War and the Reelection Motive

Abstract

This article investigates the relationship between term limits and international conflict. Theories of political survival and diversionary war both imply term limits should play a role in international relations, whereas “permanent referendum theory,” largely motivated by work in American politics, suggests otherwise. Drawing on these theories, we formulate and test competing hypotheses regarding term limits and international crises. Using dyadic militarized interstate disputes data and information on forty-eight democracies with term limits, we uncover strong evidence to support the claim that leaders reaching final terms in office are more likely to initiate conflict than those still subject to reelection. Moreover, we find that the likelihood of conflict initiation is significantly higher during times of recession, but only in the absence of binding term limits. While binding electoral terms and economic downturns are both independently associated with increased levels of conflict initiation, in concert their conditional effects actually counteract each other.

Keywords

international conflict term limits diversionary war domestic politics

What impact, if any, do term limits have on leaders and the use of force? This straightforward question has garnered scant attention from scholars of international relations.¹ But it deserves exploration. That leaders are office-seeking is now a mainstream assumption underlying many areas of domestic and party politics research. Given specified constraints of the electoral environment, leaders or parties are assumed to derive optimal policies for election or reelection (Downs 1957; Hinich and Ordeshook 1970). The salience of domestic politics in shaping international relations has become increasingly acknowledged, and work linking political structures with international conflict continues to grow. However, this research program remains largely aloof to the impact of binding term limits. Owing to this dearth of research, there are multiple hypotheses one might advance concerning the potential impact term limits have on the use of force.

The essence of political survival is that heads of state behave so as to preserve their power (de Mesquita et al. 2003). According to the “selectorate model,” democratic leaders have to satisfy broad electoral constraints and autocratic rulers have to satisfy a minimum support coalition. Thus, the driving incentive of leaders—be they democratic or not—is to maintain office, to hold onto the position which affords them their status and influence. This premise has arguably reached the level of dogma in political science. And with good reason: it has provided a useful analytical handle for grasping an array of issues within comparative politics and international relations such as democratic peace (de Mesquita et al. 2003), territorial wars (Chiozza and Choi 2003), or foreign aid allocation (de Mesquita and Smith 2007).

Term limits represent an observable and publicly known limitation on leader tenure. Where observed, term limits serve as a structural impediment to remaining in power. Viewed then from the prism of political survival, the implication is that if a term limit binds or prevents a leader’s continuation of power, an observable difference in actions should result, possibly by enabling a leader’s full scope of action.² Put another way, if the chief characteristic compelling a leader’s conduct is an operative desire to preserve her position of power, then removing this motivation should alter behavior. Based on this logic, changes in international conflict initiation (or in targeting) should result. Additionally, term limits may interact with diversionary war motives, making conflict more or less likely depending on socioeconomic contexts. This interpretation implies that term limits should transform a leader’s susceptibility to diversionary inducements.

In this article, we assess whether limits on leadership tenure—a structural constraint common to many democratic states—have bearing on international affairs, by focusing on their impact on the use of force. Expectations regarding this topic vary based on the theoretical lens one uses to interpret it. By testing the implications stemming from these various theories, we are better able to evaluate their applicability to various aspects of international relations. Term limits may influence conflict independently, as a strict interpretation of political survival would suggest. The possibility also remains that term limits alter diversionary incentives, by exempting leaders from ballot box concerns during periods of socioeconomic difficulty. And finally, the exigencies of high office may render leaders more or less permanently dependent on domestic public approval, thereby mitigating any impact of binding terms.

We seek to adjudicate these differing perspectives by adding information on term limits to a directed-dyad analysis. Specifically, we use data on binding term limits in forty-eight democracies from 1976 to 2000 and ascertain their relationship with international conflict. In addition, we add to the analysis by employing various measures for economic recessions to investigate potential conditional effects of the office-seeking assumption. This permits us to confront the evidence for term limits directly with tests of diversionary theory. Inconsistent with what we call “permanent referendum theory” (Russett 1990a), the empirical evidence we present shows strong support for the importance of term limits as an independent factor. We find leaders constrained by elections show less belligerency, while leaders subject to binding final terms and thus less restricted by the demands of the electoral process, engage in more conflictual behavior. Reaching final terms in office increases a leader’s likelihood of initiating conflict by more than 50 percent. Importantly, we find that economic recessions also strongly increase the likelihood of conflict initiation, revealing diversionary motives at work. Interestingly, if term limits bind in times of a recession, the effect vanishes and leaders can avoid satisfying diversionary motives. These results are robust across various measures of recessions and different estimation approaches. These outcomes also nicely dovetail with the existing literature, confirming a generally pacifying effect of elections, except in times of recession. Moreover, the results suggest that term limits act as a countervailing force to office-seeking behavior; they may either increase or decrease the probability of conflict, depending on the economic and political context leaders find themselves in.

Political Survival Interrupted?

The relationship between the office-seeking incentives of leaders and conflict is by now a long-standing inquiry in international relations. Much of the early work in this area examines the US case, especially after World War II (WWII). Stoll (1984) analyzes the impact of visible military force on US elections from 1947 to 1982. He finds that presidents may attempt to gain a “rally around the flag” effect by engaging in the visible use of force during reelections. Russett (1990b) finds that US participation in the initiation or escalation of a dispute is positively related to a presidential-election-year dummy variable. Work by Wang (1996) and Russett (1990a) suggests military action is more likely shortly before an election. Hess and Orphanides (1995) use a nonparametric approach to test for the increased frequency of US wars during periods when a president seeks reelection and the economy’s performance has been weak. Examining the years from 1953 to 1988, they find that the conditional probability of war approximately doubles (from 30 to 60 percent) during any year in which a president is not in a final term and which experiences an economic recession.

However, a more refined strand of this literature focuses specifically on how the risk of losing office impacts the calculus of conflict. It also moves beyond the single US case. This work spans war initiation (de Mesquita et al. 1999) to its termination (Goemans 2000), and many of its in-betweens, for example, war expansion (Siverson 1996). Chiozza and Geomans (2003) also offer empirical evidence for a reciprocal relationship between tenure and conflict. The authors find that as the risk of losing office increases, leaders become less likely to initiate a crisis and that as the risk of an international crisis increases, leaders become more likely to lose office.

One aspect of this relationship deals with war and selective incentives. As noted, the selectorate model (de Mesquita et al. 2003) is the workhorse theory in this research program. It suggests that the ballot box is of fundamental importance to heads of state; leaders must satisfy their coalitions or be punished at election time. They therefore choose policies with a view to remaining in office. Moreover, war conduct is inextricably influenced via political calculations associated with tenure because outcomes in international conflict can impact tenure duration. Seeking to preserve their power, leaders only involve themselves in disputes if such actions are likely to prolong their time in office. From this perspective, leaders are thought to choose their fights well (de Mesquita et al. 2003). In sum, heads of state selecting into international disputes do so because it advances tenure prospects. For the same reasons, leaders are thought to avoid conflicts that may cut short their time in office.

However, research in this area diverges from diversionary use-of-force literature in its expectations about the timing of war and the likelihood of losing office. It argues that the lower the risk of being put out of office, the higher is the likelihood of conflict initiation. The logic here suggests there is a premium on political security because it offers either a basis of support for conflict situations or a buffer against the poor outcomes (de Mesquita and Siverson 1995) they may entail. Gaubatz (1991) finds that states are most likely to become involved in a militarized dispute earlier rather than later in an election cycle. Huth and Allee (2002) similarly report that challengers are more likely to pursue efforts at altering the status quo after general elections.

While selectorate theory helps motivate the discussion, it does not by itself lead to precise predictions related to term limits and international conflict. It does, however, assist in properly situating the question. The selectorate model generates the expectation that nondemocratic leaders are more prone to risk conflict than are their democratic counterparts. This outcome is largely a consequence of the relatively smaller winning coalitions authoritarian rulers must satisfy. However, the private goods mechanism by which nondemocratic heads of state protect themselves from failed international adventures does not plainly apply to leaders facing final terms in office. Term limits do not affect or directly impact the size of minimum winning coalitions; they make them irrelevant.

This gives rise to an open-ended question. When binding term limits effectively unhinge leaders from the dictates of electoral accountability, how might we expect them to behave regarding the security of the state? Should we anticipate more bellicosity or restraint? The expectations of permanent referendum theory notwithstanding, leaders in final terms will enjoy more latitude to pursue favored actions pertaining to conflict scenarios; but are we able to say much about such actions ex ante? Toward addressing this issue, a host of arguments suggests that term limited leaders should behave more aggressively than those leaders still appealing to winning coalitions or the median voter.

Jackson and Morelli (2007) consider a model where the decision to attempt international war is essentially a cost–benefit calculation. In their analysis, it is the ratio of a leader’s share of the benefits from war compared to her share of costs that is a critical determinant in opting for conflict. The propensity of a leader to prefer going to war is thought to increase as the ratio of gain to risk associated with war rises. Autocratic leaders are said to have a greater inclination for war than do democratic leaders, on account of their political “bias.” That is, they have a higher benefit-to-cost ratio compared to their society as a whole and thereby a higher conflict tendency. The authors argue that in authoritarian or military regimes, rulers can keep a disproportionate share of the spoils of war. The effect is to engender a positive or greater bias for war. By contrast, democratic heads of state are more likely to overweigh the costs rather than the private benefits associated with conflict. For instance, if the risk of being voted out of office is higher after losing a war, then leaders may overemphasize the costs of conflict relative to the gains.

Extending this “political bias” analysis to incorporate term limits is straightforward. Term limits obviate the concern of being removed from office after losing a risky war. Exit from office is anticipated at the conclusion of final terms. Therefore, if the likelihood of being untimely removed from office engenders a lower war bias, then eliminating such a risk should appreciably raise a leader’s political bias. Put differently, if binding terms reduce the costs of war, thereby raising war’s benefit-to-cost ratio, then leaders should find war more attractive in final terms.

The cost argument applies similarly to Fearon’s (1994) theory of domestic audiences and international disputes. The “audience costs” associated with domestic constituencies make leaders acutely sensitive to the consequences of failed international endeavors. This mechanism is more likely to operate in democratic states, where domestic audiences play a larger role in the fate of leaders. A key implication of the theory is that domestically unaccountable leaders—that is, authoritarian rulers—should find it easier to engage in risky military endeavors and then back down if met with strong resistance.³ If arriving in a final term lowers a leader’s sensitivity to domestic audience costs, then, all else being equal, they should be more inclined to engage in military disputes or “limited probes.”

Term limits may also impact the benefits side of the political bias model. The fighting for survival explanation for war (Chiozza and Goemans 2011) argues that the risk of regular removal from office declines when leaders are the initiators of international crises, as opposed to the targets. This challenge benefit accrues, in part, because in times of war, scheduled elections often are postponed until after the war. Such benefits may be partially applicable to leaders facing binding terms, especially if that leader aims to delay elections. But more importantly, the victors in war often overthrow the norms, rules, and institutions that previously guided the regular transfer of power. The decision to initiate conflicts also opens the door to political possibilities that may otherwise be deemed too harsh during peacetime. This logic implies that if term limited leaders harbor ambitions to overturn or otherwise thwart the limits to their tenure, doing so under the cover of ongoing conflict may present the best opportunity. In conjunction, the audience cost, political bias, and fighting for survival theories imply the following hypothesis:

Hypothesis 1a: War bias hypothesis–Leaders with binding term limits will be more likely to initiate international conflict than leaders without binding term limits.

The costs associated with electoral accountability serve as a deterring device between leaders and foreign adventures. By mitigating the costs or sensitivity to them, term limits are expected to remove the disciplining effect elections have on conflict-seeking behavior. There may also be distinct benefits to initiating conflicts in final terms. Notwithstanding this line of argument, there is reason to suspect precisely the opposite. Term limits may remove incentives for initiating conflicts, prompting fewer incidences of international crisis.

Diversionary Expectations

As already noted, the literature on diversionary war generates alternative expectations about conflict situations as compared to selectorate-based arguments. In general, theories of diversionary war posit that a leader’s primary motive for involvement in international crises is to draw public attention away from domestic factors. The two causal mechanisms commonly associated with diversionary war—the scapegoat hypothesis and gambling for resurrection—both combine electoral motives with socioeconomic concerns of the electorate. The former suggests that leaders use conflict to divert focus from internal social, political, or economic problems, which are often the result of poor policies (Morgan and Bickers 1992). The latter argues that leaders anticipating being voted out of office relatively soon will rationally opt for a risky war, since even a small probability of victory may offer a boon to reelection chances (Downs and Rocke 1994). Diversionary activity is thus attributed to some incentive variable speaking to domestic discontent with the government.

If we give credence to the underlying mechanisms of diversionary war, then the imposition of term limits alters drastically our expectations about conflict initiation. Electability remains a driving force behind diversionary war. Anticipating that domestic issues will negatively impact their electability, heads of state respond by redirecting public attention outward. Diversionary conflict is the result. Binding term limits, however, relieve leaders from electability concerns. Likewise, the conflict-seeking inclination of heads of state should also be negated in final terms.

Hypothesis 1b: Diversionary negation hypothesis–Leaders with binding term limits will be less likely to initiate international conflict than leaders without binding term limits.

Hypothesis 1b takes diversionary motives as given and superimposes binding terms into the conflict calculus. The result is to produce a very straightforward hypothesis. However, the hypothesis does not specifically address motivation. By concentrating on the source of domestic discontent thought to provoke diversionary behavior, we should be better able to flesh out the implications term limits may or may not have in diversionary war settings. The remainder of this subsection delves into a specific domestic factor known to affect diversionary incentives, namely, the state of the economy.

Theories of diversionary war pervade the field of conflict studies. And rarely has so much sustained attention yielded such inconsistent results. Establishing a general relationship between domestic affairs and international conflict, while straightforward in theory, has proved highly elusive (Levy 1989; Leeds and Davis 1997). Early studies to hypothesize a link between regime type and diversionary behavior provided inconclusive results (Zinnes and Wilkenfeld 1971). As a consequence, scholars have tended to focus on the conditions under which diversionary conflicts are more likely to occur. Gelpi (1997) finds evidence that diversionary initiation of force is generally a pathology of democratic systems—one not shared by autocratic regimes. Conversely, Miller (1999) finds that democracies are less likely to escalate militarized disputes on account of diversionary impulses than are nondemocracies.

Work specifically examining the United States has proved both controversial and difficult to replicate in large-N analyses. As noted, Hess and Orphanides (1995) argue that incumbent US presidents have been more inclined to resort to force abroad during times of economic recession. And Russett’s (1990b) efforts similarly support the claim that US dispute involvement is negatively related to economic activity. Ostrom and Job (1986) review biannual, post–WWII data for the United States and find that the propensity to use major force is significantly related to an “economic misery” index. But many of these studies and their underlying empirics have been heavily criticized. The conclusions and suppositions they employ have elicited a host of efforts to replicate the findings by altering measures for key variables or repeating the analyses in different periods (e.g., James and Oneal 1991). The results of such efforts have been mixed at best and the United States is likely to be the exception rather than the norm, given its position of power.⁴

The relationship between conflict and economic activity continues to receive attention from students of international politics. Conclusions, however, are again inconsistent across studies.⁵ An early piece by Thompson (1982) finds no evidence to support the hypothesis that business cycles are systematically related to war initiation. Focusing his analysis on the United States, Fordham (1998) argues that high unemployment is more likely to prompt a diversionary incentive from left-leaning leaders, whereas right-leaning leaders favor diversionary action when facing higher inflation. According to Fordham, Republicans and Democrats face differing costs associated with the use of macroeconomic policy to combat inflation and reduce unemployment. More recently, Blomberg and Hess (2002) show that recessions do cause a substantial increase in the likelihood of external conflicts, but only in the presence of an internal conflict (e.g., ethnic conflict or genocide). Miller (1999) finds that changes in economic growth rates are negatively related to the level of force used in international militarized disputes, but the result is only significant in less democratic states. Tir (2010) finds weak support for the hypothesis that poor economic performance is associated with a higher likelihood of territorial conflict initiation. Oneal and Tir (2006) argue that economic conditions do impact the likelihood that a democracy, but not an autocracy, will initiate a fatal militarized dispute, but that the growth rates necessary to precipitate wars between democracies are rare.

While the approaches and results vary, the catalog of work on war and recessions has established the basic premise that recessions and poor economic activity tend to generate positive expectations for diversionary war. However, arguments about the state of the economy and the use of force rely critically upon the assumption that a leader’s grasp of power is acutely sensitive to economic factors. If we are to improve our understanding of the relationship between economic conditions and the resort to force, we need to better apprehend the degree to which the preservation of office depends upon economic conditions absent a dispute. Making precisely this point, Arena (2010) further notes that if readily observable factors can help determine the relative importance of economic conditions, then we can better isolate those cases where diversionary relationships are most likely to obtain compared to when they are not.

Fortunately, term limits offer a unique opportunity to probe more carefully the assumption that a leader’s fate is linked with the economy. The goal is to determine when a leader’s likelihood of retaining office is more or less dependent on extant economic influences. In a crude, yet effective fashion, term limits allow us to do this. When term limits are binding, the degree to which a leader’s probability of maintaining power depends on the economy is effectively zero. Conversely, when leaders are not in final terms of office, their electoral fate will be tied to the economy to varying degrees. Precisely how much, we are unable to say. But, when term limits are a binding constraint, a leader’s sensitivity to the state of the economy out of electoral concern is diminished to the point of irrelevancy. And when term limits are not a factor, the economy may be an influential issue in a leader’s election fate. This logic indicates that leaders facing binding term limits should be immune from the diversionary incentives typically precipitated by economic downturns.

Hypothesis 2: Immunity hypothesis–Leaders facing binding term limits will be no more likely to initiate conflict during periods of economic recession than during nonrecessionary periods.

Hypothesis 2 takes advantage of term limits as a binary proxy for a leader’s electoral sensitivity to the economy. Evidence in support of this hypothesis would indicate that diversionary motives are indeed conditioned on the need to win elections; the absence of the latter should be associated with the disappearance of the former, even in the face of a recession.

Testing the possibility that third parties anticipate the likely impact of term limits on a leader’s behavior in crisis situations invites a similar line of reasoning. That expectations should play a role in conflict is hardly controversial, but precisely how they may do so when term limits are a relevant factor is not obvious. Fortunately, we may take advantage of the “political bias” argument outlined earlier, by interpreting it from the perspective of potential challengers. The prohibition of reelection altogether (on account of term limits) affords a leader a certain freedom of response when challenged from abroad—an increased war bias. Term-limited leaders face no domestic electoral consequences from the potential high costs of warfare, which allows them to engage in aggressive foreign policy if they deem it necessary. Taken to the extreme, latitude to be reckless may be conducive for discouraging foreign provocations. In essence, the inflection of the political bias argument underlying Hypothesis 1a suggests that a heightened war bias effectively deters third parties from challenging term-limited leaders.

Relatedly, scholarship examining the explicit connection between the timing of leadership tenure and international crises has demonstrated that incumbents are more likely to participate in conflicts early in their tenure rather than later (Chiozza and Goemans 2003). Several studies argue that age and tenure are important factors in foreign affairs and that political immaturity is a liability to be exploited by targeting states. A leader’s experience in office tends to deter offensive advances from third parties. Inexperience, on the other hand, produces the opposite effect—it attracts foreign challenges. Gelpi and Grieco’s (2001) explanation for the deterring effect of experience is that resisting challenges becomes less costly the longer a leader has enjoyed the privileges of office. Focusing exclusively on the US case, Potter (2007) similarly posits that inexperience has an important explanatory role in the genesis of many American foreign policy crises. This is because young and inexperienced leaders may possess relatively weak management structures and administrations, thereby raising the likelihood of mishandling crises. Potter demonstrates that the accrual of time in office significantly lowers the chances that US presidents will be involved in militarized disputes and international crises. These expectations and results also accord with a formal model by Wolford (2007), where challenging states share an incentive to target newly elected leaders in order to gauge their resolve.

When tenure regulations permit more than a single term, leaders reaching last terms will be more practiced in the field of foreign affairs and likely possess greater diplomatic know-how than in prior terms. Taken together, the discussion on leader experience and the inflected political bias hypothesis generate the following hypothesis:

Hypothesis 3a: Experience hypothesis–Leaders facing binding term limits should less likely be targeted in international crises than are leaders without binding term limits.

In spite of these arguments and expectations, other, albeit weaker, evidence suggests the opposite conclusion. Horowitz, McDermott, and Stam (2005) find that the likelihood that a state will be challenged increases with the age of that state’s leader. This finding comports with work by Bak and Palmer (2010), which finds strong evidence that older leaders are significantly more likely to be the targets of militarized disputes.⁶ Bak and Palmer speculate this link might be due to varying levels of testosterone across leader ages. These findings alone represent somewhat of an empirical oddity, largely unsubstantiated by theoretical insight.

If the younger are typically more aggression prone, then potential challengers may avoid targeting young leaders out of fear of escalation. While a leader’s age is (trivially) increasing in the number of terms served, this age difference is not large, and such a gap does not necessarily exist when comparing leaders between countries or across time. Another possible explanation is that younger leaders and those earlier in their tenures may be more sensitive to the costs of concessions. If concessions to foreign challenges early in a leader’s tenure represent a greater domestic political liability, then young leaders may be more likely to meet aggression with strong resistance earlier rather than later in their tenure. Anticipating escalation, foreign regimes may be less willing to target leaders earlier in their elected tenure. These explanations are in no way intended to be exhaustive. It is sufficient to maintain that if third parties believe that term-limited leaders will have fewer predictable incentives to respond aggressively to challenges from abroad, then they will attempt to alter the status quo only once leaders in target states have reached final terms. While these arguments are only weakly motivated by existing theories, as a counterpoint to Hypothesis 3a, we offer the following competing hypothesis:

Hypothesis 3b: Attraction hypothesis–Leaders facing binding term limits should more likely be targeted in international crises than should leaders without binding term limits.

Permanent Referendum Theory

Work by Howell and Pevehouse (2005) notes that the subfield of American politics has developed ample theories on interactions between the executive and legislative branches. However, much of this work has ignored the international use of force as a dependent variable. As Howell and Pevehouse point out, there has been little to no cross-pollination between theories of executive and legislative interaction and the use of force.

Drawing on literature from US politics, there is equally consistent logic and evidence to suggest binding term limits ought not matter in world politics. Bruce Russett observed that the impact of elections on US dispute participation greatly decreased after 1930. He postulated that the advent of frequent polling and constant media coverage has produced “something akin to a permanent referendum on elected officials,” (1990a p. 134) where elected officials must align their politics with public opinion more or less throughout their entire tenure in office. Permanent referendum theory suggests the impact of a term limit on a leader’s behavior will be short-lived or nominal at best. Only the most parochial of elected officials will make international policy decisions with blatant disregard for extant domestic considerations. The scope of factors affecting executive decision making, including use of force, extends well beyond one-shot elections.

The theoretical basis for this view comes directly from strands of research in American politics, which establishes that electorates are indeed retrospective (Fiorina 1978) and that public support impacts a leader’s leeway of action (Neustadt 1960). This is because leaders operating within the confines of strong legislative institutions “share the public.” Popular prestige and public support are indeed precious forms of political capital, to be expended strategically for the purpose of obtaining congressional approval for various programmatic initiatives (Rivers and Rose 1985; Edwards 1990).

Investigations into referendum-voting models have explored the interactive nature of executive and legislative power. Efforts here demonstrate a strong and persistent link between the policies associated with the president’s administration and the outcome of midterm congressional elections. Citizen evaluation of the performance of the president is directly and positively related to the vote for candidates in the incumbent president’s party (Tufte 1975; Marra and Ostrom 1989). Executive policies during the first two years of an administration impact local elections. But this relationship is not strictly a one-way affair.

Midterm elections are among a host of variables shown to impact presidential actions. In fact, the constraints facing a president at any one time are more manifold than a basic “election-on, election-off” archetype. Work by Rhode and Simon (1985) suggests a leader’s ability and latitude for pursuing a political agenda is highly sensitive to both popularity as well as partisan control of Congress; this is especially true during times of economic hardship. The authors show that the five most salient factors in explaining the use of presidential veto power are public approval, congressional seat proportion, international conflict, midterm election year, and an economic–political context interaction term, but not the presidential election year covariate. Where congressional elections mattered, presidential ones did not.

The permanent referendum logic assumes all elections matter, more or less, not merely those choosing heads of state. Furthermore, it is based on the assumption that the broader electing public remains shared between heads of state and other elected officials, each possessing varying degrees of redistributive capacity. Consequently, a leader’s scope for action at any given time remains acutely sensitive to popular sentiments among a shared public. These premises and alternative distinctions lead to the following hypothesis:

Hypothesis 4: Permanent referendum hypothesis– Leaders with binding term limits will be no more or less likely to initiate international conflict than will leaders without binding term limits.

Empirical Analysis

To assess the role of term limits, we estimate the effect of the term limit variable on leader behavior or behavior of opposing actors toward these leaders in the realm of conflict. The universe of relevant cases for our investigation consists of all democratic regimes for which leaders potentially could face binding term limits.⁷ This excludes democratic leaders in regimes that do not have any regulations on the number of terms a leader can stay in office. Any effects estimated for this universe of cases only apply to leaders that operate within political institutions that feature term limits. We cannot extrapolate our findings to the behavior of leaders in other democracies, most often parliamentary regimes, without making strong assumptions about the comparability of such regimes. Furthermore, we base our analysis on a nonrandom sample of cases, which always raises the possibility of selection effects. However, our data cover a large part of the available universe of cases, and we have no reason to believe that sample selection was driven by factors affecting term limits or conflict initiation. A second and more important concern is the absence of random assignment of binding term limits. Our data (detailed in the following) record for each country-year whether a leader faces a binding term limit, that is, cannot run for reelection. According to this setup, binding term limit assignment does not follow a simple exogenous random process, but is rather determined by various observable and unobservable factors. While we cannot completely rule out that the presence of binding term limits is determined by unobserved factors that also drive the initiation of conflict, we do control for a number of important observable covariates and explore various strategies to account for dependence between observations.

Data Structure and Sources

Our data on term limits in democracies build upon that used in previous research on fiscal policy and term limits (Nogare and Ricciuti 2011). The term-limit variable is coded as 1 if in a given country-year the chief executive faces a binding term limit preventing him or her from running for reelection and as 0 otherwise. This information is largely based on the Database of Political Institutions (Beck et al. 2001). Nogare and Ricciuti go to great lengths to double-check mistakes in the original data and attempt a correct coding for the period 1977–2000. Nonetheless, we identified several cases for which the data set lists existing term limits (often for the president), while the actual chief executive can stand for reelection several times (often the prime minister). We excluded all cases for which the term limits only apply to ceremonial heads of state and heads of state with very limited powers. We also reviewed each country-year in the data set for accuracy via a process of fact checking and recoded various country-years. Additionally, we expanded the data set to include countries with binding term limits in Asia and Africa. Our data cover forty-eight democracies from 1976 to 2000.⁸ In all empirical estimations, we only use observations from that time period with a Polity2 score larger than five. As a robustness check, we also distinguish between permanent or “strong” term limits, those ruling out reelection after a fixed number of terms, and “weak” term limits that only prohibit consecutive terms.⁹

The specific unit of observation for our analysis of term limits and conflict will be the directed country-dyad-year. Increasingly, researchers have used directed dyads as the standard data structure for evaluating the strategic interaction of leaders in the international sphere. Directed dyads consist of country-year pairs that record the dyad characteristics of the sender and receiver, as well as dyad specific characteristics.

Two other papers have started to investigate the link between term limits and international crises. Haynes (2012) analyzes the effect of term limits on the credibility of threats within international disputes. He shows that term limits make presidents less effective in crisis bargaining, but have no discernible effect on crisis outcomes. A related working paper by Conconi, Sahuguet, and Zanardi (2010) engages the role of term limits for democratic peace. It argues that democratic leaders facing term limits should be as likely as nondemocratic leaders to be involved in interstate conflict. The authors use data on dyadic conflicts from 1816 to 2001, covering 177 countries. Their analysis relies on a nondirected dyadic structure, discarding information on sender and receiver. Their term-limit variable records whether any of the countries in a purely democratic dyad are term limit constrained. They find the presence of a binding term limit in a democratic dyad negates the democratic peace effect. Both sets of findings suggest promise for our investigation, but since their research focus is somewhat different, we believe for our hypotheses an alternative approach is preferable. We focus on the emergence of conflict, not on bargaining within conflict episodes, and analyze the processes of initiation and being targeted separately. Relying on a directed dyad design (and considering information on dyad members) is especially important for understanding the interaction between recessions and term limits.

Since the use of directed dyads often leads to very large data sets, applied researchers often focus on reduced samples that restrict attention to “politically relevant” dyads, as defined by major power status and contiguity. We follow this approach and furthermore only consider directed dyads from 1976 to 2000, the time frame for which information on term limits is available. Since we are not interested in evaluating broader theories of conflict, but rather only care about the effect of term limits on leader behavior, we additionally restrict our attention to dyads for which the initiator (or target) is a state with constitutionally proscribed term limits. This automatically excludes dyads between autocracies, semidemocracies, and full democracies without term limits from the sample. The goal is to create a sample in which the units with binding term limits are very similar on all other dimensions to the control units. Excluding nondemocracies and democracies without term limits as senders or receivers helps to increase comparability.

The statistical literature on international war has established that several variables significantly help explain the onset of conflict (e.g., O’Neal and Russett 1999; Bennett and Stam 2004). We include a set of standard controls. Specifically, in all models we include the Polity2 scores to measure the level of democracy, the relative military capabilities of sender versus receiver to identify dyads with military imbalances, major power status, and the physical distance between dyad members. Information on these covariates comes from the militarized interstate dispute (MIDS) data set (Ghosn et al. 2003).

To test the interaction between term limits and diversionary behavior during economic downturns, we also need additional information on the state of the economy. As discussed, one of the seminal contributions to the analysis of recessions and conflict is the Hess and Orphanides (1995) work on US conflict initiation. In their article, Hess and Orphanides use data on real output, industrial production, unemployment, and the Gallup Poll survey to date recessions for the United States. Unfortunately, data on variables other than output do not exist or are prohibitively difficult to obtain for the forty-eight countries we have identified, given the period we analyze. However, we are able to obtain real gross domestic product (GDP) data for these countries for the years 1976 to 2000. The Economic Research Service (of the USDA) publishes the “International Macroeconomics Dataset,” which includes annual real GDP data, reported in billions of dollars at 2005 prices.

Hess and Orphanides (1995) compute the change in output, ΔQ, for each year of their data set, where Q is real GDP. They then compare ΔQ in each year to the sample mean of the very same variable for the time period chosen. Thereafter, they create a recession dummy variable which is equal to 1 in each year that ΔQ was below average in the previous year, and 0 otherwise. Importantly, because the Hess and Orphanides approach merely reflects years of below average economic performance, it does not conform well to typical definitions of recessions. An additional problem with the Hess and Orphanides approach for capturing recessions is its high sensitivity toward bias. The comparison of ΔQ in each year to the sample mean over the whole sample period implies that just one unusually large value of ΔQ may bias the results, since a large value will raise the sample mean and increase the likelihood that any given year is counted as a recessionary observation. For these reasons, we employ alternative approaches to capture and reflect recessions, using data that are easily available.

The main method we use to date recessions follows an “output gap” approach. This idea is based on the notion that countries have some level of “potential” output that their factors of production are capable of producing at any given moment. One conventional way to compute potential GDP is using the Hodrick and Prescott (1997) filter on actual data for real GDP (e.g., see Gali and Gertler 1999). In essence, the HP filter is a mathematical tool that separates the cyclical component of a time series from its raw data, which effectively removes short-term fluctuations that are associated with the business cycle, thereby revealing long-term trends. Thus, we estimate the output gap by computing the HP detrended log of real GDP for the countries in the sample. Finally, we set our recession dummy variable equal to 1 for observations where actual output is below potential output (i.e., a negative output gap), and 0 otherwise. Alternatively, we also consider another popular way of classifying recessions. Shiskin (1974) suggested that a good rule-of-thumb indicator of recessions is two quarters of successive negative real economic growth. Given our real GDP data, our second recession dummy variable is equal to 1 if there is negative real economic growth in a particular year, and 0 otherwise.

Results

Turning to Hypotheses 1a and 1b, we analyze the effect of a binding term limit on the probability of initiating conflict in a dyadic setting. To test this specific premise, we create a sample of dyads in which the initiator belongs to our sample of democracies with potentially binding term limits and the receiver is of any regime type. Our dependent variable is the initiation of a militarized dispute.¹⁰

We first present results for a standard logit model with robust standard errors clustered on the dyad and a relogit model to correct for rare-events bias (King and Zeng 2001). We also consider an alternative term-limit dummy variable that includes only incidents of strong term limits and omits those cases in which political leaders can rerun for office after a specified waiting time (weak term limits). Finally, we also construct an ordinal variable that takes the value 0 for leaders able to run for office in the next term, 1 for leaders that cannot run for office the immediate next term, but potentially at future dates (the previously excluded “weak” cases) and 2 for leaders facing a permanent term limit.¹¹ This Term Limit ordered variable provides the opportunity to assess the logic of term limits at differing levels of intensity.

As noted, we employ the usual covariates from the conflict literature mentioned earlier as controls. Table 1 shows the coefficient estimates, standard errors, and p-values for the logit and relogit models.

Table 1.

Conflict Initiation and Term Limits.

	(1) Logit, clustered SE	(2) Relogit, clustered SE	(3) Logit, clustered SE	(4) Logit, clustered SE
Distance	−0.000278** (0.000106)	−0.000278** (0.000106)	−0.000273** (0.000105)	−0.000277** (0.000105)
Democracy receiver	−0.0589*** (0.0156)	−0.0591*** (0.0156)	−0.0586*** (0.0155)	−0.0589*** (0.0155)
Relative capabilities	−0.00178 (0.00119)	−0.00142 (0.00119)	−0.00180 (0.00121)	−0.00182 (0.00121)
Major power sender	0.625 (0.667)	0.607 (0.666)	0.447 (0.611)	0.565 (0.626)
Major power receiver	−0.333 (0.950)	−0.251 (0.949)	−0.400 (0.931)	−0.362 (0.936)
Term limit sender	0.738⁺ (0.416)	0.742⁺ (0.415)
“Strong” term limit sender			0.700⁺ (0.416)
Term limit (ordered) sender				0.390⁺ (0.205)
Constant	−4.155*** (0.540)	−4.126*** (0.540)	−3.981*** (0.495)	−4.099*** (0.503)
Observations	7381	7381	7381	7381
AIC	795.3	.	795.1	794.7
BIC	843.6	.	843.4	843.1

Note: AIC = Akaike information criterion; BIC = Bayesian information criterion; SE = Standard errors are given in parentheses.

⁺ p < .10. *p < .05. **p < .01. ***p < .001.

The results suggest that the distinction between term limited and reelection-seeking leaders has a statistically significant impact on the decision-making calculus of conflict initiation. The coefficient for the term-limit dummy has a positive sign and is significant below the 10 percent level for both the standard logit and the relogit model. Additionally, several control variables affect incidence of militarized disputes. Physical distance reduces the likelihood of conflict initiation, while interacting with a democratic country also reduces the chances of a militarized dispute. Notwithstanding the substantial effects of distance and democracy, the term limit impact is not only statistically significant but also substantive importance. Democratic leaders facing a binding term limit are on average 0.6 percentage points more likely to initiate conflict, compared to their reelection facing counterparts. Given the low unconditional baseline probability for conflict of 1 percent in the sample, the institutional constraint of term limits increases the likelihood of conflict by more than 50 percent. Similarly, the results are robust to using the alternative term-limit measures. When we reclassify cases in which leaders can rerun for office at later dates, we still obtain nearly identical estimates for the term limit effect. Our ordinal measure of term limits is equally positive and statistically significant below the 10 percent level. These results clearly provide support for Hypothesis 1a over 1b. Leaders reaching final terms appear to have a higher bias for conflict initiation.

A typical problem in the analysis of dyadic data is the dependence of observations. One sensible approach to deal with the dependence structure is the use of Bayesian hierarchical models. In particular, we use a limited dependent variable setup in which each dyad will have its own random intercept, drawn from a normal distribution. More specifically, the indicator variable y_ijt takes the value 1 if country i initiated an MID with country j in year t, and zero otherwise. The linear predictor η _ijt , which is linked to the probability of an MID via a canonical link function g(·), is a function of covariates in the following form:

η_{i j t} = α_{i j} + X' β + T_{i} + ϵ_{i j t},

where X ′β are dyad-specific covariates and T_i is the term limit dummy. The random dyad effects α _ij are drawn from a normal distribution, as are the observation specific error terms ∊ _ijt. The inclusion of dyad-random effects allows for partial pooling and can easily be integrated in a Bayesian context (Gelman and Hill 2007). Table 2 reports posterior modes and 95 percent highest posterior density (HPD) intervals for all variables.¹²

Table 2.

Conflict Initiation, Bayesian RE.

	Posterior mode	Lower 95 percent HPD	Upper 95 percent HPD
(Intercept)	−3.735	−4.430	−3.540
Distance	−0.000	−0.000	−0.000
Democracy receiver	−0.051	−0.055	−0.019
Relative capabilities	0.000	−0.001	0.000
Major power sender	−0.033	−0.571	0.388
Major power receiver	−0.141	−0.907	0.114
Term limit sender	0.476	0.194	0.761
N	7,381

Note: HPD = highest posterior density.

The Bayesian hierarchical model confirms the results from the standard logit approach. The posterior 95 percent HPD interval for the term-limit variable is clearly bound away from zero and positive. Accounting for random dyad effects, our confidence in the term-limit effect is even higher than in the classical models—with a larger than 95 percent posterior probability, term limits increase the likelihood of conflict initiation.

Overall, these findings suggest leaders are more likely to initiate conflict during final terms in office, once elections/reelections are safely behind them. The logit, relogit, and Bayesian hierarchical models all indicate that arriving in final and binding terms strongly increases the probability of conflict initiation. Our analysis of these cases dismisses any notion that the requirement of public approval alone should precipitate militarized involvements. On the contrary, we find that casting off the imperative of seeking public mandate affords leaders a free hand to pursue conflict initiatives. Or, as has been suggested, it may also render them less politically vulnerable to audience costs associated with bad outcomes or the failure to deliver good ones. Thus, leaders are more disposed to hazard war when the risk of losing office is no longer an issue. The consequence is that elections are indeed a restraint to bellicosity.

Hypothesis 2 states diversionary conflict may be precipitated by economic downturns. Therefore, leaders are more likely to initiate diversionary conflict during economic hardship, but only when reelection is a concern. If term limits have an effect on the decision making of leaders, we expect the role of recessions to change, conditional on the presence of a binding term limit. To assess this possibility, we include a dummy variable for the presence of a recession and an interaction term between a binding term limit and our recession dummy.¹³ Table 3 shows the results for the logit and relogit estimation including these two additional predictors.

Table 3.

Conflict Initiation and Term Limits.

	(1) Logit, clustered SE	(2) Relogit, clustered SE	(3) Logit, clustered SE	(4) Logit, clustered SE
Distance	−0.000281** (0.000106)	−0.000281** (0.000105)	−0.000277** (0.000104)	−0.000280** (0.000105)
Democracy receiver	−0.0613*** (0.0159)	−0.0615*** (0.0159)	−0.0608*** (0.0158)	−0.0612*** (0.0158)
Relative capabilities	−0.00180 (0.00120)	−0.00145 (0.00120)	−0.00183 (0.00122)	−0.00185 (0.00122)
Major power sender	0.614 (0.635)	0.582 (0.634)	0.424 (0.587)	0.533 (0.599)
Major power receiver	−0.328 (0.941)	−0.247 (0.940)	−0.397 (0.924)	−0.361 (0.927)
Term limit sender	1.216** (0.386)	1.208** (0.385)
Recession	0.721* (0.301)	0.703* (0.301)	0.659* (0.283)	0.707* (0.296)
Term Limit × Recession	−0.868* (0.413)	−0.850* (0.413)
“Strong” Term Limit Sender			1.164** (0.388)
“Strong” Term Limit × Recession			−0.897* (0.426)
Term Limit (ordered) Sender				0.631*** (0.191)
Term Limit (ordered) × Recession				−0.462* (0.218)
Constant	−4.564*** (0.507)	−4.498*** (0.507)	−4.338*** (0.493)	−4.477*** (0.479)
Observations	7,381	7,381	7,381	7,381
AIC	793.1	.	793.3	792.6
BIC	855.2	.	855.5	854.8

Note: AIC = Akaike information criterion; BIC = Bayesian information criterion. Standard errors are given in parentheses.

+p < .10. *p < .05. **p < .01. ***p < .001.

In line with diversionary expectations, the constituent coefficient for economic downturns has a positive effect on conflict initiation, when there are no binding term limits, that is, the term limit variable and interaction are both zero.¹⁴ This effect is statistically significant below the 5 percent level across all models. Furthermore, the finding is robust to alternative measures of recession. (The Supplementary Online Appendix reports results using the Shiskin measure). Importantly, the term-limit variable indicates a statistically significant and substantively meaningful effect during times of no recession, which is now even more statistically significant than in the preceding model. This confirms the results of the prior analysis; that is, electoral constraints reduce belligerence. The interaction term between term limits and the recession dummy is estimated to be negative and statistically significant at the 5 percent level, evidence of a meaningful interaction. (Similar results are obtained for the other term limit and recession measures.)

To correctly assess the substantive conditional effect of term limits, we simulate predicted probability differences in the two scenarios. Table 4 reports the mean effect of term limits and the associated 95 percent confidence intervals with and without a concurrent recession. Because an interaction term always implies a conditional effect for the conditioning variable, it is important to also assess the symmetric effect on the conditioning variable (Berry et al. 2003). Hence, the table also shows the effect of recessions with and without a binding term limit on conflict initiation.

Table 4.

Conditional First Difference Effects.

Main variable	Conditioning variable	Estimate	Lower 95 percent CI	Upper 95 percent CI
Term limit effect	Recession = 0	0.0103	0.0028	0.0215
Term limit effect	Recession = 1	0.004	−0.006	0.020
Recession effect	Term Limit = 0	0.004	0.0007	0.009
Recession effect	Term Limit = 1	−0.001	−0.008	0.008

Note: CI = confidence interval.

As is evident from the first row, in the absence of a recession, term limits increase the probability of conflict initiation by approximately 1 percentage point; that is, leaders who are unconstrained by concerns of the electorate are more likely to explore and pursue confrontational policies in the international sphere. Note that this effect is nearly twice as large compared to our estimation in Table 1. Conditioning on the presence of recessions sharpens our findings with regard to term limits. By contrast, leaders facing reelection are appreciatively more pacific absent a recession. During a recession, however, (second row) the joint impact is much more muted and is not statistically significant. Similarly, a recession increases the probability of conflict by about 0.4 percentage points for leaders that may seek reelection (third row). When the length of a leader’s tenure is still at the discretion of voters, recessions induce diversionary behavior. However, once the necessity of reelection is removed (fourth row), the effect turns negative on average, and again loses statistical significance.

These conditional effects tease out the role of electoral motives and economic downturns and imply important differences, depending on the leader’s constraints. Recessions only increase bellicosity if leaders face future elections. Without the link between the electorate’s preferences and the leader’s hold on office, economic downturns do not necessarily induce conflictual behavior, showing that the causal channel of recessions on conflict operates through an electoral link. Similarly, the vote-winning requirement reduces the probability of conflict initiation, but only when the economy is doing well. In times of recession, non-term-limited leaders are pushed to pursue more conflictual policies. These findings are highly consistent with existing theoretical models that anticipate a clear pacifying or restraining effect of elections.

To further probe the robustness of these results, we again implement a Bayesian hierarchical model with dyad random effects. Again, we find recessions have a clear positive effect on conflict initiation, while the constituent term limit effect is also positive. The interaction term is estimated to be negative. For each term, the 95 percent HPD interval does not include zero. Simulations based on this model for the conditional effects confirm the results of the classical estimation. (Detailed results can be found in the Supplementary Online Appendix.) These findings provide clear evidence in favor of Hypothesis 2, a qualified interpretation of the classical diversionary war argument. Importantly, the qualification is grounded purely in election motives. We are able to accept the premise that recessions during election-seeking cycles promote more conflicts. However, the interaction between binding terms and economic downturns shows that term limits can negate the diversionary effect of recessions. Conversely, the pacifying effects of electoral accountability only apply in nonrecessionary times. Our results indicate a rich contextual effect of institutional constraints, given the wider political and economic environment, and provide important points of connection between disparate strands of the literature.

Hypothesis 3a and 3b address the possibility that external actors anticipate a distinction between the behavior of leaders in final terms from those who are not. If this is indeed the case, then we should observe an effect of binding term limits on the probability of being targeted in an international conflict. To test this expectation, we analyze the effect of term limit variables on the probability of being targeted in dyadic conflicts using a similar approach as above. We create a sample in which the target in the directed dyad belongs to our group of countries with potentially binding term limits, while the sender can be of any regime type. In this case, being the target of an MID is the dependent variable. We include the same control variables as in the previous models. Table 5 shows the results for the logit and relogit models with robust standard errors clustered at the dyad level.

Table 5.

Conflict Target and Term Limits.

	(1) Logit, clustered SE	(2) Relogit, clustered SE	(3) Logit, clustered SE	(4) Logit, clustered SE
Distance	−0.000377*** (0.0000904)	−0.000374*** (0.0000904)	−0.000372*** (0.0000906)	−0.000375*** (0.0000901)
Democracy Sender	0.671** (0.248)	0.663** (0.248)	0.676** (0.258)	0.672** (0.253)
Relative Capabilities	−0.0168 (0.0163)	−0.0141 (0.0163)	−0.0172 (0.0170)	−0.0169 (0.0166)
Major Power Sender	0.855 (0.885)	0.865 (0.884)	0.814 (0.893)	0.831 (0.887)
Major Power Receiver	−0.468 (0.587)	−0.458 (0.586)	−0.583 (0.565)	−0.519 (0.566)
Term Limit Receiver	0.277 (0.435)	0.281 (0.435)
“Strong” Term Limit Receiver			0.168 (0.471)
Term Limit (ordered) Receiver				0.121 (0.227)
Constant	−9.580*** (2.325)	−9.478*** (2.322)	−9.503*** (2.332)	−9.535*** (2.321)
Observations	7,381	7,381	7,381	7,381
AIC	653.1	.	653.5	653.3
BIC	701.4	.	701.8	701.6

Note: AIC = Akaike information criterion; BIC = Bayesian information criterion; Standard errors are given in parentheses.

+p < .10. *p < .05. **p < .01. ***p < .001.

In contrast to the initiation case, term limits have no discernible, independent effect on the probability of being the target of an MID. The estimated coefficient is positive, that is, suggests that term limited leaders are more likely to be targeted, opposite to theoretical expectations based on Hypothesis 3a. Once we repeat the analysis using the Bayesian hierarchical model with dyad random effects, we find term limits have a clear positive effect on a leader's probability of being targeted (results available in the Supplementary Online Appendix). With more than 95 percent posterior probability, the parameter of our term limit dummy lies in the interval [0.338, 1.219]. Overall, the results indicate some mixed support for Hypothesis 3b, but given the theoretical and empirical ambiguities, future research will have to decipher this particular puzzle in more detail.

Taken together, our findings in support of Hypotheses 1a and 2 directly contradict the permanent referendum argument (Hypothesis 4). Despite the widespread notion of constant presidential attention to domestic public opinion polls, in the international setting we find that institutional constraints play a powerful role in shaping the behavior of leaders.

Discussion

Although the assumption that leaders behave so as to stay in power does not easily lend itself to direct verification, we have attempted to gain leverage on this premise by deriving falsifiable implications from it. The basic question directing this inquiry is whether a change in the election-seeking status of a leader impacts the initiation of international conflict as well as the likelihood of being a target of such conflict. Or simply put, do term limits matter in international relations? Cost- and benefit-based political bias rationales for conflict as well as diversionary theories of war suggest that term limits should matter—albeit for different reasons. However, the permanent referendum hypothesis casts doubt on such reasoning, positing that term limits alone should have only a nominal effect on conflict calculus, if any.

The short answer to the question above is that term limits do matter. Altering election constraints affects conflict in ways that make it difficult to sustain a permanent referendum theory. Term limits are a domestic political institution with the capacity to impact the realm of foreign affairs. The evidence comports with theories of office-seeking behavior, broadly speaking. When term limits interrupt or otherwise obviate the dictates of political survival, leader behavior alters in meaningful ways.¹⁵ We document a clear positive effect of binding term limits on the probability of conflict initiation. The effect is furthermore substantively meaningful. Term limited leaders are, on average, more than half a percentage point more likely to initiate conflict, which constitutes an increase of more than 50 percent relative to the baseline probability. Legacy issues and party loyalty may alter this result, and future work should seek to incorporate these impacts. While a change in term-limit status alters strongly the likelihood a leader will initiate conflict, there is less robust evidence suggesting it also makes him or her more likely to be a target of conflicts. Both the logit and relogit models reject Hypothesis 3b, which expects an increase in challenges to term-limited leaders. But the Bayesian hierarchical model offers support for this premise. If anything, these mixed results provide motivation for future research to investigate this matter in more detail.

Most importantly, a binding term limit interacts with and dampens the diversionary impact of recessions. Across all our models and several robustness checks, we consistently find a meaningful interaction between term limits and economic downturns. There are two observationally equivalent interpretations of these results. The immunity hypothesis suggests that not having to seek reelection mitigates the need for diversionary war during times of recession. Certainly, recessions do not favor incumbents during times of election. But in the absence of election motives, heads of state may dispense with diversionary endeavors. Alternatively, recessions may pose an obstacle to conducting wars engendered via higher “political bias” during final terms. This logic suggests that leaders with binding term limits find it harder to take their nations to war during times of economic hardship, an impediment they do not face during nonrecessionary periods. For instance, leaders may find it increasingly difficult to finance military conflict during times of economic downturn. Nonetheless, our findings indicate that the relationship between negative economic factors and decisions to entertain international crises is subtler than originally thought and is a puzzle that certainly merits more attention in the literature.

And finally, an implication of this study is that wars attributed to diversionary motives are quite distinct from those initiated from positions of political security. Although both are generally wars of choice, they each deserve independent scrutiny. First, unpacking diversionary war is a complicated process. Certainly, democratic leaders have a menu of options for responding to and diverting attention from recessions, beyond starting international crises. Moreover, economic downturns are not the only variable shown to produce diversionary incentives for international crises. Future work should aim at incorporating some of these considerations. Second, wars and disputes attributed to “political bias” mechanisms also warrant more investigation, including determining what types of conflicts (e.g. territorial) are more susceptible to this form crisis posturing, and if these conflicts are initiated out of a motivation to subvert domestic constraints such as term limits. Just as not all democracies are the same, neither are the wars they choose to conduct.

Footnotes

Authors’ Notes

A prior version of this article was presented at the Midwest Political Science Association Meeting, Chicago–April 22, 25, 2010. We are indebted to Chiara Nogare and Roberto Ricciuti for generously sharing their data on term limits. Thanks are owed to Raymond Mercado, Carmela Lutmar, seminar participants at Duke and Humboldt University of Berlin, and especially to Hein Goemans. We are grateful to the editor and anonymous JCR reviewers for their tremendously valuable recommendations. Remaining errors and shortcomings are ours alone. Data and code for replication of all results are available on the JCR website.

Declaration of Conflicting Interests

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

Funding

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

Notes

References

Arena

2010. “Why Not Guns and Butter: Responses to Economic Turmoil.” Foreign Policy Analysis 6 (4): 339–48.

Bak

Palmer

. 2010. “Testing the Biden Hypotheses: Leader Tenure, Age, and International Conflict.” Foreign Policy Analysis 6 (3): 257–73.

Beck

Thorsten

Clarke

George

Groff

Alberto

Keefer

Philip

Walsh

Patrick

. 2001. “New Tools in Comparative Political Economy: The Database of Political Institutions.” World Bank Economic Review 15 (1): 165–76.

Bennett

Scott

Stam

Allan

. 2004. The Behavioral Origins of War. Ann Arbor: University of Michigan Press.

Berry

William D.

Golder

Matt

Milton

Daniel

. 2012. “Improving Tests of Theories Positing Interaction.” Journal of Politics 74 (3): 653–71.

Besley

Timothy

Case

Anna

. 1995. “Does Electoral Accountability Affect Economic Policy Choices? Evidence from Gubernatorial Term Limits.” Quarterly Journal of Economics 110 (3): 769–98.

Besley

Timothy

Case

Anna

. 2003. “Political Institutions and Policy Choices: Evidence from the United States.” Journal of Economic Literature 41 (1): 7–73.

Blomberg

S. B.

Hess

G. D.

. 2002. “The Temporal Links between Conflict and Economic Activity.” Journal of Conflict Resolution 46 (1): 74–90.

Brambor

Thomas

Clark

William Roberts

Golder

Matt

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

10.

Chiozza

Giacomo

Choi

Ajin

. 2003. “Guess Who Did What: Political Leaders and the Management of Territorial Disputes, 1950-1990.” Journal of Conflict Resolution 47 (3): 443–67.

11.

Chiozza

Goemans

H. E.

. 2003. “Peace through Insecurity.” Journal of Conflict Resolution 47 (4): 443.

12.

Chiozza

Goemans

H. E.

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

13.

Conconi

Paola

Sahuguet

Nicolas

Zanardi

Maurizio

. 2010. “Democratic Peace and Electoral Accountability.” CREI Working Paper, Centre de Recrca en Economia Internacional, Barcelona. http://www.crei.cat/conferences/The_Political_Economy_of_Economic_Development/Paola_Conconi(1).pdf

14.

de Mesquita

B. B.

Morrow

J. D.

Siverson

R. M.

Smith

. 1999. “An Institutional Explanation of the Democratic Peace.” American Political Science Review 93 (4): 791–807.

15.

de Mesquita

B. B.

Siverson

R. M.

. 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.

16.

Mesquita

Bueno

Bruce

Smith

Alastair

. 2007. “Foreign Aid and Policy Concessions.” Journal of Conflict Resolution 51 (2): 251–84.

17.

Mesquita

Bueno

Bruce

Smith

Alastair

Siverson

Randolph M.

Morrow

James D

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

18.

Downs

Anthony

. 1957. An Economic Theory of Democracy. New York: Harper.

19.

Downs

G. W.

Rocke

D. M.

. 1994. “Conflict, Agency, and Gambling for Resurrection: The Principal-Agent Problem Goes to War.” American Journal of Political Science 38 (2): 362–80.

20.

Edwards

G. C.

1990. At the Margins: Presidential Leadership of Congress. New Haven, CT: Yale University Press.

21.

Fearon

James D.

1994. “Domestic Political Audiences and the Escalation of International Disputes.” American Political Science Review 88 (3): 577–92.

22.

Fiorina

M. P.

1978. “Economic Retrospective Voting in American National Elections: A Micro-analysis.” American Journal of Political Science 22 (2): 426–43.

23.

Fordham

1998. “Partisanship, Macroeconomic Policy, and US Uses of Force, 1949-1994.” Journal of Conflict Resolution 42 (4): 418–39.

24.

Gali

Gertler

. 1999. “Inflation Dynamics: A Structural Econometric Analysis.” Journal of Monetary Economics 44 (2): 195–222.

25.

Gaubatz

K. T.

1991. “Election Cycles and War.” Journal of Conflict Resolution 35 (2): 212–44.

26.

Gelman

Andrew

Hill

Jennifer

. 2007. Data Analysis Using Regression and Multi- level/Hierarchical Models. New York: Cambridge University Press.

27.

Gelpi

Chris

. 1997. “Democratic Diversions: Governmental Structure and the Externalization of Domestic Conflict.” Journal of Conflict Resolution 41 (2): 255–82.

28.

Gelpi

Chris

Grieco

Joseph

. 2001. “Attracting Trouble: Democracy, Leadership Tenure, and the Targeting of Militarized Challenges, 1918-1992.” Journal of Conflict Resolution 45 (6): 794–817.

29.

Ghosn

Faten

Palmer

Glenn

Bremer

Stuart

. 2003. “The MID3 Data Set, 1993–2001: Procedures, Coding Rules, and Description.” Conflict Management and Peace Science 21 (2): 133–54.

30.

Goemans

H. E.

2000. War and Punishment: The Causes of War Termination and the First World War. Princeton University Press.

31.

Haynes

Kyle

. 2012. “Lame Ducks and Coercive Diplomacy: Do Executive Term Lim- its Reduce the Effectiveness of Democratic Threats?” Journal of Conflict Resolution 56 (5): 771–98.

32.

Hess

G. D.

Orphanides

. 1995. “War Politics: An Economic, Rational-voter frame- work.” American Economic Review 85 (4): 828–46.

33.

Hinich

Melvin

Ordeshook

Peter

. 1970. “Plurality Maximization vs Vote Maximization: A Spatial Analysis with Variable Participation.” American Political Science Review 64 (3): 772–91.

34.

Hodrick

R. J.

Prescott

E. C.

. 1997. “Postwar U.S. Business Cycles: An Empirical Investigation.” Journal of Money, Credit and Banking 29 (1): 1–16.

35.

Horowitz

McDermott

Stam

A. C.

. 2005. “Leader Age, Regime Type, and Violent International Relations.” Journal of Conflict Resolution 49 (5): 661–85.

36.

Howell

W. G.

Pevehouse

J. C.

. 2005. “Presidents, Congress, and the Use of Force.” International Organization 59 (1): 209–32.

37.

Huth

P. K.

Allee

T. L.

. 2002. “Domestic Political Accountability and the Escalation and Settlement of International Disputes.” Journal of Conflict Resolution 46 (6): 754–90.

38.

Jackson

M. O.

Morelli

. 2007. “Political Bias and War.” The American Economic Review 97 (4): 1353–73.

39.

James

Oneal

J. R.

. 1991. “The Influence of Domestic and International Politics on the President’s Use of Force.” Journal of Conflict Resolution 35 (2): 307–32.

40.

King

Gary

Zeng

Langche

. 2001. “Logistic Regression in Rare Events Data.” Political Analysis 9 (2): 137–63.

41.

Leeds

B. A.

Davis

D. R.

. 1997.“Domestic Political Vulnerability and International Disputes.” Journal of Conflict Resolution 41 (6): 814–34.

42.

Levy

Jack S.

1989. “The Diversionary Theory of War: A Critique.” In Handbook of War Studies, edited by Midlarsky

Manus I.

, 259–88. Boston, MA: Unwin-Hyman.

43.

Marra

R. F.

Ostrom

C. W.

Jr . 1989. “Explaining Seat Change in the US House of Representatives, 1950-86.” American Journal of Political Science 27 (3): 541–69.

44.

Miller

R. A.

1999. “Regime Type, Strategic Interaction, and the Diversionary Use of Force.” Journal of Conflict Resolution 43 (3): 388–402.

45.

Morgan

T. C.

Bickers

K. N.

. 1992. “Domestic Discontent and the External Use of Force.” Journal of Conflict Resolution 36 (1): 25–52.

46.

Neustadt

R. E.

1960. Presidential Power. New York: Wiley.

47.

Nogare

Chiara Dalle

Ricciuti

Roberto

. 2011. “Term Limits: Do They Really Affect Fiscal Policy Choices?” European Journal of Political Economy 27 (4): 681–92.

48.

O’Neal

J. R.

Russett

. 1999. “Assessing the Liberal Peace with Alternative Specifications: Trade Still Reduces Conflict.” Journal of Peace Research 36 (4): 423–42.

49.

Oneal

J. R.

Tir

. 2006. “Does the Diversionary Use of Force Threaten the Democratic Peace? Assessing the Effect of Economic Growth on Interstate Conflict, 1921–2001.” International Studies Quarterly 50 (4): 755–79.

50.

Ostrom

C. W.

Job

B. L.

. 1986. “The President and the Political Use of Force.” American Political Science Review 80 (2): 541–66.

51.

Potter

P. B. K.

2007. “Does Experience Matter?” Journal of Conflict Resolution 51 (3): 351–78.

52.

Rivers

Rose

N. L.

. 1985. “Passing the President’s Program: Public Opinion and Presidential Influence in Congress.” American Journal of Political Science 29 (2): 183–96.

53.

Rhode

D. W.

Simon

D. M.

. 1985. “Presidential Vetoes and Congressional Response: A Study of Institutional Conflict.” American Journal of Political Science 29 (3): 397–427.

54.

Russett

1990a. “Economic Decline, Electoral Pressure, and the Initiation of Interstate Conflict.” In Prisoners of War, edited by Gochman

Charles S.

Sabrosky

Alan Ned

, 123–40. Lexington, MA: Lexington Books.

55.

Russett

B. M.

1990b. Controlling the Sword: The Democratic Governance of National Security. Cambridge, MA: Harvard University Press.

56.

Shiskin

Julius

. 1974. “The Changing Business Cycle.” New York Times, December 1, 222.

57.

Siverson

R. M.

1996. “Thinking about Puzzles in the Study of International War.” Conflict Management and Peace Science 15 (2): 113–32.

58.

Stoll

R. J.

, 1984. “The Guns of November: Presidential Reelections and the Use of Force, 1947-1982.” Journal of Conflict Resolution 28 (2): 231–46.

59.

Thompson

W. R.

1982. “Phases of the Business Cycle and the Outbreak of War.” International Studies Quarterly 26 (2): 301–11.

60.

Tir

Jaroslav

. 2010. “Territorial Diversion: Diversionary Theory of War and Territorial Conflict.” The Journal of Politics 72 (2): 413–25.

61.

Tufte

E. R.

1975. “Determinants of the Outcomes of Midterm Congressional Elections.” American Political Science Review 69 (3): 812–26.

62.

Wang

K. H.

1996. “Presidential Responses to Foreign Policy Crises.” Journal of Conflict Resolution 40 (1): 68–97.

63.

Wolford

2007. “The Turnover Trap: New Leaders, Reputation, and International Conflict.” American Journal of Political Science 51 (4): 772–88.

64.

Zinnes

Wilkenfeld

. 1971. “An Analysis of Foreign Conflict Behavior of Nations.” In Comparative Foreign Policy: Theoretical Essays, edited by Hanrieder

W. F.

, 167–213. New York: McKay.

Supplementary Material

Please find the following supplemental material available below.

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

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

0.00 MB

0.11 MB