Popular vs. elite democratic structures and international peace

Abstract

Structural theories of international peace among democratic regimes have relied on two distinct explanatory logics: democratic institutions may cause a state’s foreign policy to tend toward peace by exposing policymaking elites to pressure from ordinary citizens (the popular logic) or to pressure from other governmental agencies (the elite logic). These logics are often conflated in scholarly studies of war and peace, but we attempt to isolate the popular logic for empirical testing by developing a novel measure of institutionalized popular influence, the Institutional Democracy Index (IDI). Whereas previous usage of the Polity index to operationalize democratic structures has succeeded in testing the elite logic more than the popular logic, we use the IDI to analyze long-established democracies’ involvement in international conflict between 1961 and 2001. What we find are significant differences within the family of democratic regimes that point to a monadic structural explanation of peace: more popular democracies are less warlike with respect to all other regimes, not just other democracies. By capturing variance among democratic regimes in their structures of inclusion (especially formal rules pertaining to voter access, electoral formulae, and cameral structures), the IDI enables us to observe crucial differences between the conflict propensities of more popular and more elite types of democracy.

Keywords

democratic institutions democratic peace militarized conflict

Introduction

According to a view long associated with Immanuel Kant, republican states should wage war less often than monarchies because of popular influence over foreign-policy decisions. As Kant explained in 1795, ‘if […] the consent of the citizenry is required in order to determine whether or not there will be war, it is natural that they consider all its calamities before committing themselves to so risky a game’ (Kant, 1983: 113). Kant drew this conclusion by explicitly contrasting the interests of ‘the citizenry’ with those of princes. Previously, however, defenders of the United States Constitution of 1787 had warned against the dangers of popular influence in general and over foreign policy in particular (The Federalist, no. 64; see Pole, 2005: 344–345). The purpose of republican institutions, they contended, was less to channel or reflect the popular will than to ‘refine’ it through policymaking elites (The Federalist, no. 10; see Pole, 2005: 52).

The contrast between popular and elite agency over public affairs was essential to both these lines of argument, and has been prominent in recent democratic theory (e.g. McCormick, 2011; Maloy, 2013). It has been generally downplayed, however, in modern studies of the democratic peace. Within structural accounts of peace, as distinct from cultural or normative explanations, the logic of popular influence has often been conflated with a logic of ‘checks and balances’, or intra-elite constraints. Specifically, the commonly used Polity index does not measure institutionalized popular influence, and therefore the popular logic has rarely been tested empirically. Doing so requires new measures, which we introduce in an original dataset, the Institutional Democracy Index (IDI).

Institutionalized popular influence is an underlying factor which can vary significantly even across democracies (Lijphart, 1999). By focusing attention here, we build on an insight from previous studies that institutional variety among democracies is not epiphenomenal in determining interstate war and peace (Morgan & Campbell, 1991; Ireland & Gartner, 2001; Reiter & Tillman, 2002; Clark & Nordstrom, 2005; Kisangani & Pickering, 2009). Since our hypothesized causal mechanism involves stable institutional structures that facilitate popular influence over policy, we consult the literature on comparative democratic institutions in developing innovative measures to use instead of the Polity index. The Kantian logic implies not only that democracies will be more peaceful than non-democracies, but also that more popular democracies will wage war less often than more elite democracies (Caranti, 2006). Testing this logic on long-established democracies between 1961 and 2001, we find that higher levels of institutionalized popular influence are associated with significantly lower international conflict propensities, irrespective of the regime type of potential adversaries. In other words, after distinguishing between popular and elite democracies, we find evidence consistent with not only a dyadic but also a monadic structural explanation of peace.

Literature and theory

In analyzing the relationship between democratic political systems and international peace, we distinguish between states that do not (or rarely) enter war out of a peaceful disposition from states that merely enter peaceful relationships with certain partners. The former is the basis of unilaterally peaceful behavior (‘the monadic peace’) whereas the latter is the basis of peaceful conduct towards kindred states (‘the dyadic peace’). In contrast to the prevailing skepticism towards prospects for the former (e.g. Quackenbush & Rudy, 2009), we argue that a monadic peace may still be possible, depending on states’ domestic political structures. In this respect we depart from the literature’s increasing focus on a dyadic democratic peace, the role of settled borders, reverse causality¹ (e.g. Gibler & Tir, 2010), and normative factors including shared economic norms related to capitalist development (e.g. Mousseau, 2009).

Our effort to revive the monadic theory is informed by the strengths and weaknesses of previous scholarly studies of domestic structural factors conducive to peace.² The so-called Kantian peace has been analyzed in three phases, whereby (a) domestic political structure features alongside (b) membership in international organizations and (c) participation in international trade (Russett & Oneal, 2001). The first, regime-based component of this theory has proven to be the most resilient to critical scrutiny (Ward, Siverson & Cao, 2007). Yet popular influence and intra-elite checks often walk arm-in-arm in structural explanations (e.g. Reiter & Stam, 2002: 5–6). Using the Polity index, quantitative analyses have supported the dyadic explanation that democracies are not less warlike in general but are unlikely to make war on other democracies (e.g. Maoz & Russett, 1993; Dafoe, 2011).

Two criticisms of the democratic-regime component of the Kantian peace have, however, yielded important refinements to the theory. One is the leader-incentive approach (Bueno de Mesquita et al., 1999, 2003: ch. 6), whose primary concern is that the structural logic is inherently monadic and therefore at odds with the dyadic data (Bueno de Mesquita et al., 2003: 222–223). Cultural and normative factors may be invoked to explain democracies’ covert operations and imperial wars against non-democracies (pp. 221–222), but a simpler, purely structural explanation would revolve around two concepts, the selectorate and the winning coalition. On this account, the key structural feature separating democracies from autocracies involves the institutional incentives confronting leaders. Since decisionmakers in a democracy must maintain a larger winning coalition to retain power, they must be more discriminating about whom to fight and be more committed when they do fight. A dyadic explanation for the democratic peace arises from the assumption that, aware of the heightened prudence and commitment of democracies, leaders in democratic dyads seek non-military resolutions of disputes (Reiter & Stam, 2002).³ Using a subset of Polity data to operationalize their key independent variable, Bueno de Mesquita et al. (2003: 134–135) find that the size of a winning coalition can account not only for the dyadic democratic peace but also for democracies’ imperial wars against non-democracies (2003: 251–252).

A second type of response to the interpretive difficulties of the Kantian peace has been to retreat from the logic of popular influence to the logic of intra-elite constraints. An early example hypothesized that a ‘greater number of institutions’ involved in foreign policy decisions would mean more constraints and therefore less war (Morgan & Campbell, 1991: 191–192). This approach specifically advocated downplaying electoral institutions in favor of separation of powers and party competition (pp. 192–193). Subsequent studies of parliamentary democracies have considered whether partisan composition of a government (minority, majority, or coalition) affects how quickly it initiates conflict (Ireland & Gartner, 2001). With majority and coalition governments, the results have been mixed. One study found that increasing the number of parties in a ruling coalition has no significant effect on war and peace (Reiter & Tillman, 2002) while another found majority governments more likely than multiparty coalitions to initiate ‘diversionary’ military actions (Kisangani & Pickering, 2009). Other notable findings include a statistically significant negative association between a high degree of legislative control over foreign policy and involvement in military conflict (Reiter & Tillman, 2002; Clark & Nordstrom, 2005); greater congruence with the public’s foreign-policy preferences among governments with retrospective legislative oversight, significant press freedoms, and comprehensive freedom of information laws (Colaresi, 2012); and, with respect to electoral systems, a lower propensity for conflict in countries using proportional representation (PR) compared to those with single-member districts (SMD) (Leblang & Chan, 2003).

So far these empirical results have been more suggestive than definitive, but several important principles have emerged from the literature on democratic variation. First, a monadic peace is back on the agenda (see also MacMillan, 2003: 235): in all the studies mentioned above, certain institutional features have been hypothesized, and sometimes confirmed, to make democracies more peaceful with respect to all potential targets, not just other democracies. Second, precise measures of institutional variation among democracies have been constructed by departing from the Polity index.

Matching theory and empirics

We now argue that the logic of intra-elite checks, distinct from the logic of popular influence, is entrenched in the Polity index, which in turn has served as a crutch for dyadic structural explanations of the democratic peace. Polity has long been the staple source of regime measures for structural analyses (e.g. Maoz & Russett, 1993; Russett & Oneal, 2001; Reiter & Stam, 2002; Pickering & Kisangani, 2005; Prins & Daxecker, 2008). Even the leader-incentive approach, though presented as an alternative to prior structural analyses, has relied heavily on the same source of data (Bueno de Mesquita et al., 2003: 134–135). However, as a recent analysis has shown (Kennedy, 2009), testing the leader-incentive theory with structural data drawn from indexes other than Polity yields inconsistent and even contrary results. When another dyadic analysis based on Polity data (Rousseau et al., 1996) was replicated using the dichotomous Democracy–Dictatorship index (see Przeworski et al., 2000; Cheibub, Gandhi & Vreeland, 2010), statistical support for the dyadic peace virtually disappeared (Elkins, 2000).

Polity’s popularity as a measurement tool is understandable: it captures some of the deepest, most entrenched institutional structures of a regime, and it covers a wide range of democratic and non-democratic states over 200 years. The problem is that Polity’s measures (see Marshall & Jaggers, 2010) operationalize separation of powers and checks and balances more than inclusion and popular influence. The executive constraint variable in particular tends to exercise an overweening influence on a regime’s overall Polity score (Gleditsch & Ward, 1997). Indeed, one of the most common criticisms of the Polity index is that it fails to reflect significant variations among democratic regimes in their degree of inclusiveness (e.g. Coppedge, Alvarez & Maldonado, 2008: 645). Given that the original Kantian version of the structural explanation was based on the notion that those ordinary citizens who bear the costs of war also have a say in whether to go to war, the widespread use of Polity data means that the logic of popular influence has been insufficiently tested in previous statistical analyses of the democratic peace. The same holds true even for studies using other measures of democratic structure that, like Polity, choose to operationalize checks and balances among competing agencies of government at the expense of mechanisms of popular pressure (e.g. Morgan & Campbell, 1991).

We endorse the intuition that the logic of popular influence provides a potentially more coherent structural explanation for peace than the logic of intra-elite checks (Mansfield & Snyder, 2005: 24). Because of the Polity index’s distinctive character as a measure of intra-elite checks, the monadic theory of popular influence cannot be ruled out by previous studies. The fact that alternative measures of democratic institutions have recently found evidence of a monadic peace encourages us to probe the issue on a broader scale. We therefore return to Kant’s original theory: institutions that facilitate popular pressure on elite policymakers tend to constrain a state’s war-making activities.

Theoretical expectations

Before we can derive testable hypotheses from this logic, we must notice that it articulates a tendency toward peace rather than predicting that ordinary citizens never want war. Consider warlike democracies such as Israel and the United States: do they contradict the popular logic? Kant argued that those who would suffer and sacrifice most should be reluctant to initiate war, not that they should never regard war as necessary or worthwhile. Thus the popular logic does not preclude democratic publics from ever responding to provocation in a warlike manner. When vital economic and security interests are perceived to be at stake, democracies would be expected to resort to self-defense just as readily as all other states. The popular logic is therefore consistent with democratic wars fought in the name of self-defense, or of citizens’ more or less expansive notions of other vital interests.

Beginning in 2001, for example, numerous democracies participated in major wars in Afghanistan and Iraq. A palpable sense of self-defense, and a feeling of having been provoked, loomed large in the public discourse leading up to those wars, especially in the United States. The subsequent decade-plus has witnessed a robust debate around the world about the interests that were at stake and the justifiability of the costs that were incurred. Recently the war-making activities of the United States have subsided somewhat, while those of other states have ceased altogether. Our primary argument is that, if the basic institutions of democratic regimes vary in their capacity for channeling popular influence over foreign policy, there should be corresponding variation in the initiation of war.

It would be a mistake, however, to assume that this monadic logic is somehow exclusive of or incompatible with the logic of dyadic peace; instead, it leads us to use the distinction between popular and elite democracies to refine dyadic expectations. The possibility of a dyadic peace for elite democracies and a monadic peace for popular democracies would imply that movement from an elite institutional structure to a popular institutional structure should affect conflict propensity differently for different classes of potential adversaries. If an elite democracy is much more likely to initiate a dispute against an autocratic state than against a democratic state, reforms that create a more popular institutional structure should have a relatively greater conflict-dampening effect vis-à-vis the most autocratic potential adversaries. In other words, by focusing on the structure of democratic institutions as our key explanatory concept, we must account for the fact that its relationship with conflict will be conditional on the regime characteristics of potential adversaries.

In short, the logic of our argument encompasses not only a broad, monadic peace for popular democracies, but also a more limited, dyadic peace for elite democracies. Our theory, which relies on the extent of popular institutions as the primary explanatory concept, therefore calls for hypotheses that are formulated in conditional and dyadic terms, with pairwise relations between an initiator and a target of conflict. Thus the difference in the likelihood of conflict initiation between popular and elite democracies should be greater when the target is a non-democracy as opposed to another democracy, and the difference in the likelihood of being targeted between democracies and non-democracies should be greater when the initiator is an elite rather than popular democracy. In accordance with best practice when specifying an interactive hypothesis (Berry, Golder & Milton, 2012), we explain the conditional relationship with respect to both constituent concepts: institutionalized popular influence in potential initiators and overall level of democracy in potential targets.

Hypothesis 1: To the extent that a democracy’s institutional structures reflect more popular than elite influence over policy, it should be less likely to initiate a militarized conflict against states of any regime type, but this association should be stronger when the potential conflict target is a non-democracy and weaker when it is a democracy.

Hypothesis 2: A democratic regime should lower a state’s likelihood of becoming the target of a militarized conflict initiated by another democracy, but this association should be stronger when the potential initiator is an elite democracy and weaker when the potential initiator is a popular democracy.

In summary, the impressive statistical record of the dyadic democratic peace overall (see Dafoe, 2011), stemming from heavy reliance on the Polity index, provides evidence that states with institutionally constrained executives tend to be more peaceful than those with unconstrained executives – but only with respect to similar states. To uncover whether democratic institutions can reduce conflict propensity with respect to all potential adversaries, we propose returning to the logic of institutionalized popular influence and developing new and more appropriate measures for that mechanism in the next section. Our more direct approach to testing the popular logic may help us to determine whether the failure of democracies as a family to avoid war in general is explicable by differences within the family, that is, between certain democracies that make more war and others that make less.

Research design

We use directed dyad-year data spanning 1961–2001 to examine the role of institutions conducive to popular influence on the initiation of dyadic militarized conflict. The directed dyad-year is appropriate as our unit of analysis because it allows us to examine the relationship between popular institutional structures and conflict initiation to make easy comparison with previous studies and, most importantly, to test whether a monadic peace for popular democracies coexists with a dyadic peace for democracies as a family. Accordingly, for our statistical analysis, we examine directed dyad-years where state A is the potential conflict initiator and state B is the potential conflict target. Our analysis covers those countries for which we have institutional democracy data as potential initiators (21 states), while all states are included as potential conflict targets.⁴ Although limiting the scope of our analysis, this restriction on the number of conflict initiators to include only long-established democracies is useful to mitigate potential sources of omitted-variable bias, particularly bias due to development and culture.⁵ In the supplemental appendix, we also specify models including dummy variables identifying the United States and Israel as potential initiators, given the former’s uniqueness as the world’s most powerful state and the latter’s unique position in the Middle East.⁶ Our lower temporal bound, 1961, is the first year for which we have (lagged) institutional democracy data, and our analysis ends in 2001, the last year for which conflict data are available.

To measure conflict, we use the militarized interstate disputes (MID) dyadic data version 3.1 (Ghosn, Palmer & Bremer, 2004). MIDs are defined as the threat, display, or use of militarized force by one state against another. We code two dependent variables to capture the initiation of a dyadic MID. First, we examine MID initiation in terms of the first actor in a new dyadic dispute. This dichotomous variable is equal to 1 in dyad years where state A is coded as on side A in a new dyadic MID against state B, signifying that it is on the side that took the first militarized action. Given that many MIDs are relatively minor threats or displays of force, we also examine use of force in order to isolate more serious militarized disputes. This variable is equal to 1 if state A is on side A of a MID in which force is used.⁷ Both of these dependent variables are dichotomous; accordingly, we use logistic regression in all models.⁸ Given the potential for duration dependence regarding dyadic conflict, we include a counter for years since last initiation, as well as squared and cubed terms of this variable (Carter & Signorino, 2010). We lag most explanatory variables to preclude simultaneity bias, noting the lags in our tables of results. Finally, given possible un-modeled heterogeneity by dyad, we include random intercepts by dyad.⁹

The institutional democracy index

Our primary explanatory variable measures the democratic structure of potential conflict initiators by focusing on institutionalized popular influence over foreign policy. The logic of popular influence which was adumbrated by Kant requires some elaboration before it can be translated into appropriate operational measures for empirical testing. A republican government reflects public reticence to undertake the burdens of war more than an autocratic regime, presumably because the former includes the views or interests of ordinary citizens to some degree. The logic of popular influence therefore operates through structures of inclusion. Given Polity’s shortcomings in distinguishing degrees of institutionalized inclusion, we need a more precise scheme for measuring democratic structure.

Since elections as mechanisms of consent can vary in degrees of inclusion, we look especially to institutional variations in voting procedures and electoral systems. Previous studies have examined impacts of electoral rules, electoral behavior, and partisan and coalition dynamics on international peace and conflict (Ireland & Gartner, 2001; Reiter & Tillman, 2002; Leblang & Chan, 2003; Clark & Nordstrom, 2005; Kisangani & Pickering, 2009), with special emphasis on indicators for voting participation (e.g. the number of voters who turn out in any given election) and partisan competition (e.g. the relative strengths of various parties in any given legislature or cabinet). Such dynamic constraints (Clark & Nordstrom, 2005) on the ability of political leaders to make war, however, are not fundamental and fixed rules of the political game: they may and do vary from one election to another. What we seek to measure, instead, is the deep structure of a political system. Instead of voter-turnout data for any given election, what we want to know is how the state defines an eligible voter and sets up the voting process; instead of partisan seat-shares in any given legislature, we need to know how the state sets up procedures for allocating seats in the legislature and for making decisions there.

Voting access

The first component of the IDI reflects key aspects of the legal framework shaping voters’ participation in elections. Popular access to elections is a crucial aspect of variation in structures of inclusion. Instead of measuring this through data on voter turnout (e.g. Reiter & Tillman, 2002; Leblang & Chan, 2003), our voting access (VotAcc) variable measures the basic procedures that structure the voting process. In addition to the standard of universal female and male adult suffrage used by most democracy indexes, two kinds of structures distinguish those democracies with the broadest electoral access: automatic voter registration and compulsory voting (Massicotte, Blais & Yoshinaka, 2004; Birch, 2009). In the presence of these institutions, politicians must confront public opinion in the broadest possible sense. Although compulsory voting does not broaden inclusion within an authoritarian context, under a multiparty democratic system it all but guarantees that all adults participate in the electoral process, including those who are not well educated, affluent, or from the ethnic majority (Birch, 2009). Moreover, both automatic registration and compulsory voting remove the variable of voter turnout from the control of party elites and their financial backers. In the absence of these two rules, politicians themselves have opportunities and incentives to spend campaign resources on choosing their own electorate, mobilizing presumptively friendly voters and demobilizing presumptively unfriendly voters.

As important as these institutions are to broadening democratic inclusion, automatic registration and compulsory voting have never been previously tested as part of a structural explanation of the democratic peace, and therefore we have created an original dataset, as discussed below, to measure them in our time-series sample. Our voting access indicator (VotAcc) starts with 0 as the baseline score for a country without universal suffrage. We then add one point to a regime’s voting access score, up to a maximum of 3, for each of the following rules: universal adult suffrage, automatic voter registration, and compulsory voting.

Legislative representation

Our second variable reflects the schemes of voting and districting by which legislative seats are allocated. The two general alternatives are proportional representation (PR) in multi-member districts and single-member districts (SMD) with plurality voting. If the structural logic of popular influence is to hold, we assume that policymakers must confront a broad spectrum of public opinion. We identify PR as more inclusive than SMD. With more parties to choose from in PR systems, voters are more likely to identify a party that is close to their policy preferences (Gerring & Thacker, 2008: 34), and empirical research has found that PR better reflects ideological and demographic diversity than SMD (Norris, 2008; Reynolds, 2011). PR systems also generally facilitate better government responsiveness to the average citizen in terms of social policy than SMD systems (Powell, 2000, 2006; Joshi, 2012, 2013). A pivotal difference between these systems is the number of seats assigned to each electoral district, or ‘district magnitude’ (m), with very high magnitudes approximating pure proportionality and affording greater inclusion while small magnitudes have the opposite effect (Taagepera & Shugart, 1989). SMD systems which have the smallest district magnitude (m = 1) usually generate two-party systems at the district level (Sridharan & Varshney, 2001) and lead to more vote–seat disproportionality than PR systems (Cox, 1997). Crucially for the logic of institutionalized influence, SMD offers political elites the same kind of opportunity to choose their own electorate as the absence of automatic registration and compulsory voting. Gerrymandering of single-member districts which permits party elites to insulate members from electoral competition and achieve disproportional shares of offices is all but impossible under large multi-member PR districts. PR systems, by comparison, leave politicians no choice but to confront a broader spectrum of opinion.

Additionally, we believe that a measure of legislative representation must include not only electoral formulae but also cameral structure. The prime consideration here is the logical axiom (widely understood since Condorcet) that multicameral decisionmaking, with veto players arrayed against the primary representative assembly within the legislative process, always magnifies the power of cohesive minorities, whereas unicameralism always comes closest to effecting majority rule within the legislature (McGann, 2006: 90, 182). In foreign policy, the existence of a second chamber may provide a structural basis for minority power that can, particularly when allied with some policy status quo, dilute the popular character of the legislature as a whole. For these reasons, bicameralism may in fact serve more as an obstacle (or ‘veto player’) than a conduit to popular influence. Though one study (Leblang & Chan, 2003) has already found PR democracies to be more prone to peace, to our knowledge the issue of unicameral vs. bicameral legislatures has not yet been tested with respect to international conflict.

For the legislative representation (LegRep) variable, our coding starts with 0 as the baseline score representing a plurality SMD system; a multi-round or run-off SMD is coded as 0.5; parallel non-compensatory mixtures of PR and SMD, and semi-proportional multi-member district systems (such as single non-transferable vote and two- and three-member districts) are coded as 1; medium-sized PR districts with less than eight members per district are coded as 1.5; and a score of 2 is assigned to states with large PR districts (averaging eight members or more) including compensatory mixed systems. Regardless of the electoral formula, we then add another point for a unicameral rather than a bicameral legislature, such that indicator scores vary between 0 and 3.

Constitutional structure

Given variation across regimes in the balance of foreign-policy power between executive and legislative authorities, we create a categorical variable to weight VotAcc and LegRep. Specifically, we distinguish the relative shares in foreign policy held by executive and legislative power, coding states as parliamentary, presidential, or mixed. In accordance with the Democracy–Dictatorship (DD) index (Cheibub, Gandhi & Vreeland, 2010: 79–82), we code any regime in which the chief executive’s survival in office depends on the legislature as parliamentary; where it does not, it is coded as presidential. Any regime in which there is both an independent president and a dependent prime minister, and each has some significant authority, is coded as mixed.

To create our aggregate IDI, we sum the VotAcc and LegRep variables, weighted by the relative strength of the executive and legislature. We expect executive influence, which we measure in terms of voting access (VotAcc), to be relatively more important in presidential regimes, and legislative influence (represented by LegRep) to be relatively more important in parliamentary regimes. Finally, VotAcc and LegRep are equally important in mixed regimes. Accordingly, we code our weighted IDI as follows:¹⁰

For presidential regimes,

I D I = (V o t A c c \times 3 / 2) + (L e g R e p \times 1 / 2);

For mixed regimes,

IDI = V o t A c c + L e g R e p;

For parliamentary regimes,

IDI = (V o t A c c \times 1/2) + (L e g R e p \times 3 / 2).

¹¹

This weighting formula is designed to account for the relative degree of legislative and executive influence, but as a robustness check we also report statistical results using an unweighted version of the IDI – where the unweighted version is calculated as equal to VotAcc + LegRep for all states – in the supplemental appendix. In both cases, the weighted and unweighted IDI varies between a minimum score of 0 and maximum score of 6.

To compile the IDI, we collected yearly data on democratic institutions for states that have been continuously democratic since at least 1950 as identified by Lijphart (1999: 50). Our dataset, which draws heavily on the data archives of the Inter-Parliamentary Union (2013), covers 21 countries over a 61-year period (1950–2010) resulting in 1,281 country-year observations.¹² These countries were chosen because our aim was to compare the role of institutional combinations across stable, consolidated democracies, those whom Kant expected to be most peaceful (Caranti, 2006: 343), rather than to compare democracies to non-democracies and semi-democracies. To illustrate variation in popular democracy, Figure 1 plots the values of the IDI over time for these states.¹³ It demonstrates that there is considerable variation in popular influence between states, while there is somewhat less variation within states over time.¹⁴

In summary, the IDI does not attempt merely to translate standard dichotomies of majoritarian vs. consensual or presidential vs. parliamentary democracy into numbers. Instead it takes the simple conceptual premise of institutionalized popular influence and operationalizes it with the addition of some unconventional measures that reflect crucial variation in structures of inclusion. For instance, by coding unicameralism (conventionally considered majoritarian) in the same direction as electoral proportionality (conventionally considered consensual), the IDI builds on recent trends in the study of comparative democratic institutions in terms of ‘centripetal’ (Gerring & Thacker, 2008) and egalitarian (McGann & Latner, 2013) characteristics, thereby applying new measures of popular government to the study of war and peace.

Primary explanatory variables

Our first explanatory variable is the IDI for the potential initiator of a dyadic MID. This measure varies from 0 (least popular – i.e. most elite) to 6 (most popular). Coding 0 as the least popular form of democracy is useful because we interact the initiator’s IDI with an indicator of the potential target’s level of democracy. Notably, we do not examine the popular or elite nature of target democratic institutions because to do so would limit our study to a total of 420 dyads that have experienced almost no conflict since World War II. Instead, we use existing measures of target democracy.

Specifically, to capture target democracy, we include the 21-point combined Polity score (Marshall & Jaggers, 2010), adjusted by the addition of 10 such that the minimum value (representing the most autocratic regime type) is equal to 0. In order to test our interactive hypotheses, we include a multiplicative interaction term for IDI x Polity. An interactive specification allows us to examine how the influence of popular (vs. elite) democratic institutions on conflict initiation varies depending on the extent of the target’s executive constraints and electoral competition – the factors comprising the Polity indicator, addressing Hypothesis 1. It also allows us to examine whether the impact of increasing target democracy is conditional on the elite or popular character of the initiator’s democratic institutions, addressing Hypothesis 2. We also present monadic (i.e. non-interactive) model specifications in the supplemental appendix.

Other explanatory variables

Our additional explanatory variables are chosen in order to prevent potentially confounding influence. In all models, we include a variable for relative capabilities within the dyad, defined as the potential initiator’s composite index of national capabilities (CINC) score divided by the sum of the initiator’s and target’s CINC scores (Singer, Bremer & Stuckey, 1972).¹⁵ To capture the opportunity of states to engage in conflict (and generally to interact), we include a variable for the (natural log of 1 plus the) distance between dyad members. To capture the complementarity of dyad members’ foreign policy preferences, and hence their willingness to engage in disputes, we include Signorino & Ritter’s (1999) weighted global S score, which measures the similarity of the states’ alliance portfolios.¹⁶

Figure 1.

Institutional Democracy Index scores over time

Our specifications also include two dummy variables to capture major epochs in international relations. The first takes the value of 1 for years during the Cold War (prior to 1992). The second takes the value of 1 for 2001, the first year of the Global War on Terror (and the last year in our data). Accordingly, the reference category is the 1992–2000 period. In additional models presented in the supplemental appendix, we add additional explanatory variables to capture trade dependence, development (in terms of gross domestic product per capita), and common membership in international governmental organizations (IGOs). All results are consistent in these models.

Analysis

Our statistical analysis suggests two primary conclusions. First, democratic institutions structured towards more popular influence are associated with a lower propensity of initiating a MID or use of force, an effect that is strongest when the potential target is most autocratic, diminishing as the target becomes more democratic. This finding supports Hypothesis 1. Second, there is a negative and significant relationship between higher levels of target democracy (in terms of institutional constraints) and the probability of MID initiation by an elite democracy. However, target democracy becomes less pacifying as initiator institutions increasingly promote popular influence. This finding supports Hypothesis 2. Taken together, our results demonstrate support for a monadic democratic peace for popular democracies, which are unlikely to initiate conflict regardless of target regime type. Conversely, elite democracies benefit only from a dyadic peace, given that their propensity to initiate a conflict diminishes as target democracy increases.

Table I presents the results of two random effects logit models.¹⁷ In Model 1, the dependent variable is MID initiation. In Model 2, the dependent variable is use of force. In both models, the coefficient for the potential initiator’s IDI is negative and significant (p ≤ 0.001). This result provides initial evidence that more popular democratic institutions are associated with a lower likelihood of initiating a MID or using force against a potential target that is among the most autocratic states (that is, its adjusted Polity score is equal to 0, the minimum value). The coefficient for the target’s adjusted Polity score is also negative and significant (p ≤ 0.001 in both models), suggesting that the most elite democracies (that is, states where the IDI is equal to 0) are less likely to initiate a MID against more democratic potential targets. The interaction term is positive but not significant in either model (p ≤ 0.305 averaged across the two models).¹⁸

An interpretation of conditional marginal effects provides further evidence in support of Hypotheses 1 and 2.¹⁹ First, the upper-left plot of Figure 2 illustrates the marginal effect of the IDI over the range of target Polity scores, holding all other variables at their medians. Dashed lines indicate the 95% confidence bounds of the estimates. The graph shows that the marginal effect of the initiator’s IDI on MID initiation is negative and significant over the entire range of target polity scores (which are indicated on the x-axis). However, the magnitude of the negative marginal effect is greatest when target Polity is equal to its minimum, diminishing towards 0 for more democratic targets.²⁰ In other words, when the potential target is very democratic, more popular influence for the initiator appears to have a less pacifying impact. However, this relatively smaller marginal effect is most likely due to the fact that longstanding democracies – regardless of their institutional makeup – are unlikely to fight each other. Importantly, the marginal effect of the initiator’s IDI remains statistically significant at the

Table I.

Random effects logit model coefficients and standard errors

	Model 1	Model 2
	DV: MID initiation	DV: Use of force
Initiator’s institutional democracy index _t– ₁	–0.536*** (0.089)	–0.567*** (0.112)
Target’s Polity score _t– ₁	–0.130*** (0.029)	–0.137*** (0.036)
Initiator IDI _t– ₁ X Target Polity _t– ₁	0.008 (0.008)	0.015 (0.010)
Relative capabilities (initiator/total) _t– ₁	–0.640 (0.378)	–0.608 (0.463)
ln Distance	–0.565*** (0.056)	–0.559*** (0.064)
Weighted global S score _t– ₁	–1.408*** (0.346)	–1.164** (0.421)
Years since dispute	–0.293*** (0.054)	–0.282*** (0.067)
Years since dispute²	0.013*** (0.003)	0.013** (0.004)
Years since dispute³	–0.000* (0.000)	–0.000* (0.000)
Cold War	–0.427* (0.216)	–0.282 (0.275)
War on terror	0.751* (0.372)	1.084* (0.444)
Constant	0.721 (0.595)	–0.410 (0.739)
Observations	116,527	116,527
Number of dyads	3,467	3,467
χ²	256.7***	161.8***
Log likelihood	–1,043	–737.1

***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05, two-tailed tests. Standard errors in parentheses. Models include random intercepts by dyad.

0.05 level over the entire range of target Polity scores. Importantly, the difference in the marginal effect from minimum to maximum target Polity is statistically significant (p ≤ 0.004). Accordingly, this plot provides support for Hypothesis 1.

The lower-left plot in Figure 2 presents the marginal effect of the target’s Polity score on MID initiation over the range of initiator IDI values (which are indicated on the x-axis). It shows a pattern similar to that discussed above; the marginal effect of target Polity is negative and significant when the initiator’s IDI is at its minimum, that is, when the potential initiator is among the most elite democracies (keeping in mind that all potential initiators are democracies). However, the marginal effect diminishes towards 0 when the potential initiator’s democratic institutions reflect greater popular influence. This supports our argument of a monadic popular-democratic peace; target democracy has a relatively smaller pacifying impact with respect to more popular democracies because popular democracies are unlikely to initiate conflict even against autocratic targets. In fact, the marginal effect of the target’s Polity score becomes statistically indistinguishable from 0 (at the 0.05 level) when the

Figure 2.

Conditional marginal effects of primary explanatory variables, with 95% confidence intervals

initiator’s IDI is higher than 4 (e.g. the value for the Federal Republic of Germany over the entire 1961–2001 period). Importantly, this change in the marginal effect of target Polity moving from minimum to maximum initiator IDI is statistically significant (p ≤ 0.014). This plot therefore provides support for Hypothesis 2.

The right-hand plots in Figure 2 illustrate the conditional marginal effects of initiator IDI and target Polity for our second dependent variable: use of force. The patterns in these plots mirror those examining MID initiation. However, in both of these plots, the marginal effects of initiator IDI and target Polity lose statistical significance at high values of the other constituent term.²¹ Given that the conditional marginal effects of our primary explanatory variables provide only limited information regarding the probability that a democracy initiates a militarized conflict, we turn to Figure 3, which graphs the substantive probabilities of MID initiation and use of force at different values of initiator IDI and target Polity.

An examination of substantive probabilities of MID initiation further confirms that popular democracies are less likely than elite democracies to initiate militarized disputes against any target regime type. Figure 3 highlights the probabilities of MID initiation for the most elite democracies (i.e. when the IDI score is held at the minimum, presented in the top row) and the most popular democracies (i.e. when the IDI score is held at the maximum, presented in the bottom row) over the entire range of target Polity scores, with all control variables held at their medians, and holding the random effect at 0. The left-hand plots present these probabilities for MID initiation (from Model 1), while the right-hand

Figure 3.

Probabilities of conflict initiation, with 95% confidence intervals

plots present probabilities for use of force (from Model 2). Notably, the probability that a popular democracy initiates a MID or use of force (presented in the bottom row) is less than the equivalent probability for elite democracies (presented in the top row) over the entire range of the x-axis, which represents the target’s Polity score. In fact, while the lines representing MID initiation and use of force by the most elite democracies are decreasing over the target’s Polity score, the lines for popular democracies appear to be flat and very close to 0. To illustrate, the probability of MID initiation against a target with the minimum Polity score (i.e. the most autocratic potential target) is equal to 0.001 for the most elite democracies, yet equal to 0.00003 for the most popular democracies (97% lower, a statistically significant difference, p ≤ 0.012). For targets with a maximum Polity score, the probability that an elite democracy initiates a MID is equal to 0.00005, whereas the probability that a popular democracy initiates a MID is equal to 0.000001 (a difference that is not statistically significant, p ≤ 0.092). Probabilities for use of force follow the same pattern, although they are slightly smaller in all cases because use of force is less common. Although the substantive probabilities that a popular democracy initiates a MID or use of force are technically decreasing over the range of target Polity scores, the change in probability from the most autocratic to the most democratic potential target is not statistically significant. In other words, our results suggest that popular democracies are similarly unlikely to initiate MIDs or uses of force against any target regime type. As such, we find evidence of a monadic democratic peace for popular democracies.

Importantly, the substantive probabilities discussed above are very small because we hold all control variables at their medians. Yet, the median dyad has experienced 18 years of peace and is separated by over 4,000 miles. Probabilities are considerably higher for states with fewer peace years or less distance. For example, when holding both peace years and distance at 0, the probability that the most elite democracy (IDI = 0) initiates a MID against the most autocratic target (Polity = –10) is equal to 0.34 (p ≤ 0.005), whereas the probability that the most popular democracy (IDI = 6) initiates a MID against the most autocratic target is equal to 0.04, but not statistically distinct from 0 (p ≤ 0.085). To illustrate the importance of underlying conflict propensity, the supplemental appendix presents substantive probabilities, and differences therein, for median dyads as well as those with considerably dissimilar foreign policy preferences (specifically, with S scores at the 5th percentile). This table shows that the interaction effect is robust with respect to MID onset regardless of control variable values, but with respect to the use of force, is statistically significant only for dyads with a higher underlying probability of conflict. We suspect that this latter finding follows from the fact that uses of force are particularly rare events. However, future research could benefit from exploring this conditionality in greater detail.

Conclusion

Results of our analysis support the hypothesis that popular democracies are less likely than elite democracies to initiate militarized disputes irrespective of the regime type of the potential conflict target. One implication may be that increasing the popular orientation of democratic institutions provides a peace dividend, particularly with respect to autocratic adversaries. Yet our quantitative analysis, while illuminating ceteris paribus tendencies within the diverse family of democratic states, cannot prove (nor did Kant himself make the claim) that democratic publics never want war; the results merely suggest that what democratic publics want matters more in some institutional frameworks than in others. Context matters and democratic institutions are far from the only factor in war and peace – as the relatively high conflict propensity of a unicameral and proportional regime such as Israel illustrates.²² For tighter causal stories, different modes of analysis can play a supplementary role to our statistical tests.

The major takeaway here is that democratic governments are not all created equal. For scholars, our results underscore the fact that attention to the conceptualization and measurement of democracy matters. Our institutional democracy index (IDI) offers a new method to evaluate some of the central tenets of democratic-peace theory, and specifically to distinguish between popular- and elite-based theories linking democracy to peace, a distinction that has often been muddled in previous studies that pair popular-democratic theory with elite-democratic indicators.

Our results also have important implications for policymakers. Our focus on institutional characteristics of democracy allows us to highlight specific structures that may influence the peacefulness of international relationships. While leaders cannot easily or quickly instill democratic norms in citizens, they can modify institutional aspects of democracy such as electoral rules and leadership structures. For the United States, this point is salient with regard to its own institutional structures, given its tendency to become involved in relatively frequent military conflicts. Similarly, our insights may be useful for US policymakers seeking to foster democracy in developing states, as well as domestic actors seeking political reform in those same states. If more popular democratic institutions promote sustainable peace, as our analysis suggests, that would be a significant consideration to weigh against the more elite institutional options.

Footnotes

Acknowledgments

A previous version of this article was presented at the 2013 meeting of the American Political Science Association. We thank the panel participants for helpful comments. We also thank Hilla Aurén and David Searcy for valuable research assistance, Michael Albert, Alan Gilbert, and Jacob Mauslein for constructive feedback on a previous version of this article, and the editor and anonymous JPR reviewers for excellent comments that considerably improved the manuscript.

Replication data

The online appendix, dataset, codebook, and code files for the empirical analysis in this article can be found at .

Notes

References

Chunrong

Norton

Edward C

(2003) Interaction terms in logit and probit models. Economic Letters 80 (1): 123–129.

Berry

William D

Golder

Matt

Milton

Daniel

(2012) Improving tests of theories positing interaction. Journal of Politics 74 (3): 653–671.

Birch

Sarah

(2009) Full Participation: A Comparative Study of Compulsory Voting. New York: United Nations University Press.

Brambor

Thomas

Clark

William R

Golder

Matt

(2006) Understanding interaction models: Improving empirical analysis. Political Analysis 14 (1): 63–82.

Bueno de Mesquita

Bruce

Smith

Alastair

Siverson

Randolph

Morrow

James

(1999) An institutional explanation of the democratic peace. American Political Science Review 93 (4): 791–807.

Bueno de Mesquita

Bruce

Smith

Alastair

Siverson

Randolph

Morrow

James

(2003) The Logic of Political Survival. Cambridge, MA: MIT Press.

Caranti

Luigi

(2006) Perpetual war for perpetual peace? Reflections on the realist critique of Kant’s project. Journal of Human Rights 5 (3): 341–353.

Carter

David B

Signorino

Curtis S

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

Cheibub

José Antonio

Gandhi

Jennifer

Vreeland

James R

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

10.

Clark

David

Nordstrom

Timothy

(2005) Democratic variants and democratic variance: How domestic constraints shape interstate conflict. Journal of Politics 67 (1): 250–270.

11.

Colaresi

Michael

(2012) A boom with review: How retrospective oversight increases the foreign policy ability of democracies. American Journal of Political Science 56 (3): 671–689.

12.

Coppedge

Michael

Alvarez

Angel

Maldonado

Claudia

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

13.

Cox

Gary

(1997) Making Votes Count: Strategic Coordination in the World’s Electoral Systems. New York: Cambridge University Press.

14.

Dafoe

Allan

(2011) Statistical critiques of the democratic peace: Caveat emptor. American Journal of Political Science 55 (2): 247–262.

15.

Elkins

Zachary

(2000) Gradations of democracy? Empirical tests of alternative conceptualizations. American Journal of Political Science 44 (2): 287–294.

16.

Gerring

John

Thacker

Strom C

(2008) A Centripetal Theory of Democratic Governance. New York: Cambridge University Press.

17.

Ghosn

Faten

Palmer

Glenn

Bremer

Stuart

(2004) The MID3 data set, 1993–2001: Procedures, coding rules, and description. Conflict Management and Peace Science 21 (2): 133–154.

18.

Gibler

Douglas

Tir

Jaroslav

(2010) Settled borders and regime type: Democratic transitions as consequences of peaceful territorial transfers. American Journal of Political Science 54 (4): 951–968.

19.

Gleditsch

Kristian S

Ward

Michael D

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

20.

Inter-Parliamentary Union (2013) Parline Database (www.ipu.org).

21.

Ireland

Michael

Gartner

Scott

(2001) Time to fight: Government type and conflict initiation in parliamentary systems. Journal of Conflict Resolution 45 (4): 547–568.

22.

Joshi

Devin

(2012) Social and adversarial varieties of democracy: Which produces fewer criminals? Development and Society 41 (2): 229–252.

23.

Joshi

Devin

(2013) The protective and developmental varieties of liberal democracy: A difference in kind or degree? Democratization 20 (2): 187–214.

24.

Kant

Immanuel

(1983) Perpetual Peace and Other Essays on Politics, History, and Moral Practice. Indianapolis, IN: Hackett.

25.

Kennedy

Ryan

(2009) Survival and accountability: An analysis of the empirical support for ‘selectorate theory’. International Studies Quarterly 53 (3): 695–714.

26.

King

Gary

Zeng

Langche

(2001) Explaining rare events in international relations. International Organization 55 (3): 693–715.

27.

Kisangani

Emizet

Pickering

Jeffrey

(2009) The dividends of diversion: Mature democracies’ proclivity to use diversionary force. British Journal of Political Science 39 (3): 483–515.

28.

Leblang

David

Chan

Steve

(2003) Explaining wars fought by established democracies: Do institutional constraints matter? Political Research Quarterly 56: 385–400.

29.

Leeds

Brett Ashley

Ritter

Jeffrey M

Mitchell

Sara McLaughlin

Long

Andrew G

(2002) Alliance treaty obligations and provisions, 1815–1944. International Interactions 28 (3): 237–260.

30.

Lijphart

Arend

(1999) Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries. New Haven, CT: Yale University Press.

31.

MacMillan

John

(2003) Beyond the separate democratic peace. Journal of Peace Research 40 (2): 233–243.

32.

Maloy

(2013) Democratic Statecraft: Political Realism and Popular Power. New York: Cambridge University Press.

33.

Mansfield

Edward D

Snyder

Jack L

(2005) Electing to Fight: Why Emerging Democracies Go to War. Cambridge, MA: MIT Press.

34.

Maoz

Zeev

Russett

Bruce

(1993) Normative and structural causes of democratic peace, 1946–1986. American Political Science Review 87 (3): 624–638.

35.

Marshall

Monty G

Jaggers

Keith

(2010) Polity IV project: Political regime characteristics and transitions, 1800–2009 (http://www.systemicpeace.org/polity/polity4.htm).

36.

Massicotte

Louis

Blais

André

Yoshinaka

Antoine

(2004) Establishing the Rules of the Game: Election Laws in Democracies. Toronto: University of Toronto Press.

37.

McCormick

John P

(2011) Machiavellian Democracy. New York: Cambridge University Press.

38.

McGann

Anthony

(2006) The Logic of Democracy: Reconciling Equality, Deliberation, and Minority Protection. Ann Arbor, MI: University of Michigan Press.

39.

McGann

Anthony

Latner

Michael

(2013) The calculus of consensus democracy: Rethinking patterns of democracy without veto players. Comparative Political Studies 46 (7): 823–850.

40.

Morgan

T Clifton

Campbell

Sally Howard

(1991) Domestic structure, decisional constraints, and war: So why Kant democracies fight? Journal of Conflict Resolution 35 (2): 187–211.

41.

Mousseau

Michael

(2009) The social market roots of democratic peace. International Security 33 (1): 52–86.

42.

Norris

Pippa

(2008) Driving Democracy: Do Power-Sharing Institutions Work? New York: Cambridge University Press.

43.

Peterson

Timothy M

Graham

Leah

(2011) Shared human rights norms and military conflict. Journal of Conflict Resolution 55 (2): 248–273.

44.

Pickering

Jeffrey

Kisangani

Emizet F

(2005) Democracy and diversionary military intervention: Reassessing regime type and the diversionary hypothesis. International Studies Quarterly 49 (1): 23–43.

45.

Pole

, ed. (2005) The Federalist. Indianapolis, IN: Hackett.

46.

Powell

G Bingham

(2000) Elections as Instruments of Democracy: Majoritarian and Proportional Visions. New Haven, CT: Yale University Press.

47.

Powell

G Bingham

(2006) Election laws and representative governments: Beyond votes and seats. British Journal of Political Science 36 (2): 291–315.

48.

Prins

Brandon

Daxecker

Ursula

(2008) Committed to peace: Liberal institutions and the termination of rivalry. British Journal of Political Science 38 (1): 17–43.

49.

Przeworski

Adam

Alvarez

Michael

Antonio Cheibub

José

Limongi

Fernando

(2000) Democracy and Development: Political Institutions and Well-Being in the World, 1950–1990. New York: Cambridge University Press.

50.

Quackenbush

Stephen

Rudy

Michael

(2009) Evaluating the monadic democratic peace. Conflict Management and Peace Science 26 (3): 268–285.

51.

Reiter

Dan

Stam

Allan

(2002) Democracies at War. Princeton, NJ: Princeton University Press.

52.

Reiter

Dan

Tillman

Erik

(2002) Public, legislative, and executive constraints on the democratic initiation of conflict. Journal of Politics 64 (3): 810–826.

53.

Reynolds

Andrew

(2011) Designing Democracy in a Dangerous World. New York: Oxford University Press.

54.

Rousseau

David

Gelpi

Christopher

Reiter

Dan

Huth

Paul

(1996) Assessing the dyadic nature of the democratic peace, 1918–1988. American Political Science Review 90 (3): 512–533.

55.

Russett

Bruce

Oneal

John

(2001) Triangulating Peace: Democracy, Interdependence, and International Organizations. New York: WW Norton.

56.

Signorino

Curtis S

Ritter

Jeffrey M

(1999) Tau-b or not tau-b: Measuring the similarity of foreign policy positions. International Studies Quarterly 43 (1): 115–144.

57.

Singer

J David

Bremer

Stuart

Stuckey

John

(1972) Capability distribution, uncertainty, and major power war, 1820–1965. In: Russett

Bruce

(ed.) Peace, War, and Numbers. Beverly Hills, CA: Sage, 19–48.

58.

Sridharan

Varshney

Ashutosh

(2001) Toward moderate pluralism: Political parties in India. In: Diamond

Larry

Gunther

Richard

(eds) Political Parties and Democracy. Baltimore, MD: Johns Hopkins University Press, 206–237.

59.

Taagepera

Rein

Shugart

Matthew

(1989) Seats and Votes: The Effects and Determinants of Electoral Systems. New Haven, CT: Yale University Press.

60.

Ward

Michael D

Siverson

Randolph M

Cao

Xun

(2007) Disputes, democracies, and dependencies: A reexamination of the Kantian peace. American Journal of Political Science 51 (3): 583–601.

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