Irredentism and Institutions

Abstract Why do states engage in irredentism? Expanding on previous scholarship, this article advances a new theory with rationalist microfoundations that accounts for the incentives of both elites and citizens to support irredentism in democracies and dictatorships. Our model suggests irredentism is more likely when it enables political elites to provide a specific mix of private goods, public goods, and welfare transfers to citizens who desire them at the lowest tax rate. This leads to the prediction that irredentism is most likely in majoritarian democratic electoral systems and military dictatorships, and least likely in proportional electoral systems and single-party dictatorships. We test and find supportive evidence for these expectations using a comprehensive dataset covering all observed and potential irredentist cases from 1946 to 2014.

much more common in majoritarian systems, where the demographically dominant ethnic group is significantly larger than other groups, and when the dominant group suffers perceptions of "status inconsistency" as a result of economic parity with other, smaller ethnic groups in the country (Siroky and Hale 2017). Although these explanations account for the political incentives of elites to annex ethnic kin in neighboring territory, they fail to clarify the incentives and costs that the masses consider when deciding whether to support irredentism. Our framework identifies the incentives to support irredentist initiatives not only of elites, but also of the masses, rather than simply assuming elites always dupe the public when it is expedient to do so.
We expand the scope of institutional influence on irredentism while relaxing the restrictive assumptions behind ethnic outbidding and mass support for nationalism. In doing so, the theory provides stronger microfoundations for the study of irredentism. After specifying the preferences of the central actors in our model, we explain how political institutions compel elites to translate those preferences into particular national tax and redistributive policies. These policies, we argue, influence the appeal of irredentism to citizens and to leaders. We formally derive actor utilities, and our model offers precise predictions about the conditions and configurations under which we should expect to observe irredentist behavior. To assess these expectations, we deploy global data covering all observed and potential irredentist events from 1946 to 2014 (Siroky and Hale 2017), matched to institutional data across the world for the same time period (Bormann and Golder 2013;Geddes, Wright, and Frantz 2014). The next section develops the argument and then introduces the model.

The Argument
Irredentism represents a continuum and has been discussed as everything from "soft" rhetorical territorial claims all the way up to military action. Here, we define it precisely as a state's use of military force against a neighboring state to annex ethnic kin and capture territory. Irredentism is a central government decision to add a region to its current state on the basis of ethnic similarity using military force and to simultaneously remove that region from the proprietorship of another country (Ambrosio 2001, 7;Gellner 1983, 1, 57;Horowitz 1991, 10;Neuberger 1991, 103;Petacco 1998;Siroky and Hale 2017, 117). This emphasis on military action, as opposed to rhetorical claims, allows for a more replicable, discernible measure of irredentism.
We theorize that political elites must balance their desire to maximize state revenue against the need to provide their political support base with some combination of private goods, welfare transfers, and public goods (hereafter, referred to collectively as "PTP goods"). While the annexation of foreign territory may offer an attractive option to increase revenue, it is equally clear that this benefit must be balanced against the expense of supplying PTP goods to any newly incorporated citizens, not to mention the risks and costs associated with war itself.
In our stylized framework, "private goods" consist of direct reallocations of excludable, rival resources from a political leader (or from the state coffers) directly to key supporters. These private goods are generally consumed by a very narrow cadre of beneficiaries and typically provide little value to the public at large. "Transfers," which we add to selectorate theory's framework with insights from Milesi-Ferretti, Perotti, and Rostagno (2002), are reallocations of rival, excludable resources that are targeted far more broadly than private goods toward individuals who are eligible on the basis of some shared, generally nonethnic characteristic (and irrespective of their location); examples include unemployment benefits, elderly medical benefits, and income-based welfare transfers. Finally, "public goods" are reallocations of nonrivalrous, nonexcludable resources toward particular geographic localities intended for consumption by the general public. Public goods include schools, fire departments, or road signs that benefit only a particular district and whose usefulness decreases as an individual is located farther from the public good's locus of distribution (Milesi-Ferretti, Perotti, and Rostagno 2002, 612-13).
Selectorate theory deploys only a single distinction between public goods and private goods. While this classification, as discussed later, can help us understand irredentist variation within dictatorships-that is, between military and single-party dictatorships-it is not well equipped to account for the differences between majoritarian and proportional democracies. Both possess large selectorates and large winning coalitions, rendering them observationally equivalent in terms of irredentism. 4 We argue that majoritarian and proportional systems incentivize different types of redistributive policies, with critical implications for irredentism. Distinguishing among these different types of redistributive policies (public goods from transfers) is thus essential to developing a comprehensive theory of irredentism that is equally predictive across and within both democratic and autocratic regime types.
Like political elites, citizens seek to enhance their own utility, which increases with the state's provision of PTP goods and decreases with the tax rate. This limits the PTP goods that a state can provide. Citizens also possess heterogeneous preferences for the provision of public goods, which are often exacerbated by ethnic and cultural differences (Alesina and Spolaore 1997;Desmet et al. 2011;Dzutsati 2021Dzutsati , 2022Hentschel 2019). All else equal, a higher degree of ethnic differentiation implies more heterogeneous preferences for public goods and makes it more costly for the state to provide them to citizens efficiently, requiring a higher tax rate. Crucially, political institutions shape how political elites implement such preferences in national policy, determining the equilibrium bundle of PTP goods for citizens. In this way, we suggest, institutions influence the cost-benefit calculus of pursuing irredentism.
Our framework distinguishes four main "ideal types" of institutional structures. 5 Among democracies, we differentiate proportional democracies from majoritarian systems; and among dictatorships, we examine single-party and military regimes. In doing so, we build on previous work, particularly studies on varieties of democracy and interstate conflict initiation (Maoz and Russett 1993;Morgan and Campbell 1991;Morrow et al. 2006;Pickering and Mitchell 2017, 4-5), the relationship between dictatorship and conflict initiation (Geddes, Wright, and Frantz 2014;Lai and Slater 2006;Peceny and Butler 2004;Pickering and Kisangani 2010;Pickering and Mitchell 2017, 5-6), and the impact of proportional and majoritarian systems on redistributive outcomes and ethnic conflict (Lijphart 2004;McGillivray 2004;Siroky and Hale 2017).
Our model predicts irredentism is most likely in majoritarian electoral systems and military dictatorships, and that it is least likely in single-party dictatorships and proportional electoral systems. Majoritarian systems encourage state leadership to provide more public goods than transfers (Milesi-Ferretti, Perotti, and Rostagno 2002, 610). Since incorporating ethnic kin shifts the median preference for public goods further toward that of the ethnic majority, the state can provide public goods to citizens that are closer to the preferences of the ethnic majority at a lower cost, which encourages irredentism. Proportional systems, on the other hand, are associated with greater state expenditures on transfers rather than on public goods (Milesi-Ferretti, Perotti, and Rostagno 2002, 610). Incorporating ethnic kin means they will have to share more valuable transfer benefits with a larger number of individuals, providing a check on the desire to incorporate them into a unified polity through irredentism.
In single-party dictatorships, a small party elite covetously safeguards its positions and understands that the incorporation of ethnic kin through annexation could produce new competitors, 4 Selectorate theory does not distinguish between majoritarian and proportional systems in detail beyond suggesting proportional systems may be predisposed to bloc voting, as well as smaller winning coalitions, in situations where there are strong patron-client relations (Bueno de Mesquita et al. 2003, 55, 64, 488, footnote 9). 5 Geddes, Wright, and Frantz (2014, 315) categorize 9.37 per cent of the autocracies coded in their dataset as military, 12.96 per cent as monarchies, 1.44 per cent as oligarchies, 32.54 per cent as single-party dictatorships, 25.09 per cent as personalist dictatorships, and 8.6 per cent as a combination of party/personalist dictatorships. Although they find it meaningful for their purposes to disentangle military dictatorships, monarchies, and personalist dictatorships, Geddes and colleagues acknowledge that all are characterized by the small selectorates and small winning coalitions. As far as our theoretical framework is concerned, these types are interchangeable. We therefore include each within the "military" category. for anyone could potentially be a member of the party. Annexation is thus threatening to the elite, which restrains irredentism. By comparison, in military dictatorships, where leadership consists of some proportion of the military officer corps (typically, a small part of the population), the threat of unwanted competition through annexation is substantially smaller, making irredentism more likely. In sum, utilizing this typology, we expect irredentism to be most likely in majoritarian democracies and military dictatorships, and less likely in single-party dictatorships and proportional democracies. The next section describes the full theoretical model and its core hypotheses.
Describing the Model Our model begins with a world where state borders emerge endogenously in distinct sizes and potentially separate different ethnic groups into a variety of majority-minority configurations (Cederman 2001). Since state borders do not match the geography of ethnic groups ("nations") one to one, opportunities for irredentism arise. 6 In order to assess our theory about the effects of institutions on incentives for (and against) irredentism, we introduce and systematically vary political institutions ("regime types") across states, which interact over time, yielding several testable implications that we subsequently evaluate using global data.
In our model, much like the world we observe today and historically, national borders include some ethnic kin but exclude others. In these situations, leaders must determine whether pursuing irredentism is in their interest. In developing the theory, we first focus on the incentives of leaders and citizens, and then introduce institutional constraints on irredentist action.

Preferences: Leaders and Citizens
Our theoretical framework, building on "selectorate theory," assumes leaders seek to remain in power and must provide the requisite goods and resources to a specific subset of citizens whose support is critical to remain in office. However, the particular mix of PTP goods provided to core supporters differs dramatically across different types of political institutions. The selectorate (S ) consists of the group of individuals in a country who have some degree of say in who becomes a leader. What is important about being in the selectorate is that it gives one an opportunity to be in the winning coalition (W ), which is defined as a subset of the selectorate whose support is essential for leadership to maintain office (Bueno de Mesquita et al. 2003, 51). Thus, when making decisions, state leaders focus on how such a decision will impact its core political support and, in our model, the median core supporter within W (Bueno de Mesquita et al. 2003, 51-5;Downs 1957;Levi 1989;Riker 1962). The reason is straightforward-political survival: if leaders fail to maximize the utility of median supporters in W, they risk being unseated by a political competitor. Citizens would prefer to receive lavish private goods from the state as a reward for loyalty to the regime. However, when citizens are not in a position to receive abundant private goods, or when private goods must be divided among too many supporters in W that they cease to be valuable (as in most consolidated democracies), then citizens prefer public goods (Bueno de Mesquita et al. 2003, 129-32;Milesi-Ferretti, Perotti, and Rostagno 2002).
We expand upon this well-established framework by incorporating preference heterogeneity among ethnic groups in the same country and by integrating welfare transfers into selectorate theory's distinction between public and private goods in order to investigate institutional 6 It is beyond the theoretical ambitions of this project to precisely model the mechanisms through which ethnic kin find themselves outside of national borders. We recognize, with much of the constructivist literature, that such identities are fluid and that ethnic identities themselves are often a byproduct of iterative interactions between individuals, groups, and the state. Our objective here is to establish a model world wherein individuals identifying with the dominant ethnic identity of a particular nation find themselves outside of their respective nation-state's borders-without reference to specific processes creating such a situation. incentives for and against irredentism. Our theory suggests that disparate preferences for transfers and public goods across institutional types play a significant role in explaining irredentism. Also critical to our model, individuals are members of ethnically defined nations, which are "relatively large and territorially concentrated ethnic group[s] with a sense of common history and putative homeland" (Hechter 2000, 14). Within the same nation, we assume there is more "commonality of tastes" among individuals for public goods than there is across distinct nations. Consistent with this notion, scholars have found that ethnically heterogeneous settings often encounter substantial difficulty effectively providing public goods (see, for example, Alesina and Spolaore 2003;Alesina, Baqir, and Easterly 1999;Dinesen, Schaeffer, and Sønderskov 2020;Habyarimana et al. 2007;Hechter 2000, 23;Horowitz 1985, 134;Miguel 2004;Miller and Page 2007, 256). 7 To stay in power and effectively provide the PTP goods desired by W, leaders seek to maximize state revenue, mostly through taxation but sometimes by foreign conquest (Alesina and Spolaore 2003, 71-2;Bolton, Roland, and Spolaore 1996;Wittman 1991). Although leaders can raise revenue by increasing taxes, 8 additional taxation generally decreases citizen productivity and displays diminishing returns. Moreover, if leaders fail to orient taxes to a rate that satisfies their core median supporter in W, they may be ousted through an election or a coup. Since citizens would like to receive generous PTP goods while paying as little as possible in taxes, leaders are left with the constrained optimization problem of providing these goods and services at an efficient tax rate.
However, ethnic heterogeneity makes efficient public goods provision more elusive. In order to placate the median member of the winning coalition in a diverse setting, public goods delivery from leadership often ends up at an "unhappy position in the middle" (Alesina, Baqir, andEasterly 1999, 1251-2), where no group feels it is receiving public goods in its preferred format. This can increase grievances, frustration, and polarization. For instance, ethnic groups often have disparate preferences over language issues, especially when they relate to educational and employment opportunities that shape the life chances of the group's members. 9 Incongruent preferences over the status of minority languages can set the stage for intense, intractable political disagreements that take on a conflictual and sometimes even violent form. In some settings (for example, recently in the Ukraine and Latvia), these kinds of disagreements and social cleavages have escalated to militarized disputes. Policing is another example of a public good over which different ethnic groups often have dissimilar preferences that has resulted in intense disagreements that sometimes lead to armed conflicts (for example, in Northern Kosovo and Ethiopia, [see Arriola 2013]).
How debates over such issues as the language of instruction in school and the language(s) used by the state bureaucracy are decided has an immense influence on people's lives, for individuals must invest considerable time and resources to learn a new language to gain formal employment. Minority groups have often mobilized for the right to be taught in their own language, and 7 In view of a large body of research that indicates individuals have both multiple and malleable ethnic identities (see, e.g., Hale 2004, 460-1;Lustick, Miodownik, and Eidelson 2004), we do not assume that either ethnic identities or preferences are permanently fixed, immutable, or even uniform within a particular ethnic group. Instead, we see these identities and preferences as malleable over the long term, though they tend to be quite "sticky," in that "attributes associated, or believed to be associated, with descent are … difficult to change in the short term" (Chandra 2006, 414). Attributes like language are particularly sticky (Chandra 2006, 414;Hale 2004, 468-9;Wucherpfennig et al. 2012, 85). However, we do not assume every member of an ethnic group responds to these constraints similarly. One of the great advantages of our computational model is that within ethnic groups, individuals internalize these constraints in different ways. Those with "thick" ethnic attributes will have stronger ethnic preferences. Although preferences vary systematically across ethnic groups, within groups, these impacts also vary. Furthermore, we assess the sensitivity of our model results by changing the degree to which preferences vary across groups systematically (see Table 11 and Figure 3 in the Online Appendix). In this way, the use of computational modeling allows us to incorporate both "primordialist" and "constructivist" assumptions into our analysis. 8 In our model, leaders can choose to raise or lower tax rates by incremental amounts at each stage in the model's evolution over time. 9 Alesina, Baqir, and Easterly (1999, 1251-2) convincingly illustrate the contentious nature of the language of instruction in diverse areas of the United States. majority ethnic groups have also engaged in collective action to prevent such changes to the status quo. Just as minorities perceive an injustice, majorities may resent paying for a public good from which they do not benefit.
These sorts of situations have the potential to increase the appeal of irredentism to political leaders for three primary reasons: (1) irredentism provides a mechanism for the state to increase the tax revenue of the country, particularly when the irredentist target is wealthy (described further later); (2) it results in more efficient public goods provision, as "the per capita costs for the taxpayer decline as the number of payers increases" (Alesina and La Ferrara 2005, 18); and, most important for our theory, (3) it shifts W's median preference for public goods closer to the general preferences of the ethnic majority. In this way, the state can provide public goods that more closely match the preferences of the demographic majority, thereby conciliating the perceived status inconsistency and maximizing the individual utilities of those in the winning coalition by providing public goods that cost less to a more homogeneous winning coalition. Since majoritarian systems prioritize public goods spending over transfers (Milesi-Ferretti, Perotti, and Rostagno 2002, 610), they will be particularly drawn to irredentism.
Political leaders also assess the economic wealth of the enclave relative to the potential irredentist state. Although wealth does not automatically attract irredentism, and poverty does not necessarily repel it, leaders are far less likely, on average, to pursue annexation of a poor coethnic enclave than a wealthy one (Horowitz 1985, 286). Finally, all else equal, irredentism is less attractive when neighbors have a greater ability to defend themselves.

Institutions: Majoritarian, Proportional, Single Party, and Military
Political institutions critically inform state leadership decisions by influencing the distribution of PTP goods to citizens and, in turn, by shaping the cost-benefit calculus of irredentism for leaders. Our theory and analysis focus on four "ideal-type" institutional arrangements: majoritarian democracy, proportional democracy, single-party dictatorship, and military dictatorship. This section explains the intuition connecting each institutional type to the provision of PTP goods and to the likelihood of irredentism.
In democracies, S is the total number of citizens with the right to vote, and W is the group of individuals whose support is necessary for the leader to stay in power. In an idealized majoritarian democracy with a single electoral district, the size of W is typically very large (51 per cent of the population-the size necessary to achieve an electoral majority). In a similarly idealized proportional democracy with one electoral district, individual legislators can be elected to office with less than an absolute majority but are powerless to enact policy without forming coalitions representing W, that is, a majority of the voting population. In both majoritarian and proportional democracies, then, the state is unlikely to supply private goods to W, as such resources would have to be divided up between far too many individuals for them to be valuable (Bueno de Mesquita et al. 2003, 91). Instead, the state is more likely to spend revenue on public goods and/or welfare transfers that provide greater value to individuals in W.
However, the particular citizen preferences to which leaders must be responsive depend on whether a democracy is majoritarian or proportional. In order to capture this distinction, we augment current theory, which focuses on public goods, by also considering welfare transfers. Whereas proportional systems incentivize politicians to channel government spending toward programs that increase the well-being of particular social constituencies across the country (welfare transfers), majoritarian systems tend to spend on programs and public goods that benefit specific localities (Milesi-Ferretti, Perotti, and Rostagno 2002, 609-10). 10 10 Milesi-Ferretti, Perotti, and Rostagno present a stylized model according to which national legislatures consist of three representatives, with the following four key features: (1) majoritarian systems consist of three electoral districts and elect one individual per district; (2) proportional systems consist of one national-level district and elect three individuals to that district; Irredentism is particularly attractive in majoritarian systems because they prioritize public goods over transfers. Adding ethnic kin with similar preferences for public goods-via irredentism-significantly benefits the material welfare of individuals within the ethnic majority by moving the preferences of the median citizen for public goods closer to the general preferences of individuals in the ethnic majority. This larger and more homogeneous polity can more efficiently cater to its winning coalition's preferences for public goods, as political leaders have greater incentives to disregard the preferences of minority ethnic groups. For their part, minority ethnic groups continue to pay taxes and thereby to underwrite the costs of public goods from which they derive progressively less benefit. 11 While proportional democratic systems also have large S and W, like majoritarian systems, proportional representation prioritizes social spending on welfare transfers over public goods (Milesi-Ferretti, Perotti, and Rostagno 2002, 609-10). Moreover, the benefits of welfare transfers are more broadly distributed across the population and are not subject to the same kind of ethnic preferences as public goods. W is less likely to want goods and services that correspond to ethnic differences, and the benefit of irredentism in terms of more efficient public goods provision to the ethnic majority is less pronounced. In proportional democracies, increasing the number of ethnic kin through annexation may still move the preferences of the median citizen for public goods closer to the general preferences of individuals in the ethnic majority in W, but because transfers are prioritized over public goods, the gain is marginal compared to majoritarian systems. Furthermore, if ethnic kin are added, highly prized transfers must now be shared with a larger number of individuals. Since proportional systems in democratic regimes provide less benefit from irredentism to the demographic majority, the model predicts much less irredentism among proportional democratic systems than majoritarian democracies.
In authoritarian systems, S and/or W are restricted much more substantially than they would typically be in any democratic electoral system. This makes it possible for members of W to be rewarded for their loyalty with considerable direct, private benefits from state largesse that are not made available to the rest of S and that are not typically viable in democratic electoral regimes. State leaders offer a mix of private and public goods that will solidify support and maximize the utility of W. As authoritarian leaders often tax citizens at a heavy rate to provide generous private payoffs to a small number of citizens, the state generally invests little in public goods or in transfers.
While most authoritarian regimes dole out private rewards generously to W, the size of W relative to S is extremely small in single-party dictatorships compared to military dictatorships. For example, it may be the case that all citizens have the right to vote but true decision-making power is wielded by a small number of people who are members of the party elite. In single-party dictatorships, individuals in W understand that there is a large pool of individuals that would readily take their place, as "practically anyone can be brought into the coalition and everyone is (3) citizens are members of one of three (nonethnic) social groups of unequal size, each of whom is eligible for a different type of transfer; and (4) the geographic (spatial) distribution of individuals belonging to these nonethnic social groups is assumed to be random across the country. Assuming the distribution of transfer groups is constant, all three representatives in majoritarian systems will advocate for the interests of the same largest transfer group. Since only one social transfer group is represented, transfers will not be politicized. Citizens vote for representatives who will advocate for public goods over transfers in order to direct government spending more toward their own district. The result is higher spending on public goods relative to transfers. In proportional systems, on the other hand, because more than one individual is elected in each district, more than one transfer social group is represented in the legislature. Transfers are politicized in this setting. Citizens vote for representatives who will increase their own individual share of transfer spending, which results in less spending on public goods relative to transfers. The authors test this theory with a time-series analysis spanning Organisation for Economic Co-operation and Development and Latin American countries and find supportive evidence. We offer interested readers a more detailed, technical discussion of this setup in Section 4b of the Online Appendix (pp. 9-12).

11
Our theory is broadly compatible with the idea of "contingent prize allocation" (Smith, Bueno de Mesquita, and LaGatta 2016). Related work examines the relationship between "selectorate" institutions and war aims (Morrow et al. 2006). However, we focus on a small subset of the type of conflict examined by this previous work and emphasize domestic redistributive consequences.
replaceable" (Bueno de Mesquita et al. 2003, 68). As a result, members of W covetously safeguard their positions. Since annexation is potentially threatening to W, irredentism should be much less likely in single-party authoritarian systems than in military dictatorships. By comparison, where the size of W relative to S is larger, such as in military dictatorships (where S is the officer corps of the military, a very small proportion of the population, but W is a majority of that officer corps), the threat of unwanted competition through annexation is substantially smaller. As a result, we predict, irredentism will be significantly more likely in military dictatorships than in single-party dictatorships. 12 In sum, the model predicts that two regime types-military dictatorships and majoritarian democracies-have a relatively high likelihood of irredentism, whereas proportional democracies and single-party dictatorships share relatively low odds of irredentism. Figure 1 summarizes the difference between the size of S and W across these four ideal-type regimes: majoritarian democracies, proportional democracies, single-party dictatorships, and military dictatorships. Each star represents one person, the darker oval represents S, and the lighter oval represents W. Figure 2 illustrates and summarizes the implications that the size of S and W have for the state's decisions to spend revenue on PTP goods and, in turn, for incentives to engage in irredentism. In military dictatorships and single-party dictatorships, relatively little revenue is delegated to public goods or transfers, for electoral considerations play little substantive role in policymaking decisions. Instead, the state invests heavily in providing private goods to core supporters. Irredentism is thus driven by the degree to which integrating new ethnic kin into the Fig. 1. Selectorate (dark oval) and winning coalition (light oval), by institutional type with each star representing a citizen Note: Majoritarian electoral systems and proportional electoral systems have similarly sized selectorates and winning coalitions. What explains variation in propensities for irredentism in these institutional forms is rather the varying way interests get translated into different propensities by their electoral institutions for the state to provide public goods versus transfers.

12
Political leadership aims to maximize the expected utility of the median member of its winning coalition and then weighs the likelihood of victory in a conflict by assessing its own resources against those of the host state. If it anticipates an acceptable likelihood of victory, political leadership attempts to annex the kin group militarily, with a probabilistic likelihood of victory based on the two countries' relative power capabilities. If it wins the conflict, political leadership annexes the territory into its state. state might potentially threaten core supporters' access to the private goods that the state provides them. As this threat is higher in single-party dictatorships, the likelihood of irredentism is lower in single-party dictatorships than in military dictatorships.
On the other hand, in democratic electoral systems, the size of S and W are large enough that the distribution of private goods is not politically beneficial to state leadership, and so leadership focuses on providing transfers and public goods to its population. On average, the electoral incentives of majoritarian electoral systems encourage the production of public goods relative to transfers, whereas proportional electoral systems encourage more transfers than public goods (Milesi-Ferretti, Perotti, and Rostagno 2002, 610). Since preferences for public goods are influenced by ethnicity (Alesina, Baqir, and Easterly 1999;Baldwin and Huber 2010, 645;Fernández and Levy 2008;Kimenyi 2006), incorporating ethnic kin into majoritarian electoral systems moves the median voter's preferences for public goods closer to the ethnic majority. This allows the state to provide public goods more efficiently at a lower tax rate and thus makes irredentism much more attractive. As proportional electoral systems invest less heavily in public goods, these incentives are not as pronounced, which leads to less irredentism compared to majoritarian systems. 13 In sum, the model predicts the following systems as being from most to least likely to engage in irredentism: (1) military dictatorships and majoritarian electoral systems, followed by (2) proportional electoral systems and single-party dictatorships. 14 A major implication is that there is nothing intrinsic to democracies inhibiting irredentism, nor anything to dictatorships inclining states toward it.

Empirical Validation
To assess our theoretical predictions, we utilized comprehensive data on each country's political institutions (Bormann and Golder 2013;Geddes, Wright, and Frantz 2014), focusing on four

13
The focus of our theory is on decision-making processes within the potential irredentist state itself rather than on the condition of the kin group within the host state. This is consistent with our observation that there are many unrequited instances of irredentism. Examples include the Toubou in Northern Chad [who did not wish to be retrieved by Libya], the Pathans in Pakistan, and the Croats in Austria [who are uninterested in joining Croatia]). While it is more attractive in general to be retrieved by a rich state, rich neighbors do not always entice and poor ones do not always deter (on this, see Horowitz 1985, 286). Our models account for the economic condition of the group in the enclave. 14 Full technical details of the computational model are in Sections 1 through 5 in the Online Appendix. main (mutually exclusive) types 15 : (1) majoritarian electoral systems (Majoritarian); (2) proportional electoral systems (Proportional); (3) dominant, single-party dictatorships (Single Party); and (4) both military and personalist dictatorships (Military). 16 We then matched this with data on the entire universe of potential and actual irredentist cases from 1946 to 2014 (Siroky and Hale 2017). 17 Table 1 presents descriptive statistics illustrating the number of irredentist conflicts by institutional regime type, which aligns with our theoretical expectations. Majoritarian democracies and military dictatorships exhibit higher levels of irredentist conflicts than the others, with 5.8 per cent and 4.5 per cent of potential opportunities resulting in irredentism, respectively. Meanwhile, proportional democracies (1.2 per cent) and single-party dictatorships (3.1 per cent) display considerably lower rates of irredentism. Now, we turn to the statistical models, which include additional factors that have been highlighted in the literature. The main model includes, first, the margin of the largest ethnic group in the irredentist state, as more homogeneous states are thought to be more likely to engage in irredentism (Margin) (Carment and James 1997;Horowitz 1985, 281-8;Lake and Rothschild 1998;Siroky and Hale 2017). 18 It also includes an indicator for whether the ethnic enclave is ethnically homogeneous or dispersed (heterogeneous) within the host state (Dispersed) (Horowitz 1985, 285;Moore and Davis 1998;Siroky and Hale 2017); whether the ethnic enclave is discriminated in the host state (Discriminated) (Davis, Jaggers, and Moore 1997;Horowitz 1985, 291;Moore and Davis 1998, 93-4;Saideman and Ayres 2008;Siroky and Hale 2017); and the wealth ratio in the host state relative to the irredentist state (Wealth Ratio). 19 Since dyadic regime characteristics have also often been emphasized in the literature, and because anocratic dyads appear particularly predisposed to interstate violence (Fearon and Laitin 2003;Gurr 2000;Mansfield and Snyder 2002a;Mansfield and Snyder 2002b;Muchlinski 2014;Salehyan and Gleditsch 2006;Siroky and Hale 2017), we include a set of indicators to capture whether the dyad is comprised of two anocratic regimes (Anocracy/Anocracy), an anocratic irredentist state and a nonanocratic host state (Anocracy/No), a nonanocratic irredentist state and an anocratic host state (No/Anocracy), or two nonanocratic states (No/Nothe omitted reference category). 20 Finally, the main models include five further measures: (1) the population of the host state (Host Population); (2) the population of the irredentist state 15 We model institutions as exogenous to the processes under study, while recognizing that unmodeled characteristics, particularly relating to ethnicity, might link the decision to adopt a particular electoral system to irredentism. The relatively rare nature of irredentism does not allow us to run fixed effects, as there is no variation on the dependent variable within some of our triads. Yet, we do employ regional variables in the Online Appendix (see Appendix Table 3), which account for many potentially confounding characteristics separating regions, such as Western Europe from Eastern Europe. We also explicitly model a variety of potentially relevant ethnic controls, such as the margin of the size of the largest ethnic group from the second-largest ethnic group, the degree to which a target ethnic group is dispersed in the host state, and so on. 16 We combined (Geddes,Wright, and Frantz 2014) military and personalist categories into a single category, as both systems are characterized by small selectorates and small winning coalitions. Due to a small number of cases, we eliminated their monarchy category from our primary analyses. In Online Appendix Table 4, we include the disaggregated and omitted categories, and their inclusion has little substantive impact on our results. 17 This dataset is organized using triads constituted by: (1) a potential irredentist state; (2) an enclave residing outside the irredentist state's borders whose ethnicity matches the group in power in the potential irredentist state; and (3) a host state in which that coethnic enclave resides. Each triad is observed on a yearly basis, making the unit of observation a triad-year. 18 Margin is the difference between the proportion of the largest ethnic group in the potential irredentist state and the proportion of the second-largest ethnic group in the potential irredentist state. 19 This is the annual ratio of the host state's gross domestic product (GDP) per capita divided by the GDP per capita of the potential irredentist state (Alesina and Spolaore 2003, 71-2;Alesina, Spolaore, and Wacziarg 2000;Bolton, Roland, and Spolaore 1996;Siroky and Hale 2017;Wittman 1991, 127). 20 Online Appendix Table 6 reports the model without anocratic dyads. Online Appendix Table 5a displays a nondyadic specification in which we include mutually exclusive indicators of democracy, autocracy, and anocracy for the irredentist state. Neither robustness check impacts our substantive results. Online Appendix Tables 7A through 7D provide tabulations to illustrate how our anocracy dyads map onto our four institutional designations of Majoritarian, Proportional, Single Party, and Military.

(Irredentist Population);
(3) the disparity in military capabilities between the host and irredentist states (Power Disparity), which is the natural log of the host state's Composite Index of National Capabilities (CINC) divided by the potential irredentist state's CINC; and (4) whether or not the potential irredentist (Irredentist Soviet) or (5) host (Host Soviet) states were formerly part of the Soviet Union. The Online Appendix provides a range of additional specifications and tests. 21 We estimated a logistic regression, with standard errors clustered by triad. To account for temporal dependence, we utilized cubic polynomial transformations from the number of peace years in each triad dating back to the previous irredentist conflict, if there was one, or, if not, to the first year of the dataset (Carter and Signorino 2010). 22 Table 2 displays the main results, and Figure 3 presents the predicted probabilities of irredentism for each of the four focal regime types, with 95 per cent confidence intervals. 23 Both proportional electoral systems and single-party systems are associated with a lower predicted probability of irredentism compared to systems with majoritarian and military institutions. The predicted probability of irredentism for a state with a proportional electoral system is only 0.019. Similarly, single-party dictatorships have a low predicted probability of engaging in irredentism of 0.025. Meanwhile, states with majoritarian electoral systems possess a much higher predicated probability of engaging in irredentism (0.050), and so do military dictatorships (0.052). 24 While these two institutional forms cannot be statistically differentiated from each other, they are statistically distinct from single-party dictatorships and proportional electoral systems.
This statistical evidence validates the computational model's main predictions that majoritarian democracies and military dictatorships are the most prone to irredentism, whereas singleparty dictatorships and proportional democracies are the least, underscoring the importance of These additional tests include regional controls and a control for crosscutting ethnic cleavages (linguistic and religious [see Online Appendix Table 8a]) that may be associated with moves toward both proportional representation and reduced propensities for irredentism (Selway 2011). We further account for a variety of ethnic characteristics in both the host state and enclave for a variety of reasons (see Online Appendix Table 9), including that the potential irredentist state may possibly annex other ethnic groups with its own ethnic kin during irredentism (Cederman, Weidmann, and Gleditsch 2011). We also provide robustness tests, such as rare events logistic regression (see Online Appendix Table 2), removing the anocracy and former Soviet controls (see Online Appendix Table 6), assessing a potential interaction effect between proportional systems and ethnic characteristics (see Online Appendix Table 8b), and adding a parliamentary/presidential system distinction (see Online Appendix Table 5b). As discussed and presented in the Online Appendix, none of these robustness tests have a significant impact on our main model results. In addition, we assess the performance of our model against the model presented in Siroky and Hale (2017) in Online Appendix Table 10. 22 In all cases, we begin counting the number of peace years from a triad's entry into the dataset. In most cases, this is either immediately after the Second World War or following national independence. 23 The full model results when excluding each of our four institutional types as a reference variable are in Online Appendix Table 1. 24 We note that Putin's Russia is coded as a personalistic dictatorship by Geddes, Wright, and Frantz (2014), which falls within our broader category of military dictatorships (see footnotes 5 and 16). focusing on specific institutions over the broader distinction between democracies and dictatorships.
Most of the control variables are not statistically significant, but anocratic dyads are worth noting. When both the irredentist and host states are anocratic, the probability of irredentism is dramatically increased relative to dyads where neither of the regimes is anocratic. We also see that irredentism is more likely when either the irredentist or the host state is anocratic, which is consistent with the finding in the literature that anocratic dyads are in general much more likely to engage in interstate conflict.
These results have crucial implications for our understanding of irredentism. Whereas much of the literature has posited that democracies are less likely to go to war for various reasons than autocracies, the current analysis suggests, at least with regard to irredentism, that this may not be accurate. Our results clearly demonstrate that majoritarian democracies are among the regimes most likely to engage in military conflict to annex ethnic kin. Similarly, while autocracies are often seen as being particularly likely to initiate interstate conflict, our theoretical model and accompanying empirical tests suggest single-party dictatorships are far less likely to initiate irredentist conflict than majoritarian democracies. Rather than simply assuming that dictatorships pursue irredentism more than democracies, or that citizens blindly allow themselves to be whipped into a nationalistic frenzy by political leaders, we specify the incentives of citizens to follow the ethnopopulist appeals of leaders and the incentives of leaders to pursue such policies. This reveals distinct institutional effects and establishes that the observed variation within democratic regime types and within autocratic regime types is just as noteworthy and consequential as disparity between regime types.

Conclusion
This article proposes a new institutional theory of irredentism to shed light on a fundamental problem of international security that has rarely been studied in a systematic and global manner. It endeavors to make three primary contributions. First, it provides stronger microfoundations for the political calculus of leaders and the economic interests of citizens across four distinct institutional regimes. Second, it develops a new computational model of irredentism that generates observable, empirical implications and enables us to model preference heterogeneity in a manner that is often intractable for formal mathematical models. This permits us to model irredentism as an emergent systemic process resulting from the microinteractions of citizens and leaders at The theory argues that state leaders are motivated by staying in power and maximizing revenue but are constrained in this endeavor by the need to provide social welfare and public goods to their constituents. Political leaders interested in maximizing the utility of their core constituents can provide more transfers and public goods by raising the tax rate, but this extraction lowers overall productivity and reduces the potential source of tax revenue, diminishing individual wealth and generally reducing the leadership's popularity. This can cause-and has caused-leaders to be ousted from power. On the other hand, leaders can lower taxes to help citizens retain their personal wealth, but doing so ultimately lessens the amount of revenue that they have to spend in the short term on PTP goods. Political leaders are wary of disrupting the optimal relationship between taxation and the provision of PTP goods for citizens. Any decision to annex territory and incorporate ethnic kin must therefore carefully consider the implications for their political survival. Leaders are more likely to engage in irredentism when doing so clearly increases the utility of the median core political supporter, which in our framework, depends upon the country's institutions.
Instead of assuming irredentist conflict should be less likely in democracies or democratic dyads, we theorize that different types of democracies and dictatorships possess distinct implications for irredentism. Our results demonstrate that there is as much heterogeneity within institutional types as there is across them; specifically, we show that majoritarian democracies and military dictatorships are most likely to engage in irredentism, whereas proportional democracies and single-party dictatorships are least likely. These results clearly challenge some of the conventional wisdom about how regime type is thought to shape international conflict behavior and show that both democracies and dictatorships have variants that are much more and much less predisposed to irredentism.
Our framework situates leaders and masses within these institutional settings, and imbues them with well-defined microfoundational incentives. Theories emphasizing the role of elites in driving ethnic conflicts can help scholars better understand the incentives of political leaders to mobilize masses for irredentist conflict, but "the insistent question of why the masses follow" (Horowitz 1985, 104) has yet to be directly connected to the calculus of leaders or to the constraints imposed by institutions. Most of the literature emphasizing the importance of ethnic outbidding has assumed leaders behave rationally but that the masses are mobilized based on emotional considerations, such as fear, dignity, resentment, or anger. Without denigrating the role of such emotions in helping leaders mobilize nationalist support, this study proposes a new set of microfoundations for both the elite and the masses. It affords an explanation for why, in some institutional contexts but not in others, the masses may find such ethnonationalist appeals particularly agreeable and why leaders may see it as in their interest to promote them. This approach advances the study of irredentism and opens new directions for future research by explicitly connecting microincentives and institutions to international political behavior.