![Sample our Law journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5942/TOC-Subject-Sample-2021-Law.jpg"})
Abstract
In February of 2008 the New York Times ran a series—War Torn—on Iraq and Afghanistan war veterans and their adjustment to civilian life upon return from the war zone. The authors assessed the criminal involvement of veterans by using newspaper accounts and other open source data to identify homicides in which the offender was an Afghanistan or Iraq war veteran. This particular aspect of the series drew a great deal of criticism, in part because of disagreements about the wisdom of the wars, but also because the sources of data used were perceived as less than systematic and accurate. This series and the debate that it engendered raised once again to prominence the issue of whether veterans are disproportionately involved in crime upon their return from service and specifically from combat assignments. The series also raised the question of whether media accounts of violent behavior by returning combat veterans are simply anecdotal or if they portend a more system-wide problem. This paper uses data from the Surveys of Inmates of State and Federal Correctional Facilities and the Current Population Surveys from 1985 to 2004 to estimate more systematically the prevalence and nature of the offending by military veterans in civilian society. The study seeks to avoid some of the methodological weaknesses of earlier studies that examined the criminal behavior of returning veterans. Specifically, the research considers whether criminal behavior, as reflected in the likelihood of imprisonment, is affected by military service, era of service, or service during wartime after controlling for social and demographic characteristics associated with offending. The findings indicate that military service in general is not predictive of incarceration when key demographic and social integration variables are taken into account. Service during wartime was found to be inversely related to subsequent incarceration, while veterans of the post-1973 All Volunteer Force were more likely to be incarcerated than were civilians and veterans who served during the draft era.
Acknowledgments
This research was supported, in part, through a grant from the American Statistical Association (ASA) Committee on Law and Justice Statistics. Earlier drafts of this paper were presented at the 2009 International Biennial Conference of the Inter-University Seminar on Armed Forces and Society and at the 2009 meeting of the Academy of Criminal Justice Sciences. The authors are very grateful to the conference participants and the anonymous reviewers of Justice Quarterly whose criticism and suggestions have challenged us in useful and constructive ways.
Notes
1. For example, Rohlfs (Citation2010) found dips in the arrest rates of men 18-25 during the years when the Vietnam War was at its peak, presumably due to the fact that so many young men were abroad rather than on the streets.
2. Moskos (Citation1993) observed that the proportion of all men serving in the US military was about 80% during WWII, dropped to about 50% from the Korean War to the early 1960s, fell to about 25% during the Vietnam War, dropped to 20% after the suspension of the draft in 1973, and to about 10% after the end of the Cold War.
3. There is an extensive body of psychological literature on combat veterans. Kardiner (Citation1941) was prescient in anticipating the Post-Traumatic Stress Disorder (PTSD) diagnosis nearly 40 years before it was recognized as a separate disorder by the American Psychiatric Association (in 1980). The National Vietnam Veterans Readjustment Study found that 15% of male and 8.5% of female Vietnam veterans met criteria for PTSD, while among those with high levels of war zone exposure, 35.8% of men and 17.5% of women met the criteria. The study linked PTSD to readjustment problems such as occupational instability, marital conflicts, and other family problems, but not specifically to criminal behavior (Kulka et al., Citation1990). Other studies have linked veteran readjustment problems to criminal behavior and violence. Yager, Laufer, and Gallops (Citation1984) found a significant correlation between exposure to combat and subsequent arrests and convictions (generally for nonviolent offenses) among returning veterans. Their work was based on a national sample of 1,300 men of military age during the Vietnam War who were interviewed 6-15 years after veterans in the sample had left the service and controlled for several pre-service variables. And Rohlfs (Citation2010) reanalyzed self-report data from the National Vietnam Veterans Readjustment Study and found a correlation between Vietnam War experience and self-reported acts of violence. In July Citation2009, the US Army Center for Health Promotion and Preventive Medicine completed an epidemiological study following reports of eight homicides perpetrated by six soldiers from Fort Carson, Colorado from 2005 to 2008. The study concluded that the solders involved in the crimes were, in retrospect, all at risk of engaging in violent behavior based on a clustering of known risk factors, including prior criminal behavior, drug, and/or alcohol abuse, behavioral health disorders (including reluctance to seek treatment), and multiple combat deployments.
4. Veterans Supplements were not published in 1986 and 2004. As an alternative, we matched the Veterans Supplements for 1985 and 2003 with the SISFCF data for 1986 and 2004 respectively.
5. The 1986 Survey of Inmates only examined state inmates, while the 1991, 1997, and 2004 surveys include both state and federal inmates. Sample sizes of the surveys varied as follows: 1986 (14,729), 1991 (21,157), 1997 (18,326), and 2004 (18,185).
6. The monthly income was a categorical variable in the SISFCF, which was recoded into a dichotomous variable (less than poverty level, greater than poverty level) for the purpose of this study. For the 1997 Survey of Inmates, one of the initial categories included individuals who were just above and just below the poverty level threshold. In order to determine which category the participants fell in, the welfare variable was also utilized for that survey year.
7. For most of the survey years there was no specific question as to whether they had earned their high school diploma or a college degree. Therefore, if the respondent indicated that the last year they had attended school was the 12th grade and they had completed that year, it was assumed that they had earned a HS diploma. The Bachelor’s Degree category was coded similarly.
8. Wartime service was coded utilizing two different methods, depending on the survey year. For the 1997 and 2004 Survey of Inmates, there was a specific dichotomous variable which specified whether a veteran had seen combat. There was no specific combat variable in the 1986 or 1991 Surveys, so we coded wartime service if the respondent served in the military during the following dates: World War II—December 1941-September 1945; Korean War—June 1950-July 1953; Vietnam—August 1964‐April 1975; Gulf War—August 1990‐August 1993; Afghanistan/Iraq—September 2001 to present; Iraq—March 2003 to present. The EOS variable was calculated based on the inmate’s military entry date. Inmates were assigned to three different categories, conscript (entered military prior to February 1973), AVF (entered military in or after February 1973), and no military service.
9. Department of Defense 1348.33-M, Manual of Military Decorations & Awards (C6.5.1.1.2). The US Office of Personnel Management (Citation2009) considers holding a campaign or expedition medal equivalent to service during wartime in granting certain forms of veterans preference.
10. As expected, the wartime service variable captures more cases than the combat variable. The actual combat variable in the 1997 and 2004 SISFCFs counted 269 military veterans with combat experience while wartime service tallied 508 inmates. The wartime service variable identified 156 of 269 actual combat veterans. The correlation between the two variables (.409) was significant at the .01 level, using Kendall’s tau-b measure of concordant and discordant pairs. The coefficient of .409 suggests that by knowing the exact dates of service you can identify about 40% of those who experience actual combat. Additionally, characteristics of the set of inmates counted by the wartime service variable were similar to members of the set reporting actual combat, and the two sets are more similar to each other than to the general inmate population. Moreover, the wartime service variable behaved very much like the actual combat variable when entered into different multivariate models. For example, in logistic regression models with all four cross sections of data, the wartime service variable was significant (.001) with a coefficient of −1.175 (SE .064), and an odds ratio (Exp (B)) of .309. In the regression with only 1997 and 2004 data (the “actual” combat data), wartime service was significant (.001) with a coefficient of −1.138 (SE .064), and an odds ratio of .321. The similarity in values in different model specifications supports the assumption that the wartime service variable did indeed measure combat experience in years when it was not directly reported.
11. The 1997 and 2004 SISFCFs include a specific question on combat experience. The 1986 and 1991 SISFCFs did not include this question. Combat experience for these surveys was coded “yes” if the inmate was in the military during WWI, WWII, Korean War, or Vietnam War eras.
12. See, for example, Ellis, Beaver, and Wright (Citation2009).
13. We examined tolerance values and Variance Inflation Factor (VIF) values for all variables in the model. Menard (Citation1995) suggests that tolerance values less than 0.1 indicate collinearity problems while Myers (Citation1990) posits that VIF values greater than 10 should raise caution. The demographic and social integration variables are all within the range of tolerance and VIF values to suggest minimal collinearity bias (tolerance values range from .838 to .911; VIF values range from 1.03 to 1.19). The military service variables (military service, EOS, and wartime service), as expected, are more prone to collinearity as they are ostensibly measuring similar phenomena. Military service and EOS have very low tolerance values (.114 and .150) and relatively high VIF values (8.768 and 6.649). For this reason, we enter these two variables in separate models. The wartime service variable tolerance value (.485) and VIF (2.060) do not raise similar concerns regarding collinearity. An examination of variance proportions confirms dependency between military service and EOS but not between wartime service and the other military variables. But, as wartime service has tolerance and VIF values midway between the other sets of variables, we enter it in separate models to be on the safe side.
14. The most recent offenses of the participants were broken into four categories as follows: (1) Violent offenses (murder, manslaughter, kidnapping, rape, other sexual offense, robbery, assault, other violent offense); (2) property offenses (burglary, larceny, motor vehicle theft, arson, fraud, stolen property, other property offense); (3) drug offenses (drug possession, drug trafficking, other drug offense); and (4) sex offenses (rape, other sexual offense). These categories are not mutually exclusive—sex offenders are included in both the violent offense category and the sex offender category.
15. When Any Military Service is included in the model, the effect of this variable, expressed as odds ratios, remain essentially 1:1. In the model that includes all ages, military veterans are 1.038 times as likely to be incarcerated as non-veterans; in the model that includes only persons age 32 and older, the odds ratio increases slightly to 1.083. There are similar small changes in the coefficients. In regards to EOS, in the model with all ages included, draft-era service decreases incarceration odds by half (0.519) while AVF-era service increases odds by 2.311. When the analysis is limited to persons 32 and older, draft-era service decreases odds of incarceration by .660 while AVF-era service increases odds of incarceration by 2.701. The sign (+ or −) of the coefficients remains the same and the change in magnitude is only slight. The Wartime Service variable is largely unaffected by the test: combat experience decreases odds of incarceration by .461 in the all age model, and by .448 in the model that includes only those 32 and older. When Any Military Service is included in the model, the effect of this variable, expressed as odds ratios, remain essentially 1:1. In the model that includes all ages, military veterans are 1.038 times as likely to be incarcerated as non-veterans; in the model that includes only persons age 32 and older, the odds ratio increases slightly to 1.083. There are similar small changes in the coefficients.
16. When both EOS and wartime service are entered in the model, draft-era service is no longer a significant predictor of incarceration. The Wald statistic is a very low 6.4, suggesting that the coefficient is not significantly different than zero and draft-era service is not a significant predictor of incarceration.
17. Nagelkerke R 2 is a pseudo R2 that indicates the proportional reduction in the absolute value of the log-likelihood measure and provides a measure of how much the model improves as a result of the inclusion of the predictor variables. It is interpreted similar to the R 2 in linear regression by providing a gauge of the substantive significance of the model (Field, Citation2000).