Show
One of the most coveted research agendas in the literature on international crises revolves around how citizens respond to their leaders’ decisions to threaten or use force. The United States being so formidable and the literature on it being so prolific, these studies tend to focus on the incentives, behaviors, and consequences for US presidents. This goes for scholars of American politics and international relations alike, representing a confluence of studies attempting to understand the conditions under which the public rallies around the executive’s decision to project force abroad. Scholars have also examined crises in progress, analyzing how the prospect of successful military operations and wartime casualties affect US public opinion (Mueller 1973; Gartner and Segura 1998; Gelpi, Feaver, and Reifler 2005/2006). Notwithstanding that these studies have significantly advanced our knowledge about the dynamics between conflictual foreign policy and public opinion, there is a clear theoretical and empirical blind spot—the domestic consequences of crisis termination.1 The separate literature on interstate conflict, including formal models of crisis bargaining, has emphasized the importance of understanding the causes and consequences of crisis termination as part of strategic interactions over the course of an international crisis. In this article, we extend theories relevant to public opinion, presidential behavior, and international crises to develop expectations on and explore the effects of crisis terminations. We apply the logic of leader approval and competence signaling to the end stage of crises, arguing that it is an apt venue for impacting public attitudes. Assuming that peace is a popular public good, it follows that citizens generally prefer a crisis termination to an ongoing crisis. Assuming a rational president conscientious of domestic approval, we conceptualize crisis terminations as strategic events in their timing and framing. Using International Crisis Behavior (ICB) and presidential approval data (Brecher et al. 2015; Woolley and Peters 2010), our empirical findings demonstrate that crisis terminations lead to a short-term surge in approval akin to rally effects. We also specify and test conditioning factors, finding that election periods amplify the positive effect while a poor economy and high newsworthiness dampen it. Results are robust to several alternative specifications and potential issues of endogeneity, which we discuss and demonstrate. We proceed with a survey of existing scholarship to posit our general and conditioning hypotheses, a discussion of our empirical approach and findings, and conclude with a brief discussion. Presidential Approval and US Foreign PolicyPresidential approval is a prevailing concept in political science and American politics. The density of academic research, public polls, and media attention to the topic well demonstrates its significance. Approval ratings function as real-time referenda, not unlike a parliamentary vote of confidence in which the president faces monthly appraisals from the electorate about his ability to govern effectively (Crespi 1980). High approval is thought to facilitate two advantages. First, it signals higher chances of reelection (Lewis-Beck and Rice 1982; Weatherford 1986; Zolghadr, Niaki, and Niaki 2017). Even when a president in his second term cannot be reelected per se, he is similarly incentivized to pave the way for his party to preserve power, to bolster his political influence in the term’s aftermath, and to shore up his legacy. Second, leaders with higher approval might enjoy greater influence over policy and legislation (Rivers and Rose 1985; Brewer 2001; Bond, Fleisher, and Wood 2003; Canes-Wrone and de Marchi 2003; Gronke, Koch, and Wilson 2003; Barrett and Eshbaugh-Soha 2007; McKay 2011; Godwin and Ilderton 2014).2 Citing broad support from the public, one can more credibly argue that his agenda is congruent with voters’ interests. As single-minded seekers of reelection (Mayhew 1974), there is consensus that executives attentively heed and carefully calibrate their behaviors to public approval levels as they unfold in political circumstances. The natural questions that follow are how leaders calibrate their behaviors and whether and how it works. One approach that engages these questions is diversionary theory, looking closely at the strategic feedback loop between domestic (dis)approval and executive calculus in foreign crises. In response to substandard domestic conditions (read palling public support), a leader might initiate or invite foreign adventurism to improve his standing and probability of survival in power (Morgan and Bickers 1992; Richards et al. 1993; Downs and Rocke 1994; Smith 1996; James and Rioux 1998). Assuming that presidents have greater freedom of action in foreign than domestic policy, the latter falling under the jurisdiction of Congress and requiring greater executive efforts and fewer chances to frame and credit claim (Marra and Simon 1990; Leeds and Davis 1997), foreign policy is an attractive arena for this. Further, given how high-stakes, contentious, and potentially costly they are, crises up the ante for leaders to leverage public opinion. Diversionary theory is beset, however, with mixed findings and cynical assumptions. That leaders would risk blood and treasure for a selfish bump in polls is a difficult assumption to swallow. Taking a cue from interstate conflict literature that models crises as processes, we extend the analysis of presidential approval and strategic behavior to crisis terminations. Extant research recognizes three distinct temporal nodes in the sequence of crises–initiation, duration, and termination. Surprisingly, little empirical research has been conducted on presidential approval at the final phase. This gap is all the more glaring considering that much theory is founded on strong assumptions about what happens when a crisis is terminated. For instance, Koch and Sullivan (2010) theorize that public support affects the timing of termination, but that it is conditioned on partisanship, with left governments becoming more likely to withdraw as approval ratings decline, but right governments becoming less likely. Situated in this gap, we contend that crisis terminations are an ideal venue for leader competence signaling and opinion rallies. Terminations appeal to broader domestic audiences than risky conflict initiations, overcome the strategic conflict avoidance argument (Leeds and Davis 1997; Miller 1999), and require less cynical assumptions about leaders’ self-interest. They might also be easier than starting wars in the game of political elites. In short, voters prefer peace and security, and presidents prefer to keep voters happy when possible. Consequently, in the Herculean task of navigating the political arena, delivering good policy, and reaping due credit, we posit crisis terminations as an efficient tool. Crisis Termination and Public OpinionJust as with crisis initiations and intervals, there are potential costs and benefits to crisis terminations as well. Termination might involve concessions, a diminished reputation for resolve, domestic doubts about competency, and admission of expended costs without attaining objectives. Notably, crisis bargaining and audience cost literatures argue that mishandled crises can induce reputation costs (Fearon 1994; Reiter 2003). At the same time, it might involve victory, an enhanced reputation for foreign policy efficacy, domestic assurances about competency, and a ‘‘halo effect’’ for attaining objectives (Jentleson 1992). This implies that terminations (like onsets) meet the criteria for a foreign policy tool to influence public opinion. They are salient, can demonstrate competence, and thus can activate a boost of approval. In fact, relative to crisis initiations, the conventional domain of diversionary theory and studies on rally effects, terminations bear more features that recommend strategic competence signaling with fewer theoretical burdens. First, while initiating a conflict might gratify an invested or hawkish subset of the population, it also risks broad unpopularity for its impending costs and uncertainty. On the other hand, terminating a crisis resolves the tensions of risk, cost, and uncertainty, thus appealing to a broad domestic audience even if dissatisfying a hawkish few. Indeed, there are few circumstances in which one would expect widespread disapproval, either at mass or elite levels, of a decision to end a brinkmanship or hemorrhage of blood and treasure. Second, terminating a crisis for self-interest entails softer assumptions about the rational ruthlessness necessary to drag the nation to war for a marginal edge in reelection. Terminations conclude risks, not launch high-stakes gambles. They demobilize troops, not send them to the front lines. They involve clarity and certainty of terms, not an ambiguous campaign of motivated bias. Terminations truncate the ills of war, not initiate them. Third, we argue that terminating a crisis is more subject to an executive’s control, an attractive quality for managing public opinion and approval. Crisis onsets may befall (or elude) a leader, but terminations are always a choice. Even when leaders may wish to initiate a crisis, scholars point to strategic conflict avoidance as a potential moderator (Leeds and Davis 1997; Miller 1999; Clark 2003; Chiozza and Goemans 2004; Fordham 2005). At precisely the point when an executive might wish to divert attention abroad, opportunities for foreign adventurism dry up. Since crisis terminations inherently imply an ongoing contention, this moderator becomes moot. Furthermore, executives can leverage informational and institutional advantages to strategically time and frame a termination. Though crisis terminations entail risks and costs like crisis onsets, leaders can often leverage their advantages to end them under relatively preferable terms and with tools of spin and optics. We expect that presidents gauge the character of the dispute; the costs, benefits, and probability of success; the public’s awareness and mood; and the timbre of opposition. Importantly, presidents gauge these factors as they evolve over time and in comparison to counterfactuals, selecting the ideal juncture and context in which to terminate to boost approval. In light of their incentives, resources, and a degree of executive control, we argue that leaders can time and frame crisis terminations to their advantage. With fewer constraints on his freedom of action in foreign policy, the president can leverage his available diplomatic, intelligence, and military tools to calibrate the timing of termination. Admittedly, presidents’ control over such events is imperfect. While a president may not be able to choose the exact date and time of a termination, he can maneuver and negotiate to get close. Thus, leaders possess a certain plastic control at multiple junctures of a crisis–capitulating, continuing, coercing, offering side payments, credit-claiming, and framing. This yields an endogenous setting favorable to publicize and frame any advantages, good outcomes, or demonstrations of competence. Though risks, costs, and tradeoffs might manifest at the time of termination, executives have distinct framing advantages that give them the opportunity to diminish audience costs and accentuate aspects of success. Even alongside competing influences over timing, even amid suboptimal outcomes, executives might salvage assessments through presentation (Druckman 2001; Eshbaugh-Soha and Peake 2011; Payne 2019/2020). For instance, at the end of the rather unpopular Vietnam War, President Nixon claimed victory and credit for a successful termination. He insisted that all the goals he set, and the conditions he demanded, to attain “peace with honor” were met (Nixon 1973). Shortly following the withdrawal, Nixon’s approval increased by 6 percent despite the conflict’s divisive tone. The most obvious circumstance under which a president might terminate a crisis is when the outcome has reached its peak of domestic favorability, when he can publicize success and frame a good outcome. Even if a crisis is deteriorating, though, a leader can frame the termination to his advantage. Though terminating under suboptimal terms at time t might not be ideal in comparison to optimal terms, if the true comparison is terminating that crisis later—when more costs have accumulated, when bargaining conditions have deteriorated, when they have less purchase in a given domestic environment, when partisan opposition becomes more vehement, etc.—ending a crisis can still be a rational choice. A bad outcome now is preferable to a worse outcome later. If the public begins to chafe from war weariness, then a poor outcome now is preferable to a poorer outcome later when more costs have accumulated. One can imagine a number of other possible tradeoffs. Rather, ending with a poor outcome in absolute terms might be better conceptualized as strategically ending with a superior outcome in relative terms. Though out-of-sample from our upcoming statistical analysis, Trump’s treatment and framing of the fatal drone strike on Iranian Quds commander Qasem Soleimani is a strong anecdote. Whether the start date of the crisis ends up being coded as December 27, 2019,3 December 314 or January 3, 2020 when the strike itself was performed, Trump’s approval was universally higher upon crisis termination (Gallup 2021). Despite that it was largely lambasted by critics, despite a poll revealing that 53 percent of respondents disapproved of his handling of the crisis (Jackson 2020), Trump enjoyed anywhere from a 1 to 8 percent boost in approval upon termination depending on how the crisis is coded. In fact, according to Gallup his approval reached an all-time high for his presidency during the two weeks following the crisis. In an election year, even a modest 1 percent shift can be meaningful. Notably, one day after the crisis, Trump declared victory over Iran and touted his foreign policy prowess during his first campaign rally for the 2020 election (Baker 2020). Consequently, for their public popularity and presidential advantages, we expect that crisis termination will have a halo effect akin to rally effects that occur at its onset. Despite the dearth of substantive analysis on the effects of conflict termination, we maintain that it mirrors rally effects since the two causal mechanisms are driven by the same factors—maximizing rationality and executive advantages. The scholarship on rally effects demonstrates that crisis initiation correlates with a small, short-term surge in approval followed by its rapid decay for both the conflict and the president (Krosnick and Brannon 1993; Baum 2002; Hetherington and Nelson 2003; Kriner 2006). The surge can be attributed to the relative absence of partisan dissent in the early stages of a crisis and the potential amplifying presence of in-group/out-group (IOG) sentiment. First, in a rapidly unfolding crisis, the president will enjoy an informational advantage based on his privileged position to incoming intelligence. Potential opposition, equipped with less information and less reliable sources to bridge the gap, is less likely to voice criticism. Unable to present a balanced debate lacking critical opposition, the media is prone to depict a united front. Finally, lacking existing frames to aid in information processing, given that the event is foreign in origin, the public is more reliant on elite and media cues and is thus more likely to passively accept the appearance of bipartisan support (Baker and Oneal 2001; Althaus and Kim 2006; Eichenberg, Stoll, and Lebo 2006). Over time, the information asymmetry shrinks, opposition dissents more confidently and boisterously, perhaps caskets begin to arrive in the event of militarization, and the public wearies of conflict. Second, when IOG logic obtains, a rational leader might justify force more readily by framing it in the image of the enemy to activate social-psychological cues of rivalry. Thus, the causal logic of a rally operates on a rational president leveraging an informational advantage. If IOG dynamics are present, then the leader can further frame the crisis termination as a display of competence and a victory (or hedging of losses) over an opponent. Overall, this should induce a surge, or halo effect, in approval. The subsequent decay represents the resumption of politics as usual as the event fades from the forefront of public consciousness. In one of the few case studies on public approval following war termination, Althaus and Kim (2006) observe higher approval for President George H. W. Bush following the Persian Gulf War, giving us initial affirmation. We also observe other anecdotal evidence, such as the boosts in Obama’s approval following the death of Gaddafi (achieved militarily) and following the United States–Russia framework to remove the stockpile of chemical weapons in Syria (achieved diplomatically). This leads to our first hypothesis. Though intuitive, it comes alongside affirmative findings for competence signaling, diversionary conflict initiation, and rally effects and establishes a baseline correlation for the conditional hypotheses. H1—General Effect: Compared to levels in an ongoing international crisis, termination will have a positive effect on presidential approval. Conditioning FactorsWhile we expect crisis termination to have a net positive effect on presidential support in general, we identify some conditioning factors that might influence the reliability and magnitude of this trend—the degree and disposition of public attention and partisan divisiveness. The first factor that conditions public response to crisis termination is the degree of public attention to the event. This harmonizes with Baker and Oneal (2001) asserting that the perceived political importance of crisis initiation and the prominence of media reporting mediate the magnitude of a rally effect. We identify two mechanisms that influence public attention. First, election years unfailingly correspond to greater public focus and political involvement. There is a ritual regularity of campaigns, political advertisements, debates, news coverage, and editorials that increases the salience of nearly all political events. Incumbents look to leverage every instance of competence, while competitors look to emphasize any example of weakness. Given its penchant for conflict, the news media provides ample outlets for such commentary. As one scholar notes, “one autumn out of four, election forecasting surpasses baseball as America’s national pastime”(Lewis-Beck 1985, 53). Due to this focused attention, we expect the surge in approval for a crisis termination to swell higher. H2—Election Year Effect: Crises terminated in an election year will have a stronger positive effect on presidential approval than those terminated outside of an election year. Outside of the campaign environment of election years, there is still a variance of public attention to different crises and mechanisms for increasing it among the public. Following a sizable literature in political communication studies, we argue that crises figuring more frequently and centrally in the news will be more accessible and relevant to the public mind (McCombs and Shaw 1972; Iyengar 1990; Iyengar and Simon 1993; Edwards, Mitchell, and Welch 1995; Zhou and Moy 2007). Consequently, we hypothesize that greater news coverage will increase public response to the president’s termination of a crisis. H3—Media Coverage Effect: The greater the volume of major news media coverage of a crisis termination, the stronger the positive effect on presidential approval. One of the most robust correlates of presidential approval is the economy (Beck 1991; Erikson, MacKuen, and Stimson 2000; Lewis-Beck and Stegmaier 2000; Nadeau and Lewis-Beck 2001; Lewis-Beck, Nadeau, and Elias 2008). When the economy is good, approval predictably rises. When it is poor, approval reliably declines. Given how heavily the public weights economic considerations, we do not expect to observe a significant surge of approval for crisis termination when the economy is poor. This might be because the gravity of the termination is discounted in the face of economic discontent, even if voters support the event in general. Another possibility is that regardless of the president’s foreign policy success, once he is deemed inept with the economy, no additional information will offset the public’s evaluation. Thus, a public mired in a poor economy will be more miserly in approval when the president demonstrates competence in another area. Conversely, citizens are likely more generous in approving evaluations under the condition of a good economy. H4—Economy Effect: Crises terminated under poor economic conditions will have a muted positive or null effect on presidential approval compared to those terminated under good economic conditions. Finally, we consider the political atmosphere in which a crisis terminates. Just as presidential support surges during a rally event with bipartisan support, so too do we expect the degree of unity or divisiveness to condition the public’s response at the end of a crisis. Berinsky (2007) demonstrates that as elite opinion converges, the public responds with mainstream support. When it diverges on partisan lines, the public divides as well, resulting in sorted subsets of the population that will always (dis)approve of the president’s actions no matter the tone or salience of an event. Thus, polarization mitigates the effects of an aggregate swell in approval following a salient event (Jacobson 2003; Lebo and Cassino 2007). Since rallies tend to involve elite and public reactions to exogenous events, the causal arrow runs from crisis onset to a partisan response to public opinion. Since crisis terminations tend to be endogenously arranged, we expect that the degree of standing partisan divisiveness will condition public evaluations without a temporally primary movement from the crisis itself. H5—Partisan Conflict Effect: Crises terminated under conditions of higher partisan divisiveness will have a weaker positive effect on presidential approval than those terminated under conditions of low partisan conflict. Research DesignSampleThe unit of analysis is the president-crisis-month. The list of crises is taken from the ICB database (Brecher and Wilkenfeld 2000; Brecher et al. 2015),5 which includes 61 events across 269 crisis-months between April 1953 and February 2016.6 Crises are international interactions of heightened tension. In his explanation of the ICB project, James (1988) describes them as transition points in conflict processes, holding the potential of deescalation or war (see also James 2019). They begin as events short of war (i.e., Poplar Tree), are infused with perceptions of increased probabilities of war (such as the Iran Hostage crisis), and at times do escalate to war (for instance, the Gulf War). All crises are itemized in table 1.7 Since we are interested in how presidential approval changes at the time of crisis termination in comparison to previous crisis-months, single-month crises without a meaningful baseline category are excluded from the analysis.8 Table 1. US crises in the regression sample
VariablesThe dependent variable is monthly presidential job approval, published by the American Presidency Project database (Woolley and Peters 2010).9 There is evidence that presidential approval may be time-trended (Marra and Simon 1990; Brace and Hinckley 1991, 1993; Eisenstein and Witting 2000; Paldam 2008; Nadeau et al. 2019). It begins high during a new leader’s honeymoon, then declines with a temporary swell during reelection campaigns, and repeats with a steeper decay should the incumbent win office again. To ensure that our estimates of public assessments embedded in crisis terminations are not influenced by any time-dependent variation in presidential approval, we perform robustness checks with the dependent variable detrended for each president. The correlation between raw and detrended presidential approval is 0.74. Crisis Termination, our primary independent variable, is an indicator that records whether an international crisis ended in a given month. Importantly, terminations are compared to a baseline of ongoing conflict, not to a baseline of peace. Each of our four conditional hypotheses introduces a moderating variable. First, Election Period is a binary variable coded as one for the months between January and November of a presidential election year. Second, NY TimesCoverage records the cumulative number of New York Times reports over the course of a crisis. We first counted how many times the NY Times featured a given crisis for each month, then calculated the running sum of the reports by the month of termination. The information comes from the NY Times Index data in the Comparative Agendas Project (The Policy Agendas Project at the University of Texas at Austin 2017). Third, the Economic Misery index measures citizens’ objective economic conditions by adding the seasonally adjusted unemployment and inflation rates, drawn from the Bureau of Labor Statistics (Okun 1962). Fourth, the monthly Partisan Conflict index measures the extent to which politicians disagree on an issue based on “the frequency of newspaper coverage of articles reporting political disagreement about government policy—both within and between national parties—normalized by the total number of news articles within a given period” (Azzimonti 2014, 2). Because the public largely obtains information about how divisive parties’ issue positions are through the media, this measure is more suitable for our theory than the polarization measure based on roll-call votes. These moderators are interacted with the Crisis Termination variable. We control for a list of confounding factors of presidential approval: (1) the Economic Misery index lagged by one month, (2) NY TimesCoverage, (3) Crisis Duration, (4) a scale of Crisis Intensity that measures the severity of violence (four categories from minor to major) from the ICB dataset,10 (5) an indicator for when the crisis involves a Strategic Rivalry, (6) a binary Cold War variable that is coded as zero for the years since 1989 and one otherwise, and (7) three temporal dummies: First Year in Office, Election Period, and Second Term. Finally, we control for president-specific fixed effects to account for unobserved heterogeneities for each president with respect to approval ratings. Table A1 in the appendix features descriptive statistics for all variables. Table 2. Effect of crisis termination on presidential approval
Table 3. Moderating effects of crisis termination on presidential approval
EstimatorWe found evidence that OLS regressions for approval in our sample have a heteroskedastic and autocorrelated error structure, both clearly demonstrated in figure 1. The left graph shows that the error variance is increasing in the predicted values of the outcome in a systematic way, also evidenced by both Breusch–Pagan and White tests, in which we can reject the null hypothesis of homoskedasticity. The right graph shows that the residuals at time t are highly correlated to those at t−1. In a residual analysis, we found only first-order autocorrelation to be statistically significant. To address both issues, we use Stata’s ‘‘ivreg2’’ command that estimates linear regression models with heteroskedastic and autocorrelation-consistent (HAC) estimates (Baum, Schaffer, and Stillman 2010). Using the robust and bandwidth options, we implement HAC covariance estimation with the default Bartlett (Newey–West) kernel (Newey and West 1994). We also employ multiway clustering, rendering standard errors robust to arbitrary within-group and cross-panel correlation (Thompson 2009; Cameron, Gelbach, and Miller 2011). Altogether, the estimator is efficient in the presence of persistent and grouped disturbances in panel data (Driscoll and Kraay 1998). ResultsTable 2 features the empirical results of the baseline models testing Hypothesis 1. In all specifications, crisis termination has a positive and significant effect on presidential approval compared to the pre-termination period. Substantively, the average effect of a 3 percent increase in the main, full model compares quite admirably to rally effects from crisis onsets. Indeed, Lian and Oneal (1993) find that only salient, well-publicized mobilizations of force attain 2–3 percent approval boosts. Thus, despite the difficulty of moving public opinion theoretically, even with conservative estimation and an extensive battery of controls econometrically, we find robust support for H1. Table 3 presents the results of the multiplicative interaction effects between moderators and termination. Findings indicate that crisis terminations receive a stronger boost of approval during election periods (H2), but weaker rallies receive greater media coverage (H3) and during economic slumps (H4). Interpreting the statistical output of interactions is complex, so we present graphs of a series of post-estimation analyses in figure 2. The four graphs present the effect of a termination (y-axis) conditioned by each of the moderators (x-axis). The top-left graph shows that ending a crisis has a positive effect on presidential approval for both election and non-election periods. The size of the positive effect, however, is significantly greater in the election period, supporting Hypothesis 2. Substantively, when a crisis is terminated during an election period versus during regular political times, it swells approval by nearly 10 points. The top-right graph reveals a pattern opposite to our theoretical expectation in Hypothesis 3. The boosting effect of ending a crisis diminishes as the cumulative count of NY Times coverage increases. For every standard deviation increase in cumulative coverage, approval drops at the time of termination by just over 1 point. This finding seems to indicate that the more the media reports on a foreign conflict, the less the public rewards the president’s foreign policy conclusion. We offer five possible explanations for this. First, the unexpected results could stem from the nature of the coverage. We might expect more media coverage to boost approval if it is focused on themes such as enemy atrocities, the importance of US interests at stake in the crisis, or the reputational importance of a complete victory. Conversely, if news coverage focuses on risks, casualties, tragedies, the intractability of political disagreements, or the complexity of foreign involvement, greater coverage might diminish approval.11 Second, it might be that the very foreign crises that receive the most media attention are those that are the most contentious and difficult to resolve. These bear newsworthy characteristics—salience, conflict, drama—that comport with American media’s commercial logic (Groeling and Baum 2008; Eshbaugh-Soha and Peake 2011; Hayes and Guardino 2013). Third, we acknowledge that the news media does not always portray the president’s foreign policy narrative as the executive expects or desires. With trends of increasing negative news, it might be that media depictions of conflict termination are not deferential to the president’s framing strategy (Schudson 2002; Sparrow 2008). Fourth, since many crises are short-lived, we wonder if there is a non-monotonic relationship between media coverage and public opinion that our variable does not capture. When media attention is quite low and quite high, rallies are unlikely due to unawareness or high contention. When media attention is moderate, it might provide the conducive conditions of public attention and executive autonomy sans opposition influence that our theory outlines. Finally, usage of the NY Times proxy is common in the literature, but largely due to data availability rather than a theoretically ideal motivation. Given arguments of increasing media fragmentation and the role of foreign elites in filling the news hole on foreign crises, we concede that our measure might be weak (Shapiro and Jacobs 2000; Searles and Smith 2016). Ultimately, we note that Baker and Oneal (2001) yield the same unexpected finding (also measuring NY Times coverage) in their study of rally effects, perhaps inspiring a puzzle for readers to take up. In the bottom-left graph of table 3, the moderating effect of economic misery is less obvious because the 95 percent confidence intervals encroach upon the zero line in much of its range. Overall, the decreasing effect of misery is consistent with Hypothesis 4, the positive effect of termination being significant only in the mid-range of the misery index. Even though the evidence for the decreasing positive effect seems weak, an ample portion of observations in the misery index covers the significant range of positive influence. That is, the effect is statistically significant at the 0.1 level in the range of the misery index between 3.5 and 15.6 that contains 92 percent of observations in the regression sample, at the 0.05 level between 4.6 and 14.3 of the index (87 percent of observations), and at the 0.01 level between 7.9 and 11.5 of the misery index (41 percent of observations). Despite the lack of stars in this variable, it appears that the public rewards the president for crisis terminations up to a certain economic threshold, after which citizens reject or overlook this signal of competence. Consequently, we interpret some significant empirical evidence for Hypothesis 4. Finally, the bottom-right graph shows that the positive effect of a crisis termination decreases as the level of partisan conflict increases, consistent with Hypothesis 5, but is not statistically significant anywhere along its range. Thus, we find no evidence that partisan conflict impacts public opinion following a crisis termination. Given its seeming ubiquity of impact on public attitudes in other studies, this finding calls for further research. We wonder if the breadth and depth of perspectives on peace as a public good is not offset even by partisan opposition. If this is the case, then the potency of crisis terminations as a strategic presidential tool would be even greater. Results using the more conservative detrended dependent variable, shown in table A2, yield robust evidence for each of the above findings. Though coefficients are slightly weaker, statistical significance persists. Overall, we find consistent support for three of the hypotheses, controverting evidence for one, and null findings for the last, providing a clearer sketch of operational conditions in crisis termination and public opinion. Validating VignettesTo capture the micro-dynamics in presidential approval over the evolution of an international crisis, we trace President Obama’s daily approval over the course of two crises: the Libyan civil war (from February to October in 2011) and the Syria chemical weapons crisis (from August to September in 2013). Though the leader is held constant, the events vary in duration, character, and exogenous factors. Figure 3 shows the changes in Obama’s approval rating during the Libyan crisis. He hovered at approximately 50–55 percent approval at the onset of the crisis. Excepting the spike catalyzed by Osama Bin Laden’s death, the downward pattern is apparent. When the coalition forces’ military actions begin to produce some visible outcomes toward the end of the crisis, the downward pattern stagnates, fluctuating around or just below 45 percent. Of particular note, when Gaddafi was removed and killed, we observe a significant boost in Obama’s approval. Regarding our control variables, the inflation and unemployment rate did not change dramatically over the duration of this nine-month crisis (though there was a slight increase in inflation in the early period of the crisis), inhibiting clear analysis on that effect. The public sentiment of war weariness and the public condemnation of the humanitarian crisis in Libya might have dragged down presidential approval. In spite of all these unfavorable conditions, the well-ended crisis termination indeed boosted presidential approval, illustrative of our theoretical assertion. Next, we trace a shorter crisis involving the Assad regime’s chemical weapons attack on civilians in 2013, depicted in figure 4. This case provides variation in temporal dynamics and in the tools a rational and opportunistic president utilizes in conflict termination calculus, featuring multilateral cooperation instead of military force. When the crisis started in mid-August, Obama’s approval plunged about 4–6 percent until early September. He appeared determined to condemn the incident and take tangible military action, but the public was not supportive of another military conflict in the Middle East, evidenced by a disturbing plunge in approval at the first show of force. A small surge followed when Obama publicized concrete evidence of the attack, likely depicting a rally of humanitarian concern as information trickled in. Still, when he asked for Congressional approval for military strikes, public support continued to actively decay. When Russia made a proposal for a peaceful resolution to remove the stockpile of chemical weapons in Syria, the public mood shifted to become more amenable. Ultimately, when the Assad regime complied with the United States–Russia framework, Obama’s ratings recovered up to 50 percent. Our theory stresses that presidents strategically select the timing, framing, and format of termination to their benefits in response to contextual factors. While the Syria case is less obvious, it is more nuanced. Though the crisis endured a mere two months, a noteworthy dynamic between presidential approval and presidential behavior is evident, concluding with a surge rewarding Obama’s perceived competence in countering the Syrian crisis with a multilateral solution, despite a sour initial assessment. In general, these vignettes further substantiate our theoretical conjectures and empirical findings. Addressing EndogeneityWe acknowledge the contention that crisis termination is endogenous to presidential approval. Our theory indicates that the timing of crisis terminations, as well as the format and content of publicity, are influenced by domestic political conditions, the international security environment, and standing public opinion. The president has a strong incentive to terminate a crisis when the public is most likely to welcome the decision and in a way that ameliorates public concerns about performance, costs, and outcomes. It seems to follow that presidential approval, serving as a gauge of public mood on the index of presidential performance, can influence the decision to terminate a crisis. We are skeptical about the endogenous effect for several reasons. First, it is not obvious whether the president terminates a crisis to maintain high approval or to avoid the further downturn of low approval. That is, a theoretical path from presidential approval to crisis termination is dubious because the preexisting approval level does not have a unidirectional or systematic impact on termination decisions. Second, this mechanism is embedded in our theory, constituting one of the main reasons for the general hypothesis that termination should have a positive effect on approval. In modeling the president’s strategic selection of termination timing based on a variety of factors, our theory enfolds a degree of endogeneity by design. Third, crisis termination is not solely determined by the intention and behavior of the United States. The overall strategic environment and the domestic political circumstances of an adversary influence the US president’s latitude over the timing and format of termination. Thus, presidential approval is not the singular and may not even be the most decisive factor in a given crisis ending. Nonetheless, to ensure that our empirical findings are not driven by the potentially endogenous relationship, we run two-stage residual inclusion models. The first-stage probit equation estimates the probability of crisis termination using several strong exogenous predictors of termination. The first variable is the Role of Global Organization, measuring the level of key organizations’ involvement spanning zero (nonexistence), one (general organization involved), and two (United Nations Security Council Resolution). We expect that the United States’ unilateral decision to terminate a crisis should become more constrained the greater the influence of global organizations. Results show that this has a negative impact on termination. The second exogenous variable is the Centrality of Violence, an ordinal measure ranging from one to four assessing the role of violence as a preeminent management technique. Because of its higher costs and risks, we expect that termination is more likely when the United States opts to use violence rather than diplomatic means, which is corroborated in the results. In addition, we include cubic polynomials of crisis duration in the first-stage equation. The second-stage equation estimates the effect of crisis termination on presidential approval while controlling for the error correlation between the two equations, a source of simultaneity bias. The second-stage results robustly corroborate our findings. Detailed results are reported in tables A3 and A4 and figure A1 in the appendix. Obama’s daily presidential approval during Syrian chemical weapon crisis (August 21, 2013–September 21, 2013). ConclusionWe set out to explore the relatively uncharted territory of the effects of crisis termination on the public’s presidential approval, theorizing a halo effect akin to rally effects. Building on the rich research on US presidential behavior, public opinion, and international crises, we conceptualized crisis termination as patently popular with the public. We further argued that terminations are a strategic event available to a vote-maximizing executive keenly attuned to its timing and adept in framing. We conjectured that presidents could use their many positional advantages to identify, coordinate, and frame a crisis termination under conditions conducive to reputation. Consequently, we hypothesized a surge in presidential approval following a termination. Recognizing the degree of contingency in presidential leadership and public assessment, we identified several conditioning variables likely to influence this pattern. The degree of public attention to the crisis and its termination, operationalized by election years and the amount of cumulative NY Times coverage, was expected to amplify this swell in approval. Conversely, we expected a poor economy and a context of high partisan conflict to stifle the surge. Our empirical analyses demonstrated support for all but two of the hypothesis, that of NY Times coverage yielding an opposite pattern and partisan conflict having no dampening impact on public assessments. This study contributes to the literature by extending theory and presenting the first empirical model (to our knowledge) of presidential calculus and public opinion at the node of crisis terminations, an area that we believe has the potential to become an important sub-literature. We have demonstrated that there is a robust statistical connection between crisis terminations and public approval, a connection we argue is driven by the same rally mechanisms associated with conflict onsets. We have also demonstrated, however, that many factors and conditions interact with this process, leading us to see fertile ground for and fully invite future research. First, our study is confined to the US context, being well-trodden and data-rich. Further research should extrapolate and contextualize this framework to other polities with similarly powerful and directly-elected executives or even more divergent settings. Insofar as the United States is a preponderant nation, more capable of achieving good outcomes, exploring how termination rallies vary by strength and satisfaction is another important question to take up. Third, we acknowledge that other conditions and contexts might impact the dynamics of crisis termination and public opinion, both in the United States and elsewhere. Identifying and modeling these, both theoretically and empirically, would be a welcome endeavor. We have already identified two puzzles ripe for further research. First, cumulative media reports, which we assume increase crisis salience in public consciousness, have a negative impact on the halo effects of termination. Future work could explore how variation in the qualitative nature of media coverage impacts evaluations of crisis terminations. It could also innovate in the measurement of media coverage, both with the NY Times coverage variable’s distribution and with new data sources. Second, that partisan conflict does not seem to mediate public approval upon crisis terminations is curious, perhaps reinforcing the utility of these events for a strategic executive. We also suspect that relational contexts, such as economic or territorial rivalries, substantively impact public perception and approval that extends to crisis termination. We see potential in research analyzing the attenuating or amplifying effects of IOG dynamics following crisis termination, wondering if they vary in competition with enduring rivals, allies, or perhaps new actors. Taking a cue from the interstate conflict literature, one might expect that the way a crisis is initiated conditions the way the public perceives its termination, calling for a holistic modeling of the crisis sequence. These, and no doubt other omitted variables, await the good work of normal science. Another promising direction for future research would be to analyze the magnitude and duration of termination opinion rallies, contextualizing them in the broader typology of political rally events such as crisis onsets, campaigns, debates, and conventions. For those who do survey work, administering experiments on public opinion at crisis terminations would be a tremendous contribution to our understanding of its inner workings. Finally, while we see strong face validity that crisis terminations might function as diversionary venues, we do not test whether leaders terminate in response to poor domestic conditions. While it appears that terminations perform the requisite operation of diversion, future work could examine if leaders use them as such a tool. We urge further explorations to expand our understanding of the full cycle of foreign crises, including a leader’s incentives and behaviors and the public’s ensuing response. NotesKerry Chávez is a PhD candidate in Political Science at Texas Tech University. Her research focuses on the domestic politics, strategies, and technologies of conflict and security. James Wright is a PhD candidate in Political Science at Texas Tech University. His research centers on American politics, with a focus on voter behavior and tangential interests in American foreign policy. The authors would like to thank Daehee Bak for his instrumental role and mentorship in this study. References, . . “Priming Effects in Complex Information Environments: Reassessing the Impact of News Discourse on Presidential Approval.” (): –., . . “Patriotism or Opinion Leadership? The Nature and Origins of the ‘Rally ‘round the Flag’ Effect.” Journal of Conflict Resolution (): –.. . “The Economy and Presidential Approval: An Information Theoretic Perspective.” InEconomics and Politics: The Calculus of Support , edited by , , , –. :University of Michigan Press ., , . . “The Marginal and Time-Varying Effect of Public Approval on Presidential Success in Congress.” (): –., . . “The Structure of Presidential Approval: Constraints Within and Across Presidencies.” (): –., , , , . . “International Crisis Behavior Data Codebook, Version 12.” ., , . . “Robust Inference with Multiway Clustering.” Journal of Business & Economic Statistics (): –.. . “Everybody Loves a Winner: On the Mutual Causality of Presidential Approval and Success in Congress.” Congress & the Presidency (): –., . . “New Evidence Undercutting the Linkage of Approval with Presidential Support and Influence.” (): –., . . “Conflict, Agency, and Gambling for Ressurection: The Principal-Agent Problem Goes to War.” American Journal of Political Science (): –., . . “Consistent Covariance Matrix Estimation with Spatially Dependent Panel Data.” Review of Economics and Statistics (): –., , . . “War President: The Approval Ratings of George W. Bush.” Journal of Conflit Resolution (): –.. . “Presidential Approval Ratings—Donald Trump.” . ., , . . “Success Matters: Casualty Sensitivity and the War in Iraq.” (): –., . . “Presidential Defense: Decisions and Strategies to Preserve the Status Quo.” Political Research Quarterly (): –., . . “Crossing the Water’s Edge: Elite Rhetoric, Media Coverage, and the Rally-Round-the-Flag Phenomenon.” (): –., , . . “Follow the Leader? Presidential Approval, Presidential Support, and Representatives’ Electoral Fortunes.” (): –., . . “News Coverage of the Gulf Crisis and Public Opinion: A Study of Agenda-Setting, Priming, and Framing.” (): –.. . “What Do We Know About Crisis, Escalation and War? A Visual Assessment of the International Crisis Behavior Project.” Conflict Management and Peace Science (): –., . . “International Crises and Linkage Politics: The Experience of the United States, 1953-1994.” Political Research Quarterly (): –.. . “The Pretty Prudent Public: Post Post-Vietnam American Opinion on the Use of Military Force.” International Studies Quarterly (): –., . . “Should I Stay or Should I Go Now? Partisanship, Approval, and the Duration of Major Power Democratic Military Interventions.” (): –., . . “The Impact of the Gulf War on the Ingredients of Presidential Evaluations: Multidimensional Effects of Political Involvement.” American Political Science Review (): –., . . “The Aggregated Consequences of Motivated Reasoning and the Dynamics of Partisan Presidential Approval.” (): –., , . . “Economics, Party, and the Vote: Causality Issues and Panel Data.” American Journal of Political Science (): –., . . “Veto Power and Legislation: An Empirical Analysis of Executive and Legislative Bargaining from 1961 to 1986.” Journal of Law, Economics, and Organization (): –., , . . “Foreign Policy and Presidential Popularity: Creating Windows of Opportunity in the Perpetual Election.” Journal of Conflict Resolution (): –., , , . . “Are Election Results More Unpredictable? A Forecasting Test.” Political Science Research and Methods (): –.. “January 23, 1973: Address to the Nation Announcing an Agreement on Ending the War in Vietnam.” . ., , , , . . “Good Times, Bad Times, and the Diversionary Use of Force: A Tale of Some Not-so-free Agents.” Journal of Conflict Resolution (): –., . . “Passing the President’s Program: Public Opinion and Presidential Influence in Congress.” American Journal of Political Science (): –., . . “Who Leads and Who Follows? U.S. Presidents, Public Opinion, and Foreign Policy.” InDecisionmaking in a Glass House: Mass Media, Public Opinion, and American and European Foreign Policy in the 21st Century , edited by , , , –. : .. . “Who Speaks for the People? The President, the Press, and Public Opinion in the United States.” Presidential Studies Quarterly (): –., , . . “Modeling and Forecasting in US Presidential Election Using Learning Algorithms.” Journal of Industrial Engineering International (): –.AppendixTable A1.
Table A2. Moderating effects of crisis termination on detrended presidential approval
Table A3. Effect of crisis termination on presidential approval (two-stage)
Table A4. Moderating effects of crisis termination on presidential approval (two-stage)
Table A5. Robustness checks for crisis selection—baseline hypothesis
Table A6. Robustness checks for crisis selection—moderating hypotheses
|