Lame Ducks and Coercive Diplomacy

Abstract

The literature on “audience costs” in International Relations suggests that a democratic leader’s electoral accountability lends him or her a significant advantage in crisis-bargaining situations. But if elections are the primary source of a democratic leader’s accountability, a democracy’s advantage in crisis situations should disappear when its leader is no longer eligible for reelection. Accordingly, this article asks whether “lame-duck” presidents, those who are constitutionally prohibited from reelection, are systematically less effective than their reelectable counterparts in crisis-bargaining situations. Using a data set of all post–World War II militarized interstate disputes initiated by presidential democracies, the author tests several hypotheses regarding lame-duck crisis behavior derived from Fearon’s bargaining model. The findings generally support the audience costs model. Observed patterns of threat reciprocation and crisis hostility are consistent with Fearon’s predictions. Crisis outcomes, however, do not appear to be affected by a democratic leader’s lame-duck status.

Keywords

audience costs domestic politics bargaining elections foreign policy international crises

The conclusion that democratic states are uniquely effective bargainers in international crises has gained significant traction in the field (but cf. Weeks 2008; Downes and Sechser 2012). James Fearon’s pioneering work, the basis for much of this literature, argued that the ability to generate “audience costs” enhances the capacity of democratic leaders to effectively and credibly commit to publicly stated bargaining positions (Fearon 1994; see also Smith 1998; Guisinger and Smith 2002). Having staked their country’s honor and their own political reputation in a dispute, democratic leaders are able to convince their adversaries that they cannot compromise lest their angered, embarrassed, or disillusioned constituents subsequently vote them out of office. Early empirical work supported Fearon’s basic argument by demonstrating that democracies bargain and signal more effectively than do nondemocracies during interstate disputes (Eyerman and Hart 1996; Partell and Palmer 1999). More recent work, however, has conclusively shown that all democracies are not equally capable of generating audience costs (Brule, Marshall, and Prins 2010; Prins 2003). This variation across democracies has opened new avenues for testing the audience costs hypothesis and revealed important patterns that are of intrinsic interest to scholars studying democratic foreign policy behavior.

For Fearon, the democratic advantage in crisis-bargaining situations results primarily from the accountability of democratic leaders to their constituents. Elections are the vital source of this accountability, as they enable the voting population to punish unpopular or incompetent leaders by removing them from office. The audience costs logic would thus argue that elections are essential to democratic credibility in international bargaining situations. As such, the effectiveness of a democracy’s threats during interstate bargaining should systematically vary as its leaders become more or less vulnerable to electoral punishment (Leeds and Davis 1997). This article tests this basic proposition, asking whether democratic chief executives who are prohibited from reelection by constitutional term limits are stripped of the bargaining advantages normally enjoyed by democratic leaders.

This article examines the effect of a president’s “lame-duck” status on democratic crisis behavior and dispute outcomes. I define a lame duck as a democratic chief executive who is constitutionally prohibited from contesting the next election for his or her current office. Lame ducks are, by definition, personally invulnerable to electoral punishment and are consequently inhibited from generating domestic audience costs to bolster their commitments during crisis negotiations. This should produce significant variations in the type and effectiveness of threats issued across lame duck and reelectable presidents. Statistical analysis of militarized disputes initiated by presidential democracies strongly supports this intuition. I show that lame ducks issue less effective threats because their targets, doubting the lame duck’s credibility and resolve, are more likely to resist their demands. I also demonstrate that crises initiated by lame ducks are less likely to escalate to higher levels of hostility because lame ducks often back down and de-escalate the dispute in response to a target’s resistance. The outcomes resulting from crises initiated by lame ducks, however, do not significantly differ from those initiated by electorally accountable presidents. These findings broadly support Fearon’s audience costs hypothesis, utilizing a more valid measure of a democratic leader’s susceptibility to audience costs than previous tests. Broadly, the article finds that electoral cycles and constitutional electoral laws can exert a significant influence on crisis bargaining in presidential democracies (Gaubatz 1991; but cf. Leeds and Davis 1997; Potter 2007).

The remainder of this article will review the relevant literature, critiquing prior empirical tests of Fearon’s model and demonstrating the need for renewed examination. I then derive testable hypotheses regarding the effects of constitutional term limits and lame-duck leadership on interstate crises. I test these hypotheses using a data set of post–World War II (WWII) militarized interstate disputes (MIDs) initiated by presidential democracies. I also present a brief case study of the 1989 US invasion of Panama to illustrate the key audience costs dynamics of lame-duck crisis bargaining.

Audience Costs and Crisis Bargaining: Theory and Tests

The audience costs model follows much of the classic bargaining literature in arguing that the state which is better able to credibly and irrevocably commit to its position will generally compel its adversary to concede (Schelling 1960, 1966). A state that credibly commits to its position effectively raises its own costs of making concessions, potentially to the point that it would prefer to fight rather than concede. In this sense, committed states are “tying their hands” and confronting the adversary with the unpleasant choice of either making concessions or escalating the dispute to war (Fearon 1997). Because the high costs of war make it an extremely inefficient means of settling the dispute, a target facing a highly committed challenger is likely to make the concessions necessary to meet its adversary’s demands (Fearon 1995).

Under the logic of audience costs theory, although an international reputation for resolve during crisis bargaining is potentially important, the credibility of threats or commitments are determined in isolation from one another (Press 2005; but cf. Sartori 2005; Sechser 2010). If analyzed from the perspective of the policy makers for whom retaining office is an important if not paramount concern, the domestic political consequences of international crises are extremely pertinent (Bueno de Mesquita and Siverson 1995). Fearon’s most important contribution was clarifying specifically how publicity and domestic political audiences can serve as a vital source of credibility in interstate bargaining. Democratic leaders, Fearon argued, are significantly more vulnerable to domestic backlash in the wake of foreign policy failures and are thus more able to generate audience costs.

Audience costs theory argues that democratic publics tend to view policy makers that initiate or escalate international crises only to subsequently back down as either incompetent or having sacrificed the nation’s honor and reputation (Tomz 2007). This negative public response combined with institutionalized mechanisms of government accountability mean that leaders of democratic governments are more likely to lose office following an unpopular policy choice. Therefore, policy makers with office-seeking preferences will rarely retreat from public threats and commitments. Because target states can observe the public commitments made by a democratic leader and understand the domestic vulnerability created by this publicity, threats and commitments made by democracies are more likely to be seen as credible.

The audience costs model and its extensions have been subjected to numerous empirical analyses. Early examinations focused on Fearon’s basic assertion that democracies as a whole are better able to generate audience costs and should thus prove more effective bargainers in interstate crises. Many of these tests used either binary measures of democracy (Eyerman and Hart 1996) or disputants’ relative levels of democracy (Partell and Palmer 1999) as proxies for the susceptibility of leaders to audience costs. Although useful first cuts, the reliance upon measures of democracy alone to capture audience costs is unsatisfactory. More recent scholarship has demonstrated that institutional variation across democracies can significantly affect the ability of democratic leaders to generate audience costs and bargain effectively in interstate disputes. These varying institutional constraints have been argued to arise from a multitude of potential factors. These include constraints on executive decision-making authority (Partell 1997), legislative constraints on foreign policy (Choi 2010; Reiter and Tillman 2002), public and intralegislative constraints (Reiter and Tillman 2002), and the competitiveness of political participation (Prins 2003). This work makes a strong case that variation across different forms of democratic institutions affect the ability to generate audience costs.

Kenneth Schultz’s extremely influential book performs similar tests (Schultz 2001a). While Schultz’s argument ultimately differs from Fearon’s in its specification of opposition parties as the vital signaling mechanism, Schultz’s model still relies upon elections as the crucial source of punishment following a foreign policy failure. For Schultz, elections are necessary but not sufficient to signal audience costs. In his large-N empirics, Schultz, like the authors just mentioned, relies upon largely static dimensions of democratic governance in his operationalization of the independent variable—specifically the competitiveness of executive recruitment and the regulation of political participation. Yet, even an emphasis on varying institutional constraints fails to capture the full extent to which democratic leaders vary in their vulnerability to punishment by domestic audiences.

David Clark and Timothy Nordstrom importantly demonstrated that democratic constraints can be either structural, and thus relatively constant over time, or dynamic, and thus persistently in flux. Dynamic constraints are those that “change over time as the result of the normal political process” (Clark and Nordstrom 2005, 255). Several tests have examined the effects of different types of dynamic constraints on democratic foreign policy behavior. Christopher Gelpi and Joseph Grieco argued that a democratic leader’s tenure in office has a significant effect on the probability that it will be targeted by another state’s challenge (Gelpi and Grieco 2001). David Brule and his colleagues showed that economic growth, presidential approval, and congressional support all influence American dispute behavior and effectiveness (Brule, Marshall, and Prins 2010). Brandon Prins and Christopher Sprecher show that both government coalition type and legislative polarization strongly influence the effectiveness of democratic threats (Prins and Sprecher 1999). Finally, David Leblang and Steve Chan examined the effect of factors such as unified government and upcoming elections on democratic war involvement (Leblang and Chan 2003).

These quantitative tests, combined with experimental and qualitative analyses (Auerswald 2000; c.f. Trachtenberg 2012; Brown and Marcum 2011; Tomz 2007), form the empirical backing for audience costs theory. Yet, too often these tests examine constructs that do not truly capture a democratic leader’s vulnerability to electoral punishment. Proxies such as the constraints on the executive, government coalition type, and legislative control indicate constraints on a leader’s decision-making capabilities during the crisis, not that the leader is more or less vulnerable to punishment by domestic audiences after the fact. Indeed, constraints on an executive’s decision-making authority may in fact decrease vulnerability to subsequent electoral punishment by spreading the responsibility for failed policies over a broader subset of government officials. This article addresses these shortcomings, testing Fearon’s audience costs model with a specification of the independent variable that more effectively captures a democratic leader’s susceptibility to domestic punishment. It uses one type of Clark and Nordstrom’s dynamic constraints, focusing on variations in the vulnerability of democratic chief executives to electoral punishment arising from constitutional term limits. The enhanced validity of this specification of the independent variable will in turn produce a more valid test of the relevant theoretical propositions, increasing our confidence in results that are often undermined by confounding factors and selection effects (Schultz 2001b).

Theoretical Propositions and Testable Hypotheses

Elections are the vital mechanism through which democratic publics can punish inept or unpopular leaders and are thus crucial to the democratic advantage in generating audience costs. Indeed, Fearon argues that the democratic bargaining advantage derives from the fact that “democratic leaders can more credibly jeopardize their tenure before domestic audiences” (Fearon 1994, 587). To be sure, democratic leaders are subject to substantial nonelectoral domestic audience costs. Fearon even asserts elsewhere that the “responsiveness” of democratic leaders need not arise from future elections and the prospect of retrospectively voting publics expelling them from office (Fearon 1999). Negative public opinion deriving from foreign policy failures can inhibit the pursuit of a leader’s objectives irrespective of electoral constraints by decreasing the incentives for other public officials to act cooperatively. Policy failures may injure a leader’s entire party, compromising the political prospects of the political organization in which the leader retains a substantial interest even after leaving office. Additionally, policy makers are often extremely concerned with their legacy and the judgment of history. The need to maintain or generate a positive legacy may impose significant constraints upon a democratic leader that is no longer subject to electoral sanction.

These alternative forms of vulnerability to domestic audiences notwithstanding, the possibility of electoral punishment remains the most direct and important means through which democratic publics can hold leaders accountable. This article focuses on one clear source of variation in the vulnerability of democratic leaders to electoral punishment—constitutional term limits imposed upon chief executives in presidential democracies. Reelectable and lame-duck presidents differ systematically only in their individual vulnerability to electoral punishment. Alternative specifications of audience costs often systematically covary with other variables that affect the outcomes of interest. This is significantly less likely under the present analysis, as the regular transitions between lame-duck and reelectable presidents allow most potentially confounding factors to be held constant. This minimizes the possibility of omitted variable bias and endogeneity influencing the results.

The empirical portion of this study focuses on presidential democracies. The mechanisms of electoral accountability in presidential and parliamentary systems of democracy differ in two important ways (Lijphart 1992). First, presidential systems exhibit a clear separation of the legislative and executive branches of government. Presidents, unlike prime ministers, are not members of the legislature. Significantly, this means that presidential power is not contingent upon legislative support or confidence. Prime Ministers are perpetually subject to votes of no confidence in the legislature and are thus always immediately vulnerable to electoral sanctioning for perceived policy failures. Because prime ministers are constantly subject to a vote of no confidence, there is very little systematic variation in their vulnerability to electoral punishment. Conversely for presidents, who are subject only to intermittent elections at predetermined points in time, the immediacy of electoral constraints may vary.

A second source of variation between presidential and parliamentary systems arises from constitutionally imposed term limits. Prime ministers are almost never constitutionally prohibited from retaining office for a virtually indefinite period of time, provided they maintain the support of their party and the broader population. Most presidential democracies, on the other hand, have strict constitutional limits placed upon the number of terms one can serve as chief executive. Because presidents have constitutionally imposed limitations placed on their tenure, there are significant periods of presidential leadership during which the chief executive is unambiguously invulnerable to electoral sanctioning. This is often referred to as the “last period problem” in democratic politics (Zupan 1990; Carey 1996).

The logic of audience costs theory would have us expect that lame-duck presidents, those who are constitutionally prohibited from seeking reelection in the next scheduled election for the presidency, are significantly less able to generate domestic audience costs. Adversaries in interstate disputes, aware of a lame duck’s invulnerability, should be more likely to discount the commitments made by lame-duck presidents. Since their commitments are systematically less credible than those of a reelectable president, lame ducks should be largely stripped of the advantage democracies typically hold in crisis-bargaining situations.

Hypotheses on Threats and Crises Initiated by Lame - Ducks

The discussion above yields several testable hypotheses regarding threats made and crises initiated by lame-duck presidents. First, because lame-duck presidents are systematically less able to generate audience costs during interstate crisis bargaining, targets should be more likely to resist threats made by lame ducks. States targeted by a presidential democracy’s threats should be more likely to believe that the initiator will back down when it is led by a lame duck. The domestic political ramifications of making such concessions are simply less damaging for lame-duck presidents than for their reelectable counterparts. Believing that the democratic initiator is less resolved, the target should be more likely to resist the threat. This yields Hypothesis 1:

Hypothesis 1: Threats issued by lame duck presidents are significantly more likely to be resisted by target states than threats issued by reelectable presidents.

Second, the ultimate outcome resulting from a crisis should be significantly less favorable to the democratic initiator when it is led by a lame duck. Lame ducks should face greater resistance to their threats, irrespective of their actual resolve, because target states clearly perceive the lame duck’s reduced domestic constraints. Additionally, lame ducks should be more likely to back down even when they are resolved at the outset of the crisis. Their invulnerability to electoral punishment removes one of the key impediments that inhibit electorally accountable leaders from making concessions to an adversary. Consequently, the political outcomes arising out of crises initiated by lame ducks should be less favorable to the initiator than those initiated by reelectable executives. This yields Hypothesis 2:

Hypothesis 2: Crises initiated by lame-duck presidents will produce an outcome significantly less favorable to the initiating state than crises initiated by reelectable presidents.

The audience costs model is ultimately indeterminate with respect to the effect of lame-duck status on crisis hostility levels. An examination of the effect of lame-duck status on crisis hostility levels therefore cannot be taken as a true “test” of audience costs theory, as the theory in question could plausibly predict both a positive and a negative relationship. Conducting these tests, however, will help determine which of the theory’s countervailing mechanisms are most influential during crisis-bargaining situations.

Fearon makes two critical claims that bear on the effect of audience costs on crisis hostility. First, in what I call the “signaling hypothesis,” he claims that states facing low audience costs will be required to escalate to higher levels of hostility in order to demonstrate resolve. Fearon argues that “when large audience costs are generated by escalation, fewer escalatory steps are needed credibly to communicate one’s preferences” (Fearon 1994, 585). According to this logic, crises initiated by lame-duck presidents should escalate to higher levels of hostility compared to those initiated by reelectable presidents. Lame ducks simply need to work harder to demonstrate resolve. Additionally, as Hypothesis 1 argues, because the target state is more likely to doubt the credibility of a lame duck’s threats, the target will be more likely to resist and escalate the crisis. This logic yields Hypothesis 3A, or the signaling hypothesis:

Hypothesis 3A: Crises initiated by lame-duck presidents will escalate to higher levels of hostility than crises initiated by reelectable presidents. Both the lame-duck initiator and the target should exhibit higher levels of hostility.

Fearon’s second claim, what I call the “selection hypothesis,” contends that states facing low audience costs are more likely to initiate crises casually or recklessly, knowing that they can later back down and de-escalate with few domestic costs (see also Chiozza and Goemans 2003; Gelpi and Griesdorf 2001, 636). Fearon (1994, 585) writes that “if democratic leaders tend to face more powerful domestic audiences, they will be significantly more reluctant than authoritarians to initiate ‘limited probes’ in foreign policy.” This same logic would predict that lame ducks are more likely than reelectable presidents to make limited probes. These types of crises would be less likely to escalate, as the initiator is hoping primarily to glean information or secure a cheap victory through minor threats and limited escalations. Essentially, leaders facing low audience costs are less constrained and thus select themselves into initiating crises despite a relative lack of resolve. And because lame ducks face lower domestic costs for making concessions, they should de-escalate crises more quickly and more readily than electorally accountable presidents. The target’s hostility level, however, should either increase or be unaffected. This yields Hypothesis 3B, or the selection hypothesis:

Hypothesis 3B: Crises initiated by lame-duck presidents will exhibit significantly lower overall levels of hostility than crises initiated by reelectable presidents. The lame-duck initiator, not the target, will account for this reduced hostility.

The next section elaborates upon the operationalization of the relevant variables, describes the data set, and specifies the statistical models through which these hypotheses will be tested. Statistical analysis follows, presenting the results and discussing the theoretical and substantive implications of the findings. The Empirical section concludes with a brief illustrative case study of the crisis leading up to the 1989 US invasion of Panama.

Empirical Analysis: Data and Concepts

Statistical analysis of the hypotheses above relies on the Correlates of War (COW) project’s MID data set, version 3.1 (Ghosn, Palmer, and Bremer 2004). The unit of analysis is the dyadic MID, with one initiating state and one target state per observation. MIDs include all “cases in which the threat, display or use of military force short of war by one member state is explicitly directed towards the government, official representatives, official forces, property, or territory of another state” (Jones, Bremer, and Singer 1996). This definition is nicely suited for this project, as it refers to explicit and overt threats that will expose leaders to potential audience costs. The MID data set also records significant information on the target’s response and the outcome of the dispute.

I compiled a data set of all post–WWII (1946–2001) MIDs initiated by democracies in which the executive head of government’s tenure is not subject to legislative confidence. To extract these cases from the universe of post–WWII MIDs, I relied on Jose Antonio Cheibub’s coding of democracies as presidential, parliamentary, and semipresidential (Cheibub 2007). I extracted all post–WWII country years that Cheibub coded as presidential democracy. I also extracted the country years in which countries coded as semipresidential democracies operated under constitutions in which the president was the dominant actor in foreign policy and was not vulnerable to a legislative vote of no confidence.1 I further parsed the cases according to the Polity IV measure of aggregate democracy (Marshall and Jaggers 2002; Marshall et al. 2002). Of the cases specified earlier, I included only those in which the Polity IV aggregate democracy (democracy minus autocracy) score is 5 or greater. I then compiled a data set of all MIDs initiated within these country years.²

This process resulted in a population of 212 dyadic MIDs. Importantly, this includes multilateral disputes, which are often excluded from analysis because of the potentially confounding effects of late joiners, multiple initiators, and multiple target states. In this overall population of MIDs, each initiator and target was included as a separate participant, producing multiple dyadic MIDs from each multilateral dispute. These cases must be treated carefully. Given the already truncated data set, I opted not to simply exclude MIDs with multilateral initiating coalitions. Instead, to ensure robustness of the results while not blindly excluding potentially valuable data points, I tested each statistical model against four different populations of observations parsed according to the criteria below.

The first specification, labeled “model 1,” uses every dyadic MID falling under the criteria delineated earlier, allowing all cases in which there were multiple initiating and target states in a single crisis. The model 1 data set includes all 212 MIDs identified earlier. The inclusion of certain multilateral disputes is highly problematic. North Atlantic Treaty Organization’s 1999 campaign against Serbia, for example, saw thirty-eight states on the initiating side overall, and four that were included in this data set. In such cases, the effect of individual countries on the initiating side is difficult to parse out. Nonetheless, this data set maximizes the number of observations for analysis and may present a useful empirical baseline.

The second specification, labeled “model 2,” is reduced to include only one initiator from each dispute in which the initiating coalition was three or greater. For MIDs in which coalitions of three or more states issued a threat, the single most powerful initiating state alone is coded as the initiator, while the target state/states remains unchanged. MIDs with multilateral initiating coalitions are thus treated as a dyadic dispute between the target/targets and the strongest member of the initiating coalition. Model 2 does not affect disputes with multiple targets, and each individual target is included as a separate observation. This produces a data set of 199 bilateral MIDs.

The third specification, labeled “model 3,” includes only those MIDs in which there were one or two states on the initiating side. Model 3 again codes disputes with multiple targets as a series of bilateral disputes against each individual target. This yields a population of 173 observations. This specification was chosen because a target state should still be able to assess the importance of audience costs incurred by each initiator in the crisis when it is targeted by only two other states. As the number of initiators grows beyond two, it becomes increasingly difficult for a target to determine a threat’s credibility based on the electoral incentives of a single state in the threatening coalition.

Model 4 represents the most truncated data set and includes only strictly bilateral disputes. All disputes that, at any point, involved multiple states on either the initiating side or the target side are excluded. Model 4 produces a population of 135 observations and is the sparest of the specifications used in this article. The regressions below are run on all four populations to test the robustness of their results.

In any event, the regressions on models 1 through 3 include the “NumA” measure from the MID data set, which captures the number of states on the initiating side, to control for the effects of multilateral initiating coalitions. Because the extremely large coalitions in a few cases produce significant outliers, I took the square root of the NumA measure to produce a tighter distribution. In model 3, which includes only disputes with one or two initiating states, this variable amounts to a fixed effects control for bilateral initiation. For models 1 and 2, it accounts for the overall size of the initiating coalition.

Independent and Control Variables

Independent variable

The operationalization of the “lame-duck” variable is of vital importance. I code lame-duck status as a binary dummy variable taking a value of zero if the president is eligible for reelection, and a value of one if the president is ineligible for reelection in the next scheduled election for the presidency (codings were derived primarily from Nohlen 2005a, 2005b; Nohlen, Grotz, and Hartman 2001a, 2001b). Therefore, a president’s lame-duck period begins from the date of the election for his or her final term and extends throughout the final term in office.

Notably, lame-duck status here refers only to presidents who are ineligible for reelection in the next scheduled election for the presidency, even though they may be eligible for subsequent presidential elections after a delay of some interval. This is relevant in cases where a country’s constitution allows only nonconsecutive presidential terms. Chile, Peru, and Uruguay, among others, currently allow presidents to serve an unlimited number of nonconsecutive terms. The operationalization of lame-duck status specified above would code all presidents under such systems as lame ducks. Although they can eventually return to office, they cannot do so for the next term. In such cases, the immediacy of the potential electoral punishment is sufficiently remote that it is qualitatively different from those presidents who can run in the next scheduled election. Regardless, I also include a fixed effects dummy variable, coded 1 if the president is eligible for reelection to a nonconsecutive term and coded 0 otherwise, to control for such cases. Table A1 shows all lame-duck presidents that initiated at least one MID between 1946 and 2001.

Control variables

The regressions below include a vector of control variables that may be associated with the response variables. Previously, I have already specified two important controls—the number of initiating states, and a fixed effects control for presidents that remain eligible for an additional, but nonconsecutive term. Additional controls are described and justified in the following.

I code a dummy variable capturing immediate territorial contiguity of the states in the dyad derived from the MID data set’s measure of Direct Contiguity. The dummy is coded 1 if the two states share a land border or are separated by only a river at their border. It is coded 0 otherwise. This captures the effects of direct contiguity rather than mere proximity.

The target’s democracy score according to the Polity IV data set is included to control for the unique effects of bargaining between two democratic states. I use the aggregate democracy score derived by subtracting the target’s autocracy score from its democracy score, yielding a measure that ranges from −10 (pure autocracy) to 10 (pure democracy). This is a standard coding that utilizes an aggregate measure incorporating variation along multiple dimensions of democracy.

A dummy variable for crises initiated by the United States is included as a fixed effects control for any effects unique to American threats. Because the United States is the initiating state for over half (112 of the 212) of the total population of dyadic MIDs in the data set, any artifacts arising from the unique nature of American threats would have a disproportionate influence on the overall results.

The relative power of the two disputants is coded as the proportion of the dyad’s aggregate power that is controlled by the initiating state. Relative power levels are derived from the COW data set’s Composite Index of National Capability (CINC) score (v. 4; Singer, Bremer, and Stuckey 1972; Singer 1987). The initiator’s proportion of overall dyadic power is derived by dividing the initiator’s CINC score by the sum of the initiator’s CINC and the targets CINC (initCINC/(initCINC + targCINC). This again reflects standard practice.

A control for alliance ties is taken from the COW Formal Alliances data set (v. 3.03), which codes bilateral alliance ties along four ascending levels of commitment (Gibler and Sarkees 2004; Gibler 2009). I transform this into a dummy variable coded 1 if the states in the dispute dyad have any form of alliance ties between them, and zero if they lack any such alliance. Only two cases of the overall population of 212 observations had either “entente” or “nonaggression” ties. All the rest were mutual defense treaties. There is thus no loss of precision in coding this variable dichotomously.

Finally, I include the nature of the revision type pursued by the initiating state in order to control for the magnitude of the initiator’s demands on the target. This variable is derived from the MID data set and is coded trichotomously. Revision type is coded 0 if the value was missing, N/A, or “other,” coded 1 if the initiator sought a “policy” revision from the target and 2 if the initiator sought a revision of either “territory” or “regime/government.”

Empirical Analysis: Model Specification and Findings

Hypothesis 1 argues that threats issued by lame-duck presidents are more likely to be resisted by target states than threats issued by reelectable presidents due to a lame duck’s inability to generate audience costs and demonstrate the threat’s credibility. To test this hypothesis, I used the MID data set’s “Recip” score as the dependent variable. This variable is coded 1 if the target responded to the initiator’s threat with any type of militarized action. It is coded 0 if there was no such militarized response. This is a traditional measure of threat success, as a target’s militarized reciprocation of a threat indicates resistance to the initiator’s demands. A positive reciprocation score thus indicates the target’s resistance and the failure of the threat alone to compel the target’s capitulation. Hypothesis 1 would expect a positive and significant coefficient on the lame-duck variable. Because the response variable is dichotomous, I used a logit regression. Tables 1 and 2 present the results.

Table 1.

Lame Ducks and Threat Reciprocation, 1946–2001

Independent Variable	Model 1	Model 2	Model 3	Model 4
Lame duck	0.9** (.354)	.889** (.356)	.913** (.413)	1.021** (.444)
Nonconsecutive	−0.562 (.534)	−0.568 (.535)	−0.403 (.542)	−0.77 (.563)
Alliance ties	−0.326 (.474)	−0.347 (.477)	−0.402 (.513)	−0.307 (.55)
Contiguity	1.233** (.485)	1.22** (.482)	1.492*** (.512)	1.437*** (.52)
Target’s polity score	−0.009 (.031)	−0.006 (.032)	0.008 (.033)	0.007 (.034)
Revision type	0.333 (.223)	0.333 (.231)	0.417 (.267)	0.249 (.282)
US dummy	1.017** (.473)	.908** (.448)	1.323** (.523)	0.672 (.522)
Initiator’s power advantage	−0.162 (.689)	−0.172 (.715)	−0.867 (.684)	−0.34 (.722)
Number of initiating states	3.261** (1.378)	3.953** (1.901)	8.465*** (1.518)	—
Constant	−5.056*** (1.469)	−5.709*** (1.975)	−10.306*** (1.672)	−1.944*** (.677)
n	212	199	173	135
Wald χ²	23.66	20.77	51.63	19.08
Pseudo R ²	.2273	.2079	.2179	.0961
% Correctly predicted	71.23%	70.35%	67.63%	71.85%
Null model (% modal)	54.25%	51.76%	54.34%	67.41%

Note. Numbers in parentheses are robust standard errors.

**p < .05. ***p < .01.

Table 2.

Marginal Effect of Lame-Duck Status on Target’s Threat Reciprocation (predicted probabilities and first differences)

	Model 1	Model 2	Model 3	Model 4
Pr(recip \| reelectable)	.371 (.073)	.352 (.075)	.124 (.051)	.137 (.056)
Pr(recip \| lame duck)	.586 (.091)	.562 (.092)	.251 (.087)	.293 (.097)
ΔPr(recip = 1 \| lameduck_0→1)	.215 (.083)	.21 (.081)	.127 (.069)	.157 (.08)

Note. Numbers in parentheses are standard errors. Binary and ordinal control variables held constant at median values. Continuous controls held constant at means.

The results strongly support Hypothesis 1 and allow a firm rejection of the null hypothesis. In Table 1, all four models show a positive coefficient on the lame-duck variable and are well above the 95 percent confidence threshold. In models 1 and 3, the significance levels approach 99 percent, at 98.9 percent and 98.7 percent confidence, respectively. These results show that threats initiated by lame ducks are significantly more likely to be reciprocated by the targets. Table 2 gives the predicted probabilities of threat reciprocation for both reelectable and lame-duck presidents, as well as the first differences. Holding all else constant, the models predict that threats issued by lame ducks will be reciprocated between 12 percent and 21 percent more frequently than threats issued by reelectable presidents.

Hypothesis 2 argues that crises initiated by lame ducks should be less likely to produce outcomes favorable to the initiator. To test this, I use the MID data set’s “Outcome” variable to construct a trichotomous indicator of dispute outcomes. The variable is coded 0 if the MID data scores the outcome a “victory for side B” or “yield by side A” with side A being the initiator and side B being the target. It is coded 2 if the outcome is a “victory for side A” or a “yield by side B.” It is coded 1 for all other outcomes, including “stalemate,” “compromise,” or “unclear.” Thus, larger values (2) indicate a favorable outcome for the initiator, while smaller values (0) indicate a favorable outcome for the target. This again reflects a standard coding of crisis outcomes. Because the dependent variable is trichotomous and ordinal, I used an ordered logit regression.3 Hypothesis 2 would expect a negative and significant coefficient on the lame-duck variable. The results are presented in Table 3.

Table 3.

Lame Ducks and Crisis Outcomes, 1946–2001

Independent Variables	Model 1	Model 2	Model 3	Model 4
Lame duck	−0.871 (.627)	−0.354 (.715)	−0.809 (.942)	−0.31 (.86)
Nonconsecutive	−0.686 (.436)	−0.704 (.491)	—	—
Alliance ties	.992* (.548)	0.764 (.651)	0.176 (1.1)	0.662 (1.147)
Contiguity	−.843* (.488)	−.875* (.471)	−1.561 (1.255)	−1.185 (1.077)
Target’s polity score	−0.016 (.033)	0.023 (.037)	0.079 (.053)	0.029 (.051)
Revision type	1.84*** (.409)	1.639*** (.48)	1.753** (.856)	0.61 (.612)
US dummy	−0.122 (.628)	0.463 (.589)	1.09 (1.163)	0.188 (1.182)
Initiator’s power advantage	1.967*** (.687)	1.919** (.854)	1.506 (1.675)	2.593 (1.856)
Number of initiating states	−0.36 (.479)	−0.09 (1.228)	—	—
Cut 1	−2.396 (.891)	−2.192 (1.405)	6.046 (1.563)	5.658 (1.536)
Cut 2	5.056 (.854)	5.769 (1.427)	—	—
n	212	199	173	135
Wald χ²	35.71	31.91	12.96	10.54
Pseudo R ²	.2235	.1827	.1767	.0958

Note. Numbers in parentheses are robust standard errors.

*p < .10. **p < .05. ***p < .01.

The results shown in Table 3 do not allow us to reject the null hypothesis that lame-duck status has no effect on crisis outcomes. Hypothesis 2 is not supported by the analysis. Although all four models show the expected negative coefficient on the lame-duck variable, in none does it reach statistical significance. Because the outcome in the vast majority of cases (143 of the 212 or over 67 percent in the largest of the four data sets) is coded as a stalemate, the model is unlikely to produce a significant correlation. In models 3 and 4, moreover, the distribution of the trichotomous-dependent variable was so highly centered around 1, and the data set is sufficiently truncated, that none of the disputes were coded 0 for a target victory. The dependent variable was thus effectively dichotomous. Ultimately, although the highly clustered distribution of the data for the dependent variable muddies the analysis, the results in Table 3 do not indicate a significant relationship between lame-duck status and crisis outcomes.⁴

To test Hypotheses 3A and 3B, I used the MID data set’s “HostLev” measure which scores the highest level of hostility exhibited by each participant in the dispute, as well as the highest level of overall hostility exhibited during the entire dispute.⁵ This variable takes five potential values, ranked ordinally from one (no militarized action) to five (war). There are thus three different potential specifications of the hostility level variable: the initiator’s hostility, the target’s hostility, and the overall dispute hostility. I tested the effect of lame-duck status on all three of these response variables. This allows us to observe the effect of lame-duck initiators on overall crisis hostility levels, while also determining whether the initiator or target was responsible for any variation in escalation patterns.

Hypothesis 3A would expect the coefficient on the lame-duck variable to be positive and significant across all regressions in Table 4. According to the signaling hypothesis, the initiator and target, as well as the overall dispute should exhibit higher levels of hostility when a lame duck is in charge of the initiating state. Hypothesis 3B, conversely, would expect a negative coefficient on the lame-duck variable for the regressions on initiator and overall dispute hostility. With respect to target hostility, Hypothesis 3B would predict either no effect or a positive relationship arising from the logic of Hypothesis 1. Because the hostility variable is coded ordinally, I again used an ordered logit regression. I ran twelve regressions in total, with the four different data sets (models 1–4) each tested on the three different hostility dependent variables. The regression results are shown in Table 4.

Table 4.

Lame Ducks and Crisis Hostility, 1946–2001

	MID hostility				Target hostility				Initiator hostility
Indep Vars	Model 1	Model 2	Model 3	Model 4	Model 1	Model 2	Model 3	Model 4	Model 1	Model 2	Model 3	Model 4
Lame duck	−.791** (.313)	−.75** (.327)	−.581* (.34)	−0.224 (.392)	−0.044 (.329)	0.063 (.343)	0.086 (.388)	1.111** (.447)	−.837*** (.301)	−.853*** (.316)	−.729** (.332)	−0.485 (.383)
Nonconsecutive	.901* (.482)	.957* (.502)	0.895 (.571)	0.95 (.619)	−0.34 (.515)	−0.328 (.509)	−0.226 (.518)	−0.847 (.578)	.904* (.48)	.931* (.497)	0.88 (.562)	1.016* (.597)
Alliance ties	0.489 (.455)	0.379 (.481)	0.286 (.56)	0.318 (.522)	0.066 (.391)	0.051 (.427)	−0.03 (.47)	−0.258 (.577)	025 (.399)	0.101 (.421)	0.005 (.507)	0.034 (.46)
Contiguity	−0.171 (.422)	−0.101 (.441)	−0.135 (.49)	−0.453 (.498)	1.124*** (.404)	1.288*** (.425)	1.453*** (.441)	1.802*** (.51)	−0.193 (.406)	−0.201 (.432)	0.271 (.481)	−0.574 (.479)
Target’s polity	0.019 (.027)	0.03 (.03)	0.031 (.032)	0.029 (.03)	−0.02 (.028)	−0.011 (.029)	0.011 (.03)	0.003 (.034)	0.039 (.025)	.048* (.027)	0.046 (.03)	.047* (.028)
Revision type	0.389 (.207)	0.278 (.211)	0.257 (.234)	0.272 (.261)	0.174 (.195)	0.162 (.203)	0.294 (.234)	0.105 (.274)	.535*** (.205)	.379* (.21)	0.294 (227)	0.289 (.245)
US dummy	0.509 (.404)	0.469 (.441)	0.402 (.448)	0.1 (.537)	.72* (.377)	.869** (.442)	.958** (.449)	0.853 (.558)	0.083 (.396)	−0.203 (.45)	−0.193 (.46)	−0.373 (.245)
Initiator’s power advantage	−1.192* (.628)	−1.364** (.657)	−1.389** (.692)	−1.202* (.727)	−0.18 (.511)	−0.328 (.534)	−0.905 (.56)	−0.468 (.67)	−0.662 (.575)	−0.808 (.614)	−0.867 (.657)	−0.968 (.706)
Number of initiating states	0.326 (.237)	0.782 (.59)	0.257 (.993)	—	.684*** (.22)	1.038** (.468)	3.726*** (1.043)	—	0.174 (.27)	0.678 (.416)	−1.15 (.962)	—
Cut 1	−3.442 (.737)	−3.103 (.887)	−4.211 (1.352)	−3.587 (.85)	1.612 (.563)	2.134 (.69)	4.907 (1.201)	1.98 (.659)	−5.679 (1.014)	−3.373 (.795)	−5.411 (1.395)	−3.976 (.858)
Cut 2	−0.53 (.601)	−0.228 (.776)	−1.38 (1.254)	−1.13 (.682)	1.786 (.564)	2.319 (.691)	5.072 (1.204)	2.018 (.656)	−3.343 (.67)	0.052 (.661)	−2.062 (1.236)	−1.121 (.664)
Cut 3	3.854 (.673)	4.394 (.905)	—	—	2.54 (.562)	2.974 (.69)	5.703 (1.222)	2.578 (.663)	−0.111 (.587)	4.162 (.74)	—	—
Cut 4	—	—	—	—	6.158 (.736)	7.089 (1.047)	—	—	3.8 (.666)	—	—	—
n	212	199	173	135	212	199	173	135	212	199	173	135
Wald χ²	17.33	15.67	13.39	8.75	22.37	19.89	26.2	24.37	23.08	22.02	21.59	12.9
Pseudo R ²	.0457	.0552	.0483	.0431	.0585	.0659	.0659	.0912	.0548	.0637	.0716	.0592

Note. Numbers in parentheses are robust standard errors.

*p < .10. **p < .05. ***p < .01.

The results in Table 4 largely support Hypothesis 3B and strongly disconfirm Hypothesis 3A. The evidence indicates that lame-duck presidents are simply more willing to back down rather than escalate if their initial threat is unsuccessful. The results show that the initiator alone accounts for a general reduction in the overall level of hostility in disputes initiated by lame ducks. In the first four columns, the dependent variable is the highest level of hostility in the overall dispute. The coefficient for the lame-duck variable is negative in all four models. It achieves 95 percent significance in models 1 and 2, and 90 percent confidence in model 3. The lame-duck coefficient for model 4 remains negative but is not statistically significant. This largely supports Hypothesis 3B.

The middle four columns show no consistent significant relationship between lame-duck initiators and target hostility. The target thus cannot account for the decreased crisis hostility found in the first four columns. In three of the four models, moreover, the coefficient on the lame-duck variable is positive. Only in model 4, however, do the findings show the target to be significantly more likely to escalate to a higher level of hostility. These findings are largely indeterminate, as the signaling, selection, and null hypotheses could all account for a weak and inconsistent positive relationship between lame-duck initiators and target hostility levels.

The final four columns show that the initiator’s hostility level decreases significantly when led by a lame-duck president. All four coefficients are negative and three are highly significant. Models 1 and 2 show significance at a 99 percent confidence interval and while model 3 reaches 95 percent confidence. The lame-duck coefficient in model 4 indicates a negative relationship with roughly 80 percent confidence. Taken together with the target hostility results, this indicates that the initiator alone accounts for virtually all of the decreased crisis hostility levels observed in columns 1 through 4. Tables 4 and 5 essentially show that crises initiated by lame ducks exhibit significantly lower levels of hostility, and that the initiator is responsible for this decreased hostility. This is precisely what Hypothesis 3B predicts. Target states may exhibit higher levels of hostility in some instances, but this finding is not robust.

Table 5.

Marginal Effect of Lame-Duck Status on Initiator’s Crisis Hostility Level (first differences)

	Model 1	Model 2	Model 3	Model 4
ΔPr(inithost = 1 \| lameduck_0→1)	.006 (.008)	—	—	—
ΔPr(inithost = 2 \| lameduck_0→1)	.030 (.017)	.038 (.022)	.023 (.018)	.017 (.021)
ΔPr(inithost = 3 \| lameduck_0→1)	.166 (.063)	.163 (.062)	.144 (.07)	.089 (.077)
ΔPr(inithost = 4 \| lameduck_0→1)	−.185 (.067)	−.189 (.067)	−.168 (.079)	−.107 (.091)
ΔPr(inithost = 5 \| lameduck_0→1)	−.017 (.011)	−.012 (.008)	—	—

Note. Numbers in parentheses are standard errors. Binary and ordinal control variables held constant at median values. Continuous controls held constant at means.

Table 5 shows the first difference effects of lame-duck status on the initiator’s hostility levels. These results generally confirm the findings in Table 4. As Hypothesis 3B expects, the probability of an initiator’s escalation to a hostility level of 4 or 5 (the use of force or outright war) is decreased with a lame-duck initiator. Models 1 through 3 show, for example, that a lame-duck president is roughly 17 percent to 19 percent less likely to escalate a crisis to the use of force (inithost = 4) when compared to a reelectable president. Conversely, the probability of an initiator remaining at a low hostility level of 1 through 3 (no militarized action, threat of force, display of force) is increased with a lame-duck president.

Robustness Checks

The results presented above reveal interesting patterns in lame-duck crisis bargaining. But how durable are these findings? To ensure robustness, all regressions were run across multiple data sets, labeled models 1 through 4 above. Multiple respecifications of these four data sets were then used to ensure additional robustness. These results are not presented due to space constraints but are described below. First, I recoded the lame-duck variable to include presidents eligible for an additional but nonconsecutive term as nonlame ducks. Doing so largely strengthens the results above. The significance of the overall dispute hostility results was increased, with several of the coefficients reaching 99 percent significance. The magnitudes of the lame-duck coefficients in the reciprocation model decreased, but they all remained significant with 95 percent confidence.

In models 1 through 3, I also controlled for disputes with multiple target states by effectively treating MIDs with multiple targets as a single dyadic dispute. I extracted the single most powerful of the multiple targets, and included it alone as the target for that dispute. The results remained largely unchanged, with the lame-duck coefficient taking the same sign and general significance levels for the reciprocation and outcome models. The hostility level models changed slightly, as the lame-duck variable no longer had a statistically significant effect on the overall dispute’s hostility level. The lame-duck coefficients for target hostility became more regularly positive but remained insignificant. Finally, the lame-duck variable continued to exert a significant negative effect on the initiator’s hostility level. These results still support Hypothesis 3B.

To be sure that the results were not driven by “joiners” on the initiating side that entered the dispute after it was already in progress, I dropped all observations in which the initiator entered the MID later than the overall dispute’s start date. Again, this applied only to models 1 through 3. Only seven observations of the overall model 1 population of 212 were coded as joiners, and only one of these was included in models 2 and 3 populations. Dropping these observations did not have any meaningful effect on the lame-duck results.

Parliamentary democracies

Finally, the lame-duck hypotheses were tested using an expanded data set that included parliamentary democracies as nonlame ducks. This data set included all MIDs initiated by states with a Polity IV aggregate democracy score of five or greater. Every new observation added to the original data set was considered a parliamentary democracy. Two coding schemes were then used. First, parliamentary initiators were simply coded as nonlame ducks, keeping the independent variable dichotomous. Second, the lame-duck variable was coded as a trichotomous ordinal variable, with parliamentary initiators taking a value of 0, reelectable presidents a value of 1, and lame-duck presidents a value of 2. This would capture the idea that prime ministers are even more susceptible to audience costs than reelectable presidents due to their constant vulnerability to votes of no confidence. All tests were rerun using both of these codings.

These tests produced mixed but interesting results. First, regarding Hypothesis 1, the lame-duck variable no longer had a significant effect on threat reciprocation. This is surprising, given the impressive robustness of the results in Tables 1 and 2 and induces some caution regarding confidence in Hypothesis 1. Second, the expanded data set produced sufficient variation in crisis outcomes that the lame-duck variable exerted a highly significant negative effect on the probability of initiator victory, as Hypothesis 2 predicts. Using the trichotomous lame-duck coding, three of the four models show lame-duck status to decrease the probability of initiator victory with 99 percent confidence. Using the dichotomous coding, model 1 showed a negative effect with 95 percent confidence, while models 2 and 3 showed this effect with 90 percent confidence. With respect to hostility levels, the negative effect of lame-duck status on initiator hostility levels largely held up. Using the dichotomous coding, two of the models showed a negative effect with 95 percent confidence and one with 90 percent confidence. These results were weakened with the trichotomous coding, as all four models showed a negative relationship with between 75 percent and 90 percent confidence. The lame-duck variable only sporadically had a significant effect on overall dispute hostility, while target hostility remained unaffected.

These results are slightly less supportive of audience costs theory than those presented earlier. Hypothesis 1 is significantly weakened when tested against a data set including parliamentary democracies. Hypothesis 2, however, is significantly strengthened. The selection hypothesis, Hypothesis 3B, is still largely supported by these revised tests, but to a lesser degree than the tests presented in Table 4. Most notably, the trichotomous coding of lame-duck status shows a less significant effect on initiator hostility than the binary coding. The data continue to allow a firm rejection of the signaling hypothesis, however, as we see no evidence of increased hostility levels with lame-duck initiators. And while these findings remain broadly supportive audience costs theory, the discrepancies with the results presented are puzzling. Parliamentary democracies are democracies nonetheless, and audience costs theory alone can likely say very little about why the results change when parliamentary democracies are included. Unpacking this variation is a promising avenue for future research.

The evidence presented above indicates that on average, lame-duck presidents are less resolved and less effective bargainers in interstate crises. But can we find evidence, even if only illustrative, that policy makers actually perceive the incentives and mechanisms specified earlier? Can we match the external validity of the statistical findings with a comparable demonstration of the internal validity of audience costs theory? The following section presents a case illustration of the American invasion of Panama as a particularly clear-cut example of how the electoral incentives created by executive term limits affect crisis bargaining.

Case Study: Reagan, Bush, and the Invasion of Panama

The crisis leading up to the December 1989 American invasion of Panama nicely illustrates many of the audience costs dynamics captured in the preceding lame-duck hypotheses.⁶ The crisis between the United States and Panama was initiated late in Ronald Reagan’s second term, after the midterm elections and at the height of his lame-duck period. It then persisted throughout the first year of George H. W. Bush’s administration. The progression of a single crisis across two administrations, one a lame duck and the other reelectable, makes this case an ideal laboratory for examining the effect of lame-duck status on crisis behavior. Most potential, confounding variables are held constant across the Reagan and Bush portions of the crisis, and the fact that Bush was Reagan’s vice president should especially produce continuity. But as we shall see, Bush took a much harder line than Reagan throughout the crisis, demonstrating a far greater willingness to use force. The electoral constraints Bush faced throughout the 1988 election year and his first year in office, constraints that were entirely inoperative for Reagan, were a critical source of their divergent approaches.

Throughout 1986 and 1987, formerly close United States–Panama relations rapidly deteriorated. General Manuel Noriega, the head of the Panama Defense Forces who was linked to the infamous Medellin drug cartel in Colombia, had begun subverting the civilian government, harassing American military personnel, and was repeatedly violating the 1977 Panama Canal Treaties. Both Congress and the Reagan administration enacted measures aimed at coercing Noriega to step down. Then in February 1988, without the knowledge or approval of the Reagan administration, two Florida grand juries simultaneously indicted Noriega on drug trafficking charges. Reagan seized the opportunity by imposing sanctions on Panama and dispatching 1,300 additional American troops to the Canal Zone. Meanwhile, the administration was negotiating with Noriega to secure a return to democratic government. Throughout the negotiations, the indictments remained the biggest sticking point. Noriega would cede power voluntarily if they were dropped, and it was on this issue that Reagan and Bush diverged (Shultz 1993, 1053, 1056).

Reagan, a lame duck invulnerable to domestic electoral punishment, took an extremely dovish stance throughout the crisis. For Reagan, the indictments were unenforceable as long as Noriega remained in Panama. If the United States could secure Noriega’s ouster by dropping these meaningless and impotent indictments, Reagan argued, it would be unthinkable to run the risks of escalation. But with the War on Drugs becoming a prominent issue in the American political landscape, dropping the indictments would provoke severe domestic backlash. Reagan, already in the final year of his presidency, was entirely unmoved by such concerns. As Reagan’s Secretary of State George Shultz later wrote, Reagan “knew a negotiated outcome would be unpopular even if it succeeded, but he was firm” (Shultz 1993, 1064).

Reagan held this position despite resistance from many within his administration, particularly the vice president. Bush was vociferous in his opposition to any negotiated solution that involved dropping the indictments. As Eytan Gilboa writes, “Bush was then in the middle of his presidential campaign, and for him the prospect of letting a drug-dealing dictator out of the indictment looked like political suicide” (Gilboa 1995, 549). Bush’s position was based only in part on the merits of the case. James Baker, Bush’s secretary of state later wrote that “the Vice President would also suffer politically” from a negotiated deal. “It was bad policy and bad politics” (Baker 1995, 180). Reagan’s secretary of state George Shultz claims that Bush “felt the schedule for Noriega’s departure—not leaving until September (1988)—was wild from a campaign standpoint” (Shultz 1993, 1063). Having yet to establish Reagan’s reputation for toughness, Bush was compelled by domestic political concerns to take a hard-line on Panama.

Less than five months into Bush’s presidency, following fraudulent Panamanian elections in May 1989 and weeks of subsequent violence, Bush deployed 1,700 additional troops to Panama in an attempt to force Noriega’s ouster. Then in December 1989, following a particularly belligerent speech by Noriega, the Panamanian legislature declared that a state of war existed between the United States and Panama. Harassment of American military personnel in Panama escalated dramatically, with one incident resulting in the death of an American marine. The following day, Bush ordered the invasion of Panama.

This case starkly illustrates many of the key dynamics of electoral cycles and audience costs. Reagan, having a well-established reputation for toughness and immune to electoral punishment, consistently pursued a soft-line approach to the negotiations with Noriega. When Reagan ruled out the use of force early in the crisis, American diplomacy devolved into a “comedy of mixed signals that encouraged Noriega to fight on” (Kempe 1990, 232). Bush, unwilling to appear soft on drugs and vulnerable to criticism of his toughness, was then compelled by domestic considerations to take an uncompromising stance toward Noriega. Ultimately, Bush’s firmness was unable to overcome the Reagan administration’s mixed signals. But the divergent approaches of the two presidents, and the clear influence of electoral incentives in generating these approaches, nicely illustrate the critical theoretical logic linking presidential term limits, domestic audience costs, and credibility in interstate crisis bargaining.

Implications and Conclusions

This article contributes to the growing literature on democratic foreign policy and crisis bargaining by parsing out the mechanisms that enable some democracies to more effectively demonstrate resolve. The results largely support Fearon’s audience costs model, specifically the contention that a democratic leader’s electoral accountability is a vital source of credibility in bargaining situations. This offers one of the strongest extant tests of Fearon’s audience costs model, utilizing an extremely valid specification of the crucial independent variable. By coding for presidents who are constitutionally prohibited from reelection, I isolate those cases in which a democratic leader is immune to the electoral punishment that is central to Fearon’s model. Prior tests, focusing on factors such as general levels of democracy, constraints on executive decision making, or levels of voter participation are less able to directly capture the audience costs mechanism at work.

These findings also contribute to the literatures on the institutional democratic peace (Bueno de Mesquita et al. 1999; Maoz and Russett 1993), and the “democratic victory” theory (Lake 1992; Reiter and Stam 2002; Desch 2002; Downes 2009). If institutional constraints explain the lack of war between democracies, then the results above demonstrate a potential vulnerability of democratic peace theory. Additionally, if democracies tend to win most of their wars in part because they select into more winnable conflicts, then term limits may undermine this beneficial selection mechanism by removing the vital constraints which induce executive caution.

Additionally, these results yield important policy implications. My findings indicate that term limits significantly reduce the effectiveness of democratic leaders in pursuing their foreign policy interests. This may also lead to riskier foreign policy behavior, as lame-duck presidents cannot be punished by their constituents in the wake of policy failures. The benefits of term limits may outweigh such costs, especially in unconsolidated democracies where the norms and rules of leadership turnover are inchoate and unentrenched. But a thorough understanding of the risks attending lame-duck presidencies is vital to harnessing the unique characteristics of democracy while bargaining in the international arena.

Footnotes

Appendix A

Table A1.

Lame-Duck Militarized Interstate Disputes (MIDs) Initiators

President	Country	Lame-duck term	No. COW disputes
Dwight Eisenhower	United States	November 6, 1956 to January 20, 1961	2849, 2049, 2854, 125, 2870, 2002, 2876, 253
Richard Nixon	United States	November 7, 1972 to August 9, 1974	353
Ronald Reagan	United States	November 6,1984 to January 20, 1989	2232, 2740, 2741, 2742, 2775
Bill Clinton	United States	November 5, 1996 to January 20, 2001	4174, 4216, 4273, 4217, 4227, 4254, 4137, 4125, 4213, 4261, 4186
Alfonso Portillo	Guatemala	January 14, 2000 to January 14, 2004	4151
Roberto Suazo Cordova	Honduras	January 27, 1982 to January 27, 1986	2352
Jose Azcona del Hoyo	Honduras	January 27, 1986 to January 27, 1990	2771, 3905
Rafael Callejas Romero	Honduras	January 27, 1990 to January 27, 1994	3988, 4010
Carlos Roberto Reina	Honduras	January 27, 1994 to January 27, 1998	4012
Carlos Flores Facusse	Honduras	January 27, 1998 to January 27, 2002	4141, 4259, 4327
Alfredo Felix Cristiani Burkhad^a	El Salvador	January 6, 1989 to January 6, 1994	3904
Armando Calderon Sol^a	El Salvador	January 6, 1994 to January 6,1999	4153
Violeta Barrios de Chamorro^a	Nicaragua	April 24, 1990 to January 10, 1997	4011, 4171
Arnoldo Aleman^a	Nicaragua	January 10, 1997 to January 10, 2002	4147, 4140
Jose Figueres Ferrer^a	Costa Rica	November 8, 1953 to May 8, 1958	2042
Jose M. Figueres	Costa Rica	May 8, 1994 to May 8, 1998	4146
Julio Cesar Turbay Ayala	Colombia	August 7, 1978 to August 7, 1982	3120
Virgilio Barco Vargas	Colombia	August 7, 1986 to August 7, 1990	2812, 2768
Andres Pastrana Arango	Colombia	August 7, 1998 to August 7, 2002	4263
Romula Betancourt	Venezuela	February 13, 1959 to May 11, 1964	1114
Raul Leoni	Venezuela	May 11, 1964 to May 11, 1969	1166, 2922, 2940, 2239
Rafael Caldera Lopez^a	Venezuela	May 11, 1969 to May 12, 1974	2240
Carlos Andres Perez^a	Venezuela	May 12, 1974 to May 12, 1979	2317
Luis Herrera Campins	Venezuela	May 12, 1979 to February 2, 1984	2237, 2323, 3085
Jaime Lusinchi	Venezuela	February 2, 1984 to February 2, 1989	2356
Ramon Velazquez	Venezuela	June 5, 1993 to February 2, 1994	4219
Rafael Caldera Lopez	Venezuela	February 2, 1994 to February 2, 1999	4009, 4149, 4154, 4172
Jaime Roldos Aguilera	Ecuador	August 10, 1979 to May 24, 1981	3105
Sixto Duran Ballen	Ecuador	August 10, 1992 to August 10, 1996	4013
Fabian Alarcon	Ecuador	February 11, 1997 to August 10, 1998	4144, 4189
Fernando Belaunde Terry	Peru	June 28, 1980 to June 28, 1985	2118, 2119
Juscelino Kubitschek de Oliveira	Brazil	January 31, 1956 to January 31, 1961	2860
Carlos Ibanez del Campo	Chile	November 4, 1952 to November 4, 1958	2845, 1099
Jorge Alessandri Rodriguez	Chile	November 4, 1958 to November 4, 1964	1100
Eduardo Frei Montalva	Chile	November 4, 1964 to November 4, 1970	1608, 1609
Raul Alfonsin	Argentina	January 10, 1983 to July 8, 1989	2087, 2579, 2813
Leonid Kuchma	Ukraine	November 14, 1999 to January 23, 2005	4186
Roh Tae-woo	South Korea	February 25, 1988 to February 25, 1993	3984
Kim Young-sam	South Korea	February 25, 1993 to February 25, 1998	4021, 4126
Kim Dae-jung	South Korea	February 25, 1998 to February 25, 2003	4125
Carlos Garcia^b	Philippines	November 14, 1961 to December 31, 1961	2027
Fidel Ramos	Philippines	June 30, 1992 to June 30, 1998	4028, 4128, 4329

^aEligible for an additional, nonconsecutive term.

^bInitiated dispute on November 26, 1961, after losing election for second term.

Acknowledgments

The author would like to thank Todd Sechser, Dale Copeland, William Quandt, Phil Potter, Brandon Yoder, Kyle Lascurettes, Kate Sanger, Steven Liao, and two anonymous reviewers for helpful comments on earlier versions of the article. The author would also like to acknowledge the generous support of the Eisenhower Institute of Gettysburg College. All errors and omissions remain the author's.

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.

Notes

References

Auerswald

David

. 2000. Disarmed Democracies: Domestic Institutions and the Use of Force. Ann Arbor: University of Michigan Press.

Baker

James A.

1995. The Politics of Diplomacy: Revolution, War and Peace 1989-1992. New York, NY: G.P Putnam’s.

Brown

Jonathan

Marcum

Anthony

. 2010. “Avoiding Audience Costs: Domestic Political Accountability and Concessions in Crisis Diplomacy.” Security Studies 20 (2): 141–170.

Brule

David

Marshall

Bryan

Prins

Brandon

. 2010. “Opportunities and Presidential Used of Force: A Selection Model of Crisis Decision-Making.” Conflict Management and Peace Science 27 (5): 486–510.

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.

Bueno de

Mesquita

James Morrow

Bruce

Siverson

Randolph

Smith

Alastair

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

Carey

John

. 1996. Term Limits and Legislative Representation. Cambridge: Cambridge University Press.

Cheibub

Jose

. 2007. Presidentialism, Parliamentarism, and Democracy. Cambridge: Cambridge University Press.

Chiozza

Giacomo

Goemans

Hein

. 2003. “Peace through Insecurity: Tenure and International Conflict.” Journal of Conflict Resolution 47 (4): 443–67.

10.

Choi

Seung-Whan

. 2010. “Legislative Constraints: A Path to Peace.” Journal of Conflict Resolution 54 (3): 438–70.

11.

Clark

David

Nordstrom

Timothy

. 2005. “Democratic Variants and Democratic Variance: How Domestic Constraints Shape Interstate Conflict.” Journal of Politics 67 (1): 250–70.

12.

Desch

Michael

. 2002. “Democracy and Victory: Why Regime Type Hardly Matters.” International Security 27 (2): 5–47.

13.

Downes

Alexander

. 2009. “How Smart and Tough are Democracies? Reassessing Theories of Democratic Victory in War.” International Security 33 (4): 9–51.

14.

Downes

Alexander

Sechser

Todd

. 2012. “Democracies and the Credibility of Coercive Threats.” International Organization 66 (3).

15.

Eyerman

Joe

Hart

Robert

. 1996. “An Empirical Test of the Audience Cost Proposition: Democracy Speaks Louder than Words.” Journal of Conflict Resolution 40 (4): 597–616.

16.

Fearon

James

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

17.

Fearon

James

. 1995. “Rationalist Explanations for War.” International Organization 49 (3): 379–414.

18.

Fearon

James

. 1997. “Signaling Foreign Policy Interests: Tying Hands versus Sinking Costs.” Journal of Conflict Resolution 41 (1): 68–90.

19.

Fearon

James

. 1999. “Electoral Accountability and the Control of Politicians: Selecting Good Types versus Sanctioning Poor Performance.” In Democracy, Accountability, and Representation, edited by Przeworski

Adam

Stokes,

Susan C.

Manin

Bernard

, 55–97. Cambridge: Cambridge University Press.

20.

Gaubatz

Kurt Taylor

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

21.

Gelpi

Christopher

Grieco

Joseph

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

22.

Gelpi

Christopher

Griesdorf

Michael

. 2001. “Winners or Losers? Democracies in International Crisis, 1918-94.” American Political Science Review 95 (3): 633–47.

23.

Ghosn

Faten

Palmer

Glenn

Bremer

Stuart

. 2004. “The MID3 Data Set, 1993-2001: Procedures, Coding Rules, and Description.” Conflict Management and Peace Science 21 (2): 133–54.

24.

Gibler

Douglas

. 2009. International Military Alliances from 1648-2008. Washington, DC: Congressional Quarterly Press.

25.

Gibler

Douglas

Sarkees

Meredith

. 2004. “Measuring Alliances: The Correlates of War Formal Interstate Alliance Dataset, 1816-2000.” Journal of Peace Research 41 (2): 211–22.

26.

Gilboa

Eytan

. 1995. “The Panama Invasion Revisited: Lessons for the Use of Force in the Post Cold War Era.” Political Science Quarterly 110 (4): 539–62.

27.

Guisinger

Alexandra

Smith

Alastair

. 2002. “Honest Threats: The Interaction of Reputation and Political Institutions in International Crises.” Journal of Conflict Resolution 46 (2): 175–200.

28.

Jones

Daniel

Bremer

Stuart

David

Singer J.

. 1996. “Militarized Interstate Disputes, 1816-1992: Rationale, Coding Rules, and Empirical Patterns.” Conflict Management and Peace Science 15 (2): 163–212.

29.

Kempe

Frederick

. 1990. Divorcing the Dictator: America’s Bungled Affair with Noriega. New York, NY: G. P. Putnam’s.

30.

Lake

David

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

31.

Leblang

David

Chan

Steve

. 2003. “Explaining Wars Fought by Established Democracies: Do Institutional Constraints Matter?” Political Research Quarterly 56 (4): 385–400.

32.

Leeds

Brett Ashley

Davis

David

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

33.

Lijphart

Arend, ed.

1992. Parliamentary Versus Presidential Government. Oxford, UK: Oxford University Press.

34.

Maoz

Zeev

Russett

Bruce

. 1993. “Normative and Structural Causes of Democratic Peace, 1946-1986.” American Political Science Review 87 (3): 624–38.

35.

Marshall

Monty G.

Jaggers

Keith

. 2002. Polity IV Project: Political Regime Characteristics and Transitions, 1800-2002, Dataset Users’ Manual. University of Maryland: Integrated Network for Societal Conflict Research and The Center for Systemic Peace.

36.

Marshall

Monty G.

Gurr

Ted R.

Davenport

Christian

Jaggers

Keith

. 2002. “Polity IV: 1800-1999.” Comparative Political Studies 35 (1): 40–45.

37.

Nohlen

Dieter

, ed. 2005a. Elections in the Americas: A Data Handbook. Vol. 1. Oxford, UK: Oxford University Press.

38.

Nohlen

Dieter

, 2005b. Elections in the Americas: A Data Handbook. Vol. 2. Oxford, UK: Oxford University Press.

39.

Nohlen

Dieter

Grotz

Florian

Hartman

Christof

, eds. 2001a. Elections in Asia and the Pacific: A Data Handbook. Vol. 1. Oxford, UK: Oxford University Press.

40.

Nohlen

Dieter

Grotz

Florian

Hartman

Christof

, eds. 2001b. Elections in Asia and the Pacific: A Data Handbook. Vol. 2. Oxford, UK: Oxford University Press.

41.

Partell

Peter

. 1997. “Executive Constraints and Success in International Crises.” Political Research Quarterly 50 (3): 503–28.

42.

Partell

Peter

Palmer

Glenn

. 1999. “Audience Costs and Interstate Crises: An Empirical Assessment of Fearon’s Model of Dispute Outcomes.” International Studies Quarterly 43 (2): 389–405.

43.

Potter

Philip B. K.

2007. “Does Experience Matter? American Presidential Experience, Age, and International Conflict.” Journal of Conflict Resolution 51 (3): 351–78.

44.

Press

Daryl G.

2005. Calculating Credibility: How Leaders Assess Military Threats. Cornell, NY: Cornell University Press.

45.

Prins

Brandon

. 2003. “Institutional Instability and the Credibility of Audience Costs: Political Participation and Interstate Crisis Bargaining, 1816-1992.” Journal of Peace Research 40 (1): 67–84.

46.

Prins

Brandon

Sprecher

Christopher

. 1999. “Institutional Constraints, Political Opposition, and Interstate Dispute Escalation: Evidence from Parliamentary Systems, 1946-89.” Journal of Peace Research 36 (3): 271–87.

47.

Reiter

Dan

Stam

Allan

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

48.

Reiter

Dan

Tillman

Erik

. 2002. “Public, Legislative, and Executive Constraints on the Democratic Initiation of Conflict.” Journal of Politics 64 (3): 810–26.

49.

Sartori

Anne

. 2005. Deterrence by Diplomacy. Princeton, NJ: Princeton University Press.

50.

Schelling

Thomas C.

1960. The Strategy of Conflict. Cambridge: Harvard University Press.

51.

Schelling

Thomas C.

1966. Arms and Influence. New Haven, CT: Yale University Press.

52.

Schultz

Kenneth

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

53.

Schultz

Kenneth

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

54.

Sechser

Todd

. 2010. “Goliath’s Curse: Coercive Threats and Asymmetric Power.” International Organization 64 (4): 627–60.

55.

Shultz

George P.

1993. Turmoil and Triumph: My Years as Secretary of State. New York: Charles Scribner.

56.

Singer

David J.

1987. “Reconstructing the Correlates of War Dataset on Material Capabilities of States, 1816-1985.” International Interactions 14 (2): 115–32.

57.

Singer

David J.

Bremer

Stuart

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.

Smith

Alastair

. 1998. “International Crises and Domestic Politics.” American Political Science Review 93 (2): 623–38.

59.

Trachtenberg

Marc

. 2012. “Audience Costs: An Historical Analysis.” Security Studies 21 (1): 3–42.

60.

Tomz

Michael

. 2007. “Domestic Audience Costs in International Relations: An Experimental Approach.” International Organization 61 (4): 821–40.

61.

Weeks

Jessica

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

62.

Zupan

Mark

. 1990. “The Last Period Problem in Politics: Do Congressional Representatives Not Subject to a Reelection Constraint Alter Their Voting Behavior?” Public Choice 65 (2): 167–80.

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.

0.04 MB

0.00 MB