Social Science Research 39 (2010) 477–490 Contents lists available at ScienceDirect Social Science Research journal homepage: www.elsevier.com/locate/ssresearch Where did the soldiers go? The effects of military downsizing on college enrollment and employment q Meredith Kleykamp * University of Kansas, 1415 Jayhawk Blvd., Fraser 716, Lawrence, KS 66045-7556, USA a r t i c l e i n f o Article history: Available online 13 September 2009 Keywords: Military downsizing Employment College enrollment Stratification a b s t r a c t This paper examines how the military drawdown in the early 1990s influenced aggregate trends in employment and college enrollment, evaluating whether the loss of military jobs resulted in observable increases or decreases in employment rates and/or college enrollment rates. Contrary to the expectation of worsening employment among black men in particular, the drawdown had little effect on employment. However, changes in military service did have a considerable impact on college enrollment among black men. The loss of military jobs was actually associated with substantial increases in college going; college enrollments among black men may have been as much as 10% points lower had they served in the military at the same levels observed in the early 1980s. Ó 2009 Elsevier Inc. All rights reserved. 1. Introduction Researchers have portrayed military service as a key source of economic equality and social mobility for disadvantaged groups, such as racial and ethnic minorities (Angrist 1998, p. 82; Bryant et al., 1993; Mare and Winship, 1984; Moskos and Butler, 1996; Seeborg, 1994; Segal et al., 1978). However, increasing returns to education (or increasing penalties for those without higher education) throughout the 1990s and 2000s may imply a shift in the expected returns to military service (Autor et al., 2008). Black men serve in the military at higher rates than their white or Hispanic peers, and some estimates suggest that military service accounts for more than 10% of all employment among young black men (Williams, 1994, p. 35). As a consequence of high levels of service, the expansion and contraction of the volunteer military force likely has substantial impacts on young, male, high school graduates. Specifically, changing rates of military service likely influence participation in two key institutions: the labor market and higher education. This paper investigates how changing rates of military service, driven by the expansion of the armed forces in the early 1980s followed by the loss of roughly 500,000 military jobs over a 5-year drawdown period in the early 1990s, affected participation rates in the labor market and education system, and whether the consequences of military change differed by race. Because pooling the entire adult population may fail to pick up the substantial effect of military employment in key subgroups, the analysis employs aggregate data on rates of military service, stratified by age, education and race/ethnicity to examine the consequences of military downsizing for college enrollment rates and employment–population ratios. While centering on the employment and enrollment consequences of the military drawdown in the early 1990s, this analysis informs contemporary research on college enrollment. Existing research offers several explanations for the pattern of shrinking then rising gaps between black and white enrollment rates: rising parental education, increased college costs, q I am grateful for the insightful comments and suggestions of Bruce Western, Marta Tienda, Jake Rosenfeld, Devah Pager and Paul DiMaggio and the anonymous reviewers. Previous drafts of this research were presented at the Population Association of America annual meetings. * Fax: +1 785 864 5280. E-mail addresses: [email protected], [email protected] 0049-089X/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.ssresearch.2009.09.001 478 M. Kleykamp / Social Science Research 39 (2010) 477–490 changes in the generosity of college benefits, especially Pell grants, and local labor market conditions, including differential returns to college (Black and Sufi, 2002; Kane, 1994; Cameron and Heckman, 2001). Many of these measures capture the direct or opportunity cost of college enrollment; notably absent are measures reflecting the opportunity costs from military service, which may siphon off potential college students. Past research has found substantial associations between aggregate levels of military service and college enrollment rates (Mattila, 1982; Wachter and Wascher, 1984) but recent research has not incorporated the competition from military enlistment into models of college enrollment. As a result, trends in military participation that disproportionately draw in black men as the military expands, and force out black men as it contracts, are likely to influence trends in college enrollments, if military and college are competing alternatives for young adults interested in building human capital, skills and education. If, however, the military is more like ‘‘an employer of last resort”, the displacement effects of military expansion and contraction would likely affect the most marginal workers and lead to observable effects on labor force participation.1 2. Background The recent drawdown and changes in military employment have not garnered much scholarly attention. Research on US military downsizing has focused more on the policy process and on the consequences for military readiness (McCormick, 1998) than on the macro-level effects of downsizing on other institutions. Comparative work has also engaged similar questions outside the US (Segal and Babin, 2000). Related research has examined military downsizing in Russia, with an emphasis on the experiences of downsized Russian army officers and their wives (Rohall et al., 2005). But this work, while highlighting the labor market outcomes of these downsized officers, does not incorporate a key insight in this paper, that the effect of downsizing is not limited to those forced out of an institution, but also extends to those not allowed entry. 2.1. Consequences of military change Sociological analyses of the labor market often emphasize the role of labor demand in explaining poor economic opportunities for minorities, focusing on the decline of manufacturing jobs and the decline in demand for low-skill labor in the central city (cf. Wilson, 1980, 1987, 1996). Others argue more generally that market shifts toward higher technology increased demand for high-skill labor, at the expense of low-skill workers (Bernhardt et al., 2001). Such examples, focusing on influences on employer demand, de-emphasize how changes in the supply (both the size and ‘‘quality”) of those seeking employment and higher education influence the civilian labor market and the higher education system. Returning WWII veterans used GI Bill benefits to expand the level of training and education in the labor force (Angrist, 1993; Bound and Turner, 2002; Stanley, 2003; Turner and Bound, 2003). The downsizing of the US military over a relatively short period of time in the late 1980s and early 1990s provides a similar significant historical event: decreased military manpower needs represented a positive shock to the supply of potential civilian workers and students. Downsizing displaced thousands of young men from military service in the early 1990s, with approximately 500,000 military ‘‘jobs” shed in the first half of the decade. It is unclear how those displaced sorted into college or employment. What fraction of individuals who would otherwise have been members of the military joined the civilian workforce, or enrolled in college? This sorting depends partly on whether downsizing occurred by excluding new recruits who might instead choose to go onto college or enter the open labor market, or by encouraging or even forcing out military personnel. Little data on the exact numbers of individuals in each of these two groups is available. Most policy analysts concur that the majority of those displaced came from the ‘‘newly unqualified group”, since recruiting goals were drastically reduced between 1989 and 1995. But several incentive programs were instituted to encourage voluntary separations, in the hope of avoiding involuntary cuts. A Congressional Budget Office (CBO) report examining the drawdown of the officer corps estimated that only 10–20% of the reduction in the officer corps came from drawdown-induced separations (Congressional Budget Office, 1999). However, other sources suggest figures closer to 35–40% of the overall personnel cuts stemmed from separations (Brasher, 2000). The later figures did not account for the natural rate of separation in the absence of the drawdown; many of those who availed themselves of the separation incentives would still have left the service in their absence, thus the raw figures likely overstate the true ‘‘effect” of the drawdown on separations (Asch and Warner, 2001). Understanding the net effects of the 1990s drawdown on employment or college enrollment at an aggregate level depends on the choices made by those who would have enlisted but would be considered unqualified by the military after the drawdown due to higher entrance requirements after the drawdown—the ‘‘newly unqualified” (and the size of that group) and the opportunities and choices of those who leave the military post-drawdown, the ‘‘newly veteranized” (and the size of that group). A consideration of these factors guides the development of hypothesized outcomes of the most recent military drawdown. 1 Prior research goes far toward dispelling the myth of the military as a last-chance employer. Early work investigating salutary trends in college enrollment at the same time as deleterious trends in employment noted that the military and school ‘‘cream skim” the most productive black workers, leaving behind a less employable population (Mare and Winship, 1984). M. Kleykamp / Social Science Research 39 (2010) 477–490 479 2.2. Choices of the newly unqualified 2.2.1. Employment The drawdown was accomplished by limiting new enlistments, encouraging early voluntary separation, and in limited cases, forcing some individuals to leave military service. But, the bulk of the reduction in force was accomplished by restricting entry (Barley, 1998; Stacey and Anderson, 1992) and in so doing the military was able to tighten enlistment standards and to increase the quality of enlistees. Researchers and policymakers recognized the potential negative effect of the military drawdown. Concern centered on women and minorities who were likely to bear an unequal share of the consequences of such a policy change (Ahituv et al., 1994; Barley, 1998; Richards and Bowen, 1993; Stacey and Anderson, 1992; Williams, 1994) because both groups serve disproportionately in the support occupations at risk of privatization such as medical and clerical specialties (Boesel, 1992). Tightened enlistment standards would disproportionately impact African Americans; black men tend score lower than whites on the Armed Forces Qualification Test (hereafter AFQT) on average and estimates suggested that approximately 40% of the ‘‘newly unqualified” would be black at a time when roughly 20% of new recruits were black (Boesel, 1992). Ahituv et al., (1994, p. 264) warned, ‘‘For blacks, military experience provides a gateway to full-time employment more so than for whites and Hispanics. This fact has potential ominous consequences for minority men’s future employment opportunities given the recent downsizing of the military”. Military downsizing was predicted to have dire consequences for black men’s employment, yet no single study to date has investigated the impact of the drawdown on employment among this group. Earlier experiences with military downsizing after Vietnam led Ellwood and Wise (1987) to examine the effects of reduced military hiring after Vietnam on youth employment. Using aggregate data on employment and military service by state, race and age, from 1972 to 1982, they modeled the exogenous influence of decreased military participation on male civilian employment. They found that for every black youth employed by the military, the civilian employment offset was almost 1, whereas for whites, the offset was less than one and closer to zero. They inferred that if fewer blacks were in the military, employment among black youth would appear substantially worse, while the loss of military employment might actually increase white civilian employment rates. However, the only statistically significant point estimates were from model specifications that did not control for endogeneity between enlistment decisions and employment decisions. Applied to the 1990s drawdown, their results suggest that employment among black men would be expected to decline as a consequence of reductions in military opportunities; military enlistees were likely to be otherwise unemployed. They did not investigate the consequences for school enrollment in their models. However, the post-Vietnam volunteer military imposes non-trivial enlistment eligibility requirements (minimum education levels and AFQT scores in particular) and military recruits are positively selected from the general population (Department of Defense, 2006; Segal and Segal, 2004). Even among the newly unqualified, the vast majority would be high school graduates, albeit with lower AFQT scores than those remaining eligible to enlist. These newly ineligible may elect to search for civilian jobs in lieu of military employment and training, and are likely to be highly employable. Although the newly unqualified may have lower AFQT scores than those accepted for enlistment, they still may have productivity comparable to average entry-level jobseekers. Because military downsizing is expected to have the largest impact on black men, reduction in military participation rates should send proportionally more blacks than whites into the labor force. Prior research suggests black enlistees are more positively selected than are white enlistees (Mare and Winship, 1984; Teachman et al., 1993) and that black veterans should be more attractive labor force candidates than civilian blacks. In contrast to the predictions of some researchers before the drawdown, reduced military opportunity may in fact have increased black employment rates as these selective and highly employable black men enter the civilian labor force. Less prior research informs hypotheses about the consequences of the drawdown for Hispanic men. Because of the under representation of Hispanic men in the military, reductions in military participation rates may show only small if any increases in employment rates. Prior research is less informative of the degree of selectivity of Hispanic enlistees relative to their nonserving peers. They are likely to be positively selected on productivity, given the enlistment requirement of a high school degree and low rates of high school completion in the Hispanic population. Low overall military participation among whites combined with high rates of school enrollment and employment suggest the reduction in military forces will have a small impact on the white military participation rate (hereafter MPR) and a correspondingly low impact on employment among whites. Hypothesis 1. The newly unqualified are not likely to swell the ranks of the unemployed and reduced military service should have between no and a small positive effect on employment. The effect should be largest (and positive) among black men. 2.2.2. College enrollment Many military recruits join the armed forces as a means of building human capital through the acquisition of skills and training, or by earning educational benefits to be used after service. In the absence of military opportunities for education and training, the newly unqualified may opt to enroll instead in college to build human capital. Drawing on job market signaling theory in economics, DeTray (1982) argued military service signals a set of human and social capital attributes that would normally be unobservable by employers among non-veterans. Signaling theory implies that those who may have sought additional credentialing through military service may instead seek educational credentials to signal their productivity. The vast majority of the newly ineligible are expected to be high school graduates who should qualify for enrollment in non-competitive 2- or 4-year institutions. 480 M. Kleykamp / Social Science Research 39 (2010) 477–490 Research indicates that military and college are competing alternatives for high school graduates (Hexter and El-Khawas, 1988; Kilburn and Asch, 2003; Kleykamp, 2006). At the time of the drawdown, Ludwig and Hexter (1992) speculated that the newly unqualified would likely seek employment, lacking financial resources and adequate academic preparation for postsecondary schooling. They estimated that only one-third of the newly disqualified would opt for further education in lieu of military service (it is unclear from the report on what precise basis the estimates are made). But, earlier research on the determinants of college enrollment rates has consistently found a negative relationship between military service and college enrollment rates (Mattila, 1982; Wachter and Wascher, 1984). While the presence of a military draft increases college enrollments (as individuals seeks draft exemptions), higher military service rates drive down college enrollments (at least in the 1960–1980s period covered by existing research). Following from early findings, reductions in military opportunities should increase college enrollment rates among young men. As with employment, the consequences of the drawdown for college enrollment have not been analyzed. According to both human capital theory and signaling theory, reduction in military participation rates among blacks should increase enrollment rates as college-ready blacks enroll in college, turning to other sources of college financing such as loans or working while enrolled. The positive selectivity of blacks enlisting in the military implies many newly unqualified blacks are likely to be college-ready. College enrollment rates among Hispanic men may be more influenced by changes in military opportunity, given lower baseline enrollment rates among Hispanic men as compared with whites or blacks. As with employment, low overall military participation among whites combined with high rates of school enrollment and employment suggest the reduction in military forces will have a small impact on white MPR and a correspondingly low impact on enrollment among whites. Hypothesis 2. The newly unqualified are likely to be college-eligible, thus, in the absence of enlistment opportunities, they are expected to enroll in college, thereby increasing college enrollment rates. The largest effects are expected among black and Hispanic men. 2.3. Choices of downsized veterans Although not a sizeable fraction of those affected by the drawdown, those induced to leave the military because of the drawdown (the newly ‘‘veteranized”) may have gained training and skills at least partly transferable to the civilian sector and should therefore increase employment rates in general, according to human capital theory. In contrast, some research suggests military training and skills are not highly transferable to the civilian labor market, and therefore displaced individuals with non-transferable skills may instead enroll in college to further enhance their productivity (Mangum and Ball, 1987, 1989). If veterans do not enroll in school, they may face extended unemployment as they seek civilian work (Mare et al., 1984). Service members are offered educational benefits to apply towards college post-service such as the Montgomery GI Bill (MGIB). Displaced service members are likely to increase enrollments as they utilize these benefits. Both H1 and H2 are expected to hold whether downsizing happens by constraining enlistment or buy forcing out active duty military members. On balance, the competing expectations about how downsized veterans influence employment suggest this population is not likely to have a substantial influence on civilian employment (some positive, some negative effect). Downsized veterans are expected to increase college enrollments, either because they face limited employment opportunities, or because they can utilize GI Bill benefits which make college affordable. 3. Data and methods Data for analyzing the effects of changing military participation on employment and college enrollment trends come from two sources: the October supplement to the Current Population Survey (CPS) data and from administrative data files on all military members for the years 1980–2000. The military data have a limited set of variables: age, education, race/ethnicity as well as rank and service branch. Using sampling weights in the October CPS and the counts of individuals from the military data, I construct a dataset of rates of military service, employment (employment–population ratios) and college enrollment for specific race, age and education subpopulations over the 1980–2000 period. All rates are computed using the total (military + civilian) population as the denominator, departing from the practice of using only the civilian population in nationally reported statistics. Prior research has demonstrated non-trivial differences between the two approaches in the analysis of employment and wages, especially among the subgroups emphasized in this analysis (Booth and Segal, 2005; Kleykamp, 2007). Following the approach in Western et al. (2006), participation rates are estimated for four age groups: (1) 18–19, (2) 20–24, (3) 25–29 and (4) 30–34; three race/ethnic groups: (1) White, (2) Black, and (3) Hispanic; three education groups: (1) less than HS, (2) HS, GED and some college (3) Bachelor’s degree or higher, over a 21-year period. This scheme produces a dataset with a total of 4 3 3 21 = 756 ‘‘observations” (defined by distinct race–age–education–year categories) of various participation rates. Cells with improbable, non-zero rates of participation (for example, 18–19 year old college graduates) were dropped from the analysis, leaving 693 valid observations for analysis.2 In modeling college enrollment rates, I use 2 These cases were likely coding errors; all cases stem from the military data files. Because these data were a true census of those on active duty, each year had more than a million observations, increasing the likelihood of a few coding errors in the raw data. 481 M. Kleykamp / Social Science Research 39 (2010) 477–490 Trends in Rates of Participation by Race, 1973-2000 Employment College Enrollment 90 10 Military 20 15 % Population Enrolled in College 80 70 % Population Employed 60 6 4 50 10 2 % Active Duty Military 8 25 among HS graduates 1970 1980 1990 2000 1970 Year 1980 1990 Year White Black 2000 1970 1980 1990 2000 date Hispanic Fig. 1. Military participation, employment and college enrollment rates (among HS graduates) males age 17–34, by race. a subset of these observations, considering only college enrollment among high school graduates because including high school dropouts conflates changes in college going with changes in high school completion. Because this period saw a rise in the number of Hispanic immigrants without a high school diploma, unconditional measures of college enrollment are particularly problematic. 3.1. Trends in employment, enrollment and military service Fig. 1 presents trends in participation rates in employment, college enrollment and military service for white, black and Hispanic young men. Black men are more likely to be currently serving in the military than their white or Hispanic peers. In fact, in the 1980s black men were twice as likely as white men to be on active duty, with nearly 8% of all black men 18–34 serving compared with only 4% of young white men and even fewer Hispanics of the same age.3 The dramatic difference in employment rates between black and white men is also evident in Fig. 1. A consistent gap of roughly 15% points remains between the black and white employment rates. A clear disparity between Hispanics and their black and white peers in college enrollment rates is evident beginning in the mid-1980s. Because these trends report enrollment only among high school graduates, this portrait fails to capture the overall low Hispanic enrollment rates which result from the persistent low high school graduation rate of Hispanics compared with whites (Kao and Thompson, 2003). Among whites, involvement in the military fell in the post-Vietnam years and following the Cold war drawdown, but patterns differ for black men. Rates increased for black men in the late 1970s and began to fall in the early 1980s, but dropped especially rapidly with the military downsizing beginning around 1990, halving from 1989 to 2000. Given the simultaneous decline in military participation and increase in college enrollment and employment, additional analysis is needed to reveal whether the decline in military participation may have led to increased school enrollment and employment. 3.2. Analytic strategy To evaluate the association between military participation rate and rates of employment and college enrollment among high school graduates, I use measures of the age–race–education specific employment or college enrollment rates as dependent variables in separate statistical models. These are calculated as the number of employed (or enrolled) individuals divided by the total (civilian plus military) population in a given age–race–education cell, rather than the rates among 3 Hidden in these statistics on military service rates are disparities in eligibility to serve (not addressed with these data), based on such factors as health, criminal record, AFQT score and others. Fewer black men are considered eligible to serve than are their white peers (Asch et al., 2009; Gorman and Thomas 1993). In reality then, military service rates conditioned on service eligibility are even more disparate. Some have suggested that in the 1980s as many as 30–40% of young black men eligible to serve in the military were doing so (Binkin and Eitelberg, 1982). 482 M. Kleykamp / Social Science Research 39 (2010) 477–490 the civilian population as reported in most published statistics. I calculate military service rates as the number of active duty military members divided by the total population of a given cell. Because the dependent variable is a proportion bounded by zero and one, I use the logit-transformed proportion employed or enrolled in college in regression models. Simply regressing the transformed employment or enrollment rates on the military participation rate fails to account for endogeneity between military participation, school enrollment and civilian employment decisions. Naïve estimates associating reduced military participation rates with increased employment rates that ignore the negative association between employment and enlistment rates will be biased downward. Changes in military staffing may influence the civilian labor supply and the civilian labor market substantially influences military recruiting. Enlistments tend to rise when employment prospects for young men are poor, and retention is higher in slack labor markets (Dale and Gilroy, 1984, 1985; Daula and Moffitt, 1995; Daula, 1981). In tight labor markets, military recruiting suffers, as evidenced by the recruiting crisis in the late 1990s (National Research Council, 2003). Like employment, changes in the educational environment, notably changes in the cost of attending college, also affect military recruiting and enlistments. When college costs are high, military service may be a path to funding a college education. If the cost of college declines, or if alternate sources of funding become available, such as with Pell grants then military service and the GI Bill funds earned may be less lucrative (Kilburn and Klerman, 1999). To mitigate endogeneity concerns, analyses employ a 1-year lagged measure of the military participation rate to ensure a temporally prior measure of military participation as a predictor of either college enrollment or employment rates.4 The analysis also estimates three plausible model specifications to account for endogeneity: a baseline specification using prime age unemployment and a linear time trend to model business cycle fluctuations, an instrumental variables specification and a fixed-effects model. Because each of the three models accounts for simultaneity in slightly different ways, yet none taken alone provide ideal controls, the range of results from all models is presented following the approach advocated by Leamer (1983). Consistent substantive results across all three specifications should strengthen confidence in the empirical results. Despite the lagged military participation measure, the baseline model likely suffers endogeneity bias as discussed earlier. If true, the baseline model estimates are expected to be biased toward zero and larger, negative point estimates are expected from the IV or fixed-effects models that more powerfully account for simultaneity between military participation and employment and/or enrollment rates. Exploiting the exogenous change in military force size, I estimate an instrumental variables (IV) specification using the total size of each of the four branches of the military as instruments for the MPR.5 The absolute size of the military and the relative size of each branch are exogenously determined by policymakers responding to national security interests and lobbying pressures.6 Thus, the change in the size of the military can be seen as a kind of natural experiment where exogenous variation in size of the military can identify the effect of changing military participation on employment and school enrollment. For a given year t, age group i and education level j, the proportion employed (or enrolled in school or combining school and work) p, the instrumental variables (IV) model is: logitðptij Þ ¼ a0 þ a1 Mtij þ cij þ etij Mtij ¼ b0 þ b1 St þ dij þ etij where M tij ¼ EðM ij Þ, the predicted military participation rate from the regression purged of the endogeneity to employment, and St is a matrix of the instruments, annual measure of the size of each of the four military branches. The coefficients cij and dij are age–race–education fixed effects. If overall military force size (not military spending) is exogenous, then branch-specific force size should also be exogenous and uncorrelated with rates of employment or enrollment. Tests confirm that in most cases all instruments are exogenous.7 In a few instances the size of the Marine Corps appeared to be endogenous. It is included as an instrument in all models, as estimates from models including or omitting it as an instrument were not substantively different. To test for weak instruments, several authors suggest using the first stage F-statistic (Bound et al., 1995; Staiger and Stock, 1997; Stock et al., 2002). F-Statistics from nearly all models are larger than 10, the general criteria for ‘‘strong” instruments. The third approach employs a fixed-effects specification, presenting full results from the most stringent fixed-effects model as well as the range of estimates from all of the lesser, but included fixed-effects models. The fixed-effects models resolve endogeneity bias in two ways. First, the inclusion of a year fixed-effect captures unmeasured variation that affects 4 After running models with different lag structures such as a two-, 3- or 4-year lagged MPR, I find that the effects of MPR on employment or enrollment are slightly attenuated with higher order lags, but the direction of the point estimates remain negative. Given the lack of substantive guidance for higher order lag terms, I use the single-year lag. 5 Because the drawdown phased the reduction in the size of the military over more than five years, it is not possible to use a simple changepoint model including a pre- or post-drawdown indicator. Additionally, the initial phase of the drawdown was interrupted by the Gulf War in 1991 which used stoploss orders to restrict voluntary separations from the military (McCormick, 1998). 6 The literature on military Keynsianism suggests that military spending expands as a response to a stalled economy (military spending and employment rates are not exogenous). However, Keynesian military spending is not necessarily driven by increases in the size of the military force, and typically results from spending on equipment and private contractors rather than additional active or reserve forces. For this reason I use a measure of active duty force size, not overall military spending as an exogenous instrument. 7 Here the Hansen J over-identification test and the Anderson–Rubin test for instrument identification (whether the instrument is relevant) are used in the statistical software package Stata, version 10/MP. 483 M. Kleykamp / Social Science Research 39 (2010) 477–490 Table 1 Key regression coefficients from OLS, instrumental variables and fixed effects model specifications by race, men aged 18–34. Baseline Employment White Black Hispanic 0.86 (0.94) 1.50 (0.94) 3.32 (2.63) IV 7.06 (2.72)*** 4.70 (0.89)*** 21.90 (4.04)*** College enrollment (among HS graduates) White 4.71 9.87 (2.79)*** (1.92)* Black 12.05 5.95 (1.81)*** (1.49)*** Hispanic 5.61 1.93 (3.56) (4.05) FE FE No year (Range from all models) Including year (Range from all models) Sample size 3.83 (0.83)*** 3.88 (0.60)*** 14.46 (2.28)*** (3.83, 2.02) 2.02 (0.94)* 1.58 (1.04) 3.45 (2.82) (2.02, 0.32) 231 (1.58, 0.83) 231 (6.13, 3.45) 231 (7.62, 8.36) 147 (9.54, 4.69) 147 (9.38, 1.96) 147 8.08 (1.78)*** 7.32 (0.83)*** 0.26 (2.33) (3.88, 3.39) (14.46, 12.46) (8.08, 2.53) (7.32, 5.84) (0.26, 1.34) 7.62 (2.51)** 9.54 (1.85)*** 9.38 (4.38)* Note. Detailed results for each model are available from the author upon request. Robust standard error in parentheses. * p < 0.05. ** p < 0.01. *** p < 0.001. both military participation and employment or enrollment, and second, group fixed-effects capture time-invariant group characteristics related to military participation and employment or enrollment. For example, if young black men are more responsive to labor market conditions than their white peers in making decisions about military enlistment, this would be absorbed by the group-specific fixed effects. This year fixed-effect should reduce bias from the simultaneity of enrollment and employment decisions by absorbing any unmeasured change in either employment conditions or enrollment context; it should also capture secular trends in incarceration, which disproportionately take young black men out of the labor force, school and the military.8 My strategy estimates employment and college enrollment models separately. Including a time fixed-effect in models of employment or enrollment captures the unmeasured influence of changing labor market conditions on decisions to enroll in school, or the influence of changing enrollment context (for example, changing availability and generosity of financial aid) on decisions to work. For a given year t, age group i and education j, the logit-transformed proportion employed (or enrolled in school) p, the fixed-effects model is: logitðptij Þ ¼ b0 þ b1 MPRt1;ij þ cX 0tij þ etij where Xtij is the vector of year specific race–age–education fixed effects. I estimate models separately for black, white and Hispanic men that include measures of the military participation rate (measured as a proportion), a categorical measure of age and a categorical measure of highest degree earned. Because of the grouped nature of the data, a saturated model would include the interaction of education age year and all lesser included terms. The most determined model I estimate is a fixed-effects model that includes the age education interaction and all main effects, as well as a year dummy variable. All models use frequency weights reflecting the actual numbers of people in each age–education–year cell to which the various participation rates apply, given the vastly differing size of the groups reflected by the age–race–education groups. 4. Results Table 1 reports coefficient estimates of the effects of changing military participation rates on employment and college enrollment rates among white, black and Hispanic young men. The table presents the range of estimates for the effect of military participation on employment and college enrollments from the models described above to identify whether the point estimates are in a consistent direction, and of a consistent magnitude (full model estimates for the instrumental variables model and fixed-effects model that includes a time are reported in Appendix Tables A1–A4). The rows in Table 1 report which outcome is modeled (proportion employed or enrolled) while the columns reflect the functional model specification employed (i.e., baseline, instrumental variables (IV), or fixed-effects (FE) specification). All models use a logit-transformed dependent variable. The results for the most determined model are presented with associated standard errors. I also present 8 The dramatic rise in incarceration affects mostly those without a high school diploma, and thus this trend is likely to have more limited influence among high school graduates who would have been at risk of military enlistment. The limitation of the college enrollment model to high school graduates also mitigates any concerns about the role of incarceration in explaining observed patterns. 484 M. Kleykamp / Social Science Research 39 (2010) 477–490 the range of coefficient estimates for military participation rate from less restrictive fixed-effects models (i.e. those only including the main effects). In virtually every specification, a reduction in military participation is associated with increased employment and school enrollment rates. The notable exception is for Hispanics. In the IV model predicting enrollment, the coefficient estimates are positive, but not statistically different from zero. Among the models for black and white men, the magnitude of the effects is sensitive to model specification, as is the direction in the case of the enrollment models for white men. Here, the less determined fixed effects models estimate a positive coefficient for the effect of MPR on enrollment. The IV specification estimates larger effects of MPR on employment but this is not consistently true for the enrollment models. The addition of the year term to the fixed-effects specification reduces the magnitude of the effect of military participation rate on employment and enrollment, except among black and Hispanic men. In this case, inclusion of the year term increases the magnitude of the MPR estimate on college enrollment. Table 1 provides point estimates and indicators of statistical significance, but the use of the logit-transformed dependent variable complicates direct interpretation of the magnitude of the effects. To better understand the magnitude of the effects I generate model predictions under two scenarios. First, observed MPRs are used to predict employment and enrollment under the instrumental variables model and the most stringent fixed-effects model that includes the year term. Next, MPRs are fixed at their pre-drawdown levels in 1980, and used to generate predicted employment and college enrollment rates. Predictions incorporate the full model results; Figs. 2 and 3 present the predictions graphically. Fig. 2 presents race-specific results for the effect of MPR on employment for both the IV and fixed-effects specifications. The IV model estimates on the left reveal little substantive effect of changing MPRs on employment among whites, but more substantial consequences among black and Hispanic men. Comparing the predicted employment rates using either the 1980 MPR (the dashed line) or the observed MPR (the solid line), had the MPR remained at the relatively high 1980 level we could expect lower employment rates among black and Hispanic men by roughly 5% points. The fixed-effects model estimates on the right yield virtually no substantive consequences of MPR on employment across groups once the year-specific variation is accounted for. That the modest effects of MPR change on employment rates observed from the IV model disappear in the fixed-effect model suggests that the instrument may not adequately account for endogeneity. Fig. 3, which presents the race-specific predicted college enrollment rates in the same ways as Fig. 2, reveals more substantial race differences in the impact of changing MPR on college enrollment rates and some differences between the IV and fixed-effects specifications. According to the IV estimates presented on the left, had MPR remained at the 1980 rates (dashed lines), college enrollment among black men would have been approximately 6–7% points lower than those predicted based on the observed MPR (solid lines). Unlike the pattern in the employment models from Fig. 2, the fixed-effects predictions on the right suggest a larger effect of MPR on enrollment with a nearly 10% point difference in the two predictions by 2000. For Predicted Employment Rates 1980-2000 Fixed-effects models 90 80 H 70 H W 60 70 80 % Population Employed W 60 % Population Employed 90 IV models B 50 50 B 1980 1985 1990 1995 2000 Year Observed MPR-White Observed MPR-Black Observed MPR-Hisp. 1980 1985 1990 1995 2000 Year MPR at 1980 level-White MPR at 1980 level-Black MPR at 1980 level-Hisp. Fig. 2. Predicted employment rates for men aged 18–34, by race, from instrumental variables and fixed-effects models. 485 M. Kleykamp / Social Science Research 39 (2010) 477–490 Predicted Enrollment Rates 1980-2000 Fixed-effects models B 20 25 15 H W H 15 20 % Population Enrolled W B 10 10 % Population Enrolled 25 IV models 1980 1985 1990 1995 2000 Year Observed MPR-White Observed MPR-Black Observed MPR-Hisp. 1980 1985 1990 1995 2000 Year MPR at 1980 level-White MPR at 1980 level-Black MPR at 1980 level-Hisp. Fig. 3. Predicted college enrollment rates among HS graduates for men aged 18–34, by race, from instrumental variables and fixed-effects models. example, among black men, the predicted enrollment based on observed MPR was roughly 22%, while that based on the 1980 MPR was approximately 12%. Endogeneity bias is likely to be larger in the employment than enrollment models, consistent with prior studies confirming the strong connection between employment conditions (or unemployment rates specifically) and military enlistment and retention (Dale and Gilroy, 1984, 1985). There is much less research substantiating claims of a relationship between changes in the costs and conditions of schooling and military enlistment and retention. Like the IV predictions, the fixed-effects predictions on the right, show little to no change in enrollment among white men. And in both the IV and fixed-effects specifications, the difference in predicted enrollment rates among Hispanic men is roughly 4% points by the end of the series in 2000. In both the IV and fixed-effects models, the difference of roughly 10% points in predicted enrollment rates among black men swells just after the drawdown begins in the early 1990s. Thus, the timing of loss of military jobs to black men overlaps with the most substantial predicted increase in college enrollment among high school graduates and did not appear to lead to the dire employment consequences predicted by scholars. While Figs. 2 and 3 help present the predicted difference in percentage points (a common metric), a 5% point change translates to roughly a 6–7% change in employment rates (change of 5 from a base of 65–80% employment), but a 22–30% change in enrollment (change of 5 from a base of 17–22% enrollment). For black men, the absolute predicted difference of 10% points, suggests a college enrollment rate lower by 45%. 4.1. Sources of black enrollment From these data it is unclear whether the effect of the drawdown on enrollment among black men stems from individuals choosing college instead of joining the military (the newly unqualified) or whether it is from large numbers of individuals leaving the military and using educational benefits to go on to college (downsized veterans). Prior research on the planning and implementation of the drawdown suggests that the majority of the force reduction was accomplished by limiting the entry of new service members, rather than forcing out large numbers of individuals who desired to stay in the military. Thus, increased enrollments post-drawdown likely came from those now choosing to go immediately into college as opportunities for entry into military service were curtailed. Additional analyses (not reported but available on request) incorporating data on GI Bill usage rates among eligible beneficiaries provides insight into one mechanism that may account for increased school enrollments among black men after the military drawdown, namely that veterans leaving the ranks utilized their MGIB benefits to swell college enrollments. I re-estimated the race-specific enrollment models including a measure of GI Bill usage (these data were limited to the years 1985–2000 because the MGIB program replaced an earlier, less generous education benefit package in 1985). The effect of MPR on overall enrollment among black men is much the same with and without the inclusion of controls for GI Bill 486 M. Kleykamp / Social Science Research 39 (2010) 477–490 utilization and does not appear to be the primary source of the relationship. It appears the ‘‘newly unqualified” are driving the observed association between military participation and college enrollment. Other unreported analyses model 2- and 4-year college enrollment separately, to assess whether the predicted increases may have stemmed from junior or community college enrollments. While increases were predicted in both types of institutions, much larger differences were predicted among 4-year colleges (around 8% points) than 2-year college enrollments (roughly 2% points), and the overall trend increased precipitously around 1990 in 4-year, but not 2-year enrollments among black men. While the expansion of community colleges contributed to the increase in college enrollments, it likely was not the driving force. However, expansion of 4-year programs may have facilitated the enrollments among the newly unqualified, but the data used in this analysis cannot tease out differences in type of 4-year institutions. 5. Discussion The findings of this analysis suggest that trends in military participation are statistically and substantively related to observed trends in employment and college enrollment among young men. In most cases the substantive impact of the military drawdown on employment or college enrollments is moderate to small. Declining military participation appears to have led to a small increase in employment rates in the 1990s, and these results do not vary substantially by race. However, changes in military service did have a considerable impact on college enrollment among black men, precisely the group policy analysts feared would suffer the most from the military drawdown. Their employment rates did not suffer appreciably, and the loss of military jobs was actually associated with substantial increases in college going. Unlike prior research that suggests military service serves as an alternate source of civilian employment for those who cannot find suitable employment in the labor market, these results do not find evidence that the dramatic reduction in military ‘‘jobs” had a substantial correlation with employment in the civilian job market. Given the findings relating reduced military participation with increases in college enrollment among black men, high rates of military service by this group in the 1970s and 1980s may have disrupted educational trajectories. Although many who enlist subsequently go on to college or are enrolled part-time while enlisted, some who intended to go on to college after military service fail to do so upon leaving the military. Previous research on veterans of the post-Korean war era suggests that many individuals who want to go to college after military service failed to do so upon completion of their service (MacLean, 2005). The strong association between reduced military participation and increased enrollments was little changed after accounting for usage of the MGIB, adding further evidence to support the contention that enrollment increases stemmed from individual choosing college when military service was no longer available. If young black men enlist to increase their human capital, or to earn a credential signaling their productivity to employers, when military opportunities declined, they increasingly turned to schools to obtain these credentials. If this is indeed the mechanism at work, then it suggests that military enlistment may have hindered opportunities for some black men. If a non-trivial fraction of men who join the military have high educational expectations but fail to accomplish those goals after completing their service, as happened among earlier cohorts of veterans (MacLean, 2005), then military service would seem to divert them from pursuing higher education. The drawdown took place at a point in time when the returns to a college education were increasing. Had these men served and perhaps delayed or deferred further education, their military experience may not have produced the positive returns earned among earlier cohorts. Further research is needed to investigate how military service alters educational expectations, and how consistent actual post-service educational behaviors are with pre-enlistment educational intentions. The results of this study offer important insights beyond the immediate question about the ramifications of the drawdown. Current interest in racial gaps in college enrollment, especially among men, has not fully explained the pattern of shrinking, then growing black–white gaps in college going. This literature may benefit from a more detailed accounting of the opportunity costs of college enrollment by including indicators of military enlistments and military service opportunities. Given the historical trend of over-representation of blacks in the military and the under representation of black men on college campuses, it should come as no surprise that military service has some displacement effect on college enrollments. Moreover, past research showing how increased participation in military and schooling and the selective nature of those two institutions in the 1960–1980s influenced employment trends (Mare and Winship, 1984) provides precedent for the combined influence of changing participation in selective institutions. My findings suggest declining military participation among black men may have contributed, for a brief period, to reducing black–white gaps in college enrollment although further research is needed to explicitly evaluate this claim. But by the mid 1990s, African Americans no longer expressed disproportionate interest in military service and no longer enlisted at rates substantially higher than whites (Segal and Segal, 2004.) But, as the military expands, it may again compete with college to enlist young men and women. In doing so, military service may displace non-trivial numbers potential college students from enrollment. On the other hand, the strengthening of the GI Bill program in recent months will likely lead to additional enrollments of veterans. Future research on college enrollments should include better accounts of how a changing military contexts shape higher education. Highlighting the military’s role as an institution competing with the labor market and educational system broadens our conceptual framework of labor force and college participation rates. Currently, economic conditions combined with the high 487 M. Kleykamp / Social Science Research 39 (2010) 477–490 costs of college have boosted military enlistment rates. Based on the results from this investigation, we can expect a consequent reduction in college enrollments among young men, likely exacerbating gender differences in college enrollments. Appendix A See Tables A1–A4. Table A1 Coefficient estimates for instrumental variables models predicting logit-transformed employment rates. White MPRt Age 20–24 Age 25–29 Age 30–34 Less than HS BA or higher Constant Observations R-squared * ** *** *** Black Hispanic *** 7.06 (2.22) 0.75*** (0.04) 1.45*** (0.04) 1.69*** (0.06) 0.61*** (0.08) 0.34*** (0.05) 0.70*** (0.12) 4.70 (0.89) 0.90*** (0.05) 1.43*** (0.05) 1.66*** (0.05) 0.99*** (0.07) 0.71*** (0.07) 0.09 (0.08) 21.90*** (4.04) 1.02*** (0.07) 1.30*** (0.08) 1.37*** (0.09) 0.79*** (0.13) 0.06 (0.09) 1.13*** (0.18) 231 0.941 231 0.921 231 0.789 p < 0.05. p < 0.01. p < 0.001. Table A2 Coefficient estimates for instrumental variables models predicting logit-transformed college enrollment rates (among HS graduates only). MPRt Age 20–24 Age 25–29 Age 30–34 BA or higher Constant Observations R-squared White Black Hispanic 9.87*** (2.79) 0.95*** (0.07) 2.67*** (0.05) 3.48*** (0.07) 0.35*** (0.07) 0.74*** (0.14) 5.95*** (1.49) 0.85*** (0.06) 2.29*** (0.07) 3.02*** (0.10) 0.64*** (0.11) 0.21 (0.13) 1.93 (4.05) 0.88*** (0.06) 2.01*** (0.09) 2.61*** (0.12) 0.53*** (0.12) 0.36* (0.18) 147 0.958 147 0.888 147 0.874 Note. Robust standard errors in parentheses. Reference category is Age 18–19, HS/GED/some college. * p < 0.05. ** p < 0.01. *** p < 0.001. 488 M. Kleykamp / Social Science Research 39 (2010) 477–490 Table A3 Coefficient estimates for fixed-effects models predicting logit-transformed employment rates. MPRt1 Age 20–24 Age 25–29 Age 30–34 Less than HS BA or higher Age 20–24 Less than HS Age 20–24 BA or Higher Age 25–29 Less than HS Age 25–29 BA or Higher Age 30–34 Less than HS Age 30–34 BA or Higher Constant Year dummies included Observations R-squared White Black Hispanic 2.02* (0.95) 0.75*** (0.03) 1.63*** (0.03) 1.84*** (0.04) 0.18*** (0.05) 0.28*** (0.03) 0.07 (0.05) 0.07 (0.06) 0.48*** (0.05) 0 0.00 0.65*** (0.05) 0.31*** (0.04) 0.46*** (0.05) 1.58 (1.04) 0.86*** (0.05) 1.53*** (0.05) 1.71*** (0.06) 0.70*** (0.11) 0.69*** (0.10) 0.01 (0.10) 0.02 (0.18) 0.28** (0.10) 0 0.00 0.03 (0.11) 0.34** (0.13) 0.49*** (0.15) 3.45 (2.82) 0.78*** (0.08) 1.32*** (0.10) 1.53*** (0.11) 0.46** (0.16) 0.04 (0.17) 0.44*** (0.12) 0 0.00 0.34* (0.13) 0.16 (0.20) 0.13 (0.14) 0.42* (0.19) 0.46* (0.20) Yes 231 0.986 Yes 231 0.941 Yes 231 0.875 Note. Robust standard errors in parentheses. Reference category is age 18–19, HS/GED/some college. Year fixed effects omitted from table. * p < 0.05. ** p < 0.01. *** p < 0.001. Table A4 Coefficient estimates for fixed-effects models predicting logit-transformed college enrollment rates (among HS graduates only) MPRt1 Age 20–24 Age 25–29 Age 30–34 BA or higher Age 20–24 BA or Higher Age 25–29 BA or Higher Age 30–34 BA or Higher Constant Year dummies included Observations R-squared Note. Robust standard errors in parentheses Reference category is Age 18–19, HS/GED/some college. Year fixed effects are omitted from table. * p < 0.05. ** p < 0.01. *** p < 0.001. White Black Hispanic 7.62** (2.51) 0.88*** (0.07) 2.71*** (0.05) 3.51*** (0.07) 0.42** (0.13) 0 0.00 1.01*** (0.11) 1.00*** (0.13) 0.55*** (0.11) 9.54*** (1.85) 0.72*** (0.09) 2.29*** (0.09) 3.21*** (0.13) 0.29* (0.13) 0.68** (0.20) 0 0.00 0.67*** (0.17) 0.86*** (0.24) 9.38* (4.38) 0.82*** (0.07) 2.19*** (0.10) 2.91*** (0.14) 0.11 (0.26) 0 0.00 0.47 (0.31) 0.74* (0.30) 0.45 (0.31) Yes 147 0.985 Yes 147 0.936 Yes 147 0.896 M. Kleykamp / Social Science Research 39 (2010) 477–490 489 References Ahituv, Avner, Tienda, Marta, Xu, Lixin, Hotz, V. Joseph, 1994. Initial Labor Market Experiences of Minority and Nonminority Men. In: Industrial Relations Research Association 46th Annual Proceedings, pp. 256–265. Angrist, Joshua D., 1993. The effect of veterans benefits on education and earnings. Industrial & Labor Relations Review 46, 637–652. Angrist, Joshua D., 1998. Estimating the labor market impact of voluntary military service using social security data on military applicants. Econometrica 66, 249–288. Asch, Beth J., Buck, Christopher, Klerman, Jacob Alex, Kleykamp, Meredith, Loughran, David S., 2009. Military Enlistment of Hispanic Youth: Obstacles and Opportunities. RAND, Santa Monica, CA. Asch, Beth J., Warner, John T., 2001. An Examination of the Effects of Voluntary Separation Incentives. RAND, Santa Monica, CA. Autor, David H., Katz, Lawrence F., Kearney, Melissa S., 2008. Trends in U.S. wage inequality: revising the revisionists. The Review of Economics and Statistics 90, 300–323. Barley, Stephen R., 1998. Military downsizing and the career prospects of youths. The Annals of the American Academy of Political and Social Science 559, 141–157. Bernhardt, Annette D., Morris, Martina, Handcock, Mark S., Scott, Marc A., 2001. Divergent Paths: Economic Mobility in the New American Labor Market. Russell Sage Foundation, New York. Binkin, Martin, Eitelberg, Mark J., 1982. Blacks and the Military. Brookings Institution, Washington, DC. Black, Sandra E., Amir, Sufi, 2002. Who Goes to College? Differential Enrollment by Race and Family Background. NBER Working Paper No. W9310. http:// www.nber.org/papers/w9310 (retrieved 22.02.09). Boesel, David, 1992. Cutting recruits: a profile of the newly unqualified. In: Stacey, N., Anderson C. L., and United States (Eds.), Military Cutbacks and the Expanding Role of Education. Office of Educational Research and Improvement Office of Research. U.S. Department of Education Office of Educational Research and Improvement Office of Research, Washington, DC, pp. 5–20. Booth, Bradford, Segal, David R, 2005. Bringing the soldiers back in: implications of inclusion of military personnel for labor market research on race, class, and gender. Race, Gender & Class 12, 34–57. Bound, John, Jaeger, David A., Baker, Regina M., 1995. Problems with instrumental variables estimation when the correlation between the instruments and the endogenous explanatory variable is weak. Journal of the American Statistical Association 90, 443–450. Bound, John, Turner, Sarah, 2002. Going to war and going to college: did World War II and the G.I. Bill increase educational attainment for returning veterans? Journal of Labor Economics 20, 784–815. Brasher, Bart, 2000. Implosion: Downsizing the U.S. Military, 1987–2015. Greenwood Press, Westport, CT. Bryant, Richard R., Samaranayake, V.A., Wilhite, Allen, 1993. The effect of military service on the subsequent civilian wage of the Post-Vietnam veteran. The Quarterly Review of Economics and Finance 33, 15–31. Cameron, Stephen V., Heckman, James J., 2001. The dynamics of educational attainment for black, Hispanic, and white males. Journal of Political Economy 109, 455–499. Congressional Budget Office, 1999. The Drawdown of the Military Officer Corps. http://www.cbo.gov/doc.cfm?index=1772&type=0 (retrieved 2.06.09). Dale, Charles, Gilroy, Curtis, 1984. Determinants of enlistments—a macroeconomic time-series view. Armed Forces & Society 10, 192–210. Dale, Charles, Gilroy, Curtis, 1985. Enlistments in the all-volunteer force—note. American Economic Review 75, 547–551. Daula, Thomas, Moffitt, Robert, 1995. Estimating dynamic-models of quit behavior—the case of military reenlistment. Journal of Labor Economics 13, 499– 523. Daula, Thomas V., 1981. The Retention of First-Term Soldiers Theory and Empirical Evidence. Ph.D. dissertation Thesis, Economics, Massachusetts Institute of Technology, Cambridge, MA. Department of Defense, 2006. Population Representation in the Military Services, Fiscal Year 2004. Office of the Assistant Secretary of Defense (Force Management and Personnel), Washington, DC. DeTray, Dennis, 1982. Veteran status as a screening device. American Economic Review 72, 133–142. Ellwood, David T., Wise, David A., 1987. In: Wise, D.A. (Ed.), Military Hiring and Youth Employment. University of Chicago Press, Chicago, pp. 79–118. Gorman, Linda, Thomas, George W., 1993. General intellectual achievement, enlistment intentions, and racial representativeness in the United-States military. Armed Forces & Society 19, 611–624. Hexter, Holly, El-Khawas, Elaine, 1988. Joining Forces: The Military’s Impact on College Enrollments. American Council on Education, Washington, DC. Kane, Thomas.J., 1994. College attendance by blacks since 1970: the role of college cost. Family background and the returns to education. Journal of Political Economy 102, 878–911. Kao, Grace, Thompson, Jennifer S., 2003. Racial and ethnic stratification in educational achievement and attainment. Annual Review of Sociology 29, 417– 442. Kilburn, M. Rebecca, Asch, Beth J., 2003. Recruiting Youth in the College Market: Current Practices and Future Policy Options. RAND, Santa Monica, CA. Kilburn, M. Rebecca, Klerman, Jacob Alex, 1999. Enlistment Decisions in the 1990s: Evidence from Individual-Level Data. RAND, Santa Monica, CA. Kleykamp, Meredith, 2006. College, jobs or the military? Enlistment during a time of war. Social Science Quarterly 87, 272–290. Kleykamp, Meredith, 2007. Military service as a labor market outcome. Race, Gender & Class 14, 65–76. Leamer, Edward E., 1983. Let’s take the con out of econometrics. American Economic Review 73, 31–43. Ludwig, Meredith, Hexter, Holly, 1992. The impact of military drawdowns on student assistance programs. In: Stacey, N., Anderson, C.L. (Eds.), Military Cutbacks and the Expanding Role of Education. U.S. Department of Education Office of Educational Research and Improvement Office of Research, Washington, DC, pp. 5–20. MacLean, Alair, 2005. Lessons from the cold war: military service and college education. Sociology of Education 78, 250–266. Mangum, Stephen L., Ball, David E., 1987. Military skill training—some evidence of transferability. Armed Forces & Society 13, 425–441. Mangum, Stephen L., Ball, David E., 1989. The transferability of military-provided occupational training in the post-draft era. Industrial and Labor Relations Review 42, 230–245. Mare, Robert D., Winship, Christopher, 1984. The paradox of lessening racial inequality and joblessness among black youth: enrollment, enlistment, and employment, 1964–1981. American Sociological Review 49, 39–55. Mare, Robert D., Winship, Christopher, Kubitschek, Warren N., 1984. The transition from youth to adult: understanding the age pattern of employment. American Journal of Sociology 89, 326–358. Mattila, J. Peter, 1982. Determinants of male school enrollments: a time-series analysis. Review of Economics and Statistics 64, 242–251. McCormick, David, 1998. The Downsized Warrior: America’s Army in Transition. New York University Press, New York. Moskos, Charles C., Butler, John S., 1996. All That We Can Be: Black Leadership and Racial Integration the Army Way. Basic Books, New York. National Research Council, 2003. In: Sackett, P.R., Mavor, A.S. (Eds.), Attitudes, Aptitudes, and Aspirations of American Youth: Implications for Military Recruiting, National Academies Press, Washington, DC. Richards, Kara B., Bowen, Gary L., 1993. Military downsizing and its potential impacts for Hispanic, black, and white soldiers. The Journal of Primary Prevention 14, 73–92. Rohall, David E., Hamilton, V. Lee, Segal, David R., Jessica, Y.Y. Kwong, 2005. Job-search strategies in time and place: a study of post-service employment among former Russian army officers. In: Klaus Warner, Schaie, Elder, Glen H. (Eds.), Historical Influences on Lives and Aging. Springer Publishing Co., New York (Chapter 4). Seeborg, Michael C., 1994. Race, poverty and enlistment: some evidence from the national longitudinal survey of youth. Journal of Economics (MVEA) 20, 15–24. 490 M. Kleykamp / Social Science Research 39 (2010) 477–490 Segal, David R., Babin, Nehama, 2000. Institutional change in armed forces at the dawning of the 21st century. In: Quah, Stella R., Sales, Arnaud (Eds.), The International Handbook of Sociology. Sage, London, pp. 218–235. Segal, David R., Bachman, Jerald G., Dowdell, Faye, 1978. Military service for female and black youth—perceived mobility opportunity. Youth & Society 10, 127–134. Segal, David R., Segal, Mady Wechsler, 2004. America’s military population. Population Bulletin 59, 3–40. Stacey, Nevzer, Anderson, Clinton Lee, 1992. Military Cutbacks and the Expanding Role of Education. U.S. Dept. of Education Office of Educational Research and Improvement Office of Research, Washington, DC. pp. iii, 212. Staiger, Douglas, Stock, James H., 1997. Instrumental variables regression with weak instruments. Econometrica 65, 557–586. Stanley, Marcus, 2003. College education and the midcentury GI Bills. Quarterly Journal of Economics 118, 671–708. Stock, James H., Wright, Jonathan H., Yogo, Motohiro, 2002. A survey of weak instruments and weak identification in generalized method of moments. Journal of Business & Economic Statistics 20, 518–529. Teachman, Jay D., Call, Vaughn R.A., Segal, Mady W., 1993. The selectivity of military enlistment. Journal of Political & Military Sociology 21, 287–309. Turner, Sarah, Bound, John, 2003. Closing the gap or widening the divide: the effects’ of the GI Bill and World War II on the educational outcomes of black Americans. Journal of Economic History 63, 145–177. Wachter, Michael L., Wascher, William L., 1984. Leveling the peaks and troughs in the demographic cycle: an application to school enrollment rates. Review of Economics and Statistics 66, 208–215. Western, Bruce, Kleykamp, Meredith, Rosenfeld, Jake, 2006. Did Falling Wages and Employment Increase U.S. Imprisonment? Social Forces 84, 2291–2312. Williams, Roger C., 1994. An estimate of black gross job losses due to reduced defense expenditures. The Review of Black Political Economy 22, 31–41. Wilson, William J., 1980. The Declining Significance of Race: Blacks and Changing American Institutions. University of Chicago Press, Chicago. Wilson, William J., 1987. The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy. University of Chicago Press, Chicago. Wilson, William J., 1996. When Work Disappears: The World of the New Urban Poor. Alfred A. Knopf: Distributed by Random House Inc., New York, NY.
© Copyright 2026 Paperzz