tended to support these claims. Numerous studies have shown that full democracies are least likely to experience conflict (e.g. Gleditsch & Ruggeri, 2010; Hegre et al., 2001), others that full autocracies may be as peaceful as full democracies (Buhaug, 2006; Goldstone et al., 2010). Yet most studies have converged on the finding that anocracies are the most conflict-prone (Auvinen, 1997; Fearon & Laitin, 2003; Hegre et al., 2001; Henderson & Singer, 2000; Muller & Weede, 1990; Reynal-Querol, 2002), suggesting that the regime type–conflict relationship follows an inverted U.

However, Vreeland (2008) has shown that previous research has suffered from an endogeneity problem: the middle values of the Polity Index (Marshall, Gurr & Jaggers, 2016), a proxy of democracy used by most studies to identify anocracies, by definition partly reflect violence (see also Strand, 2007). Removing the elements of violence from the index, the middle values seem not to be associated with conflict. Moreover, Vreeland has shown that substituting the revised Polity Index (‘xPolity’) with alternative measures makes little difference: anocracies have no greater risk of conflict than other regimes (see also Gleditsch & Ruggeri, 2010; Treier & Jackman, 2008).

Yet the results based on xPolity and other graded indicators of democracy suffer more fundamental problems. To disentangle the anocracy–conflict relationship, we must first clearly define anocracies themselves (Ari et al., 2016: 5). Available indicators of political regimes, however, paint a different picture: they merely provide a score reflecting the overall level of democracy without linking specific characteristics to the different levels. While a country with a value of 6 on the Polity scale is considered more democratic than one with a 5, this score tells us little about the actual qualitative differences between the two regimes.

Furthermore, the graded measures of political regimes are aggregates of many subcomponents, and different combinations of values often lead to the same aggregate scores (Cheibub, Gandhi & Vreeland, 2010; Gleditsch & Ward, 1997). Thus, a country scoring low on competitiveness of executive recruitment and high on executive constraints might have the same overall score as another scoring high on competitiveness of executive recruitment and low on executive constraints. Consequently, such ‘mashup indices’ (Ravallion, 2011) generate heterogeneous pools of autocracies, democracies, and – particularly – hybrid regimes.

These problems have prevented the identification of the exact regime features associated with conflict. Fortunately, recent developments in regime data collection have turned towards greater disaggregation, producing data suitable for our research goal. We present these data in the subsequent section; we now establish a theoretical link between electoral regimes and conflict.

Electoral contestation and civil conflict

As emphasized, understanding the relationship between the electoral characteristics of regimes and civil peace is essential to understanding the democracy–peace nexus. This follows from the concept of democracy, where elections constitute the essential core (Huntington, 1991: 9; Schumpeter, 1976: 269). The key electoral elements are national elections for the legislature and (directly or indirectly) the executive, participation of multiple parties/candidates, and uncertainty about the outcome (Przeworski et al., 2000: 15–16). Together with freedom of expression, which supports the meaningfulness of elections, these elements pertain to what Dahl (1971) has labeled ‘contestation’ (see also Coppedge, Alvarez & Maldonado, 2008). Accordingly, we posit that the arguments linking democracy to conflict should center on these core features.

Dahl (1971) refers to an additional electoral dimension: inclusiveness (or participation). Here, however, we exclusively focus on contestation. Variation in inclusiveness is almost absent after 1945, undermining the relevance of this dimension for understanding contemporary civil conflicts. More importantly, a unidimensional focus on contestation ensures better alignment between our concepts, theory, and measures (Coppedge, Alvarez & Maldonado, 2008). Analyzing both dimensions would involve specifying the theoretical mechanisms pertaining to each and how they interact in producing conflict, and testing their separate (or interacting) effects in empirical analysis. Here, trading extensiveness for relevance and parsimony, we exclusively focus on contestation.

Electoral regimes and electoral events

In our framework, political regimes³ are distinguished by the presence/absence of periodic elections and their competitiveness, that is, whether opposition groups are allowed to compete in elections, have a chance of winning, and enjoy freedom of expression. While we use elections to categorize political regimes (hence, ‘electoral regimes’), we do not focus on the short-term effects of electoral events on violence. Instead, we focus on the sustained effects of electoral regimes, which affect the behavior of opposition actors beyond the actual electoral events. For example, if regimes allow periodic electoral competition, opposition actors are incentivized to await elections and pursue their interests nonviolently between them (via political parties, lobbyism, etc.). Conversely, if regimes constrain electoral competition, opposition actors are disincentivized to wait, instead contesting power violently between elections (see also Harish & Little, 2017).

Studies have shown that electoral events in non-democratic regimes increase the risk of ethnic civil war (Cederman, Gleditsch & Hug, 2013), recurrence of civil conflict (Brancati & Snyder, 2013), and small-scale civil violence (Salehyan & Linebarger, 2015). However, these effects have primarily been attributed to the short-term prospects for mobilization around electoral events (Cederman, Gleditsch & Hug, 2013: 390; Salehyan & Linebarger, 2015: 26, 30; see also Kuntz & Thompson, 2009: 254), which do not exist between elections. Hence, while electoral events and regimes are intrinsically related (i.e. the former defines the latter), their immediate and transient effects on peace are likely different.

In a related study of regime survival, Knutsen, Nygård & Wig (2017) suggest that elections serve as ‘focal points’ for mobilizing opposition. Mobilization in autocracies constitutes a typical collective action problem: people are reluctant to express their grievances due to fear of punishment and lack of knowledge as to whether others share their grievances. The salient political nature and time-limited character of elections renders overcoming such collective action problems possible, increasing the prospects of coordinated regime challenge. Yet, elections and associated institutions (e.g. parties and legislatures) also provide information, legitimacy, and means for co-optation for the incumbents, which, in the longer run, prevent anti-government challenge. In extensive empirical analysis, Knutsen, Nygård & Wig (2017) find that electoral events indeed increase the immediate risk of regime breakdown, whereas periodic elections stabilize autocracies in the longer term.

Focusing more directly on civil violence, Harish & Little (2017) present a formal model suggesting a similar pattern. Their model builds on the same assumption that elections generate favorable opportunities for mobilization, thereby increasing the effectiveness of violent anti-government action. Aware of this, the opposition should refrain from using violence between elections and displace it instead to electoral events. Compared to a baseline without periodic elections (where the effectiveness of anti-government violence is more constant), long non-electoral periods intersected with short electoral events should lead to a net positive effect on peace.

Electoral regimes are thus likely to have separate effects (from electoral events) on conflict. The likelihood of conflict, however, depends not only on the presence/absence of periodic elections but also on the actual qualities of regimes.

Electoral regime types and incentives for violence

To explain the electoral regime–conflict relationship, we take the perspective of opposition and focus on the following mechanisms: (i) ‘substitution’, (ii) legitimacy of anti-government violence, and (iii) ‘self-selection’. While discussing electoral autocracies, we also consider (iv) how elections and nominal electoral institutions (parties, legislatures, etc.) can strengthen the incumbents and prevent anti-government violence. Below, we first introduce the general mechanisms and then elaborate in detail how they operate in the context of particular regime types.

Nonviolent institutional action (e.g. election participation) involves fewer costs than violent non-institutional action (civil conflict) due to the resources and risks involved in armed conflicts. Unlike election participation, an armed conflict with a government requires extensive finances, arms, and recruits, sufficiently motivated to risk injury and death (Hendrix, 2010: 274). Chenoweth & Stephan (2011: 32–39) highlight how violent rebellions face greater obstacles to recruitment – due to the risks and costs associated with armed conflict – on average attracting many times smaller membership than nonviolent campaigns. Hence, we assume that a nonviolent institutional action⁴ is preferred to a violent non-institutional one when the probability of success of the two actions is at least on par. We define the success of an opposition action broadly as gaining access to government power or changing the incumbent’s policies.

The probability of the success of nonviolent institutional action should vary positively with contestation. If incumbents allow electoral competition, nonviolent action may be an effective choice, whereas violent action may be more attractive if incumbents restrict electoral competition. Thus, ‘fighting and voting can be seen as strategic substitutes’ (Dunning, 2011: 329), and free competition should incentivize participation in elections by increasing the opposition’s probability of success of gaining access to power via electoral means. Further, post-conflict literature highlights how democratic institutions mitigate commitment problems among the opposition (e.g. Brancati & Snyder, 2013; Durant & Weintraub, 2014; Flores & Nooruddin, 2012; Matanock, 2017). When elections occur in post-conflict settings, ex-warring parties are uncertain about each other’s willingness to comply with the formal electoral process, resulting in neither side credibly committing to the elections and a higher risk of conflict recurrence. Such commitment problems can be mitigated by fair electoral competition (Flores & Nooruddin, 2012: 561–562) accompanied by institutions ensuring compliance with elections (e.g. professionalized bureaucracies, impartial courts, and free media; see Brancati & Snyder, 2013: 829–831) or particular electoral design constraining the executive and enabling ‘the rule of law for elites’ (Durant & Weintraub, 2014: 527). This arguably generalizes to non-post-conflict settings, where constrained contestation results in uncertainties regarding the respect for electoral outcomes, disincentivizing the opposition from committing to elections and increasing the chance of substitution with violence.

The probability of the success and costs associated with nonviolence/violence also depends on the legitimacy of anti-government violence. If everyone can pursue interests via the same institutional means and elections constitute the sole legitimate path to power, violence against incumbents may be unpopular – and therefore costly and ineffective (Buhaug, 2006: 696; see also Cheibub & Hays, 2017: 91). Conversely, if only particular political actors are allowed to compete for power (e.g. as in single-party regimes) and electoral competition provides little chance of gaining access to power, anti-government violence may have popular support and be more likely to succeed. In cases where elections and legislatures are entirely absent, incumbents cannot even play the ‘popular legitimacy card’, as there are simply no institutions for popular representation (Shock, 2005).

When electoral competition is constrained or banned, opposition often operates underground, which involves multiple barriers that do not exist while joining legal organizations such as political parties in democracies. Chenoweth & Stephan (2011) discuss such barriers in the context of violent campaigns, many of which can be generalized to the context of underground activities (violent or not). Like joining an armed fight, membership of illegal oppositions entails higher risks and costs (e.g. capture and imprisonment) than membership of legal organizations. Furthermore, given their clandestine nature, would-be members have limited information about the groups they join, meaning higher uncertainty (about the size and strength of the group, its goals, etc.). Because of their secret character, underground groups themselves must follow careful recruitment practices, selecting only those with high resolve, reliability, and other qualities pertaining to clandestine activities. Prospective members must therefore be highly motivated and willing to take risks. Arguably, such qualities are conductive to violence. Individuals committing to illegal underground groups are likely to be more risk-taking and less averse to anti-government violence. Thus, compared to legal political organizations, clandestine opposition groups should be staffed with people more likely to accept and use political violence, ‘self-selecting’ into conflict.

The discussion above thus suggests a general pattern: constrained electoral contestation incentivizes opposition actors to use force, implying that (H1) the probability of civil conflict onset is negatively associated with electoral contestation.

We now detail each of the mechanisms in the contexts of particular regime types, categorized based on different degrees of contestation. In line with extant typologies of electoral regimes (Diamond, 2002; Howard & Roessler, 2006; Møller & Skaaning, 2013), we distinguish between five generic categories: non-electoral autocracies (no elections); single-party autocracies (single-party elections); multi-party autocracies (multi-party non-competitive elections); minimalist democracies (multi-party competitive elections); and polyarchies (multi-party competitive elections with freedom of expression).

As the label implies, non-electoral autocracies hold no national elections (e.g. present-day Saudi Arabia, Qatar) or postpone them for an unknown length of time (Philippines under Marcos’s martial law, present-day Eritrea). In such contexts, opposition actors are incentivized to use force; yet, doing so can incur great costs. As electoral institutions are absent, the incumbents must engage the opposition using other means, usually repression (Davenport, 2007: 490; Fjelde, 2010: 200), which can take extensive forms, since regime survival rests less on popular support (Hegre, 2014: 163; Schock, 2005: 31–32, 49–50).

This has led previous research to claim that full autocracies are ‘peaceful’; however, there is good reason to doubt this claim. If repression effectively prevented violence, highly repressive regimes would have little conflict. History reveals this hardly to be the case. The top ten countries in the Correlates of War data (elaborated below) with the highest incidence of large-scale civil war (Sudan, Philippines, Colombia, Burma, Indonesia, Angola, Sri Lanka, Ethiopia, Chad, and Afghanistan) are/were some of the most repressive in the world.

Highly repressive states free of conflict do exist (e.g. Saudi Arabia), but such cases are atypical regarding wealth and state capacity. Unlike the cases above, states like Saudi Arabia can afford a large professional security apparatus capable of maintaining order. Indeed, state capacity, backed with a strong economy, features as a key factor preventing anti-government violence in the qualitative literature (e.g. Bellin, 2004). Large-N studies on political regime and conflict also tend to focus on ‘state weakness’, highlighting the role of state capacity in identifying and subduing anti-government mobilization (e.g. Fearon & Laitin, 2003).

The low likelihood of conflict in full autocracies, as identified in previous research, might therefore be due to state capacity, a feature that is not a conceptual attribute of democracy (e.g. Mazzuca, 2010).⁵ Indeed, we see little theoretical reason why fully autocratic regimes – understood in the electoral sense – should be peaceful. Given that institutional competition over power is entirely absent, the incentives to substitute it with violence should be greater and have greater popular support than in any other regime type. Further, given that political opposition is entirely banned, most regime contenders must operate underground, attracting members more likely to accept and use anti-government violence. Thus, we expect that (H2) non-electoral autocracies have a higher probability of conflict onset than any other electoral regime type.

The discussion above suggests that H1a should be tested while controlling for state capacity, but how exactly does state capacity relate to democracy – or, more importantly, to electoral features of regimes – and civil war? Arguably, the choice to introduce elections depends on the incumbents’ capacity to repress opposition. Bellin (2004: 146) and Fjelde (2010: 199) highlight that repression is rather costly and that only high-capacity states can exclusively rely on it to quell the opposition. If state capacity is low and resources are insufficient to coerce opponents to ‘peace’, incumbents must introduce elections. Likewise, the competitiveness of such elections depends on the incumbents’ capacity to repress: the lower the capacity, the more competitive the election required. Thus, we must hold state capacity constant to identify the association between electoral regimes and conflict. The interplay between state capacity and electoral competition is an interesting question in its own right. Yet our focus here is on electoral features of regimes, categories conceptually most central to democracy.

Civil conflict research often treats electoral autocracies as transitional regimes, but research on electoral authoritarianism suggests that this regime type is often a deliberate choice aimed at undermining the opposition and keeping autocrats in power. Regular elections and nominal legislatures can serve as effective institutional settings to co-opt challengers, for example, via spoils allocated through ‘elected’ seats (Boix & Svolik, 2013; Gandhi & Przeworski, 2006). Autocrats can also use electoral campaigns to signal strength: ‘By rolling out an impressive electoral campaign machinery and whipping up popular support, the regime credibly signals to (1) the opposition that armed confrontations are futile and (2) the internal elite that coups will be opposed by numerous supporters’ (Knutsen, Nygård & Wig, 2017: 103; also Cheibub & Hays, 2017). Elections can also help reveal the strength of challengers (Little, 2012) and identify them – and they can then be repressed, discredited, or co-opted (Magaloni, 2008: 71). Finally, incumbents can use elections and parties to channel popular mobilization into regime-supportive activities and to create forums where ‘political aspirations and demands from competing factions can be discussed [within regimes’ parties] without challenging the foundations of the regime’ (Fjelde, 2010: 199; see also Davenport, 2007: 491; Gandhi & Przeworski, 2006). Altogether, this suggests that nominal democratic institutions can help autocrats to ‘divide and rule’, weakening the opposition and preventing anti-government violence.

From the opposition perspective, electoral autocracies can also reduce incentives for violence. Compared to non-electoral regimes, the spectrum of means to compete over power increases with one additional choice (Shock 2005: 15). While they do not allow the opposition to gain actual power, nominal democratic institutions do provide a ‘venue within which discussion/aspirations/activism can take place – in a sense, it may be the only “show in town”, but at least there is a show’ (Davenport, 2007: 490). Elections in autocracies can even result in some degree of popular representation. In Vietnam, for example, a single-party regime, constituency delegates not only raised issues relevant to their constituents in the national assembly but also directly questioned the performance of the executive branch (Malesky & Schuler, 2010).

Additionally, elections and associated institutions, even if only nominal, can improve the legitimacy of incumbents (Schedler, 2002), thereby undermining support for anti-government violence. Further, in some (multi-party) electoral autocracies, opposition actors are formally allowed to organize and campaign. Compared to underground groups, legal opposition parties have a more diverse membership base, which is less likely to accept (or be motivated to use) anti-government violence. Hence, electoral autocracies should generally be less prone to conflict than non-electoral regimes.

Electoral autocracies come in two basic forms: single-party, where access to power is constrained by banning opposition parties (e.g. present-day Cuba, Vietnam) and multi-party, where competition for power is constrained via informal mechanisms such as intimidation, bribery, and electoral manipulation (e.g. present-day Russia, Zimbabwe). Based on our conceptual framework, we consider non-electoral autocracies, single-party autocracies, and multi-party autocracies as representing three different levels of electoral contestation. On this basis, we hypothesize that (H3) single-party autocracies have a lower probability of civil conflict onset than non-electoral autocracies and (H4) that multi-party autocracies have a lower probability of civil conflict onset than single-party autocracies.

However, the difference in the risk of conflict in single- and multi-party autocracies may not be substantial. While multi-party autocracies, unlike single-party regimes, allow the opposition to organize in parties and compete in national elections, thereby reducing incentives for (and delegitimizing) anti-government violence, electoral competition in multi-party autocracies remains constrained, providing incentives for (and legitimizing) anti-government violence. Further, formally allowed to organize and campaign, opposition actors in multi-party autocracies have better opportunities to mobilize (e.g. Fjelde, 2010: 201). Such opportunities are also greater because multi-party settings provide less effective institutional context for monitoring, co-opting, and repressing the opposition (Davenport, 2007; Fjelde, 2010). Counteracting mechanisms may therefore be at play, generating similar net likelihood of conflict in single- and multi-party autocratic regimes.

The only reliable way to reduce the incentives for anti-government violence is, thus, to allow credible competition over power via nonviolent institutional means. In such contexts, citizens are also less likely to accept violence against a democratically elected government, and opposition actors are more likely to consist of diverse memberships, less willing to use anti-government violence. Thus, we hypothesize that (H5) regimes with minimally competitive elections have lower probability of civil conflict onset than multi-party autocracies.

Finally, we expect polyarchies – regimes where freedom of expression strengthens electoral competition – to have the lowest probability of conflict. In such regimes, citizens can express their preferences without undue restrictions. This increases the credibility of electoral processes and contestation more broadly, further strengthening the legitimacy of regimes and reducing incentives for violence to a minimum. Thus, we hypothesize that (H6) polyarchies have a lower probability of conflict onset than any other electoral regime type.

Research design

We initially tested the hypotheses in a standard country-year logit regression. The sample included all annual observations of states as listed in the LIED dataset (which are based on Gleditsch and Ward’s list of independent states; see Skaaning, Gerring & Bartusevičius, 2015: 1502). In robustness checks, we implemented a number of additional analyses, including tests with alternative measures of the key variables and alternative analysis techniques.

The Lexical Index of Electoral Democracy

To measure electoral regimes, we employed the Lexical Index of Electoral Democracy (LIED) (Skaaning, Gerring & Bartusevičius, 2015), which allows testing our hypotheses more directly than other indicators of democracy. LIED exclusively focuses on the features of electoral regimes and adheres to the minimalist definition of democracy: ‘regime where leaders are selected through contested elections held periodically before a broad electorate’ (Skaaning, Gerring & Bartusevičius, 2015: 1495). The index combines dichotomous indicators of different regime qualities in a systematic fashion, where a series of necessary and jointly sufficient conditions are arranged in a seven-level ordinal scale: (0) no elections; (1) no-party/single-party elections; (2) multi-party elections for legislature; (3) multi-party elections for legislature and executive; (4) minimally competitive, multi-party elections for legislature and executive; (5) minimally competitive, multi-party elections with full male or female suffrage for legislature and executive; (6) minimally competitive, multi-party elections with universal suffrage for legislature and executive.

The second level has very few observations and offers little relevant nuance regarding theoretical distinctions.⁶ Hence, we collapsed it with the third level to form the multi-party autocracy category. The two highest levels concern electoral inclusiveness. Since we exclusively focus on contestation, we combined them with the original fourth to form the minimalist democracy category.

The original index does not explicitly account for freedom of expression. One could thus argue that the highest level potentially covers some hybrid regimes (e.g. Diamond, 2002; Schedler, 2002). We therefore introduced an additional category that supplements the criteria for minimalist democracy with a high degree of respect for freedom of expression. This addition qualifies the competitiveness of elections and signifies a continuous check on government powers. As such, the index accounts for all of the aspects of Dahl’s contestation, and includes five categories:

L0 = no elections (non-electoral autocracy)

L1 = no-party or single-party elections (single-party autocracy)

L2 = multi-party elections for legislature and (direct or indirect) executive (multi-party autocracy)

L3 = minimally competitive, multi-party elections for legislature and (direct or indirect) executive (minimalist democracy)

L4 = minimally competitive, multi-party elections for legislature and (direct or indirect) executive and extensive freedom of expression (polyarchy)

The coding of L0–L2 is clearcut, as the corresponding features are directly observable; however, the coding of minimally competitive elections is based on more elaborate rules. LIED follows the three criteria for contestation outlined by Przeworski et al. (2000: 16): ex-ante uncertainty (elections are sufficiently free to allow the opposition a chance of winning), ex-post irreversibility (the winners take office), and repeatability (the electoral regime is not interrupted). However, whereas Przeworski et al. exclusively rely on government turnover under the same electoral rules as an indicator of contested elections, LIED also includes information from country-specific sources about voter intimidation and electoral fraud (Skaaning, Gerring & Bartucevičius, 2015: 1501).⁷

To operationalize freedom of expression, we employed the Global Media Freedom Dataset (Whitten-Woodring & Van Belle, 2017), which distinguishes between three degrees of media freedom, the highest signifying ‘criticism of government and government officials is a common and normal part of the political dialogue in the mediated public sphere’. For observations with minimally competitive elections not covered by this dataset, we implemented additional coding based on the same criteria using country-specific sources.

In contrast to the Polity and Freedom House measures, violent conflict is not directly captured by any of the LIED components. Civil conflict can influence whether elections take place and reduce their competitiveness, but minimally competitive elections can co-exist with civil war, as many of the examples in the dataset signify. Together with the lag structure of our empirical model, the risk of endogeneity related to this issue is arguably small (at least smaller than in prior research).

LIED combines different features into regime types based on theoretical considerations over the centrality of particular features to the concept of electoral democracy. This cumulative ordering implies that the index performs both classificatory (each level connects to a particular combination of regime features) and discriminating functions (distinguishing between levels). Hence, the modified LIED allows not only assessing the overall relationship between democracy and conflict but also identifying the associations between particular configurations of institutional characteristics (e.g. non-competitive multi-party elections) and conflict.

Outcome variable

LIED covers 1800–2013, expanding the conventional post-1945 time-span threefold. This allows us to substantially widen the temporal scope of our inferences and account for factors related to particular historical periods (due to limited data on other covariates, the final analysis spans 1817–2006). To match this scope on the left side of the equation, we employed the data on civil wars (Categories 4–5) from the Correlates of War (COW) Intra-State War Data (4.1) (Sarkees & Wayman, 2010). COW defines civil war as sustained combat between a state government and non-state actor(s) resulting in at least 1,000 battle-related deaths annually and taking place within the state territory. Since we focused on civil war onset, country-years after the onset year were set to 0 (the dataset contains new civil war onsets during ongoing years of other civil wars in the same countries). Following conventions, we also applied the two-year intermittency rule. The final sample included 282 separate onsets (Table I reports civil war onsets for different regimes in different periods).

Table I.

Cross-tabulation: LIED and civil war onset – historical periods

	Civil war
	1817–80		1881–1945		1946–2006		1817–2006
LIED	0	1	0	1	0	1	0	1
L0	1,680	27	809	29	1,339	55	3,828	111
L1	315	11	579	15	1,934	30	2,828	56
L2	1,175	25	1,103	19	1,548	40	3,826	84
L3	337	3	957	3	1,739	23	3,033	29
L4			300	0	2,056	2	2,356	2

Control variables

We limited control variables to the likely confounders (i.e. potentially relating to both electoral regimes and civil war). To avoid missing observations and potential collinearity problems, we only included the most likely confounders identified in the literature: wealth, economic growth, oil wealth, state capacity, political regime characteristics unrelated to elections, and instability.

Wealth and economic growth have been linked both to democracy (e.g. Przeworski et al., 2000) and conflict (e.g. Hegre & Sambanis, 2006). Therefore, we introduced the natural log of GDP/cap and annual GDP/cap growth (both lagged). Data on GDP/cap came from Bolt & van Zanden (2013). While theirs is the most complete set of data on GDP/cap for our analysis period, it contains missing observations. Therefore, we applied a standard interpolation procedure to impute values between available observations, increasing the number of observations from 10,125 to 11,479. Excluding interpolated values from the analysis below does not affect our substantive findings. Furthermore, research has shown that democracy negatively relates to energy resource wealth, particularly oil, and that oil wealth positively relates to conflict (e.g. Ross, 2012). We therefore controlled for the total oil income per capita from Haber & Menaldo (2011) (lagged).

As argued above, the relationship between political regime and conflict can also be confounded by state capacity. Therefore, we introduced the Composite Index of National Capability (CINC) (lagged) from the National Material Capabilities dataset (Singer, 1987). CINC is based on the values of total population, urban population, iron and steel production, energy consumption, military personnel, and military expenditures. We also used the CINC subcomponent: military personnel. This measure accounts more directly for a state’s military capacity and, by extension, its repressive strength.

The electoral qualities relate to other characteristics of regimes that may confound the relationship of interest. Therefore, we also included the Polity2 index from the Polity IV project (Marshall, Gurr & Jaggers, 2016). As some of its subcomponents reflect quality of elections and political competition, Polity2 partly overlaps with LIED.⁸ Yet Polity2 also captures non-electoral features of regimes, such as factionalism and constraints on the executive that may confound the LIED–civil war relationship.

To assess whether the relationship between electoral regimes and conflict is not driven by changes in electoral regimes, we also introduced a measure of regime change/instability. For the full-sample analysis, we used a measure indicating whether a country experienced a one-point increase or decrease in LIED in the preceding three years (alternatively, in the preceding year). For analysis of post-1945 data, we used democratization and autocratization measures (one- and three-year versions) from Cederman, Hug & Krebs (2010).

We also controlled for population size (ln) with data from Singer (1987) and temporal dependence with cubic polynomials (Carter & Signorino, 2010).⁹ Robustness tests included other potential confounders. The Online appendix (OA henceforth) provides summary statistics.

While we control for likely confounders and apply fixed effects and time-lags, we cannot entirely rule out the possibility of recursive effects and omitted variable bias. In OA (Tables OA6–OA7 and Figures OA1–OA3), we therefore report an analysis of factors associated with LIED.

Results

Main analyses

Table II reports the main results. We first tested H1, treating LIED as an interval/ratio variable. Subsequently, we tested H2–H6, using a ‘dummy-coded’ version of LIED, treating each level as a separate category.

LIED has a negative, highly significant coefficient when regressed separately.¹⁰ The same is true when

Table II.

Logistic regression estimates of civil war onset: 1817–2006

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
LIED	–0.34*** (0.05)	–0.21*** (0.06)	–0.20** (0.07)	–0.21** (0.07)	–0.21** (0.07)	–0.30*** (0.09)	–0.30*** (0.09)		–0.43*** (0.08)
L1								–0.41* (0.20)		–0.47* (0.18)	–0.41* (0.20)
L2								–0.37^† (0.20)		–0.56** (0.18)	0.03 (0.22)
L3								–1.09** (0.33)		–1.59*** (0.31)	–0.71* (0.29)
L4								–2.02** (0.62)		–2.66*** (0.54)	–0.93^† (0.55)
Population	0.27*** (0.03)	0.22*** (0.04)	0.22*** (0.04)	0.21*** (0.04)	0.25*** (0.06)	0.24*** (0.06)	0.24*** (0.06)	0.25*** (0.06)	0.25*** (0.05)	0.26*** (0.05)	0.25*** (0.06)
GDP/cap (ipol)		–0.48*** (0.10)	–0.48*** (0.10)	–0.48*** (0.11)	–0.47*** (0.11)	–0.46*** (0.11)	–0.46*** (0.11)	–0.42*** (0.11)			–0.42*** (0.11)
GDP growth(ipol)			–1.76^† (1.06)	–1.71 (1.06)	–1.70 (1.06)	–1.29 (1.09)	–1.21 (1.08)	–1.17 (1.09)			–1.17 (1.09)
Oil income				–0.03 (0.08)	–0.03 (0.08)	–0.03 (0.08)	–0.03 (0.08)	–0.03 (0.08)	–0.16 (0.12)	–0.14 (0.12)	–0.03 (0.08)
CINC					–2.55 (2.14)	–2.71 (2.18)	–2.53 (2.18)	–2.49 (2.15)	–2.94 (2.03)	–2.74 (2.00)	–2.49 (2.15)
Polity2						0.02 (0.02)	0.02 (0.02)	0.03^† (0.02)	0.02 (0.02)	0.04* (0.02)	0.03^† (0.02)
Instability(3y)							0.14 (0.16)	0.12 (0.16)	0.12 (0.15)	0.10 (0.15)	0.12 (0.16)
Constant	–5.02*** (0.36)	–1.22 (0.80)	–1.32 (0.81)	–1.14 (0.84)	–1.57^† (0.90)	–1.34 (0.92)	–1.45 (0.93)	–1.73^† (0.94)	–4.51*** (0.51)	–4.58*** (0.51)	–
N	13,637	10,664	10,528	10,223	10,205	10,115	10,115	10,115	11,442	11,442	10,115
chi2	263.21	240.71	236.47	233.13	233.45	228.61	229.36	235.96	231.22	242.79	235.96

L1 = single-party autocracy; L2 = multi-party autocracy; L3 = minimalist democracy; L4 = polyarchy. Non-electoral autocracy (L0) is the base category in Models 8 and 10. In Model 11 the base categories are: L0 for L1; L1 for L2; L2 for L3; and L3 for L4 (resulting in four different constants: –1.73^† (0.94); –2.13* (0.95); –2.10* (0.95); and –2.82** (1.02), respectively). Standard errors in parentheses. Cubic polynomials not reported. ^† p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001.

LIED is regressed together with the controls. Results also remain similar when using the alternative measures of capacity and instability described above (not reported here). Turning to the dummy-coded version of LIED (Model 8), we find that, compared to non-electoral autocracies, all regime types have significantly lower probability of civil war. The coefficient for polyarchy, however, should be treated with caution, as there are only two civil war onsets in such regimes in the whole sample (addressed below using alternative measures of conflict).

Due to combined missing observations among the covariates, Models 7–8 contain 10,115 observations of a possible 16,153. While such missingness is not untypical (even for post-1945 analyses), we attempted to increase the sample by removing GDP/cap and growth, covariates with most missing values (note that state wealth is partly controlled for by the remaining CINC). This generated 1,327 additional observations, further strengthening the association between LIED and civil war (Models 9–10).

In Models 8 and 10, the base category is non-electoral autocracy. A stronger test of H2–H6 is a successive comparison of each regime type to one level below. Model 11 shows results with GDP/cap and growth included back in the block. The coefficients for single-party autocracies, minimalist democracies, and polyarchies remain significant. In contrast, while significant when compared to the base category of non-electoral autocracies, the coefficient for multi-party autocracy turns insignificant when compared to single-party autocracy. Before discussing these results in the next section, we now turn to robustness tests.

Robustness tests

An almost-200-year time span increases the scope of our inferences and allows us to rule out the possibility that these findings are confounded by idiosyncrasies of particular periods. However, electoral regimes and their links to conflict are potentially different now than they were 200 years ago, pointing to potential heterogeneity problems. To address this, we limited the sample to post-1945.

Table OA2 reports the estimates. The coefficient of the interval LIED remains significant at 0.1% despite the sample size dropping to 6,823 and the number of civil wars from 232 to 134. The estimates for the dummy LIED are similarly in line with previous results. All regimes have lower probability of civil war than non-electoral autocracies (although the coefficient for single-party autocracy is now above 10%).

Analyzing post-1945 allows exploring additional confounders. Recent research has identified a robust relationship between horizontal inequalities and civil conflict (e.g. Cederman, Gleditsch & Buhaug, 2013). Horizontal inequalities may also be associated with electoral qualities of regimes, as political and economic discrimination is often maintained by constraining contestation. Therefore, we controlled for horizontal ethno-political inequality – captured by ‘max discrimination’ (the relative demographic size of the largest ethnic group subject to active discrimination) – and horizontal economic inequality, proxied by ‘max low ratio’ (the relative income gap between the poorest group and the national average). Following previous research, we also controlled for inequality and ethnic diversity in the total population with the Gini coefficient and an ethnic fractionalization index. The four indices come from Cederman, Gleditsch & Buhaug (2013). The estimates for our key variable essentially remain the same. All LIED levels also have significant coefficients (the coefficient for single-party autocracy is now significant at 5%). As previously, a stronger test comparing each regime type to one level below indicates that all levels, besides L2 (multi-party autocracy), have significant coefficients.

We subsequently introduced alternative measures of the dependent variable (Table OA3). First, we substituted the COW civil war with internal war from the UCDP/PRIO Armed Conflict Dataset v.4-2011, 1946–2010 (Gleditsch et al., 2002; Themnér & Wallensteen, 2011). The coefficient of the interval LIED remains negative and significant at 5%. Similarly, all levels of LIED retain negative signs, although the coefficients for single- and multi-party autocracies are not significant when compared to the reference category, and the coefficient for polyarchy is not significant compared to minimalist democracy. Second, we substituted the COW category of civil war with a lower-intensity armed conflict (from the same UCDP/PRIO dataset). The coefficient of the interval LIED remains negative and significant, although p = 0.059, and the estimate itself is smaller. This difference potentially pertains to the fact that regime-related factors affect whole populations, having the potential to cause large-scale violence. Conversely, minor conflicts often concern particular segments of populations, such as peripheral ethnic groups (e.g. Buhaug, 2006) fighting states over their specific status (not the status of the states). To assess this claim, we replicated our analysis using the UCDP/PRIO’s governmental war data and found that the coefficient of LIED – compared to the aggregate category – became considerably larger and more significant (Table OA3, Models 17g–19g).

Figure 1.

Estimated probabilities (with 95% CI) of civil war onset as a function of interval (upper panel) and dummy-coded LIED

As previously, all levels of the dummy LIED have negative signs. The coefficients for single- and multi-party autocracies remain insignificant compared to the base category, but the coefficient for polyarchy attains significance compared to minimalist democracy (polyarchies contain eight onsets of UCDP/PRIO’s internal war and 14 onsets of minor conflict).

We also controlled for time-invariant country-level factors via country-fixed effects. Despite a two-fold reduction of the sample (analysis of full data dropped 83 polities), the estimates correspond to the previous results (Tables OA4–OA5). The interval LIED is negative and significant in all models but one. In most specifications, all LIED levels also retain negative signs, although only the coefficient for minimalist democracy is consistently significant across all specifications.

Discussion

The results support our central expectation that contestation positively relates to civil peace. The analysis of interval LIED suggests that this relationship is potentially linear, concurring with Gleditsch & Ruggeri (2010: 303) that once we control for state capacity, democracy relates to civil war in a negative monotonic way. However, Figure 1 demonstrates that this relationship is more nuanced than the linear model suggests. Single-party autocracies have a substantially lower risk of conflict than non-electoral regimes; however, multi-party autocracies only have a slightly lower risk of conflict than single-party autocracies. Subsequently, minimalist democracies have substantially lower risk of conflict than multi-party autocracies. This indicates that the key elements accounting for civil peace are the presence of (any form of) elections and minimal electoral competition. More generally, these results demonstrate the usefulness of LIED: it allows us to assess conflict risks in particular regime types and reveal key electoral features that are associated with civil peace.

These patterns are consistent with our theoretical claims. If electoral contestation is absent, the opposition has incentives to use force. Elections and associated institutions – even if non-competitive – apparently reduce such incentives, probably because they offer alternatives to violence, facilitate autocratic divide-and-rule strategies, and diminish acceptance of anti-government violence. The fact that multi-party autocracies are only slightly less prone to conflict than single-party regimes indicates that regimes allowing opposition parties but no genuine competition introduce counteracting mechanisms that partly cancel out the positive effects on civil peace associated with increased political openness.

Irrespective of their form (single- or multi-party), electoral autocracies therefore appear to be intrinsically conflict-prone (even if less conflict-prone than non-electoral regimes): civil peace is more likely to prevail when the opposition has a meaningful chance of gaining power via electoral means. The less constrained the electoral contestation, the less attractive, costly, and illegitimate anti-government violence. Polyarchies, where electoral contestation is qualified by high levels of freedom of expression, are therefore least prone to conflict (although not much less conflict-prone than minimalist democracies).

As hinted above, non-electoral regimes are highly repressive, and the risk of conflict in such regimes might therefore not pertain to ‘substitution’, legitimacy, or ‘self-selection’, but rather to the repression–dissent link. Contrary to extant claims in the literature, pervasive repression, coupled with low state capacity, may not only fail to prevent violence but actually provoke it: ‘an overwhelming use of coercive force is a costly strategy with a high risk of backfiring. It depletes bases of support and strengthens the cause of potential conspirators to depose the dictator’ (Fjelde, 2010: 199). Additionally, repression potentially ‘increases the costs and decreases the anticipated success of nonviolent relative to violent resistance’ (Rørbæk, 2016: 1), further heightening the risk of conflict. We claim not that alternative or supplementary mechanisms are implausible, noting only that our research design cannot adjudicate between causal explanations. Large-N cross-national design allows the identification of patterns consistent with the theorized claims. Identifying the actual mechanisms driving these patterns, however, requires scaling down and alternative analytical approaches (e.g. congruence analysis or process tracing).

Irrespective of the actual mechanism accounting for our findings, the patterns revealed in our analysis lead us to concur with Hegre et al.: ‘There is a democratic civil peace’ and ‘[t]he most reliable path to stable domestic peace in the long run is to democratize as much as possible’ (2001: 44). Our analysis indicates, however, that this path may not be smooth – some moves ‘up the ladder’ of contestation are more promising than others. Apparently, there are two critical thresholds where the risk of civil war decreases substantially: one between non-electoral and electoral regimes, the other between regimes with/without competitive elections.

Footnotes

Replication data

All analyses were conducted using Stata 14.2. The dataset and do-file for the empirical analysis in this article, along with the Online appendix, can be found at .

Acknowledgments

We thank the editor, the reviewers, and the participants of 46th Annual Meeting of the Danish Political Science Association (Vejle, October 2014), 56th Annual Convention of the International Studies Association (New Orleans, February 2015), and the CODE workshop on ‘Political Repression and Violence’ (Aarhus, March 2016) for valuable comments. We are particularly thankful to Alexander Taaning Grundholm and Jakob Tolstrup for scrupulous feedback and insightful suggestions.

Funding

This research was supported by grants from the Danish Council for Independent Research (0602-02208B) and Innovation Fund Denmark (4110-00002B).

ORCID iD

Henrikas Bartusevičius

Notes

References

Ari

Baris

Gleditsch

Kristian

Hegre

Håvard

Wig

Tore

(2016) Democratization and civil conflict. APSA CD 14(2): 4–7.

Auvinen

Juha

(1997) Political conflict in less developed countries. Journal of Peace Research 34(2): 177–195.

Bellin

Eva

(2004) The robustness of authoritarianism in the Middle East: Exceptionalism in comparative perspective. Comparative Politics 36(2): 139–157.

Boix

Carles

Svolik

Milan

(2013) The foundations of limited authoritarian government: Institutions, commitments, and power-sharing in dictatorships. Journal of Politics 75(2): 300–316.

Boix

Carles

Miller

Michael

Rosato

Sebastian

(2013) A complete data set of political regimes, 1800–2007. Comparative Political Studies 46(12): 1523–1554.

Bolt

Jutta

Zanden

Jan Luiten van

(2013) The first update of the Maddison Project: Re-estimating growth before 1820. Maddison-Project Working Paper WP-4 (http://www.ggdc.net/maddison/maddison-project/publications/wp4.pdf).

Brancati

Dawn

Snyder

Jack

(2013) Time to kill: The impact of election timing on post-conflict stability. Journal of Conflict Resolution 57(5): 822–853.

Buhaug

Halvard

(2006) Relative capability and rebel objective in civil war. Journal of Peace Research 43(6): 691–708.

Carter

David

Signorino

Curtis

(2010) Back to the future: Modeling time dependence in binary data. Political Analysis 18(3): 271–292.

10.

Cederman

Lars-Erik

Gleditsch

Kristian Skrede

Buhaug

Halvard

(2013) Inequality, Grievances, and Civil War. New York: Cambridge University Press.

11.

Cederman

Lars-Erik

Gleditsch

Kristian Skrede

Hug

Simon

(2013) Elections and ethnic civil war. Comparative Political Studies 46(3): 387–417.

12.

Cederman

Lars-Erik

Hug

Simon

Krebs

Lutz

(2010) Democratization and civil war: Empirical evidence. Journal of Peace Research 47(4): 377–394.

13.

Cheibub

José

Hays

Jude

(2017) Elections and violence in Africa. Political Science Research and Methods 5(1): 81–102.

14.

Cheibub

José

Gandhi

Jennifer

Vreeland

James

(2010) Democracy and dictatorship revisited. Public Choice 143(1): 67–101.

15.

Chenoweth

Erica

Stephan

Maria

(2011) Why Civil Resistance Works: The Strategic Logic of Nonviolent Conflict. New York: Columbia University Press.

16.

Coppedge

Michael

Alvarez

Angel

Maldonado

Claudia

(2008) Two persistent dimensions of democracy: Contestation and inclusiveness. Journal of Politics 70(3): 632–647.

17.

Dahl

Robert

(1971) Polyarchy: Participation and Opposition. New Haven, CT: Yale University Press.

18.

Davenport

Christian

(2007) State repression and political order. Annual Review of Political Science 10: 1–23.

19.

Diamond

Larry

(2002) Thinking about hybrid regimes. Journal of Democracy 13(2): 21–35.

20.

Dunning

Thad

(2011) Fighting and voting: Violent conflict and electoral politics. Journal of Conflict Resolution 55(3): 327–339.

21.

Durant

Clark

Weintraub

Michael

(2014) How to make democracy self-enforcing after civil war: Enabling credible yet adaptable elite pacts. Conflict Management and Peace Science 31(5): 521–540.

22.

Fearon

James

Laitin

David

(2003) Ethnicity, insurgency, and civil war. American Political Science Review 97(1): 75–90.

23.

Fjelde

Hanne

(2010) Generals, dictators and kings: Authoritarian regimes and civil conflict 1973–2004. Conflict Management and Peace Science 27(3): 195–218.

24.

Flores

Thomas Edward

Nooruddin

Irfan

(2012) The effect of elections on postconflict peace and reconstruction. Journal of Politics 74(2): 558–570.

25.

Gandhi

Jennifer

Przeworski

Adam

(2006) Cooperation, cooptation, and rebellion under dictatorships. Economics & Politics 18(1): 1–26.

26.

Gleditsch

Kristian Skrede

Ruggeri

Andrea

(2010) Political opportunity structures, democracy, and civil war. Journal of Peace Research 47(3): 299–310.

27.

Gleditsch

Kristian Skrede

Ward

Michael

(1997) Double take: A reexamination of democracy and autocracy in modern polities. Journal of Conflict Resolution 41(3): 361–383.

28.

Gleditsch

Nils Petter

Wallensteen

Peter

Eriksson

Mikael

Sollenberg

Margareta

Strand

Håvard

(2002) Armed conflict 1946–2001: A new dataset. Journal of Peace Research 39(5): 615–637.

29.

Goldstone

Jack

Bates

Robert

Epstein

David

Gurr

Ted

Lustik

Michael

Marshall

Monty

Ulfelder

Jay

Woodward

Mark

(2010) A global model for forecasting political instability. American Journal of Political Science 54(1): 190–208.

30.

Haber

Stephen

Menaldo

Victor

(2011) Do natural resources fuel authoritarianism? A reappraisal of the resource curse. American Political Science Review 105(1): 1–26.

31.

Harish

Little

Andrew

(2017) The political violence cycle. American Political Science Review 111(2): 237–255.

32.

Hegre

Håvard

(2014) Democracy and armed conflict. Journal of Peace Research 51(2): 159–172.

33.

Hegre

Håvard

Sambanis

Nicholas

(2006) Sensitivity analysis of empirical results on civil war onset. Journal of Conflict Resolution 50(4): 508–535.

34.

Hegre

Håvard

Ellingsen

Tanja

Gates

Scott

Gleditsch

Nils Petter

(2001) Toward a democratic civil peace? Democracy, political change, and civil war, 1816–1992. American Political Science Review 95(1): 33–48.

35.

Henderson

Errol

Singer

David

(2000) Civil war in the post-colonial world, 1946–92. Journal of Peace Research 37(3): 275–299.

36.

Hendrix

Cullen

(2010) Measuring state capacity: Theoretical and empirical implications for the study of civil conflict. Journal of Peace Research 47(3): 273–285.

37.

Howard

Marc

Roessler

Philip

(2006) Liberalizing electoral outcomes in competitive authoritarian regimes. American Journal of Political Science 50(2): 365–381.

38.

Huntington

Samuel

(1991) The Third Wave: Democratization in the Late Twentieth Century. Norman, OK: University of Oklahoma Press.

39.

King

Gary

Roberts

Margaret

(2015) How robust standard errors expose methodological problems they do not fix, and what to do about it. Political Analysis 23(2): 159–179.

40.

Knutsen

Carl Henrik

Nygård

Håvard Mokleiv

Wig

Tore

(2017) Autocratic elections: Stabilizing tool or force for change? World Politics 69(1): 98–143.

41.

Kuntz

Philipp

Thompson

Mark

(2009) More than just the final straw: Stolen elections as revolutionary triggers. Comparative Politics 41(3): 253–257.

42.

Little

Andrew

(2012) Elections, fraud, and election monitoring in the shadow of revolution. Quarterly Journal of Political Science 7(3): 249–283.

43.

Magaloni

Beatriz

(2008) Credible power sharing and the longevity of authoritarian rule. Comparative Political Studies 41(4/5): 715–741.

44.

Malesky

Edmund

Schuler

Paul

(2010) Nodding or needling: Analyzing delegate responsiveness in an authoritarian parliament. American Political Science Review 104(3): 482–502.

45.

Marshall

Monty

Gurr

Ted

Jaggers

Keith

(2016) Polity IV project: Dataset user’s manual (http://www.systemicpeace.org/inscr/p4manualv2016.pdf).

46.

Matanock

Aila

(2017) Bullets for ballots. International Security 41(4): 93–132.

47.

Mazzuca

Sebastian

(2010) Access to power versus exercise of power. Studies in Comparative International Development 45(3): 334–357.

48.

Møller

Jørgen

Skaaning

Svend-Erik

(2013) Democracy and Democratization in Comparative Perspective. London: Routledge.

49.

Muller

Edward

Weede

Erich

(1990) Cross-national variation in political violence: A rational action approach. Journal of Conflict Resolution 34(4): 624–651.

50.

Przeworski

Adam

Alvarez

Michael

Cheibub

Jose

Limongi

Fernando

(2000) Democracy and Development. Cambridge: Cambridge University Press.

51.

Ravallion

Martin

(2011) Mashup indices of development. World Bank Research Observer 27(1): 1–32.

52.

Reynal-Querol

Marta

(2002) Political systems, stability and civil wars. Defense and Peace Economics 13(6): 465–483.

53.

Ross

Michael

(2012) The Oil Curse: How Petroleum Wealth Shapes the Development of Nations. Princeton, NJ: Princeton University Press.

54.

Rørbæk

Lasse

(2016) Ethnic exclusion and civil resistance campaigns: Opting for nonviolent or violent tactics? Terrorism and Political Violence, DOI: 10.1080/09546553.2016.1233872.

55.

Salehyan

Idean

Linebarger

Christopher

(2015) Elections and social conflict in Africa, 1990–2009. Studies in Comparative International Development 50(1): 23–49.

56.

Sarkees

Meredith

Wayman

Frank

(2010) Resort to War: 1816–2007. Washington, DC: CQ.

57.

Schedler

Andreas

(2002) The menu of manipulation. Journal of Democracy 13(2): 36–50.

58.

Schumpeter

Joseph

(1976) Capitalism, Socialism and Democracy. New York: Harper & Bros.

59.

Shock

Kurt

(2005) Unarmed Insurrections: People Power Movements in Nondemocracies. Minneapolis, MN: University of Minnesota Press.

60.

Singer

David

(1987) Reconstructing the Correlates of War dataset on material capabilities of states, 1816–1985. International Interactions 14(2): 115–132.

61.

Skaaning

Svend-Erik

Gerring

John

Bartusevi

Henrikas

(2015) Lexical index of electoral democracy. Comparative Political Studies 48(12): 1491–1525.

62.

Strand

Håvard

(2007) Retreating from a Civil Democratic Peace? Revisiting the Relationship between Political Institutions and Civil War. Oslo: University of Oslo.

63.

Themnér

Lotta

Wallensteen

Peter

(2011) Armed conflict, 1946–2010. Journal of Peace Research 48(3): 525–536.

64.

Treier

Shawn

Jackman

Simon

(2008) Democracy as a latent variable. American Journal of Political Science 52(1): 201–217.

65.

Vreeland

James

(2008) The effect of political regime on civil war: Unpacking anocracy. Journal of Conflict Resolution 52(3): 402–425.

66.

Whitten-Woodring

Jenifer

Van Belle

Douglas

(2017) The Correlates of Media Freedom: An introduction of the Global Media Freedom Dataset. Political Science Research and Methods 5(1): 179–188.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB