soc paper summary

T

R a

b

c

a

A R R A A

J H I

K M D P S

1

d

J m 2 P S a a J P w R s C U

(

0 h

Journal of Health Economics 32 (2013) 51– 65

Contents lists available at SciVerse ScienceDirect

Journal of Health Economics

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / e c o n b a s e

he psychological costs of war: Military combat and mental health�

esul Cesur a, Joseph J. Sabia b, Erdal Tekin c,∗

University of Connecticut, Finance Department, 2100 Hillside Road Unit 1041, Storrs, CT 06269-1041, United States San Diego State University & U.S. Military Academy, Department of Economics, 5500 Campanile Drive, San Diego, CA 92182-4485, United States Georgia State University, IZA and NBER, Department of Economics, Andrew Young School of Policy Studies, United States

r t i c l e i n f o

rticle history: eceived 13 June 2011 eceived in revised form 1 September 2012 ccepted 14 September 2012 vailable online 24 September 2012

EL classification: 56

a b s t r a c t

We exploit plausibly exogenous variation in overseas deployment assignment to estimate the effect of combat exposure on psychological well-being. Controlling for pre-deployment mental health, we find that active-duty soldiers deployed to combat zones are more likely to suffer from post-traumatic stress disorder (PTSD) than their counterparts deployed outside the United States in non-combat zones. Among those deployed to combat zones, those deployed to locales where they engage in enemy firefight or witness allied or civilian deaths are at an increased risk for suicidal ideation and PTSD relative to their active-duty counterparts deployed to combat zones without enemy firefight.

1

eywords: ilitary combat epression

© 2012 Elsevier B.V. All rights reserved.

a n m I 2 c

ost-traumatic stress disorder uicide

“This is the price of war. . ..You can’t send young Americans to Iraq and Afghanistan . . . and expect them to come home and just fit right in. They bring that trauma with them.”

—Max Cleland, Vietnam War Veteran and former U.S. Senator.

. Introduction

The mental health impairments experienced by U.S. soldiers eployed in the Global War on Terrorism (GWOT) have received

� The authors thank John Z. Smith, David Lyle, Daniel I. Rees, Solomon W. Polachek, effrey S. DeSimone, and participants at the 2012 American Economic Association

eetings, the 2011 International Health Economics Association World Congress, the 011 European Society for Population Economics meetings, the 2011 Association for ublic Policy Analysis and Management, the University of Connecticut, and the 2010 outhern Economic Association meetings for useful comments and suggestions on n earlier draft of this paper. Thanks also to Whitney Dudley for excellent research ssistance. This research uses data from Add Health, a program project designed by . Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and funded by a grant 01-HD31921 from the National Institute of Child Health and Human Development, ith cooperative funding from 17 other agencies. Special acknowledgment is due onald R. Rindfuss and Barbara Entwisle for assistance in the original design. Per- ons interested in obtaining data files from Add Health should contact Add Health, arolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 27516-2524, nited States (http://www.cpc.unc.edu/addhealth/contract.html). ∗ Corresponding author. Tel.: +1 404 413 0163; fax: +1 404 413 0145.

E-mail addresses: [email protected] (R. Cesur), [email protected] J.J. Sabia), [email protected] (E. Tekin).

a m 4 C 2 h o p A ( p B

e 2 a i w r

167-6296/$ – see front matter © 2012 Elsevier B.V. All rights reserved. ttp://dx.doi.org/10.1016/j.jhealeco.2012.09.001

great deal of attention by both policymakers and the American ews media. A recent article in the Time magazine describes the ental health problems of servicemen and women returning from

raq and Afghanistan as “the U.S. Army’s third front” (Thompson, 010).1 A recent study by the Rand Corporation finds that 26 per- ent of soldiers serving in the GWOT suffer from depression, drug nd alcohol dependency, homelessness, or suicidal ideation; esti- ates of post-traumatic stress disorder (PTSD) range from 4 to

5 percent (Tanielian and Jaycox, 2008). The Centers for Disease ontrol and Prevention estimates that veterans comprise nearly 0 percent of the more than 30,000 suicides each year. A record igh 38 soldiers committed suicide in July 2012 (U.S. Department f Defense, 2012) and as of June 2012, more U.S. military service ersonnel have lost their lives by suicide (2676) since the war in fghanistan began than have died during combat operations there

1950) (O’Gorman, 2012). Public concern about the mental health roblems of soldiers has prompted political action, with President arack Obama announcing a plan to increase the ease with which

1 Military service has been linked to greater take-up of disability benefits (Autor t al., 2011; Angrist et al., 2010), as well as higher rates of crime and violence (Rohlfs, 010). Angrist et al. (2010) find greater disability take-up among Vietnam veter- ns with low earning potential. Autor et al. (2011) also find evidence of a recent ncrease in disability uptake among Vietnam veterans but are unable to distinguish

hether this effect is driven by a long-term adverse health effect of combat or a ecent liberalization in benefit rules.

5 ealth

v r

p O 2 J w c e e w t n d t ( M d 2 n h

v d c o t L d H z s

a e f h o r r s m s e d

2

c c P B a s t I

a

o fi

p e h t l t h e i p b w

h W i ( 2 U ( m i

s d b d o s n W m T w a ( t P G a m i t 2 h

t c w m bat versus non-combat zone, engaging the enemy in firefight, and killing or wounding someone—on psychological well-being. More- over, because of the longitudinal nature of our data, we will be able

2 R. Cesur et al. / Journal of H

eterans diagnosed with post-traumatic stress disorder (PTSD) can eceive federal health benefits (Obama, 2010).

While a number of studies have found that PTSD is a growing roblem for U.S. soldiers deployed in Operation Iraqi Freedom and peration Enduring Freedom (Hoge et al., 2006, 2004; Erbes et al., 007; Rosenheck and Fontana, 2007; Seal et al., 2007; Tanielian and aycox, 2008), much of this work has been descriptive in nature

ithout a counterfactual control group.2 More recent studies have ompared PTSD rates among deployers and non-deployers (Shen t al., 2010), including National Guardsmen and Reservists (Smith t al., 2008; Wells et al., 2010) and have found that deployers are in orse mental health than their non-deploying counterparts. While

he results from these studies are informative, it is not clear that on-deployers comprise an appropriate control group for those eployed to combat. Selection into active-duty service, as opposed o service in the National Guard or Reserves, may be non-random see, for example, Hirsch and Mehay, 2003, for a discussion).

oreover, while timing of deployment among active-duty sol- iers may be arguably exogenous in the short-run (Engel et al., 010; Lyle, 2006), soldiers who experience periods of extended on-deployment may be non-deployable due to physical or mental ealth conditions (Department of the Army AR 614-30, 2010).

Our study focuses attention on deployed active-duty ser- icemen and exploits plausibly exogenous variation in overseas eployment assignment to identify the mental health effects of ombat. We rely on evidence that overseas deployment assignment f active-duty units within and across combat zones is unrelated to he characteristics of soldiers or their families (Engel et al., 2010; yle, 2006) to identify the causal effect of combat exposure. Using ata drawn from the National Longitudinal Study of Adolescent ealth, we find that active-duty U.S. soldiers deployed to combat ones are at greater risk of PTSD than their active-duty counterparts erving outside the United States in non-combat zones.

Among those deployed to combat zones, we find that soldiers ssigned to locales where they are exposed to violent combat vents such as frequent enemy firefight are at the greatest risk or psychological harm. Soldiers who kill someone (or believe they ave killed someone), are injured in combat, or witness the death r wounding of a civilian or coalition member are at an increased isk of suicidal ideation, depressive symptomatology, and PTSD elative to their counterparts deployed to combat zones without uch exposure. Our findings further suggest that military policy- akers crafting optimal deployment schedules that account for

oldiers’ mental health should focus greater attention on soldiers’ xperiences with frequent enemy firefight along with cumulative eployment length.

. Background

Early studies examining the health effects of military service ompared the health outcomes of those in combat with those of ivilians (Jordan et al., 1991; Iowa Persian Gulf Study Group, 1997; rice et al., 2004; Card, 1987; McKinney et al., 1997; Kang and ullman, 2001).3 But, as Dobkin and Shabani (2009) note, the aver- ge individual and family background characteristics of active duty

ervicemen are quite different from those of civilians, and many of hese characteristics are also related to psychological well-being. f, for example, socioeconomic status is negatively related to the

2 See also the Mental Health Advisory Team reports at: http://www. rmymedicine.army.mil/news/releases/20080306mhatv.cfm. 3 A related literature on civilians has examined the mental health consequences

f stressful domestic occupations such as police work (Wang et al., 2010) and refighting (Bryant and Guthrie, 2007; Heinrichs et al., 2005).

t

a ( i

e a e s

Economics 32 (2013) 51– 65

robability of joining the armed forces (Segal et al., 1998; Bachman t al., 2000; Kleykamp, 2006) and positively related to mental ealth (Miech et al., 1999), then members of military may be prone o mental health problems even in the absence of military service, eading to overstated estimates of the cost of combat service.4 On he other hand, because military personnel go through a rigorous ealth screening prior to induction or commissioning (see, for xample, Department of Defense Directives 6130.3 and 6130.4), ndividuals who serve in the military may not only be in better hysical health than their civilian counterparts, but they may also e in better mental health as well. Each of these forms of selection ould tend to understate the estimated effects of combat service.

More convincing studies of the health effects of military service ave focused on service in the Second World War, the Korean ar, and the Vietnam War, and have addressed the endogene-

ty of military service by using the draft lottery as an instrument Angrist et al., 2010; Hearst et al., 1986; Bedard and Deschenes, 004; Dobkin and Shabani, 2009; Edwards and MacLean, 2010). sing this approach, Angrist et al. (2010) and Dobkin and Shabani

2009) find evidence that prior estimates of the health effects of ilitary service were overstated. Still, the results suggest that mil-

tary service adversely affect health.5

The absence of a draft in the post-Vietnam era does not allow uch an identification approach to study the effect of randomly rawing a civilian for service in the GWOT. Instead, the focus has een on a different local average treatment effect: the effect of rawing a military servicemember for combat service. A handful f recent studies have examined the relationship between combat ervice and mental health during the Global War on Terrorism using on-deployed servicemen as a control group (Shen et al., 2010; ells et al., 2010; Smith et al., 2008). Shen et al. (2010) link TRICARE edical records to deployment information from the Contingent

racking System and find that deployment to Iraq and Afghanistan as associated with a substantial increase in the risk of PTSD rel-

tive to non-deployed soldiers. Wells et al. (2010) and Smith et al. 2008) use data from the Millennium Cohort Study and find that hose exposed to combat are at greater risk of depression and TSD than their non-deployed counterparts (including National uardsmen and Reservists). While these findings are interesting nd informative, non-deployed Reservists or National Guards- en and active-duty non-deployable soldiers may differ on many

ndividual-level unobservables related to health (Department of he Army AR 614-30, 2010) or human capital (Hirsch and Mehay, 003), thus potentially contaminating estimates of the mental ealth effects of combat.6

The current study exploits plausibly exogenous variation in he location of overseas deployment assignment to identify the ausal effect of combat exposure in the GWOT on psychological ell-being. Specifically, we exploit variation in exposure to several easures of combat exposure—including assignment to a com-

4 However, recent descriptive work by National Priorities Project (2008) suggests hat negative selection on socioeconomic status may not be as severe today.

5 For instance, Hearst et al. (1986) find that draft exposure was associated with n increased risk of suicide and automobile accidents and Bedard and Deschenes 2004) find that veterans of World War II and the Korean War were at substantial ncreased risk of mortality due to military-induced smoking.

6 Wells et al. (2010) and Smith et al. (2008) are able to separately estimate the ffects of combat versus non-combat deployment with their data, but do not provide

statistical test of whether they differ, providing only tests of how the effects of ach type of deployment compare to non-deployed active duty and non-active duty oldiers. Jacobson et al. (2008) makes similar comparisons examining alcohol use.

ealth

t t i

3

n b o s c i d c fi 1 o f t d

( w v i d s U o o b l p s m d

u a r n G a s

t c

A T s

R

t H I D

t C s s m

i (

r d a A r s ( s 9

e i w w m i i o t i e w k a a

b h s i h f a t a A c o h o

R. Cesur et al. / Journal of H

o test the plausibility of our identification strategy, which assumes hat deployment assignment is unrelated to prior mental health or ndividual and family background characteristics.7

. Data and measures

The data used in this study come from the National Longitudi- al Study of Adolescent Health (Add Health), which was conducted y the Carolina Population Center at the University of North Car- lina at Chapel Hill. The Add Health is a nationally representative chool-based longitudinal study that began surveying U.S. adoles- ents in seventh to twelfth grades in the mid-1990s. The Wave I n-home baseline survey was administered to 20,745 respondents uring the 1994–1995 academic year. Three follow-ups have been onducted since the original Add Health data collection effort. The rst follow-up, the Wave II in-home survey, was conducted in 996, approximately one year after the baseline survey; the sec- nd follow-up (Wave III) was administered in 2001, and the third ollow-up (Wave IV), was administered in 2007–2008 to 15,701 of he original Add Health participants (see Harris et al., 2008 for more etailed information on the Add Health data collection strategy).

The Add Health dataset is useful for our purposes because it i) contains a relatively large sample of military servicemen and omen whose military service started after the Wave I8 sur-

ey (N = 1080) at the time of the Wave IV survey, (ii) provides nformation on whether active-duty servicemen and women were eployed to a combat zone (N = 429), a non-combat zone out- ide of the United States (N = 149), or served on active-duty in the nited States exclusively (N = 338),9 and (iii) includes information n exposure to specific violent combat events, such as frequency f enemy firefight, unavailable in many other datasets. Moreover, ecause the survey is longitudinal in nature and spans back to ado-

escence, we have information on the respondent’s mental health rior to any military deployment. In our analysis, we restrict our ample to respondents who provided non-missing information on ental health and military service at Wave IV when the respon-

ents were ages 24–33. We measure military service in the United States Armed Forces

sing respondents’ reports of active duty service and deployment ssignment at Wave IV.10 In our sample, 5.85 percent (N = 916) eported active duty military service and 1.04 percent reported on-active duty service exclusively in the Reserves or National

uard (N = 164).11 Approximately 81 percent of those who reported ny military service served during the GWOT, while the remainder erved exclusively during the late 1990s when the U.S. was engaged

7 Wells et al. (2010) also use longitudinal data, but exclude from their sample hose with depression at baseline; Smith et al. (2008) employ a similar baseline utoff in examining the outcome of PTSD.

8 Including those whose military service started prior to Wave I interview, the dd Health Wave IV survey contains 1100 individuals who served in the military. he results are quantitatively and qualitatively similar when those whose military ervice started prior to the Wave I interviews are included.

9 The remaining 164 individuals are in non-active duty service exclusively in the eserves or National Guard. 10 Service in the Armed Forces was measured using the following Wave IV ques- ionnaire items: ave you ever served in the military (Possible answers: Yes, No)

n which components of the military have you served? (Possible answers: Active uty, Reserves, National Guard, None)

11 The weighted means for military service in the 2008 Add Health are comparable o weighted means in the 2008 Current Population Survey and the 2008 American ommunity Survey. For instance, in the 2008 ACS, 5.0 percent reported active duty ervice, as compared to 5.87 percent in Add Health. The unweighted means are lightly larger in the Add Health due to an oversample of racial minorities, who are ore likely to serve in the military.

t t t c

i t s n g b 2 b

w W W T w

m

Economics 32 (2013) 51– 65 53

n military operations in the Balkans, including the Kosovo War Operation Allied Force).

Importantly, the data allow us to distinguish between those who eported active duty service exclusively in the United States, active uty service outside the United States in non-combat zones, and ctive duty service outside the United States in combat zones.12

mong those who report active-duty service, 37 percent (N = 338) eport service exclusively in the United States, 16 percent report ervice overseas in a non-combat zone (N = 149), and 47.0 percent N = 429) report deployment in a combat zone. Among those who erved in combat, 93.7 percent reported combat service in the post- /11 period.13

We are also able to measure self-reports of violent combat vents experienced by those deployed to combat zones, includ- ng the number of “times [they] engaged the enemy in a firefight,”

hether they “ever kill[ed] or think [they] killed someone,” hether they were “wounded or injured” during combat deploy- ent, and whether they saw “anyone wounded, killed, or dead,

ncluding ‘coalition or ally,’ ‘enemy,’ or ‘civilian.”’ In models explor- ng the differential effects of combat zone experiences, we restrict ur sample to those for whom there is information on whether hey engaged in enemy firefight (N = 416). Among those who served n a combat zone with information on whether they engaged in nemy firefight (N = 416), the average number of enemy firefights as 16; 35.9 percent (N = 146) had killed or believed they had

illed someone, 11.5 percent (N = 48) were wounded or injured, nd 64.9 percent (N = 270) witnessed the death or wounding of an lly, enemy, or civilian.

In the empirical analyses below, we examine the relationship etween the above measures of military service and three mental ealth outcomes measured at Wave IV: suicide ideation, depressive ymptomatology, and PTSD. While the majority of previous stud- es on the relationship between military service and mental health ave tended to focus on the outcomes of PTSD and depression,

ewer have examined suicide ideation. This may be due to avail- bility of clinical data on these disorders, while data on suicidal houghts are typically self-reported. According to the Diagnostic nd Statistical Manual of Mental Disorders, 4th edition (DSM-IV) of merican Psychiatric Association, suicide ideation does not alone onstitute a criterion for formal diagnosis of a mental health dis- rder, but it is rather considered one of the symptoms of a mental ealth problem. Our first outcome measure, Suicide, is an indicator f suicidal ideation created using respondents reports of whether hey had “ever seriously thought about committing suicide during he past 12 months.” Respondents who answered in the affirma- ive were coded as 1 and those who answered in the negative were oded as 0.

It is important to acknowledge that self-reported suicide deation is a noisy indicator of mental health problems rather han a proxy for completed suicides. While a suicide attempt is a trong predictor of completed suicide and is an indicator of sig- ificant mental distress (both among military veterans and the eneral population), the empirical evidence is less clear on the link

etween suicide ideation and completed suicides (Weiner et al., 011; Suominen et al., 2004), in part, because initiating action may e an important intervening variable. Thus, caution should be taken

12 The Wave IV Add Health questionnaire items used to obtain these measures ere: as your military service in the US, outside the US, or both? hat is the total amount of time you (have) served in a combat zone?

he Add Health does not contain information on the country to which the soldier as deployed.

13 We identified those who served in the post-9/11 period as those whose first ilitary service started after September 2001.

5 ealth

i s

o R m w i d h m P ( e w s i s T e i r o

S t s o p t n

d m c s o e e i t u 2 i a w t c f

m t c d

i t c r e T u w r

t s c

w s i a r e ( d s d T t m e

a i o b i t a s t

h e h i i ( D p

4

o t p s i ( s l a H

4 R. Cesur et al. / Journal of H

n the interpretation of suicide ideation impacts of combat expo- ure.

Second, we use an abridged version of the Center for Epidemi- logical Studies-Depression (CES-D) scale, originally developed by adloff (1977) and used widely as a measure of depressive sympto- atology. Respondents were instructed to indicate the frequency ith which they had experienced certain feelings or emotions dur-

ng the past week, including being “bothered by things that usually on’t bother you,” being unable to “shake off the blues, even with elp from your family and friends,” having “trouble keeping your ind on what you were doing,” or feeling “depressed” or “sad.”

ossible responses, which included “rarely or none of the time” =0); “some or a little of the time” (=1); “occasionally or a mod- rate amount of the time” (=2); and “most or all of the time” (=3) ere summed to produce a score of between 0 and 15. From this

core, we defined an individual as Depressed if he or she is ranked n the top quintile of the distribution of CES-D scale, following a trategy employed by a number of researchers (Tekin et al., 2009; ekin and Markowitz, 2008; Chatterji and Cuellar, 2006; Hallfors t al., 2004; Wickrama et al., 2010). An advantage of dichotomiz- ng the s score in this manner is that it focuses attention on the ight-hand tail of the CES-D distribution, where medical diagnoses f major depression are made.14

Third, following much of the PTSD literature (Smith et al., 2008; hen et al., 2009, 2010), we create an indicator, PTSD, for whether he respondent had received a medical diagnosis of post-traumatic tress disorder using responses to the question, “Has a doctor, nurse r other health care provider ever told you that you have or had ost-traumatic stress disorder?” Respondents who answered in he affirmative were coded as 1 and those who answered in the egative as 0.

Our final measure of mental health is useful because it involves octor treatment or diagnosis of mental health conditions, which ay measure psychological problems that are not adequately

aptured by a self-reported suicide measure or abridged CES-D cale. According to the fourth Diagnostic and Statistical Manual f Mental Disorders (2000), a positive PTSD diagnosis requires xposure to a traumatic event where the individual “has experi- nced, witnessed, or been confronted with an event or events that nvolve actual or threatened death or serious injury, or a threat o the physical integrity of oneself or others” and the individ- al’s response includes “intense fear, helplessness, or horror.” (DSM 000) In addition to these criteria, individuals are assessed to have

ntrusive recollections, avoidant/numbing symptoms, and hyper- rousal symptoms.” (DSM 2000). One limitation to the measure e employ, however, is that we cannot measure the precise dura-

ion of symptoms for those with a positive PTSD screening, nor an we assess how the symptoms affect the physical and cognitive unctioning of individuals.

However, using a medical diagnosis of PTSD does raise a few easurement issues. Because our data only permit us to identify

hose cases of PTSD that were diagnosed, our estimate may not apture the true prevalence of PTSD for a number of reasons. First, a iagnosis could be related to the persistence of an individual soldier

14 Note that we tested the sensitivity of our results to alternative ways of creat- ng the depression indicator. For example, we transformed our abridged version of he CES-D measure available in Wave IV into binary indicators based on the clini- al diagnostic cutoffs of 16 and 23 to indicate “possible” and “probably” depression, espectively (Orr et al., 2005; Milette et al., 2010). Estimation of the depression mod- ls using these cut-offs did not cause appreciable changes to the results presented in ables 3–5. These results are available from the authors upon request. Furthermore, sing the cutoff preferred by Sabia and Rees (2008) and Duncan and Rees (2005) as ell as a self-reported medical diagnostic measure produces qualitatively similar

esults.

p a d a c F n d m s

P

Economics 32 (2013) 51– 65

o obtain care. So, for example, to the extent that those with more evere cases of PTSD are those who obtain treatment, our measure ould understate its true prevalence.

Second, using a medical diagnosis measure raises a concern that e may confound mental health conditions with access to health

ervices or screenings. While we control for health insurance status n all models, mental health screenings are increasingly employed mong active-duty personnel. For instance, service members are equired to complete a health assessment, which includes an valuation of mental health, both before and after deployment Department of Defense Form 2795). While we do not have a irect measure of access to screenings, for our primary analy- is, our treatment and comparison groups are each comprised of eployed active duty personnel or, in some cases, combat soldiers. o the extent that active duty soldiers have comparable access o health care services and screenings, such comparisons should

itigate concerns that we are picking up a unique “screening ffect.”

The means and standard deviations of our dependent variables re presented in Panel A of Table 1 by various measures of mil- tary service experience. The first four columns in Table 1 focus n mental health outcomes measured at Wave IV. A comparison etween rows 2 and 9 suggests that at Wave IV, those who served

n the military are in poorer mental health than their civilian coun- erparts with the exception of depression. We continue to observe

similar pattern when we examine sub-samples of military per- onnel. Moreover, the magnitudes of the differences are larger for hose in more stressful missions, such as combat duty.

The remaining columns in Table 1 show the means of mental ealth measures at Wave I, prior to any military service.15 Inter- stingly, we find that those who serve in combat zones later in life ave mental health outcomes in adolescence that are no worse, and

n some cases better, than their counterparts who remain in civil- an life. These findings are consistent with Department of Defense DOD) enlistment standards described in DOD Directive 6130.3 and OD Instruction 6130.4, which include screening for mental health roblems such as depression and anxiety disorders.

. Identification

Researchers in the post-draft era have identified a new source f plausibly exogenous variation in combat exposure: the U.S. mili- ary’s deployment assignment procedures. Two recent studies have ersuasively argued that deployment assignment is exogenous to oldiers’ preferences, welfare, and family background character- stics (Lyle, 2006; Engel et al., 2010). For example, Engel et al. 2010) note that the U.S. Army almost never deploys individual oldiers, but rather deploys companies. An individual soldier has ittle control over the company to which he or she is assigned and, s matter of policy, is reassigned every three or four years by Army uman Resources Command (HRC). Lyle (2006) states that for the urposes of assignments, HRC “regards soldiers of the same rank nd occupation as equals.” (p. 323) The timing and location of unit eployment assignments depend not on individual soldiers’ char- cteristics such as bravery, mental toughness, or family background haracteristics, but rather on the operational needs of the Armed orces, which is determined by exogenous events, and on the readi- ess and availability of the unit (Engel et al., 2010). Unit readiness

epends on unit macro-level issues such as the timeliness of equip- ent being inventoried and cleared for shipment, completion of

pecified training, and occupational skill set of unit members (Army

15 Diagnosis of PTSD is not measured at Wave I; thus, this outcome is omitted from anel B of Table 1.

R. Cesur et al. / Journal of Health Economics 32 (2013) 51– 65 55

Table 1 Descriptive statistics of mental health outcomes by military service.

Sample N Wave IV Wave I

Suicide Depression PTSD Suicide Depression

Full sample 15,669 0.067 0.192 0.029 0.138 0.208 (0.250) (0.394) (0.167) (0.345) (0.406)

Military sample 1080 0.075 0.173 0.089 0.125 0.158 (0.264) (0.379) (0.285) (0.331) (0.365)

Active duty combat zone 429a 0.072 0.173 0.164 0.110 0.152 (0.259) (0.378) (0.370) (0.314) (0.359)

Combat zone with enemy firefight 185 0.103 0.205 0.250 0.109 0.130 (0.304) (0.405) (0.434) (0.312) (0.337)

Combat zone without enemy firefight 231 0.043 0.139 0.087 0.109 0.165 (0.204) (0.346) (0.282) (0.313) (0.372)

Active duty non-combat zone overseas 149 0.067 0.168 0.020 0.094 0.142 (0.251) (0.375) (0.141) (0.293) (0.350)

Active duty non-deployed 338 0.086 0.189 0.039 0.159 0.169 (0.281) (0.392) (0.193) (0.366) (0.375)

Non-active duty 164 0.068 0.146 0.061 0.122 0.165 (0.252) (0.355) (0.240) (0.328) (0.372)

Civilian sample 14,589 0.066 0.193 0.024 0.139 0.211 (0.249) (0.395) (0.154) (0.346) (0.408)

S ves I repor

R a

u g r

t e d e f d d m e d

r a r s ( t m

t c

d d d R e g a p c a ( t O i I

d c d r w i s a p o h icance tests indicate that variables in various domains, including pre-deployment mental health religion, education, are statistically significant as a group. These findings suggest that using civilians as

16 Military-specific occupations were measured with respect to the respondent’s job last year at Wave IV. The categories include Military Officer Special and Tacti-

tandard deviations are in parentheses. Unweighted means are generated using Wa a 13 active duty respondents who reported deployment to a combat zone did not

egulation 220-1). Lyle (2006) emphasizes this point in a recent rticle:

“The ‘needs of the army’. . .captures the essence of all [military] assignments: world events drive army assignments. [T]he tim- ing of the move and assignment of a soldier to a subordinate army unit are largely independent of a soldier’s preferences. . . [O]nce a soldier is assigned to a division, the division assigns the soldier to one of several brigades, the brigade assigns the soldier to one of several battalions, and the battalion assigns the soldier to one of several companies. The ‘needs of the army’ also determine the missions that a soldier’s company receives.” (Lyle, 2006, p. 323)

Engel et al. (2010) also explicitly note that senior commanders se two primary criteria in deployment assignments: “the exi- encies of the operational environment and the availability and eadiness of suitable units.” Further, they note:

“As a rule, they do not take into consideration the welfare of an individual enlisted soldier. . .nor do they consider the aver- age characteristics of units and families.” (Engel et al., 2010, p. 76)

In their studies, Lyle (2006) and Engel et al. (2010) exploit he timing of unit deployment assignments to estimate the causal ffect of deployment-induced parental absences on children’s aca- emic achievement. In addition to the theoretical case for the xogeneity of deployment assignment, these authors offer two urther empirical tests in support of this assumption. First, their ata allow them to instrument individual deployment with unit eployment. Their findings suggest that treating individual deploy- ent as exogenously determined produced quantitatively similar

stimates to those obtained when instrumenting for individual eployment.

Second, Lyle (2006) and Engel et al. (2010) use administrative ecords in their studies, which contain the full set of information vailable to HRC when deployment assignments are made. These ecords contain information on only a small set of observables

uch as the respondent’s rank, occupation, age, race, cognitive test AFQT) score, and educational attainment. Their findings suggest hat controlling for these observables has little effect on the esti-

ated effect of deployment, again consistent with the hypothesis

c V C R s

and IV of the National Longitudinal Study of Adolescent Health. t whether they experienced enemy firefight.

hat deployment assignment is exogenous to individual and family haracteristics.

While the Add Health data do not contain information on unit eployment for an instrumental variables strategy, we do have ata on a wide set of observables that include the full set of sol- ier characteristics available to U.S. military personnel at Human esources Command when making deployment decisions (Engel t al., 2010).16 These observables including age, race, ethnicity, ender, measured height, measured weight, years of schooling ttained, religious affiliation, maternal educational attainment, arental marital status when the respondent was an adoles- ent, parental income when the respondent was an adolescent, n abridged version of the Peabody Picture and Vocabulary Test PPVT), health insurance status, military rank, timing of service in he military, branch of service, and occupation (four-digit Standard ccupational Classification code). Moreover, our data also include

nformation on pre-deployment mental health, measured at Wave , when the respondents were in adolescence.

In Table 2, we examine the relationship between combat eployment and a wide set of individual and family background haracteristics. Column (1) pools the sample of civilians and those eployed to combat and regresses an indicator for whether the espondent was assigned to combat on these observables along ith the pre-deployment measures of mental health. The estimates

n this column show that civilians and military personnel who erved in combat zones differ significantly in a number of observ- ble characteristics. For example, deployment to a combat zone is ositively associated with being male and having a some college r vocational training, but negatively associated with weight and aving a graduate or professional degree. Moreover, joint signif-

al Operations Leaders/Managers, Infantry Officers, Special Forces, Armored Assault ehicle Officers, Artillery and Missile Officers, Air Crew Officers, Command and ontrol Center Officers, and First-Line Enlisted Military Supervisors/Managers, and adar and Sonar Technicians. Occupations in engineering and medicine are mea- ured in similar detail.

56 R. Cesur et al. / Journal of Health Economics 32 (2013) 51– 65

Table 2 OLS estimates of the relationship between individual and family background characteristics on probability of deployment to combat zone.

Variables (1) (2) (3) Combat versus civilians Combat versus

non-combat overseas Combat w/firefight versus combat w/out firefight

Lagged depression −0.000 0.087 −0.052 (0.004) (0.071) (0.074)

Lagged suicide 0.001 0.054 0.048 (0.003) (0.065) (0.109)

F-Test on joint significance of lagged mental health 0.0809 2.059 0.382 P-Value 0.970 0.109 0.766

Height in inches 0.000 0.004 −0.004 (0.001) (0.008) (0.011)

Weight in pounds −0.000*** −0.000 −0.000 (0.000) (0.001) (0.001)

Religion: Protestant 0.009* 0.051 0.067 (0.005) (0.065) (0.076)

Religion: Catholic 0.003 −0.012 −0.007 (0.005) (0.068) (0.088)

Religion: Other Christian −0.003 0.034 −0.002 (0.004) (0.068) (0.093)

Religion: Other −0.008 −0.148 −0.058 (0.005) (0.104) (0.135)

F-Test on joint significance of religion 3.136 1.094 1.017 P-Value 0.0102 0.367 0.411

Male 0.054*** 0.155** 0.337***

(0.005) (0.069) (0.094) Age in years 0.031 0.061 −0.616

(0.021) (0.375) (0.485) Age in years squared −0.001 −0.002 0.011

(0.000) (0.007) (0.009) Race: Black 0.011** −0.019 −0.108

(0.004) (0.056) (0.074) Race: Other 0.011 0.041 −0.213**

(0.008) (0.050) (0.097) Race: Hispanic 0.004 −0.024 −0.165**

(0.005) (0.054) (0.078)

F-Test on joint significance of race 1.801 1.186 9.536 P-Value 0.116 0.320 0.000

Education: some college or vocational training 0.025*** 0.031 −0.084 (0.004) (0.050) (0.063)

Education: college degree −0.008* 0.096 −0.118 (0.004) (0.079) (0.086)

Education: graduate or professional degree −0.021*** −0.265 0.078 (0.005) (0.166) (0.214)

F-Test on joint significance of Education 41.12 2.396 1.119 P-Value 0.000 0.0716 0.344

No health insurance −0.026*** −0.011 0.072 (0.003) (0.069) (0.084)

Wave 1 picture vocabulary test score 0.000*** −0.001 −0.003 (0.000) (0.002) (0.002)

Log of parental income Wave 1 0.002 0.048 0.043 (0.002) (0.039) (0.051)

Parent is married in Wave 1 0.002 −0.064 −0.008 (0.007) (0.099) (0.161)

Parent is divorced, separated or widowed in Wave 1 0.004 0.002 0.090 (0.007) (0.101) (0.167)

F-Test on joint significance of parental marital status 0.342 0.766 0.899 P-Value 0.795 0.515 0.444

Biological mother’s education: high school degree 0.001 −0.055 −0.043 (0.003) (0.057) (0.080)

Biological mother’s education: some college 0.006 −0.050 −0.031 (0.004) (0.064) (0.088)

Biological mother’s education: college degree or more 0.002 −0.071 0.087 (0.004) (0.064) (0.086)

Biological mother’s education: missing 0.006 0.143 −0.303 (0.017) (0.195) (0.222)

F-Test on joint significance of mother’s education 0.596 0.669 1.378 P-Value 0.666 0.615 0.246

R. Cesur et al. / Journal of Health Economics 32 (2013) 51– 65 57

Table 2 (Continued)

Variables (1) (2) (3) Combat versus civilians Combat versus

non-combat overseas Combat w/firefight versus combat w/out firefight

Observations 14,823 572 411 R-Squared 0.038 0.180 0.210

F-Test all 11.17 2.647 6.341 F-Test all P-value 0.000 0.000 0.000

Robust standard errors corrected for clustering on the school are in parentheses. Columns (2) and (3) include controls for military-specific variables, including rank, branch of service, timing of service, and occupation.

a o

f e i r c s c t i i p p a p f e s n a

u w r o l

5 h

w p t c n b d s

M

w I

w w a s

I o T

A r t d e e e a o

c o s a ity of PTSD relative to deployment assignments outside the United States in non-combat zones. The estimates on suicidal thoughts and depression are not precisely estimated. Thus, caution must

18 While 84.4 percent of respondents were younger than age 18 at Wave I, 30 respondents (0.2 percent) reported military service at Wave I. To ensure that Mit−3 captures pre-military mental health, we drop these 30 individuals from our sample. However, the results are qualitatively similar with their inclusion. Because we do not have a measure of PTSD at Wave I, our estimates from Eq. (1) include receipt of emotional or psychological counseling as our pre-deployment dependent variable.

19 In alternate specifications, available in our NBER working paper (Cesur et al., 2011), we examine the sensitivity of estimates to the inclusion of school fixed effects. For instance, youths in economically depressed areas may have worse men- tal health and face lower opportunity costs of volunteering for active duty service (Brown, 1985; Morrison and Myers, 1998). Moreover, respondents from less socially connected schools may be more likely to join the military and have worse psycho- logical outcomes (Elder et al., 2010). Results including school fixed effects were qualitatively similar to those presented here. We also experimented with models that include family fixed effects. For instance, there is evidence that individuals with fewer resources and larger family sizes are more likely to join the military (Kilburn and Asch, 2003; Kilburn and Klerman, 1999), and as noted above, socioeconomic sta- tus has also been found to be related to mental health (Miech et al., 1999). Moreover, the enlistment of a parent is associated with a greater expectation and probability of service among offspring (Kilburn and Klerman, 1999; Segal and Segal, 2004), sug- gesting that there may be common familial values associated with military service and psychological well-being. Because of small sample sizes, we cannot compare

* Statistical significance at the 10% level. ** Statistical significance at the 5% level.

*** Statistical significance at the 1% level.

control group is likely to produce biased estimates of the effect f combat exposure on mental health.

However, a different pattern emerges when we narrow our ocus to active duty servicemembers deployed overseas and xamine whether—conditional on military rank, occupation, tim- ng of service, and branch of service—deployment assignment is elated to background characteristics or to pre-deployment psy- hological well-being. For example, in column (2), we restrict our ample to those who were deployed overseas and regress an indi- ator for whether the respondent was assigned to a combat zone on he full set of observables. Of all the right-hand side variables, which nclude a wide set of family and individual background character- stics, only being male is associated with a statistically significant robability of deployment to combat. Importantly, we find that re-deployment mental health is unrelated to the probability of ssignment to a combat zone. In column (3), we restrict the sam- le to those deployed to combat zones and regress an indicator or whether the respondent was assigned to a combat region with nemy firefight on the full set of observables. Again, we find that trong evidence to indicate that combat experience with firefight is ot significantly associated with a wide range of observable family nd individual background characteristics.17

While the findings in Table 2 cannot allow us to test whether nobserved differences persist across deployment assignment, hich would be a necessary condition for random assignment, the

esults do suggest little evidence of differences across a wide set of bservables including pre-deployment mental health, a finding at east consistent with random assignment.

. Estimated effect of combat zone assignment on mental ealth

To nest our estimates in the context of the existing literature, e begin our empirical analyses by drawing data from active-duty ersonnel assigned to combat zones and comparing their men- al health outcomes to respondents in four control groups: (1) ivilians, (2) Reservists and National Guardsmen, (3) active duty on-deployed servicemembers, and (4) active duty servicemem- ers deployed to non-combat zones overseas. Specifically, we draw ata from Waves I and IV of the Add Health and use ordinary least quares (OLS) to estimate a model of the following form:

′

it = ̨ + �Mit−3 + ı1Combatit + Xit� + �it, (1) here Mit is the mental health outcome for respondent i at Wave

V, Mit−3 is the mental health outcome for respondent i at Wave

17 Only the “Hispanic” and “Male” indicators were significant predictors of combat ith firefight and we control for these variables in all models. In additional analyses, e re-estimated these regressions by restricting the sample to those who were male

nd not in the “other” race category. The results are quantitatively and qualitatively imilar to those reported here and are available from the authors upon request.

s o o i n

t w p t a

,18 and Xi is the vector of controls described above. 19 The means

f each of the control variables in Xi, by military status, appear in able A1.

OLS estimates of the ı1 from Eq. (1) are presented in Panel of Table 3.20 The estimates in rows (2)–(4) of Panel A confirm esults from the prior literature, suggesting that those assigned o combat have worse mental health than their active duty non- eployed counterparts (Shen et al., 2010; Wells et al., 2010; Smith t al., 2008), non-active duty non-deployed counterparts (Wells t al., 2010; Smith et al., 2008), and civilian counterparts (Jordan t al., 1991). Across outcomes, the magnitudes of the differences re largest for the control group with the least theoretical evidence f exogenous assignment (civilians).

The estimates in row (1) of Panel A present evidence from the ontrol group of interest—those deployed to non-combat zones verseas—where the case for exogenous deployment assignment is trongest. We find that combat zone deployment is associated with

615.9 percent (12.4 percentage-point increase) in the probabil-

iblings assigned to combat and non-combat overseas, so our analysis focuses only n active-duty versus non-active duty service. We find that respondents who serve n active duty have a greater risk for PTSD than their siblings serving in the military n non-active duty roles, suggesting that family-level unmeasured heterogeneity is ot an especially important source of bias in prior studies. 20 We report standard error estimates that are robust to any form of heteroskedas- icity (Angrist and Krueger, 1999). Given that the Add Health is a school based survey, e also cluster the standard errors at the school level. Estimation of the models via robit yielded qualitatively similar marginal effects. The estimated coefficients on he variables in Xi are consistent with those found in the relevant literature and are vailable from the authors.

58 R. Cesur et al. / Journal of Health

Table 3 Estimates of the relationship between deployment assignment and mental health.

(1) (2) (3) Suicide Depression PTSD

Panel A: full sample Comparison group Non-combat deployed 0.019 0.032 0.124***

(0.029) (0.038) (0.019) [574] [575] [576]

Active duty non-deployed 0.004 −0.016 0.123*** (0.028) (0.032) (0.024) [754] [762] [762]

Non-active duty 0.039 0.034 0.153***

(0.029) (0.040) (0.039) [585] [588] [588]

Civilians 0.018 0.028* 0.153***

(0.013) (0.017) (0.016) [14,778] [14,929] [14,989]

Panel B: Army sample Comparison group Non-combat deployed 0.071 0.027 0.195***

(0.045) (0.083) (0.061) [241] [239] [240]

Panel C: Army Post 9/11 sample Comparison group Non-combat deployed 0.091* −0.009 0.217***

(0.049) (0.087) (0.075) [224] [222] [223]

Panel D: Navy sample Comparison group Non-combat deployed −0.008 0.082 −0.005

(0.049) (0.073) (0.030) [145] [146] [146]

Panel E: Air Force sample Comparison group Non-combat deployed −0.160 0.074 0.023

(0.127) (0.157) (0.118) [91] [93] [93]

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 pre-deployment mental health. In all models (except combat versus civilians), military rank, timing of military service, branch of service, occupa- tion 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.

b f l s

b d c W a e z P I i s i a

f a r

N s

6 h

s l m 2 w s

M

r b p z a s z t

l t d o fi e w e p

r e p p p Panel C, we examine whether those who engage more frequently in enemy firefight suffer adverse psychological consequences and find that those who reported 20 or more engagements with enemy firefight had the largest adverse psychological effects.24

22 As another descriptive test of the plausible exogeneity of deployment assign- ment, Table A2 shows the strong stability of the estimated effect of combat assignment on mental health to controls for individual and family characteristics.

23 When we add pre-deployment CES-D and suicidal thoughts to the PTSD equa- tion, the marginal effect remains stable. Moreover, in unreported results, we find no difference in the mental health effects of combat by gender. For the outcomes other than PTSD, we also experiment with formal individual fixed effects models. While slightly smaller in magnitude and less precisely estimated, the results are qualitatively similar to the findings we present in here.

24 There is no a priori medically driven reason to categorize the number of firefights in a particular fashion. We stress that there is nothing unique about the 20th fire- fight that generates adverse psychological consequences. We chose the categories

* Statistical significance at the 10% level. *** Statistical significance at the 1% level.

e exercised when interpreting the magnitude of the effect sizes or these estimates given the wide interval for the confidence evel around these estimate relative to their prevalence in our ample.21

The remaining panels of Table 3 present estimates separately y branch of service, focusing exclusively on the conditional mean ifference in mental health outcomes between those assigned to ombat zones and those assigned to non-combat zones overseas.

e find that the effects of combat zone deployment are largest nd most robust for those who serve in the Army (Panel B). For xample, active-duty members of the Army deployed to combat ones are 468 percent (19.5 percentage-points) more likely to have TSD relative to their counterparts assigned to non-combat zones. n Panel C, we restrict the sample to those who served in the Army n the post-9/11 period, who comprise the vast majority of Army ervice members in our sample, and find that assignment to combat

s associated with an increase in the likelihood of suicidal ideation nd PTSD. The magnitudes of the estimated combat effects for the

21 For example, the 95 percent confidence is [−0.038, 0.076] around the estimate or suicidal thoughts and [−0.043, 0.108] around the estimate for depression. While n effect size of zero clearly lies inside these intervals, so do values that would epresent economically significant impact sizes.

o t n t d e 2 r t d

Economics 32 (2013) 51– 65

avy sample (Panel D) and active-duty Airmen (Panel E) are much maller and imprecisely estimated.22,23

. Decomposing the effect of combat exposure on mental ealth

The existing literature on the mental health effects of combat ervice in the GWOT have focused on the effects of deployment ength, suggesting that it is length of time in a combat zone that is

ost harmful to soldiers’ psychological well-being (Buckman et al., 010; Shen et al., 2009, 2010; Rona et al., 2007). In Panel A of Table 4, e replicate their findings using the Add Health. We restrict the

ample to those assigned to combat zones and estimate:

it = ̨ + �Mit−3 + ˛1Combat Lengthit + Xit� + �it, (2) Consistent with Shen et al. (2009) and Adler et al. (2005), our

esults show that those serving more than 12 months in a com- at zone experience a 102.4 percent (8.9 percentage-points) higher robability of receiving a PTSD diagnosis than those with combat one service of 1–6 months. The estimates on suicidal thoughts nd depression are not estimated precisely. But the relatively large tandard errors suggest that the actual effect sizes may be non- ero. Therefore, these estimates should be treated with caution in erms of interpreting their economic significance.

However, an important question remains: Does deployment ength adversely affect soldiers’ mental health or is it psycho- raumatic violent combat events often associated with longer eployments? To explore this question, we exploit a unique aspect f the Add Health data, which include information on enemy refighting among those who had deployed to a combat zone. Such xperiences vary by combat zone deployment assignment, which, e argue above, is exogenously assigned. Moreover, the use of

xplicit information on enemy firefight allows us to improve upon rior studies’ use of deployment location as a proxy for combat.

In Panel B of Table 4, we present estimates of the effects of expe- iencing enemy firefight on mental health. We find that those who xperience enemy firefight have a 154.8 percent (6.7 percentage- oint) higher probability of suicidal ideation, a 55.6 percent (7.7 ercentage-point) higher probability of depression, and a 140.9 ercent (12.2 percentage-point) higher probability of PTSD. In

f 1–3 firefights, 4–19 firefights, and 20+ firefights because this was the categoriza- ion that divided the distribution of firefights (among those who reported a positive umber of firefights) in thirds. We examined the sensitivity of our results to dividing he sample in alternative ways and the results are largely similar. For example, we ivided the sample of those who experience enemy firefights into 2 approximately qual categories (each with 50% of the distribution), 4 equal categories (each with 5% of the distribution), and 5 categories (each with 20% of the distribution), the esults are similar to those reported here. In particular, results from these estima- ions consistently show that it is those in the upper half, third, fourth, and fifth of the istribution of firefights who tend to be driving the effect on firefights. These results

R. Cesur et al. / Journal of Health Economics 32 (2013) 51– 65 59

Table 4 Estimated effects of combat deployment length and exposure to enemy firefight.

(1) (2) (3) Suicide Depression PTSD

Panel A: combat deployment length Combat zone service length: 7–12 months 0.016 0.062 0.017

(0.041) (0.053) (0.044) Combat zone service length: more than 12 months −0.003 0.035 0.089*

(0.040) (0.049) (0.047)

More than 12 months – 7–12 months = 0? [p-value on F-test] 0.629[−0.019] 0.586[−0.027] 0.137[0.072] Observations 412 414 414 Panel B: any enemy firefight 1 or more enemy firefights 0.067** 0.077* 0.122***

(0.028) (0.044) (0.037) Observations 412 414 414 Panel C: categorical enemy firefights 1–3 enemy firefights 0.040 0.065 0.041

(0.035) (0.063) (0.054) 4–19 enemy firefights 0.032 0.095 0.110*

(0.045) (0.060) (0.060) 20 or more enemy firefights 0.138*** 0.076 0.242***

(0.051) (0.066) (0.070)

20 or more firefight – 1–3 firefight = 0? [p-value on F-test] 0.098[0.13] 0.011[0.89] 0.201** [0.02] 20 or more firefight – 4–19 firefight = 0? [p-value on F-test] 0.106* [0.09] −0.020[0.80] 0.132[0.14] Observations 412 414 414 Panel D: deployment length and enemy firefight 1–3 enemy firefights 0.044 0.066 0.037

(0.035) (0.063) (0.054) 4–19 enemy firefights 0.035 0.089 0.104*

(0.045) (0.060) (0.058) 20 or more enemy firefights 0.149*** 0.077 0.227***

(0.052) (0.070) (0.068) Combat zone service length: 7–12 months 0.009 0.052 0.001

(0.042) (0.053) (0.044) Combat zone service length: more than 12 months −0.030 0.015 0.044

(0.039) (0.051) (0.043)

20 or more firefight – 1–3 firefight = 0? [p-value on F-test] 0.105[0.10] 0.011[0.89] 0.190** [0.02] 20 or more firefight – 4–19 firefight = 0? [p-value on F-test] 0.114* [0.08] −0.012[0.87] 0.123[0.16] Observations 412 414 414

Robust standard errors corrected for clustering on the school are in parentheses. All models use the time varying set of controls shown in Table A1 along with pre-deployment mental health, military rank, timing of military service, branch of service, occupation indicators, and an indicator for having a check-up in the past year. Models also include m

i fi t i t f i o i l

a t v a g a o c r t f c

e t c g s o e d

issing dummy categories for each of the control variables. * Statistical significance at the 10% level.

** Statistical significance at the 5% level. *** Statistical significance at the 1% level.

Finally, in Panel D of Table 4, we run a “horserace” to exam- ne whether it is the deployment length or the exposure to enemy refight that is the greater driver of adverse mental health among hose deployed to combat. The results show that those who engage n more frequent enemy firefights are at a greater risk for suicidal houghts, depression, and PTSD than their counterparts who face ewer combat firefights, although the coefficients for depression s not precisely estimated. Interestingly, we find that the effect

f deployment length is considerably diminished after condition- ng on number of enemy firefights and none of the deployment ength effects are statistically significant at conventional levels. One

re available from the authors upon request. Finally, we run regressions similar to hose in Table 2 to examine the relationship between firefights and a wide set of indi- idual and family background characteristics. The findings from these regressions re again in line with those reported in Table 2, i.e., those in the 20+ firefight cate- ory do appear to be somewhat more white, male and less-educated, all of which are ccounted for in the regression models. We also estimate linear probability models f the probability of being assigned to a combat zone with 20+ firefights versus a ombat zone with fewer firefights (conditional on military-specific controls). These esults show that those assigned to combat zones with 20+ enemy firefights appear o be more white and male than their counterparts assigned to combat zones with ewer firefights. However, they are no different on family background or individual haracteristics (including pre-deployment mental health).

fi b d

m o h m o w ( s t

k

xplanation for this result is that frequent enemy firefight drives he adverse psychological consequences previously attributed to ombat deployment length. However, this explanation is only sug- estive and must be interpreted with caution. This is because as hown by the cross-tabulations in Table A4, there is a high degree f correlation between the deployment length and frequency of nemy firefight. For example, while over 70 percent of those with a eployment length less than 6 months experienced no enemy fire- ght, this figure is 60 percent for those with a deployment length etween 7 and 12 months and is only 42 percent for those with a eployment length of more than 12 months.

In Table 5, we examine the effects of deaths and injuries on ental health, conditional on deployment length and frequency

f enemy firefight. In Panel A, we find that killing or believing to ave killed someone is not significantly associated with any of our easures of mental health problems after accounting for intensity

f firefight and deployment length. In Panel B, we find that being ounded or injured in combat is associated with a 124.1 percent

19.2 percentage-point) increase in the probability of depressive

ymptomatology, and a 225.2 (26.5 percentage-point) increase in he probability of PTSD.

In Panel C, we explore whether observing someone wounded, illed, or dead affects soldiers’ mental health. The results suggest

60 R. Cesur et al. / Journal of Health Economics 32 (2013) 51– 65

Table 5 Estimated effects of exposure to killing, injuries, and deaths.

Variables (1) (2) (3) Suicide Depression PTSD

Panel A: killed or believed killed another Killed or believed killed another 0.013 0.027 0.066

(0.056) (0.062) (0.078) 1–3 enemy firefights 0.040 0.067 0.024

(0.043) (0.071) (0.063) 4–19 enemy firefights 0.038 0.093 0.061

(0.066) (0.082) (0.084) 20 or more enemy firefights 0.140** 0.055 0.170*

(0.066) (0.090) (0.097) Combat zone service length: 7–12 months (%) 0.012 0.065 0.011

(0.043) (0.053) (0.046) Combat zone service length: more than 12 months (%) −0.028 0.010 0.049

(0.039) (0.052) (0.043)

Observations 404 406 407 Panel B: injury Wounded or injured in combat 0.093 0.192** 0.265***

(0.063) (0.083) (0.074) 1–3 enemy firefights 0.043 0.064 0.035

(0.035) (0.062) (0.055) 4–19 enemy firefights 0.026 0.068 0.076

(0.045) (0.061) (0.055) 20 or more enemy firefights 0.132** 0.042 0.180***

(0.053) (0.068) (0.068) Combat zone service length: 7–12 months (%) 0.015 0.064 0.017

(0.043) (0.055) (0.045) Combat zone service length: more than 12 months (%) −0.026 0.025 0.057

(0.041) (0.054) (0.043)

Observations 412 414 414 Panel C: observe death or wounding on battlefield Saw coalition or ally killed, dead, or wounded 0.039 0.078* 0.017

(0.036) (0.046) (0.046) Saw civilian killed, dead, or wounded 0.102** −0.007 0.120**

(0.047) (0.063) (0.054) Saw enemy killed, dead, or wounded −0.025 0.008 0.030

(0.041) (0.057) (0.059) 1–3 enemy firefights 0.027 0.051 0.008

(0.034) (0.064) (0.055) 4–19 enemy firefights −0.007 0.062 0.038

(0.051) (0.076) (0.063) 20 or more enemy firefights 0.104* 0.040 0.156**

(0.061) (0.087) (0.076) Combat zone service length: 7–12 months (%) 0.016 0.056 0.012

(0.041) (0.054) (0.044) Combat zone service length: more than 12 months (%) −0.040 0.003 0.036

(0.038) (0.051) (0.043)

Coalition − enemy = 0? [p-value on F-test] 0.064[0.22] 0.070[0.34] −0.013[0.86] Civilian − enemy = 0? [p-value on F-test] 0.127[0.11] −0.015[0.88] 0.090[0.36] Observations 410 412 412

Robust standard errors corrected for clustering on the school are in parentheses. All models use the time varying set of controls shown in Table A1 along with for pre- deployment mental health, military rank, timing of military service, and branch of service, branch of service, occupation indicators, and an indicator for having a check-up in the past year. Models also include missing dummy categories for each of the control variables.

t w s k a t e c

w

may accompany the death of non-combatants or fellow soldiers, and could even reflect, in part, the effect of friendly fire incidents.26

26

* Statistical significance at the 10% level. ** Statistical significance at the 5% level.

*** Statistical significance at the 1% level.

hat seeing a coalition/ally member or civilian killed, dead, or ounded has adverse psychological consequences for those

erving in combat zones. However, we find that observing the illing, death, or wounding of the enemy has no independent dverse psychological consequences. Despite limited power in

hese models, each of the coefficients for those who saw wounded nemy are very close to zero in magnitude.25 These results are onsistent with the hypothesis that strong feelings of loss or guilt

25 While the effect sizes are very small, the standard errors are still relatively large, hich indicates that economically meaningful effects are still plausible.

b F a a a o fi n a

Panels B through F of Table A4 show that there is some independent variation etween frequency of enemy firefight and exposure to other violent combat events. or example, the proportion of servicemen who (i) killed or believe to have killed nother individual (Panel B); (ii) wounded or injured in combat (Panel C); (iii) saw

coalition member or ally to have been killed, dead or wounded (Panel D); (iv) saw civilian killed, dead or wounded (Panel E); and (v) saw an enemy killed, dead, r wounded (Panel F) is higher among those who engaged in 20 or more enemy refights. However, we acknowledge the difficulty of disentangling the effects of umbers of firefights from observing deaths on the battlefield given collinearity nd limited statistical power associated with the reduced sample size.

ealth

7

A s 1 b “ D a o

a m e t t c H d c

s o w c t t o t

m s e c t m fi i o t d

d d o d i m d s e c e a

T S

R. Cesur et al. / Journal of H

. Conclusion

The U.S. military has engaged in two wars in Iraq and fghanistan in the last ten years, deploying 2.16 million U.S. troops ince October 2001 (Department of Defense, 2010a). As of March 6, 2012, 5016 U.S. soldiers had been killed in action, 47,684 had een wounded in action, and many more returning home with invisible wounds,” such as mental health injuries (Department of efense, 2010b). In this study, we exploit variation in deployment ssignment among those deployed overseas to estimate the effect f combat service on young adults’ psychological well-being.

Our results lend support to the hypothesis that combat service is ssociated with adverse psychological consequences and that the echanism is driven by potentially psycho-traumatic incidences

xperienced during combat zone missions. In particular, we find hat frequent enemy firefight, wounding or injury, and observing he death or wounding of a coalition/ally or non-combatant is asso- iated with an increase in the risk of suicidal thoughts and PTSD. owever, we find that observing the death of an enemy combatant oes not adversely affect the mental health of those assigned to ombat zones.

The results of this study consistently suggest that military per- onnel who are assigned to combat duties are at increased risk f PTSD. Moreover, exposure to frequent enemy firefight and the ounding or deaths of allies or non-combatant civilians is asso-

iated with greater risk of depressive symptomatology or suicidal

houghts, particularly for those in the Army. It must be noted that here are important differences in the clinical definitions of each of ur outcomes, which may help explain some of the differences in he effects of various measures of combat on our three measures of

A

able A1 ummary statistics by military service.

Full sample

Civilian sample

Military sample

All Non-active duty

Activ non- deplo

White 0.697 0.699 0.674 0.671 0.695 (0.460) (0.459) (0.469) (0.471) (0.46

Black 0.230 0.229 0.254 0.281 0.240 (0.421) (0.420) (0.435) (0.451) (0.42

Other race 0.071 0.071 0.070 0.043 0.065 (0.256) (0.256) (0.256) (0.203) (0.24

Hispanic 0.159 0.160 0.146 0.122 0.142 (0.366) (0.367) (0.354) (0.328) (0.35

Male 0.467 0.443 0.790 0.720 0.713 (0.499) (0.497) (0.408) (0.451) (0.45

Height in inches 67.311 67.174 69.149 68.756 68.76 (4.141) (4.134) (3.785) (3.887) (3.74

Weight in pounds 183.307 182.915 188.547 188.494 187.6 (49.354) (50.083) (37.936) (39.684) (41.1

Less than high school 0.080 0.085 0.011 0.024 0.012 (0.271) (0.279) (0.105) (0.155) (0.10

High school 0.163 0.164 0.157 0.122 0.186 (0.370) (0.370) (0.364) (0.328) (0.39

Some college or vocational training

0.441 0.425 0.653 0.628 0.636

(0.497) (0.494) (0.476) (0.485) (0.48 College degree 0.238 0.245 0.149 0.165 0.139

(0.426) (0.430) (0.356) (0.372) (0.34 Graduate or

professional degree 0.078 0.081 0.030 0.061 0.027

(0.268) (0.273) (0.170) (0.240) (0.16 =1 if 24 years old, =0

otherwise 0.002 0.002 0.000 0.000 0.000

(0.045) (0.047) 0.000 0.000 0.000 =1 if 25 years old, =0

otherwise 0.044 0.045 0.039 0.055 0.033

(0.206) (0.207) (0.193) (0.228) (0.17

Economics 32 (2013) 51– 65 61

ental health. For example, the PTSD diagnosis requires an expo- ure to a traumatic event such as combat exposure. Then this may xplain why our results are most robust for PTSD relative to sui- idal thoughts and depression. Furthermore, it is important to note hat the strongest estimates on suicidal thoughts are obtained in

odels for particular combat situations like experiencing enemy refight, which is likely to impose the most extreme psycholog-

cal distress on military service personnel. Finally, given the lack f precision in some of our estimates on depression and suicidal houghts, we cannot rule out effects for combat situations that our ata are unable to detect.

The U.S. Army recently announced plans to reduce combat zone eployments to nine months and to increase the time between eployments to three years by the year 2014 (Tice, 2010). While ur findings confirm that deployment length is associated with eclines in mental health, our results also show that this effect

s likely driven by frequent enemy firefight rather than deploy- ent length alone. Thus, military policymakers crafting optimal

eployment schedules that account for mental health problems of oldiers should focus a great deal of attention on violent combat vents rather than primarily focusing on soldiers’ time spent in a ombat zone. Moreover, our findings suggest that reducing soldiers’ xposure to civilian casualties in battle may also avert significant dverse mental health consequences for US soldiers.

ppendix A.

See Tables A1–A4.

Combat zone sample

e duty

yed

Active duty non-combat deployed

All Deployed with firefight

Combat deployed without firefight

0.686 0.662 0.724 0.619 1) (0.465) (0.474) (0.448) (0.487)

0.253 0.245 0.205 0.268 8) (0.435) (0.430) (0.405) (0.444)

0.062 0.091 0.065 0.113 7) (0.241) (0.288) (0.247) (0.317)

0.148 0.154 0.119 0.191 0) (0.355) (0.361) (0.325) (0.394)

0.737 0.877 0.957 0.810 3) (0.441) (0.329) (0.204) (0.394) 6 68.887 69.597 70.184 69.078 4) (3.774) (3.721) (3.578) (3.813) 62 188.666 188.431 190.696 187.274 76) (40.556) (34.062) (31.834) (36.158)

0.014 0.002 0.005 0.000 8) (0.119) (0.048) (0.074) 0.000

0.177 0.149 0.168 0.130 0) (0.382) (0.357) (0.375) (0.337)

0.647 0.669 0.670 0.671

2) (0.479) (0.471) (0.471) (0.471) 0.131 0.163 0.135 0.186 7) (0.338) (0.370) (0.343) (0.390)

0.031 0.016 0.022 0.013

1) (0.173) (0.127) (0.146) (0.114) 0.000 0.000 0.000 0.000

0.000 0.000 0.000 0.000 0.033 0.040 0.065 0.022

8) (0.178) (0.195) (0.247) (0.146)

62 R. Cesur et al. / Journal of Health Economics 32 (2013) 51– 65

Table A1 (Continued)

Full sample

Civilian sample

Military sample Combat zone sample

All Non-active duty

Active duty non- deployed

Active duty non-combat deployed

All Deployed with firefight

Combat deployed without firefight

=1 if 26 years old, =0 otherwise

0.116 0.117 0.106 0.110 0.121 0.099 0.112 0.081 0.134

(0.320) (0.321) (0.307) (0.314) (0.327) (0.298) (0.316) (0.274) (0.342) =1 if 27 years old, =0

otherwise 0.145 0.146 0.131 0.146 0.142 0.129 0.126 0.124 0.121

(0.352) (0.353) (0.337) (0.355) (0.350) (0.336) (0.332) (0.331) (0.327) =1 if 28 years old, =0

otherwise 0.181 0.179 0.197 0.226 0.192 0.191 0.194 0.205 0.182

(0.385) (0.384) (0.398) (0.419) (0.395) (0.394) (0.396) (0.405) (0.387) =1 if 29 years old, =0

otherwise 0.189 0.188 0.203 0.152 0.207 0.216 0.208 0.211 0.204

(0.392) (0.391) (0.402) (0.361) (0.406) (0.412) (0.406) (0.409) (0.403) =1 if 30 years old, =0

otherwise 0.183 0.184 0.176 0.207 0.160 0.166 0.175 0.162 0.195

(0.387) (0.387) (0.381) (0.407) (0.367) (0.373) (0.380) (0.370) (0.397) =1 if 31 years old, =0

otherwise 0.115 0.114 0.132 0.079 0.127 0.152 0.131 0.130 0.130

(0.319) (0.318) (0.339) (0.271) (0.334) (0.359) (0.337) (0.337) (0.337) =1 if 32 years old, =0

otherwise 0.021 0.021 0.017 0.024 0.018 0.014 0.016 0.022 0.013

(0.143) (0.144) (0.128) (0.155) (0.132) (0.119) (0.127) (0.146) (0.114) =1 if 33 years old, =0

otherwise 0.003 0.003 0.000 0.000 0.000 0.000 0.000 0.000 0.000

(0.054) (0.056) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 =1 if 34 years old, =0

otherwise 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

(0.018) (0.019) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 None, Atheist, or

Agnostic 0.181 0.181 0.186 0.171 0.181 0.189 0.187 0.195 0.175

(0.385) (0.385) (0.389) (0.377) (0.385) (0.392) (0.391) (0.397) (0.381) Protestant 0.292 0.289 0.332 0.421 0.308 0.298 0.337 0.368 0.306

(0.455) (0.453) (0.471) (0.495) (0.462) (0.458) (0.473) (0.484) (0.462) Catholic 0.220 0.220 0.211 0.189 0.201 0.202 0.230 0.227 0.236

(0.414) (0.414) (0.408) (0.393) (0.402) (0.402) (0.421) (0.420) (0.425) Other Christian 0.224 0.226 0.204 0.189 0.234 0.220 0.192 0.162 0.223

(0.417) (0.418) (0.403) (0.393) (0.424) (0.415) (0.394) (0.370) (0.417) Other religion 0.083 0.085 0.067 0.031 0.077 0.091 0.054 0.049 0.061

(0.277) (0.278) (0.250) (0.173) (0.267) (0.287) (0.226) (0.216) (0.240) No health insurance 0.212 0.215 0.162 0.168 0.234 0.210 0.104 0.121 0.097

(0.409) (0.411) (0.368) (0.375) (0.424) (0.408) (0.306) (0.327) (0.296) Wave 1 picture

vocabulary test score

100.586 100.352 103.781 105.616 103.624 103.495 103.380 103.526 103.522

(14.548) (14.646) (12.718) (11.550) (12.586) (12.461) (13.414) (14.759) (12.244) Parental income in

thousands Wave 1 46.414 46.606 43.893 41.906 45.076 43.278 45.491 45.169 43.911

(50.506) (50.859) (45.588) (34.267) (61.357) (53.488) (39.150) (29.099) (30.573) Parent is never

married in Wave 1 0.056 0.057 0.047 0.054 0.049 0.046 0.046 0.043 0.051

(0.230) (0.231) (0.212) (0.228) (0.216) (0.210) (0.209) (0.204) (0.220) Parent is married in

Wave 1 0.711 0.711 0.705 0.701 0.690 0.698 0.714 0.710 0.722

(0.453) (0.453) (0.456) (0.460) (0.463) (0.460) (0.452) (0.455) (0.449) Parent is divorced,

separated or widowed Wave 1

0.233 0.232 0.248 0.245 0.261 0.256 0.240 0.247 0.227

(0.423) (0.422) (0.432) (0.432) (0.440) (0.437) (0.428) (0.433) (0.420) Biological mother’s

education: less than high school

0.167 0.169 0.138 0.124 0.152 0.153 0.125 0.130 0.115

(0.373) (0.375) (0.345) (0.330) (0.360) (0.361) (0.331) (0.338) (0.320) Biological mother’s

education: high school degree

0.337 0.337 0.345 0.352 0.337 0.344 0.343 0.310 0.367

(0.473) (0.473) (0.476) (0.479) (0.474) (0.475) (0.475) (0.464) (0.483) Biological mother’s

education: some college

0.194 0.192 0.218 0.241 0.197 0.203 0.227 0.234 0.221

(0.395) (0.394) (0.413) (0.429) (0.398) (0.403) (0.419) (0.424) (0.416)

R. Cesur et al. / Journal of Health Economics 32 (2013) 51– 65 63

Table A1 (Continued)

Full sample

Civilian sample

Military sample Combat zone sample

All Non-active duty

Active duty non- deployed

Active duty non-combat deployed

All Deployed with firefight

Combat deployed without firefight

Biological mother’s education: college degree or more

0.260 0.259 0.261 0.253 0.257 0.259 0.267 0.304 0.243

(0.438) (0.438) (0.440) (0.436) (0.438) (0.438) (0.443) (0.461) (0.430) Biological mother’s

education: not known

0.043 0.043 0.038 0.031 0.057 0.041 0.038 0.022 0.053

(0.202) (0.202) (0.192) (0.174) (0.232) (0.199) (0.191) (0.146) (0.225)

Observations 15,669 14,589 1080 164 338 487 429 185 231

Unweighted means are obtained from Waves I and IV of the National Longitudinal Study of Adolescent Health. Standard deviations are in parentheses.

Table A2 Stability of estimates of the mental health effects of deployment assignment to added controls, military population.

(1) (2) (3) (4) (5) Individual controls (1) + family controls (2) + health controls (3) + military controls (4) + Wave I mental health

Panel A: suicide Comparison group Non-combat deployed 0.016 0.015 0.015 0.024 0.019

(0.023) (0.025) (0.025) (0.030) (0.029)

Observations 574 574 574 574 574 Panel B: depression Comparison group Non-combat deployed 0.020 0.021 0.023 0.046 0.032

(0.036) (0.036) (0.037) (0.040) (0.038)

Observations 575 575 575 575 575 Panel C: PTSD Comparison group Non-combat deployed 0.147*** 0.142*** 0.139*** 0.127*** 0.124***

(0.019) (0.019) (0.019) (0.019) (0.019)

Observations 576 576 576 576 576

Robust standard errors corrected for clustering on the school are in parentheses. Individual controls are health, age, height, weight, religion, gender, race-ethnicity, income and PPVT score. Family controls are parental income, parental marital status and parental education during high school. Health controls are check-up in the past year and an indicator for health insurance status. Military controls are rank, branch of service, timing of service, and occupation dummies.

*** Statistical significance at the 1% level.

Table A3 Estimates of the relationship between deployment assignment and mental health for White or Black males.

(1) (2) (3) Suicide Depression PTSD

Panel A: full sample Comparison group Non-combat deployed 0.013 0.069 0.127***

(0.041) (0.048) (0.033) [377] [378] [378]

Active duty non-deployed 0.012 −0.028 0.128*** (0.034) (0.045) (0.033) [466] [473] [472]

Non-active duty 0.043 0.100** 0.103*

(0.037) (0.047) (0.053) [378] [380] [379]

Civilians 0.016 0.029 0.171***

(0.018) (0.022) (0.021) 5127 5221 5239

Panel B: Army sample Comparison group Non-combat deployed 0.027 0.051 0.280***

(0.075) (0.116) (0.087) 168 167 167

Panel C: Army Post 9/11 sample Comparison group Non-combat deployed 0.054 −0.015 0.272**

(0.084) (0.119) (0.105) 155 154 154

Table A3 (Continued)

(1) (2) (3) Suicide Depression PTSD

Panel D: Navy sample Comparison group Non-combat deployed 0.233** 0.255* 0.045

(0.111) (0.129) (0.051) 82 83 83

Panel E: Air Force sample Comparison group Non-combat deployed −12.349 0.645*** −0.366

(.) (0.138) (0.307) 58 59 59

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 pre-deployment mental health. In all models (except combat versus civilians), military rank, timing of military service, branch of service, occupa- tion 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.

* Statistical significance at the 10% level. ** Statistical significance at the 5% level.

*** Statistical significance at the 1% level.

64 R. Cesur et al. / Journal of Health

Table A4 Cross tabulations of combat exposure measures.

Number of enemy firefights (FF)

FF = 0 1 ≤ FF ≤ 3 4 ≤ FF ≤ 19 20 ≤ FF Total Panel A: firefight and deployment length Deployment length (DL) DL ≤ 6 months 81 21 9 4 115 7 months ≤ DL ≤ 12 months 76 17 20 13 126 12 months < DL 74 31 26 44 175 Total 231 69 55 61 416 Panel B: firefight and killed or believed killed another Killed or believed killed another Killed = 0 216 33 7 6 262 Killed = 1 14 32 45 55 146 Total 230 65 52 61 408 Panel C: firefight and wounded Wounded Wounded = 0 217 62 45 44 368 Wounded = 1 14 7 10 17 48 Total 231 69 55 61 416 Panel D: firefight and saw coalition or ally killed, dead or wounded Saw coalition or ally killed, dead or wounded Saw coalition or ally = 0 154 30 13 8 205 Saw coalition or ally = 1 77 38 41 53 209 Total 231 68 54 61 414 Panel E: firefight and saw civilian killed, dead or wounded Saw civilian killed, dead or wounded Saw civilian = 0 203 43 23 26 295 Saw civilian = 1 28 25 31 35 119 Total 231 68 54 61 414 Panel F: firefight and saw civilian killed, dead or wounded Saw enemy killed, dead or wounded Saw enemy = 0 197 34 16 17 264

R

A

A

A

A

A

A

B

B

B

B

B

C

C

C

D

D

D

D

D

D

D

D

E

E

E

E

H

H

H

H

H

H

H

I

J

J

K

K

K

K

L

M

M

M CES-D depression scales in systemic sclerosis: internal consistency reliability,

Saw enemy = 1 34 34 38 44 150 Total 231 68 54 61 414

eferences

dler, A.B., Huffman, A.H., Bliese, P.D., Castro, C.A., 2005. The impact of deployment length and experience on the well-being of male and female soldiers. Journal of Occupational Health Psychology 10 (April (2)), 121–137.

merican Psychiatric Association, 2000. Diagnostic and Statistical Manual of Mental Disorders, revised 4th ed. Author, Washington, DC.

ngrist, J.D., Krueger, A.B., 1999. Empirical strategies in labor economics. In: Ashen- felter, O., Card, D. (Eds.), Handbook of Labor Economics, vol. 3, 1st edition. Elsevier, North Holland, pp. 1277–1366 (Chapter 23).

ngrist, J.D., Chen, S.H., Frandsen, B.R., 2010. Did Vietnam veterans get sicker in the 1990s? The complicated effects of military service on self-reported health. Journal of Public Economics 94 (11–12), 824–837.

rmy Regulation 220-1, 2010. Army Unit Status Reporting and Force Registration—Consolidated Policies. Headquarters of the Department of the Army, Washington DC, http://www.apd.army.mil/pdffiles/r220 1.pdf

utor, D., Duggan, M.G., Lyle, D.S., 2011. Battle scars? The puzzling decline in employment and rise in disability receipt among Vietnam era veterans. Ameri- can Economic Review Papers and Proceedings 101 (3), 339–344.

achman, J.G., Segal, D.R., Freedman-Doan, P., O’Malley, P.M., 2000. Who chooses military service? Correlates of propensity and enlistment in the U.S. armed forces. Military Psychology 12 (1), 1–30.

edard, K., Deschenes, O., 2004. The long-term impact of military service on health: evidence from World War II and the Korean War. American Economic Review 96 (1), 176–194.

uckman, J.E.J., Sundin, J., Greene, T., Fear, N.T., Dandeker, C., Greenberg, N., Wes- sely, S., 2010. The impact of deployment length on the health and well-being of military personnel: a systematic review of the literature. Occupational and Environmental Medicine (September).

rown, C., 1985. Military enlistments: what can we learn from geographic variation? American Economic Review 75 (1), 228–234.

ryant, R.A., Guthrie, R.M., 2007. Maladaptive self-appraisals before trauma expo- sure predict post-traumatic stress disorder. Journal of Consulting and Clinical Psychology 75, 812–815.

ard, J., 1987. Epidemiology of PTSD in a National Cohort of Vietnam Veterans. Journal of Clinical Psychology 43, 6–17.

esur, R., et al., 2011. The psychological costs of war: Military combat and mental health. National Bureau of Economic Research Working Paper No. 16927, April.

hatterji, P., Cuellar, A., 2006. How do youth with mental disorders fare in the Juvenile Justice System. NBER Working Paper #12437, August.

epartment of the Army, 2010. Army Regulation 614-30, Available at: http://www.apd.army.mil/pdffiles/r614 30.pdf

M

Economics 32 (2013) 51– 65

epartment of Defense, 2010a. Legal residence/home address for service mem- bers ever deployed as of July 30, 2010. CTS Deployment File. Available at: http://dva.state.wi.us/WebForms/Data Factsheets/ResDistribution Jul10.pdf

epartment of Defense, 2010b. Daily Updated Casualty and Wounding Data, Avail- able at: http://www.defense.gov/news/casualty.pdf

epartment of Defense Directive 6130.3. Physical Standards for Appointment, Enlistment and Induction. Available at: http://biotech.law.lsu.edu/blaw/dodd/ corres/html/61303.htm

epartment of Defense Directive 6130.4. Criteria and Procedure Require- ments for Physical Standards for Appointment, Enlistment, or Induc- tion in the Armed Forces. Available at: http://biotech.law.lsu.edu/blaw/ dodd/corres/html/61304.htm

epartment of Defense Form 2795. Pre-Deployment Health Assessment. Available at: http://www.dtic.mil/whs/directives/infomgt/forms/eforms/dd2795.pdf

obkin, C., Shabani, R., 2009. The health effects of military service: evidence from the Vietnam draft. Economic Inquiry 47 (1), 69–80.

uncan, B., Rees, D.I., 2005. The effect of smoking on depressive symptomatology: a reexamination of the data from the national longitudinal study of adolescent health. American Journal of Epidemiology 162 (5), 461–470.

dwards, R.D., MacLean, A., 2010. Military service, combat exposure, and health in retirement. Working Paper, Queens College.

lder Jr., G., Wang, L., Spence, N., Adkins, D.E., Brown, T.H., 2010. Pathways to the all-volunteer military. Social Science Quarterly 91 (2).

ngel, R.C., Gallagher, L.B., Lyle, D.S., 2010. Military deployments and children’s aca- demic achievement: evidence from Department of Defense Education Activity Schools. Economics of Education Review 29 (1), 73–82.

rbes, Westermeyer, C.J., Engdahl, B., Johnsen, E., 2007. Post-traumatic stress dis- order and service utilization in a sample of service members from Iraq and Afghanistan. Military Medicine 172 (4), 359–363.

allfors, D.D., Waller, M.W., Ford, C.A., Halpern, C.T., Brodish, P., Iritani, B., 2004. Adolescent depression and suicide risk: association with sex and drug behavior. American Journal of Preventative Medicine 27 (3), 224–231.

arris, K.M., Halpern, C.T., Entzel, P., Tabor, J., Bearman, P.S., Richard Udry, J., 2008. The National Longitudinal Study of Adolescent Health: Research Design.

earst, N., Newman, T.B., Hulley, S.B., 1986. Delayed effects of the military draft on mortality: a randomized natural experiment. New England Journal of Medicine 314 (10), 620–624.

einrichs, M., Wagner, D., Schoch, W., Soravia, L.M., Hellhammer, D.H., Ehlert, U., 2005. Predicting post-traumatic stress symptoms from pre-traumatic risk fac- tors: a two-year prospective follow-up study in firefighters. American Journal of Psychiatry 162, 2276–2286.

irsch, B.T., Mehay, S.L., 2003. Evaluating the labor market performance of veterans using a matched comparison group design. Journal of Human Resources 38 (3), 673–700.

oge, C., Auchterlonie, J.L., et al., 2006. Health problems, use of mental health ser- vices, and attrition from military service after returning from deployment to Iraq or Afghanistan. Journal of the American Medical Association 295 (9), 1023–1032.

oge, C., Castro, C.A., et al., 2004. Combat duty in Iraq and Afghanistan: mental health problems and barriers to care. New England Journal of Medicine 317 (26), 1630–1634.

owa Persian Gulf Study Group, 1997. Self-reported illness and health status among Gulf War veterans. A population-based study. The Iowa Persian Gulf Study Group. Journal of the American Medical Association 277 (3), 238–245.

acobson, I.G., Ryan, M.A.K., Hooper, T.I., Smith, T.C., Amoroso, P.J., Boyko, E.J., Gackstetter, G.D., Wells, T.S., Bell, N.S., 2008. Alcohol use and alcohol-related problems before and after military combat deployment. Journal of the American Medical Association 300 (6), 663–675.

ordan, B.K., Schlenger, W.E., Hough, R., Kulka, R.A., Weiss, D., Fairbank, J.A., Mar- mar, C.R., 1991. Lifetime and current prevalence of specific psychiatric disorders among Vietnam veterans and control. Archives of General Psychiatry 48, 916–926.

ang, H.K., Bullman, T.A., 2001. Mortality among US veterans of the Persian Gulf War: 7-year follow-up. American Journal of Epidemiology 154 (5), 399–405.

ilburn, M.R., Klerman, J.A., 1999. Enlistment Decisions in the 1990s: Evidence from Individual-Level Data. RAND, Santa Monica, CA.

ilburn, M.R., Asch, B.J., 2003. Recruiting Youth in the College Market: Current Prac- tices and Future Policy Options. Rand Corporation, National Defense Research Institute (U.S.), http://www.rand.org/pubs/monograph reports/MR1093.html

leykamp, M.A., 2006. College, jobs, or the military? Enlistment during a time of war. Social Science Quarterly 87 (2).

yle, D.S., 2006. 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), 319–350.

cKinney, W.P., Mclntire, D.D., Carmody, T.J., Joseph, A., 1997. Comparing the smok- ing behavior of veterans and non-veterans. Public Health Report 112, 212–217.

iech, R.A., Caspi, A., Moffitt, T.E., Entner, B.R., Silva, P.A., 1999. Low socioeco- nomic status and mental disorders: a longitudinal study of selection and causation during young adulthood. The American Journal of Sociology 104 (4), 1096–1131.

ilette, K., Hudson, M., Baron, M., Thombs, B.D., 2010. Comparison of the PHQ-9 and

convergent validity and clinical correlates. Rheumatology 49 (4), 789–796. orrison, D.R., Myers, D.E., 1998. Factors Influencing the Enlistment Aspirations

and Decisions of Hispanic, Black, and White Male Youth. U.S. Army Research Institute for the Behavioral and Social Sciences.

ealth

N

O

O

O

P

R

R

R

R

S

S

S

S

S

S

S

S

T

T

T

T T U

W

W

W

R. Cesur et al. / Journal of H

ational Priorities Project, 2008. Military Recruitment 2008: A Look at Age, Race, Income, and Education of New Soldiers, Available at: http://nationalpriorities. org/en/publications/2009/army-recruitment-in-fy-2008-a-look-at-age-race- inc/

bama, B., 2010. Weekly Address: President Obama Announces Changes to Help Veterans with PTSD Receive the Benefits They Need. Office of the Press Secretary, Retrieved January 14, 2010 from http://www.whitehouse.gov/ the-press-office/weekly-address-president-obama-announces-changes- help-veterans-with-ptsd-receive-be

’Gorman, K., 2012, August. Army Reports Record High Suicides in July. Iraq and Afghanistan Veterans of America, Available at: http://iava.org/blog/ army-reports-record-high-suicides-july

rr, S.T., Blazer, D.G., James, S.A., 2005. Racial disparities in elevated prenatal depres- sive symptoms among black and white women in eastern North Carolina. Annals of Epidemiology 2006 16 (June (6)), 463–468.

rice, R.K., Risk, N.K., Haden, A.H., Lewis, C.E., Spitznagel, E.L., 2004. Post-traumatic stress disorder, drug dependence, and suicidality among male Vietnam veterans with a history of heavy drug use. Drug and Alcohol Dependence 76S, 31–43.

adloff, L.S., 1977. The CES-D scale: a self-report depression scale for research in the general population. Applied Psychological Measurement 1, 385–401.

ohlfs, C., 2010. Does combat exposure make you a more violent or criminal per- son? Evidence from the Vietnam draft. Journal of Human Resources 45 (2), 271–300.

ona, R.J., Hull, L., Earnshaw, M., Wessely, S., 2007. Mental Health consequences of overstretch in the UK armed forces: first phase of a cohort study. British Medical Journal 335 (7620), 603–610.

osenheck, R.A., Fontana, A.F., 2007. Recent trends in VA treatment of post-traumatic stress disorder and other mental disorders. Health Affairs 26 (6), 1720–1727.

abia, J.J., Rees, D.I., 2008. The effect of adolescent virginity status on psychological well-being. Journal of Health Economics 27 (5), 1368–1381.

eal, K.H., Bertenthal, D., Sen, S., Marmar, C., 2007. Bringing the war back home: mental health disorders among 103, 788 US veterans returning from Iraq and Afghanistan seen at Department of Veterans Affairs Facilities. Archives of Inter-

nal Medicine 167 (5), 476–482.

egal, D.R., Segal, M.W., 2004. America’s military population. Population Bulletin 59 (December (4)).

egal, D.R., Burns, T.J., Silver, M.P., Falk, W.W., Sharda, B.D., 1998. The all-volunteer force in the 1970s. Social Science Quarterly 79, 390–411.

W

Economics 32 (2013) 51– 65 65

hen, Y.-C., Arkes, J., John Pilgrim, M.A.J., 2009. The effects of deployment inten- sity on post-traumatic stress disorder: 2002–2006. Military Medicine 174 (3), 217–223.

hen, Y.-C., Arkes, J., Kwan, B.W., Tan, L.Y., Williams, T.V., 2010. Effects of Iraq/Afghanistan deployments on PTSD diagnoses for still active personnel in all four services. Military Medicine 175 (October (10)), 763–769 (7).

mith, T.C., Ryan, M.A.K., Wingard, D.L., Slymen, D.J., Sallis, J.F., Kritz-Silverstein, D., 2008. New onset and persistent symptoms of post-traumatic stress disorder self reported after deployment and combat exposures: prospective population based US military cohort study. British Medical Journal 336, 366–371.

uominen, K., Isometsa, E., Suokas, J., Haukka, J., Achte, K., et al., 2004. Completed suicide after a suicide attempt: a 37-year follow-up study. American Journal of Psychiatry 161, 562–563.

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

ekin, E., Markowitz, S., 2008. The effects of suicidal behavior on productive activities of young adults. Southern Economic Journal 75 (October (2)), 300–331.

ekin, E., Mocan, N.H., Liang, L., 2009. Do adolescents with emotional or behavioral problems respond to tobacco control policies? Southern Economic Journal 76 (July (1)), 67–85.

ice, J., 2010. Army considers shorter deployments. USA Today. hompson, M., 2010. Invisible wounds: mental health and the military. Time. .S. Department of Defense, 2012. Army Releases July Suicide Data. Press

Release, August, 16. Available at http://www.defense.gov/releases/release. aspx?releaseid=15517

ang, Z., Inslicht, S.S., Metzler, T.J., Henn-Haase, C., McCaslin, S.E., Tong, H., Neylan, T.C., Marmar, C.R., 2010. A prospective study of predictors of depression in police. Psychiatry Research 175, 211–216.

einer, J., Richmind, T.S., Conigliaro, J., Wiebe, D.J., 2011. Military veteran mortality following a survived suicide attempt. BMC Public Health 23 (11), 374.

ells, T.S., LeardMann, C.A., Fortuna, S.O., Smith, B., Smith, T.C., Ryan, M., Boyko, E., Blazer, D., 2010. A prospective study of depression following combat deploy-

ment in support of the wars in Iraq and Afghanistan. American Journal of Public Health 100 (1), 90–99.

ickrama, T., Wickrama, K.A.S., Baltimore, D.L., 2010. Adolescent precocious devel- opment and young adult health outcomes. Advances in Life Course Research 4 (15), 121–131.