Combat exposure, cigarette consumption, and substance use.

Cesur, Resul ; Chesney, Alexander ; Sabia, Joseph J. 等

Abusing alcohol and drugs has been part of military culture historically: troops do it for fun, to ease the stresses of war or to be part of the brotherhood.

(Pauline Jelinek, Huffington Post, September 2012)

I. INTRODUCTION

While much media attention has been paid to the number of soldiers killed and wounded in the Global War on Terrorism (GWOT), policymakers have increasingly turned their attention to the many hidden costs of war imposed on U.S. military personnel. In 2008, the RAND Corporation published an influential report showing that one-quarter of American soldiers returning from combat deployments in Iraq and Afghanistan suffer from "invisible wounds" of war caused by the stresses and psychological trauma of combat exposure (Tanielian and Jaycox 2008). The symptoms are often manifested in the form of post-traumatic stress disorder (PTSD), depression, traumatic brain injury, and suicide ideation (Cesur, Sabia, and Tekin 2013).

A number of recent surveys document substantial rates of risky health behaviors among soldiers serving in the GWOT. One in eight veterans of the Iraq and Afghanistan wars received an alcohol-related counseling referral (National Council on Alcoholism and Drug Dependence 2012), over one-quarter suffered from some combination of drug and alcohol dependency, homelessness, and depression (Tanielian and Jaycox 2008), and nearly 40% of soldiers and Marines reported smoking cigarettes, leading the Institutes of Medicine (2009) to recommended a transition toward a tobacco-free military.

Combat exposure could causally affect the likelihood of subsequent risky health behaviors for a number of reasons. First, individuals exposed to combat often experience heightened risk-taking and an adrenaline rush while in combat. Prolonged exposure may induce risk taking (Killgore et al. 2008) and thrill seeking (Vaughan 2006) as well as increase perceived fearlessness and pain thresholds (Joiner 2005). These emotions may increase risky health behaviors among combat veterans.

The stress of combat may also lead veterans to engage in risky health behaviors as a coping mechanism (Institutes of Medicine 2009). (1) There is compelling evidence that the combat exposure increases the risk for stress-related disorders and poorer mental health. Combat service has been linked to increased risk for PTSD (Bedard and Deschenes 2006; Cesur, Sabia, and Tekin 2013; Hoge, Auchterlonie, and Milliken 2006; Hoge et al. 2004), depression (Cesur, Sabia, and Tekin 2013), and suicide ideation (Gold et al. 2000; Grossman and Siddle 1999; Hearst, Newman, and Hulley 1986; Page, Engdahl, and Eberly 1991). Thus, engaging in risky health behaviors may be a means through which service members cope with or escape from stress-related ailments. These behaviors could also be a "cry for help" or even reflect an attempt at further self-injury (Jacobson et al. 2008; Joiner 2005).

A final mechanism through which combat could affect risky behaviors is via income effects. In the short run, increased combat pay could increase substance use to the extent that these substances are normal goods. However, in the longer run, military service reduces men's subsequent civilian wages, which could have the opposite effect (Angrist 1990).

However, if individuals with a greater propensity of undertaking risky health behaviors are more likely to be enlisted, the observed associations between military service and detrimental health behaviors may simply reflect spurious correlations. To disentangle the health effects of war from selection effects, studies of prior U.S. wars (World War II, Korea, and Vietnam) used the draft lottery (Angrist 1990) and cohort-level variation in the probability of exposure to war (Bedard and Deschenes 2006; Rohlfs 2010). Using these identification strategies, studies have found that military service is directly linked to cigarette consumption (Bedard and Deschenes 2006), binge drinking (McFall, Mackay, and Donovan 1992), and drug use (Price et al. 2004), but essentially unrelated to the probability of AIDS-related intravenous drug use (Hearst, Newman, and Hulley 1986) and alcohol consumption (Dobkin and Shabani 2009).

Despite this body of work, much less research has been performed by health economists on the causal behavioral health effects of combat exposure in GWOT. One important practical reason for this is the abolition of the draft lottery. Most studies of the health consequences of military service in GWOT compare the risky behavior outcomes of deployed soldiers to non-deployed personnel, including Reservists and National Guardsmen (Hoge, Auchterlonie, and Milliken 2006). For instance, Thomsen et al. (2011) found that deployed marines and sailors are more likely to use illegal drugs than their nondeployed counterparts; Smith etal. (2011) and Hoerster et al. (2012) found some evidence that deployment is associated with higher rates of smoking2; and Jacobson et al. (2008) and Hooper et al. (2008) predicted that deployment is associated with an increased risk of alcohol abuse. However, Reservists and National Guardsmen differ from active-duty deployed soldiers on a variety of characteristics that are also related to risky behaviors (Hirsch and Mehay 2003), and active duty soldiers with periods of extended nondeployment may be nondeployable due to health conditions (Department of the Army, AR 614-30, 2010). Therefore, given that (1) selection into the Reserves/National Guard may differ from selection into active duty service and (2) nondeployments may occur due to health-related reasons, the conclusions from these studies should be cautiously interpreted.

Following the identification strategy employed by Lyle (2006) and Cesur, Sabia, and Tekin (2013), the present study uses variation in overseas deployment assignment among active duty deployed personnel to identify the effect of combat exposure on subsequent cigarette consumption, alcohol consumption, and illicit drug use. We find that combat exposure is associated with substantially increased risk of subsequent risky health behaviors, even after controlling for predeployment risky behaviors. Our results suggest that the mental health effects of war service can explain up to two-thirds of the estimated association between combat exposure and risky health behaviors.

The remainder of the article is organized as follows. The next section describes the data sets and introduces the variables used in the analysis. Section III lays out the econometric model and articulates our identification strategy. Section IV presents the results, and Section V concludes.

II. DATA AND VARIABLES

We start our empirical investigation using data extracted from the National Longitudinal Study of Adolescent to Adult Health (NLSAAH), and supplement our analysis with a representative sample of active duty military personnel, the 2008 Department of Defense (DOD) Health and Related Behaviors (HRB) Survey.

Collected by the University of North Carolina at Chapel Hill, the NLSAAH is a nationally representative school-based survey of junior high and high school students who were first interviewed during the 1994-1995 school year. Three subsequent follow-up surveys were conducted, including the third follow-up (Wave IV), when respondents were in their late twenties and early thirties. Surveys were administered privately via the Computer Assisted Self-Interviewing (CASI) system to minimize under-reporting of sensitive health behaviors.

The estimation sample includes 565 individuals aged 25-32 years at the third follow-up survey (1) who had served as an active duty service member and been deployed overseas at least once by the time of Wave IV survey; (2) whose military service started after the Wave I interview; and (3) who provided nonmissing information on combat exposure and the outcomes under study. (3) As discussed by Cesur, Sabia, and Tekin (2013, 2015) and Cesur and Sabia (forthcoming), a key advantage of the NLSAAH is that it contains information on observables available to Human Resources Command when making deployment assignments: service members' military branch, rank, occupation, and length of military service.

Among the 565 individuals in our sample, 224 were current active duty service members at the time of the Wave IV survey and 341 were veterans whose service had ended prior to the Wave IV survey. (4) When we examine deployment assignments among service members in our sample, we find that 185 were assigned to combat zones with enemy firefight engagement, 231 were assigned to combat zones without enemy firefight, and 149 were deployed overseas to noncombat zones.

Our key independent variable, Combat Exposure, is a dichotomous variable set equal to 1 if the respondent reported deployment assignment to combat zone where he or she "engage[d] the enemy in firefight"; it is set equal to 0 if the respondent reported overseas deployment assignment to a noncombat zone or to a combat zone without enemy firelight. (5)

We construct three health outcomes in the NLSAAH using data from the third follow-up survey. First, with regard to smoking, respondents are asked: During the past 30 days, on how many days did you smoke cigarettes?

Respondents who reported positive days of smoking in the past 30 days were coded as 1 and those that reported 0 days of smoking were coded as 0.

Second, we measure binge drinking using service members responses to the following questionnaire items: During the past 30 days, on how many days did you drink? Think of all the times you have had a drink during the past 30 days. How many drinks did you usually have each time? A "drink" is a glass of wine, a can or bottle of beer, a wine cooler, a shot glass of liquor or a mixed drink.

Respondents who answered that they drank on at least one day in the past 30 days and reported usually having five or more drinks if male or four or more drinks if female were coded as 1. Others were coded as 0. Because of the NLSAAH survey instrument, our binge drinking measure is not the standard binge drinking measure used in the public health literature (capturing any binge drinking in the last 30 days), but rather captures more frequent binge drinking.

Finally, drug use was measured using responses to the following questionnaire items: During the past 30 days, on how many days did you use marijuana? During the past 30 days, on how many days did you use your favorite drug [includes sedatives, tranquilizers, stimulants, pain killers, steroids, cocaine, crystal methamphetamine, ecstasy (MDMA), inhalants, LSD, heroin, PCP, or other illegal drugs]? (6)

Using responses to these measures, we generate indicators of any illicit drug use, marijuana use, and hard (nonmarijuana) drug use in the prior 30 days.

Next, we use data from the 2008 DOD HRB Survey, conducted by the RTI International to estimate the health and well-being of active duty service members. The HRB survey is comprised of 28,546 active duty military service members ages 18 to 50. Participants were selected to represent men and women in all pay grades all over the world, but excluded personnel who were absent without official leave (AWOL), attending a service academy, or were incarcerated at the time of data collection. The vast majority of surveys were answered by participants at military installations, while a small number were answered by mail for those who could not attend such sessions. The survey was a pencil-and-paper survey and while self-administered, the lack of a CASI system of data collection, may result in underreporting of behaviors. (7) However, measurement error should not bias our estimates as long as misreporting of health behaviors is not associated with combat deployment. Our main sample consists of 14,740 active duty respondents from all branches of the United States Armed Forces--Army: 3,253; Navy: 4,242; Marine Corps: 3,014; and Air Force: 4,070--who were deployed overseas and provided nonmissing information on health behaviors.

We measure combat exposure in the DOD survey in a similar way to the NLSAAH. If the respondent reported being exposed to enemy firefight during deployment, Combat Exposure is coded to 1. If the respondent reported overseas deployment without enemy firefight, Combat Exposure is coded to 0. Among the sample of overseas-deployed service members, 7,166 (48.6%) reported combat exposure. (8,9)

As in the NLSAAH, respondents to the DOD survey were asked about their participation in risky health behaviors. The questionnaire items on cigarette consumption and marijuana use in the last 30 days are identical to the NLSAAH and coded analogously. For other drug use and binge drinking, the questionnaire items in the DOD survey are different than in the NLSAAH. With regard to other drug use, respondents were asked about the use of cocaine, LSD, PCP, ecstasy (MDMA), other hallucinogens (peyote, mescaline, and psilocybin), methamphetamine, heroin, GHB/GBL, and inhalants in the last 30 days. (10) Binge drinking was defined in a more standard fashion: "consuming five or more drinks (four or more for women)" on "at least one occasion during the past 30 days."

The DOD data address the two shortcomings of the NLSAAH data: (1) it has a large, representative sample of active duty deployed service members, which will allow obtaining the estimates by branch of service more precisely; (2) and the age range of the participants (18-50) makes the estimates obtained from the DOD survey more generalizable to active duty service members. However, there are a few disadvantages. First, because the DOD survey is comprised only of current active duty service members, this could introduce sample selection bias if prior combat exposure could affect who remains in the sample. For example, if some individuals choose to leave the military and return to civilian life after their term of service has expired, and these individuals are those whose risky health behaviors are most severely affected by combat duty, then this sample selection could result in understating the adverse health effects of combat service. Second, while the DOD survey contains information on military branch, rank, timing of service, and installation-level Major Command, (11) a key disadvantage is the lack of information on occupation.

III. EMPIRICAL APPROACH

We begin with the NLSAAH data, restrict the sample to active duty deployed service members, and estimate the following linear probability model:

(1) [R.sub.i] = [[beta].sub.0] + [[beta].sub.1][C.sub.i] + [[beta]'.sub.2][M.sub.i] + [[beta]'.sub.3][X.sub.i] + [[epsilon].sub.i]

where [R.sub.i], is a dichotomous indicator for whether respondent i has engaged in a particular risky behavior (smoking, binge drinking, and drug use), [C.sub.i] is a dichotomous indicator for whether the respondent has been assigned to a combat zone with enemy firefight, [M.sub.i] is a vector of relevant military characteristics (branch, rank, occupation, timing of service, and length of service), and [X.sub.i] is a set of individual and family background characteristics listed in Table Al, including predeployment smoking, drinking, and illicit drug use. In order for our estimate of [[beta].sub.1] to be unbiased, it must be the case that deployment assignment of active duty deployed personnel is exogenous to risky health behaviors.

There is substantial theoretical evidence--and some descriptive empirical evidence--in support of this assumption's validity (Cesur and Sabia forthcoming; Cesur, Sabia, and Tekin 2013; Engel, Gallagher, and Lyle 2010; Lyle 2006; Negrusa and Negrusa 2014). Human Resources Command, in fact, deploys companies or units and generally not individual service members. (12) Senior commanders determine unit deployment assignment based on the exigencies of the operational environment, which are driven by world events, and "the availability and readiness of suitable units" (Engel, Gallagher, and Lyle 2010, 76) determined by equipment availability, timing of training completion, and the occupational skill set of unit members (Army Regulation 220-1; Cesur, Sabia, and Tekin 2013, 2015). Neither service members' underlying propensity to engage in risky health behaviors nor their individual or family traits play a role in the location of unit deployment assignments. (13) The U.S. Armed Forces considers service members of identical military occupation and rank as perfectly equivalent in the production of national security. (14)

In Table A3, we use data from the NLSAAH to descriptively explore the credibility of the exogeneity of deployment assignment, in a spirit similar to Cesur, Sabia, and Tekin (2013, 2015) and Cesur and Sabia (forthcoming). Consistent with those studies, we find that conditional on military rank, timing of service, branch, and occupation, deployment assignment is orthogonal to a wide set of background characteristics: age, race, height, weight, years of schooling attained, religious affiliation, maternal educational attainment, parental marital status when the respondent was an adolescent, parental income when the respondent was an adolescent, health insurance status, as well as the respondent's predeployment risky health behaviors (smoking, binge drinking, and drug use), measured analogously to the outcome variables, when the respondents were in high school. (15,16,17)

One notable exception to this pattern of results is gender, where we find that men are more likely than women to report assignment to combat. However, we do not interpret this as evidence of endogenous selection into combat, but rather an explicit artifact of military policy. Prior to January 2013, the U.S. military policy dramatically limited women's combat roles. The 1948 Women's Armed Services Integration Act specifically excluded women from nearly all combat operations. A 1993 Defense Department directive lifted the ban on most combat in aviation positions, though it maintained restrictions on some Air Force positions, including those that give direct support to ground troops. Between 1993 and 2013, women continued to be banned from most direct combat missions, as per Pentagon policy, with a 2012 Pentagon report stating that women were ineligible for service in approximately 20% (or 230,000) of positions in the Armed Forces. (18,19) In January 2013, the ban on combat service for women was fully lifted. However, because our data come from prior to 2013, it is not surprising that men are more likely to serve in combat roles than women. When we restrict our analysis sample to men and repeat the analysis presented in this paper, the pattern of results is similar. (20)

Next, we turn to the DOD, which, as noted above, lacks information on military occupation. Given the lack of potentially important information on this measure, we first ensure that those who are deployed to combat and see enemy firefight are statistically equivalent on observables to those deployed to noncombat zones or combat zones without enemy firefight. We estimate a nearest neighbor matching model, where we first use a probit model to estimate the probability of assignment to combat:

(2) [C.sub.i] = [[delta].sub.0] + [[delta].sub.1] [Z.sub.i] + [v.sub.i]

where [Z.sub.i] includes military rank, individual's Major Command (MAJCOM), frequency of deployments, age, race, marital status, gender, and educational attainment. Our nearest neighbor matching procedure imposes common support on observables and requires predicted probabilities of combat exposure within 0.00015. (21) These parameters were chosen to ensure that on the above-mentioned observables, there were no statistical differences across the combat exposed and noncombat exposed samples. After matching, mean differences in the risky behavior outcomes were calculated and standard errors generated via bootstrapping. (22)

IV. RESULTS

A. Descriptive Statistics

Table 1 shows descriptive statistics for the key variables by Combat Exposure. (23) The first three columns present means using the NLSAAH data and the final three columns for the DOD HRB survey. Rates of smoking and drug use were higher in the NLSAAH than the DOD HRB survey (0.391 vs. 0.263 for smoking; 0.118 vs. 0.013 for marijuana use; and 0.044 vs. 0.039 for other drug use), which is not surprising given that (1) younger service members are more likely to engage in these behaviors than older individuals, (2) the NLSAAH data include information on former service members who are no longer on active duty (and not subject to, e.g., random drug testing), and (3) there are differences in survey administration. (24) One notable difference is our measure of binge drinking, but this can be explained by differences in the measures, as the NLSAAH measures capture typical monthly binge drinking. Across each of these datasets, rates of smoking, binge drinking, and drug use are greater for those who were exposed to combat relative to those who were deployed but not exposed to combat. Table 2 also shows that combat deployment appears most frequent in the Army relative to other branches of service.

B. Main Results

The first row of Table 2 presents the estimates of [[beta].sub.1] from Equation (1) using the NLSAAH. (25) In panel A, we find that combat exposure is associated with a 10.4 percentage point increase in the probability of smoking (column 1), a (statistically insignificant) 2.6 percentage point increase in the probability of binge drinking (column 2), and a 6.6 increase in the percentage point probability of illicit drug use (column 3). When we separate drug results by marijuana or other drug use, we find that combat exposure is associated with a (statistically insignificant) 3.8 percentage point increase in the likelihood of marijuana use (column 4), and a (statistically insignificant) 3.4 percentage point increase in the probability of other drug use (column 5). Relative to the means shown in Table 1, these marginal effects are quite large. (26)

The final two rows of panel A separate the control group into those who were assigned to combat zones without enemy firefight and those who were deployed overseas to noncombat zones. While less precisely estimated, the findings continue to show that combat exposure is positively related to the outcomes under study. The magnitude of the estimated relationship appears to be somewhat larger when using the noncombat zone deployed individuals as a control group, but the differences in estimated effects across the two separate control groups are statistically indistinguishable.

The remaining panels of Table 2 present NLSAAH estimates by branch of service--Army (panel B), Marines (panel C), Navy (panel D), and Air Force (panel E). The results, which are quite imprecise given the small sample sizes, on the whole, continue to point to a positive relationship between combat exposure and the probability of engaging in risky behaviors. The estimates are positive in 12 of 16 cases, though largest in the Navy.

Given the limited power of the NLSAAH, particularly with regard to branch-specific estimates, we next turn to the DOD survey. Tables 3 and 4 present evidence on the success of our matching procedure. After matching, we find that those who were assigned to combat zones are statistically equivalent with regard to military rank, MAJCOM, number of deployments, educational attainment, gender, age, race, and marital status. Thus, on observables related to occupation, our procedure has ensured common support.

Table 5 presents PSM estimates using the DOD HRB data. Row (1) presents estimates from the pooled sample and the remaining rows show estimates for the Army (row 2), Marines (row 3), Navy (row 4), and Air Force (row 5). For the pooled sample, the findings suggest that combat exposure is associated with a 3.6 percentage point increase in the probability of smoking, a 4.1 percentage point increase in the probability of binge drinking, a 2.8 percentage point increase in the probability of any drug use, a 1.3 percentage point increase in the likelihood of marijuana use, and a 2.9 percentage point increase in the probability of using other drugs. (27) Risky behavioral effects appear to be largest for the Army and Navy, consistent with much of the recent literature that has found larger health effects for these branches (Cesur, Sabia, and Tekin 2013). (28) One reason for smaller effects of combat exposure for those in the Air Force may be less physical proximity to, and perhaps psychological consequences of, combat exposure. (29) Taken together, the findings across both datasets suggest that combat exposure increases the likelihood of smoking and drug use across branches; moreover, there is some evidence that combat exposure increased subsequent binge drinking among soldiers.

The primary mechanism through which combat exposure has been hypothesized to affect mental health (Cesur, Sabia, and Tekin 2013), relationship health (Negrusa and Negrusa 2014), and domestic violence (Cesur and Sabia forthcoming) is the psychological stress-related consequences of war. First, we replicate the work of Cesur, Sabia, and Tekin (2013) and Cesur and Sabia (forthcoming) in Table 6, and show that combat exposure is positively related to a variety of stress-related disorders, including PTSD, anxiety disorders, and suicide ideation. (30)

Next, in Table 6, we explore how psychological health may mediate the relationship between combat exposure and risky behaviors in the NLSAAH. Panel A shows the results without including controls for mental health. In panel B, we add controls for PTSD, suicidal ideation, and the Cohen Stress Scale to the right-hand side of Equation (1) and explore how the estimate of Pi changes. We find that psychological stress-related mediators reduce the magnitude of the estimated effect of combat exposure on smoking by about 19% and the effect of combat exposure on any drug use by about 34%, respectively, suggesting that combat-induced psychological stress may be one important mechanism through which combat affects risky health behaviors.

Table 7 repeats the same exercise using the DOD HRB data and a nearest neighbor matching strategy. In each panel, the first row presents the estimates without the mediating controls as matching variables, while the second row adds these controls. The results point to a similar pattern, though the magnitude of the estimated effect of combat exposure on risky health behaviors falls by approximately 50%, and in some cases nearly 70%, after controlling for PTSD, suicide ideation, and stress. One explanation for why the mediators may be more important in the DOD as compared to the NLSAAH data is that the PTSD and stress scales are more detailed in the former dataset (see footnotes 27 and 28). (31)

Next, using the NLSAAH data, we explore the role of our mental health mediators using an alternate measure of combat exposure, Combat Length. In Table 8, we replace Combat Exposure with binary combat deployment length indicators. (32) The coefficient estimates presented in Table 8 represent the impact of the associated combat zone deployment length category in comparison to being deployed overseas to a noncombat zone. The results show (1) the impact of combat zone service on outcome measures seems to be uniform (i.e., statistically indistinguishable) for different combat zone deployment lengths; and (2) the decrease in effect sizes, when controlling for psychological stressors, follows a similar pattern with the results presented in Table 6.

Finally, we use yet another alternate measure of combat, Combat Harm. This measures whether the respondent was wounded in combat or had killed (or believed he had killed) someone on the battlefield, and has been used in a number of studies examining the health-related impacts of combat (Cesur and Sabia forthcoming; Cesur, Sabia, and Tekin 2013). Table 9 shows findings using the NLSAAH and Table 10 using the DOD data. We also estimate the impact of Combat Harm among those who experienced combat firelight in the DOH HRB data. The results presented in Table 11 show that Combat Harm leads to an increased likelihood of undertaking risky health behaviors among service members who were assigned to a combat zone with firefight. Our findings generally point to continued evidence that combat is associated with substantially increased risks of cigarette consumption, binge drinking, and drug use, with the mental health effects of combat continuing to be a strong mediator.

V. CONCLUSIONS

This study uses variation in overseas deployments among active duty U.S. service members to estimate the relationship between combat exposure and risky health behaviors. Our findings provide consistent evidence that combat exposure is associated with increased risk for subsequent cigarette consumption, binge drinking, and illicit drug use. The results are generally largest for those serving in the Army, Navy, and Marines relative to the Air Force. We also find that the psychological consequences of war can explain up to two-thirds of the estimated relationship between combat exposure and risky behaviors.

The findings presented in this study suggest that there may be substantial health behavioral costs to U.S. service members exposed to combat in the GWOT and that future estimates of the costs of war should consider these costs. Moreover, our estimates may be biased downwards if attrition through death is more likely among those exposed to combat. That is, if those who die are more likely to be exposed to more frequent or severe stresses that trigger risky behaviors, our estimates may understate the health behavioral effects of combat exposure. Moreover, because binge drinking and drug use often produce important externalities, such as domestic violence (Angelucci 2008; Cesur and Sabia forthcoming; El-Bassel et al. 2005; Exum 2002; Klosterman and Fals-Stewart 2006; Kyriacou et al. 1999; Markowitz and Grossman 1998, 2000; Stuart et al. 2008), increased medical care expenditures (Bates, Cesur, and Santerre 2015), traffic fatalities (Carpenter and Dobkin 2009; Cook and Durrance 2013; Young and Bielinska-Kwapisz 2006), and crime (Carpenter 2005, 2007; Cook and Durrance 2013; Mocan and Tekin 2005), these costs are likely to extend beyond the private costs to service members, and their families and communities.

Finally, it is important to note that our empirical approach identifies the health behavioral effects of facing enemy firefight among active duty deployed service members. This local average treatment effect could differ from the effect of conscription of civilians into war service via draft lottery. For example, if those who select into military service are more (or less) likely to be able to cope with the stresses of war, then the risky behavioral effects of war service in the general population may differ from our findings.

ABBREVIATIONS

AWOL: Absent Without Official Leave

CASI: Computer Assisted Self-Interviewing

GWOT: Global War On Terrorism

HRB: Health and Related Behavior

NLSAAH: National Longitudinal Study of Adolescent to Adult Health

PTSD: Post-Traumatic Stress Disorder APPENDIX TABLE A1 Means of Alternate Combat Measures and Control Variables in NLSAAH Data Combat Combat Variable A11 Exposure = 1 Exposure = 0 Mediating stress indicators and alternate combat measures PTSD 0.122 0.250 0.061 (0.328) (0.434) (0.239) Stress 4.32 4.58 4.19 (3.01) (3.05) (2.99) Suicide ideation 0.069 0.103 0.053 (0.254) (0.304) (0.224) Combat harm 0.299 0.787 0.068 (0.458) (0.410) (0.253) Control variables Height in inches 69.462 70.184 69.111 (3.772) (3.578) (3.819) Weight in pounds 188.4 189.7 187.7 (38.1) (34.7) (39.7) Religion: Protestant 0.317 0.368 0.292 (0.466) (0.484) (0.455) Religion: Catholic 0.223 0.227 0.221 (0.417) (0.420) (0.416) Religion: other Christian 0.193 0.162 0.208 (0.395) (0.370) (0.406) Religion: other 0.073 0.049 0.084 (0.260) (0.216) (0.278) Male 0.853 0.957 0.803 (0.354) (0.204) (0.399) Age in years 28.662 28.530 28.726 (1.711) (1.773) (1.678) Race: Black 0.251 0.205 0.274 (0.434) (0.405) (0.446) Race: other 0.081 0.065 0.090 (0.274) (0.247) (0.286) Race: Hispanic 0.159 0.119 0.179 (0.366) (0.325) (0.384) Education: some college or 0.671 0.670 0.671 vocational training (0.470) (0.471) (0.471) Education: college degree 0.174 0.157 0.182 (0.379) (0.365) (0.386) No health insurance 0.119 0.119 0.118 (0.324) (0.325) (0.324) Wave 1 picture vocabulary 96.113 97.930 95.229 test score (29.404) (27.513) (30.278) Log of parental income Wave 1 2.651 2.800 2.578 (1.683) (1.643) (1.699) Parent is married in Wave 1 0.620 0.622 0.618 (0.486) (0.486) (0.486) Parent is divorced, 0.205 0.216 0.200 separated, or widowed in (0.404) (0.413) (0.401) Wave 1 Biological mother's 0.342 0.308 0.358 education: high school degree (0.475) (0.463) (0.480) Biological mother's 0.221 0.232 0.216 education: some college (0.415) (0.424) (0.412) Biological mother's 0.266 0.303 0.247 education: college degree or (0.442) (0.461) (0.432) more Currently in the military 0.397 0.438 0.376 (0.490) (0.498) (0.485) Months served in the military 69.628 72.481 68.240 (33.724) (34.060) (33.516) Rank: specialist/corporal 0.338 0.303 0.355 (0.474) (0.461) (0.479) Rank: sergeant 0.368 0.400 0.353 (0.483) (0.491) (0.478) Rank: staff sergeant 0.149 0.178 0.134 (0.356) (0.384) (0.341) Rank: first class sergeant 0.087 0.103 0.079 or higher (0.282) (0.304) (0.270) Army 0.418 0.541 0.358 (0.494) (0.500) (0.480) Marines 0.181 0.254 0.145 (0.385) (0.437) (0.352) Navy 0.251 0.119 0.316 (0.434) (0.325) (0.465) Air Force 0.165 0.114 0.190 (0.371) (0.318) (0.392) Service exclusively after 0.234 0.287 0.208 11-Sep (0.424) (0.453) (0.406) Observations 565 185 380 Note: The means are generated using data drawn from Waves I and IV of the National Longitudinal Study of Adolescent to Adult Health. TABLE A2 Means of Alternate Combat Measures and Control Variables in DOD HRB Survey Combat Combat Variable All Exposure = 1 Exposure = 0 Mediating stress indicators and alternate combat measures PTSD 0.098 0.132 0.065 (0.297) (0.339) (0.247) Stress 0.135 0.152 0.120 (0.342) (0.359) (0.325) Suicide ideation 0.042 0.045 0.038 (0.200) (0.207) (0.192) Combat harm 0.139 0.260 0.009 (0.346) (0.448) (0.097) Control variables CONUS 0.693 0.755 0.635 (0.461) (0.430) (0.481) Rank E4-E6 0.531 0.513 0.548 (0.499) (0.500) (0.498) Rank E7-E9 0.160 0.164 0.157 (0.367) (0.371) (0.363) Rank W1-W5 0.033 0.049 0.018 (0.179) (0.215) (0.133) Rank 01-03 0.103 0.103 0.102 (0.303) (0.304) (0.303) Rank 04-010 0.096 0.108 0.083 (0.294) (0.311) (0.277) Number of deployments 1.576 1.885 1.284 (1.175) (1.050) (1.212) High school education 0.215 0.211 0.219 (0.411) (0.408) (0.414) Some college 0.486 0.480 0.492 (0.500) (0.500) (0.500) College degree and above 0.276 0.282 0.270 (0.447) (0.450) (0.444) Male 0.783 0.830 0.739 (0.412) (0.376) (0.439) Age 31.372 31.586 31.169 (7.678) (7.580) (7.764) Age squared 1043.114 1055.109 1031.766 (509.601) (504.462) (514.191) Black 0.176 0.160 0.190 (0.380) (0.367) (0.392) Asian 0.052 0.036 0.067 (0.222) (0.187) (0.250) Race other 0.118 0.106 0.130 (0.323) (0.308) (0.337) Married 0.633 0.656 0.611 (0.482) (0.475) (0.488) Divorced 0.109 0.115 0.102 (0.311) (0.319) (0.303) Observations 14740 7166 7574 Note: The means are generated using data drawn from the 2008 DOD HRB Survey. TABLE A3 Evidence on Exogeneity of Deployment Assignment in NLSAAH Data (1) (2) (3) Combat Combat Exposure = 1 Exposure = 1 Combat Versus Versus Exposure = 1 Combat Combat Zone Versus Variables Exposure = 0 without Exposure Noncombat Zone Predeployment 0.008 0.064 -0.050 smoking (0.062) (0.073) (0.076) Predeployment binge 0.040 0.057 -0.005 drinking (0.051) (0.066) (0.067) Predeployment drug 0.031 0.056 0.020 use (0.049) (0.065) (0.070) F-Test on joint 0.629 1.437 0.154 significance of 0.597 0.236 0.927 prior behaviors p Value Log height 0.235 0.075 0.632 (0.645) (0.793) (0.863) Log weight -0.181 -0.057 -0.294 (0.137) (0.179) (0.191) Religion: Protestant 0.054 0.078 0.053 (0.053) (0.071) (0.069) Religion: Catholic 0.025 0.044 -0.009 (0.061) (0.074) (0.082) Religion: other -0.022 0.007 -0.047 Christian (0.066) (0.084) (0.085) Religion: other -0.096 -0.009 -0.192 (0.094) (0.141) (0.136) F-Test on joint 1.409 0.572 1.235 significance of 0.235 0.684 0.300 religion p Value Male 0.244 *** 0.292 *** 0.365 *** (0.067) (0.096) (0.107) Age in years -0.276 -0.479 0.066 (0.373) (0.460) (0.517) Age in years squared 0.005 0.008 -0.001 (0.007) (0.008) (0.009) Race: Black -0.054 -0.053 -0.037 (0.054) (0.066) (0.076) Race: other -0.078 -0.082 0.008 (0.072) (0.086) (0.124) Race: Hispanic -0.128 ** -0.135 * -0.070 (0.059) (0.080) (0.085) F-Test on joint 1.646 1.053 0.284 significance of 0.182 0.372 0.837 race p Value Education: some -0.012 -0.044 0.037 college or (0.048) (0.062) (0.072) vocational training Education: college -0.023 -0.079 0.015 degree (0.073) (0.088) (0.109) F-Test on joint 0.0537 0.444 0.165 significance of 0.948 0.643 0.848 education No health insurance 0.002 0.056 -0.004 (0.060) (0.084) (0.088) Wave 1 picture 0.001 0.001 0.000 vocabulary test (0.001) (0.001) (0.001) score Log parental income 0.038 0.042 0.082 (0.032) (0.050) (0.051) Parent is married -0.030 0.064 -0.198 (0.109) (0.141) (0.127) Parent is divorced, 0.058 0.170 -0.086 separated, or (0.113) (0.142) (0.129) widowed F-Test on joint 1.463 1.892 1.858 significance of 0.236 0.156 0.161 parental marital status p Value Mother's education: -0.005 -0.003 0.014 some college (0.052) (0.068) (0.062) Mother's Education: 0.047 0.093 -0.003 college degree or (0.053) (0.058) (0.068) more F-Test on joint 0.489 1.595 0.0328 significance of mother's education p Value 0.614 0.208 0.968 Observations 565 416 334 R-squared 0.251 0.311 0.359 F-Test all 3.521 3.232 3.431 F-Test all p value 0.000 0.000 0.000 Notes: Robust standard errors corrected for clustering on the school are in parentheses. All models include controls for military- specific variables, including rank, branch of service, timing of service, and occupation. Regressions are estimated using data drawn from Waves I and IV of the National Longitudinal Study of Adolescent to Adult Health. Models also include missing dummy categories for each of the control variables. * Significant at 10%; ** significant at 5%; *** significant at 1% levels. TABLE A4 Means of Outcomes for Active Duty Samples Ages 24-32 Years NLSAAH Survey Combat Combat Variable All Exposure = 1 Exposure = 0 Smoking 0.338 0.400 0.303 (0.474) (0.493) (0.461) Binge drinking 0.166 0.180 0.158 (0.373) (0.386) (0.366) Any drug use 0.058 0.124 0.021 (0.234) (0.331) (0.144) Marijuana use 0.032 0.076 0.007 (0.176) (0.267) (0.084) Other drug use 0.031 0.062 0.014 (0.174) (0.242) (0.118) Army 0.429 0.556 0.357 (0.496) (0.500) (0.481) Marines 0.130 0.185 0.098 (0.337) (0.391) (0.298) Navy 0.241 0.148 0.294 (0.429) (0.358) (0.457) Air Force 0.232 0.148 0.280 (0.423) (0.358) (0.450) Observations 224 81 143 DOD HRB Survey Combat Combat Variable All Exposure = 1 Exposure = 0 Smoking 0.297 0.329 0.267 (0.457) (0.470) (0.443) Binge drinking 0.497 0.531 0.465 (0.500) (0.499) (0.499) Any drug use 0.043 0.059 0.027 (0.202) (0.236) (0.163) Marijuana use 0.012 0.018 0.006 (0.107) (0.132) (0.075) Other drug use 0.039 0.055 0.024 (0.194) (0.228) (0.153) Army 0.237 0.390 0.094 (0.425) (0.488) (0.292) Marines 0.167 0.226 0.111 (0.373) (0.418) (0.314) Navy 0.303 0.123 0.472 (0.460) (0.328) (0.499) Air Force 0.293 0.262 0.323 (0.455) (0.440) (0.468) Observations 5777 2791 2986 Notes: The means from the first three columns are generated using data drawn from Wave IV of the National Longitudinal Study of Adolescent to Adult Health; the means from the final three columns are generated using data drawn from the 2008 DOD HRB Survey. Note that some service members in the NLSAAH report multiple branches of service so the proportions may sum to greater than 1. TABLE A5 Estimates of Relationship between Combat Exposure and Combat Stress (1) (2) (3) (4) (5) All Army Marines Navy Air Force Outcome Panel A : NLSAAH (OLS full controls) PTSD 0.139 *** 0.166 0.111 0.017 0.058 (0.030) (0.067) (0.124) (0.062) (0.098) [563] [234] [102] [142] [93] Suicide 0.054 * 0.115 ** -0.045 0.033 -0.036 ideation (0.028) (0.049) (0.075) (0.119) (0.076) [563] [234] [102] [142] [93] Psychological 0.664 ** 0.557 -0.566 -0.468 2.009 * stress (0.264) (0.475) (0.913) (0.737) (1.093) [563] [234] [102] [142] [93] Outcome Panel B: DOD HRB (PSM) PTSD 0.066 *** 0.103 *** 0.105 ** 0.074 *** 0.041 ** (0.009) (0.037) (0.029) (0.026) (0.016) [4,876] [467] [552] [859] [1,204] Suicide 0.09 0.025 0.018 0.007 0.003 ideation (0.006) (0.024) (0.019) (0.018) (0.011) [4,876] [467] [552] [859] [1,204] Psychological 0.059 *** 0.78 * 0.036 0.056 ** 0.049 ** stress (0.011) (0.045) (0.035) (0.027) (0.020) [4,876] [467] [552] [859] [1,204] Notes: Panel A: robust standard errors corrected for clustering on the school are in parentheses. Number of observations is in brackets. All models use the full set of controls shown in Table A1 along with predeployment risky behaviors. In all models, military rank, timing of military service, branch of service, occupation indicators, and an indicator for having a check-up in the past year are controlled for. Models also include missing dummy categories for each of the control variables. Panel B: Bootstrapped standard errors are in parentheses. Number of observations is in brackets. All models use the full set of controls shown in Table A2. * Significant at 10%; ** significant at 5%; *** significant at 1% levels. TABLE A6 Propensity Score Matching Estimates of Relationship between Combat Exposure and Risky Behaviors for HRB Survey by Age (1) (2) (3) (4) (5) All Army Marines Navy Air Force Outcome Panel A : age 18-23 Smoking 0.050 0.333 * -0.011 0.000 -0.079 (0.053) (0.175) (0.088) (0.185) (0.126) [519] [53] [188] [54] [76] Binge 0.04 0.222 0.011 0.185 0.132 drinking (0.054) (0.137) (0.073) (0.175) (0.142) [519] [53] [188] [54] [76] Any drug use 0.050 *** 0.037 0.106 *** -0.037 -0.079 (0.032) (0.120) (0.040) (0.089) (0.054) [519] [53] [188] [54] [76] Marijuana use 0.015 -0.037 0.032" 0.037 -0.053 (0.020) (0.075) (0.025) (0.071) (0.046) [519] [53] [188] [54] [76] Other drug 0.050 0.037 0.096 ** 0.000 -0.053 use (0.031) (0.113) (0.037) 0.092 (0.051) [519] [53] [188] [54] [76] Outcome Panel B: age 24-32 Smoking 0.046 * 0.050 0.089 -0.024 0.095 (0.025) (0.099) (0.088) (0.077) (0.064) [1,489] [163] [252] [161] [400] Binge 0.061 * 0.050 0.013 -0.016 0.065 drinking (0.032) (0.105) (0.096) (0.095) (0.065) [1,489] [163] [252] [161] [400] Any drug use 0.037 ** 0.038 0.000 0.031 0.015 (0.013) (0.061) (0.029) (0.038) (0.023) [1,489] [163] [252] [161] [400] Marijuana use 0.016 *** -0.013 0.000 0.024 0.000 (0.006) (0.027) (0.009) (0.023) (0.011) [1,489] [163] [252] [161] [400] Other drug 0.035 *** 0.025 0.000 0.024 0.019 use (0.013) (0.056) (0.029) (0.036) (0.022) [1,489] [163] [252] [161] [400] Outcome Panel C: age 33+ Smoking 0.025 0.041 -0.013 0.0188 0.080 * (0.025) (0.098) (0.099) (0.071) (0.042) [1,700] [144] [152] [321] [403] Binge 0.055 * 0.278 *** 0.039 0.013 0.025 drinking (0.031) (0.102) (0.085) (0.071) (0.066) [1,700] [144] [152] [321] [403] Any drug use 0.009 0.014 0.000 0.013 0.035 * (0.011) (0.054) (0.025) (0.029) (0.021) [1,700] [144] [152] [321] [403] Marijuana use 0.003 0.000 0.000 0.006 0.000 (0.003) (0.013) (0.000) (0.013) (0.003) [1,700] [144] [152] [321] [403] Other drug 0.008 0.014 0.000 0.013 0.035 use (0.011) (0.054) 0.025 (0.027) (0.021) [1,700] [144] [152] [321] [403] Notes: Bootstrapped standard errors are in parentheses. Number of observations is in brackets. All models use the full set of controls shown in Table A2. * Significant at 10%; ** significant at 5%; *** significant at 1% levels.

REFERENCES

Angelucci, M. "Love on the Rocks: Domestic Violence and Alcohol Abuse in Rural Mexico." The B.E. Journal of Economic Analysis & Policy, 8(1), 2008, article 43.

Angrist, J. D. "Lifetime Earnings and the Vietnam Era Draft Lottery: Evidence from Social Security Administrative Records." American Economic Review, 80(3), 1990, 313-36.

Bates, L. J., R. Cesur, and R. E. Santerre. "Short-Run Marginal Medical Costs from Booze and Butts: Evidence from the States." Southern Economic Journal, 81(4), 2015, 1074-95.

Bedard, K., and O. Deschenes. "The Long-Term Impact of Military Service on Health: Evidence from World War II and the Korean War." American Economic Review, 96(1), 2006, 176-94.

Bray, R., M. Pemberton, L. L. Hourani, M. Witt, K. L. Rae Olmsted, J. M. Brown, B. L. Weimer, M. E. Lane, M. E. Marsden, S. Scheffler, R. Vandermaas-Peeler, K. R. Aspinwall, E. Anderson, K. Spagnola, K. Close, J. L. Gratton, S. Calvin, and M. Bradshaw. 2008 Department of Defense Survey of Health Related Behaviors among Active Duty Military Personnel: A Component of the Defense Lifestyle Assessment Program. Research Triangle Park, NC: RTI International, 2009.

Carpenter, C. "Heavy Alcohol Use and the Commission of Nuisance Crime: Evidence from Underage Drunk Driving Laws." American Economic Review: Papers and Proceedings, 95(2), 2005, 267-72.

--. "Heavy Alcohol Use and Crime: Evidence from Underage Drunk Driving Laws." Journal of Law and Economics, 50(3), 2007, 539-57.

Carpenter, C., and C. Dobkin. "The Effect of Alcohol Consumption on Mortality: Regression Discontinuity Evidence from the Minimum Drinking Age." American Economic Journal: Applied Economics, 1(1), 2009, 164-82.

Cesur, R., and J. J. Sabia. "When War Comes Home: The Effect of Combat Service on Domestic Violence." Review of Economics and Statistics, Forthcoming.

Cesur, R., J. J. Sabia, and E. Tekin. "The Psychological Costs of War: Military Combat and Mental Health." Journal of Health Economics, 32(1), 2013, 51-65.

--. "Combat Exposure and Migraine Headache: Evidence from Exogenous Deployment Assignment." Economics & Human Biology, 16, 2015, 81-99.

Cook, P. J., and C. P. Durrance. "The Virtuous Tax: Lifesaving and Crime-Prevention Effects of the 1991 Federal Alcohol-Tax Increase." Journal of Health Economics, 32(1), 2013, 261-7.

Dobkin, C., and R. Shabani. "The Health Effects of Military Service." Economic Inquiry, 47(1), 2009, 69-80.

El-Bassel, N., L. Gilbert, E. Wu, H. Go, and J. Hill. "Relationship between Drug Abuse and Intimate Partner Violence: A Longitudinal Study among Women Receiving Methadone." American Journal of Public Health, 95(3), 2005, 465-70.

Engel, R. C., L. B. Gallagher, and D. S. Lyle. "Military Deployments and Children's Academic Achievement: Evidence from Department of Defense Education Activity Schools." Economics of Education Review, 29(1), 2010, 73-82.

Exum, M. L. "The Application and Robustness of the Rational Choice Perspective in the Study of Intoxicated and Angry Intentions to Aggress." Criminology, 40(4), 2002, 933-66.

Gold, P. B., B. E. Engdahl, R. E. Eberly, R. J. Blake, W. F. Page, and B. C. Frueh. "Trauma Exposure, Resilience, Social Support, and PTSD Construct Validity among Former Prisoners of War." Social Psychiatry and Psychiatric Epidemiology, 35, 2000, 36-42.

Grossman, D., and B. Siddle. "Psychological Effects of Combat," in Encyclopedia of Violence, Peace and Conflict, Vol. 3, edited by L. Kurtz, and J. Turpin. San Diego, CA: Academic Press, 1999, 139-49.

Hearst, N., T. B. Newman, and S. B. Hulley. "Delayed Effects of the Military Draft on Mortality: A Randomized Natural Experiment." New England Journal of Medicine, 314, 1986, 620-24.

Hirsch, B. T., and S. L. Mehay. "Evaluating the Labor Market Performance of Veterans Using a Matched Comparison Group Design." Journal of Human Resources, 38(3), 2003, 673-700.

Hoerster, K., K. Lahavot, T. Simpson, M. McFall, G. Reiber, and K. Nelson. "Health and Health Behavior Differences: US Military, Veteran, and Civilian Men." American Journal of Preventive Medicine, 43(5), 2012, 283-489.

Hoge, C. W., C. A. Castro, S. C. Messer, D. McGurk, D. I. Cotting, and R. L. Koffman. "Combat Duty in Iraq and Afghanistan, Mental Health Problems and Barriers to Care." New England Journal of Medicine, 351(1), 2004, 13-22.

Hoge, C. W., J. L. Auchterlonie, and C. S. Milliken. "Mental Health Problems, Use of Mental Health Services, and Attrition from Military Service after Returning from Deployment to Iraq or Afghanistan." Journal of the American Medical Association, 295(9), 2006, 1023-32.

Hooper, R., R. J. Rona, M. Jones, N. T. Fear, L. Hull, and S. Wessely. "Cigarette and Alcohol Use in the UK Armed Forces and Their Association with Combat Exposure: A Prospective Study." Addictive Behaviors, 33(8), 2008, 1067-71.

Institutes of Medicine. Combating Tobacco Use in Military and Veteran Populations. National Academy of Sciences, 2009. Accessed April 1, 2013. http://www.nap .edu/catalog/12632.html.

Jacobson, I. G., M. A. K. Ryan, T. I. Hooper, T. C. Smith, P. J. Amoroso, E. J. Boyko, G. D. Gackstetter, T. S. Wells, and N. S. Bell. "Alcohol Use and Alcohol-Related Problems before and after Military Combat Deployment." Journal of the American Medical Association, 300(6), 2008, 663-75.

Joiner, T. E. Why People Die by Suicide. Cambridge, MA: Harvard University Press, 2005.

Kessler, R. C., G, Andrews, L. J. Colpe, E. Hiripi, D. K. Mroczek, S. L. Normand, E. E. Walters, and A. M. Zaslavsky. "Short Screening Scales to Monitor Population Prevalences and Trends in Non-Specific Psychological Distress." Psychological Medicine, 32, 2002, 959-76.

Killgore, W. D. S., D. I. Cotting, J. L. Thomas, A. L. Cox, D. McGurk, A. H. Vo, C. A. Castro, and C. W. Hoge. "Post-Combat Invincibility: Violent Combat Experiences Are Associated with Increased Risk-Taking Propensity Following Deployment." Journal of Psychiatric Research, 42(13), 2008, 1112-21.

Klosterman, K. C., and W. Fals-Stewart. "Intimate Partner Violence and Alcohol Use: Exploring the Role of Drinking in Partner Violence and Its Implications for Intervention." Aggression and Violent Behavior, 11(6), 2006, 587-97.

Kyriacou, D. N., D. Anglin, E. Taliaferro, S. Stone, T. Tubb, J. A. Linden, R. Muelleman, E. Barton, and J. F. Kraus. "Risk Factors for Injury to Women from Domestic Violence." New England Journal of Medicine, 341, 1999, 1892-98.

Lyle, D. S. "Using Military Deployments and Job Assignments to Estimate the Effect of Parental Absences and Household Relocations on Children's Academic Achievement." Journal of Labor Economics, 24(2), 2006, 319-50.

Markowitz, S., and M. Grossman. "Alcohol Regulation and Domestic Violence towards Children." Contemporary Economic Policy, 16, 1998, 309-20.

--. "The Effects of Beer Taxes on Physical Child Abuse." Journal of Health Economics, 19(2), 2000, 271-82.

McFall, M. E., P. W. Mackay, and D. M. Donovan. "Combat-Related Posttraumatic Stress Disorder and Severity of Substance Abuse in Vietnam Veterans." Journal of Studies on Alcohol, 53, 1992, 357-63.

Mocan, N. H., and E. Tekin. "Drugs and Juvenile Crime: Evidence from a Panel of Siblings and Twins," in Substance Use: Individual Behavior, Social Interactions, Markets, and Politics, edited by B. Lindgren and M. Grossman. Amsterdam, The Netherlands: Elsevier, 2005, 91-120.

Morrell, D. "Military Encounters Resistance to Ban on Smoking." 2009. Accessed June 10, 2013. http://www .boston.com/news/world/middleeast/articles/2009/08/ 30/study&uscore;proposing&uscore;smoking&uscore; ban&uscore;in&uscore;military&uscore;draws& uscore;protests/?camp=pm

National Council on Alcoholism and Drug Dependence. "Veterans and Alcohol," 2012. Accessed July 10, 2013. http://www.ncadd.org/index.php/leam-about-alcohol/ seniors- vets-and- women/198- veterans-and-alcohol

Negrusa, B., and S. Negrusa. "Home Front: Post-Deployment Mental Health and Divorces." Demography, 51(3), 2014, 895-916.

Page, W. F., B. E. Engdahl, and R. E. Eberly. "Prevalence and Correlates of Depressive Symptoms among Former Prisoners of War." Journal of Nervous and Mental Disease, 179, 1991, 670-77.

Price, R. K., N. K. Risk, A. H. Haden, C. E. Lewis, and E. L. Spitznagel. "Post-Traumatic Stress Disorder, Drag Dependence, and Suicidality among Male Vietnam Veterans with a History of Heavy Drug Use." Drug and Alcohol Dependence, 76S, 2004, 31-43.

Rohlfs, C. "Does Combat Exposure Make You a More Violent or Criminal Person? Evidence from the Vietnam Draft." Journal of Human Resources, 45(2), 2010, 271-300.

Smith, T. C., I. G. Jacobson, T. I. Hooper, C. A. LeardMan, E. J. Boyko, B. Smith, G. D. Gackstetter, T. S. Well, P. J. Amoroso, G. C. Gray, J. R. Riddle, M. A. K. Ryan, and the Millennium Cohort Study Team. "Health Impact of US Military Service in a Large Population-Based Military Cohort: Findings of the Millennium Cohort Study, 2001-2008." BMC Public Health, 11, 2011, 69-78.

Stuart, G. L., J. R. Temple, K. W. Follansbee, M. M. Bucossi, J. C. Hellmuth, and T. M. Moore. "The Role of Drug Use in a Conceptual Model of Intimate Partner Violence in Men and Women Arrested for Domestic Violence." Psychology of Addictive Behaviors, 22(1), 2008, 12-24.

Tanielian, T., and L. H. Jaycox. Invisible Wounds of War: Psychological and Cognitive Injuries, Their Consequences, and Services to Assist Recovery. Santa Monica, CA: Rand Corporation, 2008.

Thomsen, C. J., V. A. Stander, S. K. McWhorter, M. M. Rabenhorst, and J. S. Milner. "Effects of Combat Deployment on Risky and Self-Destructive Behavior among Active Duty Military Personnel." Journal of Psychiatric Research, 45, 2011, 1321-31.

Vaughan, D. "Risky Business." Today's Officer, 2006, 21-24.

Weathers, F. W., B. T. Litz, J. A. Huska, and T. M. Keane. The PTSD Checklist-Civilian Version (PCL-C). Boston, MA: National Center for PTSD, 1994.

Williams, T. "Suicides Outpacing Deaths for Troops." New York Times, 2012.

Young, D. J., and A. Bielinska-Kwapisz. "Alcohol Prices, Consumption and Traffic Fatalities." Southern Economic Journal, 72, 2006, 690-70.

Cesur: Assistant Professor, Finance Department, University of Connecticut, Storrs, CT 06269-1041. Phone (860) 486-6315, Fax (860) 486-0634, E-mail [email protected]

Chesney: Graduate Student, Department of Economics, San Diego State University, San Diego, CA 92182-4485. Phone (619) 594-1675, Fax (619) 594-5062, E-mail steelcia2011 @gmail.com

Sabia: Associate Professor, Department of Economics, San Diego State University, San Diego, CA 92182-4485. Phone (619) 594-2407, Fax (619) 594-5062, E-mail [email protected]

doi: 10.1111/ecin.12312

(1.) Military policymakers have, in fact, argued for not attempting to curb legal risky health behaviors to the extent that they help service members to cope. For instance, the office of former U.S. Secretary of Defense Robert Gates announced the Secretary's opposition to a military smoking ban on the following grounds:

[Secretary Gates] knows that the situation [service-members] are confronting is stressful enough as it is. I don't think he is interested in adding to the stress levels by taking away one of the few outlets they may have to relieve stress. (Morrell 2009)

(2.) However, Hooper et al. (2008) found that military service in the United Kingdom is essentially unrelated to cigarette consumption.

(3.) The results we describe below are quantitatively and qualitatively similar when those (11 individuals) who reported that their military service started prior to the Wave I interviews are included. Because the NLSAAH survey does not provide the age at which the survey respondents completed high school, we excluded these individuals from the analysis.

(4.) When we restrict the NLSAAH sample to those who are currently on active duty at the time of the Wave IV survey, the pattern of results is similar to those reported on the full sample in the paper (albeit less precisely estimated because the sample size is substantially smaller).

(5.) In our prior study (Cesur, Sabia, and Tekin 2013), our key measure of deployment assignment was an indicator of deployment to a combat zone as compared to a noncombat zone. We focus on Combat Exposure for two reasons: (1) it is the comparable measure of lifetime combat deployment available in the DOD HRB Survey, the parallel dataset used in this paper, and (2) to focus on high intensity combat experiences found to be the most stressful by Cesur, Sabia, and Tekin (2013). However, in unreported results available upon request, we examine the effect of combat zone deployment (whether or not it was accompanied by enemy firefight) on risky health behaviors. These estimates documented that those assigned to combat zones are significantly more likely to engage in risky health behaviors as compared to their counterparts assigned to noncombat zones. However, the estimated effect is effect is largest for those assigned to combat zones with firefight.

(6.) "Favorite" drug in the NLSAAH data is defined as the drug that the respondent uses most frequently during their lives. Therefore, measurement error may be introduced if the respondent used a nonmarijuana drug in the last 30 days but this drug was not the drug that he or she had designated as the "favorite" most frequently used drug in his or her lifetime.

(7.) See Bray et al. (2009) for more detailed information on the DOD HRB data collection strategy.

(8.) Exposure to combat fire is defined as answering yes to either one of the following experiences in these questions: Thinking about all of your deployments (combat and noncombat), how many times have you had each of the following experiences? I, or members of my unit, received incoming fire from small arms, artillery, rockets, or mortars. My unit fired on the enemy.

(9.) The DOD survey measures combat versus noncombat zone deployment only in the previous 12 months. The use of Combat Exposure allows our measures to be consistent across datasets.

(10.) Respondents were, however, told to exclude "steroids, sexual enhancers, and analgesics" from their report of illicit drug use.

(11.) A Major Command (MAJCOM) is a subdivision for a particular military installation responsible for a specific combat/support mission. These MAJCOMs include U.S. Army Training and Doctrine Command, U.S. Army Europe, U.S. Army Pacific, 8th Army, U.S. Fleet Forces Command, Commander Pacific Forces, Naval Medical Command, Com mander Naval Installations Command, Marine Corps Installations East, Marine Corps Installations West, Air Combat Command, Air Education and Training Command, Air Force Materiel Command, Air Force Space Command, Air Mobility Command, Pacific Air Forces, and U.S. Air Forces Europe.

(12.) Lyle (2006) and Engel, Gallagher, and Lyle (2010) are able to empirically test this theoretical point in their data by instrumenting individual soldier deployment with unit deployment. The results using the instrument are qualitatively and quantitatively similar to treating individual deployment as exogenously determined.

(13.) Engel, Gallagher, and Lyle (2010) note that "as a rule, [HRC] do[es] not take into consideration the welfare of an individual enlisted soldier ... nor do they consider the average characteristics of units and families" when making assignment decisions (Engel, Gallagher, and Lyle 2010, 76).

(14.) Lyle (2006) and Engel, Gallagher, and Lyle (2010) persuasively argue that Army Human Resources Command (AHRC) "regards soldiers of the same rank and occupation as equals" (323).

(15.) Column (1) presents results when comparing those exposed to combat to those deployed to either combat zones without actual combat exposure or to noncombat zones overseas; column (2) compares those exposed to combat to only those deployed to combat zones without exposure; and column (3) compares those exposed to combat to only those deployed to noncombat zones.

(16.) Note that weight, educational attainment, and health insurance status were measured at Wave IV. Hence, they may be affected by combat exposure. In unreported specifications, we estimated our models by excluding these variables to get a handle on whether body weight, schooling, and health insurance status mediate the relationship between combat exposure and risky health behaviors. These results are nearly identical with the main estimates.

(17.) We also reestimate the effect of each background characteristic on deployment assignment (not controlling for the others) conditional on military observables, as well as gender for the reason noted above. These estimates, which are available from the authors upon request, reflect a similar pattern of results.

(18.) During GWOT, women were, for the first time, allowed to be permanently attached to a battalion in roles such as "radio operators, medics, tank mechanics and other critical jobs." (Williams 2012).

(19.) Service members are eligible to be assigned to all positions for which they are qualified, except that women shall be excluded from assignment to units below the brigade level whose primary mission is to engage in direct combat on the ground. (Burelli 2013)

(20.) There is evidence that non-Hispanic service members are likely to be exposed to combat, so we control for ethnicity indicators in all models, but note that if we limit the sample to non-Hispanics as well as non-Hispanic males only, all of the pattern of results discussed below holds. In unreported specifications, we also estimated the effect of each background characteristic on deployment assignment without controlling for other observable characteristics (other than the military observables and gender) and find a similar pattern of results.

(21.) We also experimented with trimming 5%-10% of observations with predicted probabilities furthest from the highest and lowest predicted probabilities. Our findings were robust to caliper parameters and matching method employed (such as radial matching).

(22.) An alternative to the above approach would be to simply estimate an OLS Equation (2) using the DOD data and only conditioning on the Zs. The findings using this approach were not quantitatively or qualitatively different from those obtained using the above model.

(23.) Tables A1 and A2 provide the means of the independent variables (as well as Combat Harm, an alternate combat exposure measure and combat stressors, PTSD, Depression and Suicidal Ideation, discussed below) by Combat Exposure for the NLSAAH and DOD HRB samples, respectively.

(24.) In Table A4, we compare the means of the outcomes for the DOD data for those aged 24-32 years and for the active duty sample in the NLSAAH. The means are more similar with this "apples to apples" comparison. Interestingly, rates of nonmarijuana drug use, which includes harder drugs such as cocaine, methamphetamine, and heroin, are comparable to or even greater than rates of marijuana use. One explanation for this is that many harder drugs are more quickly eliminated from the body than marijuana, decreasing the likelihood of detection from random drug tests.

(25.) To economize on space, we only present the estimated coefficients on the variable of interest, Combat Exposure. Estimated coefficients on the control variables are available upon request.

(26.) In unreported specifications, we also estimated the models presented in row (1) with military controls only to test whether the results differ when the full set of personal and family background characteristics, including predeployment risky behaviors, are excluded as controls to the estimating equation. These estimates, which are available from the authors upon request, produced nearly identical results, consistent with the hypothesis that combat zone deployment is exogenous to personal and family background characteristics, including predeployment risky health behaviors.

(27.) As noted above, one concern with the above DOD HRB data is the lack of information on occupation. However, when we compare nearest neighbor matching estimates in the NLS AAH with and without controls for occupation, the results from this exercise (which are available upon request) produce estimates that are statistically equivalent and quantitatively nearly identical, suggesting that controls for military rank, number of deployments, educational attainment, and MAJCOM may be sufficient proxies for military occupation such that the DOD estimates are not plagued by significant bias.

(28.) Table A6 presents estimates of the effect of combat exposure in the DOD HRB survey by age: those aged 18-23, aged 24-32 (to match the NLSAAH sample), and aged 33 and older. The results suggest the largest impacts of combat exposure on risky behaviors for those in their mid-twenties to early thirties.

(29.) We also produced the OLS estimates of the impact of Combat Exposure on risky health behaviors in the DOD HRB survey. These estimates are very similar to the results presented in Table 4 and are available from the authors upon request.

(30.) In both datasets, we measure suicide ideation dichotomously following Cesur and Sabia (forthcoming) using the individuals' report of whether s/he ever seriously thought about committing suicide during the past year. In the NLSAAH, we measure PTSD via respondent's self-report of whether "a doctor, nurse, or other health-care provider ever told you that you have or had PTSD"; and we use the Cohen Stress Scale, which is preconstructed by the NLSAAH. In the DOD HRB survey, the PTSD measure is not a self-reported diagnosis, but rather generated via a PTSD Checklist-Civilian Version test (Weathers et al. 1994), and the stress variable was created based on survey questions pertaining to a six-item Kessler psychological distress scale, i.e., the K-6 scale, (Kessler et al. 2002). See Cesur and Sabia (forthcoming) for a detailed description of how the Cohen Stress Scale in the NLSAAH, and how the PTSD and K-6 scale in the DOD HRB are constructed. The means of each of these variables, by combat exposure, are available in Tables A1 and A2 for the NLSAAH and the DOD HRB surveys, respectively.

(31.) Another might be that the mechanisms differ by age. We test this hypothesis by limiting the DOD sample to those aged 24-32 years to match the NLSAAH sample (see Table A6). Results continue to show that the mediators continue to explain a larger share of the estimated effects in the DOD HRB sample as compared to NLSAAH sample.

(32.) The DOD HRB data do not provide information on combat deployment length. Instead, we employed the total number of combat and peacekeeping missions' deployments after September 11 as an alternative combat exposure indicator. These results suggest a direct association between number of deployments and risky health behaviors. TABLE 1 Means of Risky Behaviors and Branch of Service, by Combat Exposure NLSAAH Survey Combat Combat Variable All Exposure = 1 Exposure = 0 Smoking 0.391 0.478 0.348 (0.488) (0.501) (0.477) Binge drinking 0.202 0.243 0.182 (0.402) (0.430) (0.387) Any drug use 0.142 0.184 0.121 (0.349) (0.388) (0.327) Marijuana use 0.118 0.142 0.106 (0.323) (0.350) (0.308) Other drug use 0.044 0.070 0.032 (0.206) (0.256) (0.175) Army 0.418 0.541 0.358 (0.494) (0.500) (0.480) Marines 0.181 0.254 0.145 (0.385) (0.437) (0.352) Navy 0.251 0.119 0.316 (0.434) (0.325) (0.465) Air Force 0.165 0.114 0.190 (0.371) (0.318) (0.392) Observations 565 185 380 DOD HRB Survey Combat Combat Variable All Exposure = 1 Exposure = 0 Smoking 0.263 0.282 0.246 (0.440) (0.450) (0.431) Binge drinking 0.445 0.476 0.415 (0.497) (0.500) (0.493) Any drug use 0.043 0.059 0.029 (0.204) (0.235) (0.168) Marijuana use 0.013 0.020 0.007 (0.113) (0.138) (0.082) Other drug use 0.039 0.054 0.025 (0.194) (0.226) (0.156) Army 0.223 0.356 0.097 (0.416) (0.479) (0.297) Marines 0.206 0.284 0.134 (0.405) (0.451) (0.340) Navy 0.292 0.122 0.452 (0.455) (0.327) (0.498) Air Force 0.279 0.239 0.317 (0.448) (0.426) (0.465) Observations 14740 7166 7574 Notes: The means from the first three columns are generated using data drawn from Wave IV of the National Longitudinal Study of Adolescent to Adult Health; the means from the final three columns are generated using data drawn from the 2008 DOD HRB Survey. Note that some service members in the NLSAAH report multiple branches of service, so that the proportions may sum to greater than 1. TABLE 2 Estimates of the Relationship between Combat Exposure and Risky Behaviors in NLSAAH (1) (2) (3) Variables Smoking Binge Drinking Any Drug Panel A: full sample Combat exposure 0.104 0.026 0.066 (0.043) (0.043) (0.038) [563] [554] [565] Combat exposure (comparison 0.085 * 0.033 0.063 * group: combat zone deployed (0.051) (0.053) (0.035) without combat exposure) [414] [406] [416] Combat exposure (comparison 0.149 *** 0.032 0.067 group: noncombat zone (0.053) (0.054) (0.050) deployed) [333] [329] [334] Panel B: Army Combat exposure 0.109 0.034 0.007 (0.073) (0.074) (0.056) [235] [231] [236] Panel C: Marines Combat exposure 0.073 -0.029 0.003 (0.134) (0.095) (0.088) [102] [101] [102] Panel D: Navy Combat exposure 0.276 -0.093 0.183 (0.127) (0.103) (0.091) [141] [140] [142] Panel E: Air Force Combat exposure -0.218 0.341 ** 0.029 (0.232) (0.167) (0.125) [93] [90] [93] (4) (5) Variables Marijuana Other Drug Panel A: full sample Combat exposure 0.038 0.034 (0.038) (0.021) [560] [565] Combat exposure (comparison 0.032 0.033 group: combat zone deployed (0.034) (0.026) without combat exposure) [412] [416] Combat exposure (comparison 0.040 0.035 * group: noncombat zone (0.052) (0.020) deployed) [331] [334] Panel B: Army Combat exposure -0.026 0.058 * (0.054) (0.031) [233] [236] Panel C: Marines Combat exposure 0.000 0.154 * (0.085) (0.090) [101] [102] Panel D: Navy Combat exposure 0.153 0.009 (0.084) (0.048) [141] [142] Panel E: Air Force Combat exposure 0.066 -0.039 (0.112) (0.052) [93] [93] Notes: Robust standard errors corrected for clustering on the school are in parentheses. Number of observations is in brackets. All models use the full set of controls shown in Table A1 along with predeployment risky behaviors. In all models, military rank, timing of military service, branch of service, occupation indicators, and an indicator for having a check-up in the past year are controlled for. Models also include missing dummy categories for each of the control variables. * Significant at 10%; ** significant at 5%; *** significant at 1% levels. TABLE 3 Evidence on Matching on Observables in DOD HRB Survey All Combat Noncombat p Value Rank E4-E6 0.54 0.54 0.84 Rank E7-E9 0.16 0.16 0.93 Rank WI-W5 0.03 0.04 0.00 Rank 01-03 0.11 0.11 0.35 Rank 04-010 0.10 0.09 0.24 Number of deployments in lifetime 1.62 1.61 0.39 High school education 0.20 0.21 0.22 Some college 0.49 0.50 0.82 College degree or above 0.29 0.27 0.11 Male 0.77 0.78 0.36 Age 31.65 31.37 0.16 Age squared 1062 1046 0.21 Race (Black) 0.18 0.17 0.96 Race (other) 0.71 0.74 0.05 Married 0.54 0.54 0.84 Divorced 0.16 0.16 0.93 Currently in contiguous United 0.71 0.74 0.05 States (CONUS) Army Combat Noncombat p Value Rank E4-E6 0.52 0.51 0.84 Rank E7-E9 0.16 0.17 0.65 Rank WI-W5 0.07 0.05 0.46 Rank 01-03 0.13 0.15 0.36 Rank 04-010 0.06 0.09 0.32 Number of deployments in lifetime 1.25 1.18 0.79 High school education 0.16 0.16 0.85 Some college 0.48 0.45 0.72 College degree or above 0.35 0.35 0.83 Male 0.73 0.71 0.76 Age 31.65 31.13 0.54 Age squared 1059 1027 0.52 Race (Black) 0.25 0.26 0.88 Race (other) 0.56 0.65 0.03 Married 0.52 0.51 0.84 Divorced 0.16 0.17 0.65 Currently in contiguous United 0.56 0.65 0.03 States (CONUS) Marines Combat Noncombat p Value Rank E4-E6 0.49 0.54 0.30 Rank E7-E9 0.10 0.10 0.84 Rank WI-W5 0.06 0.06 0.67 Rank 01-03 0.15 0.13 0.43 Rank 04-010 0.10 0.09 0.60 Number of deployments in lifetime 1.27 1.30 0.37 High school education 0.37 0.42 0.40 Some college 0.36 0.33 0.83 College degree or above 0.27 0.25 0.54 Male 0.83 0.84 0.94 Age 28.68 28.69 0.79 Age squared 877 878 0.80 Race (Black) 0.08 0.09 0.54 Race (other) 0.91 0.91 0.95 Married 0.49 0.54 0.30 Divorced 0.10 0.10 0.84 Currently in contiguous United 0.91 0.91 0.95 States (CONUS) Navy Combat Noncombat p Value Rank E4-E6 0.58 0.60 0.57 Rank E7-E9 0.16 0.15 0.72 Rank WI-W5 0.01 0.01 0.65 Rank 01-03 0.09 0.08 0.79 Rank 04-010 0.09 0.10 0.73 Number of deployments in lifetime 1.82 1.86 0.70 High school education 0.23 0.26 0.43 Some college 0.50 0.46 0.26 College degree or above 0.25 0.25 0.76 Male 0.82 0.86 0.24 Age 32.05 32.09 0.90 Age squared 1084 1087 0.90 Race (Black) 0.17 0.16 0.75 Race (other) 0.75 0.83 0.01 Married 0.58 0.60 0.57 Divorced 0.16 0.15 0.72 Currently in contiguous United 0.75 0.83 0.01 States (CONUS) Air Force Combat Noncombat p Value Rank E4-E6 0.52 0.57 0.11 Rank E7-E9 0.18 0.16 0.22 Rank WI-W5 0.00 0.00 Rank 01-03 0.12 0.09 0.10 Rank 04-010 0.12 0.12 0.51 Number of deployments in lifetime 1.61 1.60 0.82 High school education 0.12 0.14 0.48 Some college 0.54 0.59 0.23 College degree or above 0.33 0.27 0.08 Male 0.80 0.81 0.76 Age 32.24 32.06 0.59 Age squared 1093 1085 0.67 Race (Black) 0.13 0.12 0.67 Race (other) 0.68 0.68 0.88 Married 0.52 0.57 0.11 Divorced 0.18 0.16 0.22 Currently in contiguous United 0.68 0.68 0.88 States (CONUS) Notes: Nearest neighbor matching is employed using data drawn from the 2008 DOD HRB Survey. TABLE 4 Evidence on Matching on Major Command in DOD HRB Survey All p Combat Noncombat Value Army U.S. Army Training and Doctrine Command (a) 0.02 0.03 0.00 U.S. Army Europe (a) 0.03 0.03 0.02 U.S. Army Pacific (a) 0.03 0.03 0.09 Eighth Army (a) 0.03 0.03 0.32 Navy U.S. Fleet Forces Command (a) 0.1 0.07 0 Commander Pacific Forces (a) 0.04 0.04 0.08 Naval Medical Command (a) 0.03 0.02 0.01 Commander Naval Installations Command (a) 0.04 0.03 0.09 Marines Marine Corps Installations East (a) 0.12 0.15 0 Marine Corps Installations West (a) 0.07 0.08 0 Air Force Air Combat Command (a) 0.08 0.07 0.02 Air Education and Training Command (a) 0.04 0.03 0.04 Air Force Materiel Command (a) 0.04 0.03 0.74 Air Force Space Command (a) 0.04 0.03 0.3 Air Mobility Command (a) 0.07 0.06 0.07 Pacific Air Forces (a) 0.06 0.04 0.03 U.S. Air Forces Europe (a) 0.04 0.04 0.54 Branches p Combat Noncombat Value Army U.S. Army Training and Doctrine Command (a) 0.11 0.12 0.300 U.S. Army Europe (a) 0.17 0.13 0.24 U.S. Army Pacific (a) 0.15 0.13 0.58 Eighth Army (a) 0.11 0.09 0.22 Navy U.S. Fleet Forces Command (a) 0.24 0.19 0.13 Commander Pacific Forces (a) 0.14 0.14 0.91 Naval Medical Command (a) 0.14 0.15 0.65 Commander Naval Installations Command (a) Marines Marine Corps Installations East (a) 0.33 0.31 0.85 Marine Corps Installations West (a) Air Force Air Combat Command (a) Air Education and Training Command (a) 0.09 0.1 0.37 Air Force Materiel Command (a) 0.1 0.11 0.47 Air Force Space Command (a) 0.08 0.1 0.12 Air Mobility Command (a) 0.21 0.19 0.29 Pacific Air Forces (a) 0.15 0.16 0.49 U.S. Air Forces Europe (a) 0.11 0.1 0.41 Notes: Nearest neighbor matching is employed using data drawn from the 2008 DOD HRB Survey. (a) These represent various major commands for the Army, Navy, Marines, and Air Force, respectively. TABLE 5 Propensity Score Matching Estimates of Relationship between Combat Exposure and Risky Behaviors, DOD HRB Survey (1) (2) (3) (4) (5) Binge Sample Smoking Drinking Any Drug Marijuana Other Drug All 0.036 *** 0.041 *** 0.028 *** 0.013 *** 0.029 *** (0.014) (0.014) (0.007) (0.003) (0.007) [4,876] [4,876] [4,876] [4,876] [4,876] Army 0.103 ** 0.095 * 0.008 -0.008 0.017 (0.044) (0.049) (0.034) (0.018) (0.032) [467] [467] [467] [467] [467] Marines 0.018 0.029 0.033 0.011 0.029 (0.041) (0.048) (0.020) (0.014) (0.028) [552] [552] [552] [552] [552] Navy -0.011 0.046 0.034 ** 0.025 ** 0.039 *** (0.040) (0.039) (0.016) (0.001) (0.015) [859] [859] [859] [859] [859] Air 0.058 ** 0.026 0.014 0.001 0.016 force (0.029) (0.035) (0.013) (0.012) (0.012) [1,204] [1,204] [1,204] [1,204] [1,204] Notes: Bootstrapped standard errors are in parentheses. Number of observations is in brackets. All models use the full set of controls shown in Table A2. * Significant at 10%; ** significant at 5%; *** significant at 1% levels. TABLE 6 Exploration of Whether Psychological Stress Mediates the Relationship between Combat Exposure and Risky Behaviors in NLSAAH (1) (2) (3) (4) (5) Binge Any Other Smoking Drinking Drug Marijuana Drug Panel A: without controlling for stress variables Combat exposure 0.106 ** 0.026 0.067 * 0.039 0.034 (0.042) (0.043) (0.038) (0.038) (0.021) [561] 1553] [563] [559] [563] Panel B: controlling for stress variables Combat exposure 0.086 * 0.024 0.044 0.023 0.021 (0.045) (0.042) (0.037) (0.037) (0.021) PTSD 0.108 * 0.005 0.111 * 0.083 0.047 (0.065) (0.079) (0.063) (0.058) (0.046) Suicide 0.088 0.059 0.059 0.064 0.009 ideation (0.078) (0.093) (0.069) (0.065) (0.048) Stress 0.001 -0.003 0.008 0.003 0.008 ** (0.008) (0.006) (0.005) (0.005) (0.003) [561] [553] [563] [559] [563] Notes: Robust standard errors corrected for clustering on the school are in parentheses. Number of observations is in brackets. All models use the full set of controls shown in Table A1 along with predeployment risky behaviors. In all models, military rank, timing of military service, branch of service, occupation indicators, and an indicator for having a check-up in the past year are controlled for. Models also include missing dummy categories for each of the control variables. * Significant at 10%; ** significant at 5%; *** significant at 1% levels. TABLE 7 Exploration of Whether Psychological Stress Mediates the Relationship between Combat Exposure and Risky Behaviors Using PSM in DOD HRB Survey (1) (2) (3) Smoking Binge Drinking Any Drug Stress controls? Panel A: all No 0.036 *** 0.041 *** 0.028 *** (0.014) (0.014) (0.007) [4,876] [4,876] [4,876] Yes 0.020 0.019 0.014 ** (0.014) (0.017) (0.007) [4,779] [4,779] [4,779] Stress controls? Panel B: Army No 0.103 ** 0.095 * 0.008 (0.044) (0.049) (0.034) [467] [467] [467] Yes 0.070 0.083 0.004 (0.050) (0.059) (0.027) [460] [460] [460] Stress controls? Panel C: Marines No 0.018 0.029 0.033 (0.041) (0.048) (0.020) [552] [552] [552] Yes 0.085 0.008 0.008 (0.058) (0.051) (0.017) [517] [517] [517] Stress controls? Panel D: Navy No -0.011 0.046 0.034 ** (0.040) (0.039) (0.016) [859] [859] [859] Yes 0.000 -0.024 0.019 (0.038) (0.041) (0.019) [833] [833] [833] Stress controls? Panel E: Air Force No 0.058 ** 0.026 0.014 (0.029) (0.035) (0.013) [1,204] [1,204] [1,204] Yes 0.016 0.000 0.001 (0.035) (0.035) (0.009) [1,216] [1,216] [1,216] (4) (5) Marijuana Other Drug Stress controls? Panel A: all No 0.013 *** 0.029 *** (0.003) (0.007) [4,876] [4,876] Yes 0.008 * 0.014 ** (0.004) (0.007) [4,779] [4,779] Stress controls? Panel B: Army No -0.008 0.017 (0.018) (0.032) [467] [467] Yes 0.013 0.000 (0.018) (0.026) [460] [460] Stress controls? Panel C: Marines No 0.011 0.029 (0.014) (0.028) [552] [552] Yes 0.008 0.000 (0.017) (0.018) [517] [517] Stress controls? Panel D: Navy No 0.025 ** 0.039 *** (0.001) (0.015) [859] [859] Yes 0.019 * 0.022 (0.011) (0.018) [833] [833] Stress controls? Panel E: Air Force No 0.001 0.016 (0.012) (0.012) [1,204] [1,204] Yes -0.003 -0.003 (0.006) (0.008) [1,216] [1,216] Notes: Bootstrapped standard errors are in parentheses. Number of observations is in brackets. All models use the full set of controls shown in Table A2. * Significant at 10%; ** significant at 5%; *** significant at 1% levels. TABLE 8 The Effect of Combat Zone Deployment Length on Health Behaviors in NLSAAH (1) (2) (3) Smoking Binge Drinking Any Drug Panel A: without controlling for stress variables Combat zone service length: 0.094 * 0.024 0.064 1-6 months (0.054) (0.055) (0.042) Combat zone service length: 0.042 0.045 0.077 * 7-12 months (0.060) (0.060) (0.045) Combat zone service length: 0.121 ** 0.008 0.058 1-2 years (0.051) (0.058) (0.050) Combat zone service length: 0.040 0.126 * 0.051 more than 2 years (0.085) (0.068) (0.063) [563] [554] [565] Panel B: controlling for stress variables Combat zone service length: 0.088 0.025 0.054 1-6 months (0.054) (0.056) (0.043) Combat zone service length: 0.027 0.048 0.053 7-12 months (0.060) (0.061) (0.047) Combat zone service length: 0.Q99 * 0.012 0.025 1-2 years (0.053) (0.056) (0.050) Combat zone service length: 0.029 0.129 * 0.030 more than 2 years (0.088) (0.070) (0.066) [563] [554] [565] (4) (5) Marijuana Other Drug Panel A: without controlling for stress variables Combat zone service length: 0.058 0.019 1-6 months (0.042) (0.021) Combat zone service length: 0.058 0.046 7-12 months (0.047) (0.028) Combat zone service length: 0.051 0.004 1-2 years (0.046) (0.024) Combat zone service length: 0.031 0.006 more than 2 years (0.059) (0.028) [560] [565] Panel B: controlling for stress variables Combat zone service length: 0.052 0.011 1-6 months (0.043) (0.021) Combat zone service length: 0.044 0.028 7-12 months (0.049) (0.028) Combat zone service length: 0.033 -0.016 1-2 years (0.045) (0.029) Combat zone service length: 0.022 -0.009 more than 2 years (0.062) (0.027) [560] [565] Notes: Robust standard errors corrected for clustering on the school are in parentheses. Number of observations is in brackets. All models use the full set of controls shown in Table A1 along with predeployment risky behaviors. In all models, military rank, timing of military service, branch of service, occupation indicators, and an indicator for having a check-up in the past year are controlled for. Models also include missing dummy categories for each of the control variables. * Significant at 10%; ** significant at 5%; *** significant at 1% levels. TABLE 9 Estimates of the Effect of Combat Harm on Risky Behaviors, NLSAAH Data (1) (2) (3) (4) (5) Binge Smoking Drinking Any Drug Marijuana Other Drug Panel A: without controlling for stress variables Combat 0.129 0.028 0.003 0.001 0.012 harm (0.046) (0.047) (0.035) (0.033) (0.021) [555] [547] [557] [553] [557] Panel B: controlling for stress variables Combat 0.107 ** 0.027 -0.031 -0.022 -0.006 harm (0.049) (0.048) (0.035) (0.034) (0.021) PTSD 0.085 0.008 0.125 ** 0.090 0.054 (0.064) (0.079) (0.062) (0.057) (0.047) Suicide 0.074 0.054 0.062 0.062 0.012 ideation (0.078) (0.092) (0.068) (0.065) (0.048) Stress 0.002 -0.003 0.009 0.004 0.008 ** (0.008) (0.007) (0.005) (0.005) (0.004) [555] [547] [557] [553] [557] Notes: Robust standard errors corrected for clustering on the school are in parentheses. Number of observations is in brackets. All models use the full set of controls shown in Table A1 along with predeployment risky behaviors. In all models, military rank, timing of military service, branch of service, occupation indicators, and an indicator for having a check-up in the past year are controlled for. Models also include missing dummy categories for each of the control variables. * Significant at 10%; ** significant at 5%; *** significant at 1% levels. TABLE 10 PSM Estimates of the Effect of Combat Harm on Risky Behaviors, DOD HRB Data (1) (2) (3) Smoking Binge Drinking Any Drug Stress controls? Panel A: all No 0.051 ** 0.088 *** 0.088 *** (0.020) (0.020) (0.016) [2,484] [2,484] [2,484] Yes 0.013 0.075 *** 0.069 *** (0.019) (0.025) (0.025) [2,336] [2,336] [2,336] Stress controls? Panel B: Army No 0.046 0.091 * 0.097 *** (0.052) (0.049) (0.037) [659] [659] [659] Yes 0.061 0.046 0.061 (0.048) (0.048) (0.042) [564] [564] [564] Stress controls? Panel C: Marines No 0.054 0.058 0.090 *** (0.044) (0.050) (0.033) [557] [557] [557] Yes 0.032 0.089 0.041 (0.055) (0.059) (0.041) [495] [495] [495] Stress controls? Panel D: Navy No -0.015 0.022 0.088 * (0.060) (0.070) (0.052) [275] [275] [275] Yes 0.000 0.016 0.047 (0.057) (0.061) (0.049) [249] [249] [249] Stress controls? Panel E: Air Force No 0.120 *** 0.054 0.012 (0.045) (0.046) (0.033) [339] [339] [339] Yes -0.013 -0.033 0.000 (0.055) (0.063) (0.032) [305] [305] [305] (4) (5) Marijuana Other Drug Stress controls? Panel A: all No 0.048 *** 0.083 *** (0.007) (0.016) [2,484] [2,484] Yes 0.025 *** 0.069 *** (0.007) (0.016) [2,336] [2,336] Stress controls? Panel B: Army No 0.012 0.097 *** (0.012) (0.036) [659] [659] Yes 0.014 0.064 (0.015) (0.040) [564] [564] Stress controls? Panel C: Marines No 0.050 *** 0.090 *** (0.017) (0.090) [557] [557] Yes 0.032 ** 0.044 (0.014) (0.041) [495] [495] Stress controls? Panel D: Navy No 0.095 *** 0.088 * (0.028) (0.052) [275] [275] Yes 0.071 *** 0.040 (0.023) (0.049) [249] [249] Stress controls? Panel E: Air Force No 0.036 * 0.060 (0.019) (0.029) [339] [339] Yes 0.007 -0.007 (0.017) (0.031) [305] [305] Notes: Bootstrapped standard errors are in parentheses. Number of observations is in brackets. All models use the full set of controls shown in Table A2. * Significant at 10%; ** significant at 5%; *** significant at 1% levels. TABLE 11 PSM Estimates of the Effect of Combat Harm on Risky Behaviors among Those Assigned to Combat with Firelight, DOD HRB Data (1) (2) (3) Smoking Binge Drinking Any Drug Stress controls? Panel A: all No 0.024 0.099 *** 0.083 *** (0.025) (0.027) (0.018) [1,945] [1,945] [1,945] Yes 0.001 0.084 *** 0.045 ** (0.024) (0.027) (0.020) [1,863] [1,863] [1,863] Stress controls? Panel B: Army No 0.039 0.111 ** 0.105 *** (0.052) (0.053) (0.038) [571] [571] [571] Yes 0.041 0.058 0.021 (0.063) (0.059) (0.040) [487] [487] [487] Stress controls? Panel C: Marines No -0.004 0.066 0.079 ** (0.047) (0.056) (0.036) [488] [488] [488] Yes 0.016 0.057 0.063 (0.056) (0.059) (0.046) [380] [380] [380] Stress controls? Pane1 D: Navy No 0.136 -0.045 -0.045 (0.152) (0.163) (0.111) [87] [87] [87] Yes 0.000 -0.025 0.025 (0.209) (0.199) (0.122) [78] [78] [78] Stress controls? Panel E: Air Force No -0.009 -0.029 -0.039 (0.082) (0.092) (0.057) [199] [199] [199] Yes 0.089 0.100 -0.044 (0.097) (0.102) (0.047) [197] [197] [197] (4) (5) Marijuana Other Drug Stress controls? Panel A: all No 0.033 *** 0.077 *** (0.009) (0.018) [1,945] [1,945] Yes 0.011 0.045 ** (0.007) (0.020) [1,863] [1,863] Stress controls? Panel B: Army No 0.007 0.101 *** (0.018) (0.037) [571] [571] Yes 0.008 0.025 (0.014) (0.040) [487] [487] Stress controls? Panel C: Marines No 0.050 *** 0.075 ** (0.016) (0.034) [488] [488] Yes 0.021 0.063 (0.019) (0.046) [380] [380] Stress controls? Pane1 D: Navy No 0.045 -0.045 (0.069) (0.110) [87] [87] Yes 0.051 0.000 (0.038) 0.127 [78] [78] Stress controls? Panel E: Air Force No 0.000 -0.029 (0.023) (0.057) [199] [199] Yes 0.000 -0.044 (0.017) (0.049) [197] [197] Notes: Bootstrapped standard errors are in parentheses. Number of observations is in brackets. All models use the full set of controls shown in Table A2. * Significant at 10%; ** significant at 5%; *** significant at 1% levels.