Abstract
This article examines the effects of peacetime cold war military service on the life course according to four potentially overlapping theories that state that military service (1) was a disruption, (2) was a positive turning point, (3) allowed veterans to accumulate advantage, and (4) was an agent of social reproduction. The article argues that the extent to which the effect of military service on veterans' lives corresponds with one or another of the preceding theories depends on historical shifts in three dimensions: conscription, conflict, and benefits. Military service during the peacetime draft era of the late 1950s had a neutral effect on the socioeconomic attainment of enlisted veterans. However, it had a positive effect on veterans who served as officers, which partly stemmed from status reproduction and selection. Yet net of pre-service and educational differences by rank, officers in this peacetime draft era were still able to accumulate advantage.
Keywords: social inequality, military service, life course, military rank
Since the early 1970s, researchers have debated the effects of military service on socioeconomic attainment. Different researchers have found that military service has positive, negative, and neutral associations with earnings and occupational status depending on veterans' characteristics and on era of service. According to early research, veterans tended to work at jobs with the same status and income as nonveterans.1 This research, however, was based on information provided by veterans who served in different periods. Later research found that veterans of World War II, the Korean War, and the post–Korean cold war earned more and had higher status than did nonveterans.2 According to recent research, World War II veterans earned no more, and possibly earned less, than did equivalent nonveterans.3 Vietnam veterans had lower income and lower occupational status than did nonveterans of the same period.4 Veterans who were drafted to serve in Vietnam paid a penalty compared to those who were not drafted.5 White veterans of the post-Vietnam era, or the All-Volunteer Force (AVF), earn less than white nonveterans, while non-white veterans earn more than their nonveteran counterparts.6 More recently, research has found that the effects of military service differed by rank.7 These findings suggest that the effects of military service differ across different groups and historical periods.
This article examines the effects of military service by focusing on how such service has differed across historical time. It suggests that the effect of military service is influenced by its larger social context: the forces that lead people to serve in the military, the political factors that shape the military experience, and the opportunities provided to service members during and after their tours of duty. It argues that the effects of military service differ according to the presence or absence of three different characteristics: conscription, conflict, and benefits. To demonstrate the importance of these characteristics, the article focuses on veterans who served in a period that has received little attention, the cold war period between the Korean and Vietnam Wars, which was relatively free of conflict. The men who came of age during this era were eligible to be drafted. These veterans did not have access to the 1944 Servicemen's Readjustment Act and its successors, commonly known as the GI Bill, as did veterans of earlier and later eras. Therefore, the following analysis provides an assessment of the effect of military service during an era without conflict or benefits but with conscription.
Much previous research has focused on how military service affected the lives of U.S. veterans who served during the major wars since 1940: World War II and the Korean and Vietnam Wars.8 In addition, research has explored the effect of the post-Vietnam AVF on veterans' lives.9 Yet relatively little research has looked at how military service affected the lives of the five million individuals who entered the military between 1954 and 1964, the decade between the Korean and Vietnam Wars.10 According to the 2000 census, more than one-third of men in their early sixties, those approaching retirement, served in the peacetime cold war armed forces. Little is known about how these veterans' military service affected their later civilian lives. Researchers have commonly combined information about these peacetime veterans with information about those who served during wartime.11 This article looks at how military service affected the later civilian work lives of some of the millions of men who served in the armed forces during the peacetime draft era. In particular, it examines occupational status and earnings measured when these veterans were in their mid-thirties and early fifties. Previous research has left open a number of questions related to service in this era. Did veterans of this period pay a penalty in their later work lives as it appears wartime veterans did? Did they obtain through their service the credentials or skills they needed to do well at work?
The article also examines an empirical puzzle that has recently begun to surface in the literature. Veterans who served as officers appear to disproportionately benefit from their service.12 Yet research has not examined how and why the effect of serving as an officer may differ from that of serving in the enlisted ranks. Part of the lack of research regarding this question is due to limitations in the data. Few data sets provide information about rank along with veteran status. This article uses one of those few data sets.
The article proceeds according to the following plan. First, it outlines four competing theories of how military service in general affects veterans' outcomes. Next, it suggests a framework for understanding how historical variation in three dimensions affects the likelihood that veterans' outcomes will correspond with one or the other of the initial four theories. Then, it uses data to test whether veterans' outcomes correspond to the predictions described above in a particular historical period: the peacetime draft era of the late 1950s. Finally, it relates these findings to previous research, suggesting that future research pay attention to historical variation in the three characteristics that shape the impact of military service.
In brief, the findings suggest that military service during the peacetime draft era had a neutral effect on the socioeconomic attainment of enlisted veterans and a positive effect on the attainment of veterans who served as officers. During this portion of the draft era, there were no differences between men who served and those who did not in terms of the class background. However, even during this draft era, men were less likely to serve in the armed forces if they had greater academic abilities and achievements. Veterans ultimately had higher socioeconomic attainment in their later civilians lives if they served as officers. Some, though not all, of the apparent positive association between officer status and socioeconomic attainment stemmed from the fact that officers differed from enlisted men on all pre-service measures when they entered the military. Thus, some of the officer premium stemmed from status reproduction and selection. Yet net of these pre-service differences, officers in this peacetime draft era were still able to accumulate advantage.
Disruption, Turning Point, Cumulative Advantage, and Status Reproduction
Previous research has developed four potentially overlapping theories of the effects of military service on later attainment. According to the first theory, military service negatively affects veterans' lives and serves as a disruption. According to the second theory, it improves the lives of recruits who were on a negative trajectory, serving as a positive turning point. According to the third theory, it positively affects the lives of recruits who enter the armed forces with pre-service privileges, allowing these veterans to accumulate advantage. According to the fourth theory, it has no effect on veterans' lives and serves as an agent of social reproduction. A counterpoint to these four theories suggests that military service has no net effect on veterans' lives after accounting for the individual characteristics that cause people to enter the armed forces.
The view of military service as a disruption emphasizes lost time. All else being equal, veterans should earn less when they are young because, on average, they have less experience than nonveterans of the same age in the civilian labor market.13 All veterans should, therefore, experience military service as a disruption. Military service should have a negative effect on veterans because the timing of service competes with educational attainment and the accumulation of work experience. The timing of military service, beginning when people turn eighteen, competes with the timing of college. Therefore, one might expect veterans to have lower educational attainment than nonveterans and thus lower socioeconomic standing. Though recent military recruits have earned more than their civilian counterparts,14 during the cold war, soldiers earned less than equivalent civilians. At all times, veterans begin their first civilian jobs when they are older than nonveterans with the same characteristics.15 Therefore, they should accumulate, on average, less civilian experience over their work life. If employers base wages and salaries on years of civilian experience, then, other things being equal, veterans should earn less and have lower status than equivalent nonveterans. This negative account makes a sharp distinction between military and civilian training.
The view of military service as a positive turning point focuses on the extent to which military service helps people overcome the limitations associated with their individual or family characteristics. For example, some research has found that African American and Mexican American veterans earned more than their nonveteran counterparts because military service provided a “bridging environment” that enabled them to learn skills that they would not otherwise have learned.16 Veterans who had a pre-service history of delinquency may have reduced chances of later criminal offending.17 In addition, research has found that military service benefited veterans with less education, such as high school dropouts.18 Research in this vein has argued that veterans from disadvantaged backgrounds benefit from several attributes of military service. People who join the military leave home and become independent for the first time, often going to a new state or country. Military service expands their horizons and cuts these service-members off from their families and friends, creating a break from a negative past.19 These individuals learn such skills as working with other people, operating within a bureaucracy, and meeting a deadline. Military service helps them move from a negative context to a positive one, ultimately serving as a positive turning point. Whatever the reason, military service may substitute for or facilitate the attainment of a college education among veterans from lower status backgrounds or with lower prior academic achievements.
By contrast, the theory of cumulative advantage suggests that military service disproportionately benefits veterans with greater pre-service resources and accomplishments. This view of military service builds on the theory of the Matthew effect. In his initial identification of the effect, Merton drew on the biblical quote from the book of Matthew, “for unto everyone that hath shall be given; … but from him that hath not shall be taken away even that which he hath.”20 More recently, the Matthew effect, or the theory of cumulative advantage, has been applied to the study of aging to suggest that people who start with greater initial resources tend to accumulate benefits throughout the life course, while those with fewer resources do not.21 When applied to the analysis of military service, the theory suggests that people from more privileged families are more likely to benefit from their service in their later civilian lives than are people from less privileged families.22 Thus, this theory presents the mirror image of the turning point theory, in which relatively disadvantaged veterans disproportionately benefit from their service.
The apparent negative or positive association between military service and socioeconomic attainment may stem from differences between veterans' and nonveterans' family backgrounds. Veterans may end up with the same occupational status and income as they would have had they not served in the military. The armed forces may thus reproduce prior civilian status. The theory of status reproduction refers to the process by which the class position into which an individual is born influences his or her later-life circumstances, primarily through cultural tastes.23 The military version of this theory is based on the following chain of reasoning: pre-service status affects status within the military, which, in turn, determines post-service status.24 People are born into different class positions that determine their military service and rank, which, in turn, affect their later-life socioeconomic attainment. According to this view, military service should have no association with later-life outcomes net of family background. This view predicts that people born to more well-off families are unlikely to enlist in the armed forces.25 In times of war, these people can call on a doctor to exempt them from the draft.26 Some recruits, however, do come from higher status families. They usually attend officer training or military academies, leading to higher rank.27 They are the veterans most likely to use GI Bill benefits to subsidize their college education.28 By contrast, people born to poor or low-status families do not have the same early advantages and, as a consequence, will remain in the lower end of the occupational status and income distributions.29 During times of war, they are the first to enlist in the military. If drafted, they usually comply with their orders. They are assigned to dangerous combat positions where they are more likely to be killed.30 In line with this explanation, the Vietnam War has been called a “class war,” in which the sons of working-class parents were more likely to fight and die in Southeast Asia.31 If the armed forces simply reflect the effects of family background, the role of military service is consistent with the theory of status reproduction.
Alternatively, men may decide to serve in the armed forces if they have individual characteristics that make it less likely that they would work in jobs with higher or lower income and status even without military service. This view suggests that the apparent association between military service and socioeconomic outcomes is a product of selection.32 According to this view, veterans have lower or higher IQ scores and high school rank than nonveterans before they enter the armed forces.33 Recruits with greater intellectual abilities are assigned to positions with higher rank while in the military. Their pre-service characteristics, in turn, determine their ultimate socioeconomic attainment. They advance as far as they would have had they not served in the military. Thus, military service has a neutral association with later-life accomplishments net of veterans' prior academic achievements and aspirations.
Table 1 summarizes the predictions of the four theories regarding the effect of military service on socioeconomic attainment according to pre-service status. The rows of the table represent the pre-service status of recruits, while the columns represent their relative post-service status. Each cell contains the theory that predicts the relationship specified by the intersection of column and row. For example, if veterans with high pre-service socioeconomic status achieve higher post-service status relative to comparable nonveterans, this would be consistent with the theory of cumulative advantage. By contrast, if veterans with low pre-service status achieve higher post-service status relative to similar nonveterans, this would be consistent with the turning point theory.
Table 1. Theoretical Predictions Regarding the Relationship between Pre- and Post-Service Status.
Pre-Service Status | Post-Service Status | ||
---|---|---|---|
Lower | Same | Higher | |
Low | Cumulative advantage | Status reproduction | Turning point |
High | Disruption | Status reproduction | Cumulative advantage |
Three Dimensions of Military Service
This article argues that the extent to which the effect of military service on veterans' lives corresponds with one or another of the preceding theories depends on historical context. The context of service in the U.S. armed forces has not been constant across the past six decades but has varied according to three dimensions. The first dimension concerns the method of recruitment at a particular time, whether draft or volunteer. It affects the types of people who enter the armed forces. It also affects the experience of entering the military, increasing the possibility that military service will disrupt the lives of recruits who are drafted. The second dimension concerns the state of military conflict when recruits enter the military, that is, whether the nation is at war or peace. The presence or absence of conflict profoundly affects the experience that service members have while they are in the military. Recruits who enter during wartime are more likely to have a negative experience and thus to experience their service as a disruption. The third dimension consists of the extent to which the government provides benefits for veterans. This dimension affects the experience of veterans after they leave the armed forces. In contrast to the previous two dimensions, the presence of benefits makes it more likely that veterans will experience an improvement in their status after they leave the military. In reality, the three dimensions overlap. For example, the U.S. government has been more likely to operate a draft during wartime. In addition, the extent of benefits may influence the types of people who volunteer for the armed forces. Nevertheless, the article argues that it is useful to separate the three dimensions to derive hypotheses and motivate analyses regarding the effects of service.
Table 2 summarizes variation in these dimensions over time. For example, during the major wars of the past six decades, World War II, and the Korean and Vietnam Wars, recruits entered the military through a draft, experienced conflict, and received educational benefits.34 However, during the other periods, soldiers saw variation in all three characteristics. For example, during the post-draft era inaugurated in 1973, soldiers have entered voluntarily and received benefits. By contrast, during the period 1955 to 1963, soldiers were drafted but neither experienced a war nor received benefits. This article concentrates on the experiences of this latter group of veterans when compared to nonveterans who were eligible to serve during the same period. Five million men served during this peacetime draft period, and, for the most part, they did not receive benefits. Who were they, and how did their service affect their later lives?
Table 2. Context of Military Service in Three Dimensions.
Draft | War | Benefits | |
---|---|---|---|
1941–1945 | ✓ | ✓ | ✓ |
1946–1949 | ✓ | ✓ | |
1950–1953 | ✓ | ✓ | ✓ |
1954 | ✓ | ✓ | |
1955–1963 | ✓ | ||
1964 | ✓ | ✓ | |
1965–1973 | ✓ | ✓ | ✓ |
1974–1990 | ✓ | ||
1991 | ✓ | ✓ | |
1992–2001 | ✓ | ||
2002– | ✓ | ✓ |
The Impact of the Draft
The conscription dimension is important for understanding the nature of the relationship between military service and veterans' outcomes because it affects the types of people who enter the military and the average characteristics of soldiers. For the first half of the preceding sixty years, the United States recruited soldiers with a draft. Since 1973, however, recruitment has been voluntary.35 The following analyses examine veterans who served during several years of the draft period. The draft primarily affects the extent to which the military is an agent of status reproduction. When the armed forces recruit service members by means of conscription, they should tend to draw enlistees with a broader range of familial and personal characteristics than when they recruit volunteers. In the ideal, the presence of a draft means that there are no class differences between those who serve in the armed forces and those who do not. It also means that men with a wider range of abilities serve in the military. The draft-era armed forces draw, in the ideal, from a broad spectrum of the population, which leads to negligible differences between veterans and nonveterans in terms of their resources and abilities. These characteristics lead to the following hypothesis:
Hypothesis 1: During the draft era, there were no differences between veterans and nonveterans in terms of their family and individual characteristics.
During conscription, men who volunteered had more control over the timing and type of their service than did draftees. They could determine when they entered and had relatively greater choice of military branch and occupation. For instance, they could enter the Navy or Air Force, branches that provide more technical training and less exposure to dangerous situations than the Army or Marines. Draftees could only enter the Army and, in certain cases, the Marines. Men who feared being drafted could volunteer for service, thus avoiding the possibility of being drafted.36 It is difficult to assess the extent to which veterans who reported being volunteers should be considered draft motivated and thus not true volunteers. Nonetheless, these draft-motivated volunteers had more control over their service than did draftees. Previous research has focused on the extent to which the draft represented a tax on draftees, leading them to earn less when they re-entered the civilian labor market than non-draftee veterans and than nonveterans.37 Indeed, draftee Vietnam veterans earned less than did non-draftee veterans.38 In addition, they may have had lower income and occupational status than did nonveterans throughout the work life. These findings suggest the following hypothesis:
Hypothesis 2: Draftee veterans have had lower socioeconomic attainment than volunteers and than comparable nonveterans.
Previous research has suggested that veterans have different characteristics from nonveterans even in times of the conscription. In the United States, the draft operated with a number of exemptions that were more likely to benefit people from more- rather than less-privileged backgrounds.39 Men from less-privileged backgrounds were more likely to serve in the armed forces than were those from more-privileged backgrounds, even when service was mandatory. Thus, some research has argued that military service is governed by class bias even under a draft.40 Other research has suggested that the physical and mental requirements for entering the armed forces ensured that veterans had greater capabilities than nonveterans. Indeed, veterans of World War II were more physically fit and may have had greater earning power than nonveterans.41
Two recent articles have also noted that officers benefited from their service in their later civilian attainment.42 During the draft era, at least some of this effect may have stemmed from the different average characteristics of veterans who served as officers. As noted above, men with greater pre-service socioeconomic and intellectual resources may have been more likely to serve in positions with higher rank while in the draft-era armed forces. Thus, an observed benefit of officer status may have stemmed from the reproduction of pre-service socioeconomic status or from selection on the basis of individual characteristics. Alternatively, it may be that veterans who served as officers benefited from their time in the military net of their greater pre-service advantages. If veteran officers came from higher status backgrounds and achieved relatively higher status after leaving the military, the previously observed officer premium during the draft era may correspond with the theory of cumulative advantage. This chain of reasoning leads to the following two hypotheses:
Hypothesis 3a: During the draft era, veterans who served as officers came from families with greater socioeconomic resources and had greater academic and educational abilities.
Hypothesis 3b: Net of pre-service differences, these officer veterans had greater socioeconomic attainment in their later lives than did comparable nonveterans.
Peacetime Veterans
The conflict dimension shapes the effect of military service on veterans' lives for several reasons. Most basically, it determines whether service members serve during war or peace. Soldiers who enter during wars are much more likely than those who enter during peacetime to see combat, with a number of important consequences. First, though mortality does exist during peacetime, soldiers who are exposed to combat are much more likely to die than those who are not.43 Second, they are more likely to suffer physical wounds, which can have a long-term impact on their physical and mental health and later civilian careers.44 Third, soldiers exposed to combat are much more likely to suffer negative psychological consequences, such as posttraumatic stress disorder and other associated mental disorders, which can have a long-term impact on veterans' lives.45 By contrast, soldiers who serve during peacetime are relatively free of the physical and mental health consequences of wartime service. This suggests that peacetime veterans should not be negatively affected by their service. The impact of peacetime service should therefore correspond with the status reproduction theory. The following analysis focuses on peacetime service and examines the following hypothesis:
Hypothesis 4: Among those who were eligible to serve during peacetime, veterans had no worse outcomes than did nonveterans.
Veterans without the GI Bill
The benefits dimension has justly been accorded much attention in the literature and concerns the effect of funds provided by the government on veterans' lives. To counteract the potentially negative effect of service detailed in the preceding discussions of conscription and conflict, the U.S. government has provided funds to help veterans further their education.46 World War II veterans benefited from educational and other funds provided by the GI Bill. Research suggests that these benefits had large-scale effects, transforming the shape of civic engagement in American society.47 Veterans who used military educational benefits after serving in every period beginning with World War II have attained more education and have had higher earnings than those who did not.48 An analysis based on the 1973 Occupational Changes in a Generation Survey shows that the GI Bill increased attainment of higher education by 15 percent to 20 percent among men born in the 1920s and early 1930s.49 Veterans who took advantage of the GI Bill also worked at higher-status jobs than did other veterans and nonveterans.50
Yet these benefits were not available to veterans of all eras. Veterans of the peacetime draft era did not have access to the same educational funding as did most other veterans since World War II. The GI Bill funds were unavailable between 1955 and 1965, the peacetime cold war. When Congress reinstated the GI Bill, the new law retroactively provided educational funding to veterans who served in the late 1950s and the early 1960s.51 However, veterans of this era who had planned to enter college before their service were less likely to go on to get a college degree than were comparable nonveterans.52 This finding suggests that, for veterans who served in an era when benefits were not available, military service may have been a disruption. It leads to the following hypothesis:
Hypothesis 5: Among those who were eligible to serve when veterans' benefits were not available, veterans should have worse socioeconomic outcomes than comparable nonveterans before taking into account differences in total educational attainment.
Data and Method
The Sample
The following analyses are based on data derived from the Wisconsin Longitudinal Study (WLS). The WLS contains information provided by 10,317 men and women, a one-third random sample of those who graduated from Wisconsin high schools in 1957. WLS data were collected from the 1957 graduates or their parents in 1957, 1964, 1975, 1992, and 2004. These data provide a full record of social background, youthful aspirations, schooling, military service, and socioeconomic attainment throughout the life course. More than half of the male WLS graduates served in the armed forces. Fewer than 1 percent of the female graduates served in the military. Therefore, the analytical sample is limited to the 4,992 men in the survey.
The WLS provides a unique data set with which to test the preceding hypotheses. As indicated by the survey's name, the data are longitudinal. They include measures of teenagers' academic achievements, aspirations, and family background collected before the respondents were eligible to enter the military. These multiple measures provide a method of controlling for preexisting differences between veterans and nonveterans, not based on retrospective data. The data also include multiple measures of veterans' and nonveterans' attainment decades after military service was complete. The respondents provided information about their occupational and economic attainment when they were in their thirties, the middle of their careers. In addition, they provided information when they were in their early fifties, near the end of their careers. Thus, the data also provide a view of the long-term effects of military service across the work life. In addition, the data include more detailed information about the characteristics of service, particularly military training, draft status, and rank, than do most other population data sets regardless of whether the data are longitudinal or cross-sectional.
The WLS data are also well suited to demonstrating the effects of the three dimensions of military service. The survey respondents became eligible to serve in the armed forces during a period of time that has received little empirical attention. As mentioned above, the late 1950s were years during which there was no conflict, yet the armed forces filled their ranks by means of a draft. Thus, millions of men served in this period. These data provide information about an era during which there was no war nor were there benefits, but there was a draft. Yet when compared with the wartime draft eras, World War II and the Korean and Vietnam Wars, and the peacetime voluntary period following the Vietnam War, there has been little empirical attention paid to this period in the history of the U.S. armed forces. No other data set provides such extensive information about the veterans and nonveterans of this era.
The data are limited in that they do not provide information derived from the entire country. However, the sample respondents are similar in many respects to people of their age in the country. According to the census, approximately two-thirds of the members of the birth cohort covered by the survey are white high school graduates, similar to the members of the WLS sample.53
Dependent Variables
The dependent variables are three measures of occupational status and earnings measured at two points in time: 1975 and 1992. In 1975, the respondents were in their mid-thirties, near the midpoint of their working lives. In 1992, they were in their early fifties, near the end of their careers. (See Table 3 for all variables, along with their means and standard deviations or proportions within the sample.)
Table 3. Means, Standard Deviations, and Proportions by Veteran Status.
Variable | Total Sample | Nonveteran | Veteran | |||
---|---|---|---|---|---|---|
M | SD | M | SD | M | SD | |
Family background variables | ||||||
Average family income | 6,491.76 | 6,202.70 | 6,565.16 | 6,231.09 | 6,395.55 | 6,166.39 |
Logged | 8.65 | 0.56 | 8.65 | 0.57 | 8.65 | 0.55 |
Father's Socioeconomic Index | 35.04 | 23.45 | 35.33 | 23.80 | 34.66 | 22.99 |
Father's education | 9.89 | 3.49 | 9.88 | 3.50 | 9.92 | 3.48 |
Mother's education | 10.68 | 2.83 | 10.60 | 2.86 | 10.79 | 2.79 |
Number of siblings* | 3.07 | 2.38 | 2.98 | 2.29 | 3.20 | 2.49 |
Farm origin* | 0.19 | 0.21 | 0.17 | |||
Academic achievement variables | ||||||
IQ score* | 102.21 | 15.18 | 102.93 | 15.46 | 101.27 | 14.75 |
High school rank* | 98.38 | 14.49 | 100.28 | 14.75 | 95.90 | 13.77 |
College graduate* | 0.34 | 0.40 | 0.28 | |||
Aspirations | ||||||
Serve in military* | 0.28 | 0.19 | 0.40 | |||
Continue schooling* | 0.45 | 0.53 | 0.35 | |||
Work as professional* | 0.40 | 0.46 | 0.32 | |||
Work as manager | 0.21 | 0.20 | 0.22 | |||
Served in the military | ||||||
Enlisted | 0.38 | — | — | 0.88 | ||
Officer | 0.05 | — | — | 0.12 | ||
Socioeconomic outcomes in 1974 | ||||||
Occupational education* | 40.60 | 30.85 | 43.67 | 32.25 | 36.58 | 28.42 |
Occupational income | 40.88 | 22.83 | 40.78 | 23.08 | 41.01 | 22.51 |
Income* | 16,270.34 | 11,475.79 | 16,636.75 | 11,977.50 | 15,790.09 | 10,768.68 |
Logged | 9.55 | 0.72 | 9.55 | 0.75 | 9.54 | 0.68 |
Socioeconomic outcomes in 1992 | ||||||
Occupational education* | 40.86 | 29.88 | 43.46 | 30.71 | 37.46 | 28.41 |
Occupational income | 41.73 | 23.44 | 42.02 | 23.53 | 41.35 | 23.31 |
Income* | 71,781.76 | 253,652.40 | 79,745.07 | 330,811.30 | 61,344.30 | 71,265.92 |
Logged | 10.63 | 1.11 | 10.62 | 1.14 | 10.63 | 1.07 |
n | 2,960 | 1,679 | 1,281 |
p < .05.
Occupational status
Occupational status is measured in two ways: occupational education and occupational income. Occupational education measures the percentage of workers within a given occupational category who had completed a year of college or more. It captures the credentials workers must have to enter a particular occupation. Occupational income measures the percentage of workers in the occupation who earned at least $10,000 in 1969, slightly above the median for male high school graduates.54 It measures the rewards workers receive in each occupation. Because these measures are not normally distributed, they are transformed in the following manner,
where y indicates the status measures.
Earnings
Respondents provided information about their earnings in both the 1975 and 1992 surveys. These measures are transformed using a started log (ln (1000 + earnings)). Because there are a few outliers in the transformed distributions, the measures are top coded at 2.5 standard deviations above the mean of the logged distribution. This results in top coding 1 percent to 2 percent of the distribution for each of the measures.
Independent Variables
Military service
The key variables of interest refer to military service. The most basic measure denotes whether the respondent had, in the words of the survey, “ever been on active duty in the U.S. Armed Forces or spent at least two months on active duty for training in the Reserves or National Guard.” The veteran sample is limited to men who had at least twelve months of active duty. Men with less than twelve months of duty are considered to be nonveterans.55 The measure of active duty service is derived from a question asked in 1975. If the 1975 measure is missing, this measure is updated with information from the 1992 wave of the survey. All other measures of military service are obtained from the 1975 wave of the survey. A second military measure concerns whether or not men were drafted. Men who indicated that they first entered the armed forces by being drafted are considered draftees. As noted above, some of those who reported entering as volunteers may have done so to avoid the draft. These men should be considered draft-motivated volunteers. Unfortunately, the data do not identify which veterans truly volunteered and which did so to avoid the draft. A third military measure is based on a question that was asked in 1975, which prompted the veterans to provide their rank at separation or their current rank.
The analyses also include an indicator of the length of additional training that recruits received, measured in months. Veterans were asked to report any military training programs or schools that they completed other than basic training. The measure used is based on the number of months that they reported participating in these additional military training programs or schools. The respondents could report three such training programs. In the survey, 45 percent of veterans reported just one program, 17 percent reported two programs, and 18 percent reported three programs. The respondents could describe the major field of study for their training program according to more than 600 possibilities. For the first training program, they reported 200 different major fields of study. The most commonly reported fields of study were electronics, audiovisual communication, and administration, with 4 percent of the veterans who reported having a first training program reporting training in each of these areas. Slightly less common fields of study were officers' training, mechanical maintenance, signal intelligence, law enforcement, and medical assistant, with 3 percent of veterans reporting training in each of these areas. In addition, thirty-five veterans, or 2 percent of those reporting a first field of military training, reported being trained in maneuver combat arms, while six veterans, less than 1 percent, reported being trained in combat surveillance and target acquisition. Previous research has suggested that veterans trained in non-combat specialties may benefit from their training, while those trained in combat specialties do not.56 Therefore, the article tests the robustness of the training measure by running alternate analyses that exclude the sixty-seven veterans who reported being trained in any way for combat in any of the three possible training programs. This exclusion does not affect the substantive results.
Family background
The assessment of family background is based on six measures: average family income, father's occupational status, mother's and father's education, number of siblings, and farm origin. These measures are standard measures used to capture family background.57 Family income is the average family income taken from tax records filed between 1957 and 1960. Father's occupational status is based on the graduate's 1975 retrospective report of father's occupation transformed into the Duncan Socioeconomic Index (SEI), which has a range of 0 to 100. The SEI is the result of a regression of occupational category on the average educational attainment and earnings in the occupation.58 The measures of parental education are based first on retrospective reports of parents' years of education provided in 1975 and then on reports provided in 1957. The number of siblings ever born was enumerated in the 1975 survey. This variable is top-coded at 11. Farm origin is based on geographic information collected in 1957.
Pre-service attainment and plans
As described above, previous research has found that veterans differ from nonveterans in ways that may explain their ultimate attainment.59 For example, veterans may have lower academic abilities than nonveterans and choose to enter the armed forces as an alternative to enrolling in college. These preexisting differences may then affect their ultimate socioeconomic attainment. The analyses therefore include a measure of pre-service ability, based on the Henmon-Nelson IQ tests taken in the freshman and junior years of high school. High school rank reflects percentiles that were calculated within each school, based on high school grades. Both the IQ and high school rank measures are normalized with means of 100. Veterans may also have different goals than do nonveterans. These different goals may lead them to enter the military and also determine the course of their later working lives, causing the apparent association between military service and later-life outcomes. Therefore, the analyses include four aspirations variables: military service, college education, and professional or managerial occupation. These measures are based on information that the respondents provided in 1957.
Educational attainment
Education prior to entering the armed forces is ascertained with the following question: “What was the highest grade of regular school you completed before you first entered active military service?” This measure consists of answers given in 1975 and 1992. Including the measure allows for a difference in the effects of pre- and post-service education. Total educational attainment is indicated by a measure of college graduation, regardless of whether the schooling was completed before or after service. Men who graduated from college are compared to men who had a high school degree but did not graduate from college. This measure is obtained from a combination of information provided in 1975 and 1992.
Method
As the preceding section indicates, the article assesses socioeconomic attainment using multiple measures. It tests the assumption that these multiple outcomes indicate a single underlying factor. In other words, it tests whether earnings and occupational status measured at the same point in time are related to each other and whether these socioeconomic indicators measured at one point in time are related to the same indicators measured at another point in time. To model these connections, it uses a multiple indicator multiple cause (MIMIC) framework. The MIMIC model was introduced to economics and sociology in the 1970s.60 According to Mare, the MIMIC model “is a natural, rigorous, and efficient way to represent the effects of exogenous variables on multiple endogenous outcomes.”61 The virtue of this model in the current case is that it captures the enduring nature of socioeconomic attainment over the life course, akin to the economic concept of “permanent income.” The model has several advantages over a more traditional approach with a series of regressions of each of the dependent variables on the independent variables. The traditional approach implicitly assumes that the outcomes are independent of each other. However, it is likely that these multiple measures are related to each other. The MIMIC model allows a test of these relationships. In the MIMIC model reported in the following article, all of the independent variables affect one underlying latent socioeconomic construct, which is indicated by all six of the socioeconomic measures, regardless of the period in which they were reported.62Figure 1 shows the basic MIMIC model to be estimated. The current article tests whether or not the MIMIC model fits the data better than the more common method of estimating a series of separate regressions of each outcome on all of the exogenous variables. (See the appendix for a fuller description of the various models and comparisons of the fit of the MIMIC model to alternative models.)
The results derived from the MIMIC model differ from those derived from a series of regressions run on each outcome separately in two respects. First, each independent variable is constrained to affect all of the outcomes through the latent variable. It affects all of the outcomes in the same direction and proportion relative to all of the other independent variables. The effects differ only in the relative effects of the independent variables across outcomes, which are multiplied by the loading for each of the outcomes. Second, each of the error terms of the outcomes is correlated with all of the error terms of the other outcome variables. Therefore, this model allows a test of an enduring relationship between the independent variables, in this case early life experiences and characteristics, and the outcomes, in this case occupational status and earnings measured at two points in the life course.
As the figure indicates, the dependent variable is the underlying latent construct, which is indicated by the six outcome variables: one measure of earnings and two measures of occupational status assessed at two different periods. It can also be described as an index of these six measures. According to the full model, it is affected by the eighteen independent variables: four measures of military service, six measures of family background, six measures of pre-service aspirations and attainment, and, finally, two measures of educational attainment.
Findings
Pre-Service Similarities and Differences
Hypothesis 1 predicts that there are no socioeconomic and ability differences between those who served and those who did not when the armed forces recruited by means of a draft. Thus, the average characteristics of veterans and nonveterans are of interest. Table 3 shows the means and proportions of the measures used in the analysis for the entire sample and broken out by veteran status. With few exceptions, the veterans from the WLS sample, who graduated from Wisconsin high schools in 1957, were similar to the nonveterans in the sample in terms of their early family background. They came from families that were similar to those of nonveterans in terms of family income, paternal status, and parental education. They differed from nonveterans in coming from families with, on average, more siblings. These findings are consistent with hypothesis 1. They suggest that, at least during the draft of the late 1950s, the military drew men from all levels of pre-service socioeconomic status. The military, therefore, was relatively egalitarian during this peacetime draft period with respect to family background.
However, veterans differed from nonveterans in their individual characteristics measured prior to service. Veterans had lower IQ scores, and lower high school rank, were more likely to plan to enter the military, and were less likely to plan to continue their schooling. They were more likely to aspire to managerial and less likely to aspire to professional occupations. They were less likely to have graduated from college. These differences are not consistent with hypothesis 1. Even in this period of the peacetime draft, recruits tended to have lower academic abilities and educational achievements and different aspirations than did men who did not serve. Thus, an observed effect of military service may be one of selection on the basis of individual characteristics.
Table 4 shows the means and proportions for the veteran sample, broken out by rank. Hypothesis 3a predicts that, during the draft era, officers had greater pre-service socioeconomic status and abilities than did enlisted men. Indeed, officers differed from veterans who served in the enlisted ranks on all pre-service measures. Compared to the enlisted veterans, officers came from families with more resources, had greater academic and educational attainment, and had higher aspirations. They came from families with higher income, higher paternal status, higher parental education, and fewer siblings and were less likely to have grown up on farms than were enlisted veterans. They were much more likely to have graduated from college. Nearly 90 percent of officers graduated from college either before or after their service, compared to approximately 20 percent of all other veterans. These differences are consistent with hypothesis 3a.
Table 4. Means, Standard Deviations, and Proportions by Rank.
Variable | All Veterans | Enlisted | Officers | ||||
---|---|---|---|---|---|---|---|
M | SD | M | SD | M | SD | ||
Family background variables | |||||||
Average family income | 6,395.55 | 6,166.39 | 6,035.09 | 5,941.55 | 8,957.60 | 7,084.23 | |
Logged | 8.65 | 0.55 | 8.61 | 0.54 | 8.94 | 0.57 | |
Father's Socioeconomic Index | 34.66 | 22.99 | 32.46 | 21.53 | 50.32 | 26.77 | |
Father's education | 9.92 | 3.48 | 9.59 | 3.23 | 12.25 | 4.24 | |
Mother's education | 10.79 | 2.79 | 10.61 | 2.73 | 12.09 | 2.88 | |
Number of siblings | 3.20 | 2.49 | 3.32 | 2.54 | 2.35 | 1.94 | |
Farm origin | 0.17 | 0.38 | 0.19 | 0.39 | 0.09 | 0.29 | |
Academic achievement variables | |||||||
IQ score | 101.27 | 14.75 | 99.34 | 14.05 | 114.97 | 12.11 | |
High school rank | 95.90 | 13.77 | 93.82 | 12.53 | 110.68 | 13.18 | |
College graduate | |||||||
Pre-service | 0.14 | 0.04 | 0.82 | ||||
Post-service | 0.14 | 0.15 | 0.07 | ||||
Aspirations | |||||||
Serve in military | 0.40 | 0.44 | 0.11 | ||||
Continue schooling | 0.35 | 0.28 | 0.85 | ||||
Work as professional | 0.32 | 0.25 | 0.78 | ||||
Work as manager | 0.22 | 0.23 | 0.11 | ||||
Drafted | 0.15 | 0.16 | 0.03 | ||||
Enlisted | 0.88 | 1.00 | — | ||||
Officers | 0.12 | — | 1.00 | ||||
Months of military training | 5.63 | 7.72 | 5.34 | 7.32 | 7.72 | 9.93 | |
Logged | 1.41 | 0.99 | 1.39 | 0.96 | 1.55 | 1.16 | |
Socioeconomic outcomes in 1974 | |||||||
Occupational education | 36.58 | 28.42 | 32.46 | 26.41 | 65.87 | 24.87 | |
Occupational income | 41.01 | 22.51 | 38.44 | 21.33 | 59.28 | 22.22 | |
Income | 15,790.09 | 10,768.68 | 14,743.28 | 9,286.77 | 23,230.38 | 16,295.74 | |
Logged | 9.54 | 0.68 | 9.51 | 0.64 | 9.77 | 0.89 | |
Socioeconomic outcomes in 1992 | |||||||
Occupational education | 37.46 | 28.41 | 33.55 | 26.56 | 65.29 | 25.46 | |
Occupational income | 41.35 | 23.31 | 38.99 | 22.45 | 58.09 | 22.46 | |
Income | 61,344.30 | 71,265.92 | 53,654.18 | 61,266.37 | 116,002.60 | 105,605.90 | |
Logged | 10.63 | 1.07 | 10.53 | 1.05 | 11.36 | 0.91 | |
n | 1,281 | 1,123 | 158 |
Note: All variables, except logged months of training, differ at the .01 level. Logged months of training differ at the .05 level.
These preliminary findings indicate that, during the peacetime draft era, men from families with higher socioeconomic status were as likely to serve in the military as men from families with lower socioeconomic status. However, they were more likely to achieve positions with higher status within the armed forces, that is, to serve as officers. In addition, men with greater pre-service academic and educational achievements were less likely to serve in the military than were men with lower IQ scores, high school rank, and aspirations. When they did serve, they were more likely to achieve positions with higher rank.
The Privileges of Rank
Estimates from the ordinary least squares regressions
In preliminary analyses summarized in Table 5, each outcome measure is separately regressed on military rank in a series of models with the other independent variables, which are separately entered in three groups defined as family background, pre-service academic achievement, and total educational attainment.63 Each of the rows contains one of the six socioeconomic outcomes. Column 1 indicates the direction of the association between officer status and that outcome, while column 3 indicates the direction of the association between enlisted status and the outcome when only the military measures are in the model. Columns 2 and 4 indicate which category of variables, if any, explain the initial association.
Table 5. Summary of Ordinary Least Squares Regressions of Six Socioeconomic Outcomes on Military and Other Covariates.
Officer Association | Explained by | Enlisted Association | Explained by | ||
---|---|---|---|---|---|
1974 outcomes | |||||
Occupational education | Positive | — | Negative | Pre-service achievement | |
Occupational earnings | Positive | Pre-service achievement | Negative | Pre-service achievement | |
Earnings | Positive | Total education | Negative | Family background | |
1992 outcomes | |||||
Occupational education | Positive | Total education | Negative | College graduate | |
Occupational earnings | Positive | — | Negative | Pre-service achievement | |
Earnings | Positive | — | Negative | Family background |
Across all six of the outcome measures, veterans who served as officers had higher socioeconomic attainment than did nonveterans. They worked at jobs with higher status as measured by both occupational education and occupational income and had higher earnings. These findings are consistent with previous research that shows that veterans who served as officers benefited from their service. In the case of one outcome, the positive association of officer status with later socioeconomic attainment becomes insignificant after including pre-service achievement factors. On average, officers had higher IQ scores and high school rank than did nonveterans, which explains their higher occupational earnings in 1974. In two cases, the positive association between rank and socioeconomic attainment becomes insignificant due to the fact that officers were more likely than all other respondents to graduate from college. Thus, the greater average educational attainment of officers explains their correspondingly greater earnings in 1974 and occupational education in 1992. However, in the case of the remaining three outcomes, the positive association between officer status and socioeconomic attainment remains despite the differences between officers and everyone else in terms of family background, academic achievement, and educational attainment. Even when pre-service and total educational differences between officers and all others are considered, officers had higher occupational education in 1974, higher occupational earnings in 1992, and greater earnings in 1992. These findings present some evidence that officers benefited from military service net of their preexisting advantages. They are consistent with hypothesis 3b that, during this draft era, officers were able to accumulate advantage.
By contrast, veterans who served in the enlisted ranks had lower attainment than did nonveterans across all six outcomes. However, this negative association was explained by the lower pre-service attainment and family background of veterans who served in the enlisted ranks. Enlisted men came from families with fewer financial and educational resources, which explains the fact that they had lower earnings in both 1974 and 1992. They also had lower average academic credentials as measured by their IQ scores and high school rank, which explains the fact that they had lower occupational status at both points in time than did nonveterans. These findings are consistent with hypothesis 4 and suggest that, during this peacetime era, veterans' outcomes did not differ from those of nonveterans. They suggest that the apparent negative association between military service in the enlisted ranks and later outcomes in this peacetime draft era stemmed from both selection and status reproduction.
Estimates from the MIMIC model
As indicated by the fit statistics included in the appendix, the MIMIC model, which treats the multiple outcomes as related to one another, rather than distinct, fits the data better than does a series of multivariate regressions for each outcome. Table 6 presents estimates from a series of MIMIC models in the metric of the log of earnings in 1992 when the respondents were in their early fifties. As specified by the model, the pattern of results is similar across the other five outcomes.
Table 6. Effects of Independent Variable on Socioeconomic Construct, Normalized on Log of Income in 1992.
Model 1: Military Variables | Model 2: Family Background | Model 3: Pre-Service Attainment and Plans | Model 4: All Pre-Service Characteristics | Model 5: Model 4 + Human Capital | Model 6: Model 4 + Disruption | Model 7: Full Model | |
---|---|---|---|---|---|---|---|
Military rank | |||||||
Officer | 0.125*** (0.014) |
0.099*** (0.012) |
0.041*** (0.010) |
0.034*** (0.010) |
0.036*** (0.011) |
0.036*** (0.010) |
0.039*** (0.011) |
Enlisted | −0.013 (0.009) |
−0.020* (0.009) |
−0.007 (0.009) |
0.005 (0.009) |
0.007 (0.009) |
0.017 (0.009) |
0.022* (0.010) |
Logged months of training | −0.004 (0.009) |
−0.008 (0.009) |
|||||
Drafted | −0.027** (0.008) |
−0.028** (0.008) |
|||||
Family background measures | No | Yes | No | Yes | Yes | Yes | Yes |
Aspirations | No | No | Yes | Yes | Yes | Yes | Yes |
Pre-service academics or education | No | No | Yes | Yes | Yes | Yes | Yes |
Total educational attainment | No | No | No | Yes | Yes | Yes | Yes |
Note: Standard errors are in parentheses.
Significant at .05.
Significant at .01.
Significant at .001.
The table shows that officers had higher status and earnings than nonveterans throughout the life course. Pre-service differences explain much, but not all, of the officer premium. In model 1, which does not include pre-service differences, veterans who served as officers earned 12 percent, or $4,000 more, plus or minus $400, than did nonveterans in 1992. In model 2, family background variables are added to the model, which reduces the size of the officer premium to 10 percent. To an extent, officers had higher socioeconomic attainment because they came from families with higher status, income, and education than did nonveterans. At least a portion of the officer premium stemmed from status reproduction. The officer premium was also partially explained by differences between the pre-service aspirations and achievements of officers and nonveterans as measured by IQ score, high school rank, and pre-service education. In model 4, when all of the differences between officers and non-officers are considered, the officer premium is reduced to approximately one-fourth of its initial value. Said another way, three-fourths of the officer premium stemmed from pre-service and total educational differences between officers and nonveterans. These results suggest that a substantial portion of the officer premium stemmed from selection. Yet when all pre-service differences are considered, veterans who served as officers still earned 3 percent, or $1,100 more, plus or minus $300, than did nonveterans with comparable characteristics. This finding is consistent with hypothesis 3b, which stated that, during the draft period, officers accumulated advantage.
To put this effect of officer status in context, in 1992 the men who graduated from college in this cohort earned 15 percent more than did men who graduated only from high school. Figure 2 shows the effects of college graduation, and of officer and enlisted status, on the log of earnings at both the 1992 and 1975 waves of the survey. The figure is based on the full model, in which all pre-service family and individual characteristics are held at their means or modal categories. Recall that in 1975 the members of the sample were, on average, thirty-six years old, while in 1992 they were, on average, fifty-three years old. As the figure shows, the gaps between college graduates and veteran officers, on one hand, and nonveteran high school graduates and veteran enlisted men, on the other, widened over the two waves of the survey. This may reflect the accumulation of advantage as the sample respondents aged. It may also reflect the period increase in income inequality. Previous research has noted that during the past thirty years the earnings gap between college graduates and people with less education widened.64 The analyses suggest that the gap between officers and people who did not serve as officers may also have increased during the same period. Regardless of the interpretation, the figure shows that the officer premium persisted across the life course, similar to the premium afforded to college graduates. Veterans who served as officers earned a premium relative to nonveterans when they were in their mid-thirties and when they were in their mid-fifties. This finding is consistent with hypothesis 3b, which suggested that veteran officers accumulated advantage. However, the benefit to socioeconomic attainment of serving as an officer was not as large as the benefit of graduating from college.
The effects of officer status and a college degree cannot be easily separated from one another. Approximately 40 percent of nonveterans graduated from college, compared to 20 percent of veterans who served in the enlisted ranks and 90 percent of veterans who served as officers. Therefore, the benefit that officers received was primarily in addition to the benefit they received from their college education.
The officer premium did not derive from military training. In Table 6, model 5 includes a measure of months of training that veterans received while in the armed forces. This measure did not affect socioeconomic attainment, nor did it explain the officer premium.65 Thus, the officer premium was not a result of increased human capital while in the military, at least as measured by months of training. As model 6 shows, the negative effect of being drafted suppressed the slightly positive effect of serving in the enlisted ranks. Draftees earned 1 percent less than did nonveterans with equivalent characteristics. This finding presents weak support for hypothesis 2. Relative to veterans who volunteered, veterans who were drafted experienced their military service as a disruption. This effect was counterbalanced, however, by the small positive effect of serving in the enlisted ranks.
Conclusion
The outset of the article posed the following questions: Did military service provide veterans with the credentials or skills they needed to do well at work? Did it substitute for formal higher education? The article has suggested that the answers to these questions can best be assessed with reference to historical variation in three dimensions: the conscription dimension, which determines the types of people who serve in the military; the conflict dimension, which determines the experience that those people have while in the armed forces; and the benefits dimension, which affects the resources that they receive after leaving the military. The article has focused on the post-service experiences of veterans who served during a peacetime draft era and who had limited access to educational benefits.
To some extent, military service during this draft era reproduced prior civilian status. When all men are considered, veterans did not systematically differ from nonveterans in terms of their family characteristics. Yet even during the peacetime draft era, men from families with higher socioeconomic status were more likely than men from less privileged backgrounds to become officers. These findings cast light on the debate regarding whether or not military enlistment and service were governed by class bias.66 They suggest that officer status reproduced pre-service status.
Net of these pre-service differences, veterans who served as officers had higher earnings and occupational status than did nonveterans. Thus, during this peacetime draft era, veterans from families with greater economic and educational resources benefited more from their service than did veterans from families with fewer resources. These benefits lasted throughout the life course. Even when the veterans were in their early fifties, those who had served as officers earned more and had higher occupational status than did nonveterans, net of pre-service differences and total education. This finding suggests that serving as an officer in the armed forces increased these veterans' human capital. However, this positive effect was not a consequence of the formal training provided by the military. It may have stemmed from differences in the military experience itself. In addition, officers may have formed friendships and made connections while in the armed forces that led to their greater success later in life. They may have entered a “pipeline” of individual connections that facilitated greater achievement over the life course.67 These findings suggest that military service enabled men who served as officers to accumulate advantage.
This finding is consistent with previous research. Two previous articles also find that veterans who served as officers in different periods benefited from their military service.68 This previous research has not examined the size of the officer premium relative to that received by college graduates. The analyses presented in the preceding article showed that veteran officers did not receive as large a premium as did college graduates. Therefore, military service did not substitute for formal higher education. In addition, in this peacetime draft era, veterans who served as officers were more likely than those who served in the enlisted ranks and than nonveterans to have completed college. Military service as an officer provided, for the most part, a supplemental premium for those who had a college degree. This article, therefore, contributes to an understanding of the size of the benefit of officer status relative to the size of the benefit of a college degree.
The article also contributes to an understanding of why officer status is positively associated with later attainment. It has considered three possible explanations: selection, status reproduction, and cumulative advantage. In this peacetime draft era, men with greater pre-service socioeconomic and academic advantages were more likely to become officers than to become enlisted men. Indeed, pre-service academic achievements explained a portion of the officer premium. Veteran officers had higher high school rank, IQ scores, and aspirations than did non-officers, which partially explained their greater later socioeconomic success. They were also more likely than veteran enlisted men to have graduated from college. Thus, they benefited from military skills or credentials that were, for the most part, in addition to their greater pre-service formal educational advantage. They benefited from formal education and from the privileges of rank. The article therefore suggests that the interpretation of the officer premium is partly consistent with all three explanations: selection, status reproduction, and cumulative advantage.
It is worth noting that the data did not provide evidence that military service was a positive turning point for particular types of veterans. There were no interactions between military service and individual or family characteristics. (See the appendix for more details regarding model fit.) Had they been present, these interactions would have indicated that veterans from relatively disadvantaged backgrounds experienced more positive outcomes than they would have had they not served in the armed forces. In addition, a model in which there were no interactions between military service and college graduation fit better than a model that included interactions between college graduation and the military variables, indicating that the effects of the military variables did not differ across educational level. This finding reinforces the findings of previous research that found that high school graduates and college graduates experienced the same effects of veteran status during the draft era.69 In contrast, veterans of the more recent AVF who graduated from college have experienced a wage penalty based on their service.70 Military service did not serve as a turning point for members of different subgroups during this peacetime draft era in which veterans did not receive educational benefits.
It is also worth noting that the data did not support hypothesis 5, that veterans in an era without benefits experienced military service as a disruption. It may be that the draft, which reduced average differences between veterans and nonveterans, counteracted the negative effect of the lack of benefits. It may also be that the absence of conflict during this era reduced the negative impact of military service. Thus, during the peacetime draft era without benefits, the primary effect of military service was to improve the lives of veterans who served as officers, who also entered the armed forces with greater pre-service advantages, consistent with the theory of cumulative advantage. Future research should examine the extent of the officer premium in other eras, paying specific attention to shifts in the historical contexts of conscription, conflict, and benefits.
Acknowledgments
This research was conducted using the facilities of the Center for Demography and Ecology at the University of Wisconsin–Madison, which is supported by the National Institute of Child Health and Human Development (R24 HD 47873). The work was also supported by training grants from the National Institute on Aging to the Center for Demography of Health and Aging at the University of Wisconsin–Madison (T32 AG00129) and to the RAND Corporation (5 T32 AG00024410). I am grateful for comments and suggestions from Patricia M. Shields, Jeremy Freese, Charles N. Halaby, Robert M. Hauser, Diane L. Lindstrom, Andy Sauber Olds, James Yocom, and the anonymous reviewers of Armed Forces & Society.
Biography
Alair MacLean is an assistant professor of sociology at Washington State University Vancouver. Her research centers on the intersection of social inequality, military service, and the life course. She has published articles in Annual Review of Sociology, Sociology of Education, and Research in Social Stratification and Mobility.
Appendix
As mentioned in the article, the multiple indicator multiple cause (MIMIC) model was compared to alternative models on the basis of goodness-of-fit statistics. Two models were considered as alternatives. The first model treats each of the six different socioeconomic outcomes as separate from each other. This model is similar to the standard approach of estimating regressions of each outcome one after the other on all of the independent variables. It allows the early life characteristics to affect each of the outcomes in different ways. To take just one possible example, military service may positively affect occupational status in the middle of the work life but not affect earnings near the end of the work life.
The second model is a two-factor model, in which all of the independent variables affect an underlying latent variable that is indicated by the three outcome measures at the first point in time: occupational education, occupational income, and earnings in 1975. This latent variable, in turn, affects a latent variable indicated by occupational education, occupational income, and earnings in 1992. The independent variables directly affect socioeconomic attainment at each period. They also indirectly affect 1992 socioeconomic attainment through attainment in 1975. The virtue of this approach is that it takes advantage of the fact that the data contain not just one but multiple measures of socioeconomic attainment at each period. The occupational status measures and the earnings measure are assumed to be correlated with each other at each time but not across periods.
By contrast, the MIMIC model assumes that all of the outcomes are correlated regardless of the time at which they were measured. This model is described in the methods section of the article. It is based on the following set of three equations. The first equation is for the measurement model for the outcome or indicator variables,
(1) |
where y is a 6 × 1 vector of the 6 socioeconomic outcome variables, η1 is the latent endogenous variable, λy is a 6 × 1 matrix of coefficients of the regression of the 6 y variables on η1, and ε is a 6 × 1 vector of measurement errors in the 6 y variables, which are allowed to be freely correlated. The structural model for the underlying factor is calculated by,
(2) |
where η1 is as defined above, Γ is a 1 × 18 matrix of coefficients of the relationship between the 18 ξs and η1, ξ is a 18 × 1 vector of the 18 latent independent variables. The measurement of the independent variables is specified by the following equation,
(3) |
where x is a 18 × 1 vector of the 18 independent variables, ξ is as defined above, and δ is a 18 × 1 matrix of measurement errors in the 18 x variables.71
The preferred model is chosen using the Bayesian information criterion (BIC). BIC is a measure of goodness of fit that takes into account sample size.72 BIC is calculated as follows,
(4) |
where χ2 is the likelihood ratio test statistic for comparing the model Mk to the saturated model, in which all data points are described exactly, dfk is the degrees of freedom associated with Mk, and n is the sample size. Negative values of BIC suggest that the model Mk is preferable to the saturated model. Differences of −6 constitute strong evidence that a particular model is preferable to another model.73
The basic model is then estimated in a multiple group context, in which the groups are defined by their characteristics. For example, the model is estimated as a two-group model in which the groups are veterans and nonveterans. This allows for a formal test of the interactions between group characteristics and the independent variables by constraining the effects to be the same or allowing them to differ across the multiple groups. An alternative multiple-group model, in which the multiple groups are defined by educational attainment, is also estimated. This is equivalent to an ordinary least squares regression in which educational attainment is interacted with the independent variables, thus allowing the independent variables to have different effects at different levels of educational attainment. In other words, this model suggests that the effects of military service and rank differ among high school graduates from those effects among college graduates. This model fit is worse than the fit for the other models.
The most constrained, parsimonious model of the relationship between early life experiences and later-life outcomes fits better than the more complex, detailed models that include multiplicative interactions. Table A1 shows the fit statistics for the series of models estimated using the full sample of veterans and nonveterans. As the table shows, model 3, the MIMIC model with a single underlying socioeconomic factor, fits better than both model 1, which separately estimates the equations for each of the separate outcomes, and model 2, a two-factor model in which the independent variables affect the outcomes at the two different periods differently.74 Models 4 through 7 are estimated in a multiple-group context. The multiple-group models test for interactions between the independent variables and veteran status. In all cases, the interactions between veteran status and the other independent variables lead to a deterioration in model fit. These results suggest that the background characteristics and experiences had the same effects on socioeconomic attainment regardless of whether an individual was a veteran or a nonveteran. The presence of interactions between military service and the independent variables was tested in two other ways (fit statistics available on request). First, a multiple-group model, in which the groups were volunteers, draftees, and officers, was estimated using data just from the sample of veterans. This allows a test of interactions between, for instance, officer status and length of training. None of the interactions improved the fit of the model. Second, based on the literature showing differences in the effect of military service by educational attainment, a series of models in which college graduates were treated as one group and men who did not graduate from college were treated as another was estimated. The military service variables had the same effects regardless of whether men were college graduates or not. This suggests that the veteran status variables did not interact with college graduation. Thus, the effects of military rank were the same regardless of whether veterans were college graduates or not. Therefore, the article presents results based on a single-group MIMIC model.
Table A1. Bayesian Information Criterion (BIC) Statistics of Models Comparing Veterans to Nonveterans.
χ2 | df | BIC | |
---|---|---|---|
Single group models | |||
1. Separate equations, each outcome | 3,331.09 | 15.00 | 3,210.91 |
2. Two-factor model | 407.05 | 61.00 | −81.66 |
3. MIMIC model | 48.87 | 65.00 | −471.89 |
Multiple group models | |||
4. Multiple group MIMIC model | 129.39 | 159.00 | −1,147.19 |
5. 3+ family interactions | 127.89 | 154.00 | −1,108.54 |
6. 3+ individual characteristics | 119.56 | 153.00 | −1,108.84 |
7. 3+ college graduation interaction | 124.78 | 158.00 | −1,143.77 |
BIC contrasts | |||
Two factor model vs. separate regressions (2–1) | −3,292.58 | ||
MIMIC vs. two-factor model (3–2) | −390.23 | ||
Multiple groups (4–3) | −675.30 | ||
Differences in effect of family background (5–4) | 38.64 | ||
Differences in effect of aspirations (6–4) | 38.34 | ||
Differences in effect of education (7–4) | 3.42 |
Note: MIMIC = multiple indicator multiple cause.
Notes
- 1.Mason William M. PhD diss. University of Chicago; 1970. On the Socioeconomic Effects of Military Service. [Google Scholar]
- 2.De Tray Dennis. Veteran Status as a Screening Device. American Economic Review. 1982;72(1) [Google Scholar]; Teachman Jay D, Call Vaughn RA. The Effect of Military Service on Educational, Occupational, and Income Attainment. Social Science Research. 1996;25(1) [Google Scholar]
- 3.Angrist Joshua D, Krueger Alan B. Why Do World-War-II Veterans Earn More than Nonveterans? Journal of Labor Economics. 1994;12(1) [Google Scholar]
- 4.Angrist Joshua D. Lifetime Earnings and the Vietnam Era Draft Lottery—Evidence from Social-Security Administrative Records. American Economic Review. 1990;80(3) [Google Scholar]; Mangum Stephen L, Ball David E. The Transferability of Military Provided Occupational Training in the Post-Draft Era. Industrial and Labor Relations Review. 1989;42(2) [Google Scholar]; Teachman, Call . The Effect of Military Service. [Google Scholar]
- 5.Teachman Jay D. Military Service during the Vietnam Era: Were There Consequences for Subsequent Civilian Earnings? Social Forces. 2004;83(2) [Google Scholar]
- 6.Angrist Joshua D. Estimating the Labor Market Impact of Voluntary Military Service Using Social Security Data on Military Applicants. Econometrica. 1998;66(2) [Google Scholar]
- 7.Dechter Aimee, Elder Glen H., Jr World War II Mobilization in Men's Work Lives: Continuity or Disruption for the Middle Class. American Journal of Sociology. 2004;110(3) doi: 10.1086/422662. [DOI] [PMC free article] [PubMed] [Google Scholar]; Hirsch BT, Mehay SL. Evaluating the Labor Market Performance of Veterans Using a Matched Comparison Group Design. Journal of Human Resources. 2003;38(3) [Google Scholar]
- 8.Angrist . Lifetime Earnings. [Google Scholar]; Angrist, Krueger . Why Do World-War-II Veterans Earn More. [Google Scholar]; De Tray . Veteran Status. [Google Scholar]; Mangum, Ball . The Transferability. [Google Scholar]; Teachman, Call . The Effect of Military Service. [Google Scholar]
- 9.Angrist . Estimating the Labor Market Impact. [Google Scholar]; Xie Yu. The Socioeconomic Status of Young Male Veterans, 1964–1984. Social Science Quarterly. 1992;73(2):379–96. [Google Scholar]
- 10.Flynn George Q. The Draft, 1940–1973. Lawrence: University Press of Kansas; 1993. [Google Scholar]
- 11.For an exception, see Hauser Robert M. The Earnings Trajectories of Young Men. CDE Working Paper 79-24; Madison, WI. 1979.
- 12.Dechter, Elder . World War II Mobilization. [Google Scholar]; Hirsch, Mehay . Evaluating the Labor Market Performance. [Google Scholar]
- 13.Berger Mark C, Hirsch Barry T. The Civilian Earnings Experience of Vietnam-Era Veterans. Journal of Human Resources. 1983;18 [Google Scholar]; Bryant Richard R, Samaranayake VA, Wilhite Allen. The Effect of Military Service on the Subsequent Civilian Wage of the Post-Vietnam Veteran. Quarterly Review of Economics and Finance. 1993;33(1) [Google Scholar]; Bryant Richard, Wilhite Al. Military Experience and Training Effects on Civilian Wages. Applied Economics. 1990;22(1) [Google Scholar]; Goldberg Matthew S, Warner John T. Military Experience, Civilian Experience and the Earnings of Veterans. Journal of Human Resources. 1987;32(1) [Google Scholar]; Teachman . Military Service During the Vietnam Era. [Google Scholar]
- 14.Angrist Joshua D. The Effect of Veterans' Benefits on Education and Earnings. Industrial and Labor Relations Review. 1993;46(4) [Google Scholar]
- 15.Hauser . The Earnings Trajectories. [Google Scholar]; Teachman . Military Service During the Vietnam Era. [Google Scholar]
- 16.Browning Harley L, Lopreato Sally C, Poston Dudley L., Jr Income and Veteran Status: Variations among Mexican Americans, Blacks and Whites. American Sociological Review. 1973;38(1)Lopreato Sally C, Poston Dudley L., Jr Differences in Earnings and Earnings Ability between Black Veterans and Nonveterans in the United States. Social Science Quarterly. 1977;57(4)Poston Dudley L., Jr The Influence of Military Service on the Civilian Earnings Patterns of Blacks, Mexican Americans, and Anglos. Journal of Political and Military Sociology. 1979 Spring;7 But also see Cutright Phillips. The Civilian Earnings of White and Black Draftees and Nonveterans. American Sociological Review. 1974;39(3)
- 17.Johnson Robert J, Kaplan Howard B. Psychosocial Predictors of Enlistment in the All-Voluntary Armed Forces: A Life-Event-History Analysis. Youth & Society. 1991;22(3) [Google Scholar]; Sampson Robert J, Laub John H. Socioeconomic Achievement in the Life Course of Disadvantaged Men: Military Service as a Turning Point, Circa 1940–1965. American Sociological Review. 1996;61(3) [Google Scholar]
- 18.Angrist . Estimating the Labor Market Impact. Berger, Hirsch . The Civilian Earnings Experience. Bryant, Samaranayake, Wilhite . The Effect of Military Service. But also see Teachman . Military Service During the Vietnam Era.
- 19.Brotz Howard, Wilson Everett. Characteristics of Military Society. American Journal of Sociology. 1946;51(5) doi: 10.1086/219845. [DOI] [PubMed] [Google Scholar]; Elder Glen H., Jr Military Times and Turning Points in Men's Lives. Developmental Psychology. 1986;22(2) [Google Scholar]; Elder Glen H., Jr War Mobilization and the Life Course: A Cohort of World War II Veterans. Sociological Forum. 1987;2(3) [Google Scholar]; Elder Glen H, Jr, Hareven Tamara K. Rising above Life's Disadvantage: From the Great Depression to War. In: Elder Glen H, Modell J, Parke RD., editors. Children in Time and Place. New York: Cambridge University Press; 1993. [Google Scholar]
- 20.Merton Robert King. The Matthew Effect in Science. Science. 1968;159(3810) [PubMed] [Google Scholar]
- 21.Dannefer D. Cumulative Advantage/Disadvantage and the Life Course: Cross-Fertilizing Age and Social Science Theory. Journals of Gerontology Series B–Psychological Sciences and Social Sciences. 2003;58(6) doi: 10.1093/geronb/58.6.s327. [DOI] [PubMed] [Google Scholar]
- 22.London Andrew S, Wilmoth Janet M. 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. 2006;28(1) [Google Scholar]
- 23.Bourdieu Pierre. Distinction: A Social Critique of the Judgment of Taste. Cambridge, MA: Harvard University Press; 1984. [Google Scholar]
- 24.Levy Yagil. Militarizing Inequality: A Conceptual Framework. Theory and Society. 1998;27(6) [Google Scholar]
- 25.Moskos Charles C. The American Enlisted Man: The Rank and File in Today's Military. New York: Russell Sage; 1970. [Google Scholar]
- 26.Appy Christian G. Working-Class War: American Combat Soldiers and Vietnam. Chapel Hill: University of North Carolina Press; 1993. [Google Scholar]; Fallows James. What Did You Do in the Class War, Daddy? Washington Monthly. 1975;7(8) [Google Scholar]
- 27.Moskos . The American Enlisted Man. [Google Scholar]; Stouffer Samuel Andrew. The American Soldier: Adjustment During Army Life, volume 2, Studies in Social Psychology in World War II. Princeton, NJ: Princeton University Press; 1949. [Google Scholar]
- 28.Angrist . The Effect of Veterans' Benefits. [Google Scholar]
- 29.Card Josefina J. Lives after Vietnam: The Personal Impact of Military Service. Lexington, MA: Lexington Books; 1983. [Google Scholar]; Elder . Military Times. [Google Scholar]
- 30.Mayer Albert J, Hoult Thomas Ford. Social Stratification and Combat Survival. Social Forces. 1955;34 [Google Scholar]; Zeitlin M, Lutterman KG, Russell JW. Death in Vietnam: Class, Poverty, and the Risks of War. Politics and Society. 1973;3(3) [Google Scholar]
- 31.Appy . Working-Class War. [Google Scholar]; Fallows . What Did You Do. [Google Scholar]; Zeitlin, Lutterman, Russell . Death in Vietnam. [Google Scholar]
- 32.Cohen Jere, Segal David R, Temme Lloyd V. The Educational Cost of Military Service in the 1960s. Journal of Political and Military Sociology. 1986;14(2) [Google Scholar]; Cohen Jere, Warner Rebecca L, Segal David R. Military Service and Educational Attainment in the All-Volunteer Force. Social Science Quarterly. 1995;76(1) [Google Scholar]; Teachman, Call . The Effect of Military Service. [Google Scholar]
- 33.Teachman Jay D, Call Vaughn RA, Segal Mady Wechsler. The Selectivity of Military Enlistment. Journal of Political and Military Sociology. 1993;21(2) [Google Scholar]
- 34.Flynn . The Draft, 1940–1973. [Google Scholar]
- 35.Ibid.
- 36.Ibid.
- 37.Bailey Duncan, Cargill Thomas F. The Military Draft and Future Income. Western Economics Journal. 1969;4 [Google Scholar]; Hansen W Lee, Weisbrod Burton A. Economics of the Military Draft. Quarterly Journal of Economics. 1967;81(3) [Google Scholar]; Miller James C, III, Tollison Robert. The Implicit Tax on Reluctant Military Recruits. Social Science Quarterly. 1971;51 [Google Scholar]; Oi Walter Y. The Economic Cost of the Draft. American Economic Review. 1967;57(2) [Google Scholar]
- 38.Teachman . Military Service During the Vietnam Era. [Google Scholar]
- 39.Flynn . The Draft, 1940–1973. [Google Scholar]
- 40.Mayer, Hoult . Social Stratification. [Google Scholar]; Zeitlin, Lutterman, Russell . Death in Vietnam. [Google Scholar]
- 41.Angrist, Krueger . Why Do World-War-II Veterans Earn More. [Google Scholar]; Bedard K, Deschenes O. The Long-Term Impact of Military Service on Health: Evidence from World War II and Korean War Veterans. American Economic Review. 2006;96(1) doi: 10.1257/000282806776157731. [DOI] [PubMed] [Google Scholar]
- 42.Dechter, Elder . World War II Mobilization. [Google Scholar]; Hirsch, Mehay . Evaluating the Labor Market Performance. [Google Scholar]
- 43.Merli MG. Socioeconomic Background and War Mortality During Vietnam's Wars. Demography. 2000;37(1) [PubMed] [Google Scholar]
- 44.Elder Glen H, Jr, Shanahan Michael J, Clipp Elizabeth Colerick. Linking Combat and Physical Health: The Legacy of World War II in Men's Lives. American Journal of Psychiatry. 1997;154 doi: 10.1176/ajp.154.3.330. [DOI] [PubMed] [Google Scholar]; Villa VM, Harada ND, Washington D, Damron-Rodriguez J. Health and Functioning among Four War Eras of Us Veterans: Examining the Impact of War Cohort Membership, Socioeconomic Status, Mental Health, and Disease Prevalence. Military Medicine. 2002;167(9) [PubMed] [Google Scholar]
- 45.Hoge Charles W, Castro Carl A, Messer Stephen C, McGurk Dennis, Cotting Dave I, Koffman Robert L. Combat Duty in Iraq and Afghanistan, Mental Health Problems, and Barriers to Care. New England Journal of Medicine. 2004;351(1) doi: 10.1056/NEJMoa040603. [DOI] [PubMed] [Google Scholar]; Kang Han K, Hyams KC. Mental Health Care Needs among Recent War Veterans. New England Journal Of Medicine. 2005;352(13) doi: 10.1056/NEJMp058024. [DOI] [PubMed] [Google Scholar]
- 46.Angrist . The Effect of Veterans' Benefits. [Google Scholar]; Fligstein Neil. master's thesis. University of Wisconsin; 1976. Military Service, the Draft, and the GI Bill: Their Effects on the Lives of American Males. [Google Scholar]; Mason . On the Socioeconomic Effects. [Google Scholar]; Veterans Administration. Education Tuition Assistance Study: A Report Prepared by the Veterans' Administration. Washington, DC: Veterans Administration; 1975. [Google Scholar]; Xie . The Socioeconomic Status. [Google Scholar]
- 47.Mettler Suzanne. Soldiers to Citizens: The GI Bill and the Making of the Greatest Generation. New York: Oxford University Press; 2005. [Google Scholar]
- 48.Angrist . The Effect of Veterans' Benefits. [Google Scholar]; Sampson, Laub . Socioeconomic Achievement. [Google Scholar]; Stanley M. College Education and the Midcentury GI Bills. Quarterly Journal of Economics. 2003;118(2) [Google Scholar]
- 49.Stanley . College Education. [Google Scholar]
- 50.Angrist . The Effect of Veterans' Benefits. [Google Scholar]; Fligstein . Military Service. [Google Scholar]; Nam Charles B. Impact of the ‘GI Bills’ on the Educational Level of the Male Population. Social Forces. 1964;43(1) [Google Scholar]
- 51.Veterans Administration. Education Tuition. [Google Scholar]
- 52.MacLean Alair. Lessons from the Cold War: Military Service and College Education. Sociology of Education. 2005;78(3) [Google Scholar]
- 53.Sewell William H, Hauser Robert M, Springer Kristen W, Hauser Taissa S. As We Age: A Review of the Wisconsin Longitudinal Study, 1957–2001. Research in Social Stratification and Mobility. 2004;20 [Google Scholar]
- 54.Hauser Robert M, Warren John Robert. Socioeconomic Indexes for Occupations: A Review, Update, and Critique. Sociological Methodology. 1997;27 [Google Scholar]
- 55.When the men who had less than twelve months of active duty are treated as veterans, the results are substantively similar.
- 56.Bryant, Samaranayake, Wilhite . The Effect of Military Service. [Google Scholar]; Bryant, Wilhite . Military Experience and Training. [Google Scholar]
- 57.Blau Peter Michael, Duncan Otis Dudley. The American Occupational Structure. New York: John Wiley; 1967. [Google Scholar]; Featherman David L, Hauser Robert M. Opportunity and Change Studies in Population. New York: Academic Press; 1978. [Google Scholar]
- 58.Duncan Otis Dudley. A Socioeconomic Index for All Occupations. In: Reiss A Jr, editor. Occupations and Social Status. New York: Free Press; 1961. [Google Scholar]
- 59.Cohen, Segal, Temme . The Educational Cost of Military. [Google Scholar]; Cohen, Warner, Segal . Military Service. [Google Scholar]; Teachman, Call . The Effect of Military Service. [Google Scholar]
- 60.Hauser Robert M. Disaggregating a Social-Psychological Model of Educational Attainment. Social Science Research. 1972;1 [Google Scholar]; Hauser Robert M, Goldberger Arthur S. The Treatment of Unobservable Variables in Path Analysis. In: Costner HL, editor. Sociological Methodology, 1971. San Francisco: Jossey-Bass; 1971. [Google Scholar]
- 61.Mare RD. Response: Statistical Models of Educational Stratification—Hauser and Andrew's Models for School Transitions. Sociological Methodology. 2006;36:27. [Google Scholar]
- 62.Hauser, Goldberger . The Treatment of Unobservable Variables. [Google Scholar]; Joreskog Karl G, Sorbom Dag. Lisrel 8: User's Reference Guide. Lincolnwood, IL: Scientific Software International; 2001. [Google Scholar]
- 63.I ran the ordinary least squares regressions without the veterans who had received combat training as describe above, and the point estimates were slightly different in some cases. In most cases the estimate of the effect of officer status was slightly larger (e.g., .277 vs. .264) and in one case slightly smaller (.337 vs. .328) when combat-trained veterans were excluded. However, the substantive findings did not differ when the models were estimated with and without the veterans who were trained according to combat specialties. Therefore, I present the results based on data including the combat-trained veterans. In addition, I ran the analyses without the respondents who reported that they both were officers and had been drafted. This exclusion did not influence the results. Therefore, I present findings based on data including officers who reported having been drafted.
- 64.Morris Martina, Bruce Western. Inequality in Earnings at the Close of the Twentieth Century. Annual Review of Sociology. 1999;25 [Google Scholar]; Krueger Alan B. How Computers Have Changed the Wage Structure: Evidence from Microdata, 1984–1989. Quarterly Journal of Economics. 1993;108(1):33–60. [Google Scholar]
- 65.Training had the same effect for officers and enlisted men. In the veteran sample, there is no interaction between officer status and training.
- 66.Appy . Working-Class War. [Google Scholar]; Card . Lives after Vietnam. [Google Scholar]; Fallows . What Did You Do. [Google Scholar]; Mayer, Hoult . Social Stratification. [Google Scholar]; Zeitlin, Lutterman, Russell . Death in Vietnam. [Google Scholar]
- 67.Dechter, Elder . World War II Mobilization. [Google Scholar]
- 68.Cohen, Segal, Temme . The Educational Cost. [Google Scholar]; Dechter, Elder . World War II Mobilization. [Google Scholar]; Hirsch, Mehay . Evaluating the Labor Market Performance. [Google Scholar]
- 69.Berger, Hirsch . The Civilian Earnings Experience. [Google Scholar]
- 70.Bryant, Samaranayake, Wilhite . The Effect of Military Service. [Google Scholar]
- 71.For more details on the multiple indicator multiple cause model, see Hauser . Disaggregating a Social-Psychological Model. Hauser, Goldberger . The Treatment of Unobservable Variables. Joreskog, Sorbom . Lisrel 8. Mare . Response.
- 72.Raftery AE. Bayesian Model Selection in Social Research. In: Marsden V Peter., editor. Sociological Methodology 1995. Cambridge, MA: Basil Blackwell; 1995. [Google Scholar]
- 73.Ibid.
- 74.The substantive results differ slightly when missing values are imputed using the imputation procedure in Prelis, which relies on an expectation maximization algorithm (Lisrel 8.54). The effect of being drafted is larger and statistically significant in the imputed data set. All other estimates are the same regardless of whether the models are based on listwise deleted or imputed data. Therefore, I present the more conservative estimates based on the smaller sample of complete cases.