The defining political institution of republican democracy is the election of government representatives. As Mill (1861) and others (e.g., Dahl 1989) state, it is the reelection incentive that makes leaders faithful agents of their constituents. Models of democratic politics and foreign policy typically rely on this principal--agent framework to explain the relationship between elected officials, their constituents, and foreign policy behavior (Bueno de Mesquita et al. 2003; Reiter and Stam 2002). A central premise underlying this framework is that democratic leaders have electoral incentives to represent, or at least appear faithful to, the interests and preferences of their constituents; otherwise, they face replacement. For example, Koch and Gartner (2005), in their analysis of legislative accountability and casualties, argue that casualties matter most when they are directly tied to a legislator's constituents. According to these models, elected officials are constrained in their actions by the potential reward of reelection, as well as the threat of removal from office for misbehavior.
However, not all executives are equally constrained by the reelection incentive. Some systems allow the leader to be reelected indefinitely, while others bar the executive from reelection after serving out their term, or terms. Furthermore, in some systems the executive, the opposition or both can call for elections, or even replace the executive without new elections, which creates a constant threat of replacement. For other leaders, replacement, barring a constitutional crisis of sorts, can only happen at a fixed, known point in time and only then by reelection. This suggests that democratic leaders do not face similar reelection incentives when trying to explain conflict behavior. Because of this variation in whether, when, and who can replace executives, I argue that scholars should account for the groups that can actually replace leaders and whether and when replacements can happen.
With the idea that not all democratic executives are accountable to the same types of principals, this article asks the question: how does variation in political accountability among democratic states, when coupled with the reelection incentive, affect the timing of conflict in democratic states? While a great deal of research examines whether domestic politics, in the form of incentives and constraints created by political institutions, levels of executive approval, or changing economic conditions, affect whether democracies use force, there is a paucity of research that specifically examines how the election cycle and the desire to retain office affects the timing of conflict behavior. Furthermore, much of this research does not distinguish between those executives that can and cannot run for reelection. Finally, and perhaps most importantly, this research often fails to account for the variation in who leaders are accountable to, that is, who are the principals that can directly remove them from office.
I argue that the type of electoral system is a crucial determinant of political accountability. Specifically, is the system party centered, candidate centered, or some combination of both (Norris 1997; Siavelis and Morgenstern 2008)? I argue that the degree to which parties or voters control the selection and oversight and removal of executives shapes the policy incentives that politicians confront as elections approach and subsequently shapes the state's conflict behavior. Building from the agency loss and political accountability literature, I argue that leaders in endogenous, party-centered systems are more constrained by their principals because of the multiple avenues of replacement and the greater ability of the principal to monitor and sanction the agent. (1) At the same time, leaders whose immediate principals are the electorate and who face fixed elections are less constrained given the difficulty of removal and the greater difficulties in monitoring the agent. Subsequently, executives in these systems are more likely to engage in conflict or escalate a conflict as the election draws near. (2)
Not accounting for the different principal--agent relationships among political systems and the variation in the reelection incentive of executives may explain why studies examining the role of elections and conflict are contradictory in terms of their results. While some research argues that democratic leaders are less constrained immediately following an election and become more constrained as elections near (Auerswald 1999; Gaubatz 1991; Chiozza and Goemans 2003; 2011), others argue that as elections near, conflict is likely to increase (Hess and Orphanides 1995). Furthermore, others state that there is no relationship between elections and conflict (Gowa 1998). Thus, part of the answer to explain these contradictory results is to understand how the electoral system shapes the principal--agent relationship between executives (agents) and principals (either voters or parties). To address these concerns, I first review research on leader tenure, election cycles, and conflict. I then develop a theoretical framework largely drawn from the literature on agency loss in comparative politics and construct a model of conflict behavior that accounts for how the variations in agency loss affect conflict behavior. I test my expectations on a data set of term lengths and election cycles to account for factors that help determine whether a political system is candidate or party centered and when an executive can run for reelection. The results suggest that the ability of principals to monitor and sanction their agents influences the conflict behavior of democratic leaders across election cycles. Specifically, executives up for reelection in candidate-centered systems are more likely to use conflict as an election nears than those in party-centered systems.
Democratic Political Incentives and Conflict Behavior
Much of the research on unpacking democratic politics and conflict behavior either implicitly or explicitly focuses on the ex ante constraints that democratic executives confront prior to engaging in conflict or the ex post constraints leaders potentially face once a conflict is underway. Furthermore, much of the prior research considers these constant across time as opposed to varying over time. (3) Ex ante constraints affect the ability of leaders to enact policies and engage in conflict by increasing the number of actors involved in the policy making process. Much of the early work on unpacking democracies focused on these constraints. These actors include the number of parties in and out of government (Ireland and Gartner 2001) the number of institutional players such as legislatures that are involved in the decision-making process (Auserwald 1999; Clark 2000; Leblang and Chan 2003) or the degree to which the public participates in politics as measured by the degree of suffrage in a state (Reiter and Tillman 2002).
More recent research emphasizes ex post constraints, examining how a leader's political security affects decisions over when and whether to engage in international conflicts (Bueno de Mesquita et al. 2003; Chiozza and Goemans 2003; 2011; Koch 2009; Williams 2013, Wolford 2007). This research suggests that the political insecurity of leaders is a key driver in the foreign policy behavior of states. For the most part, research on leader tenure and conflict focuses on the comparison between democracies and autocracies. Bueno de Mesquita et al. (2003), for example, explain that because of smaller winning coalitions and larger loyalty norms, authoritarian leaders are more secure and subsequently less constrained than their democratic counterparts. Subsequently, these leaders are more likely to engage in riskier foreign policies, given that they are unlike to lose office if the outcome is less than positive. Most of the studies suggest that the more insecure a leader, the more constrained they are likely to be in their foreign policy behavior.
In terms of the election cycle and conflict research, those that focus on ex post accountability suggests that democratic leaders who have just assumed power are freer to initiate conflict and that as the next election draws near political vulnerability increases and conflict behavior declines. Gaubatz (1991), in his now seminal work on election cycles, shows that democratic leaders are more likely to initiate conflict early in the election cycle. One explanation he offers for this is that prior to an election, the relative power between the masses and government shifts in favor of the masses. Therefore, leaders may move towards more peaceful policy positions, promising to keep their country out of war prior to the election. However, after the election, and well before the next election, the balance of power shifts back to the government, lifting electoral constraints on more aggressive conflict behavior from executives.
Auerswald (1999) also argues that conflict is more likely shortly after an election. Building from the idea that voters are myopic in their evaluation of public officials, he argues that voters' beliefs are activated by recent events, and they discount temporally distant events. Leaders of democracies, therefore, may delay using force until after an election to minimize the possibility of electoral punishment. In other words, democratic leaders are less constrained in their conflict behavior early in their tenure, because they think they will have plenty of time to make up for potential failures and avoid electoral punishment in the next election. Along these same lines, Chiozza and Goemans (2003) argue that as the probability of losing office increases, the lower the likelihood of conflict initiation, suggesting that democratic leaders are most secure and less constrained immediately after an election. More recently, they suggest that democratic leaders are unlikely to gamble...