The Political Cost of War Mobilization in Democracies and Dictatorships

Abstract

Recent research concludes fighting or losing an interstate war is not costlier for democratic leaders than dictators, which implies most of our institutional explanations for differences in conflict behavior across regime type rest on empirically tenuous assumptions. I argue military mobilization, a fundamental but often overlooked aspect of war, should be costlier for democrats than dictators. Waging interstate war is associated with higher military spending and, often, lower social spending. Variation across regime type in the representation of the general public, civilian elite, and military in leaders’ winning coalitions should make democrats more likely than dictators to lose power given wartime patterns of government spending. This argument finds support during the period from 1950 to 2001. My findings provide microfoundations for a number of existing empirical results and suggest that differences in the conflict behavior of democracies and dictatorships should be largest when waging war requires a significant mobilization effort.

Keywords

political survival interstate conflict domestic politics war regime type

Research on interstate conflict largely agrees that domestic politics significantly affect interstate conflict processes. Institutional explanations for differences in conflict behavior across regime type often assume the political costs of interstate war are higher in democracies than in autocracies (e.g., Russett and Oneal 2001; Reiter and Stam 2002). These theoretical arguments, at least implicitly, suggest that the political survival of a democratic leader is more sensitive to fighting and/or losing an interstate war than is the political survival of a nondemocratic leader (among others, Schultz 2001a; Bueno de Mesquita et al. 2003). Recent empirical scholarship, though, concludes that participating in an interstate war has limited influence on the survival of democrats or dictators and the outcome of an interstate war affects the political survival of autocratic leaders but not democratic incumbents (Chiozza and Goemans 2004; Debs and Goemans 2010). This implies interstate war is not costlier for democratic leaders than it is for dictators and, more importantly, that our explanations for variation in conflict behavior across regime type follow from questionable microfoundations. If at least some aspect of interstate war is not costlier for democrats than dictators, then our understanding of why democracies and nondemocracies differ in when they initiate, reciprocate, and terminate interstate conflicts and wars is flawed.

I argue a fundamental but largely overlooked characteristic of interstate war is costlier for democrats than it is for dictators. Fighting in an interstate war requires a government to increase military spending and, often, is accompanied with a decrease in nonmilitary spending (Sandler and Hartley 1995). The difference between peacetime and wartime military spending represents a government’s economic mobilization for interstate war (e.g., Bueno de Mesquita et al. 2003; Goldsmith 2007). Existing research on mobilization generally focuses on how it influences the escalation of an interstate dispute to war or a war’s outcome (among others, Organski and Kugler 1980; Slantchev 2005) and largely ignores its domestic political consequences. I argue mobilization should be more likely to result in democratic leaders being removed from power than nondemocratic leaders. This claim follows from the observation that the winning coalitions of democratic leaders largely consist of members of the public who prefer relatively lower military spending and higher social spending than the civilian elite and members of the military that largely make up the winning coalitions of nondemocratic leaders. This implies the spending patterns associated with mobilization for interstate war should be politically costlier for a democratic leader than a dictator. Consistent with this argument, I find mobilization for war increases the probability a democratic leader loses power to a greater degree than the probability a nondemocratic leader is removed from office.

The analyses presented here make multiple contributions to the literature on domestic politics and interstate conflict. Most importantly, they identify conditions under which interstate war is costlier for democratic leaders than it is for nondemocratic leaders. However, given the findings of other scholars, my results suggest the assumption that the political cost of fighting a war is necessarily higher in democracies is too strong. Instead, whether war is costlier for democrats than it is for dictators is conditional on the mobilization of economic resources. This implies differences in the conflict behavior of democracies and nondemocracies should be largest when fighting involves or is likely to involve significant mobilization efforts.

This article proceeds in six sections. The first describes research on the relationship between leader survival, interstate war, and regime type. The second section presents my argument for why mobilization for war should be costlier for democratic leaders than dictators. The third describes my empirical tests while the fourth and fifth report my results. The article concludes with a discussion of the larger implications of my findings.

Leader Survival, Interstate War, and Regime Type

Institutional explanations for variation in conflict behavior across regime type often assume that the political cost of war is greater in democracies than it is in nondemocracies. While these arguments differ in their specifics, the general logic is that variation in political accountability across regime type makes it more likely that a democratic leader will be punished by the public for leading her country into an interstate war or overseeing a losing war effort than a dictator who only needs the support of the political elite to stay in power (among others, Bueno de Mesquita et al. 1999; Russett and Oneal 2001; Reiter and Stam 2002). Scholars frequently cite Bueno de Mesquita and Siverson (1995) as empirical justification for these assumptions. However, as Debs and Goemans (2010) point out, Bueno de Mesquita and Siverson analyze the unconditional probability of leader survival across regime type, not the theoretically relevant quantity of whether democrats or dictators are more likely to be removed from office conditional on their participation in or the outcome of an interstate war.

Recent research concludes the empirical relationship between leader survival, interstate war, and regime type differs from the assumptions underlying most explanations of autocratic and democratic conflict behavior. Focusing first on war participation, Chiozza and Goemans (2004) find the probability of an incumbent retaining office is statistically unrelated to her country’s involvement in an interstate war. Distinguishing between how a leader is removed from power, Chiozza and Goemans (2011) find conflict participation does not alter the probability that a democratic leader will experience either a regular or a forcible removal from office. In contrast, autocratic leaders who initiate conflicts decrease their likelihood of both regular and irregular removal from power, as long as they do not subsequently lose an ensuing war.

Turning to war outcomes, Chiozza and Goemans (2004) and Debs and Goemans (2010) find the tenure of autocratic leaders is more sensitive to interstate war outcomes than is the political survival of democratic leaders. Their analyses indicate the probability a democratic incumbent will be removed from office is not significantly affected by her country’s performance in an interstate war. Nondemocratic incumbents, however, are rewarded with a lower probability of losing office upon winning an interstate war and are punished with a higher probability of removal from power after losing a war. Croco (2011) finds that culpable leaders—those leaders in power at the beginning of a war or who share a political connection with the leader in power at the beginning of a war—are more likely to be removed from office after losing a war in both democracies and nondemocracies.¹ Integrating insights from Croco (2011) and Weeks (2008, 2012), Croco and Weeks (n.d.) find that culpable and vulnerable leaders are more likely to lose office after losing a war than are nonculpable or nonvulnerable leaders regardless of regime type.²

A more nuanced relationship between war outcomes and political survival emerges if we consider the manner of leader removal. Chiozza and Goemans (2011) find that losing an interstate war increases the likelihood an autocrat suffers a regular or a forcible removal from office. They fail to find a significant relationship between interstate war outcomes and the likelihood a democratic leader is removed through peaceful means, but do find that leaders of parliamentary democracies face a higher probability of forcible removal upon losing a war. However, forcible removals are empirically rare in democracies: only 5.5 percent of democratic leaders between 1919 and 2003 lost office in an irregular manner (data from Debs and Goemans 2010). When considered together, then, recent empirical scholarship on the relationship between leader survival, interstate war, and regime type overwhelmingly indicates participating in or losing an interstate war is not systematically costlier for democrats than it is for dictators.³

Institutional explanations for variation in conflict behavior across regime type rely on the questionable assumptions that democratic leaders are more likely to be removed from power than autocratic leaders for participating in or losing an interstate war. If interstate war is not more costly for democratic leaders than it is for dictators, then much of what we think we know about the relationship between regime type and interstate conflict is wrong. The next section argues a common but largely overlooked feature of interstate war is costlier for democratic leaders than dictators: economic mobilization for war.

The Political Cost of War Mobilization

Governments increase the economic resources dedicated to the military when prosecuting an interstate war (Sandler and Hartley 1995). The increase in a government’s allocation of resources to the military associated with prosecuting an interstate war represents a state’s economic mobilization for war (Bueno de Mesquita et al. 2003; Goldsmith 2007). It is instructive to consider the degree to which states have mobilized for contemporary interstate wars. Between 1950 and 2001, countries fighting in an interstate war allocated, on average, 6.8 percent of their annual gross domestic product (GDP) to military spending, while countries at peace spent only 2.5 percent of their annual GDP on the military.⁴

Economic mobilization for war, or higher military spending in general, does not necessarily imply a complementary decrease in the proportion of a state’s economic resources allocated to nonmilitary spending. For example, governments can increase military spending without cutting nonmilitary expenditures by raising taxes, borrowing money, and/or printing money (Rockoff 2012). However, at least during the last sixty years fighting an interstate war has been associated with governments allocating fewer of their economic resources to nonmilitary purposes. On average, governments distributed 75.4 percent of expenditures to nonmilitary programs during peacetime but only 50.1 percent to nonmilitary spending when they were fighting an interstate war during the period from 1950 to 2001.⁵ This reduction in nonmilitary expenditures often includes cuts to social spending. From 1960 to 1999, governments allocated 16.2 percent less of their annual GDP to health care spending during an interstate war than they did during peacetime (1.6 percent vs. 1.9 percent per Kugler [2002]). Thus, the increase in military spending associated with war mobilization often is accompanied with a decrease in the proportion of expenditures dedicated to nonmilitary spending.

The political cost of mobilization refers to how the patterns of government spending associated with an interstate war effort affect the political survival of an incumbent leader. Explicit consideration of the domestic consequences of mobilization is almost nonexistent in the scholarly literature. I argue the political cost of war mobilization should be higher for democratic incumbents than for autocrats. This claim follows from a set of straightforward assumptions about how leaders allocate the economic resources available to them; variation in the composition of winning coalitions across regime type; the relative spending preferences of the general public, civilian elite, and members of the military; and the effect mobilization for interstate war has on patterns of government spending.

My argument begins with the common assumption that all incumbents rely on the support of a winning coalition to remain in power (Bueno de Mesquita et al. 2003). Following Bueno de Mesquita et al., I define an incumbent’s winning coalition as the subset of the population who have a say in choosing a government’s leadership whose support is necessary for her to retain office. I further assume that one of the principle ways an incumbent maintains the support of her winning coalition is by allocating economic resources to their preferred policies (Arena and Nicoletti 2014).

The above assumptions imply that an incumbent’s prospects of retaining office are linked to her winning coalition’s assessment of how she spends the resources available to her. Understanding how war mobilization influences a leader’s probability of survival requires that we identify the preferences of her winning coalition over how a government should distribute its economic resources. At this point, it is useful to divide a nation’s citizens between the elite, who consist of wealthy civilians and members of the military, and the more numerous and relatively poorer general public. The general public and elite systematically vary in their relative membership in autocratic and democratic winning coalitions and in their relative preferences for social and military spending. I focus first on their political influence across regime type.

The winning coalitions of democratic and nondemocratic incumbents differ in two fundamental ways. The first is that the winning coalitions of democratic leaders are larger than the winning coalitions of dictators (Bueno de Mesquita et al. 1999, 2003, 2004). I focus on the second difference. The membership of autocratic and democratic winning coalitions largely is drawn from different segments of a state’s population. Put simply, different types of people get into the winning coalitions of democrats and dictators. Dictators’ smaller winning coalitions are made up almost exclusively by members of the civilian elite and/or military (Bueno de Mesquita et al. 2003), with the relative influence of the civilian elite and military varying among nondemocratic regime types.⁶ In contrast, democratic political institutions induce relatively high levels of participation and contestation, which results in democratic leaders requiring the support of a large portion of the general public to remain in power (Dahl 1971; Boix 2003; Acemoglu and Robinson 2006). It therefore follows that democratic winning coalitions are composed of proportionately more members of the general public and fewer members of a society’s civilian and military elite than are autocratic winning coalitions.

Variation in who gets into a leader’s winning coalition across regime type implies that a democratic incumbent must be relatively more responsive to the preferences of the general public in order to retain office while a dictator must be relatively more responsive to the preferences of the elite. This insight is important for variation in the political cost of mobilization across regime type because members of the public and elite prefer different distributions of government spending. In general, compared to members of the elite, members of the public prefer their government allocate relatively more of its resources to social spending and fewer resources to military spending. This claim follows from four observations. The first two concern the preferences of the public and military over military spending.

First, military training socializes members of a state’s armed forces to value a stronger military and favor higher military spending than the civilian population (Nordlinger 1977; Geddes 2003). Second, while all citizens benefit from the military expenditures required to provide the public good of national security (Olson 1965), the public and members of the military have personal reasons to assess military spending beyond what is required for national security differently. Military spending over and above the level necessary to provide national security crowds out consumption spending popular among the public (Fordham and Walker 2005) but finances private benefits and club goods for members of the military (on the link between club goods and political survival, see Arena and Nicoletti 2014). These private and club benefits include, but are not limited to, the salaries of members of the military, access to goods and services at reduced price at base or post exchanges, and free or subsidized housing.

Research on civil–military relations supports the claim that the public and members of the military have different preferences over military spending. Bachman, Blair, and Segal (1977) find that, in general, members of the military prefer higher military spending than do members of the public. Addressing a potential guns versus butter trade-off, Holsti (1998) and Szayna et al. (2007) find that members of the public are more likely than members of the military to think that military spending should be decreased in order to increase education spending.

There are two reasons for the general public and civilian elite to have different preferences regarding social spending. First, the public derives more direct benefits from social spending than do the wealthy civilian elite, whom generally can provide themselves with the services that the public receives via the welfare state (e.g., health care and education). Notably, this is the case with all means-tested social welfare programs by definition. Second, spending on social programs typically is financed through taxes on the civilian elite (Przeworski et al. 2000; Boix 2003). Thus, the civilian elite bear the brunt of the costs of the social welfare state while deriving relatively fewer benefits than members of the public. It therefore follows that the general public would prefer a government allocate more of its resources to social spending than the civilian elite. This claim is consistent with the general, negative correlation between income and support for the welfare state in the United States and Europe (Cook and Barrett 1992; Jæger 2006) and the findings that the relatively affluent (Gilens 2009) and top 1 percent of earners (Page, Bartels, and Seawright 2013) prefer lower social welfare spending than the rest of the US public.

Variation in who gets into autocratic and democratic winning coalitions and their spending preferences imply that the political cost of mobilization for interstate war should be higher for a democratic incumbent than a dictator. This claim follows from how mobilization for war influences spending patterns and the distributions of spending that best secure the survival of democratic and nondemocratic incumbents. Economic mobilization for interstate war increases the resources allocated to the military and often is accompanied with a decrease in the proportion of resources dedicated to social spending. Leaders are less likely to be removed from power when they implement policies consistent with the preferences of their winning coalition. The spending preferences of the public, civilian elite, and military and their representation in leaders’ winning coalitions across regime type imply that democratic leaders are more likely to lose office given relatively high levels of military spending and low levels of social spending than are dictators. Accordingly, the patterns of government spending associated with mobilization for interstate war should be more likely to result in a democratic incumbent losing power than an autocratic leader.

It is worth highlighting two characteristics of my theoretical argument before moving forward. First, my argument assumes that some of the characteristics of “normal” domestic politics continue to operate during times of war. In general, an incumbent is more likely to remain in power if she allocates resources in a manner consistent with her constituents’ policy preferences. I assume that this dynamic continues to hold during an interstate war, at least to some degree. Second, my argument implies that, compared to nondemocratic regimes, democracies should generally allocate fewer of their resources to military spending and more of their resources to social spending. These predictions are not unique to my argument. They have the desirable quality, though, of being consistent with the substantial amount of empirical research that finds democracies spend more on social spending (e.g., Bueno de Mesquita et al. 2003) and less on military spending (e.g., Fordham and Walker 2005) than nondemocracies. Most relevant for my argument, Carter and Palmer (2015) find that nondemocracies increased military spending and decreased social spending to a greater degree during interstate wars than democracies between 1950 and 2001.

There are several reasons why I might fail to find empirical support for my claim that war mobilization is costlier for democratic leaders than dictators. First, my argument runs counter to the only existing research that explicitly addresses the domestic political consequences of war mobilization. Selectorate theory argues mobilization should be relatively more costly for autocratic leaders than for democratic leaders (Bueno de Mesquita et al. 1999, 2003, 2004). This claim follows from its assumptions that winning an interstate war is a public good and the economic resources not allocated to a war effort are then distributed to a leader’s winning coalition in the form of private benefits (2003, pg. 266). These assumptions imply two things. One, mobilization for interstate war increases the probability an autocratic leader will lose office because it decreases the resources available to be spent on private benefits. Two, mobilization indirectly decreases the probability a democratic incumbent will be removed from power because it increases the probability a state will win a war. As such, selectorate theory concludes the political cost of mobilization should be higher for autocratic leaders than for democratic leaders. Taking a different theoretical perspective, Goldsmith (2007) argues opposition parties provide democratic incumbents with an incentive to mobilize for war, absent among nondemocratic leaders. This is because citizens have the option of punishing an incumbent that fails to allocate the resources necessary to win a war by voting for a leader’s political opposition in a democracy. It then follows that not mobilizing economic resources for an interstate war effort should be more costly for a democrat than for a dictator. Despite their arguments, the empirical relationship between war mobilization and leader survival is an open question as neither Bueno de Mesquita et al. nor Goldsmith test their expectations.

Second, strategic behavior could result in no observable relationship between mobilization, regime type, and leader survival. If democratic incumbents are motivated by survival concerns and know they are likely to be removed from power if a war requires a significant mobilization of resources, then democratic leaders will rationally avoid participating in the interstate wars that require a substantial mobilization of resources. This would imply the relationship between mobilization and leader survival is endogenous and that my statistical analyses would be biased. However, the direction of the bias induced by the strategic behavior of democratic leaders would cut against the existence of observable evidence consistent with my theoretical argument.⁷

Third, the equilibrium war finance strategy for a democratic leader might minimize the threat to her political survival. Leaders that are able to finance an interstate war exclusively through some combination of borrowing money, raising taxes, and printing money can avoid reducing social spending. While the mobilization effort would represent an opportunity cost (as suggested by Holsti [1998] and Szayna et al. [2007]), fighting a war without reducing social spending could minimize the political cost of mobilization for a democratic incumbent.

Fourth, it could be that members of the public accept high levels of military spending when their country is fighting a war. This might occur because of a “rally around the flag” effect or a recognition that providing the public good of national security requires greater resources during an interstate war (Sandler and Hartley 1995). The latter possibility echoes the idea that the public might view mobilization not as a cost of war but as an investment that is required for their country to win an interstate war (Sullivan 2008).

Ultimately, whether democrats or dictators pay a higher political cost for mobilization is an empirical question. The next section describes the data and methods used to assess the relationship between leader survival, war mobilization, and regime type.

Research Design

The relationship between leader survival, regime type, and war mobilization is estimated using data on all political executives from 1950 to 2001. With the exception of variables measuring government spending, all data were taken from the replication materials associated with Debs and Goemans (2010).⁸ The dependent variable is the number of days a leader has been in power and is drawn from the Archigos project (Goemans, Gleditsch, and Chiozza 2009).⁹ The average leader served approximately four years and eight months, but this masks considerable variation: the fifth and ninety-fifth percentiles are equal to seventeen days and seventeen years and nine months.

Three theoretical concepts need to be operationalized in order to assess how war mobilization influences the survival of democratic and nondemocratic leaders: interstate war, military spending, and regime type. The dichotomous Interstate War is coded 1 in year t, if a leader’s country is involved in an interstate war and 0 otherwise. Interstate War is drawn from the ICB project (Brecher and Wilkenfeld 1997). There are 238 war-year observations that were overseen by 85 leaders in the data set. My primary analyses operationalize Military Spending as a state’s defense burden or the percent of a country’s GDP allocated to the military in year t. A state’s defense burden is the traditional indicator of national military spending in the defense economics literature (e.g., Sandler and Hartley 1995).¹⁰ Military Spending uses the National Material Capabilities data set (Singer, Bremer, and Stuckey 1972) for expenditure data and GDP data from Gleditsch (2002). Mean Military Spending in the data set is 2.75 (fifth percentile = 0.24; ninety-fifth percentile = 10.23).

My measure of regime type, Democracy, is coded 1 in year t if a leader’s government has a value of +7 or greater on the 21-point Polity2 index (Marshall and Jaggers 2005) and 0 otherwise. Democracy takes on a value of 1 in approximately 32 percent of the observations in the data set. Increasingly, conflict scholars disaggregate nondemocracies in their analyses (e.g., Weeks 2012). I pool types of nondemocracies for three reasons. First, my theoretical argument distinguishes between democracies and nondemocracies, not among democracies and types of authoritarian regimes. Empirically discriminating between democracies and nondemocracies, therefore, more closely fits with my theoretical argument and demonstrates greater construct validity than, for instance, distinguishing among democratic, personalist, one-party, and military regimes. Second, pooling nondemocracies does not bias my statistical analyses in favor of my argument. My statistical estimates reflect the mean difference in the effect of war mobilization on leader survival in all democracies and all nondemocracies. If the effect of mobilization on leader survival varies among autocracies, pooling them could weaken my statistical results by increasing inefficiency (King, Keohane, and Verba 1994). In this scenario, disaggregating types of authoritarian regimes would improve the efficiency of the statistical estimates, but would come at the price of moving the empirical analyses further away from the theoretical argument. If autocracies do not significantly vary in the political cost of mobilization, disaggregating them would decrease the efficiency of the parameter estimates (King, Keohane, and Verba 1994). Third, disaggregating authoritarian regimes would substantially complicate my statistical analyses due the number of interaction terms that would be required to produce unbiased estimates.

Retrieving unbiased estimates of how regime type, interstate war, and military spending influence leader survival requires that a statistical model include all possible interactions among Democracy, Interstate War, and Military Spending (e.g., Brambor, Clark, and Golder 2006; Braumoeller 2004). The most precise treatment of this issue is given by Braumoeller (2004), who writes that “In any interaction of k independent variables, a full set of $\sum_{k = 1}^{n} (\begin{matrix} n \\ k \end{matrix})$ coefficients must be estimated to avoid forcing the estimated hyperplane to assume a shape that may not conform to the general tendency of the pointcloud that it is intended to describe” (p. 811). Because omitting lower order and/or “tacit” interaction terms will lead to biased inferences unless all of the various effects are equal to zero, a more parsimonious specification that excludes these terms is methodologically inappropriate. The effect of war mobilization on leader survival across regime type therefore is modeled with Democracy, Interstate War, Military Spending, and the full set of interaction terms among these three measures.

While not inherently part of the mobilization process, participation in an interstate war often is accompanied by a reduction in the proportion of resources allocated to nonmilitary purposes. The theoretical argument developed above indicates that altering patterns of social spending, in particular, could influence the probability of leader survival. I model the relationship between leader survival and social spending in two ways. First, models include a measure of social spending as a control variable in order to isolate the effect of war mobilization on leader survival while accounting for the average influence of an indirect consequence of some mobilization efforts. Second, I also estimated a model that incorporates social spending into the set of interaction terms among Democracy, Interstate War, and Military Spending. Doing so allows for an unbiased estimate of how patterns of social spending associated with a mobilization effort influence the survival prospects of democratic and nondemocratic leaders but comes at the price of a substantially more complicated specification. My primary measure of social spending is the percentage of a state’s GDP allocated to health care in year t. Health Care Spending is drawn from Kugler (2002). I use health care spending to capture a state’s social spending because it is available for a greater number of countries for a longer period of time than other common indicators of social spending.

The statistical models control for a number of factors thought to influence leader survival. I account for whether the leader’s country was victorious (War Win), vanquished (War Loss), or obtained a draw (War Draw) in an interstate war during her time in power. Each variable is modeled as a decay function of $W a r O u t c o m e = \frac{1}{1 + t}$ , where $W a r O u t c o m e \in {W a r W i n, W a r D r a w, W a r L o s s}$ and t stands for the number of years since a given outcome was achieved. The war outcome variables are based on data from Brecher and Wilkenfeld (1997). At the state level, I control for a country’s involvement in a Civil War in a given year with a dichotomous indicator based on the Uppsala Conflict Data Program/International Peace Research Institute, Oslo (UCDP/PRIO) Armed Conflict Database (Gleditsch et al. 2002). The models also account for four aspects of a state’s economy: its (logged) GDP per capita (Maddison 2008), GDP growth (Maddison 2008), Trade Openness (International Monetary Fund 2008), and Change in Trade Openness (International Monetary Fund 2008). A state’s (logged) Population (Singer, Bremer, and Stuckey 1972) also is included as a control variable. The models account for three leader-level characteristics thought to influence an incumbent’s tenure: a leader’s Age when she entered office, the number of Previous Times in Office she served, and whether she came to power via an Irregular Entry. These three variables are drawn from the Archigos project (Goemans, Gleditsch, and Chiozza 2009). Finally, all of the models contain a set of variables that identify whether a state is located in Africa, the Americas, Asia, or the Middle East (Europe serves as the baseline category). These variables allow the statistical models to capture any region-specific variation in leader survival not explicitly accounted for by the other independent variables.

I estimate leader survival using semiparametric Cox models. Cox models are preferable to parametric event history models when the phenomenon of theoretical interest is the relationship between a set of covariates and the likelihood of a subject failing (Box-Steffensmeier and Jones 2004). The models estimated here account for nonproportional hazards between covariates and leader survival. Analysis of the Schoenfeld residuals was used to determine whether the influence of an explanatory variable on the hazard of a subject failing was constant over time (Box-Steffensmeier and Jones 2004). Following Box-Steffensmeier and Zorn (2001), any offending variable was interacted with the natural log of a leader’s tenure in office up to time t.¹¹

The Cox models estimated here also capture shared frailty among leaders of the same state. Shared frailty event history models account for unobserved heterogeneity across subgroups that makes subjects within group j more or less likely to fail than subjects in group $\tilde{j}$ (Box-Steffensmeier and Jones 2004). The shared frailty approach allows the statistical models to account for factors that make leaders of particular states systematically more or less likely to fail than leaders of other states without including control variables for every factor that might influence a leader’s survival. The frailty term v is a random variable with a mean of 1, variance of Θ, and is drawn from the γ distribution. Chiozza and Goemans (2004), Goemans (2008), and Debs and Goemans (2010) have used Cox models with shared frailty to model leadership survival.

Amelia II was used to address concerns about missing data (Honaker, King, and Blackwell 2007). Multiple imputation of missing observations helps avoid the inefficiency and selection bias associated with listwise deletion and is more accurate than single imputation (King et al. 2001; Honaker, King, and Blackwell 2007). Details about the imputation procedures are available in the Supplementary Appendix. The multiple imputation model produced five data sets of 7,935 leader-year observations (1,431 leaders from 162 countries).

Empirical Results

Table 1 reports my primary regression results. The estimates presented in Table 1 reflect the mean coefficients and corrected standard errors, as computed by Rubin’s (1987) method, yielded by the estimation of identically specified Cox models on each of the five imputed data sets.¹² Positive (negative) coefficients indicate higher values of an explanatory variable are associated with a leader facing a greater (lower) hazard of losing office. An interaction with ln(t) signed in the opposite direction of the constituent term indicates a decay in the original effect over a leader’s tenure.

Table 1.

Leader Survival, Regime Type, and Mobilization for Interstate War, 1950 to 2001.

	β	SE
Democracy	0.29	0.46c
Democracy × ln(t)	−0.01	0.07c
Interstate War	−0.77	0.42c
Military Spending	−0.07	0.02c
Democracy× Interstate War	−0.53	0.80c
Democracy × Military Spending	0.09	0.03c
Interstate War × Military Spending	−0.06	0.07c
Democracy × Interstate War × Military Spending	0.16	0.08c
Health Care Spending	−0.01	0.04
Win War	−1.40	0.84†
Draw War	−0.43	0.61
Lose War	2.06	0.50**
Civil War	0.85	0.37*
Civil War × ln(t)	−0.06	0.06
GDP per capita	2.26	0.20**
GDP per capita × ln(t)	−0.36	0.03**
Growth	−2.84	0.42**
Trade Openness	0.06	0.16
$Δ$ Trade Openness	−0.14	0.10
Population	0.03	0.04
Age	0.01	0.003**
Times in Office	−0.09	0.06
Irregular Entry	4.00	0.38**
Irregular Entry × ln(t)	−0.60	0.06**
Africa	7.27	0.57**
Africa × ln(t)	−1.20	0.08**
Americas	5.20	0.52**
Americas × ln(t)	−0.78	0.07**
Asia	6.47	0.51**
Asia × ln(t)	−1.04	0.07**
Middle East	6.24	0.65**
Middle East × ln(t)	−0.98	0.09**
Observations	7,935
Leaders	1,431
Failures	1,145
Log-likelihood	−6,640.03
Wald test	23.53	**
$Θ$	0.26	**

Note: c-parameters estimate conditional relationship whose statistical significance cannot directly be assessed.

Two-tailed: ^† p ≤ .1 *p ≤ .05 **p ≤ .01

Three things about Table 1 are worth noting. First, the statistically significant Θ indicates nontrivial variation exists in the likelihood of a leader being removed from power across states and that accounting for shared frailty among leaders of the same country is methodologically appropriate. Second, the significant regional indicators imply meaningful regional variation in leader tenure. Third, the use of interaction terms means that the effect of war mobilization on leader survival across regime type cannot be assessed using Table 1 for two reasons. One, the coefficient associated with any variable included in the set of interaction terms tells us the impact of an increase in that variable when all of the other constituent terms are equal to zero (Braumoeller 2004). Two, the standard error associated with a coefficient reflects the uncertainty around a variable’s estimated effect when the other constituent terms and interaction terms are equal to zero and does not account for the covariance among the constituent and interaction terms (Brambor, Clark, and Golder 2006).

Given the limitations of Table 1, I used a set of simulations based on the model’s parameter estimates to calculate how mobilization for interstate war influences the relative survival prospects of a democratic leader and a dictator. I took 1,000 draws from a multivariate normal distribution based on the coefficient and variance–covariance matrices of each model estimated on each of the five imputed data sets. I then calculated the probability of a democratic leader and a dictator surviving up to time t over a five-year period given four mobilization scenarios. Each scenario assumes that a leader mobilized her country’s resources and fought an interstate war in her first two years in office (mean war duration in the data), and then was able to return military spending back to its mean peacetime levels during the following three years. The values of Military Spending used in the simulations represent the average values across the five imputed data sets for a given scenario. The first scenario assumes that a leader was able to fight a war without increasing military spending from its peacetime level (i.e., Military Spending took on values of 2.63 percent). This serves as a baseline from which we can estimate how politically costly mobilizations are for a leader. The second scenario assumes that a leader fought an interstate war with a small military mobilization, which is operationalized by setting Military Spending equal to 3.62 percent (or 25 percent of the mean mobilization). The third scenario assumes that a leader engaged in an average mobilization, which was accomplished by setting Military Spending to its mean wartime values while the war was being waged (6.61 percent). The fourth scenario models a large war effort by setting Military Spending to 10.58 percent during the war, which represents twice the size of the average military mobilization. The values of the constituent variables and interaction terms were manipulated as necessary to accurately reflect the different scenarios. All of the simulations assume that the incumbent controlled a European country, won the war, and had either the mean (continuous variables) or median (nominal or ordinal variables) characteristics with respect to the other control variables.

Assessing the relative political cost of war mobilization across regime type requires the calculation of three quantities. The first is the probability that a democrat or a dictator will remain in office in each of the four mobilization scenarios. The second is the effect of mobilization on the probability that a democrat or a dictator will remain in power. This is done by calculating the difference between a given type of leader’s political survival when her country fought a war without altering peacetime allocations of military spending and (1) a small mobilization effort; (2) an average mobilization effort; and (3) a large mobilization effort. This identifies the within-regime type effect of mobilization on a leader’s survival prospects. The third quantity is the difference in the effect of a given war mobilization on the probability that a democratic leader and a dictator will remain in power. This identifies the relative political cost of mobilization across regime type and is the quantity of primary theoretical interest. I report the probabilities of leader survival for a given mobilization scenario in the Supplementary Appendix. Figure 1 presents the mean effect of war mobilization on the probability a democratic leader (row 1) and a dictator (row 2) will be removed from power given a small mobilization (column A, solid green line), an average mobilization (column B, dashed red line), and a large mobilization (column C, dotted blue line).¹³ Values above (below) zero imply that a mobilization effort increases (decreases) the probability a leader will be removed from power compared to the no-mobilization scenario. The effect of mobilization on leader survival is statistically significant at the 0.05 level if the confidence interval (shaded area) does not include the zero-line.

Figure 1.

The political cost of mobilization in democracies and dictatorships, 1950 to 2001.

Figure 1 suggests three empirical relationships between leader survival, war mobilization, and regime type. First, mobilization for interstate war is associated with an increase in the probability a democratic leader will be removed from office. Compared to the scenario in which fighting an interstate war did not alter a government’s peacetime allocation of military spending, democratic leaders face a statistically significant higher probability of losing power given a small mobilization (row 1, column A), an average mobilization (row 1, column B), or a large mobilization (row 1, column C). Second, mobilization for interstate war does not significantly affect the survival of dictators. Relative to a war that does not involve an increase in military spending, nondemocratic leaders appear to be less likely to be removed from office given a small mobilization (row 2, column A), an average mobilization (row 2, column B), or a large mobilization (row 2, column C). These differences, though, are not quite statistically significant at the 0.05 level. Third, the substantive magnitude of the effect of war mobilization on leader survival is increasing in the size of a mobilization effort. For example, compared to the no-mobilization scenario, at the end of her second year in office a democratic leader is 1.5 percentage points more likely to be removed from power given a small mobilization, 6.9 percentages points more likely given an average mobilization, and 17.1 percentage points more likely given a large mobilization. At the same time, compared to fighting a war without increasing military spending, an autocratic leader is expected to be 1 percentage point less likely to be removed from power at the end of his second year given a small mobilization, 3.5 percentage points less likely given an average mobilization, and 5.5 percentage points less likely given a large mobilization.

Figure 2 offers a direct test of my theoretical argument by reporting the difference in the changes in the probability a democratic leader and a dictator will remain in power, given the move from no mobilization to a small mobilization (Panel A), an average mobilization (Panel B), and a large mobilization (Panel C). As the confidence intervals in each panel lie completely above the zero-line, Figure 2 indicates that war mobilization increases the probability a democratic incumbent will be removed from power to a significantly greater degree than the probability a dictator will lose office. Further, the magnitude of this effect is increasing in the size of a mobilization effort. Compared to the no-mobilization scenario, the likelihood a democratic leader will be removed after her second year in power is increased 2.5 percentage points more than it is for a dictator given a small mobilization, 10.4 percentage points more given an average mobilization, and 22.6 percentage points more given a large mobilization effort.

Figure 2.

Differences in the political cost of mobilization across regime type, 1950 to 2001.

The results of the control variables largely are consistent with prior research. On average, a leader is more likely to be removed from office after losing an interstate war and less likely to lose power after winning a war than if her country had remained at peace. Obtaining a draw does not have a significant influence on a leader’s expected tenure. My results suggest that civil war and an irregular entry increase a leader’s hazard of losing office, but that these relationships weaken throughout an incumbent’s tenure (Debs and Goemans 2010). I also find that the leaders of wealthier countries face a higher probability of losing office early in their tenures, but that this effect decreases over time. Positive economic growth significantly lowers the probability an incumbent is removed from power (Chiozza and Goemans 2004; Debs and Goemans 2010). I find no significant relationship between an incumbent’s probability of survival and the number of previous times she has been in office or her state’s trade openness, change in trade openness, or (logged) population.

Additional Analyses

I conducted a number of additional analyses. The first set uses a leader-year unit-of-analysis, while the second relies on a war-leader unit-of-analysis to incorporate measures that could affect my results but are not easily included in a leader-year framework. I present here the results of postestimation simulations and report the statistical models in the Supplementary Appendix.¹⁴

Leader-Year Analyses

My primary analysis models war mobilization as a function of the level of military spending. Alternatively, war mobilization can thought of as a deviation from a state’s typical allocation of military spending. To capture this idea, I reestimated the primary model with Military Spending measured as the annual deviation from a state’s mean defense burden. The first row of Figure 3 presents the difference in the effect of mobilization on the probability that a democratic leader and a dictator will remain in power. The results in row 1 also indicate that the effect of war mobilization on leader survival is significantly greater in democracies than in dictatorships.

Figure 3.

Differences in the political cost of mobilization across regime type, 1950 to 2001. Row 1—Defense burden measured as annual deviation; row 2—parsimonious specification; row 3—education spending; row 4—health care spending included in set of interaction terms; row 5—regime type measured with continuous variable.

The control variables included in my primary model are based on the analyses of Chiozza and Goemans (2004) and Debs and Goemans (2010). I also estimated a more parsimonious model that contains only the theoretically necessary explanatory variables, variables that identify a war’s outcome, and the regional indicators. As reported in row 2 of Figure 3, the results from this model also imply that war mobilization increases the probability a democratic incumbent will be removed from power to a greater degree than a dictator.

Health care spending is used to proxy social spending in my primary analysis because cross-national data on health care expenditures are available for a longer period of time for more countries than other indicators. I estimated a model using education spending to ensure my results are not sensitive to this decision. This does not appear to be the case, as row 3 of Figure 3 indicates mobilization increases the probability a democratic incumbent will lose office to a greater degree than it does for a nondemocratic leader when social spending is modeled as education spending.

War mobilization often is accompanied by a decrease in the proportion of economic resources dedicated to social spending. We can model how this change influences the effect of war mobilization on the survival of democrats and dictators by incorporating social spending into the set of interactions among measures of regime type, interstate war, and military spending. This induces a substantial amount of multicollinearity among the constituent variables and various interaction terms but is necessary to produce unbiased estimates (Braumoeller 2004). The results yielded by these models are reported in row 4 of Figure 3 and suggest two things. First, the political cost of war mobilization is significantly higher in democracies than in dictatorships. Second, at least at the margins, democratic leaders can mitigate the relative political cost of mobilization by funding higher wartime military spending with a strategy that does not reduce social spending. This inference follows from the observation that the magnitudes of the difference in the probabilities of leader survival across regime type are relatively larger when Health Care Spending is not incorporated into the set of interaction terms.

Finally, it could be the case that my results are driven by the use of a dichotomous indicator of regime type. I therefore estimated a set of models that used the 21-point Polity2 (Marshall and Jaggers 2005) index to measure regime type. As the results reported in row 5 of Figure 3 indicate, these analyses also demonstrate that a democratic leader is more likely to lose office given mobilization for interstate war than a dictator.

War-Leader Analyses

The results reported thus far have not accounted for the possible effects of fatalities (Valentino, Huth, and Croco 2010) or a leader’s culpability (Croco 2011). Unfortunately, incorporating measures of these concepts into the leader-year analyses reported above is problematic because systematic data on participant fatalities that vary over the course of an interstate war do not exist and there is a lack of variation in the culpability measure across regime type during the period from 1950 to 2001. I therefore estimated the effect of war mobilization on leader survival using the war-leader data set associated with Croco (2011). More specifically, I used a logit model to estimate leader removal as a function of whether a leader’s state was a democracy, its mean wartime deviation of defense burden from its mean peacetime defense burden, an interaction between those two variables, the (logged) casualties that occurred during a leader’s wartime tenure, whether a leader was culpable for the war, the war’s outcome, and a leader’s (logged) prewar tenure. Figure 4 presents the difference in the effect of wartime deviations from a state’s mean defense burden on the probability of leader removal across regime type.

Figure 4.

Differences in the political cost of mobilization across regime type, 1950 to 2001. Based on logit model using a war-leader unit of analysis.

Consistent with my other analyses, increasing a state’s military spending during an interstate war increases the probability a democratic leader will be removed from power to a significantly greater degree than is the case with a dictator. Thus, my substantive findings are robust to the effects of leader culpability and fatalities, an alternative unit of analysis (war-leader vs. leader-year), and an alternative definition of interstate war (COW vs. ICB). Figure 4 also implies reducing military spending during a war to below a state’s mean defense burden decreases the probability a democratic incumbent will lose office to a greater degree than it does a nondemocratic leader.

Conclusion

The idea that interstate war is politically costlier for democratic leaders than it is for dictators underlies most institutional explanations for variation in conflict behavior across regime type. Therefore, recent research that demonstrates this assumption is not empirically supported calls into question why scholars have found that democracies are less likely to fight other democracies (Russett and Oneal 2001), more selective in the conflicts they initiate (Clark and Reed 2003), less likely to reciprocate military challenges (Schultz 2001a), more likely to fight shorter wars (Bennett and Stam 1998), and more likely to win the wars they fight (Reiter and Stam 2002). My results offer insight as to why conflict processes vary across regime type.

I find that democratic leaders pay a higher political cost for interstate war than do dictators when a war effort is associated with increases in military spending. Filson and Werner (2007) formally demonstrate that a greater sensitivity to the cost of war participation should be associated with lower probabilities of conflict initiation and reciprocation, shorter conflicts, and higher probabilities of victory. My results provide an empirical referent for Filson and Werner’s theoretical results, with a slight twist. Specifically, they suggest that differences in conflict behavior across regime type are likely conditional on the size of an expected or observed mobilization effort. If the cost of war participation in and of itself does not vary across regime type (as demonstrated by Chiozza and Goemans 2004), then it is not clear how much variation across regime type we should observe in situations where a war is, or could be, fought without a significant mobilization of resources. In contrast, if fighting a war requires a substantial mobilization effort, my results imply that incumbency-valuing democrats and dictators have strong incentives to behave differently.

My results also have at least two other implications for the relationship between regime type and interstate conflict. First, they suggest that democratic leaders have an incentive absent among autocrats to try to limit the scope of mobilization. Combined with the observation that dictators are more likely to be punished for losing an interstate war than democratic incumbents (e.g., Debs and Goemans 2010), my results imply that war should be associated with larger increases in military spending and cuts in social spending in autocracies than in democracies, at least during the period from 1950 to 2001. This is consistent with the findings of Carter and Palmer (2015) but not with selectorate theory (Bueno de Mesquita et al. 2004). However, selectorate theory’s argument relies on the assumptions that losing an interstate war is costlier for democratic leaders and mobilization is costlier for autocratic leaders. The empirical results of Chiozza and Goemans (2004), Debs and Goemans (2010), and this article suggest these assumptions do not hold. Accordingly, it is unclear why democrats have an incentive to mobilize more resources for war than dictators.

Second, my results inform the debate over why democracies are more likely to win the wars they fight than are nondemocracies (most notably, Reiter and Stam 2002). Most explanations for this empirical finding argue that, relative to dictatorships, democracies are more selective about the conflicts they initiate (Clark and Reed 2003), mobilize more of their economic resources during a war (Lake 1992), or are both more selective and try harder to win the wars they fight (Bueno de Mesquita et al. 1999). My results suggest trying to win an interstate war through a large mobilization effort is a less attractive option for a democratic leader than a dictator. Therefore, it is likely the case that democratic success in interstate wars is due to greater selectivity and not a mobilization advantage.

The political cost of interstate war mobilization is greater for democratic leaders than it is for dictators. This finding offers a theoretical explanation for a number of prominent findings in the interstate conflict literature that, importantly, does not rely on the empirically untenable assumptions that a democratic incumbent is more likely to lose office than an autocratic leader for participating in or losing an interstate war. My results also imply that whether democracies and dictatorships pursue different strategies with respect to crisis bargaining and interstate war should be conditional on the need to mobilize if ex ante bargaining fails and fighting requires a significant mobilization of resources. The findings presented here therefore suggest that explicitly considering the political cost of mobilization holds promise for improving our understanding of interstate conflict processes.

Footnotes

Author’s Note

Previously presented at the 2013 meeting of the Peace Science Society (International). Replication materials are available at the Journal of Conflict Resolution’s website.

Acknowledgments

I thank Scott Bennett, David Carter, Matt DiGiuseppe, Ben Jones, Jake Kathman, Doug Lemke, Tim Nordstrom, Glenn Palmer, James Shortle, Yael Zeira, Heather Ondercin, two anonymous reviewers, and Paul Huth for comments on earlier versions of this manuscript and Sarah Croco for sharing her data.

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

Supplementary Material

Supplementary material for this article is available online.

Notes

References

Acemoglu

Daron

Robinson

James A.

. 2006. Economic Origins of Dictatorship and Democracy. Cambridge, UK:Cambridge University Press.

Arena

Philip

Nicoletti

Nicholas P.

. 2014. “Selectorate Theory, the Democratic Peace, and Public Goods Provision.” International Theory 6 (3): 391–416.

Bachman

Jerald G.

Blair

John David

Segal

David R.

. 1977. The All-volunteer Force: A Study of Ideology in the Military. Ann Arbor: University of Michigan Press.

Bennett

D. Scott

Stam

Allan C.

. 1998. “The Declining Advantages of Democracy: A Combined Model of War Outcomes and Duration.” Journal of Conflict Resolution 42 (3): 344–66.

Boix

Carles

. 2003. Democracy and Redistribution. New York: Cambridge University Press.

Box-Steffensmeier

Janet M.

Jones

Bradford S.

. 2004. Event History Modeling. A Guide for Social Scientists. Cambridge, UK: Cambridge University Press.

Box-Steffensmeier

Janet M.

Zorn

Christopher J. W.

. 2001. “Duration Models and Proportional Hazards in Political Science.” American Journal of Political Science 45 (4): 972–88.

Brambor

Thomas

Clark

William

Golder

Matt

. 2006. “Understanding Interaction Models: Improving Empirical Analyses.” Political Analysis 14 (1): 63–82.

Braumoeller

Bear F.

2004. “Hypothesis Testing and Multiplicative Interaction Terms.” International Organization 58 (4): 807–20.

10.

Brecher

Michael

Wilkenfeld

Jonathan

. 1997. A Study of Crisis. Ann Arbor: University of Michigan Press.

11.

Bueno de Mesquita

Bruce

Morrow

James

Siverson

Randolph M.

Smith

Alastair

. 1999. “An Institutional Explanation of the Democratic Peace.” American Political Science Review 93 (4): 791–808.

12.

Bueno de Mesquita

Bruce

Morrow

James

Siverson

Randolph M.

Smith

Alastair

. 2004. “Testing Novel Implications of the Selectorate Theory of War.” World Politics 56 (3): 363–88.

13.

Bueno de Mesquita

Bruce

Siverson

Randolph

. 1995. “War and the Survival of Political Leaders: A Comparative Study of Regime Types and Political Accountability.” American Political Science Review 89 (4): 841–55.

14.

Bueno de Mesquita

Bruce

Smith

Alastair

Siverson

Randolph M.

Morrow

James

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

15.

Carter

Jeff

Palmer

Glenn

. 2015. “Keeping the Schools Open While the Troops Are Away: Regime Type, Interstate War and Government Spending.” International Studies Quarterly 59 (1): 145–57.

16.

Chiozza

Giacomo

Goemans

H. E.

. 2004. “International Conflict and the Tenure of Leaders: Is War Still Ex Post Inefficient?” American Journal of Political Science 48 (3): 604–19.

17.

Chiozza

Giacomo

Goemans

H. E.

. 2011. Leaders and International Conflict. New York: Cambridge University Press.

18.

Clark

David H.

Reed

William

. 2003. “A Unified Model of War Onset and Outcome.” Journal of Politics 65 (1): 69–91.

19.

Cook

Fay Lomax

Barrett

Edith J.

. 1992. Support for the American Welfare State. New York: Columbia University Press.

20.

Croco

Sarah E.

2011. “The Deciderâ€™s Dilemma: Leader Culpability, War Outcomes, and Domestic Punishment.” American Political Science Review 105 (3): 457–77.

21.

Croco

Sarah E.

Weeks

Jessica L. P.

. n.d. “Willing and Able: Culpability, Vulnerability, and Leaders’ Sensitivity to War Outcomes.” Unpublished manuscript, University of Maryland, College Park, MD and University of Wisconsin-Madison, Madison, WI.

22.

Dahl

Robert A.

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

23.

Debs

Alexandre

Goemans

H. E.

. 2010. “Regime Type, the Fate of Leaders and War.” American Political Science Review 104 (3): 430–45.

24.

Filson

Darren

Werner

Suzanne

. 2007. “Sensitivity to Costs of Fighting Versus Sensitivity to Losing the Conflict.” Journal of Conflict Resolution 51 (5): 691–714.

25.

Fordham

Benjamin O.

Walker

Thomas C.

. 2005. “Kantian Liberalism, Regime Type and Military Resource Allocation: Do Democracies Spend Less?” International Studies Quarterly 49:141–57.

26.

Geddes

Barbara

. 2003. Paradigms and Sand Castles: Theory Building and Research Design in Comparative Politics. Ann Arbor: University of Michigan Press.

27.

Gilens

Martin

. 2009. “Preference Gaps and Inequality in Representation.” PS: Political Science & Politics 42 (2): 335–41.

28.

Gleditsch

Kristian S.

2002. “Expanded Trade and GDP Data.” Journal of Conflict Resolution 46 (5): 712–24.

29.

Gleditsch

Nils Petter

Wallensteen

Peter

Erikson

Mikael

Sollenberg

Margareta

Strand

Haavard

. 2002. “Armed Conflict, 1945-99: A New Dataset.” Journal of Peace Research 39 (5): 615–37.

30.

Goemans

H. E.

2008. “Which Way out? The Manner and Consequences of Losing Office.” Journal of Conflict Resolution 53 (6): 771–94.

31.

Goemans

H. E.

Gleditsch

Kristian Skrede

Chiozza

Giacomo

. 2009. “Introducing Archigos: A Data Set of Political Leaders, 1875-2003.” Journal of Peace Research 46 (2): 269–83.

32.

Goldsmith

Benjamin E.

2007. “Defense Effort and Institutional Theories of Democratic Peace and Victory: Why Try Harder?” Security Studies 16 (2): 189–222.

33.

Heston

Alan

Summers

Robert

Aten

Bettina

. 2012. Penn World Table Version 7.1. Philadelphia: Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania.

34.

Holsti

Ole R.

1998. “A Widening Gap between the US Military and Civilian Society?: Some Evidence, 1976-96.” International Security 23 (3): 5–42.

35.

Honaker

James

King

Gary

Blackwell

Matthew

. 2007. “Amelia II: A Program for Missing Data.” Accessed December 23, 2013. http://gking.harvard.edu/amelia/.

36.

International Monetary Fund. 2008. International Financial Statistics. Washington, DC: International Monetary Fund [Producer and Distributor], June 2008.

37.

Jæger

Mads Meier

. 2006. “What Makes People Support Public Responsibility for Welfare Provision: Self-interest or Political Ideology? A Longitudinal Approach.” Acta Sociologica 49 (3): 321–38.

38.

King

Gary

Honaker

James

Joseph

Anne

Scheve

Kenneth

. 2001. “Analyzing Incomplete Political Science Data: An Alternative Algorithm for Multiple Imputation.” American Political Science Review 95 (1): 49–69.

39.

King

Gary

Keohane

Robert O.

Verba

Sidney

. 1994. Designing Social Inquiry: Scientific Inference in Qualitative Research. Princeton, NJ: Princeton University Press.

40.

Kugler

Jacek

. 2002. Relative Political Capacity Data Set. Unpublished Manuscript. Claremont, CA: Claremont Graduate University.

41.

Lake

David A.

1992. “Powerful Pacifists: Democratic States and War.” American Political Science Review 86 (1): 24–37.

42.

Maddison

Angus

. 2008. “Statistics on World Population, GDP and Per Capita GDP, 1-2008 AD.” Accessed December 23, 2013. http://www.ggdc.net/maddison/oriindex.htm.

43.

Marshall

Monty

Jaggers

Keith

. 2005. “Polity IV Dataset Users’ Manual.” Accessed December 23, 2013. www.cidcm.umd.edu//polity.

44.

Nordlinger

Eric A.

1977. Soldiers in Politics: Military Coups and Governments. Englewood Cliffs, NJ: Prentice Hall.

45.

Olson

Mancur

. 1965. The Logic of Collective Action: Public Goods and The Theory of Groups. Cambridge, MA: Harvard University Press.

46.

Organski

A. F. K.

Kugler

Jacek

. 1980. The War Ledger. Chicago: The University of Chicago Press.

47.

Page

Benjamin I.

Bartels

Larry M.

Seawright

Jason

. 2013. “Democracy and the Policy Preferences of Wealthy Americans.” Perspectives on Politics 11 (1): 51–73.

48.

Przeworski

Adam

Alvarez

Michael E.

Antonio Cheibub

José

Limongi

Fernando

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

49.

Reiter

Dan

Stam

Allan C.

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

50.

Rockoff

Hugh

. 2012. America’s Economic Way of War: War and the U.S. Economy from the Spanish-American War to the First Gulf War. New York: Cambridge University Press.

51.

Rubin

Donald B.

1987. Multiple Imputation for Nonresponse in Surveys. New York: Wiley.

52.

Russett

Bruce

Oneal

John R.

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

53.

Sandler

Todd

Hartley

Keith

. 1995. The Economics of Defense. New York: Cambridge University Press.

54.

Schafer

Joseph L.

Olsen

Maren K.

. 1998. “Multiple Imputation for Multivariate Missing-data Problems: A Data AnalystÂ’s Perspective.” Multivariate Behavioral Research 33 (4): 545–71.

55.

Schultz

Kenneth A.

2001a. Democracy and Coercive Diplomacy. Cambridge, UK: Cambridge University Press.

56.

Schultz

Kenneth A.

2001b. “Looking for Audience Costs.” Journal of Conflict Resolution 45 (1): 32–60.

57.

Singer

J. David

Bremer

Stuart A.

Stuckey

John

. 1972. “Capability Distribution, Uncertainty, and Major Power War, 1820-1965.” In Peace, War, and Numbers, edited by Russett

Bruce

, 19–48. Beverly Hills, CA: Sage.

58.

Slantchev

Branislav L.

2005. “Military Coercion in Interstate Crises.” American Political Science Review 99 (4): 533–47.

59.

Sullivan

Patricia L.

2008. “Sustaining the Fight: A Cross-sectional Time-series Analysis of Public Support for Ongoing Military Interventions.” Conflict Management and Peace Science 25 (2): 112–35.

60.

Szayna

Thomas S.

McCarthy

Kevin F.

Sollinger

Jerry M.

Demaine

Linda J.

Marquis

Jefferson P.

Steele

Brett

. 2007. The Civil-military Gap in the United States: Does it Exist, Why, and Does it Matter? Vol. 379. Santa Monica, CA: Rand Corporation.

61.

Valentino

Benjamin A.

Huth

Paul K.

Croco

Sarah E.

. 2010. “Bear Any Burden? How Democracies Minimize the Costs of War.” The Journal of Politics 72 (2): 528–44.

62.

Weeks

Jessica L.

2008. “Autocratic Audience Costs: Regime Type and Signaling Resolve.” International Organization 62 (1): 35–64.

63.

Weeks

Jessica L.

2012. “Strongmen and Straw Men: Authoritarian Regimes and the Initiation of International Conflict.” American Political Science Review 106 (2): 326–47.