Abstract
Public opinion studies on war attitudes say little about civilians who are related to military service members. The authors argue that military “service-connected” individuals are missing voices in the research that examines public support for war. Using over 50,000 observations from the 2010 Cooperative Congressional Election Study, the authors estimate attitudes toward the war in Iraq, the war in Afghanistan, and the use of US military troops in general. The authors find that service-connected civilians express greater support for war and the use of troops than civilians without such a connection. This study discusses the implications of these findings for theoretical advancements in the literature addressing war attitudes and the conceptualization of the “civil–military gap.”
Since World War II, polling on popular support for potential or ongoing military conflict has been a consistent approach to gauging the public’s attitudes. Using these data, scholars have posited several theories for explaining war support, or its absence, within the public. With few exceptions, studies of war attitudes have steered clear of the role of group attachments. Our purpose is to renew attention to group-based differences by focusing on the heterogeneity of connections to the military in the civilian population. Among civilians, associational proximity to a service member ranges from no personal acquaintance to immediate family member, with acquaintances, work colleagues, and close friends somewhere between these two extremes. The key difference, particularly clear from contrasting the extreme categories, is that civilians who have no connection to a service member are likely to bring different considerations to mind when evaluating a conflict than those civilians with a connection. Grouping all civilians together overlooks these differences because it assumes common exposure to the military and attachment to veterans. Whether one has an immediate family member who is a veteran merits serious consideration because what is at stake when reacting to proposals to increase troop levels or characterize invading Iraq as a mistake is unlikely to be the same for all civilians.
Why focus on this relationship in the civilian population? A 2004 report on America’s military population notes that military dependents exceed the number of personnel in uniform, with the 1.4 million active duty service members supporting 1.9 million immediate family members. 1 If we add to this figure those who live with the 23.4 million US armed forces veterans, then approximately one in five Americans have a direct or family-mediated connection with the military, the majority of whom have a family-mediated connection. 2 We term these individuals “military service-connected civilians,” owing to the distinct familial, social, and often financial pressures facing current and former military dependents. It is this distinction that motivates our group-oriented approach to evaluating contemporary civilian war attitudes.
The principle preoccupation among the earliest group-oriented studies of war attitudes was identifying which groups supported or opposed the war in Vietnam. For example, one research team evaluated Vietnam attitudes among Professors 3 , while another surveyed Protestant clergy. 4 For a variety of reasons, including their limited size in the population, these particular groups have not received sustained attention in the literature. By contrast, given the size of the service-connected civilian population, it is hard to think of another demographic group that is so pervasive and yet so consistently overlooked by research on public support for war.
We seek an answer to the question, do the immediate family members of veterans hold views toward conflict and war that are distinct from other civilians? This question is important for reasons other than the demographic share of veterans’ relatives in the general population. Our answer unpacks some of the complexity in civilian attitudes toward war and provides context for evaluating the civil–military gap in political attitudes among the masses. Do characterizations of the civil–military gap depend on which civilians one compares to veterans? The authors use survey data of a large representative sample of Americans in order to evaluate attitudes toward the war in Iraq, the war in Afghanistan, and the preference to increase US troops in Afghanistan. The authors find that individuals with immediate family in the armed services express greater support for war and increasing troops in Afghanistan than their civilian counterparts who are not related to service members. Accounting for whether a person is related to a veteran also changes our statistical estimate of support for war expressed by veterans. This study discusses the implications of these findings for theoretical advancements in the literature addressing war attitudes.
Service-Connected and Unconnected Civilians
Who is a service-connected civilian? The relationship between a civilian and the military can be visualized as a continuum from a strong relationship, such as a spouse, parent, or child currently serving on active duty to a weak one: civilians whose understanding of the military comes largely from the media. Figure 1 illustrates this continuum and suggests that civilians with the strongest link understand the military through family-mediated experiences. The military–civilian relationships are also moderated by the recentness of the connection to the military: those whose civilian experience or loved one’s military service has ended may view military issues differently than those who are currently connected. This distinction is likely to appear when discussing current conflicts, with the immediate family of active service members holding attitudes which diverge from those civilians whose family have completed their service.
Typology of civilian categories. Authors’ conceptualization.
The diversity of these connections raises a question of equivalence for these relationships: how can we compare the connection a civilian who served as a temporary employee on a military base holds to those held by civilians whose families have made a career out of military service? A related concern is that one of the largest groups of service-connected civilians, military spouses, is overwhelmingly comprised of women, while this is less true for the other groups. The variation for civilians in the continuum can be elaborated for veterans, distinguishing between active duty, reserves, and career veterans. While we agree that military–civilian relationships are diverse and gender certainly adds to the complexity of our continuum, analysis of a fully developed categorization of military–civilian relationships is beyond our purpose. The authors wish to demonstrate that two of these groups, civilians who are related to a veteran and civilians without veteran family members, are especially distinct from one another, and this has implications for their attitudes toward the current wars in Iraq and Afghanistan.
Being service connected entails having experienced demands, stressors, and pressures from social roles that distinguish the members of military families from other civilians. For example, while acknowledging some similarities in the tension between occupation and family demands experienced by military and civilian families, Segal argues that “[t]he military is unusual in the pattern of demands it makes on service members and their families [emphasis in the original].” 5 The unique constellation of demands experienced by military families include risk of injury or death to the service member, geographic mobility, periodic separation of the service member from the family, and living in foreign countries. Building on this work, Burrell et al. characterize these demands as stressors important to well-being outcomes in military families. 6 In a study focusing on the spouses of soldiers deployed to Mount Sinai in support of the Camp David Peace Accords, Wood and colleagues identified stressors beyond concern for the physical health of their loved one, including inability to access military services, marital strain, eventual reunion with their loved one, and assuming the role of a single parent and additional household duties. 7 Segal also notes the importance of social roles by elaborating on the normative pressures on spouses and children regarding their role in the military community, particularly behavior conformity corresponding to the rank of the service member. 8
The demands, stressors, and normative pressures experienced by military families create the conditions for the development of a social identity. 9 Following the distinctions offered by Brewer, the meaning of service connected as a social identity can be person based, relational social, group based, or collective. 10 The person-based meaning refers to how central membership in a military family is to a service-connected individual’s sense of self. The relational social identity captures the relationship role (and corresponding normative pressures that come with that role) that a service-connected individual has to a veteran. A group-based service-connected identity is reflected in the strategies to cope with the demands of military life, such as reliance on supportive social networks consisting of other military families and viewing the losses of the group as a personal loss. 11 Finally, a collective service-connected identity is exemplified concretely by recent pro-veteran and veteran family measures enacted by state governments, 12 nongovernment civic groups devoted to veterans and their families, and in the abstract by a sense of pride that comes from playing a key role in service to country. A service-connected identity has elements from each of these disparate identity meanings. Of course, the military family experience is likely to vary by relationship to the service member (e.g., spouse, daughter or sibling), whether the service member is actively serving or not, and whether it is time of peace or war. Despite these nuances, the differences between service-connected and unconnected civilians on war attitudes are likely greater than those between civilians who share a service-connected social identity.
Public Opinion Studies of War Support
The fundamental theoretical underpinning of civil–military gap studies is psychological attachment to a group. 13 The motivation may be to understand governance, 14 democratization, authoritarian personalities, 15 or constituent representation in foreign policy 16 ; however, the premise in civil–military studies is that military service members differ as a group from their civilian counterparts. In studies of veteran attitudes, this is generally captured with a survey question that identifies respondents who are veterans. 17 Just as previous scholars noted heterogeneity in the military by distinguishing, for example, officers from enlistees, 18 the authors call attention to the heterogeneity among civilians in order to better characterize civilian war attitudes. The authors rely on a survey measure which distinguishes service-connected from unconnected civilians instead of the conventional veteran/civilian dichotomy to evaluate the war attitudes of these two groups and show how they compare to veterans.
The authors of this study are not the first to examine civilian heterogeneity and war support. Mueller pioneered this work at the individual level by dividing respondents into categories which reflected the military service of their male relatives and served as a proxy for “self-interest’’ mentality. 19 Mueller reported that civilians with relatives serving in the armed forces expressed only small differences in war support, and these only infrequently. 20
The self-interest and civilian heterogeneity approach received its most thorough attention in Lau, Brown, and Sears’ evaluation of the effect of civilian ties to the military on support for the Vietnam war. Using 1968 CPS data, Lau et al. categorize civilians as having no connection to the military; having a relative who served, but not in Vietnam; or having a relative who served in Vietnam. 21 In line with Mueller’s self-interest argument, Lau et al. assumed that civilians related to veterans want to maximize the likelihood of the safe return of loved ones, and reasoned that this would lead to antiwar attitudes. They found that having a family member serving in Vietnam increased support for the war, contrary to their expectations. 22 One explanation for this finding is that other considerations consistent with self-interest, such as dissonance reduction and group conformity, help explain why the service connected differ from other civilians.
The authors contend that the concern for the safety of loved ones may be a primary concern, but there are other considerations consistent with self-interest motivations that have implications for war attitudes. For example, as a coping and dissonance reduction strategy a service-connected individual might be motivated to avoid statements or lines of reasoning that conclude that risk of injury or death by a loved one is in vain. Calling the US war in Iraq a mistake may cheapen the sense of achievement felt by a service-connected individual that comes from playing an integral role in supporting their loved one’s commitment. The service connected may internalize the expectation that they should be the biggest cheerleaders of veterans. Behavior or responses that are not congruent with these internalized norms are likely to be muted in response to the social pressures to conform that Segal identified as characteristic of military families. In linking self-interest motivations to war attitudes, there is good reason to expect that the service connected will be more supportive of war than other civilians.
Given the lack of clear support for separating service-connected civilians from others in public opinion analysis immediately following the Vietnam War, it is unsurprising that more recent work has avoided this approach. This neglect is unwarranted, however, since the 1970s, several societal trends have increased the distance between civilians with relatives serving in the military from those civilians who do not, in particular the end of the draft and the “Republicanization” of the South. The authors contend that these trends contribute to a civilian–civilian gap based on family connections to the military in much the same way that they have been argued to foster the civil–military gap. 23
Our central hypothesis is (Hypothesis 1) Service-connected individuals express greater support for war than civilians who are not service connected. On the questions of whether the United States made a mistake to invade Iraq or Afghanistan, we expect that service-connected civilians are most likely to respond “no,” when compared to unconnected civilians. Consistent with these expectations, we also expect service-connected individuals to prefer a strategy of increasing troops to Afghanistan, rather than reducing or even keeping the deployment size the same. Unlike earlier studies of civilian attitudes toward war, the authors of this study test this hypothesis controlling a variety of demographic confounders, and in a way that compares civilian diversity to veterans.
Theoretically, divergence between veterans and the service connected raises different concerns than a gap between veterans and civilians who are not service connected. If service-connected civilians do not differ greatly from veterans in their support for war, then support for military action can be characterized as less fractured than a situation in which service-connected civilians differ greatly from veterans (and therefore, are more similar to civilians who are not service connected). Empirically, our argument about civilian diversity implies that the difference between veterans and nonveterans sharpens depending on whether the comparison is to all nonveterans in general or civilians who are not related to a veteran in particular. The existing literature reports that veterans are more likely than nonveterans to express a favorable evaluation of war. 24 However, depending on which civilians that one compares to veterans will influence evaluations of the civil–military gap. Based on this expectation, the authors offer a second hypothesis: (Hypothesis 2): The estimated difference between veterans and civilians increases once accounting for service-connected individuals. Evaluating the first hypothesis speaks to a civilian–civilian gap, while evaluating the second hypothesis positions us to comment on alternative conceptualizations of the civil–military gap. As we show below, in addition to building on the theoretical motivation for distinguishing the service connected from other civilians, our analysis offers systematic controls of confounding factors, and places civilian differences in context with veteran attitudes.
Our Data and Approach
The authors make use of data from the 2010 Cooperative Congressional Election Study (CCES), a collaboration of 47 research teams yielding a representative sample of 55,400 US adults. 25 Each research team fielded separately at least one national sample survey of 1,000 cases, with half of the survey content controlled by the research team and the other half devoted to Common Content. For this article, the authors draw exclusively from items appearing in the Common Content, which consists of questions appearing on all team modules. All surveys were completed on the Internet under the auspices of You Gov/Polimetrix, and cases were selected using matched random samples. Sample matching is ideally suited for Internet access panels, and the opt-in Internet approach generates data similar to telephone and mail modes of surveys. 26 Additional details on the sampling methodology used for the 2010 CCES is available in Appendix A.
The 2010 CCES is particularly useful for scholars interested in military service, civilian heterogeneity, and support for war. Respondents were asked whether it was a mistake to invade Iraq and Afghanistan, and whether they would prefer to see the number of US troops in Afghanistan increased, kept the same, or decreased. Critical to our purpose is that the survey allows us to distinguish civilians who have an immediate family member in the military from civilians who do not. Another important advantage is that the data include a sizeable share of veterans (20 percent) and males who are related to veterans. Sixty-seven percent of the service connected in the sample are female, compared to 58 percent female among other civilians. This feature sets our analysis apart from the 2004 Kaiser Foundation’s Military Family Study, which includes only the spouses of active duty members of the US Army. 27 Including veterans is an improvement over the early study by Lau and colleagues.
The authors present three sets of logistic regression models of war support. In the case of the third model, increasing troops in Afghanistan, we collapse the response options “decrease” and “keep the same.” We want to capture the difference between those who support war and those who do not, and distinguishing the preference to “increase troops” from other responses is more appropriate for our purposes. As for the responses to separate evaluations of whether invading Iraq and Afghanistan was a mistake, in both cases, a value of 1 indicates the response “no” it was not a mistake, and 0 otherwise. For all three outcomes, a 1 indicates greater support for war. By including multiple indicators of war support, we ensure that any patterns we observe once accounting for the missing voices of veteran family members are robust and not sensitive to a single question. 28
For each of the three regression models, we begin with a baseline set of predictors and add indicators of veteran status and having an immediate family member in the military. The baseline model includes traditional predictors of war support, including interest in politics, partisanship, gender, race/ethnicity, age, education, marital status, and income. Interest in politics is an ordinal measure ranging from 0 to 3, with higher values corresponding to greater interest. Party identification is coded as separate indicators for republicans and independents, leaving democrats as the reference category. This means that the coefficients republican and independent should be interpreted as comparisons to democrats. We account for gender and include additional controls to avoid confounded conclusions about the effect of being related to a veteran. Coding and the distribution for all variables are included in Appendix B.
Building on the baseline model, we introduce our key independent variables of interest, which come from the question, “We’d like to know whether you or someone in your immediate family is currently serving or has ever served in the US military. Immediate family is defined as your parents, siblings, spouse, and children.” Overall, 20 percent of respondents are veterans, either currently or previously serving. A full 43 percent of respondents are service connected through an immediate family member who has served or is currently serving. The remaining 36 percent of respondents are civilians who are not related to someone in the military. From this distribution, the authors craft three mutually exclusive indicators. The authors begin by including an indicator for veteran status to the baseline model, which designates all nonveterans as the comparison group. To the veteran status model, we compare a fully specified model that includes an indicator for being service connected through a family member, leaving civilians who are not related to a service member as the reference category. The authors hypothesize that relatives of service members hold attitudes toward war that differ from civilians who are not service connected through family. As a second hypothesis, the authors anticipate that accounting for these missing voices also impacts our estimates of veteran status on war support because the contrast to civilians without a service connection is sharper than the contrast to nonveterans in general.
Findings
For ease of exposition, the authors plot the estimated coefficients in each of the three sets of multivariate regression models of war support. The authors direct readers who prefer to read regression results in tabular form to Appendix C. Figure 2 shows whether the estimates are statistically discernable from zero, represented here by the dashed vertical line. For each measure of war support, the authors estimate three model specifications beginning with the baseline model, then adding veteran status, and culminating with a full specification that includes the indicator for service-connected individuals. Each of the estimated coefficients is represented by a point with thick horizontal lines representing the 50 percent confidence interval, and a thin horizontal line capturing the 95 percent confidence interval. Estimates with confidence intervals that intersect with the dashed vertical line do not achieve conventional standards of statistical significance.
Comparing support for war among veterans, individuals related to a veteran, and civilians who are not related to a veteran. Intercepts are estimated but not shown in model results. Source: Cooperative Congressional Election Study 2010.
The baseline model estimates are represented by the lightest gray points. Many of the baseline predictors of war support are statistically significant and in the direction that we anticipated. Relative to individuals who identify with the democratic party, republican and independent identifiers are more likely to support war. Individuals who report higher levels of interest in politics are also more likely to express support for war. By contrast, and consistent with earlier studies, females, African Americans, and to a lesser extent Latinos, are less supportive of war than their respective male and white counterparts. 29 The estimates for income variables also suggest that individuals who report greater than $30,000 in annual household income are more supportive of war than individuals from the poorest households. 30
As expected, service-connected individuals are distinct from those individuals who are not related to a veteran. The central focus of this study is the missing voices represented by the dark point at the top of each panel that corresponds to the variable Service Connected. Across all measures of support for war, individuals who report having an immediate family member in the services are more likely to support war. That is the case with respect to evaluating whether invading Iraq and Afghanistan was a mistake, and the preference for increasing US troops in Afghanistan. This pattern holds even after controlling a variety of potential confounders as represented in the baseline model. The authors count this as evidence in support of Hypothesis 1.
Hypothesis 2 calls for a comparison of the estimate for veteran status in models that include and exclude service-connected individuals from the reference category. We can make this comparison in Figure 1 by focusing on the two Veteran estimates for all three outcomes. The mid-gray point corresponds to the estimate that compares veterans to all nonveterans in general, which is the equivalent of grouping civilians with and without an immediate family member in the military. In agreement with previous studies that have accounted for veteran status, we find that military service is generally associated with greater levels of support for war. 31 However, accounting for missing voices shifts the Veteran estimate to the right in the plot, or in the direction of greater support relative to civilians who are not related to veterans. This shift is robust and is observed across three measures of war support. The authors attribute this shift to the commonality in views and attitudes held by veterans and their family members. Once those common views are separated from a general civilian category, the authors find a sharper distinction between veterans and civilians, indicating support for Hypothesis 2.
The evidence that the authors have presented thus far offers support for both hypotheses. The binary nature of our outcome variables further complicates the interpretation of the estimates because they are not directly tied to the scale of the dependent variable. An important limitation of hypothesis testing based on p value or statistical significance is that it does not adequately convey the substantive effects of being service connected on war support. Moreover, large sample sizes are biased toward lower p values, and the authors want to be sure that the conclusions of this study are justified by the weight of the evidence, given the large number of cases analyzed. The authors address this by comparing predicted probabilities of war support for indicators of interest.
The authors compare the predicted probabilities of war support from two model specifications. Figure 3 plots predicted probabilities and 95 percent confidence intervals for all three outcomes. The model that distinguishes veterans and respondents who are related to a veteran from civilians who are not related to a veteran is in bold relief. For ease of comparison predictions from the model that group all civilians together are in gray. Civilians who are not related to a veteran are the baseline category represented by the open square at the bottom of each plot. The middle, open-square point represents service-connected individuals, and the top solid-triangle point corresponds to veterans.
Comparing the predicted probability of support for war among veterans, individuals related to a veteran, and civilians who are not related to a veteran.
Beginning with the left panel and the points in bold relief, the model predicts a 35 percent probability that civilians who are not related to a veteran (dark, bottom open square point) will respond “No” it was not a mistake to invade Iraq. Individuals who are related to someone who has previously served (dark, middle open square point) are associated with a 42 percent probability of saying “No,” while a 45 percent probability of responding similarly is predicted for veterans (solid triangle point). It is important that the confidence bands that correspond to the “service-connected” predictions do not overlap with the baseline prediction and confidence bands for civilians who are not service connected, suggesting distinct, and in this case greater, substantive effects. The dashed vertical lines indicate that the predictions for veterans and service connected are not statistically distinct. By contrast, there is a fair amount of separation in our comparison of interest, civilians and service connected. This is strong evidence in support of Hypothesis 1, and the pattern holds for the other two dichotomous measures of war support in Figure 3.
Hypothesis 2 requires a comparison of the points in bold relief and the points in gray. When comparing veterans to the baseline civilian category, the model predicts a difference that is double the difference predicted by comparing the veterans to nonveterans in general (the points in gray). Grouping all civilians together masks the distinct predictions we find for the service connected and masks the primary difference underpinning the civil–military gap in the public, the difference between veterans and civilians who are not related to a veteran. This is an important finding because it suggests an answer to why some studies report differences between civilians and veterans in war support and others do not. Indeed, this may be an answer to some of the mixed findings represented in the extant literature on the civil–military gap with respect to political attitudes in general.
Evaluative Discussion
Our evidence shows that distinguishing service-connected civilians from other civilians in a statistical model highlights a civilian–civilian gap in war attitudes: those civilians with a veteran immediate family member expressed more support for the wars in Iraq and Afghanistan and were more likely to favor an increase in troop levels in Afghanistan. Moreover, our models generate different estimates of the partial effect of veteran status, even controlling partisanship and gender. These findings engage Rohall, Ender, and Matthews’ (2006) article in this journal. 32 They find a civil–military fusion, rather than gap, emerging in the post 9/11 era, which they suggest is a “fourth wave in civil–military affairs,” based on a solidarity response to terrorist attacks in the United States and other countries. What evidence they observe of a civil–military gap between cadets and undergraduates is attributed to partisan and gender differences.
There are at least two alternative interpretations to what Rohall et al. call the fourth wave. First, what they detect as fusion could be the front end of decaying solidarity as a war unfolds. Second, the difference between our finding and Rohall et al. is primarily the attitudinal shift of a meaningful segment of the population whose voices have heretofore been missing in public opinion studies of war support. Drawing on survey data that taps attitudes well into the war, our findings suggest that variation in the civil–military gap is a function of civilian heterogeneity, even controlling partisan and gender differences. Does this mean that there is no fourth wave as Rohall et al. suggest? No, with our data we can only infer patterns about the civil–military gap to the 2010 US population. However, distinguishing service-connected individuals from other civilians can be a strategy for actually testing whether a civil–military fusion, and hence a fourth wave, has emerged.
Although service-connected civilians have received significant attention from military sociologists, especially with regard to the relationship between work and family life, the group has not been compared to other groups by scholars of public opinion and war support since the 1970s. 33 Despite being a large share of the civilian population, no study of Iraq or Afghanistan war support has evaluated the impact of being related to a veteran. Ours is the first large-N study to examine how the immediate relatives of veterans view these conflicts, how those views compare to civilians who are not related to veterans, and how both of these groups compare to veterans. The authors posed the question, do the immediate family members of veterans hold views toward conflict and war that are distinct from other civilians? Based on the evidence examined, we answer, “yes.” The previously missing voices of veterans’ relatives are distinct from other civilians, and accounting for these voices generates different estimates of veteran association on war support.
To be sure that the findings of this study were substantively meaningful, the authors also compared predicted probabilities of war support for the indicators of interest. The authors found that service-connected individuals are more supportive of war than their civilian counterparts who are not related to a veteran. What is important here is that the predictions are based on estimates from a statistical model that controls interest in politics, partisanship, gender, race, age, education, marital status, and income. In other words, the authors have a measure of confidence that the distinctions we identify for the service connected are independent of other factors that we know to be correlated with support for war and attitudes toward conflict in general.
The authors argue that our findings have important implications for theories of war support that characterize individuals as reasoning through the costs and benefits of conflict when forming their views toward war. We are not suggesting that a military counterculture exists, or that we are moving toward a military caste system in the United States. On the other hand, our findings do invite a critical perspective to how the civil–military gap is conceptualized. Just as studies have shown that treating veterans as a monolithic group masks important differences by branch, cohort, and rank, our study shows that important nuances are revealed in attitudinal support for war when civilian heterogeneity is incorporated into the analysis.
Future Directions for Research
Our purpose in this analysis has been to demonstrate the evidence of a civilian–civilian gap in attitudes toward the wars in Iraq and Afghanistan predicated on the presence of an immediate relative who is a veteran. The above analysis supports our hypotheses that service-connected civilians do express higher levels of support than unconnected civilians, and that including service-connected civilians improves our understanding of war support among veterans and civilians. The value of this work is not a definitive statement on a civilian–civilian gap, the civil–military gap, or the factors which influence war support within the population. Rather, we argue for a reimagining of the relationship between civilians and the military institution through the inclusion of service-connected individuals. In line with this goal, the authors conclude by offering five suggestions for future research.
Indicators of service connection may not be useful predictors in all models of war support. Building on the work by Jentleson, analysts would do well to identify which policy objectives reveal more or less distinction between service-connected civilians and other civilians. 34 The authors have scrutinized our hypotheses with three important attitudes toward ongoing conflicts in Iraq and Afghanistan. However, the CCES asks questions that are more abstract measures of war support. Namely, the approval of uses of the US military for different causes including to secure oil, destroy terrorist camps, intervene in genocide or civil war, spread democracy, help allies, and assist the United Nations (UN). Although analyzing these items is beyond the scope of this article, future research identifying the conditions under which service connections matter in these more abstract measures of war support would be a useful next step. Such a strategy fits with the caution issued by Berinsky and Druckman that how we conceptualize and measure war support has implications for the conclusion that we reach. 35
Second, cost–benefit calculations and evidence of a reasoning public might best be scrutinized by comparing different service categories of service connection. A rough start might be to replicate the experimental study by Gartner and test whether having an immediate family member who is a veteran moderates the treatment effect. 36 Our expectation, derived from this work, is that recent casualties and rising casualty trends will have a greater effect on individuals who are related to veterans even controlling interest in politics, partisanship, gender, and other standard demographic indicators. Information about the costs of war means something different, and perhaps greater, for someone who very possibly might experience or has experienced war-related loss. Simply put, there is more at stake for the service connected.
Third, extending this logic to a theory of elite-cue approaches relaxes the assumption that individuals turn to elites for direction on forming issue positions. Deference to experts is a rational response in the face of low information and high costs for learning the details of an issue. 37 However, this assumption is less tenable for the immediate family members of veterans whose indirect connection to the military probably exposes them to more war-related information than other civilians. As we noted from the outset and illustration in Figure 1, the category of civilian is varied in connection to the military, and the relationships that define service connected vary as much (sibling, spouse, parent). At least one study on coping with military life suggests that veteran spouses have strong incentives to keep informed about the military in order to manage daily life. 38 Moreover, the attitudes of service-connected civilians are more likely to be motivated in a direction away from characterizing an ongoing conflict as a mistake. To say otherwise risks conveying that a loved one’s efforts are in vain. Research by psychologists finds that motivated reasoning is especially influential when individuals feel that the stakes are high and that the outcome is personally relevant. 39 These conditions certainly seem to be satisfied prima facie for the immediate spouses of veterans in comparison to other civilians, and it may well work differently for siblings, children, or parents of veterans.
One of the standard practices in evaluating the effects of environmental cues on individual behavior is to compare individuals whose identities are made salient by cues, to those who do not hold such an identity. For example, an individual’s veteran identity might be made more salient depending on whether she resides in a community that is experiencing casualties from an ongoing war abroad, has a military base, is represented by a high-profile elected official like a US Senator who is a veteran, and whether many or few veterans live nearby. To show the effect of these myriad environmental cues on a behavioral outcome the analyst simply draws comparisons to other veterans without these cues (or fewer of them) and to nonveterans who are exposed to the same context. In other words, the authors argue that some individuals who are not veterans, like the service connected that we showcase here, may be sensitive to environmental cues that previous studies assume are only of consequence to veterans. 40 This study on this front conceives of group attachment more broadly than is typically seen in studies of contextual effects.
Building off the findings that the authors report here, a fifth line of research worth developing is one that focuses on the consequences of having multiple group attachments. For example, we find that being service connected is associated with greater support for war, while females and African Americans, and to a lesser extent Latinos, are less likely to support war. No work that we are aware of has examined whether these effects cancel out for nonwhite females, related to a veteran, or are amplified for nonwhite males related to a veteran. These kinds of studies are possible with large and national samples because the concern for small cell size after parsing data in a multivariate analysis is not an issue.
Footnotes
Appendix A
Appendix B
Appendix C
The authors would like to express their great appreciation for the assistance of John Allen Williams, Mike Koch, Steve Nicholson, and Philip Converse in completing this manuscript. We also thank Patricia Shields and the anonymous reviewers for their helpful feedback on earlier drafts of this manuscript. Authorship is equal; authors are listed in alphabetical order.
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors would like to thank the College of Letters and Sciences at the University of Wisconsin Oshkosh for its support of travel expenses and research assistance for this article.