International Review of Economics and Finance 79 (2022) 241–257

Contents lists available at ScienceDirect

International Review of Economics and Finance

journal homepage: www.elsevier.com/locate/iref

The effect of combat exposure on financial problems

Adam Ackerman a, *, Ben Porter b
a
United States Air Force Academy, USA
b
Leidos, Inc., Naval Health Research Center, USA

A R T I C L E I N F O A B S T R A C T

JEL classification: This paper examines whether combat exposure leads to financial problems among surviving
H56 deployed veterans. We use restricted panel data for the years 2001 through 2016 from 64,508
I32 deployed Millennium Cohort Study participants, and we accommodate real-world uncertainty
D14
with an information theoretic, semi-parametric Generalized Maximum Entropy model. The
D91
average predicted probability of developing a new major financial problem (such as bankruptcy)
Keywords:
and greater financial distress increases 0.44 percentage points (21 percent relative to the mean
Bankruptcy
probability) following a single combat exposure and increases 0.90 percentage points (43 percent
Financial distress
Combat exposure relative to the mean probability) following multiple combat exposures. Simulation results identify
Traumatic stress policy-relevant characteristics to target before a veteran deploys. The results point toward vet­
Veteran erans with poorer pre-deployment mental or physical health, veterans in enlisted ranks, and
veterans between the ages of 26 and 36 as being less resilient to the effects of combat exposure on
financial problems.

1. Introduction

Though over 2.7 million veterans received several hundred additional dollars from imminent danger pay and combat zone tax
exclusions for each month of post-September 11, 2001 deployment (Pleeter et al., 2011; Wenger et al., 2018), a recent study suggests
over a third of veterans may have encountered post-deployment financial problems meeting basic needs such as food, shelter, clothes,
transportation, social activities, or medical care (Elbogen et al., 2012).1 Furthermore, the study indicated financial problems are
positively associated with homelessness, criminal arrests, alcohol misuse, drug misuse, suicidal behavior, physical aggression, and
poor mental health (Elbogen et al., 2012). Veteran financial problems also create substantial risks for firms and the financial sector.
Combat exposure could have contributed to veteran financial problems as a result of neuropsychological degradation or behavioral
changes. Many combat veterans have experienced post-deployment migraine headaches as well as declines in memory and sustained
attention (Cesur et al., 2015; Vasterling et al., 2006). Additionally, combat exposure may lead to mental disorders (Cesur et al., 2013).
Almost one fourth of deployed veterans are estimated to have post-traumatic stress disorder, and more than one in ten are estimated to
have depression (Fulton et al., 2015; Gadermann et al., 2012). Veterans may employ risky or self-damaging strategies to escape

* Corresponding author.
E-mail addresses: [email protected] (A. Ackerman), [email protected] (B. Porter).
1
Research often uses different criteria to define a “veteran.” Elbogen et al. (2012) classify only individuals who separated from the US military
and reservists as veterans. Veteran can also refer to individuals serving in a specific military conflict (e.g. Vietnam Veteran). Our current study
examines all individuals who served in the US military, regardless of whether they continued military service or received an honorable or
dishonorable discharge, in support of the Global War on Terror. Our examination suggests policy implications affecting the needs of a broader
population.

https://doi.org/10.1016/j.iref.2022.02.042
Received 26 July 2020; Received in revised form 18 April 2021; Accepted 13 February 2022
Available online 15 February 2022
1059-0560/Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

emotional stress from combat exposure (US Institute of Medicine, 2013). Recent studies examining relationships between veteran
combat exposure and educational attainment, cigarette consumption, substance use, domestic violence, and divorce suggest a wide
range of non-pecuniary consequences that could harm economic well-being (Armey & Lipow, 2016; Cesur et al., 2016; Cesur & Sabia,
2016; Edwards, 2015).
While ongoing overseas conflicts have driven research interest in the costs of war, empirical limitations have led to an incomplete
picture of veteran economic well-being following combat exposure (US Institute of Medicine, 2013). An economic analysis of a
cross-sectional sample of 753 deployed and non-deployed veterans did not reveal an association between combat exposure and
reduced household income (Edwards, 2015). However, deployment- and disability-related compensation can make detecting changes
in spending, borrowing, and savings from income comparisons difficult, especially with a small sample. Furthermore, observable and
unobservable characteristics of veterans who deploy overseas differ significantly from similarly aged counterparts, both outside and
within the military, and empirical analyses of deployed veterans and non-deployed populations can be biased by pre-deployment
differences related to economic well-being (Armed Forces Health Surveillance Center, 2007).
The current study uses new data and estimation procedures to better understand how combat exposure affects the economic well-
being of surviving deployed veterans. We capture individual financial declines with restricted panel data for the years 2001 through
2016 from 64,508 veterans. Our measurement of financial decline includes objective and subjective economic well-being aspects to
bolster the policy relevance of our estimates (Ayllón & Fusco, 2017; Deaton, 2010). Our multinomial models account for a wide range
of observable pre-deployment and deployment factors related to the macro-economic environment, family, and individual. We also
used an information theoretic, semi-parametric Generalized Maximum Entropy (GME) estimator, which does not restrict the expected
error term to zero to account for real-world uncertainty. The approach reduces the sensitivity of estimates to noisy samples, infre­
quently observed outcomes, and highly correlated covariates (Golan, Judge, & Miller, 1996). Moreover, we leverage military-induced
variation in deployed veterans’ assignments, which we view as determining reported combat exposure, to address potential bias from
non-random selection on unobservable characteristics (Cesur et al., 2013). Finally, we investigate how effects of combat exposure vary
over different time periods and across veterans with different pre-deployment characteristics. The examination of individual char­
acteristics before deployment provides information relevant to proactive policy recommendations.
Our results show an assignment with combat exposure increases the predicted probability of a deployed veteran developing both
financial distress and a major financial problem. Furthermore, the estimated effect of combat exposure on financial decline increases
with multiple exposures and poorer pre-deployment health. Our estimates are robust to specifications with alternate combat exposure
measures and falsification tests. Back-of-the-envelope calculations suggest recent combat exposures added $41 million in short-term
productivity losses and as many as 3629 (1.34 per 1000 veterans) personal bankruptcies.
Our work creates a more complete picture of economic well-being following combat exposure. Similar to work suggesting combat
exposure increases the probability of homelessness (Ackerman et al., 2020), our analysis overcomes past empirical limitations and
reveals a policy-relevant aspect of economic disadvantage not previously discerned. Second, while a recent study suggests non-shared
environmental factors among siblings may explain more than a third of the variance in financial distress (Xu et al., 2017), under­
standing how specific environmental factors affect an individual’s financial outcomes is complicated by potential endogeneity. Our
analysis exploits plausibly exogenous military-induced variation in assignments, and, to our knowledge, we are the first to identify that
the policy-relevant environmental factor, combat exposure, may trigger a financial decline. Finally, we add evidence supporting a
strategy of health capital investments to increase economic resiliency to traumatic events. A wide body of economic literature doc­
uments positive effects of health capital on economic outcomes (Smith, 1999), and psychological literature suggests health capital can
improve psychological resilience to traumatic exposure (Bonanno et al., 2007). However, to the best of our knowledge, only one other
study has demonstrated health capital may mitigate the impact of traumatic exposure on economic well-being (Ackerman et al., 2020).

2. Data

The Millennium Cohort Study was designed to evaluate the effects of military exposures, and it is the largest prospective study ever
undertaken in the US military. The study enrolled US military personnel from all service branches and active duty, Reserve, and
National Guard components, and follows participants during service and after separation from the military. Questionnaires include
over 450 measures for overall well-being, military experiences, and other militarily relevant issues. Participants complete question­
naires via secure website or mail at approximately 3-year intervals between their baseline and 2016 surveys. Fig. 1 displays the study

Fig. 1. Millennium Cohort Study timeline.

242
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

timeline.
The Millennium Cohort Study currently has 201,619 participants who enrolled during four separate panels between 2001 and
2013. Panel 1 (July 2001–June 2003) enrolled 77,047 consenting participants drawn from a population-based sample of the October
2000 US military, with oversampling of females, those with previous deployment experience, and Reserve/Guard personnel. Panel 1
received five waves of questionnaires prior to 2016. Panels 2 and 3 sampled service members with 1–3 years of military service and
oversampled females and Marines. Panel 2 (June 2004–February 2006) enrolled 31,110 participants, and Panel 3 (June
2007–December 2008) enrolled 43,439 participants. Panel 2 received four waves of questionnaires prior to 2016, and Panel 3 received
three waves of questionnaires prior to 2016. Panel 4 sampled service members with 2–5 years of military service and oversampled
females and married service members. Panel 4 (April 2011–April 2013) enrolled 50,052 participants. Panel 4 received two waves of
questionnaires prior to 2016. Panel 5 is currently being collected. Over 70 percent of all participants completed at least one follow-up
questionnaire.
Questionnaires are only one part of Millennium Cohort Study data collection. The Millennium Cohort Study links questionnaires
with electronically maintained military datasets, including Defense Manpower Data Center (DMDC) information on deployments in
support of the post-9/11 conflicts. The military datasets provide information to supplement and validate many questionnaire items
(see pp 55–77 of Crum-Cianflone, 2013, for details on the study design and integration with military records). Multiple investigations
of the Millennium Cohort Study have found it to contain reliable data reporting with minimal non-response bias (Chretien et al., 2007;
LeardMann et al., 2007; Riddle et al., 2007; Ryan et al., 2007; Smith, Wingard, et al., 2007; 2007; Smith, Wingard, et al., 2007, 2007;
Wells et al., 2008). The comprehensiveness of the Millennium Cohort Study data makes it well suited for this study, and the longi­
tudinal nature enables evaluation of temporal sequences from combat exposure to different outcomes.
The available data are conditioned to represent only veterans likely eligible to experience combat and financial decline. To capture
veterans likely eligible to experience combat, sample observations include only those from 80,982 veterans with post-9/11 conflict
deployments in 2001 or later, a basic military entry date, and information to determine a veteran’s military component at the baseline
survey, and non-missing information on self-reported traumatic exposure at baseline and at least one follow-up. To capture the onset of
financial problems, sample observations are further restricted to those from 64,654 veterans not reporting the highest level of financial
distress or a major financial problem in the prior period. The sample also excludes observations from 146 veterans with missing or
invalid information for age, gender, race, or education. The data include multiple observations from the same veteran when available,
and 64,508 veterans, including 27,635 Panel 1 veterans, 10,694 Panel 2 veterans, 13,698 Panel 3 veterans, and 12,481 Panel 4
veterans, provide a total of 131,036 sample observations for analysis.
Four mutually exclusive states of financial decline are defined using a veteran’s survey responses on two items assessing financial
problems from two consecutive waves. One survey item asks veterans to indicate one of three levels of stress concerning financial
problems in the past 4 weeks, while the other survey item provides a more objective metric and asks veterans to indicate whether they
have suffered a major financial problem (such as bankruptcy) in the past 3 years. Financial decline is categorized into four levels by the
intersection of increased perceived stress (increased vs. steady/declined) and presence of major financial problems (present vs.
absent).
Though the use of two consecutive survey responses to create a single observation of financial decline is somewhat non-traditional,
the financial decline variable fits naturally with the Millennium Cohort Study’s staggered panels and is well suited to understand a
veteran’s response to combat exposure. The financial decline variable makes use of information from all available survey questions on
financial problems. Although the variable does not assess the prevalence of specific types of financial problems like bankruptcy and
foreclosure, it leverages the sequential structure of the data to capture individual, rather than average, changes in financial problems.
Considering the production of financial problems as a transition from one state of economic well-being to another, the financial decline
variable captures a first-order Markov process for economic well-being where an individual’s current state of economic well-being is
dependent on the individual’s previous state of economic well-being.
Table 1 describes the distribution of financial decline at each follow-up survey wave, with the sample size shown in the last row.
Approximately three fourths of observations fall into the state, “None”, which represents reporting neither increasing stress about
finances nor a new major financial problem since the prior survey wave. The three other states capture different financial declines since

Table 1
Distribution of financial decline, by survey wave.
State Definition Wave Wave Wave Wave Total
2 3 4 5

None No new distress or major problem 75.41 75.49 73.6 77.34 75.34
Distress Only Stress level (none, a little, or a lot) in the last 4 weeks from financial problems 21.23 21.17 22.03 19.67 21.17
reported at current survey > stress level in the last 4 weeks from financial problems
reported at prior survey. Did not also report new major problem.
Major Problem Major financial problems (such as bankruptcy) in the past 3 years reported at 1.32 1.25 1.68 1.32 1.37
Only current survey and not previous survey. Did not also report higher distress.

Major Problem & Reported new major problem and higher distress. 2.04 2.09 2.69 1.67 2.13
Distress

Observations 55,819 34,291 24,428 16,498 131,036

Source: 2001–2016 Millennium Cohort Study, Panels 1–4, authors’ unweighted computations.

243
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

the prior survey wave. The “Distress Only” state captures veterans reporting only increased financial stress in the 4 weeks prior to the
follow-up survey. The state is likely to capture financial decline in response to combat exposure; however, the increase in financial
distress may only reflect an expected increase in financial problems or a new minor financial problem. Approximately one in five
veterans at each wave report only increased financial distress. The “Major Problem Only” state captures veterans reporting only an
increase in past major financial problems in the past 3 years. The state may capture a major financial decline following a combat
exposure, but the information in the state alone is insufficient to determine whether the financial decline started prior to combat
exposure. A little more than 1 percent of veterans at each wave report a major problem only. The state, “Major Problem & Distress”,
captures new major financial problems such as bankruptcy with an increased level of financial distress, and about 2 percent of veterans
at each wave report the combined transition. The state is of principal interest because it captures a large financial decline following
combat exposure alongside increased levels of subjective financial distress. The states are nominal, rather than ordered, because the
expected relationship between each pair of states is unknown.

3. Empirical methods

Our estimation objective is to determine whether combat exposure affects financial decline. We explain background and motivation
for the use of GME multinomial models with the given data, and then we discuss assumptions needed to identify the effect of combat
exposure on financial decline. After outlining support for identifying assumptions, we present the framework to assess the effect of
combat exposure on financial decline.

3.1. Generalized Maximum Entropy estimation

Conceptual econometric models have unobserved parameters and noise components intended to reflect the true data generation
process. Real world samples with small random changes in individual financial transitions over time, infrequently observed outcomes,
and highly correlated covariates contribute to uncertainty when estimating parameters and making inferences from available infor­
mation. The GME approach builds on information theory and the Maximum Entropy (ME) principle to analyze problems with limited
and complex data.
The GME approach uses an objective function that is the sum of the coefficient distribution entropy and the error distribution
entropy. In information theory, Shannon’s entropy is a measure reflecting the expected information content of an outcome, when
considering all possible outcomes. If a probability distribution has only one possible outcome, then the distribution entropy is zero. On
the other hand, if a distribution has uniform probabilities for multiple outcomes, then the distribution entropy is maximum. The ME
approach to the multinomial choice problem selects among all possible coefficient distributions the one that maximizes uncertainty
remaining in the distribution while adhering to observed sample moment constraints. In this manner, ME estimation uses the limited
observed information to provide the most noncommittal option while avoiding parametric modeling assumptions. The GME approach
advances the ME approach by not restricting expected errors to zero, thus accounting for unknown disturbances that impose challenges
with Maximum Likelihood (ML) logit and Maximum Entropy multinomial estimators (Golan, Judge, & Perloff, 1996). The GME
multinomial estimator, in all finite samples, outperforms traditional Maximum Likelihood (ML) logit and Maximum Entropy multi­
nomial estimators in terms of efficiency and robustness (Golan, Judge, & Miller, 1996; 1996b). Traditional likelihood-based inference
tests still apply to GME multinomial estimates as ML logit estimates are equal to GME multinomial estimates in cases where all errors
are actually zero or the sample size approaches infinity. GME estimations are performed with NLOGIT 6.0 software (Greene, 2016).
Information theoretic methods of inference fall into an info-metrics framework, and Golan (2018) provides in-depth discussion of
info-metric foundations and widespread applications.

3.2. Identification

Combat exposure itself is best thought of as a random process resulting from military factors, including occupational group,
deployment location, and deployment timing, plausibly exogenous to an individual’s risk of financial decline. Much work has been
conducted identifying variable risk of location on risk of combat exposure. The estimation strategy identifies an effect of combat
exposure on financial decline that is conditional only on a few observable military characteristics, specifically an individual’s military
component, branch, rank, occupation, and timing of service. Identification requires two assumptions. First, combat exposure needs to
result from a conditionally random assignment process. Variation in combat exposure is known to exist with different deployment
assignments; recent killed- or wounded-in-action rates per thousand deployed members varied from 0 to nearly 13 across countries in a
combat zone (Pleeter et al., 2011). Similar variation in combat exposure also occurs within countries. Second, deployment assignments
and experiences, conditional on only observable military characteristics, should be exogenous to individual pre-deployment charac­
teristics. In other words, non-military individual characteristics that could determine financial problems should not explain variation
in deployment assignments and experiences.
Several researchers present a compelling argument for individual deployment assignments not being driven by forces related to
non-military individual characteristics (Cesur et al., 2013, 2015, 2016; Ackerman et al., 2020; Cesur & Sabia, 2016; Engel et al., 2010;
Loughran & Heaton, 2013; Lyle, 2006). The premise of the argument stems from evolving overseas events and military unit availability
driving the military deployment process. As overseas military operations normally involve large, complex and coordinated actions, it is
not practical for the military to task a single veteran to handle all requirements. The hierarchal nature of the military normally results
in units being tasked to conduct overseas operations, and the military, as a rule, does not use non-military personal characteristics (e.g.

244
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

number of dependents, risk-taking preferences, past financial behavior) when making deployment selections. Instead, veterans with
the same branch, rank, and occupation are viewed as interchangeable in the deployment process. See Lyle (2006) for the seminal
explanation of military deployment process assumptions, and reference Cesur et al. (2013) for the extension of the argument to assess
the experience of combat exposure.
Researchers also use data to support the conditional random assignments assumption. Closely following past studies of combat
exposure leveraging variation induced by the deployment assignment process, the model examining the conditional random assign­
ments assumption takes the form of an experiment consisting of N trials. The probabilities of a set of binary random variables, Assign1j ,
Assign2j ,…,AssignNj , which equal one if combat exposure category j (j = 1, 2, …, J) is observed for trial i (i = 1, 2,…,N), are related to a
set of explanatory variables with the equation:
( ⃒ ) ( )
(1)
′ ′
pij = Pr Assignij = 1⃒Mi , Xi = F β0j + β1j Mi + β2j Xi > 0
∑
subject to pij = 1. The three possible categories of combat exposure are no combat exposure, single combat exposure, and
j

multiple combat exposure. Combat exposure is primarily determined by a veteran indicating personal exposure to witnessing a per­
son’s death due to war, disaster, or tragic event in the past 3 years. In 150 observations where a veteran’s primary indication of combat
exposure is missing, combat exposure is determined by a veteran indicating seeing Americans who were seriously injured or killed
during deployment. Of veterans in the sample with both combat exposure measures, more than 76 percent of veterans indicating
combat exposure with the alternate measure also indicate combat exposure with the primary measure. While the alternate deployment
specific measure may be preferable, the Millennium Cohort Study did not ask about deployment specific exposure until the 2007–2008
questionnaire. To address misclassification concerns with the primary measure and improve estimates of combat exposure, a binary
variable indicating another tragic loss, defined by the death of serious illness of a loved one or family member in the past 3 years, is
included in all models. Other research with the Millennium Cohort Study indicates 98.4 percent of individuals endorsing this proxy for
combat exposure endorse additional specific measures of combat exposure (Porter et al., 2018). Additionally, while geographic
deployment location data may permit identification of risk in a particular time period, the period of the current study spans such a wide
range that the level of combat exposure associated with deployment to specific locations varies over time. For example, between 2007
and 2010, killed in action rates in Afghanistan increased seven fold while killed in action rates in Iraq decreased by two thirds (Pleeter

Table 2
Description of combat exposure, deployed time, and military control variables.
Variable Description Mean Std. Dev.

Combat exposure Reference group is no reported combat exposure

Single exposure Equal to 1 if only one reported combat exposure, 0 otherwise 0.06 0.23
Multiple exposures Equal to 1 if two or more reported combat exposures, 0 otherwise 0.12 0.32
Deployed time Reference group is no reported deployed time in period
Deployed 1–179 days Equal to 1 if 1–179 days deployed, 0 otherwise 0.22 0.41
Deployed 180–364 days Equal to 1 if 180–364 days deployed, 0 otherwise 0.20 0.40
Deployed 365 days or more Equal to 1 if 365 days or more deployed, 0 otherwise 0.06 0.24
Service entry Reference group is initial service in 2001 or earlier
Post-2001 entry Equal to 1 if initial service in 2002 or later, 0 otherwise 0.37 0.48
Years of service Reference group is less than 7 years of military service
7–12 Equal to 1 if years of military service ≥7 and < 13, 0 otherwise 0.21 0.41
v13–18 Equal to 1 if years of military service ≥13 and < 19, 0 otherwise 0.17 0.37
19 or more Equal to 1 if years of military service ≥19, 0 otherwise 0.23 0.42
Rank Reference group is enlisted
Officer Equal to 1 if officer rank either W1 or O1 or higher, 0 otherwise 0.30 0.46
Branch of service Reference group is Army
Navy Equal to 1 if in the Navy or Coast Guard, 0 otherwise 0.15 0.36
Air Force Equal to 1 if in the Air Force, 0 otherwise 0.32 0.47
Marine Corps Equal to 1 if in the Marine Corps, 0 otherwise 0.06 0.25
Component Reference group is active duty
Reserve or National Guard Equal to 1 if in Reserve or National Guard, 0 otherwise 0.37 0.48
Separated Equal to 1 if separated or unknown component, 0 otherwise (some separated respondents return to service) 0.09 0.28
Occupational classification Reference group is combat specialist
Electrical repair Equal to 1 if electrical repair occupation, 0 otherwise 0.09 0.29
Comm/intel Equal to 1 if communications/intelligence occupation, 0 otherwise 0.09 0.29
Health care Equal to 1 if health care occupation, 0 otherwise 0.10 0.30
Other technical Equal to 1 if other technical or specialty occupation, 0 otherwise 0.03 0.16
Functional support Equal to 1 if functional support occupation, 0 otherwise 0.18 0.38
Equipment repair Equal to 1 if equipment repair occupation, 0 otherwise 0.14 0.34
Craft workers Equal to 1 if craft worker occupation, 0 otherwise 0.03 0.17
Service and supply Equal to 1 if service and supply occupation, 0 otherwise 0.10 0.30
Other Equal to 1 if other occupation, 0 otherwise 0.04 0.19

Observations 131,036

DMDC = Defense Manpower Data Center.

Table displays unweighted means of variables generated using survey responses and DMDC data at the wave prior to assessing financial decline.

245
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

et al., 2011). In the current study, veterans reported a single combat exposure in the past 3 years 6 percent of the time, and veterans
reported more than one combat exposure in the past 3 years 12 percent of the time.
The military vector M includes an individual’s service entry, years of service, rank, branch, component, and military occupation at
the wave prior to the assessment of financial decline. Table 2 provides variable definitions and descriptive statistics for combat
exposure, deployed time and the vector M.
The vector X includes individual and family background variables at the survey prior to deployment assignment that could be
associated with combat assignment or financial decline. These variables represent past financial distress, gender, age, race, ethnicity,
education, family structure, physical health, mental health, health behaviors, and adverse childhood experiences. Additionally, vector
X includes macroeconomic measures of the real gross domestic product (GDP) annual growth rate, real military pay annual growth
rate, and respective 3 year lags for the survey year. Table 3 provides variable definitions and descriptive statistics for vector X.
The scalars β0j represent the function intercepts, the coefficient vectors β1j measure whether military observables explain combat
′

exposure, and the coefficient vectors β2j measures whether background variables explain combat exposure. F(∙) is the function linking
′

the probability with the covariates for an individual. If combat exposure is exogenous to an individual’s financial decline, then the

Table 3
Description of non-military background variables.
Variable Description Mean Std.
Dev.

Lag financial distress Equal to 1 if reported stress from financial problems, 0 otherwise 0.33 0.47
Female Equal to 1 if female gender, 0 otherwise 0.24 0.43
Age Age in years measured at follow-up survey 37.40 9.32
Age2 Age in years × Age in years measured at follow-up survey 1485.70 764.17
Race & ethnicity Reference group is white, non-Hispanic
Black Equal to 1 if black race and non-Hispanic ethnicity, 0 otherwise 0.09 0.29
Other Equal to 1 if other race and non-Hispanic ethnicity, 0 otherwise 0.11 0.31
Hispanic Equal to 1 if Hispanic ethnicity, 0 otherwise 0.07 0.25
Education Reference group is high school or less
Some college Equal to 1 if only some college education, 0 otherwise 0.48 0.50
College degree Equal to 1 if bachelor’s degree or higher, 0 otherwise 0.40 0.49
Family structure
Dependents Equal to number of dependents in military record 1.54 1.51
Last known value used if value missing at prior survey
Recent divorce Equal to 1 if divorced or separated in past 3 years (or ever if baseline survey), 0 otherwise 0.13 0.33
Health
PCS Physical health component summary* 54.13 6.72
MCS Mental health component summary* 52.48 8.47
Smoking Equal to 1 if past year smoked cigarettes (ever smoked 100 cigarettes if baseline survey), 0 otherwise 0.28 0.45
Heavy drinking Equal to 1 if 15 (8) drinks or more during a typical week for men (women), 0 otherwise 0.07 0.25
Childhood Experiences
Neglect Equal to 1 if respondent indicated grown-ups did not take care of them in the way that they should before the age 0.07 0.25
of 18, 0 otherwise. Examples of neglect in question include respondent not getting enough food, receiving
medical care when sick, and having a safe place to stay.
Verbal abuse Equal to 1 if respondent indicated grown-ups made them feel scared or really bad by saying mean things to them 0.18 0.39
before the age of 18, 0 otherwise
Physical abuse Equal to 1 if respondent indicated grown-ups hit (not including spanking on the bottom), beat, kicked, or 0.25 0.43
physically hurt them before the age of 18, 0 otherwise
Sexual abuse Equal to 1 if respondent indicated grown-ups touched private parts when they shouldn’t have, made them touch 0.08 0.27
their private parts, or forced them to have sex, 0 otherwise
Macroeconomic
GDP growth Real US GDP annual growth rate at follow-up survey year (Source: US Department of Commerce, 2017) 2.21 0.89
Lag GDP growth GDP growth 3 years prior to follow-up survey year 1.37 1.72
Military pay growth Real US military pay annual growth rate at follow-up survey year. (Source: Defense Finance and Accounting 0.48 1.34
Service, 2017)
Lag military pay Military pay growth 3 years prior to survey follow-up year 1.08 1.19
growth

Observations 131,036

DMDC = Defense Manpower Data Center, GDP = gross domestic product.

Table displays unweighted means of variables generated using survey responses and DMDC data at the wave prior to assessing financial decline unless
otherwise noted. Description of dummy variables for missing family structure, health, health behaviors, and childhood variable responses not dis­
played.
*The MCS and PCS are standard health instruments commonly used in veteran studies. Higher scores imply better health. The scores are derived from
36 ordinal scale questions in eight health domain: physical functioning, role limitations due to physical problems, bodily pain, general health per­
ceptions, energy/vitality, social functioning, role limitations due to emotional problems, and mental health. The transformation of the questions into
the uncorrelated component summary scales is accomplished using weights derived from a national probability sample of the US population, and the
scales are standardized so the MCS and PCS have a direct interpretation to a US population mean of 50 and standard deviation of 10 (Hays et al., 1993;
Kazis et al., 2004).

246
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

coefficient vectors β2j should not help explain combat exposure. To make model interpretation more tangible, results display the
′

estimated marginal effect of each independent variable on the overall probability of observing a category. The marginal effect for each
continuous variable is calculated with all independent variables set to the mean. The marginal effect for each categorical variable is

Table 4
Marginal effects of background characteristics on combat exposure probabilities.
(1) (2) (3)

Combat Exposure

None Single Multiple

Variables Effect SE Effect SE Effect SE

Military
Service Entry
Post-2001 entry − 0.0041 0.0032 0.0000 0.0020 0.0041 0.0025
Years of service
7–12 0.0146 0.0032 − 0.0051 0.0021 − 0.0095 0.0025
13–18 0.0056 0.0042 0.0003 0.0028 − 0.0060 0.0032
19 or more 0.0180 0.0049 0.0002 0.0033 − 0.0182 0.0038
Rank
Officer 0.0009 0.0033 − 0.0050 0.0022 0.0042 0.0026
Branch of service
Navy 0.1063 0.0024 − 0.0301 0.0016 − 0.0762 0.0019
Air Force 0.0261 0.0038 − 0.0147 0.0026 − 0.0115 0.0028
Marine Corps 0.1058 0.0032 − 0.0239 0.0020 − 0.0820 0.0026
Component
Reserve or National Guard 0.0221 0.0022 − 0.0028 0.0015 − 0.0193 0.0017
Separated 0.0902 0.0048 − 0.0222 0.0030 − 0.0680 0.0039
Occupational classification
Electrical repair 0.1062 0.0042 − 0.0209 0.0026 − 0.0853 0.0034
Comm/intel 0.0396 0.0037 − 0.0147 0.0025 − 0.0249 0.0027
Health care − 0.0669 0.0034 0.0050 0.0025 0.0619 0.0024
Other technical 0.0166 0.0059 − 0.0013 0.0039 − 0.0152 0.0045
Functional support 0.0778 0.0033 − 0.0157 0.0021 − 0.0621 0.0026
Equipment repair 0.0924 0.0037 − 0.0140 0.0023 − 0.0784 0.0030
Craft workers 0.0583 0.0062 − 0.0089 0.0039 − 0.0494 0.0050
Service and supply 0.0311 0.0035 − 0.0030 0.0023 − 0.0281 0.0026
Other 0.0166 0.0048 − 0.0013 0.0032 − 0.0152 0.0036
Non-Military
Financial distress 0.0057 0.0021 − 0.0003 0.0014 − 0.0054 0.0016
Female 0.0719 0.0022 − 0.0143 0.0015 − 0.0576 0.0016
Age 0.0007 0.0002 − 0.0006 0.0002 − 0.0001 0.0002
Race
Black 0.0236 0.0032 0.0054 0.0023 − 0.0290 0.0023
Other 0.0067 0.0030 0.0004 0.0021 − 0.0071 0.0022
Hispanic 0.0090 0.0037 0.0037 0.0025 − 0.0127 0.0027
Education
Some college − 0.0129 0.0031 0.0011 0.0019 0.0118 0.0024
College degree − 0.0034 0.0041 − 0.0019 0.0026 0.0053 0.0033
Family structure
Dependents − 0.0004 0.0008 0.0004 0.0005 0.0000 0.0006
Divorce − 0.0191 0.0030 0.0068 0.0019 0.0124 0.0024
Health
PCS 0.0016 0.0001 − 0.0005 0.0001 − 0.0011 0.0001
MCS 0.0009 0.0001 − 0.0004 0.0001 − 0.0005 0.0001
Smoking − 0.0139 0.0023 0.0061 0.0015 0.0078 0.0018
Heavy drinking − 0.0062 0.0039 − 0.0040 0.0023 0.0102 0.0031
Childhood experiences
Neglect − 0.0204 0.0044 0.0048 0.0027 0.0156 0.0035
Verbal abuse − 0.0103 0.0031 0.0059 0.0020 0.0044 0.0024
Physical abuse − 0.0276 0.0028 0.0081 0.0018 0.0196 0.0022
Sexual abuse − 0.0038 0.0039 0.0023 0.0025 0.0015 0.0030

Observations 131,036
Pseudo-R2 0.52
Pseudo-R2 (military only) 0.51

GME = Generalized Maximum Entropy, MCS, mental health component summary, PCS = physical health component summary, SE = standard error.
Columns 1–3 display GME marginal effects for multinomial combat exposure assignment model. The marginal effect for each continuous variable is
calculated with all independent variables set to the mean. The marginal effect for each categorical variable is calculated by changing all observations
from the reference value to the category value with all other independent variables set to the mean. All columns include macroeconomic variables in
Table 3. All columns include dummy variables for missing values of family structure, health, health behaviors, and childhood experiences.

247
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

calculated by changing all observations from the reference value to the category value with all other independent variables set to the
mean.
Table 4 reveals support for the assumption of conditional random assignments with model fit statistics and estimated marginal
effects of individual non-military variables on the different levels of combat exposure. The Pseudo-R2 measure captures the goodness of
fit between the sample information and the model on an increasing scale of 0–1 by comparing the (maximized) entropy of the fitted

Table 5
Marginal effects of deployment-related variables on financial decline.
(1) (2)

Military Only Controls Military and Non-Military Controls

Effect SE Effect SE

Probability (Y = None)
Combat exposure (reference is none)
Single − 0.0338 0.0053 − 0.0260 0.0051
Multiple − 0.0395 0.0041 − 0.0306 0.0039
Weighted average of exposure* − 0.0377 0.0034 − 0.0291 0.0033
Deployed time (reference is none)
Deployed 1–179 days 0.0230 0.0033 0.0231 0.0031
Deployed 180–364 days 0.0309 0.0032 0.0267 0.0032
Deployed 365 days or more 0.0450 0.0049 0.0437 0.0047
Weighted average of deployed time* 0.0291 0.0026 0.0273 0.0025
Disabling illness or injury − 0.1362 0.0059 − 0.1029 0.0058
Other tragic loss − 0.0317 0.0024 − 0.0291 0.0024
Probability (Y ¼ Distress Only)
Combat exposure (reference is none)
Single 0.0266 0.0051 0.0215 0.0049
Multiple 0.0249 0.0039 0.0193 0.0038
Weighted average of exposure* 0.0254 0.0032 0.0200 0.0032
Deployed time (reference is none)
Deployed 1–179 days − 0.0185 0.0030 − 0.0207 0.0029
Deployed 180–364 days − 0.0221 0.0030 − 0.0206 0.0029
Deployed 365 days or more − 0.0311 0.0046 − 0.0322 0.0043
Weighted average of deployed time* − 0.0220 0.0025 − 0.0227 0.0024
Disabling illness or injury 0.0868 0.0055 0.0681 0.0055
Other tragic loss 0.0233 0.0023 0.0231 0.0023
Probability (Y ¼ Major Problem Only)
Combat exposure (reference is none)
Single 0.0016 0.0013 0.0005 0.0009
Multiple 0.0043 0.0011 0.0025 0.0008
Weighted average of exposure* 0.0034 0.0009 0.0018 0.0007
Deployed time (reference is none)
Deployed 1–179 days − 0.0017 0.0007 − 0.0007 0.0006
Deployed 180–364 days − 0.0028 0.0007 − 0.0015 0.0006
Deployed 365 days or more − 0.0035 0.0010 − 0.0021 0.0008
Weighted average of deployed time* − 0.0025 0.0006 − 0.0012 0.0005
Disabling illness or injury 0.0147 0.0018 0.0090 0.0013
Other tragic loss 0.0029 0.0006 0.0015 0.0005
Probability(Y ¼ Major Problem & Distress)
Combat exposure (reference is none)
Single 0.0062 0.0016 0.0044 0.0015
Multiple 0.0107 0.0013 0.0090 0.0013
Weighted average of exposure* 0.0089 0.0010 0.0074 0.0010
Deployed time (reference is none)
Deployed 1–179 days − 0.0021 0.0008 − 0.0012 0.0008
Deployed 180–364 days − 0.0052 0.0007 − 0.0038 0.0008
Deployed 365 days or more − 0.0086 0.0008 − 0.0078 0.0009
Weighted average of deployed time* − 0.0046 0.0007 − 0.0033 0.0007
Disabling illness or injury 0.0347 0.0023 0.0258 0.0021
Other tragic loss 0.0055 0.0006 0.0046 0.0007

Observations 131,036 131,036

Pseudo-R2 0.52 0.55

GME = Generalized Maximum Entropy, SE = standard error.

GME marginal effects for multinomial models displayed. The marginal effect for each continuous variable is calculated with all independent variables
set to the mean. The marginal effect for each categorical variable is calculated by changing all observations from the reference value to the category
value with all other independent variables set to the mean. Column 1 includes military control variables in Table 2 and macroeconomic variables in
Table 3. Column 2 includes column 1 variables and all other background variables in Table 3 and dummy variables for missing values of family
structure, health, health behaviors, and childhood experiences. *Weighted average of unspecified switch from reference category computed using
sample shares of each category.

248
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

model with the entropy of the uniform distribution (Golan, 2018). The inclusion of all individual non-military background variables in
the model increases respective Pseudo-R2 goodness-of-fit measures by only one hundredth relative to the model without individual
non-military background variables. The measures indicate non-military background characteristics, as a whole, do little to explain
dispersion in combat exposure. While columns 1 through 3 show some military background variables can influence assignment
probabilities by 10 percent, columns 1 through 3 also show each individual non-military background variable on the assignment
probabilities to combat exposure is generally no more than about 2 percent. Gender is the only consistent exception, and the results
indicate female service members are less likely to be assigned to a war environment or combat exposure. A reasonable explanation in
line with exogenous selection is previous policies limiting the role of females in combat (see Cesur et al., 2016, for full discussion).
Although the potential for endogenous selection cannot be completely ruled out, the results, taken together, provide support for the
conditional random assignments assumption.

3.3. Modeling the effect of combat exposure on financial decline

The model to determine the conditional effect of reported combat exposure on financial decline takes the form of an experiment
consisting of N trials. The probabilities of a set of binary random variables, Y1j , Y2j , …, YNj , which equal one if state j (j = 1, 2, 3, 4)
corresponding to None, Distress Only, Major Problem Only, and Major Problem & Distress is observed for trial i (i = 1, 2,…,N), are related
to explanatory variables with the equation:
( ⃒ ) ( )
(2)
′ ′ ′
pij = Pr Yij = 1⃒Mi , Xi = F δ0j + δ1j D + π1j M + π2j X > 0
∑
subject to pij = 1. F(∙) is the function linking the probability of a state with the covariates for an individual. The scalars δ0j are the
j
function intercepts. The vector D includes deployment experiences including the relevant combat exposure and time spent deployed,
and the coefficient vectors δ1j show how vector D is associated with the probability of developing financial problems. The vectors M
′

and X are as previously described, and coefficient vectors π1j and π2j explain how the variables in M and X relate to the financial
′ ′

decline. The validity of the exogeneity assumption can be further assessed by estimating equation (2) without the individual non-
military background variables in vector X. If the exogeneity assumption is valid, then the marginal effects of the combat exposure
should be stable.

4. Results

Table 5 provides estimated marginal effects of combat exposure and other deployment-related variables on financial decline. After
comparing models with and without individual non-military background variables, the possibility of the effect of combat exposure
extending into other survey periods is examined. Graphical analysis examines predicted variation in the effect of combat exposure on
an increase in both a major problem and distress with background characteristics. Lastly, estimates with alternate measures of combat
exposure demonstrate the robustness of findings.
Table 5 results suggest detrimental effects to economic well-being from combat exposure. For Table 5, the model in column 1 with
only military variables and the model in column 2 with both military and other control variables lead to the same conclusions. The
estimated marginal effects of combat exposure on the state of no financial decline are negative, and the estimated marginal effects of
combat exposure on an increase in financial distress only, an increase in a major financial problem only, and an increase in distress and
a major problem are consistently positive. A single combat exposure increases the probability of developing a major problem and
distress by about half a percentage point, and multiple combat exposures increase the probability of developing a major problem and
distress by about one percentage point. The influence of combat exposure is substantial relative to the 2.1 percent average predicted
probability of developing a major problem and distress.
The inclusion of individual non-military background variables in Table 5 has little influence on the magnitude of findings, and a
comparison of estimates provides further support for the conditional independence of combat exposure to personal characteristics. An
informal gauge of the strength of likely bias from unobservable characteristics can be calculated by dividing the marginal effect of
combat exposure in column 2 by its difference with the marginal effect in column 1 (Altonji et al., 2005). Using the conditional share
weighted average effect of combat exposure on the probability of developing a major problem and distress, the computed ratio implies
selection on unobservable characteristics would need to be nearly five times greater than selection on non-military background
variables, including the previous level of financial distress, to explain away the estimated effect. The Pseudo-R2 measures indicate
including non-military background variables improves goodness of model fit, thus models including non-military background vari­
ables represent preferred specifications.
Combat exposure is not the only deployment-related factor influencing the probability of financial decline. Table 5 shows deployed
time reduces the probability of financial decline. The estimate of the conditional share weighted average effect of deployed time on
developing a major problem and distress is similar in magnitude to the conditional share weighted average effect of combat exposure.
The results suggest deployment time, possibly due to temporary increases in pay and benefits, can help counterbalance harmful effects
of combat exposure. On the other hand, physical illness or disability during the survey period, results increase the estimated proba­
bility of developing a major problem and distress by more than 2 percentage points. The influence of other tragic losses involving a
veteran’s family or loved ones on financial decline appears similar to influence of combat exposure on financial decline, and the results
suggest inclusion of other tragic losses in the model helps reduce possible conflating of estimates of combat and non-combat traumatic

249
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

experiences that could be captured by the defined combat exposure variable. As the primary interest is in the consequences of combat
experiences, the influence of all background variables on financial decline is described in Table A1.
To evaluate whether any effects of combat exposure are delayed or anticipatory, combat exposures reported in other survey periods
are included in the model. Columns 1 and 2 in Table 6 reveal that any influence of combat exposure from the prior survey periods on

Table 6
Marginal effects of combat exposure over time on financial decline.
(1) (2) (3)

Effect SE Effect SE Effect SE

Probability (Y=None)
Combat Exposure
Single − 0.0251 0.0052 − 0.0312 0.0076
Multiple − 0.0286 0.0041 − 0.0344 0.0060
1-Period Lag Combat Exposure
Single − 0.0091 0.0043 − 0.0091 0.0066
Multiple − 0.0064 0.0035 − 0.0084 0.0052
2-Period Lag Combat Exposure
Single − 0.0083 0.0057
Multiple 0.0051 0.0047
1-Period Lead Combat Exposure
Single − 0.0014 0.0064
Multiple 0.0101 0.0046
Probability (Y ¼ Distress Only)
Combat Exposure
Single 0.0205 0.0050 0.0265 0.0073
Multiple 0.0175 0.0039 0.0226 0.0057
1-Period Lag Combat Exposure
Single 0.0069 0.0041 0.0075 0.0064
Multiple 0.0056 0.0034 0.0064 0.0050
2-Period Lag Combat Exposure
Single 0.0075 0.0055
Multiple − 0.0060 0.0045
1-Period Lead Combat Exposure
Single 0.0023 0.0064
Multiple − 0.0120 0.0046
Probability (Y ¼ Major Problem Only)
Combat Exposure
Single 0.0005 0.0010 − 0.0006 0.0011
Multiple 0.0025 0.0009 0.0027 0.0011
1-Period Lag Combat Exposure
Single − 0.0005 0.0008 − 0.0003 0.0010
Multiple − 0.0001 0.0006 − 0.0002 0.0008
2-Period Lag Combat Exposure
Single − 0.0003 0.0009
Multiple 0.0010 0.0008
1-Period Lead Combat Exposure
Single − 0.0008 0.0015
Multiple 0.0003 0.0012
Probability(Y ¼ Major Problem & Distress)
Combat Exposure
Single 0.0041 0.0014 0.0061 0.0022
Multiple 0.0086 0.0013 0.0091 0.0019
1-Period Lag Combat Exposure
Single 0.0027 0.0011 0.0019 0.0017
Multiple 0.0009 0.0009 0.0022 0.0013
2-Period Lag Combat Exposure
Single 0.0011 0.0014
Multiple − 0.0001 0.0011
1-Period Lead Combat Exposure
Single − 0.0001 0.0014
Multiple 0.0016 0.0012

Observations 130,950 70,916 72,100

Pseudo-R2 0.55 0.55 0.57

GME = Generalized Maximum Entropy, SE = standard error.

GME marginal effects for multinomial models displayed. The marginal effect for each continuous variable is calculated with all independent variables
set to the mean. The marginal effect for each categorical variable is calculated by changing all observations from the reference value to the category
value with all other independent variables set to the mean. All columns include military control variables in Table 2, all other deployment-related
variables in Table 5, all other background variables in Table 3, and dummy variables for missing values of family structure, health, health behav­
iors, and childhood experiences.

250
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

the probability of developing financial problems is rather limited, and the results consistently find combat exposure in the most recent
3-year survey period associated with a decline in economic well-being. If combat exposure was correlated with pre-deployment un­
observables that contribute to financial decline, then the one period lead of combat exposure in the model would likely be significant.
Instead, the column 3 results serve as evidence of a falsification test supporting the identification strategy.
Groups of veterans with different background characteristics may be more or less resilient to combat exposure. While the marginal
effects computed with all other variables set to sample means provide insight at one important point, estimates with a range of values
for background characteristics help us understand how the predicted effect of combat exposure on development of both a major
problem and distress changes. Figs. 2 and 3, simulated using parameters estimated for Table 5 model specification with all background
variables, show ‘what if’ changes in the effects of combat exposure when veterans have different ranks, ages, and health. Unobservable
differences may also exist among some groups of veterans in the full sample; however, estimates, displayed graphically in Figure A1,
from samples divided by gender, samples divided by military service component, samples divided by combat specialist and non-
combat specialist occupations, samples divided by military service entry period, and samples divided by prior financial stress
appear consistent with estimates from the sample with all groups.
Fig. 2 reveals predicted variation in the effect of combat exposure on financial decline over rank and age. The variation in the effects
exhibit similar patterns for single and multiple exposures, but the variation is more pronounced for multiple exposures. As higher
military rank and years of service equate to higher military pay that can provide a financial buffer, it is not surprising the oldest officers
are the least vulnerable to effects of combat exposure. Interestingly, the veterans most vulnerable to effects of combat exposure are not
the youngest enlisted veterans, but enlisted veterans around 30 years old. Possible explanations for the finding include the youngest
veterans having fewer responsibilities and the military placing more restrictions on the youngest veterans (e.g., unable to live off base).
Fig. 3 reveals veterans with poorer pre-deployment self-reported health scores appear more vulnerable to the effects of combat
exposure on financial decline. The variation in the effects exhibit similar patterns for combat exposure frequency and health category,
but the variation is more pronounced for multiple exposures and mental health. Each 10-point, or 1 standard deviation for the US
population, mental component score decline translates to the marginal effect of multiple combat exposure on the probability of
developing a financial stress and a major financial problem increasing roughly 0.3 percentage points. The findings suggest increased
health capital, a non-monetary resource, can lessen the impact of traumatic exposure on financial decline.

Fig. 2. Marginal effects of combat exposure by rank at ages on probability (Major Problem & Distress). Generalized Maximum Entropy marginal
effects for categorical values of combat exposure and rank computed by setting all observations to category value and comparing to respective base
value with age set to values 21, 26, …, 56, 61, and all other variables set to sample means. Model includes military control variables in Table 2, all
deployment-related variables in Table 6, all other background variables in Table 3, and dummy variables for missing values of family structure,
health, health behaviors, and childhood experiences.

251
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

Fig. 3. Marginal effects of combat exposure at health scores on probability (Major Problem & Distress). Generalized Maximum Entropy marginal
effects for categorical values of combat exposure computed by setting all observations to category value and comparing to base value with health
score set to values 30, 35, …,65, 70, and all other variables set to sample means. Model includes military control variables in Table 2, all other
background variables in Table 3, and dummy variables for missing values of family structure, health, health behaviors, and childhood experiences.

4.1. Additional robustness checks

Alternate measures of combat exposure were substituted into the model to evaluate the robustness of the main finding, the marginal
effect of combat exposure on the probability of developing financial distress and a major financial problem. The 2007 survey intro­
duced questions regarding the frequency that veterans experienced enemy attacks or ambushes and improvised explosive devices
during deployments in the past 3 years. Columns 2 and 3 of Table 7 display estimates for the alternate combat exposure measures,
created from the 2007 and later survey question responses, that are similar to the estimates of the primary measure of combat exposure
in column 1.
The 2011 survey introduced a question enabling another falsification test of the main finding. Specifically, the survey asked
veterans about the frequency of exposure to smoke from burning trash or feces during deployments in the past 3 years. Smoke exposure

Table 7
Marginal effects of other combat measures on probability (major problem & distress).
(1) (2) (3) (4) (5)

Effect SE Effect SE Effect SE Effect SE Effect SE

Combat exposure
Single 0.0039 0.0018 0.0017 0.0022
Multiple 0.0094 0.0015 0.0087 0.0020
Other measures
Attacked
Single 0.0049 0.0019
Multiple 0.0069 0.0011
IED
Single 0.0032 0.0019
Multiple 0.0067 0.0015
Smoke
Single 0.0003 0.0033 0.0003 0.0032
Multiple 0.0037 0.0012 0.0021 0.0012

Observations 57,221 57,221 57,221 32,516 32,516

Pseudo-R2 0.56 0.56 0.56 0.55 0.55

GME = Generalized Maximum Entropy, IED = improvised explosive device, SE = standard error.
GME marginal effects for multinomial models displayed. Estimates for None, Distress Only, and Major Problem Only states not displayed. The marginal
effect for each continuous variable is calculated with all independent variables set to the mean. The marginal effect for each categorical variable is
calculated by changing all observations from the reference value to the category value with all other independent variables set to the mean. All
columns include military control variables in Table 2, all other deployment-related variables in Table 5, all other background variables in Table 3 and
dummy variables for missing values of family structure, health, health behaviors, and childhood experiences.

252
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

from burning trash or feces seems unlikely to test a veteran’s understanding of the world or change financial behaviors. Though smoke
exposure in some veterans could have caused health conditions with major financial consequences, the inclusion of controls for
disability or illness in the model make any expected influence of smoke exposure on financial decline small. Table 7 column 4 displays
results for estimations with the smoke exposure measures, created from the 2011 and later survey question responses, substituted for
the primary combat exposure measures, and Table 7 column 5 displays results for estimations with the smoke exposure measures and
the primary combat exposure measures. The results provide support for the hypothesis that combat exposure leads to financial decline.
Table 7 column 5 also indicates estimations for the primary combat exposure measures are robust when only considering a single year
of financial decline observations.

5. Summary of results and discussion

Finances are the most commonly reported stressor for the US population and most Americans experience multiple traumatic events
(American Psychological Association, 2015; Kilpatrick et al., 2013), yet, to the best of our knowledge, this is the first study to measure
financial decline following traumatic exposure. The results consistently support the hypothesis combat exposure is associated with
financial declines for deployed veterans. The results indicate, on average, the probability of developing a new major financial problem
(such as bankruptcy) and greater financial distress increases 0.44 percentage points (21 percent relative to the mean probability)
following a single combat exposure and increases 0.90 percentage points (43 percent relative to the mean probability) following
multiple combat exposures.
We translate the results to a crude lower bound cost estimate of lost productivity and a crude upper bound estimate of bankruptcies
for the 2.7 million veterans deployed from 2001 through 2016. First, assuming average rates of combat exposure reported in the study,
the results suggest combat exposure contributed to an additional 3629 veterans with major financial problems and distress. Next, we
define our average veteran to be a 29-year-old enlisted (E− 5) service member in 2016 with approximate annual basic pay of $38,000
(not counting local allowances for housing, sustenance, or deployments). Assuming financial problems result in a conservative 10
percent productivity loss, or $11,400 over a 3 year survey period (Garman et al., 1996), the marginal effects of combat exposure on
major financial problems and distress translate to a lower bound, short-term productivity loss of $41 million. Next, we assume all
respondents only considered the survey question example of bankruptcy as a major financial problem. In this case, we end up with an
upper bound estimate of 3629 (1.34 per 1000 veterans) personal bankruptcies due to combat exposure. Moreover, neither estimate
captures the financial problem related costs to individual health, military security, military recruitment, or long-term productivity.
The current study also stands out among other economic studies using longitudinal data to identify deleterious influences of combat
exposure on non-market outcomes (Cesur et al., 2013, 2015, 2016; Armey & Lipow, 2016; Cesur & Sabia, 2016) because simulation
results help identify policy-relevant characteristics that may be targeted before a veteran is sent into harm’s way. The results point
toward veterans with poorer pre-deployment mental or physical health, veterans in enlisted ranks, and veterans between the ages of 26
and 36 as being less resilient to the effects of combat exposure on financial problems. As changes in the deployment selection process
could be problematic, the results suggest policies to strengthen health or financial management skills may be best applied to targeted
populations regardless of deployment selection. Additionally, as Edwards (2015) economic study with cross-sectional data reveals no
significant association between household income and combat exposure, the current study highlights the value added by examining
variation within individuals and with alternate economic measures to understand the impact of combat exposure on economic
well-being.

Disclaimer

The author is a military service members or employees of the U.S. Government. This work was prepared as part of his official duties.
Title 17, U.S.C. §105 provides that copyright protection under this title is not available for any work of the U.S. Government. Title 17,
U.S.C. §101 defines a U.S. Government work as work prepared by a military service member or employee of the U.S. Government as
part of that person’s official duties. Report No. 20–21 was supported by the U.S. Navy Bureau of Medicine and Surgery under work unit
no. 60002. The views expressed in this article are those of the authors and do not necessarily reflect the official policy or position of the
Department of the Navy, Department of the Air Force, Department of Defense, nor the U.S. Government. The study protocol was
approved by the Naval Health Research Center Institutional Review Board in compliance with all applicable Federal regulations
governing the protection of human subjects. Research data were derived from an approved Naval Health Research Center Institutional
Review Board protocol, number NHRC.2000.0007.

Author statement

Adam Ackerman: Conceptualization; Project administration (equal); Data curation (equal); Methodology; Formal analysis; Writing
– original draft preparation; Review and editing (equal). Bennett Porter: Project administration (equal); Data curation (equal); Writing
– review and editing (equal).

Declaration of competing interest

Ben Porter also holds an affiliation with the Social Science Research Center at Mississippi State University.

253
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

Appendix

Table A1 displays the influence of individual military, individual non-military, and economic variables on financial decline. Several
military variables influence financial decline. Military service of more than 6 years, officer rank, Air Force branch, and active duty
component are each associated with increases in the no probability of financial decline. The time of military service entry is not
associated with financial decline, and variation in financial decline explained by occupational classifications is relatively limited.
The results suggest several individual non-military variables also influence financial decline. Hispanic ethnicity, higher numbers of
dependents, divorce in the past 3 years, smoking, heavy drinking, and adverse childhood experiences are associated with decreases in
the probability of no financial decline. Higher educational attainment, higher physical health component summary scores (PCS), and
higher mental health component summary scores (MCS) are associated with increases in the probability of no financial decline. The
inclusion of the pre-deployment financial distress variable improves the model accuracy, but the marginal effects require caution in
interpretation given that veterans with pre-deployment financial distress can only move one level (rather than two levels for veterans
without pre-deployment financial distress) higher on the financial distress scale. As the age variable enters the model non-linearly, the
displayed marginal effects at the mean age on financial decline require caution in interpretation.
The model suggests recent economic conditions have greater influence on financial decline than pre-deployment economic con­
ditions, but a one percent change in any economic condition is associated with less of change in the probability of financial decline than
is expected from a single combat exposure. Higher recent gross domestic product (GDP) growth and higher inflation-adjusted military
pay growth are associated with increases in the probability of no financial decline. On the other hand, the variation in financial decline
explained by pre-deployment GDP growth and inflation-adjusted military pay growth is relatively limited. A possible takeaway is that
policies increasing military pay (at levels observed in the sample) may not help prevent veteran financial problems in the long-term
and policymakers may want to consider other interventions to prevent veteran financial problems in the long-term.

Table A1
Marginal Effects of Background Variables on Financial Decline

(1) (2) (3) (4)

Pr(Y=None) Pr(Y = DO) Pr(Y = MPO) Pr(Y = D&MP)

Effect SE Effect SE Effect SE Effect SE

Other military variables

Service entry
Post-2001 entry 0.0035 0.0038 − 0.0011 0.0037 − 0.0019 − 0.0005 − 0.0005 0.0010
Years of service
7–12 0.0151 0.0036 − 0.0137 0.0035 − 0.0007 − 0.0007 − 0.0007 0.0010
13–18 0.0209 0.0050 − 0.0153 0.0048 − 0.0022 − 0.0033 − 0.0033 0.0012
19 or more 0.0169 0.0058 − 0.0093 0.0057 − 0.0037 − 0.0039 − 0.0039 0.0014
Rank
Officer 0.0674 0.0037 − 0.0556 0.0035 − 0.0035 − 0.0083 − 0.0083 0.0010
Branch of service
Navy − 0.0063 0.0036 0.0072 0.0034 − 0.0006 − 0.0004 − 0.0004 0.0009
Air Force 0.0566 0.0028 − 0.0504 0.0026 − 0.0004 − 0.0058 − 0.0058 0.0007
Marine Corps − 0.0318 0.0051 0.0359 0.0050 − 0.0018 − 0.0023 − 0.0023 0.0011
Component
Reserve or National Guard − 0.0523 0.0028 0.0407 0.0027 0.0022 0.0094 0.0094 0.0009
Separated − 0.0118 0.0050 0.0107 0.0048 0.0005 0.0006 0.0006 0.0013
Occupational classification
Electrical repair 0.0034 0.0046 − 0.0054 0.0044 0.0015 0.0005 0.0005 0.0014
Comm/intel 0.0054 0.0045 − 0.0062 0.0043 − 0.0001 0.0008 0.0008 0.0013
Health care − 0.0122 0.0048 0.0117 0.0046 0.0001 0.0005 0.0005 0.0013
Other technical − 0.0149 0.0077 0.0127 0.0073 0.0013 0.0010 0.0010 0.0020
Functional support 0.0093 0.0039 − 0.0100 0.0037 0.0003 0.0004 0.0004 0.0011
Equipment repair 0.0054 0.0042 − 0.0074 0.0040 0.0013 0.0008 0.0008 0.0012
Craft workers 0.0016 0.0070 − 0.0069 0.0066 0.0016 0.0037 0.0037 0.0020
Service and supply − 0.0124 0.0045 0.0093 0.0043 0.0006 0.0025 0.0025 0.0013
Other 0.0212 0.0062 − 0.0176 0.0059 − 0.0008 − 0.0028 − 0.0028 0.0017
Non-military background variables
Financial distress 0.1724 0.0025 − 0.1942 0.0023 0.0224 − 0.0005 − 0.0005 0.0006
Female − 0.0027 0.0030 0.0049 0.0029 − 0.0017 − 0.0005 − 0.0005 0.0008
Age 0.0029 0.0003 − 0.0029 0.0003 0.0001 − 0.0002 − 0.0002 0.0001
Race
Black Non-Hispanic − 0.0007 0.0041 − 0.0059 0.0039 0.0023 0.0043 0.0043 0.0012
Other Non-Hispanic 0.0057 0.0039 − 0.0046 0.0037 − 0.0007 − 0.0004 − 0.0004 0.0011
Hispanic − 0.0184 0.0047 0.0156 0.0045 0.0008 0.0020 0.0020 0.0012
Education
Some college 0.0122 0.0036 − 0.0109 0.0035 − 0.0003 − 0.0010 − 0.0010 0.0009
College degree 0.0262 0.0047 − 0.0217 0.0045 − 0.0016 − 0.0029 − 0.0029 0.0012
(continued on next page)

254
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

Table A1 (continued )
(1) (2) (3) (4)

Pr(Y=None) Pr(Y = DO) Pr(Y = MPO) Pr(Y = D&MP)

Effect SE Effect SE Effect SE Effect SE

Family structure
Number of dependents − 0.0121 0.0009 0.0093 0.0009 0.0009 0.0019 0.0019 0.0002
Divorce − 0.0109 0.0036 0.0041 0.0034 0.0026 0.0042 0.0042 0.0010
Health
PCS 0.0042 0.0002 − 0.0038 0.0002 0.0000 − 0.0004 − 0.0004 0.0000
MCS 0.0055 0.0001 − 0.0051 0.0001 0.0000 − 0.0005 − 0.0005 0.0000
Smoking − 0.0178 0.0027 0.0127 0.0026 0.0015 0.0035 0.0035 0.0007
Heavy drinking − 0.0191 0.0047 0.0198 0.0046 − 0.0016 0.0009 0.0009 0.0011
Childhood experiences
Neglect − 0.0099 0.0050 0.0045 0.0047 0.0025 0.0029 0.0029 0.0012
Verbal abuse − 0.0318 0.0038 0.0254 0.0036 0.0008 0.0056 0.0056 0.0011
Physical abuse − 0.0213 0.0032 0.0183 0.0031 0.0009 0.0021 0.0021 0.0009
Sexual abuse − 0.0331 0.0047 0.0279 0.0046 0.0018 0.0034 0.0034 0.0012
Macroeconomic
GDP growth 0.0220 0.0016 − 0.0181 0.0015 − 0.0004 − 0.0036 − 0.0036 0.0004
Lag GDP growth 0.0020 0.0010 − 0.0004 0.0009 − 0.0008 − 0.0007 − 0.0007 0.0003
Military pay growth 0.0219 0.0018 − 0.0178 0.0018 − 0.0003 − 0.0038 − 0.0038 0.0005
Lag military pay growth − 0.0016 0.0016 0.0003 0.0015 0.0002 0.0012 0.0012 0.0005

Observations 131,036
Pseudo-R2 0.55
DO = Distress Only, MPO = Major Problem Only, D&MP = Distress & Major Problem.
Effect = Generalized Maximum Entropy marginal effect, GDP = gross domestic product, Generalized Maximum Entropy, MCS = mental health
component summary score, PCS physical health component summary score, SE = standard error. The marginal effect for each continuous variable is
calculated with all independent variables set to the mean. The marginal effect for each categorical variable is calculated by changing all observations
from the reference value to the category value with all other independent variables set to the mean. All columns include deployment-related variables
in Table 5, and dummy variables for missing values of family structure, health, health behaviors, and childhood experiences.
Figure A1 presents the effects of combat exposure on the probability of a new major financial problems and distress for different
sub-samples. The confidence interval of the marginal effects for each sub-sample overlap with the confidence interval of the marginal
effects for the full sample. The results suggest combat exposure has a similar influence on different groups of veterans.

255
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

Fig. A.1. Marginal effects of combat exposure over sub-samples on probability (Major Problem & Distress). GME marginal effects for categorical
variables computed by setting all observations to category value and comparing to respective base value with all other variables set to sample
means. Model includes military control variables in Table 2, all deployment-related variables in Table 6, all other background variables in Table 3,
and dummy variables for missing values of family structure, health, health behaviors, and childhood experiences. .

References

Ackerman, A., Porter, B., & Sullivan, R. (2020). The effect of combat exposure on veteran homelessness. Journal of Housing Economics, 49, 101711. https://doi.org/
10.1016/j.jhe.2020.101711
Altonji, J. G., Elder, T. E., & Taber, C. R. (2005). Selection on observed and unobserved variables: Assessing the effectiveness of Catholic schools. Journal of Political
Economy, 113, 151–184. https://doi.org/10.1086/426036
American Psychological Association. (2015). Stress in America: Paying with our health. Washington, DC: American Psychological Association. https://www.apa.org/
news/press/releases/stress/2014/stress-report.pdf. (Accessed 2 April 2020).
Armed Forces Health Surveillance Center. (2007). Healthy deployers: Nature and trends of health care utilization during the year prior to deployment to OEF/OIF,
active components, US Armed Forces, January 2002–December 2006. MSMR, 14(3), 2–5.
Armey, L. E., & Lipow, J. (2016). Hard lessons: Combat deployment and veteran interest in higher education. Applied Economics Letters, 23, 768–772. https://doi.org/
10.1080/13504851.2015.1105918
Ayllón, S., & Fusco, A. (2017). Are income poverty and perceptions of financial difficulties dynamically interrelated? Journal of Economic Psychology, 61, 103–114.
https://doi.org/10.1016/j.joep.2017.03.008
Bonanno, G. A., Galea, S., Bucciarelli, A., & Vlahov, D. (2007). What predicts psychological resilience after disaster? The role of demographics, resources, and life
stress. Journal of Consulting and Clinical Psychology, 75, 671–682. https://doi.org/10.1037/0022-006X.75.5.671
Cesur, R., Chesney, A., & Sabia, J. J. (2016). Combat exposure, cigarette consumption, and substance use. Economic Inquiry, 54, 1705–1726. https://doi.org/10.1111/
ecin.12312
Cesur, R., & Sabia, J. J. (2016). When war comes home: The effect of combat service on domestic violence. The Review of Economics and Statistics, 98, 209–225. https://
doi.org/10.1162/REST_a_00541
Cesur, R., Sabia, J. J., & Tekin, E. (2013). The psychological costs of war: Military combat and mental health. Journal of Health Economics, 32, 51–65. https://doi.org/
10.1016/j.jhealeco.2012.09.001

256
A. Ackerman and B. Porter International Review of Economics and Finance 79 (2022) 241–257

Cesur, R., Sabia, J. J., & Tekin, E. (2015). Combat exposure and migraine headache: Evidence from exogenous deployment assignment. Economics and Human Biology,
16, 81–99. https://doi.org/10.1016/j.ehb.2014.01.001
Chretien, J.-P., Chu, L. K., Smith, T. C., Smith, B., & Ryan, M. A. (2007). Demographic and occupational predictors of early response to a mailed invitation to enroll in
a longitudinal health study. BMC Medical Research Methodology, 7, 6. https://doi.org/10.1186/1471-2288-7-6
Crum-Cianflone, N. (2013). The Millennium cohort study: Answering long-term health concerns of US military service members by integrating longitudinal survey
data with military health system records. In J. Amara, & A. M.) Hendricks (Eds.), Military health care: From pre-deployment to post-separation. New York, NY:
Routledge.
Deaton, A. (2010). Price indexes, inequality, and the measurement of world poverty. The American Economic Review, 100, 3–34. https://doi.org/10.1257/aer.100.1.5
Defense Finance and Accounting Service. (2017). Military pay charts [WWW Document] https://www.dfas.mil/dam/jcr:0066b0ac-2d02-47a6-8b78-1cdc91b56ef7/
2017MilitaryPayChart.pdf. (Accessed 2 April 2020).
Edwards, R. D. (2015). Overseas deployment, combat exposure, and well-being in the 2010 National Survey of Veterans. Journal of Human Capital, 9, 64–93. https://
doi.org/10.1086/680403
Elbogen, E. B., Johnson, S. C., Wagner, H. R., Newton, V. M., & Beckham, J. C. (2012). Financial well-being and postdeployment adjustment among Iraq and
Afghanistan war veterans. Military Medicine, 177, 669–675. https://doi.org/10.7205/MILMED-D-11-00388
Engel, R. C., Gallagher, L. B., & Lyle, D. S. (2010). Military deployments and children’s academic achievement: Evidence from Department of Defense education
activity schools. Economics of Education Review, 29, 73–82. https://doi.org/10.1016/j.econedurev.2008.12.003
Fulton, J. J., Calhoun, P. S., Wagner, H. R., Schry, A. R., Hair, L. P., Feeling, N., Elbogen, E., & Beckham, J. C. (2015). The prevalence of posttraumatic stress disorder
in operation enduring freedom/operation Iraqi freedom (OEF/OIF) veterans: A meta-analysis. Journal of Anxiety Disorders, 31, 98–107. https://doi.org/10.1016/j.
janxdis.2015.02.003
Gadermann, A. M., Engel, C. C. C., Naifeh, J. A., Nock, M. K., Petukhova, M., Santiago, L. P. N., Benjamin, W., Zaslavsky, A. M., & Kessler, R. C. (2012). Prevalence of
DSM-IV major depression among U.S. Military personnel: meta-analysis and simulation. Military Medicine, 177, 47–59. https://doi.org/10.7205/MILMED-D-12-
00103
Garman, E. T., Leech, I. E., & Grable, J. E. (1996). The negative impact of employee poor personal financial behaviors on employers. Journal of Financial Counselling
Planning, 7(1), 157–168.
Golan, A. (2018). Foundations of info-metrics: Modeling, inference, and imperfect information. New York, NY: Oxford University Press.
Golan, A., Judge, G. G., & Miller, D. (1996). Maximum entropy econometrics: Robust estimation with limited data. New York, NY: Wiley.
Golan, A., Judge, G., & Perloff, J. M. (1996). A maximum entropy approach to recovering information from multinomial response data. Journal of the American
Statistical Association, 91, 841–853. https://doi.org/10.1080/01621459.1996.10476952
Greene, W. H. (2016). NLOGIT 6.0: Reference guide. Econom. Softw. Inc N. Y.
Hays, R. D., Sherbourne, C. D., & Mazel, R. M. (1993). The RAND 36-item health survey 1.0. Health Economics, 2, 217–227. https://doi.org/10.1002/hec.4730020305
Kazis, L. E., Lee, A., Spiro, A., Rogers, W., Ren, X. S., Miller, D. R., Selim, A., Hamed, A., & Haffer, S. C. (2004). Measurement comparisons of the medical outcomes
study and veterans SF-36® health survey. Health Care Financing Review, 25, 43–58.
Kilpatrick, D. G., Resnick, H. S., Milanak, M. E., Miller, M. W., Keyes, K. M., & Friedman, M. J. (2013). National estimates of exposure to traumatic events and PTSD
prevalence using DSM-IV and DSM-5 Criteria. Journal of Traumatic Stress, 26, 537–547. https://doi.org/10.1002/jts.21848
LeardMann, C. A., Smith, B., Smith, T. C., Wells, T. S., & Ryan, M. A. K. (2007). Smallpox vaccination: Comparison of self-reported and electronic vaccine records in
the Millennium cohort study. Human Vaccines, 3, 245–251. https://doi.org/10.4161/hv.4589
Loughran, D. S., & Heaton, P. (2013). Post-traumatic stress disorder and the earnings of military reservists. Rand Health Q, 3, 1–3.
Lyle, 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, 319–350. https://doi.org/10.1086/499975
Pleeter, S., Gallo, A. O., Gould, B. R., Li, M. X., Liu, S. H., Simon, C. J., Witschonke, C. F., & Horowitz, S. A. (2011). Risk and combat compensation. Alexandria, VA:
Institute for Defense Analysis. IDA Paper P-4747.
Porter, B., Hoge, C. W., Tobin, L. E., Donoho, C. J., Castro, C. A., Luxton, D. D., & Faix, D. (2018). Measuring aggregated and specific combat exposures: Associations
between combat exposure measures and posttraumatic stress disorder, depression, and alcohol-related problems. Journal of Traumatic Stress, 31(2), 296–306.
Riddle, J. R., Smith, T. C., Smith, B., Corbeil, T. E., Engel, C. C., Wells, T. S., Hoge, C. W., Adkins, J., Zamorski, M., & Blazer, D. (2007). Millennium cohort: The
2001–2003 baseline prevalence of mental disorders in the US military. Journal of Clinical Epidemiology, 60, 192–201. https://doi.org/10.1016/j.
jclinepi.2006.04.008
Ryan, M. A. K., Smith, T. C., Smith, B., Amoroso, P., Boyko, E. J., Gray, G. C., Gackstetter, G. D., Riddle, J. R., Wells, T. S., Gumbs, G., Corbeil, T. E., & Hooper, T. I.
(2007). Millennium cohort: Enrollment begins a 21-year contribution to understanding the impact of military service. Journal of Clinical Epidemiology, 60,
181–191. https://doi.org/10.1016/j.jclinepi.2006.05.009
Smith, J. P. (1999). Healthy bodies and thick wallets: The dual relation between health and economic status. The Journal of Economic Perspectives Journal of the
American Economics Association, 13, 144. https://doi.org/10.1257/jep.13.2.145
Smith, T. C., Jacobson, I. G., Smith, B., Hooper, T. I., Ryan, M. A., & Millennium Cohort Team. (2007a). The occupational role of women in military service: Validation
of occupation and prevalence of exposures in the Millennium cohort study. International Journal of Environmental Health Research, 17, 271–284. https://doi.org/
10.1080/09603120701372243
Smith, B., Leard, C. A., Smith, T. C., Reed, R. J., Ryan, M. A., & Millennium Cohort Team. (2007a). Anthrax vaccination in the Millennium cohort: Validation and
measures of health. American Journal of Preventive Medicine, 32, 347–353. https://doi.org/10.1016/j.amepre.2006.12.015
Smith, B., Wingard, D. L., Ryan, M. A., Macera, C. A., Patterson, T. L., & Slymen, D. J. (2007). US military deployment during 2001–2006: Comparison of subjective
and objective data sources in a large prospective health study. Annals of Epidemiology, 17, 976–982. https://doi.org/10.1016/j.annepidem.2007.07.102
Smith, T. C., Zamorski, M., Smith, B., Riddle, J. R., LeardMann, C. A., Wells, T. S., Engel, C. C., Hoge, C. W., Adkins, J., & Blaze, D. (2007b). The physical and mental
health of a large military cohort: Baseline functional health status of the Millennium Cohort. BMC Public Health, 7, 340. https://doi.org/10.1186/1471-2458-7-
340
US Department of Commerce. (2017). Bureau of economic analysis. https://www.bea.gov/, 2.2.18.
US Institute of Medicine. (2013). Returning home from Iraq and Afghanistan: Assessment of readjustment needs of veterans, service members, and their families. Washington,
DC: National Academies Press.
Vasterling, J. J., Proctor, S. P., Amoroso, P., Kane, R., Heeren, T., & White, R. F. (2006). Neuropsychological outcomes of Army personnel following deployment to the
Iraq War. JAMA, 296, 519–529. https://doi.org/10.1001/jama.296.5.519
Wells, T. S., Jacobson, I. G., Smith, T. C., Spooner, C. N., Smith, B., Reed, R. J., Amoroso, P. J., Ryan, M. A., & Millennium Cohort Team. (2008). Prior health care
utilization as a potential determinant of enrollment in a 21-year prospective study, the Millennium Cohort Study. European Journal of Epidemiology, 23, 79–87.
https://doi.org/10.1007/s10654-007-9216-0
Wenger, J. W., O’Connell, C., & Cottrell, L. (2018). Examination of recent deployment experience across the services and components. Santa Monica, CA: RAND
Corporation. RAND Report 1928-A.
Xu, Y., Briley, D. A., Brown, J. R., & Roberts, B. W. (2017). Genetic and environmental influences on household financial distress. Journal of Economic Behavior &
Organization, 142, 404–424. https://doi.org/10.1016/j.jebo.2017.08.001

257