Life and Limb: New Estimates of Casualty Aversion in the United States

Abstract

Dramatic improvements in US military medicine have produced an equally dramatic shift in the kinds of battle casualties the US military has sustained in its most recent wars. Specifically, there has been a notable increase in the ratio of nonfatal to fatal casualties. Most studies of casualty aversion in the United States, however, have focused on fatal casualties. Using a series of survey experiments, I investigate whether respondents are equally sensitive to fatal and nonfatal casualties, differences between populations with and without close military ties, and whether views on casualties are conditioned by respondents’ level of knowledge about casualties or the individual costs of war they expect to incur. I find that, while the general public is generally insensitive to different types of casualties, respondents with close ties to the military are better able to distinguish among kinds of casualties. This advantage, however, is not due to respondents with close military ties being better informed about war casualties. Instead, those who bear the costs of war directly appear better able to distinguish among those costs.

Dramatic improvements in military medicine have produced an unprecedented shift in the nature of casualties sustained by US military personnel in armed conflict. While the historical ratio of wounded to killed was stable at 3:1 for centuries, the United States’ wounded-to-killed ratio today is closer to 10:1.¹ The increased wounded-to-killed ratio, which shot up in the recent wars in Afghanistan and Iraq, is beneficial in that many lives are saved that would have been lost in the past. However, it also translates into certain increases in the costs—both human and financial—of war. Many of today's veterans return home with extremely serious injuries that would not have been survived in previous wars. Their futures are significantly altered, as are the futures of the family members and friends who care for them. Nonfatal casualties, however, have rarely figured into how political scientists, policymakers, and pollsters frame the relationship between public opinion and the decision to use military force because the debate around “casualty aversion” has nearly always equated casualties with fatalities.

The changing nature of US military casualties raises at least three critical questions. First, to what extent are voters sensitive to the possibility of nonfatal versus fatal casualties in considering support for military action? Second, given that less than one-half of 1 percent of the US population currently serves in the armed forces, are attitudes regarding nonfatal casualty aversion segregated by social proximity to the military? And third, if there are differences between populations with and without close ties to the military, are these differences driven by those with close ties to the military being better informed about these changing costs of war, aware that they are likely to bear the brunt of paying these costs of war, or both?

Via a series of survey experiments, I find that information regarding the war wounded rarely affected support for military action among a representative sample of US respondents. Respondents with close military ties, however, were better able than respondents without such ties to rank the human costs of war. These results also shed light on why respondents with close military ties are more sensitive than the general public to the human costs of war. It is not, as one might expect, because they were better informed; respondents with close military ties were in fact generally worse than those without such ties at estimating casualty ratios. It is more likely that this difference is due to two factors: those with close military ties know they will bear the costs of war directly, and those with close military ties may also hold a particular set of values that makes voluntary participation in military service more likely. This finding is consistent with previous research that demonstrates a strong inverse relationship between the imposition of direct costs of war and support for military deployment (Horowitz and Levendusky 2011; Flores-Macías and Kreps 2015).

In addition to the human costs of war borne by military personnel and their families, the US Government bears the financial costs of pensions as well as disability and medical payments for the returned wounded. While it is difficult to put a precise number to the increasing financial costs associated with the downstream effects of improvements in military medicine, one recent effort estimates the future costs of caring for veterans of the Afghanistan and Iraq wars to be between |${\$}$|600 billion and |${\$}$|1 trillion (Bilmes 2011, 1). Previous estimates of the financial costs of war, however, have tended to ignore the war wounded. For example, the Congressional Research Service (CRS), when estimating the future costs of war, appears to focus more on factors such as transportation and meal costs for deployed military personnel than on the costs that must be borne once injured service men and women return home. This is so even though the CRS explicitly includes the military benefits system as part of its description (but not analysis) of the costs of war (Belasco 2014, 3). Similarly, with few exceptions, the scholarly literature on public opinion and the use of force has defined war casualties almost exclusively in terms of fatalities, as have organizations such as Pew and Gallup in their survey questions. This paper is the first, to my knowledge, to focus extensively on how and whether information about the war wounded might affect public support for military deployment in the United States.

The remainder of this paper proceeds as follows. First, I describe the improvements in military medicine that have produced the increased wounded-to-killed ratio. I then show that previous literature, as well as polls, have only very rarely included the war wounded in analyses of public support for war and, in particular, in analyses of casualty aversion. Following this discussion, I discuss theoretically the possible relationship between information about the war wounded and public support for military deployment, including a possible gap in civilian versus military views on casualty aversion and the use of force. Next, I describe and then present the results of a survey experiment meant to explore how the public reacts to fatal and nonfatal casualties; I also discuss results from three follow-up experiments. My conclusions follow.

Changes in Military Medicine and Associated Costs

The historical wounded-to-killed ratio was 3:1 (Clodfelter 2008, 4). Recent years have seen a dramatic shift in medical care in conflict zones that has translated to a more than threefold increase in this statistic. These improvements in military medicine are especially present in the US military. Four trends have radically improved the US wounded-to-killed ratio, particularly since the commencement of OEF and OIF (Fazal 2014). First, improvements in preventive care, including child nutrition, immunization policies, and field sanitation, have greatly reduced the likelihood that soldiers fall ill during military campaigns. Healthier soldiers are less likely to succumb to any wounds sustained; they also contribute to full complements of fighting forces, which bears on the survivability and health of the unit as a whole. Second, improvements in battlefield medicine, such as recent attention to stemming blood loss and logistical reorganization of medical forces in theater, have been pioneered by the United States; US forces have, in turn, benefited greatly from these changes. Third, modern medical evacuation practices mean that injured soldiers can often be transported to advanced medical facilities in less than an hour, compared to the days (sometimes weeks) that wounded soldiers used to lie on the battlefield. Fourth, today's personal protective equipment (PPE) means that the two parts of the body most vulnerable to fatal wounds—the head and the trunk—are better protected today than in the past. Use of such equipment on the battlefield has been credited with saving hundreds of lives (Koerner 2004; Moss 2006; Tong and Bierne 2013).

Improvements in military medicine might be only part of the story of the United States’ improved wounded-to-killed ratio. The United States’ most recent wars have been qualitatively different from many of its previous conflicts. The United States has stopped fighting conventional wars and switched to counterinsurgency. Some have attributed the United States’ improved casualty statistics to the combination of, for example, the widespread use of PPE and the changed nature of warfighting on land (Brevard, Champion, and Katz 2012). While it is true that most injuries sustained by United States forces today result from explosive weapons, the shift from guns to bombs as the primary source of wounds occurred during World War II (Fazal 2018). Given that the injuries produced by explosives are generally more serious than those produced by guns, this shift would suggest an overall decrease in the wounded-to-killed ratio, rather than the increase we actually observe. Thus, the effects of military medicine on the wounded-to-killed ratio may be even stronger than they appear.

The main result of these changes is a dramatic rise in the United States’ wounded-to-killed ratio (see Figure 1). The wounded-to-killed ratio is determined by dividing the number of military personnel wounded in action who survived by those killed in action. It does not include the disease-non-battle-injury (DNBI) rate, minor wounds followed by a return to combat within 72 hours, or other types of casualties; for example, soldiers dying or suffering from malaria or automobile accidents are excluded from this ratio. At its peak, the official US wounded-to-killed ratio was 10.4:1.

Figure 1.

US wounded-to-killed ratios, 1776–2003.

Source: Department of Defense²

A secondary result of this shift is an increase in costs of caring for wounded veterans. In large part because of these changes in military medicine, the nature of wounds sustained and survived by today's veterans has shifted. Better PPE, for example, makes traumatic brain injury (TBI) and facial disfigurement significantly more likely than in the past.³ Whereas in previous wars, soldiers who lost limbs in combat might have died due to blood loss, modern hemostatic practices stem hemorrhage and save lives. However, these lives are indelibly changed. While the costs of caring for the returned wounded are borne principally by that veteran's family and friends, these costs are also paid by taxpayers. Higher numbers of returned wounded mean more pensions paid by the Defense Department⁴ and increased pressure on the Department of Veteran's Affairs to provide new types of medical care for injured soldiers. For example, Linda Bilmes has now revised her and Joseph Stiglitz's original estimates of disability and medical care costs for veterans of OEF and OIF up by approximately |${\$}$|200 billion; this increase is driven by the unanticipated number—and unprecedented rate—of the returned wounded, which was caused in turn by the dramatic improvements in military medicine outlined above (Bilmes 2011, 3). Thus, the wounded are a much more relevant category—at least in terms of numbers—for analyses of casualty sensitivity today than in the past.

“Casualty” Aversion in Context

Perhaps the most enduring link between public opinion and the use of military force in the scholarly literature has been the casualty aversion hypothesis. Casualty aversion implies a cost–benefit analysis, where the main cost is casualties incurred by (in this case) US forces. As casualties go up, public support for war should decrease. This argument was most prominently launched by John Mueller, who identified an inverse logarithmic relationship between the number of US casualties and public support for the Korean and Vietnam Wars (Mueller 1973, 61). The intuition supporting the casualty aversion hypothesis is attractively simple, and it has surfaced as part of many arguments in international relations from the democratic peace (in Kant's version, democratic publics prefer not to go to war because they are casualty averse) to audience costs (leaders will be punished for performing poorly in war, where one measure of performance is casualties sustained) (Reiter and Stam 2002; Filson and Werner 2004; Potter and Baum 2010; Lee 2016). Several recent polls have supported the notion that the US public is casualty averse. Eichenberg and Stoll find a fourteen-point decrease in support when casualties were mentioned in pre-Iraq War polls, and in 2016 the Chicago Council on Global Affairs reported a marked preference for the use of drones or even manned aircraft—where the likelihood of casualties is extremely low—over the deployment of ground troops to combat terrorism abroad (Eichenberg and Stoll 2004, 16; Global Views 2016: US Public Topline Report 2016, 45–46).

Refinements of Mueller's original claim have focused on the timing of casualties, the notion of “proximate” casualty aversion, and the demographics of the military, among other factors. Gartner found strong support for the casualty aversion hypothesis when examining monthly data in relation to casualty reports. “Both higher monthly casualties and increasing casualty trends erode support for military action” (Gartner 2008, 103). Gartner as well as others also advanced the notion that casualties close to home matter most in inducing casualty aversion; the authors of one of these studies argue that, had there been fewer casualties of the Iraq war, President Bush would have won a larger share of the vote in the 2004 presidential election (Gartner, Segura, and Wilkening 1997; Karol and Miguel 2007; Hayes and Myers 2009). Another angle on the casualty aversion hypothesis speaks to who participates in the military and thus bears the most direct costs of deployment. In this vein, Horowitz and Levendusky find that reinstating the draft would decrease support for military deployment; they attribute this finding to the possibility of casualties distributed randomly across society, rather than concentrated to the population participating in or related to an all-volunteer force (Horowitz and Levendusky 2011).

Challengers to the casualty aversion hypothesis have taken aim at the link between casualties and public opinion on several fronts. Gelpi et al. argue that it is the prospect for military success, and not the possibility of incurring casualties, that determines public support for war (Jentleson 1992; Gelpi, Feaver, and Reifler 2009). Berinsky as well as Althaus et al. are especially critical of the claim, underlying the casualty aversion hypothesis, that voters know enough about casualties incurred such that these data can inform their view on war (Berinsky 2007; Althaus et al. 2014). Berinsky further argues that it is partisanship and, in particular, cues from partisan elites, that drive public opinion, and not public opinion that drives elite decision-making on war (Zaller 1992; Berinsky 2009; Saunders 2015).

An important strand of literature from political psychology suggests a different perspective on casualty aversion. Voters, those writing from this school of thought argue, respond to political decisions based on values rather than cost–benefit calculations (Rathbun et al. 2016; Bayram 2015, 2017). One set of values, discussed below, may be expressed via patriotism or a “warrior culture,” potentially suggesting differences in responses from voters with and without close military ties. Another literature from psychology on “mortality salience” cues respondents on life-threatening events, then gauges a range of behaviors, from the proclivity toward charitable giving to respondents’ ability to interpret facial expressions correctly (Burke, Martens, and Faucher 2010). In theory, mortality salience could help us understand whether those with close military ties will be more or less casualty sensitive than those without close military ties. In practice, however, very few studies of mortality salience have been conducted on military populations (for an exception, see Anaki, Brezniak, and Shalom 2012). What is more, mortality salience and related literature on terror management theory have been called into question of late (Klein et al. 2019).

All the literature described above defines casualties almost exclusively in terms of fatalities.⁵ This is somewhat surprising given that Mueller appears to have included both the dead and the wounded in his definition of casualties (Mueller 1973, 36 (Figure 2.1), 168, and elsewhere), including in at least some of his data analysis. It is also surprising because the dictionary definition of casualties includes the dead and wounded, and has done so for centuries.⁶ Webster's offers the following definition of casualty in the military context: “a member of the armed forces who is lost to active service through being killed, wounded, captured, interned, sick, or missing” (Neufeldt and Guralnik 1988, 218–19).

Figure 2.

Support for military deployment (all groups).

While there are exceptions to the rule that scholars of public opinion and the use of force have equated casualties with fatalities, they are relatively minor. Karol and Miguel include the wounded in their national-level analysis of casualty aversion, but not in the more fine-grained analysis of the proximate casualty aversion hypothesis that most interests them (Karol and Miguel 2007). Gelpi et al. (2009) explicitly define casualties as fatalities on the first page of their book, but do conduct an experiment as a robustness check to assess casualty sensitivity to the wounded.⁷Kriner and Shen (2016) likewise conduct experiments on nonfatal casualties as part of a larger analysis of the relationship between economic inequality and military participation (Kriner and Shen 2016). While others might mention the wounded as part of the general concept of casualty,⁸ empirical analysis has focused on fatalities as the key measure of casualties.

This contracted definition of casualty is mirrored in general public opinion surveys. The wounded were not mentioned in any of the various questions fielded from January to July 2015 compiled by pollingreport.com on the possibility of deploying ground troops in Iraq and/or to fight the Islamic State.⁹ In a review of nearly 1,000 questions fielded by organizations including Pew/USA Today, Gallup, NBC News/Wall Street Journal, Fox News, CNN, CBS News/New York Times, and ABC News/Washington Post on the Afghanistan and Iraq Wars, I could not find a single example of a question that referenced the wounded.¹⁰ Indeed, relatively few of the questions referenced fatalities at all, a finding consistent with Althaus et al.’s content analysis of news reporting that revealed only rare discussions of fatalities in the news (Althaus et al. 2014, 202, Figure 2a).

The historical 3:1 wounded-to-killed ratio made the decision to focus on fatalities rather than fatalities plus the wounded in polling and scholarship on casualty aversion relatively unproblematic—not only was this ratio fairly static over a very long period of time, the number of wounded was not much larger than the number of killed. However, with today's improvements in military medicine, the returned wounded constitute an increasingly important sector of the costs of war. Rectifying this problem, however, is challenging for analyses based on observational data. The main reason Karol and Miguel (2007) did not include the war wounded in their geo-analysis of the proximate casualty hypothesis was that the necessary data on the war wounded were not available. While Kriner and Shen (2016) were able to obtain some data along these lines, they acknowledge tremendous difficulty in this type of data collection. The George W. Bush administration was reluctant to release data on US fatalities following the start of OEF and OIF for fear of the “Dover effect”—the possible depression of public support for war in response to seeing images of military coffins being unloaded at Dover Air Force Base (WNYC 2009). However, political leaders may also anticipate a “Walter Reed Effect.”¹¹ Seeing soldiers with multiple amputations, facial disfigurement, and other more severely debilitating injuries is not only testament to the miracles of modern military medicine, but also reveals an especially saddening aspect of the consequences of war. It should not be surprising that governments—including the US government—have been shy to offer much information on the war wounded.

Theoretical Expectations

Insofar as voters in the United States are sensitive to casualties because of the effect on human life, or even because casualties are equated with financial costs in some way, this sensitivity should translate into aversion to incurring all casualties—the killed as well as the wounded. But are all voters equally sensitive to these costs? In designing the survey experiment discussed below, I considered three main possibilities.

First and most straightforward, respondents may be equally sensitive to fatal and nonfatal casualties. One way to assess this relationship is to compare across different types of casualty. Another is to compare reactions to different numbers of casualties. Research in support of the casualty aversion hypothesis finds that the number of casualties—particularly when matched to specific localities and media markets—has an inverse effect on the support for war (Mueller 1973; Gartner, Segura, and Wilkening 1997; Gartner and Segura 1998; Karol and Miguel 2007). Previous research has also examined the effect of different numbers of casualties (almost always cast in terms of fatalities) to try to identify a threshold effect for public support, although with mixed results.¹² If voters are equally sensitive to the number killed versus the number wounded, we ought not observe a statistically significant difference between those given information about the same number of killed versus wounded. However, if they are sensitive to increased numbers of casualties, this effect also ought to be observable across treatments.

Second, sensitivity to the war wounded could be conditional upon the type of wounds sustained by military personnel. Death is final and clear, but “wounds” could span many types of injuries. The injuries that are severe enough to place US military personnel in the “wounded-in-action” category today are typically life changing, but the range across which life is changed is very wide. Over 325,000 US military personnel have experienced TBI from 2000 to 2015. Another 1,600 have suffered amputation over the same period (Fischer 2015). These injuries, while often permanent, do not necessarily preclude productive economic reintegration into society for wounded veterans. However, another, unknown, percent of veterans is left unable to care for themselves as a result of battle wounds. Describing this latter category of injuries should decrease support for military deployment.

Third, two distinct logics suggest a civil–military gap in casualty toleration that could be exacerbated by information about the war wounded. The first argument is that respondents most likely to incur the costs of caring for the returned wounded—those with close ties to military personnel—will also be most likely to oppose military action that produces a large number of the wounded. The relationship between military personnel (and their families) and support for military deployment is complex (Golby, Cohn, and Feaver 2016). Previous research has differed on whether draft-age men and their families would support or oppose military deployment, with Mueller not identifying any significant difference, but Horowitz and Levendusky (2011) identifying a significant decrease in support for military deployment among respondents faced with the possibility of conscription. More generally, the conventional wisdom is that the military and their families tend to oppose the use of force, in part because of the known risks to themselves and their loved ones. This intuition is consistent with scholarship comparing the casualty sensitivity of civilian versus military elites, which has found military elites to be more casualty averse than their civilian counterparts (Feaver and Gelpi 2004, especially Chapters 4 and 5). This logic would suggest that information about the wounded would deepen any previously held opposition to military deployment in this subsample of respondents, as these are the people who would pay the costs of being wounded themselves or caring for a severely wounded family member. It would also make sense that this group of respondents is more likely than the general population to have observed the nature of wounds sustained by recent veterans, and to understand the time and care these veterans will require.

However, a countervailing effect is also possible. To the extent that the rank and file—versus the elites—of the US military subscribes to a “warrior culture” (Mead 1999/2000), they may be more tolerant of casualties than the general public. This type of logic is consistent with the argument that values, rather than cost–benefit calculations, drive policy attitudes (Rathbun et al. 2016; Bayram 2017). It is not clear whether such a cultural frame would generate greater or lesser reaction to information about the war wounded. On one hand, a wounded warrior may represent a cultural challenge to this ethos, and those wounded could be shunned in some way. This logic would suggest that information about the war wounded could significantly depress support for military deployment among those with close military ties. On the other hand, the notion, perhaps embedded in “warrior culture,” that soldiers are making a noble sacrifice may carry over into casualty sensitivity. Soldiers and their families, in other words, may proudly bear the human costs of war. Importantly, while both these logics are often discussed in US foreign policy, they have been more explored in analyses of elite rather than mass military opinion.

More generally, we should expect those respondents feeling the direct costs of war to react differently from respondents who are more removed from the costs of war. Of course, it could also be that reactions to the finality of death in contrast with the survival of the returned wounded are such that all respondents weigh fatalities much more highly than casualties, and therefore we ought not expect to observe any effect of amending the definition of casualties to include information about the wounded. In focusing on the costs of war associated with the returned wounded, I certainly do not mean to diminish the pain of losing a life. However, it is important to consider the possibility that, for some respondents, a life saved but forever altered might be a worse outcome than a life lost.

Experiment Design

To assess the impact of information on the battle wounded on public support for military deployment, I conducted a survey experiment embedded in a US national survey fielded by YouGov from February 22 to March 1, 2016. YouGov uses stratified sampling to match respondents to a representative sample of the US population; for this survey, 1,480 respondents were matched down to a final sample of 1,400. At my request, the sample was further weighted by YouGov to overrepresent respondents who knew someone—even an acquaintance—in the military. Thus, half my sample knew someone in the military. Just over a quarter of my sample had close ties to someone in the military.

All respondents received three or four (depending on respondent) pre-treatment questions. The first question asked whether they had ties to someone in the military—whether they themselves had served, whether they had a family member who had served, or whether they knew someone who had served. I was especially interested in respondents with close ties to someone in the military, where either they or an immediate family member (a parent, child, spouse, or sibling) had served. Just less than fourteen percent (13.71 percent) of respondents had served, or were currently serving (where service includes the National Guard and Reserves) in the US military, and a similar percentage (13.88 percent) had an immediate family member who was a veteran or currently serving member of the US military.¹³ While I requested oversampling of respondents with military ties, these numbers are fairly close to those in the general population. A 2012 Gallup Study reported that approximately 13 percent of US citizens were veterans, and less than half of 1 percent of US citizens serve in the military today (Newport 2012; Eikenberry and Kennedy 2013).

The next two pre-treatment questions asked respondents to identify the most critical foreign policy issue facing the United States today, and also tested their knowledge about the Islamic State, which provides the basis of the vignette used in the survey. From February 1, 2016 to March 1, 2016, the Islamic State was very much in the news, with 198 stories reported by The New York Times that included the phrase “Islamic State” during this period. Respondents attentive to the news would have learned (in reverse chronological order) about US presidential candidates’ approaches to addressing military gains made by the Islamic State (including military deployment, as discussed in the survey experiment vignette), the travails of daily life under the reign of the Islamic State that produced large numbers of refugees, and efforts to create a cease-fire that would deliver humanitarian aid to Syria. When asked about the most pressing foreign policy issues of the day, 43 percent selected militant groups such as the Islamic State, 21 percent selected immigration from Central America and Mexico, 20 percent selected cybersecurity, 5 percent chose trade with Asia and/or Europe, and another 5 percent were primarily concerned about Chinese expansion into the South and East China Seas. When asked to identify the country(ies) in which the Islamic State primarily operates, 82.5 percent of respondents correctly chose Iraq and Syria.

The survey instrument employed a 2 × 4 design, with seven treatment groups and one control group that received the following baseline scenario:

Recent events in Iraq and Syria, specifically military advances made by the Islamist militant group known as the Islamic State, have sparked debate on whether the U.S. should deploy ground troops to Syria. There is bipartisan support for at least a limited deployment. Defense experts agree that any significant military intervention would be logistically challenging, financially costly, and last at least three years.

Many experts also argue that intervention to halt the advance of the Islamic State would help protect U.S. interests. Would you support the deployment of U.S. troops to Syria to fight the Islamic State?

Each treatment group received the baseline scenario. The additional information included in the seven treatment arms varied along the dimensions of the number and/or nature of casualties, as described in Table 1. For example, in Treatment 1, respondents were also told that “Estimates suggest that at least 1,000 US military personnel would die in combat in such a mission. US battle fatalities today often result from the detonation of improved explosive devices or small arms fire.” In Treatment 6, respondents were told: “Estimates suggest that at least 1,000 US military personnel would die in combat in such a mission, and an additional 10,000 would suffer serious physical wounds. Physical battle wounds today often lead to TBI, amputation, or facial disfigurement.” Treatment 7 retains the 10:1 wounded-to-killed ratio, but paints a more severe picture of the wounded: “Physical battle wounds today often leave soldiers permanently disabled and unable to care for themselves.” Note that the language of “less” or “more” severe injuries was not used in the survey instrument, but is used here as a shorthand.

Table 1.

Experimental groups

The information on war casualties included in the experiment is as accurate as possible, to address concerns regarding external validity. The number of 1,000 fatalities is based on a running three-year average of averages for the Iraq and Afghanistan wars.¹⁴ For treatments that included information about the number of the wounded alongside the number of killed, I used a 10:1 wounded-to-killed ratio, which is at the high end of the current ratio for the United States. I chose this higher ratio to try to create an easy test for the hypothesis that the numbers of the war wounded could affect respondent support for military deployment. Information about the type of wounds sustained is accurate. For example, in 2008, there were 200,000 US troops in theater in Iraq and Afghanistan, 97 of whom received major amputations (Belasco 2009; Fischer 2015).¹⁵ I also included information on how various types of fatal and nonfatal wounds are sustained; these were taken from military medical textbooks and also drawn from my own interviews with military physicians.

The use of round numbers in the experiment may raise eyebrows, but is an accurate portrayal of how casualty numbers are projected in government and in the media. For example, the title of a Los Angeles Times article in the lead-up to the 1990 Gulf War reads “Potential War Casualties Put at 100,000” (“Potential War Casualties Put at 100,000: Gulf crisis: Fewer U.S. troops would be killed or wounded than Iraq soldiers, military experts predict.” 1990). Closer to the example used in this experimental design, a 2002 Brookings Institution report on possible US casualties in the then-imminent Iraq War also used round numbers; numbers of deployed were similarly reported as round numbers (O'Hanlon 2002; Sanger and Shanker 2002). One reason for the common use of round numbers in projected casualties is that such projections are based on formulas that encourage rounding (Dupuy 1990). Even post-hoc reporting of casualties tends to use round numbers as placeholders (Lutz and MacLeish 2020). Thus, even if respondents are more likely to react to precise casualty numbers, such information is typically presented in round numbers, therefore contributing to the study's external validity.

To ensure internal validity, I intentionally included some treatments (shaded in gray in Table 1) that were not externally valid. One possible concern with the experiment design could be that respondents were being cued by the differences in numbers of killed versus wounded (i.e., the 10:1 wounded-to-killed ratio) rather than by the different types of casualties. To establish whether this type of cuing could be occurring, I included Treatments 2 through 5, which first allow comparison across constant numbers of killed and wounded (Treatment 1 versus Treatment 2 versus Treatment 3), and then allow comparison across different numbers of the same type of wounded (Treatment 2 versus Treatment 4 and Treatment 3 versus Treatment 5). None of these various combinations and pairings yielded statistically distinct results, suggesting that respondents were not, in fact, being cued by different numbers of war casualties. The full survey instrument is included in the online appendix.

Respondents were randomly assigned either to the baseline scenario or to one of the seven treatment groups, and asked about their level of support for military deployment, based on a five-point scale ranging from “strongly oppose” to “strongly support,” with “don't know” as an additional, sixth category. Respondents in the control group as well as those receiving the first through fifth treatments, which provided limited information about war casualties, also received follow-up questions testing their expectations regarding the number of US military personnel who could be killed and/or wounded in such a deployment.

Experimental Results

Across all treatments, an average of 41 percent of respondents supported military deployment. This number is lower than that for a CNN/ORC Poll fielded approximately two months prior to my survey, when 49 percent of respondents favored sending US ground troops to Iraq or Syria to fight the Islamic State, and for a Quinnipiac University Poll also fielded two months prior to my survey, which reported 52 percent of respondents supporting the notion of sending ground troops (Agiesta and Schliefer 2015; U.S. Voters Oppose Syrian Refugees, But Not All Muslims, Quinnipiac University National Poll Finds 2015).

The Effect of the War Wounded on Support for Military Deployment

There are several comparisons of interest in analyzing the results of the survey experiment. The first is to examine support for military deployment for the baseline scenario plus all the treatment conditions, as presented in Figure 2.¹⁶ Forty-one percent of respondents in the control group supported military deployment. None of the treatments produced effects statistically significantly different from the baseline. The largest difference in support is a drop for respondents receiving information about those killed only (T1), who were about 10 percent less likely to support deployment compared to those in the control group (37 percent versus 41 percent).

One concern regarding these results could be that respondents were being cued by the larger combined number of casualties in the final treatment. The survey was designed to address this concern, and the results should allay it. If the number of casualties was driving the results, we should observe a statistically significant difference in support between the treatment group receiving the “1,000 less severe wounds” (T2) condition versus the treatment group receiving the “10,000 less severe wounds” (T4) condition, but we do not observe such a difference. In theory, it could also be that respondents were being cued by additional information in the final treatment, where they receive information about the killed and the wounded. However, the addition of information in treatments compared to the baseline did not yield any statistically significant differences. More important, such additional information would be expected to depress support for military deployment, whereas the opposite effect is observed here.

The Role of Military Ties

Perhaps the apparent inability to distinguish among different types of casualties is due to lack of knowledge and understanding of military casualties today. We might therefore expect that respondents with close ties to the military would be better able to distinguish among casualty categories. As Golby, Cohn, and Feaver (2016, 110, 127) have found, the general US public may be decreasingly knowledgeable about the military and military matters.

I hypothesized that respondents with close ties to the military would be especially sensitive to the costs of the war wounded, as they would be responsible for caring for the war wounded upon their return. The results, however, initially confound these expectations. Respondents with close military ties in nearly every treatment group were significantly more likely to support military deployment compared to respondents without such ties, as shown in Figure 3. The two exceptions were the final two treatments, where respondents were given information about those killed and wounded. Here, respondents with close military ties were less likely to support deployment compared to those without such ties.

Figure 3.

The role of military ties.

Respondents with close military ties were generally more supportive of military deployments when given information about the walking wounded as opposed to the killed; this difference was statistically significant when the numbers of wounded and killed were equal, at 1,000 (Treatment 1 versus Treatment 2). This finding is intuitive; all else equal, military personnel and their families would prefer their loved ones to return home with “less severe” wounds than not at all. Perhaps also not surprisingly, there were no statistically significant differences between those respondents with close military ties who received prompts about the killed only versus prompts about wounds that would leave a soldier unable to care for herself. Here, this subset of respondents may be weighing the emotional costs of losing a loved one permanently similarly to the financial and other costs of caring for a permanently disabled family member (or being permanently disabled themselves). This set of respondents was also generally less inclined to support military deployment when given information about wounds that would leave a soldier unable to care for herself than when given information about the walking wounded, but this difference was only statistically significant in one of three pairings (10,000 less severely wounded [Treatment 4] versus 10,000 more severely wounded [Treatment 5]).

A Rational Preference Ordering?

Figure 4 presents a subset of the results in Figure 3, but focuses on the set of comparisons I am most interested in (highlighted in yellow in Table 1). The results suggest a clear civil–military divide in the ability to rank different casualty scenarios. While those respondents without close military ties did not distinguish among the various casualty treatments (as seen most clearly in Figure 2), those with close military ties responded in ways more consistent with a basic intuition about the additive effect of casualties.

Figure 4.

Casualty aversion in context.

Among respondents with close military ties, support for deployment drops ten points when moving from the baseline to Treatment 1 (1,000 KIA), three points when moving from Treatment 1 to Treatment 6 (1,000 KIA plus 10,000 “less severe” wounds), and seven points when moving from Treatment 6 to Treatment 7 (1,000 KIA plus 10,000 “more severe” wounds). Although none of these differences are statistically significant in moving from one bar to the next, the two final treatments are statistically significantly different from the baseline, and the “KIA only” treatment is nearly statistically significantly different (p = .12).

Information Effects and Military Families

That respondents with and without close military ties responded differently to prompts about various types of casualties is not necessarily surprising. However, it raises at least two important questions. First, why were respondents with close military ties more willing to use force overall than respondents without such ties? And second, why are people with close military ties better able to rank the various combinations of the human costs of war?

Greater overall support for military deployment among respondents with close military ties may be explained both by factors that generate selection into the military and the effect of military service on political attitudes. A significant proportion of those serving in today's all-volunteer force come from military families, where there is a tradition of service that is often imbued with the notion of sacrificing for the nation. Also, as Suzanne Mettler argues, even for those military personnel (and their families) who served in eras of conscription, a consequence of military service may be increased civic engagement and commitment to the nation, which may in turn translate into a greater willingness to make sacrifices (Mettler 2005). Those with close ties to the military may be generally more supportive of military deployment than those without such ties—even when the human costs of war are presented to them in stark terms—because they view the human costs of war as worthwhile, particularly within the broader military and political context. In this regard, their values with respect to sacrifice, military action, and casualties may well differ somewhat from those in the larger population. What is more, these preferences and values could be amplified by being in a network of other people with close military ties.

Even though they were surprisingly likely to support military deployment generally, respondents with close military ties were unsurprisingly better collectively at ranking the human costs of war. Perhaps respondents with close military ties are better able than respondents without those ties to assess the human costs of war because they are better informed about these costs. If they live in military communities, such as bases, and/or are members of military families, they could have seen people returning home with injuries. They would have understood, at least to some extent, the impact of caring for the returned wounded on both veterans and their families. Some of the post-treatment questions from the survey experiment allow us to explore this line of inquiry. Respondents in the control group—who received no information about casualties—were asked how many wounded and killed they expected as a result of the military action after responding to the initial prompt regarding support for such action. On average, respondents without close military ties in the control group revealed priors that hewed more or less to a 10:1 wounded-to-killed ratio (at 9.62:1). If respondents with close military ties are better informed about military casualties, we would expect this subset of respondents to be more likely to provide the correct answer than respondents without close military ties. In fact, however, respondents with close military ties tended to underestimate the wounded-to-killed ratio (at 7.5:1), on average. Similar follow-up questions were given for treatments in which respondents received information about the killed or wounded only. Across these treatments, respondents without close military ties were consistently better than those with close military ties at estimating the wounded-to-killed ratio. Based on these data, it does not appear that respondents with close military ties are better informed about casualty ratios.

Even if they are no better able to estimate casualty ratios, we might still expect that respondents with close ties to the military would be better informed of the costs of caring for the war wounded, due to network effects. I included a series of post-treatment, post-measurement questions—asked only of respondents with close military ties—about whether these respondents knew someone who had been killed or wounded in Iraq and/or Afghanistan. As reported in Table 2, a slight majority did not know anyone who had been killed or wounded.

Table 2.

Did respondents with close military ties know someone killed or wounded in war? *

Nature of relationship	KIA (percent)	WIA (percent)
Close family member	3.3	5.6
Extended family member	5	5.8
Friend	11.3	13.4
Acquaintance	9.1	10.5
Friend of a friend	24.5	17.9
Did not know anyone	53.7	53.7

Note: *Note that these percentages will not necessarily add to 100, as some respondents knew people in more than one category, and others did not respond at all.

The anticipated network effects that could diminish support for military deployment among those with close military ties appear to have been overestimated. It is not the case, for example, that a majority of those with close ties to the military know someone who returned from war having been wounded, and thus can easily observe the costs to family and personal lives of recovering from, or living with, such wounds. What is more, when asked who was providing or paying for care for those wounded, a clear majority of respondents replied that they did not know. This evidence does not support the notion that respondents with close military ties were better than those without such ties at assessing the costs of war because they are better informed. As I argue below, it is more likely that an understanding that they will bear the direct costs of nonfatal casualties drives the differences between civilian and military populations in this regard.

Regression Analysis

Table 3 introduces standard control variables in logistic regressions on the YouGov survey results. The results line up with results from previous studies in that Democrats are significantly less likely to support military deployment compared to other demographic groups; they are also consistent with the claim that there is a shrinking gender gap in support for military deployment (Eichenberg 2003; Boettcher and Cobb 2006). For the analysis on the general population (Model 1), none of the treatments generated coefficients that even approached statistical significance. For the analysis on the subset of the population with close ties to the military, the results were (very) slightly more encouraging: respondents in the final treatment group—where fatalities plus soldiers suffering wounds leaving them unable to care for themselves are highlighted—were statistically significantly less likely to support military deployment compared to the baseline. The results regarding possible network effects were mixed, in that, among respondents with close military ties, those with ties to the Army were consistently less likely to support military deployment compared to those with ties to other branches of service. Knowing someone who had been killed or wounded, however, did not significantly affect support for deployment. This result may be an artifact of low statistical power, or it may be due to the possibility that partisanship effects overwhelm treatment effects (Golby, Cohn, and Feaver 2016, 129).

Table 3.

Logistic regressions on support for military deployment

	Close military ties only for all remaining models (2–4)
Full sample	(1)	(2)	(3)	(4)
1k KIA	−0.19 (0.24)	−0.40 (0.44)	−0.37 (0.44)	−0.34 (0.44)
	P = 0.44	P = 0.36	P = 0.40	P = 0.44
1k less severe wounds	.11 (0.24)	.19 (0.47)	.22 (0.48)	.18 (0.49)
	P = 0.65	P = 0.69	P = 0.65	P = 0.71
1k more severe wounds	−0.01 (0.24)	.15 (0.50)	.13 (0.50)	.08 (0.51)
	P = 0.97	P = 0.76	P = 0.79	P = 0.87
10k less severe wounds	.16 (0.24)	.60 (0.48)	.55 (0.48)	.48 (0.49)
	P = 0.51	P = 0.21	P = 0.26	P = 0.33
10k more severe wounds	−0.05 (0.24)	−0.40 (0.48)	−0.46 (0.49)	−0.47 (0.49)
	P = 0.83	P = 0.41	P = 0.35	P = 0.34
1k KIA + 10k less severe wounds	.26 (0.24)	−0.36 (0.45)	−0.38 (0.45)	−0.40 (0.46)
	P = 0.28	P = 0.42	P = 0.40	P = 0.38
1k KIA + 10k more severe wounds	−0.16 (0.24)	−0.74 (0.45)	−0.79 (0.45)	−0.87 (0.46)
	P = 0.50	P = 0.10	P = 0.08	P = 0.06
Close military ties	.35 (0.14)
	P = 0.01
Army			−0.55 (0.25)	−0.62 (0.25)
			P = 0.02	P = 0.01
Know someone killed				.40 (0.34)
				P = 0.25
Know someone wounded				.18 (0.35)
				P = 0.60
Female	−0.10 (0.12)	.05 (0.25)	.04 (0.25)	.01 (0.25)
	P = 0.43	P = 0.85	P = 0.89	P = 0.96
Married	.22 (0.12)	.45 (0.25)	.46 (0.25)	.45 (0.25)
	P = 0.08	P = 0.07	P = 0.06	P = 0.07
Employed (full or part-time)	.01 (0.12)	−0.16 (0.24)	−0.16 (0.24)	−0.23 (0.24)
	P = 0.91	P = 0.50	P = 0.51	P = 0.35
Democrat	−0.86 (0.13)	−1.05 (0.27)	−1.04 (0.27)	−1.01 (0.27)
	P = 0.00	P = 0.00	P = 0.00	P = 0.00
Nonwhite	−0.05 (0.14)	−0.08 (0.28)	−0.04 (0.28)	−0.07 (0.28)
	P = 0.71	P = 0.78	P = 0.89	P = 0.80
N	1201	324	324	324
Pseudo-R2	0.0426	0.0806	0.0924	0.1034
LR Chi-2	69.34	36.20	41.49	46.40
Prob > Chi2	0.0000	0.0003	0.0001	0.0000

The answer to the question of how sensitive US voters are to the wounded as an additional cost of war appears to be: not very. This finding is counterintuitive and will be surprising to some. However, it is consistent with previous research challenging the general casualty aversion hypothesis that suggests that factors such as partisanship, policy aims, and the prospects of military victory are the key drivers of public support for war (Gelpi, Feaver, and Reifler 2009; Kriner and Shen 2016). In particular, the effect of partisanship is remarkably strong. Self-identified Democrats were close to 30 percent less likely to support military action than Republicans or Independents. That partisanship exerts such a strong effect even absent the elite cues meant to tap into political polarization that have been used in prior scholarship is especially striking (Zaller 1992; Berinsky 2009; Guisinger and Saunders 2017).¹⁷ This finding supports the troubling claims made by Saunders (2015) and Kreps (2010) that public opinion may matter less than we think in informing policy; if the public is indeed indifferent toward different types of casualties, elites can proceed with military action without having to pay too much attention to public attitudes (Kreps 2010; Saunders 2015). The relationship between party identification and support for deployment, however, is not necessarily stable. In follow-up experiments (described below) conducted under the Trump Administration, the correlation between party identification and support for deployment was not statistically significant.

The most intuitive, yet unexplained, result appears to be that respondents with close military ties could better assess increasingly severe casualty combinations than respondents without close military ties. A less intuitive finding pertaining to those respondents with close military ties is that they were generally more supportive of military deployment than respondents without close military ties. This finding may well interact with the effect of values as well as partisanship. As recent research has shown, the military has become increasingly partisan even as it has become increasingly trusted (Urben, 2013, 2017; Risa Brooks 2020a, 2020b).

These results could be sensitive to a number of factors, including the nature of the scenario used in the vignette. To test for the robustness of the results, I conducted follow-up experiments that included different vignettes: one on a possible US intervention in the Yemeni civil war, and another on a hypothetical Middle Eastern country whose ongoing civil war could allow it to become a terrorist haven. In both cases, respondents continued to be generally insensitive to different combinations of casualties (see the online appendix for a fuller discussion of these results).

Direct versus Indirect Costs of War

Based on the results from the main YouGov survey using the Islamic State vignette, it does not appear that respondents with close military ties are better able to assess the human costs of war because they are better informed. Perhaps instead, they are better able to assess the human costs of war because they know they will bear these costs directly. If this is case, this claim should be testable even for those respondents without close ties to the military.

To explore the possible relationship between the direct imposition of costs and the ability to rank casualty scenarios, I fielded a third follow-up survey on 1,470 respondents from March 16, 2018 to April 4, 2018 via Qualtrics. The Qualtrics panel was nationally representative on gender, age, income, race/ethnicity, and education. The first half of the survey replicates the four key conditions (control; 1,000 KIA; 1,000 KIA + 10,000 “less severe” wounds; 1,000 KIA + 10,000 “more severe” wounds) used in the 2016 YouGov survey, based on the same Syria/IS scenario. The second half of the survey added the following sentences to each of these conditions: “The United States would finance the conflict in part through the existing budget and in part through debt. Additionally, the government would impose a 1 percent national sales tax paid by all Americans to help fund the military action.”¹⁸ This language is borrowed from Flores-Macías and Kreps (2015); I also borrow their logic in hypothesizing that, in comparing across treatments with and without the additional information about taxation, support for military deployment would be consistently lower when respondents were told that their taxes would increase.

The main comparison of note is between the two halves of the Qualtrics survey, only one of which includes information about the financial repercussions of military deployment on respondents. Once again, the results of a general population survey demonstrate the inability to order casualty outcomes according to severity. In both halves of the survey, respondents were generally more supportive of deployment the more information they were given about the most severe combinations of casualties. More germane to this particular experiment, we also observe depressed support for military deployment across every condition when respondents were informed that there would be a tax imposed to finance a conflict in Syria (see Figure 5).

Figure 5.

The human versus financial costs of war.

This finding verifies Flores-Macías and Kreps’ earlier results, and is consistent with that of Horowitz and Levendusky (2011), whose survey experiment results demonstrated decreased support when respondents were prompted with the possibility of reinstating the draft in the United States.

Conclusion

The decision to use military force is often conceived of as a cost–benefit calculation. To make this calculation, decision makers—including voters—must have a sense of costs and benefits. The costs of war have gone up with the improvements in military medicine that have produced today's dramatic escalation of the US wounded-to-killed ratio. However, knowledge of this increase appears not to have filtered down to voters in a meaningful way.

It also appears that the cost–benefit calculation differs for those who directly bear the costs of war. Those with close ties to the military surely value their own lives and the lives of their loved ones, but may also include a set of factors, flavored heavily by values of patriotism and noble sacrifice, into their own calculations of the net benefit of war. The shift to the all-volunteer force in 1973 has led to a smaller military community whose members’ attitudes may be diverging from those of the general population when it comes to key foreign policy questions. Indeed, as Kriner and Shen (2016) show, this population also may have divergent domestic preferences from the population without close military ties. This research suggests training additional attention on military personnel and their families, such as with future surveys on military bases or towns nearby military bases. It is worth knowing whether and how those with close military ties hold different preferences about military deployment than the general population, particularly in an era when an all-volunteer force has been deployed with great frequency.

Another avenue for future research would be to investigate the effects of cumulative, and/or retrospective, fatal and nonfatal casualties on public support for military action. A prior experiment by Gelpi, Feaver, and Reifler (2009, 144–45) that combines some information about the wounded with a series of questions regarding the accumulation of casualties, however, generates very similar results to mine. An experimental design based on a hometown newspaper story also could shed light on whether the general public is more responsive to cost–benefit, partisan, or values-based cues.

A key benefit of training attention on the military wounded is to expose the “hidden” costs of war. However, such hidden costs extend beyond wounded military personnel. The increased use of military contractors calls attention away from those in uniform; the public appears to be unaware, however, of casualties among military contractors except in spectacular cases such as the four contractors killed in Fallujah in 2004 (Avant 2005; Avant and de Nevers 2011). Similarly, casualties borne by allies—especially, in the United States’ most recent wars, Iraqi and Afghan security forces—receive little attention on the home front. Collateral damage in the form of civilian deaths for which US forces are responsible generate more of a public opinion outcry, but likely not enough to move the needle on support for deployment (Kreps and Wallace 2016; Sanders 2018). A more comprehensive examination of public reaction to a larger set of the hidden—and human—costs of war is likely in order.

While this paper has focused on the US case, the shift from conscripted to volunteer forces and improvements in military medicine have occurred on a global scale. For example, a number of European countries have eliminated conscription (Haltiner 1998; Jehn and Selden 2002; Desilver 2019). And most states that have been especially prone to fight wars in recent years—such as India and Israel—have seen similar improvements in their wounded-to-killed ratios (Fazal 2014, 102). Insofar as they also indicate a decline in casualty aversion, these shifts may generate important strategic implications. If casualty aversion is on the decline worldwide, the efficacy of tripwire strategies, such as sending troop deployments to tie hands and sink costs, must be reevaluated (Rovner and Talmadge 2014, 554).

The analysis here, however, also suggests a potential solution to this problem. While general population respondents were not generally sensitive to various combinations of human casualties, they responded quite strongly—and negatively—when told that they would bear the financial costs of war directly. This result underlines the somewhat unseemly importance of translating the human costs of war into dollars for popular consumption. What remains unknown is the extent to which key policymakers have absorbed the fact that the costs of war have changed due to increasing numbers of the war wounded. Until and unless these new costs are factored into equations for war at both the popular and elite levels, US decisions to use force (or not) will be impaired by an easily rectified handicap.

Tanisha M. Fazal is a Professor of Political Science at the University of Minnesota.

Notes

Author's note: For valuable comments, conversations, and suggestions, I thank Burcu Bayram, Mike Desch, Christopher Gelpi, Ron Krebs, David Nickerson, Paul Poast, Elizabeth Saunders, Caitlin Talmadge, and Geoff Wallace. I am grateful to the Kroc Institute for International Peace Studies and the Institute for the Study of the Liberal Arts, both at the University of Notre Dame, as well as the Harry Frank Guggenheim Foundation, the Notre Dame International Security Center, the University of Minnesota College of Liberal Arts, and the Talle Faculty Research Grant for financial support. Earlier versions of this paper were presented at the University of Minnesota's American Politics Colloquium and IR Faculty Workshop; the Lansing Bee Jr./Bankard Seminar in Global Politics at the University of Virginia; the International Security and Conflict Studies Speaker Series at the George Washington University; the Monday International Relations and Theory seminar at UC-Berkeley; West Point's Modern War Institute; Yale University's International Relations Colloquium; The University of Chicago's Program on International Security Policy; and, at the 2015 Annual Meeting of the American Political Science Association. Allison Hostetler and Jonathan Leslie provided valuable research assistance. The data underlying this article are available on the ISQ Dataverse at https://dataverse.harvard.edu/dataverse/isq.

Footnotes

1

See “America's Wars” for wars prior to Operation Enduring Freedom (OEF) and Operation Iraqi Freedom (OIF) and the Defense Casualty Analysis System for data on Operations Enduring Freedom and Iraqi Freedom. Accessed October 28, 2019. https://www.va.gov/opa/publications/factsheets/fs_americas_wars.pdf. https://dcas.dmdc.osd.mil/dcas/pages/casualties.xhtml.

2

http://www.defense.gov/news/casualty.pdf.

3

The effect on amputation is subtle. Earlier wars saw more soldiers sustaining killing injuries to the torso that today might produce amputation. However, military physicians in earlier wars also appear to have been more likely to use amputation in part because knowledge of hemostatics was limited. Today, military physicians are more likely to treat patients at risk for amputation, but less likely to amputate. (Fischer 2015; U.S. Military Builds on Rich History of Amputee Care2014).

4

Survivors of military personnel killed in action receive a fraction of what veterans receive in pension benefits.

5

See, for example, (Gelpi, Feaver, and Reifler 2009, 1, footnote 1). Kriner and Shen (2014, 1175) similarly define the human costs of war in terms of “combat deaths”. Horowitz and Levendusky (2011, 530) equate high casualties with the number of US troops that might die in a conflict.

6

The Oxford English Dictionary dates the first usage of the term casualty to the mid-sixteenth century, and defines it in terms of battlefield losses that include the wounded and deserted, as well as those killed.

7

Indeed, theirs is the first such attempt to assess casualty sensitivity to the wounded of which I am aware, but it constitutes a very minor discussion in their book. (Gelpi, Feaver, and Reifler 2009, 144).

8

Most of the literature does not mention the wounded at all. Those that do include the wounded in the motivation and theoretical discussion, but not in the empirics. See, for example, Boettcher and Cobb (2006, 833, 837).

9

Pollingreport.com includes ten topics under the heading of national security: Afghanistan, China, Iran, Iraq, ISIS, Korea (North and South), Libya, Russia, and terrorism. There were no questions reported for 2015 for Afghanistan, China, and Korea (North and South). Only one question on Iran and terrorism, respectively, referred to military action, but without reference to ground troops or any other specifics. No questions on Libya or Russia referred to the possibility of US military deployment. See www.pollingreport.com. Accessed August 21, 2015.

10

Pollingreport.com was the source for these questions.

11

Walter Reed National Military Medical Center is the United States’ largest military medical center. US military casualties are typically sent first to Walter Reed when they return from overseas for care.

12

For example, Gartner (2008) argues for a strong effect of casualty accrual, but Gelpi et al. (2009) dispute this claim.

13

Note, however, that these two categories overlap. In other words, respondents who serve(d) in the military may also have—and indeed, are especially likely to have—family members that serve(d) in the military. Therefore, we cannot sum these two percentages to generate the percent of respondents with close ties to the military. Accounting for this distinction, 25.35 percent of respondents in this sample had close ties to the military—they themselves serve(d) and/or they have a close family member who serves or served.

14

Annual data for US fatalities were taken from iCasualties: icasualties.org.

15

Information on minor amputations is not currently available, but interviews of military physicians lead me to include amputations (minor and major) as well as facial disfigurement as the two most visible battle injuries sustained and survived today by US military personnel. An even more common battle injury is TBI, which is not included in this experiment because it is not visible. I also do not include post-traumatic stress disorder (PTSD) in the survey experiment, even though over 100,000 new cases of PTSD were diagnosed among deployed US military personnel from 2001 to the present. I exclude PTSD because, unlike the other type of injuries, it has likely always been present in the population of military personnel, but has only recently been diagnosed (and accepted) more effectively.

16

For these and all subsequent analyses, I collapsed support to a binary measure, where support = 1 if respondents weakly or strongly supported deployment.

17

Note that the finding on partisanship is potentially compatible with some analyses of mortality salience from social psychology that suggest a relationship between mortality salience and political ideology (Jost et al. 2017). Also note, however, that Guisinger and Saunders (2017) found the Syrian Civil War—the focus of the main vignette in this article—to be a relatively low-polarization issue (Guisinger and Saunders 2017, 432).

18

This language is taken directly from Gustavo Flores-Macías and Sarah Kreps’ analysis of support for military action based on different types of war financing (Flores-Macías and Kreps 2015).