Following the Party in Time of War? The Implications of Elite Consensus

ABSTRACT

Prominent perspectives in the study of conflict point to two factors that exert substantial influence on public opinion about foreign intervention: (1) news about casualties and (2) signals from partisan elites. Past work is limited, however, in what it can say about how these two factors interact. We present an experiment designed to understand the surprisingly common scenario where elites send competing messages about whether the public should support war or oppose it—and these messages do not coincide with party divisions. We find that partisans are generally insensitive to news about casualties, but they become noticeably more sensitive when they perceive within-party disputes over support for the war. Independents, however, respond to news of casualties irrespective of what messages elites send. These findings shed light on when and how the public responds to competing and unclear cues and speak to the role of public opinion in determining conflict outcomes and democratic foreign policy-making more broadly.

KEYWORDS:

• Conflict
• foreign policy
• public opinion

Acknowledgments

The authors wish to thank John Ciorciari, Christian Davenport, James Morrow, Phil Potter, Allan Stam, the anonymous reviewers, and the editor of International Interactions for providing valuable feedback on earlier drafts. All errors and omissions remain our own.

Supplemental data

Supplemental data for this paper can be accessed on the publisher’s website.

Notes

¹ 39% of House Democrats and 58% of those in the Senate voted for the 2002 Iraq War Resolution.

² The question of when those elites detect the possibility of electoral gain is a worthy subject of future study. This electoral motivation logic extends Schultz’s (1998:832) argument regarding a democracy’s ability to signal resolve in a crisis bargaining situation. Political elites of an opposition party will speak out against potential military action when they perceive electoral gain from doing so, but will support the use of force as advocated by the state’s leadership if they do not believe opposition will yield any electoral payoff. (See also Schultz 2001 for a full treatment of the theory.)

³ See Gartner et al. (2004) and Karol and Miguel (2007) as examples of work connecting casualties not just to public opinion, but the fortunes of elected officials. Also, see Kriner (2010) for an extended treatment of how Congressional elites will use events from the conflict to shape the public’s views of it, which in turn, may compel the president to change course. It should be noted here, however, that just because public and Congressional support for the conflict has soured, it does not automatically follow that the president will decide to end the conflict. Indeed, Croco finds that culpable executives—those who were either in charge when the decision to go to war was made or are directly connected to that decision (that is, from the same party)—frequently will not give in and will fight “so long as winning the war… remains a possibility (italics in original)” (Croco 2011:461). While these leaders will be more likely to “gamble for resurrection” (Downs and Rocke 1994), nonculpable executives—those not connected to the decision to involve the state in the war—will be more likely to end involvement in the conflict. For example, even after Republicans lost Congress in the 2006 midterm elections, President Bush did not bow to pressure to end the Iraq War coming from the new Democratic majority or public opinion, as he instead ordered an additional 30,000 troops into the country in early 2007.

⁴ Full question wording is provided in the SI.

⁵ In the SI, we show that our sample compares well with Census benchmarks for race, education, gender, and age.

⁶ The underlying model is Support = B₀ + B₁American deaths + B₂Unified Elites + B₃Deaths× Elites + e, estimated separately (OLS) for partisans and independents. We calculate marginal effects of American deaths from the appropriate main effects and interactions. Corresponding statistical tables are in the SI.

⁷ To determine partisan affiliation, we asked respondents to locate themselves on the following seven-point scale: Strong Democrat, Moderate Democrat, Lean Democrat, Independent, Lean Republican, Moderate Republican, Strong Republican. All respondents who self-identified as Independent were classified as Independents; all others were classified as Partisans.

⁸ We can test H2 more formally by estimating Support = B₀ + B₁American deaths + B₂Independent + B₃Deaths× Independent + e and examining B₃. In this model, B₃ = 0.09; SE = 0.04; p < .05, consistent with the hypothesis. Corresponding statistical tables are in the SI.