7
Longitudinal Survey Data for Empirical Research on Military Environments
Researchers need appropriate data in order to evaluate the factors that affect individual and small unit behavior in military environments. With these data, they can address questions like the following:
- What demographic, educational, and family characteristics predict a successful adjustment to military life?
- How do different family contexts and concerns shape service members’ responses to particular military assignments?
- What unit-level factors affect success in the military?
- What attitudes are related to successful adjustment to military service?
- What individual-level and unit-level variables predict such undesirable events as early discharge, poor performance ratings, disciplinary actions, dysfunctional behavior (e.g., alcoholism), and suicide?
- What variables are related to poor mental health episodes among military personnel?
- How do trajectories during military service relate to readjustment after service?
In this chapter, the committee argues that these types of questions are best approached with longitudinal data, that is, repeated surveys of the same individuals over a period of years. The chapter first describes the administrative and survey data that are currently available about the armed forces in cross-section. It then describes many of the longitudinal
surveys based on civilian populations, which could provide a model for an effort to collect data regarding the military. Next it presents information about the primary existing longitudinal survey conducted by the armed forces, which focuses almost exclusively on assessing service members’ health. It concludes by recommending that the Department of Defense (1) collect additional administrative data from all applicants to the armed forces, (2) encourage civilian and military researchers to use the existing administrative and survey data, and (3) create a new longitudinal survey to further research on social and organizational factors in the context of military environments.
CROSS-SECTIONAL DATA ON THE MILITARY
The armed forces currently collect administrative data on the demographics, deployments, health, and performance of soldiers. In addition, they have long conducted surveys that provide information about service members in cross-section.
Administrative Data
The armed forces collect data regarding the backgrounds of applicants and recruits, which provide an overview of the military population. They currently collect data on geographic origin, race, age, test scores, education, and gender. While these data provide important information on the contexts from which service members come, they have not enabled researchers and policy makers to assess basic questions about the family backgrounds of recruits, which may have important implications for performance of and interactions within and between small units.
The armed forces do not, for example, gather data that would allow one to infer the socioeconomic background of applicants and service members. Since at least the 1950s, scholars have tried to assess the socioeconomic backgrounds of soldiers (Mayer and Hoult, 1955). During the wars in Iraq and Afghanistan, policy makers and journalists were concerned that service members disproportionately came from poor and disadvantaged backgrounds (Rangel, 2002). Yet scholars had only one technique for using military administrative data to draw conclusions about the socioeconomic origins of recruits. They had to assess these origins based not on direct reports but rather by combining the zip code information collected by the military with the characteristics of neighborhoods (Kane, 2006). This strategy resulted in enough observations to provide statistical power to detect differences in neighborhoods of origin between service members and civilians. However, these results could reasonably be viewed as subject to the ecological fallacy: researchers incorrectly imputing the average charac-
teristics of a geographic area to all the residents of that area. Perhaps the most famous example of this fallacy was identified in the work that first labeled it (Robinson, 1950): whereas the individual correlation between being foreign born and being illiterate was small and positive, the ecological correlation between the percent foreign born and percent illiterate by region in the United States was large and negative.
Surveys
The history of large-scale, cross-sectional surveys conducted by the U.S. military stretches back at least to the American Soldier studies that were directed by Samuel Stouffer during World War II. This series of cross-sectional surveys was administered to nearly half a million service members during the course of that war. In the late 1940s, the results were published in multiple volumes that contributed much to our understanding of military life (e.g., Stouffer, 1949a, 1949b) and produced important concepts such as the theory of relative deprivation (Sewell, 1989). The surveys themselves also helped refine survey methodology (Stouffer, 1966).
The Department of Defense continues to conduct surveys that address the same topics, and these surveys are primarily cross-sectional. Civilian researchers have used the Survey of Active Duty Military Personnel, for example, to generate insight into questions about the relationships between military and civilian society. Some scholars have used the 1999 version of that survey to examine whether gender and race moderate work satisfaction in the military (Lundquist, 2008), as well as to evaluate the culture of particular military occupations (Burland and Lundquist, 2013).
LONGITUDINAL SURVEYS BASED ON NONMILITARY POPULATIONS
Social science research on behavior in civilian environments has long made extensive use of longitudinal surveys of samples of individuals and households, and these surveys include information on former service members. In the United States, major national surveys include multiple National Longitudinal Surveys sponsored by the Bureau of Labor Statistics, multiple surveys of school populations sponsored by the Department of Education, the Panel Study of Income Dynamics (PSID) sponsored by the National Science Foundation, the Health and Retirement Study (HRS) sponsored by the National Institute on Aging, and the National Longitudinal Study of Adolescent Health sponsored by the National Institute of Child Health and Human Development. In addition, the Census Bureau operates two large-scale, ongoing longitudinal surveys (the Survey of Income and Program Participation and the Current Population Survey) that follow waves
of respondents for a year and half to four years. Other longitudinal surveys have surveyed samples of residents of American subpopulations. Notable early examples include the intergenerational Framingham Heart Study and Wisconsin Longitudinal Survey, begun in 1948 and 1957, respectively, and the National Center for Education Statistics program of longitudinal studies of factors that predict success in school (e.g., Early Childhood Longitudinal Studies).
The specifics of the existing surveys differ, but most of them share several central features, including (1) a well-designed full-probability sample designed to yield a sizable, representative sample of the population of interest, (2) a carefully constructed and pretested questionnaire intended to yield useful evidence on personal backgrounds and behaviors, and (3) periodic reinterviews of respondents aiming to trace the dynamics of their circumstances and outcomes over time. Longitudinal surveys with these features have proved to be enormously valuable resources for description of civilian work, health, education, and family experiences. They have provided essential data for studies of the determinants of individual and household behaviors, fertility decisions, success in the labor force, and cognitive achievements in school. They have facilitated research aiming to distinguish causal effects from statistical associations. It is difficult to imagine modern empirical social science without them.
Civilian longitudinal surveys are rarely based on simple random samples of the population but usually follow stratified random sampling strategies in order to generate information both about individuals and about the groups of which they are part. For example, the National Longitudinal Survey of 1979 sampled housing units and then, within those units, sampled individuals between the ages of 14 and 21. In some cases, the sample contains multiple siblings in the same family, the data for which have been used, for example, to estimate family fixed effects models (Rosenzweig and Wolpin, 1995). As another example, the High School and Beyond survey of 1980 sampled first schools and then students from within those schools. This complex structure has enabled researchers to conduct analyses both on individuals and on groups. For example, researchers have examined how educational opportunities vary both between and within schools (Gamoran, 1987).
Survey coordinators have encouraged scholars to use the data from the large longitudinal civilian surveys, such as the PSID and the National Longitudinal Surveys, by convening conferences that focus on early results. In 2005, for example, the PSID project staff convened a conference for researchers using the data from the newly launched Child Development Supplement of that survey. At such conferences, scholars benefit from having a venue to present findings and receive critiques and suggestions from other researchers who have experience using the same data.
Because they represent the entire American population, many of the existing longitudinal surveys include some respondents who have served in the armed forces. Social scientists have occasionally used the available data to compare the labor-market, health, and other outcomes of persons who have or have not performed military service. They have used several of these national longitudinal studies to evaluate outcomes related to the military, though they have assessed such phenomena not during military service but afterward among veterans.
Scholars have most often evaluated questions related to military service using the National Longitudinal Surveys’ National Longitudinal Survey of Youth 1979 (NLSY79). When it was first fielded, the NLSY79 included an oversample of 1,280 people serving in the armed forces in 1978 who had been born between 1957 and 1961. This oversample was conducted in two stages, with the first stage drawing a sample of 200 military “units.”1 In the second stage, individuals within these units were sampled based on their birth years, with oversampling to include enough women in the military sample. The sample was also stratified on the basis of military branch and geographic location. In principle, one could use this sample to estimate “unit” level behavior, though apparently no research has done this. Unfortunately, 84 percent of the oversample was dropped in the 1985 NLSY79 administration. Yet in the first year and the years after the survey began, 552 male respondents enlisted in the armed forces but were not technically included in this oversample. After 1985, therefore, the NLSY79 has continued to follow 756 men who have served in the military. A significant body of work examines how these NLSY79 veterans fare according to at least four different types of outcomes: labor market (MacLean and Parsons, 2010; Teachman and Tedrow, 2007), health (Teachman, 2011), marital status (Lundquist, 2006), and crime (Bouffard, 2005).
Researchers have recently also begun to use the HRS to examine the long-term impacts of service among the population that served at least 40 years ago. The HRS is a panel survey that was started in 1992; it features a representative national probability sample of the U.S. population who were 50 years of age or older in 1992. Since that time, the survey has added subsequent cohorts to reach a sample of more than 20,000 older people. Because rates of service in older cohorts were much higher than they are today (due to the earlier presence of the draft, which may make some of the findings not generalizable to members of the military in an all-volunteer force), the sample includes a relatively large share of veterans. Approximately 50 percent of the male HRS respondents have served in the
__________________
1The frame of military units in the NLSY79 was based on Unit Identification Codes, which identified, across the services, groups of approximately 200 military personnel (roughly equivalent to an Army company) assigned to the same geographic location.
military. Researchers have used these data to explore questions regarding health and mortality after service (London and Wilmoth, 2006; MacLean and Edwards, 2010; Wilmoth et al., 2010).
Scholars have recently begun to use the Adolescent Health Survey (Add Health) to examine how people come to enlist in the armed forces. This survey consists of a random sample of 20,745 students in grades 7-12 in 1994-1995. Add Health includes some questions regarding military service and deployment. These data have also been linked to military administrative records. Researchers have used them to assess pathways into military service (Burdette et al., 2009; Wang et al., 2012). By the 2008 wave of the survey, approximately 12 percent of the men had enlisted in the military. Rates of service among women were less than 3 percent (Wang et al., 2012).
Researchers have also evaluated the socioeconomic characteristics of recruits by using civilian surveys (U.S. Congressional Budget Office, 2007). This research is limited because these surveys typically include relatively small numbers of military recruits from which to generalize. For example, the U.S. Congressional Budget Office draws on the National Longitudinal Survey of Youth 1997, which is a survey of youth born between 1980 and 1984 and has information on only approximately 100 enlistees (U.S. Congressional Budget Office, 2007).2
Research using these civilian-based surveys has been limited in two important ways. First, the numbers of respondents with military backgrounds are relatively small and the questions posed do not focus on their military experiences. Hence, existing surveys do not provide the foundation for systematic study of behavior in military environments. Second, these civilian-based surveys are, by definition, limited to particular cohorts. The HRS, for example, contains data contributed by people who were first eligible to serve in the military in the decades prior to and including the early 1970s. The NLSY79 includes information about people who began their military service in the late 1970s and early 1980s. Add Health represents people who became eligible to serve in the armed forces in the late 1990s.
THE MILLENNIUM COHORT STUDY: A LONGITUDINAL SURVEY BASED ON THE MILITARY POPULATION
The Department of Defense has been operating a large-scale longitudinal study since 2001, the Millennium Cohort Study (developed in response
__________________
2“The NLSY sample draws disproportionately from the lower end of the income distribution. CBO [Congressional Budget Office] weighted the full sample to represent U.S. households with children in residence. As a result, the characteristics of the military subset of that sample should be representative of young enlisted personnel in 2000. Because of the ages of the young people included in the NLSY, CBO’s sample did not contain any military officers” (U.S. Congressional Budget Office, 2007, footnote 89, p. 30).
to recommendations in a 1999 report from the Institute of Medicine).3 This survey was initiated shortly before the beginning of the wars in Iraq and Afghanistan and is focused on health outcomes. The initial web-based survey consisted of 77,047 respondents who were serving at that time (Smith and Millennium Cohort Study Team, 2009). Since 2001, the survey has added three more panels of random samples of people serving in each of three subsequent years: 2004, 2007, and 2011.4 Millennium cohort respondents complete a survey every 3 years and continue to do so after leaving service. The survey data are linked to multiple administrative datasets, including information from, among many others, medical history, mortality, and deployment data (Smith and Millennium Cohort Study Team, 2009). In 2011, the survey added a family component with a goal of incorporating 10,000 family members. The goal is to have a total of approximately 200,000 respondents. The study will continue through 2022. The study design does not appear to contain a plan to add cohorts after the most recent wave in 2011 or to continue the survey past the projected end date of 2022.
The survey is primarily focused on health outcomes. It is designed to elicit information about military experiences, health outcomes, and health care (Smith and the Millennium Cohort Study Team, 2009). The questionnaire for the population serving in 2010, for example, consists of 99 questions, the majority of which concern health conditions. Many of the questions are quite detailed. One question asks, for example, if the respondent has ever been diagnosed with any of 46 different conditions, along with the year of diagnosis, as well as whether the respondent was hospitalized. Data obtained in these periodic surveys of respondents are linked with their Department of Defense administrative records, which enables research that effectively uses both data sources. The administrative records include demographic information regarding race, gender, and age.
As would be expected based on the survey’s focus on health, the data have primarily been used in research regarding health outcomes (Jacobson et al., 2012), but they have also been used to study sexual harassment (LeardMann et al., 2013) and civilian employment (Horton et al., 2013). The study recently added a family component, which will provide essential
__________________
3The Department of Defense also operates, along with the National Institute of Mental Health, another shorter-term epidemiologic study (5-year study to complete in 2014) that includes pre- and postdeployment surveys focused on suicide: the Army Study to Assess Risk and Resilience in Servicemembers (STARRS). The STARRS study team includes researchers from the Uniformed Services University of the Health Sciences; University of California, San Diego; University of Michigan; Harvard Medical School; and the National Institute of Health. For more information, see http://www.armystarrs.org/ [January 2014].
4Information about the Millennium Cohort Study available: http://www.millenniumcohort.org/aboutstudy.php [March 2014].
information on the family context that shapes service members’ behavior and outcomes.
The data do not, however, appear to include crucial information for determining the impact of social and organizational factors on the context of military environments. They do not, for example, provide information about service members’ social origins, such as parental socioeconomic status and educational attainment.
The Millennium Cohort Study is available in principle to all researchers but has been little used outside the survey team. Currently, the majority of the publications that are listed as using this survey have been written by project staff.
CREATION OF A LONGITUDINAL DATABASE
The committee was struck by the potential utility of vast amounts of administrative and survey data that are currently collected by multiple entities within the Department of Defense, but are not retained or maintained in such a way as to facilitate exploratory research programs reliant upon such data. Therefore, the committee advocates the creation of a longitudinal database to store already collected administrative and survey data and to be expanded to include data from the new longitudinal survey described below. The database should capture as wide a range of administrative and survey data as possible, to include, for example, responses from surveys given across the armed services, results of initial testing, individual demographics and biodata, duty rotations, assignments, positions, and performance evaluations. It should also facilitate unit-level research (see Chapter 6) by correlating individual and unit-level data. In short, any data pertaining to a soldier’s development, performance, and progress should be tracked and integrated. Establishing a central repository for data collected from a probability sample of all recruits (with data providing a record of career paths and achievements for recruits from all backgrounds) would facilitate combining sets of data to provide a record of career paths and achievements for recruits from all backgrounds. Consistent with Recommendation 1, the database would serve as the access point for making large amounts of appropriate data available to internal and external behavioral researchers to study basic scientific questions in military contexts. While the committee recognizes that confidentiality concerns and external access issues will need to be addressed, the implementation of such a database would provide an unprecedented collection of data to benefit basic understanding of the impact of social and organizational factors on the behaviors of individuals and groups in the context of military environments. The committee also recognizes that the creation and maintenance of such a database would require funding beyond that required for data collection.
Conclusion 7
The committee concludes that the Army does not currently collect or distribute sufficient data necessary to answer future questions about how social and organizational factors affect the behavior of individuals and small units.
As mentioned above, scholars and policy makers have long expressed interest in the socioeconomic characteristics of recruits, but have not been able to address such basic questions with existing administrative data (U.S. Congressional Budget Office, 2007; Mayer and Hoult, 1955). If the necessary data were collected, researchers could use them to specify how family background correlates with how service members behave and interact. For example, the armed forces could collect information about the socioeconomic characteristics of parents, such as parental occupational and educational attainments. They could also collect other standard demographic data, such as marital status, based on advice from military and civilian researchers. These data would enable researchers to compare the characteristics of recruits to those of the larger population as reflected in the Decennial Census, the Current Population Survey, and other civilian longitudinal studies. Such data could then be linked to other administrative records, as well as being linked with the information collected under the survey proposed below.
Recommendation 7.1
The U.S. military should collect more demographic and socioeconomic information about potential recruits than it currently does in the application process.
Research on social and organizational factors in the context of military environments could also be advanced more quickly if civilian researchers were able to use the existing data more easily. According to a summary of a recent report by a study committee of the Institute of Medicine (Office of News and Public Information, National Academies, 2013):5
Due to limited access to data and the scarcity of research focusing on economic, social, and other impacts of deployment on military service members and their families, the committee was unable to answer many questions about the readjustment needs of this population and the status and effectiveness of support programs. A large array of relevant data are being collected by several federal departments and agencies, and if it were
__________________
5See Institute of Medicine (2013).
possible to fully link and integrate this data, the aggregated information could be comprehensively analyzed to answer many key questions about readjustment, the committee said. However, numerous barriers hinder access to this data, the committee found, such as unclear procedures and steps for making data requests.
Recommendation 7.2
The U.S. Army Research Institute for the Behavioral and Social Sciences and other U.S. military funding agencies should work with project staff of the Millennium Cohort Study and with other relevant parties collecting survey and administrative data on military personnel to:
- create a longitudinal database to be composed of survey data and administrative records presently collected and data from future surveys that may be administered to military personnel;
- fund and disseminate research using the survey data and administrative records collected by the Department of Defense; and
- convene, support, and publicize conferences for researchers who are currently using this data or who are interested in using this data for future research.
Through the process outlined in this recommendation, the data would become more widely available to all researchers, and the initiative should also incorporate funding for outside researchers.
The Millennium Cohort Study has proven to be a rich source of data regarding the health of service members, yet it is not designed to elicit information about a wider array of topics of potential interest to the U.S. Army Research Institute for the Behavioral and Social Sciences or the Department of Defense. Furthermore, its activity is planned to end in 2022, with no new cohorts after 2011, and it does not contain the necessary information that would enable researchers to address directly the topics explored in this report, such as norms (Chapter 2), status in small units (Chapter 5), and multiteam systems (Chapter 6). While the committee deliberated whether a continuation of the Millennium Cohort Study would be satisfactory for these needs, the committee determined that it would not (due partly to its focus on health outcomes and partly to its planned termination but also to its focus on the individual soldier, a focus that does not include sufficient attention to unit-level measures). To evaluate the topics of research recommended in this report, the Department of Defense should launch a new longitudinal study as recommended below. This study will require substantial effort, and execution will require considerable resources. The committee believes the Department of Defense should seek assistance in
developing such a study to move it ahead effectively and expeditiously, and the new survey should draw on the substantial experience of past civilian longitudinal surveys and of the Millennium Cohort Study, as appropriate.
Recommendation 7.3
The U.S. Army Research Institute for the Behavioral and Social Sciences should establish a working group of experts in survey research, empirical social science, and military subject matter charged with development of a new longitudinal survey strategy to track both individuals and small units over time.
In addition to developing the new longitudinal survey, the working group would be charged with assessing the number of times an average active service member has to respond to surveys and, if the number is deemed excessive, suggesting ways to minimize these instances.
To be most useful, the target population should encompass all new recruits (both officers and enlisted personnel), with the baseline interview undertaken near each recruit’s time of entry into the armed forces. However, the design of the study’s starting point should consider how to capture data descriptive of entire existing units (as required for research on multiteam systems, as discussed in Chapter 6) and thereby allow for sampling soldiers by organizational level. While some stratified probability sample of new recruits would become members of the sample, additional soldiers further along in their careers would likely need to be adopted into the survey sample. The stratification could reflect both the characteristics of the units they join and individual background characteristics (such as prior educational attainments). As in the current sample design, respondents would subsequently be interviewed periodically throughout the duration of their service, as well as afterward. To facilitate study of contextual determinants of behavior, consideration should be given to supplementary data collection aiming to characterize the environments within which military respondents act. In addition, the survey data should be linked to administrative data to provide information on success in the military. To give just two examples, survey responses of respondents could be linked to their records of promotions and honors.
Bouffard, L.A. (2005). The military as a bridging environment in criminal careers: Differential outcomes of the military experience. Armed Forces and Society, 31(2):273-295.
Burdette, A.M., V.E. Wang, G.H. Elder, T.D. Hill, and J. Benson. (2009). Serving God and country? Religious involvement and military service among young adult men. Journal for the Scientific Study of Religion, 48(4):794-804.
Burland, D., and J.H. Lundquist. (2013). The dynamic lives and static institutions of the “Two Armies”: Data from the survey of active duty personnel. Armed Forces and Society, 39(1):78-101.
Gamoran, A. (1987). The stratification of high school learning opportunities. Sociology of Education, 60(3):135-155.
Horton, J.L., I.G. Jacobson, C.A. Wong, T.S. Wells, E.J. Boyko, B. Smith, M.A. Ryan, T.C. Smith, and Millennium Cohort Study Team. (2013). The impact of prior deployment experience on civilian employment after military service. Occupational and Environmental Medicine, 70(6):408-417.
Institute of Medicine. (1999). Gulf War Veterans: Measuring Health. L.M. Hernandez, J.S. Durch, D.G. Blazer II, and I.V. Hoverman, Eds.; Committee on Measuring the Health of Gulf War Veterans, Institute of Medicine. Washington, DC: The National Academies Press.
Institute of Medicine. (2013). Returning Home from Iraq and Afghanistan: Assessment of Readjustment Needs of Veterans, Service Members, and Their Families. Committee on the Initial Assessment of Readjustment Needs of Military Personnel, Veterans, and Their Families. Board on the Health of Select Populations. Washington, DC: The National Academies Press.
Jacobson, I.G., J.L. Horton, C.A. LeardMann, M.A.K. Ryan, E.J. Boyko, T.S. Wells, B. Smith, and T.C. Smith. (2012). Posttraumatic stress disorder and depression among U.S. military health care professionals deployed in support of operations in Iraq and Afghanistan. Journal of Traumatic Stress, 25(6):616-623.
Kane, T. (2006). Who Are the Recruits? The Demographic Characteristics of U.S. Military Enlistment, 2003-2005. Washington, DC: Heritage Foundation.
LeardMann, C.A., A. Pietrucha, K.M. Magruder, B. Smith, M. Murdoch, I.G. Jacobson, M.A.K. Ryan, G. Gackstetter, T.C. Smith, and Millenium Cohort Study Team. (2013). Combat deployment is associated with sexual harassment or sexual assault in a large, female military cohort. Women’s Health Issues, 23(4):e215-e223. Available: http://www.whijournal.com/article/S1049-3867(13)00038-8/fulltext [April 2014].
London, A.S., and J.M. Wilmoth. (2006). Military service and (dis)continuity in the life course: Evidence on disadvantage and mortality from the health and retirement study and the study of assets and health dynamics among the oldest-old. Research on Aging, 28(1):135-159.
Lundquist, J.H. (2006). The black-white gap in marital dissolution among young adults: What can a counterfactual scenario tell us? Social Problems, 53(3):421-441.
Lundquist, J.H. (2008). Ethnic and gender satisfaction in the military: The effect of a meritocratic institution. American Sociological Review, 73(3):477-496.
MacLean, A., and R.D. Edwards. (2010). The pervasive role of rank in the health of U.S. veterans. Armed Forces and Society, 36(5):765-785.
MacLean, A., and N.L. Parsons. (2010). Unequal risk: Combat occupations in the volunteer military. Sociological Perspectives, 53(3):347-372.
Mayer, A.J., and T.F. Hoult. (1955). Social stratification and combat survival. Social Forces, 34(2):155-159.
Office of News and Public Information, The National Academies. (2013, March 26). For Immediate Release: Evidence-Based Diagnostics and Therapies and Long-Term Forecasts of Needs Among Steps Necessary to Ease Iraq and Afghanistan Service Members’ Readjustment to Post-Deployment Life. News Release. Washington, DC: The National Academies. Available: http://www8.nationalacademies.org/onpinews/newsitem.aspx?RecordID=13499 [June 2014].
Rangel, C.B. (2002). Bring Back the Draft. New York Times, December 31, p. A19. Available: http://www.nytimes.com/2002/12/31/opinion/bring-back-the-draft.html [April 2014].
Robinson, W.S. (1950). Ecological correlations and the behavior of individuals. American Sociological Review, 15(3):351-357.
Rosenzweig, M.R., and K.I. Wolpin. (1995). Sisters, siblings, and mothers: The effect of teenage childbearing on birth outcomes in a dynamic family context. Econometrica, 63(2):303-326.
Sewell, W.H. (1989). Some reflections on the golden-age of interdisciplinary social-psychology. Annual Review of Sociology, 15(1):1-16.
Smith, T.C., and Millennium Cohort Study Team. (2009). The U.S. Department of Defense Millennium Cohort Study: Career span and beyond longitudinal follow-up. Journal of Occupational and Environmental Medicine, 51(10):1,193-1,201.
Stouffer, S.A. (1949a). The American Soldier: Adjustment During Army Life. Studies in Social Pscyhology in World War II, Vol. 1. Princeton, NJ: Princeton University Press.
Stouffer, S.A. (1949b). The American Soldier: Combat and Its Aftermath. Studies in Social Pscyhology in World War II, Vol. 2. Princeton, NJ: Princeton University Press.
Stouffer, S.A. (1966). Measurement and Prediction. New York: Wiley.
Teachman, J.D. (2011). Are veterans healthier? Military service and health at age 40 in the all-volunteer-era. Social Science Research, 40(1):326-335.
Teachman, J.D., and L.M. Tedrow. (2007). Joining up: Did military service in the early all volunteer era affect subsequent civilian income? Social Science Research, 36(4):1,447-1,474.
U.S. Congressional Budget Office. (2007). The All-Volunteer Military: Issues and Performance. Washington, DC: U.S. Congress. Available: http://www.cbo.gov/ftpdocs/83xx/doc8313/07-19-MilitaryVol.pdf [March 2014].
Wang, L., G.H. Elder Jr., and N.J. Spence. (2012). Status configurations, military service and higher education. Social Forces, 91(2):397-422.
Wilmoth, J.M., A.S. London, and W.M. Parker. (2010). Military service and men’s health trajectories in later life. Journals of Gerontology Series B-Psychological Sciences and Social Sciences, 65(6):744-755.
