Abstract
In this study, we argue that ethnic inclusiveness is an important democratic norm that fosters interstate peace. When two states are socialized into the notion of ethnic tolerance, they acquire the ability to reach cooperative arrangements in time of crisis. Based on cross-national time-series data analysis covering the period 1950–2001, we illustrate how two states that are inclusive of their politically relevant ethnic groups are less likely to experience interstate disputes than states that remain exclusive. This finding was robust, regardless of sample size, intensity of the dispute, model specification, or estimation method. Therefore, we believe in the existence of ethnic peace: ethnic inclusiveness represents an unambiguous force for democratic peace.
Introduction
Since the advent of modern democracy, its proponents have suggested that democracy might serve as a force for peace in a turbulent world. In the 18th century, Thomas Paine and Immanuel Kant both mused that the propagation of representative governments may result in greater peace. Nearly two centuries later, empirical social science began to support this argument (Babst, 1972), leading to a number of related empirical findings about democratic peace that still hold true today despite continuing criticism (e.g., James et al., 1999). For example, scholars have found that democracies are only more peaceful with each other, meaning the likelihood of interstate war is slim to none; however, democracies are not necessarily more peaceful in their relations with all countries (see Choi, 2016; Oneal and Russett, 2005; Park, 2013).
When scholars broadened their analytical focus from interstate wars to militarized interstate disputes (MIDs) that did not involve full-scale war (e.g., when one state threatens, displays, or uses force against another), they found that the democratic peace phenomenon held true. Since scholars have reached similar findings when analyzing interstate wars and MIDs, our theory section refers to them as international conflict, for the purposes of analytical simplicity; however, our empirical section distinguishes three distinctive measures: MIDs, fatal MIDs, and interstate wars. By and large, the empirical findings point toward Paine’s (1776/1997) and Kant’s (1795/2010) democratic peace, yet a theoretical consensus on how pairs of democracies actually create peace through non-violent, democratic norms has eluded scholars. 1
In this study, we contribute to the democratic peace literature through a special focus on ethnic inclusiveness as an important democratic norm. We define ethnic inclusiveness at the country level based on whether politically relevant ethnic groups have access to executive-level power by way of their representatives. The definition of ethnicity includes ethnolinguistic, ethnosomatic (or racial), and ethnoreligious groups. Although it is true that historically religious conflict has tended to be far more frequent and violent than ethnic conflict, our definition of ethnicity does not differentiate them since it embraces ethnoreligious groups as a conceptual component. Thus, we use the two terms, ethnic conflict and religious conflict, interchangeably in this study.
Ethnic groups become politically relevant as soon as there is a minimal degree of political mobilization or intentional political discrimination due to their ethnic backgrounds. Executive-level power is referred to as the political executive, such as presidential and government cabinet and executive offices, as well as the top ranks of national militaries and bureaucracies. The definition of ethnic inclusiveness does not necessarily mean equal treatment of all ethnic communities because it is limited to those groups that are politically significant in national politics. Once politically active, some groups may have more access to central-level state power via representatives than others. Put differently, ethnically inclusive governments are not dominated or monopolized by a single ethnic group, they do not discriminate against other minority groups, and they often have laws and institutions in place to help enforce and reinforce these important characteristics. Consequently, in ethnically inclusive countries, politically relevant ethnic groups have consistent access to the highest levels of state power. In contrast, ethnic exclusion occurs when an ethnic group’s members are excluded from service or representation in the executive branch of government (see Wimmer et al., 2009).
It should be noted that ethnic inclusiveness is largely hindered when ethnic groups are subjected to active, intentional, and targeted discrimination with the intent of excluding them from executive-level power and representation. Examples include African Americans until the civil rights movement and Guatemaltecan Indians until the end of the civil war. Power sharing is a good way to increase ethnic inclusiveness, as it is an arrangement that divides executive power among leaders who claim to represent particular ethnic groups. Power sharing can be either formal (e.g., Lebanon) or informal (Switzerland). 2 However, some form of inclusiveness may be achieved by accident. In states with a large number of ethnic groups, inclusiveness in the legislature may be achieved without explicit intentionality, but simply as a byproduct of election results (e.g., Sri Lanka and Papua New Guinea). Demography may matter in another way. Countries tend to deny ethnic groups that have transnational ethnic kin risk, on the grounds that their loyalty is questioned (Cederman et al., 2018; Weiner, 1971). In particular, countries are likely to exclude and possibly even discriminate against groups that are ethnically linked to states with which the host country has ongoing rivalries (Mylonas, 2012).
We have observed that while democracies may be more likely to mandate ethnic inclusiveness,
3
non-democracies often deliberately exclude certain minorities (e.g., Rohingya refugees in Myanmar). We therefore argue that pairs of countries that are internally ethnically inclusive are less likely to engage in external conflict with each other. This argument was inspired by Russett’s (1993: 133–134) recognition of the irony that the initial creation of democratic institutions may contribute to the explosion of ethnic conflicts . . . that does not mean, however, that the solution lies in less democracy. Rather, it likely lies in devising institutions, and nurturing norms and practices, of democratic government with respect for minority rights.
Along these lines, we consider ethnic inclusiveness to be a non-violent, democratic norm of state behavior that is exhibited when state leaders prefer peaceful coexistence between different groups, incorporate diverse ethnic groups into the political system, and create laws that guarantee such treatment and protection in the future.
We argue that as a non-violent, democratic norm, ethnic inclusiveness is a clear, viable force for dyadic peace (peaceful coexistence between two states). We tested this argument through a dyadic analysis of international conflict and found that more politically inclusive states were less prone to entering into a conflict with each other. Specifically, based on a cross-sectional, time-series data analysis covering the period 1950–2001, we report evidence that dyadic states that are inclusive of their politically relevant ethnic groups are less likely to experience international conflict. Some difficulties in identifying the causal effect of ethnic inclusiveness on international conflict arise from potential temporal dependence. To address this concern, we employed multiple statistical methods that are frequently used for causal inference in the literature of international conflict. The finding was robust and consistent, regardless of the estimation method. Accordingly, our empirical analysis leads us to believe that ethnic peace is an unambiguous force for non-violence in contemporary international politics.
How ethnic inclusiveness creates interstate peace
Although scholars consistently explore the connection between ethnic composition and internal or external state conflict (e.g., James, 2010; Woodwell, 2004), little has been done to integrate this work into the equally vast literature on democratic peace. When studies of democratic peace incorporate ethnic composition into their analysis, the role tends to be incidental and its effects on conflict are assumed to be distinct from democracy. Conversely, we make the case that ethnic power relations and democratic peace are intimately connected.
Ethnic inclusiveness as a democratic norm
In order for a democratic norm to be included in the democratic peace model, two selection criteria should be met. First, an appropriate measure of a democratic norm should reflect a self-imposed norm of state conflict behavior. In other words, norms of behavior should not be inferred from non-behavioral measures such as political institutions or stability, instead these norms should reflect the conflict behavior of national leaders. Second, the norm of the behavior being measured should be democratic in nature. Specifically, democratic norms should be assumed to exist when state leaders conduct their foreign policy in a just, fair, and equitable fashion. We think that these criteria can be captured by the degree of inclusiveness of politically relevant ethnic (religious, racial, or linguistic) groups because they reflect the self-imposed behavior of state leaders (James, 2010). What is so special about ethnic groups as opposed to other politically relevant groups? Our argument, which will be explored in some depth, is that compared with political incidents that do not involve ethnic tension, ethnically framed politics tend to be more salient, ethnically framed disputes more volatile, and ethnically framed conflicts more violent. Therefore, genuine democratic norms can be most clearly seen in a state’s treatment of its ethnic groups.
The ability of ethnic framing to make issues more salient, disputes more volatile, and conflicts more violent has been empirically demonstrated in a number of studies. The perception of real socio-cultural boundaries eliminates the option of recruiting members of other communities to a common cause, making all members of the other group potential targets (Juergensmeyer, 2003; Tilly, 2003: 21). In addition to affecting joint-interest sharing, socio-cultural boundaries also facilitate the development of negative stereotypes. Asal and Rethemeyer (2008) describe a process they term ‘othering’ whereby religious and ethnic groups construct outlandish, negative, and dehumanizing images of other religious and ethnic groups with which they have tensions and conflict. Similarly, Gressang (2001) reasons that religiously motivated violence is easier to commit because it is performed for a supernatural audience perceived to be in support of a cause. Thus, ethnic framing allows antagonized groups to kill more indiscriminately. Therefore, it stands to reason that ethnic framing also makes conflicts more violent than they otherwise would be. Kaufman (1996) demonstrates that, because other ethnic groups are easily demonized, conflicts involving clashes of ethnic identity more often result in extreme forms of violence. Midlarsky (2005) also argues that the ability of a state to perpetrate genocide depends on the extent to which the victimized ethnic group can be clearly portrayed as criminals or traitors.
Importantly, however, even disputes that are not intrinsically ethnic can be framed as if they were. Some political disputes are ethnic in nature, focusing on a divergence of religious or cultural practices, discrimination, or historical grudges, all of which are perceived to be irreconcilable. Other disputes are non-ethnic, focusing on political, economic, or material issues such as land or resources. Although this distinction is rather general, it is empirically supported by Sambanis’ (2001) work on the differing causes of ethnic and non-ethnic civil wars. That said, either of these broad types of dispute can be framed in ethnic or religious language. Group leaders perceive and experience the strategic benefits, in terms of recruitment and morale, of framing their cause in ethnic or religious terms (Hobsbawm, 1992). Regardless of how authentic an appeal to ethnicity may be, such appeals are likely to be used and are often likely to be successful in motivating a group to act against another (Smith, 2000: 57). Thus, there is an umbrella of ethnic relevance that covers a broad spectrum of conflicts, making them more salient, volatile, and violent.
Because ethnicity serves to make tensions and conflicts more severe, it follows that it is most difficult to develop democratic norms among and between different ethnic and religious groups. When a dispute arises over a given political, material, or ideological point of contention, there is, at the outset, the prospect of a non-violent resolution if both sides can cooperate and work toward a compromise. On the other hand, if the dispute is ethnically or religiously motivated or framed, there is less perceived justification for cooperation and less ground for compromise. Therefore, ethnic relations can serve as a good test for the development of democratic norms. It can thus be assumed that governments exhibiting intentional inclusivity of politically relevant ethnic groups abide genuinely by democratic norms and embrace other politically relevant groups. In other words, the presence or absence of democratic norms is most evident, meaningful, and beneficial in terms of ethnic relations.
Ethnic inclusiveness provides theoretical justification and support for the existing literature on democratic norms because government inclusiveness closely approximates the norms of bounded competition and contingent consent as originally explicated by Dixon (1994) and Schmitter and Karl (1991). In an inclusive political environment where compromise, cooperation, and non-violent conflict resolution are commonplace, political leaders of ethnic groups must have developed ways of peacefully competing with other ethnic groups and other politically relevant groups, in other words, bounded competition. For instance, even though India is a highly ethnically diverse country where nearly 20 distinct politically relevant ethnic groups like the Sikhs and Muslims inherently face a high degree of political competition for national representation, it also has a relatively inclusive government in which even the highest offices of government consistently reflect the country’s ethnic diversity. 4 Because Indian presidential candidates are chosen by party leaders and elected by an electoral college, the ethnic diversity of the presidency reflects the behavioral norms of state leaders. These norms are also reflected in India’s extensive constitutional guarantees of nondiscrimination and equality for different ethnic groups. Further, because group leaders in such countries have a reasonable expectation of access to high-level governmental power, these leaders are more willing to accept the rule of other politically relevant groups—contingent consent.
Conversely, an absence of these democratic norms results in governments that exclude a larger share of politically relevant ethnic and other groups. The case of Sudan, from its independence in 1956 until the separation of South Sudan in 2011, provides a useful example. Although Sudan had nearly as many politically relevant ethnic groups as India, their behavior toward each other was much less democratic and less peaceful. Deng (1995) makes a convincing case that the competition of ethnic and religious identities among Sudanese ethnic groups proved inimical to the repeated attempts toward democratic development in the Sudan. Undoubtedly, when compounded by the continual struggle over resources, these ethnic identities made Sudan’s internal politics more salient, its disputes more volatile, and its conflicts more violent. This was consistently reflected in the exclusionary behavior of Sudanese state leaders. The Arab-dominated government systematically excluded the southern, animist and Christian groups from the political process, 5 stifling any prospect for the development of bounded competition and contingent consent, and eventually resulting in the move—backed by the surrounding countries—for an independent South Sudan, achieved in 2011.
We reason that ethnic inclusiveness reduces dyadic conflict on the international level. If state leaders show a consistent preference for compromise and peaceful dispute resolution vis-à-vis their most contentious internal political groups—ethnic groups—these democratic preferences should be normalized to all arenas of potential disputes, including disputes with other politically relevant internal groups as well as disputes with other states. Simply put, peaceful conflict resolution is expected when two countries in crises mutually recognize ethnic inclusiveness as a democratic norm (see Dixon, 1994). Dyads with good records of political inclusiveness will exhibit a mutual recognition of their counterpart’s preference for peaceful dispute resolution, making interstate wars less likely. In this way, peace happens in much the same way as postulated by other normative explanations of democratic peace. On the other hand, when a country with high ethnic inclusiveness confronts a country that has a history of mistreating different ethnic groups, mutual trust and respect do not arise so that the risk of interstate war grows.
The picture of norm externalization on the international level becomes even clearer when considering the continuity between domestic and international ethnic relations. Because national borders are not drawn to encompass only one ethnic group or to encompass the entirety of any ethnic group, and because the territories of ethnic groups often overlap, ethnically relevant issues naturally extend across political borders (e.g., the Kurdish–Turkish conflict). This gives rise to irredentism and provides a strong link between internal and external conflict. Internal ethnic conflict moves to the interstate level for a number of reasons. State leaders may be motivated by the ethnic composition of their population to intervene in the ethnic conflict of a bordering state (Carment et al., 2006). A state may also be more likely to intervene if it has a history of ruling multiethnic territory (Salehyan, 2009).
This explains why contiguous states like Bosnia and Herzegovina and Serbia are more likely to go to war, particularly at their outset, than noncontiguous states (Mansfield and Snyder, 1995). The Bosnian War was known to have been fought along ethnic lines and was consequently fought more severely after the government of the Yugoslav republic of Bosnia-Herzegovina declared its independence from Yugoslavia in 1992. As the Bosniaks and Serbs had not developed democratic norms of dealing with each other even under Yugoslav rule due to their lack of bounded competition or contingent consent, the Bosnian Serbs felt the need to create their own state and the Republic of Serbia felt compelled to back them. This is the abridged story behind many interstate wars. The same pattern of incongruence between ethnic and national borders is apparent in a long list of wars like those over Kashmir, the conflicts in the Horn of Africa, the Uganda–Tanzania War, all of the Arab–Israeli wars, to name a few.
However, just as ethnic tension and conflict can straddle the domestic and international levels, so too can peace. If a state’s leaders have democratic norms through which the government consistently includes politically relevant ethnic groups, it is likely to have well-developed norms for dealing with groups outside of its borders. Bosnia and Herzegovina, for example, became significantly more inclusive of its ethnic groups following the Dayton Agreement in 1995 and has since fostered a relatively peaceful democracy where the Serbs, Bosniaks, and Croats can all expect a certain level of representation. However, there are also other reasons why peace has prevailed between Bosnia and its neighbors (specifically Serbia) that might offer more causal weight. A major reason, of course, stemmed from the international intervention to end the war. This brought about a significant presence of third-party peacekeepers, including NATO and, later, EU forces, which would have had a deterrent effect. A desire for EU accession may also be a contributing factor, but this brings us back to the traditional sources of democratic peace. While the Bosnia and Herzegovina example is fresh in our memory, Switzerland, for example, has long been inclusive of its significant ethnic German, Italian, and French minorities—for whom the thought of ethnic war with Germany, Italy, or France seems outlandish.
Given that many interstate conflicts are extensions of domestic ethnic tensions, state leaders are likely to develop perceptions of a potential opponent’s democratic or non-democratic norms. If State A’s leaders share an ethnic bond with a group that stands to benefit or suffer from the presence or lack of State B’s democratic norms, State A should have a better perception of the behavior of State B’s leaders. In other words, whereas conventional understandings of perceptions describe their formation as being a mostly logical process, acknowledging ethnicity allows us to see that the process more often involves a great deal of ethnicity-related issues.
The above discussion leads to our main hypothesis as follows:
H1: Interstate dyads exhibiting high internal ethnic inclusiveness are less likely to experience interstate conflict.
Research design
To empirically test the ethnic inclusiveness hypothesis, we collated three sets of samples during the period 1950–2001: (a) all possible dyad-years, (b) politically relevant dyad-years, and (c) autocratic dyad-years. These samples were analyzed in the context of a standard statistical model of democratic peace, specifically Oneal and Russett’s (2005) model. Since the democratic peace model has been widely replicated and proven to be highly robust, possible coding errors or faulty model building on our part were minimized. The data availability of the dependent variable and the main independent variable determined the time frame for analysis. We do not believe that contemporary developments would have changed the main findings of our study: an increasing number of countries have promoted ethnic inclusiveness, especially since the decolonization period, so the likelihood of international conflict should be reduced for the years subsequent to 2001, the last year of our analysis. This is in line with the visual presentation of Cederman et al. (2018: 1291) that depicts an increasing trend of ethnic inclusiveness during the period 1946–2013.
Our baseline model included ethnic inclusiveness plus seven control variables to predict the outbreak of international conflict; all independent variables were lagged one year to mitigate problems of reverse causality.
To verify the robustness of our findings, we employed three different dichotomous dependent variables that were collated from the Correlates of War dataset on interstate disputes, which reflect different intensities of conflict. The first dependent variable was dichotomized for the onset of a MID of any severity. A MID is here defined as ‘a set of interactions between or among states involving threats to use military force, displays of military force, or actual uses of military force’ (Gochman and Maoz 1984: 587). The second dependent variable was dichotomized for the onset of a fatal MID where at least one soldier is killed per dyad-year. It should be noted that the manner in which national leaders respond to military fatalities continues to stimulate scholarship, thus the onset of a fatal MID was implemented as an additional check for robustness. The third dependent variable was dichotomized for the onset of an interstate war, which is defined as ‘a) sustained combat, involving b) regular armed forces on both sides and c) 1,000 battle fatalities among all of the system members involved’ (Sarkees, 2000: 125).
The main independent variable, ethnic inclusiveness, was operationalized with the Ethnic Power Relations dataset. 6 Since the data includes an indicator that measures the proportion of the ‘excluded’ population in the total population that is ethno-politically relevant, we called it ethnic exclusiveness (as opposed to inclusiveness) and anticipated that as the level of this variable increased, so would the likelihood of international conflict. In this way, we could preserve the integrity of the original data, while at the same time testing our core theoretical argument surrounding ethnic peace. To correct the positive skew of the data, we followed Wimmer et al.’s (2009) approach by using its logged transformation, which is consistent with the assumption that increases in the share of the excluded population have a greater effect on the likelihood of conflict at lower levels of exclusiveness. The logged ethnic exclusiveness variable ranged from 0–4.58 with a mean value of 0.93. 7 As shown in Online Appendix 1, 8 we found 15 democratic countries whose ethnic exclusiveness was below half (2.29) of the maximum value. Those countries were expected to experience a lower level of international conflict, compared with countries that exhibited high levels of ethnic exclusiveness.
Consistent with the way that democratic peace researchers convert democracy from a monadic- to a dyadic level, the ethnic exclusiveness variable used the weak link measure (see Dixon, 1994). In each dyad, the state with the lower exclusiveness score was taken to be the stronger determinant of future interactions. Hence, the more exclusive that state was, the less it would be constrained from engaging in a dispute, and therefore the more conflictual the dyad would be.
The seven control variables were economic interdependence, joint membership in intergovernmental organizations, contiguity, major power, geographic distance, national capability ratio, and allies. The economic interdependence variable also assumed the weak link, relying on the less interdependent state to determine the likelihood of a dyadic dispute. Increased interdependence will lead to a decreased possibility of conflict. The economic interdependence measure was the lower trade-to-GDP ratio. The joint membership in intergovernmental organizations (IGOs) variable was measured by the total number of IGO memberships that both countries in the dyad shared. The more joint memberships in IGOs, the more constrained the two states would be from engaging in a MID, thus leading to a more peaceful dyad.
To see whether power politics espoused by realists explained leaders’ conflict behavior better than ethnic exclusiveness, five realist variables were included as confounding factors. While the contiguity and major power variables are likely to increase the risk of conflict, such variables as geographic distance, national capability ratio, and allies are projected to decrease it. Some suggest that when either neighboring states or a major power state are involved in a crisis, the likelihood of conflict increases. For this reason, conflict studies underscore the significance of these two factors due to their impact on conflict in international relations (Russett and Oneal, 2001). Major power and contiguity were operationalized as dichotomous variables for their occurrence. Since geographic proximity increases a state’s opportunities and willingness to seek aggressive military campaigns, the geographic distance variable was included in the analysis. The national capability ratio variable was introduced to control for power preponderance. It was anticipated that an asymmetric power relationship would reduce conflict, since strong states can achieve their political goals without resorting to force, while weak states are unlikely to challenge stronger ones. To operationalize this variable, we took the natural log of the ratio of higher to lower national capability on population, industry, and military strength. The ‘allies’ variable was incorporated to account for the argument that a military alliance creates a dampening influence on conflict.
The data analysis employed standard logistic regression with peace years correction (a.k.a. logit splines), generalized estimating equations (GEEs), and rare event logit. It should be noted that the use of standard logit with peace years correction follows Oneal and Russett’s (2005) study; GEEs and rare event logit were introduced to further verify the statistical robustness of the main findings. In fact, by incorporating years of peace and its three splines into their analysis, Oneal and Russett dealt with temporal dependence issues. The logit splines model also controlled for the number of the states in the international system. According to Raknerud and Hegre (1997), the probability of a dispute, for any given pair of non-relevant states, has declined over time as membership in the international system has expanded. The system size variable took into account the dramatic growth in the number of sovereign countries since World War II. The GEEs model was adjusted for first-order autoregressive correlation (AR1) within each unit, estimating statistical significance using Huber/White/Sandwich robust standard errors that took into account the clustering of dyadic data.
Empirical results
This section is divided into four sub-sections following the convention of democratic peace research programs: basic analysis, substantive effects, robustness tests, and testing democratic norms and institutions in the same model. Relying on all dyads, politically relevant dyads, and autocratic dyads, basic analysis offers insight into the effect of ethnic exclusiveness on the outbreak of MIDs, fatal MIDs, and wars. Substantive effects are provided to ensure that the basic analysis findings are not statistical artifacts. Additional robustness tests were implemented to make sure that the results reported were not sensitive to different estimation methods and model specification. The last sub-section looks at the effect of ethnic exclusiveness in the presence of institutional constraints in the same model.
Basic analysis
Table 1 shows the empirical results for our base models, 9 in which we tested the relationship between the main predictor, the percentage of politically relevant ethnic groups that were excluded from political power, and the outcome variables: the onset of MIDs, fatal MIDs, and wars. As shown in Model 1, the coefficient on the ethnic exclusiveness variable is significantly different from zero and its sign is in the expected direction. 10 This result substantiates our argument that all dyads exhibiting domestic government exclusiveness (inclusiveness) increase (decrease) their probability of the onset of MIDs. In a broad sense, it appears that democratic norms are an important framework for reducing the risk of interstate conflict. In addition to the exclusiveness variable, the model also includes a number of control variables. Lending support to existing research on liberal peace, we found that economic interdependence exerted a pacifying effect on the onset of MIDs. However, the effect of IGO membership appears to be counterintuitive, as it actually increased the likelihood of a dispute. This implies that nation-states may use IGOs as a way of advancing their national interests. 11 The hypotheses about the five realist variables were supported: while contiguity and major power were positively associated with the onset of MIDs, geographic distance, national capability ratio, and allies were linked negatively.
Ethnic exclusiveness and international conflict.
Note: Numbers in parentheses are robust standard errors adjusted for clustering on dyads.
IGOs = intergovernmental organizations.
p < 0.05; **p < 0.01; ***p < 0.001, one-tailed tests.
Models 2 and 3 replicate Model 1 after replacing the dependent variable all MIDs with fatal MIDs and wars, respectively. The ethnic exclusiveness variable remained significant, regardless of dependent variable. As government exclusiveness increased, the pair was more likely to engage in either fatal MIDs or wars.
Although the specification of models 1 to 3 is considered to be standard in the conflict literature, it is not immune from criticism: (a) contiguity may be conceptually related to geographic distance, (b) major power has no theoretical meaning aside from its reflecting a state’s frequent involvement in international conflict, and (c) system size has no theoretical justification for being included in a dyadic analysis. The omission of these three control variables from models 1 to 3 should assuage the above concerns; the results are reported in models 4 to 6. The streamlined model specification did not cause the ethnic exclusiveness variable to become insignificant; in other words, ethnic exclusiveness still matters.
Our analysis has so far relied upon all dyads; however, the use of politically relevant dyads has some advantages. Politically relevant dyads are considered to have highly destabilizing effects on the international system because they consist of either a major power or have states that share a common border, situations that often result in a heightened chance of conflict (Russett and Oneal, 2001: 102–103). The results for models 7 to 9 are based on politically relevant dyads. It appears that ethnic exclusiveness is still positively associated with interstate conflict, whether it is all MIDs, fatal MIDs, or wars.
How do we know that the driving force behind the main findings reported in Table 1 is the ethnic exclusiveness variable? Put differently, because many countries with inclusive minority policies are liberal democracies, there is a possibility that the latter may be accounting for the empirical pattern discovered in the analysis. To address this issue, we re-ran the models in Table 1 with a sample of autocracies that was created after dropping all democratic dyad-years from all possible dyad-years sample. As shown in Online Appendix 4, the re-estimated results are very similar to those reported in Table 1: ethnic exclusiveness plays an important role in increasing the likelihood of international conflict.
Substantive effects
In this type of statistical analysis, if a sample size is very large, small p-values can occur even though the difference between the true value of the parameter and the null hypothesized value is small. In other words, with a large number of observations, statistical significance does not necessarily indicate a statistically meaningful or important finding in any practical sense. With this concern in mind, the substantive effects of variables of interest required further examination. Table 2 reports the substantive effects for ethnic exclusiveness, economic interdependence, and IGO membership that appeared in models 1 to 3 in Table 1; the substantive effects for the five realist variables are not reported to save space. Table 2 shows evidence that compared with a typical dyad, dyads with more ethnic exclusion are more likely to experience a dispute. For example, according to Model 1, a change of one standard deviation of ethnic exclusiveness increases the likelihood of a militarized dispute by 19%, and two standard deviations by 40%. Economic interdependence still reduces the likelihood of a dispute, while joint membership in IGOs increases the likelihood of a dispute. A similar pattern for these three variables is reported in the rest of the substantive effects analysis. One intriguing finding is that the percent change in the likelihood of a fatal dispute or a war, on average, appears to be larger than that for all MIDs. This may imply that ethnic exclusion makes dyadic states more conflict-prone as the severity of conflict increases. All in all, the empirical results in Table 2 verify that the pacifying role of ethnic inclusiveness, in the contemporary era, is not likely to be a statistical artifact.
Substantive effects.
Note: 1The baseline values are as follows: mean for continuous variables, 0 for Allies, 0 for Contiguity, and 0 for Major Power.
IGOs = intergovernmental organizations.
Robustness tests
It is plausible that alternative statistical estimation techniques could cause the significance of ethnic exclusiveness to disappear if this factor is not robust. Table 3 therefore evaluates the robustness of the results reported for models 1 to 3 following implementation of two other advanced statistical estimations: GEEs and rare event logit. GEEs correct for first-order autocorrelation as well as for heteroskedasticity. Rare event logit models are introduced to address the issue of excessive zeros that are present in conflict data. Table 3 shows evidence that, irrespective of estimation method, the aggravating effect of the ethnic exclusiveness variable on international conflict is confirmed in a consistent manner.
Ethnic exclusiveness and international conflict: robustness tests.
Note: Numbers in parentheses are semi-robust standard errors adjusted for clustering on dyads with the GEE models: robust standard errors adjusted for clustering on dyds with the relogit models.
IGOs = intergovernmental organizations.
p < 0.05; **p < 0.01; ***p < 0.001, one-tailed tests.
Ethnic exclusiveness versus institutional constraints
Based on the Polity datasets, democratic peace studies emphasize various institutional factors, especially the degree of constraint imposed on the chief executive in identifying the main determinants of international conflict. Accordingly, the democracy variable was constructed to capture the effect of democratic institutions rather than norms. The variable is a continuous democracy score of the least democratic state in the dyad, ranging from ‒10 (most non-democratic) to 10 (most democratic). Given the institutional features incorporated into this measure, we called this measure “institutional constraints.” In this study, we have argued from a democratic norm perspective that ethnic inclusiveness within democratic dyads is an important contributor to peace. It would therefore be interesting to examine the effect of ethnic inclusiveness in the presence of institutional constraints in the same model. In doing so, we might learn whether both democratic norms and institutional variables have a pacifying effect on international conflict. It should be noted that our ethnic inclusiveness variable is empirically distinct from the institutional constraints variable—the traditional measure of democracy—as the correlation between the lowest ethnic inclusiveness and the lowest institutional constraints is only 0.07. Table 4 shows the results obtained using the three different estimation techniques and the three different dependent variable measures. It appears that, across models, both ethnic exclusiveness and institutional constraints achieve significance with the sign in the expected direction, implying that both democratic institutions and norms serve as causes of international conflict.
Ethnic exclusiveness and institutional constraints.
Note: Numbers in parentheses are robust standard errors adjusted for clustering on dyads.
IGOs = intergovernmental organizations.
p < 0.05; **p < 0.01; ***p < 0.001, one-tailed tests.
Following in the footsteps of the democratic peace literature that assesses either normative or institutional features of democracies, we explored the independent effect of ethnic exclusiveness. However, it was also worth examining whether an interaction term between ethnic exclusiveness and institutional constraints would contribute to a better fit of the democratic peace model. This examination compared a standard additive model with a multiplicative interaction model. By comparing the overall fit of the two competing models, we intended to determine whether adding an interaction term contributed enough additional information to assist in explaining the absence of conflict among democratic dyads. When we implemented two commonly used comparative statistics—an Akaike’s information criterion (AIC) test and a Bayesian information criterion (BIC) test—the results were found to be inconsistent. For example, the AIC value was 9645.992 for additive Model 1 (Table 4) and 9641.131 for its multiplicative interaction model (the estimates are not shown to save space). The smaller value of the multiplicative interaction model indicates that the multiplicative specification does a better job than the additive specification in explaining leaders’ conflict behavior. However, the BIC value was 9805.859 for additive Model 1 and 9811.656 for its multiplicative interaction model. Since the model with the lowest BIC value indicates the better result, the multiplicative specification was not deemed a desirable choice. Because these two comparative statistics failed to show the superiority of the multiplicative interaction in a consistent manner, we have not reported the coefficients and standard errors.
Conclusion
The democratic peace literature is in need of revitalization through more accurate empirical testing of its normative arguments. For democratic peace researchers who wish to assert that democratic peace models include a variable for measuring democratic norms and that ‘the norms themselves may be more important than any particular institutional structure or formal constitutional provision’ (Russett, 1993: 31; also see Russett and Oneal, 2001), the examination of ethnic power relations provides a constructive, empirical way to revitalize the topic. When conflicts are framed in an ethnic manner, they grow more severe. Likewise, ethnic inclusiveness is a good test of whether the norms of bounded competition and contingent consent are present. When a state fosters inclusivity and tolerance toward its own ethnic minorities, states with similar norms will perceive a similarity in conflict management techniques. Thus, we argue that it is the norms of inclusiveness that create peace between dyadic pairs. Furthermore, acknowledging ethnicity allows us to show how the formation of preferences can be imbued with the greater salience of ethnic sentiment. In these ways, we can offer further theoretical justification for norm-based approaches to democratic peace theory. However, there has been very little engagement between the ethnopolitics and democratic peace research communities because the ethnopolitics literature tends to be focused on domestic conflict processes.
Our results provide robust empirical support for this line of reasoning. Ethnic exclusiveness was a strong and consistent predictor of interstate conflict, even when controlling for other liberal and realist correlates of conflict. Since this finding was robust, regardless of the sample size, the intensity of a dispute, the model specification, and the estimation method, we believe in ethnic peace that is created through ethnic inclusiveness at the domestic and then consequently in international arenas.
As with the bulk of research on democratic peace, our results suggest that Paine (1776/1997), Kant (1795/2010) and others were correct to some degree in asserting the important relationship between democracy and peace. Although all of the Kantian peace variables in our model had some bearing on the absence of international conflict, our ethnic inclusiveness variable was the most robust. This suggests that there is something more essential about how a state treats its politically relevant ethnic groups than other measures of democracy like institutions, economic interdependence, or IGO membership. Ethnic inclusiveness indeed provides an effective way of explaining war aberrations between democracies such as Austria/Italy (after World War II), UK/Ireland, Argentina/ UK (Falklands), India/Pakistan, which critics of the democratic peace refer to as instances of violent conflict between formally democratic countries. Although proponents of democratic peace may regard those wars as exceptions, our analysis indicates that they were instances of conflict between countries with high versus low levels of ethnic inclusiveness (e.g., when a conflict occurred in 1947, 1965, and 1971, India was politically inclusive while Pakistan was highly exclusive). The fact that ethnic inclusiveness is a non-violent, democratic norm should be a hopeful message for proponents of democratic peace, because political inclusion is a relatively more attainable goal than other policy goals implied by institutional or economic arguments. Put simply, if states wish to have more peaceful relations abroad, they should begin by treating their own people at home more inclusively.
Our theory and empirical findings start from the proposition that ethnic inclusiveness is part of everyday perceptions of democracy, and provide important implications as they can help us better understand the democratic peace phenomenon from a more accurate normative perspective. The good news is that history appears to be on our side as ethnic inclusiveness has become more prevalent over time. Similarly, Gurr (2000), Cederman et al. (2017) and Cederman et al. (2018), show that as many new regimes have preferred inclusive to exclusive practices, especially since the decolonization period, ethnic warfare has declined since the mid-1990s. With the spread of ethnic inclusiveness, the world may have a golden opportunity to achieve perpetual peace. Yet, it is still early to assert that ethnic inclusiveness is present in all parts of the world. As Cederman et al. (2013) observe, some regions such as sub-Saharan Africa have greatly benefited from inclusive practices, while others have hardly joined the trend toward declining exclusion, as illustrated by the Middle East and North Africa. Furthermore, we have seen democratic states where ethnic inclusiveness is not promoted. If our theory and empirical findings hold true, those democratic countries are unlikely to contribute to peaceful coexistence. Lastly, it concerns us that the surge of populist nationalism in Western democracies, which increasingly puts a certain ethnic group first at the expense of other groups, and destabilizes the benign force of IGOs such as the European Union, may turn the recent wave of inclusion into a wave of exclusion (Cederman et al., 2018).
Our study focused on the normative character of ethnic inclusiveness, since the democratic peace literature falls short of explaining the role of democratic norms in dyadic peace. We looked into the importance of political inclusion: how leaders from different ethnic groups participate in the decision-making of central government. Yet, we also recognize that ethnic inclusiveness may be projected as a form of democratic institution if we shift our analytical focus from political participation to power-sharing regimes. By power sharing, political leaders may accommodate demands from different ethnic groups, which should lead to peaceful coexistence. We note that while the democratic peace literature is rather limited on ethnic inclusiveness, other studies observe that democratic polities have increased ethnic group inclusiveness and decreased overt discrimination throughout history (Gurr, 1993, 1994). In particular, literature on ethnic conflict has long recognized the importance of democratic institutions. Lijphart (1977) and Lehmbruch (1967), for example, argue that power-sharing governmental arrangements that guarantee some representation for ethnic groups—parliamentary, consociationalist, and so on—have a pacifying effect on ethnically diverse states. Similarly, Bormann et al. (2018) show that formal power-sharing institutions reduce the likelihood of ethnic conflict mostly through the power-sharing behavior that these institutions induce. The existing literature, however, has yet to emphasize power sharing as a key institutional factor of interstate peace. This is a promising research project that requires a full-length article and thus we leave it for future research.
As power sharing strengthens democratic institutions, state leaders tend to be more constrained in their decisions to go to war (see Gelpi and Griesdorf, 2001). In the same way that democratic norms can be imbued with the salience of ethnic sentiments, so too can democratic institutions. In the existing democratic peace literature, the constraints that are imposed on presidents and prime ministers are assumed to be procedural in nature. In reality, state leaders in ethnically inclusive governments are likely to have ethnic attachments that have an impact on their attitudes toward nearby states and also their conflict behavior. In a state like Ethiopia, for example, if the Eritrean minorities had achieved some representation in the executive branch of the Ethiopian People’s Revolutionary Democratic Front government in its formative years during the early 1990s, those leaders may have been more likely to constrain the decision to go to war with Eritrea in the late 1990s. But the Eritrean groups had difficulty gaining representation at any level of government throughout the 1990s, which further exacerbated the internal conflict that gave rise to the interstate war between Ethiopia and the new state of Eritrea. This is another avenue of research on domestic accommodation, which we were unable to explore in this study and thus leave for future research.
Supplemental Material
sj-pdf-1-ips-10.1177_0192512119884738 – Supplemental material for Democratic peace: Does ethnic inclusiveness reduce interstate conflict?
Supplemental material, sj-pdf-1-ips-10.1177_0192512119884738 for Democratic peace: Does ethnic inclusiveness reduce interstate conflict? by Seung-Whan Choi and Henry Noll in International Political Science Review
Footnotes
Acknowledgements
We would like to thank David Carment, Rebecca Clendenen, Anahit Gomtsian, Patrick James, and Keith Simonds for their constructive comments and suggestions.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Notes
Author biographies
References
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.
