Institutional Castling: Military Enlistment and Mass Incarceration in the United States.

The military is a major state provider of employment, occupational training, and educational subsidies. Yet military downsizing and its increased selectivity during penal expansion may have cleaved off employment opportunities for disadvantaged men. We show how institutional castling-the shifting prominence of competing institutions in the lives of specific demographic groups-has affected the underlying risk of military employment and penal confinement. Black veterans who have dropped out of high school are less likely to be incarcerated than their nonveteran counterparts, and declines in the employment rates of military servicemembers with less than a high school education are associated with large increases in incarceration rates. The military's critical role in providing institutional protection from the penal system has eroded for young, undereducated African American men.

The United States has seen tremendous growth in incarceration since 1970. Although crime rates are at historic lows, incarceration rates remain at unprecedented highs (Travis, Western, and Redburn 2014). Mass incarceration—rates of imprisonment significantly above historical and societal levels that lead to the systematic incapacitation of particular sociodemographic groups within a society (Garland 2001a)—has been fueled by a set of social policies that disproportionately affect young, undereducated, nonwhite men. Changes in mandatory minimum sentencing laws, increased prosecutorial discretion, and more severe criminal sanctions for low-level drug offenses have contributed to the expansion of the criminal justice system, creating staggering race and class inequality in incarceration (Western 2006; Alexander 2010; Pettit 2012). On any given day, more than one-third of young black men who dropped out of high school are behind bars and face a lifetime risk of spending time in prison or jail at close to 70 percent (Pettit and Western 2004; Pettit, Sykes, and Western 2009; Pettit 2012; Sykes and Pettit 2014).

At the same time, the U.S. military changed profoundly in the size and composition of its personnel. Beginning in 1973, for example, the American armed forces transitioned from a staffing policy based on a selective service draft to one that positioned the military as a competitor for volunteers within the civilian labor market (Fredland et al. 1996). For the first time in the nation’s history, black men came to be overrepresented in uniform relative to their concentration in the civilian population (Fernandez 1996; Oi 1996; Segal and Segal 2004), and military employment emerged as a major player in the labor market for young men of color. By 1979, one in four working black men age sixteen to twenty-four were employed by the armed forces, as were one in six of their Latino counterparts (Grissmer 1992, 37).

Yet, conventional wisdom holds that the armed forces and the criminal justice system are two distinct arms of the state performing separate and isolated functions. This perception persists despite evidence linking military participation with subsequent criminal activity (Culp et al. 2013; Tsai et al. 2013) and the potential loss of turning points from crime and social disadvantage (Elder 1986; Sampson and Laub 1993; 1996). Importantly, the armed forces have played a key role in employing large percentages of moderately skilled men from disadvantaged backgrounds (Grissmer 1992; Segal and Segal 2004) and, until 2015, were the largest employer in the United States (DMDC 2015; Lundquist, Pager, and Strader 2018). We believe that these institutions are directly connected by virtue of their relationship to the labor market experiences of young black men because changes in the processes that govern selection into one of these institutions are likely to reverberate across other institutions (Han 2018; Mare and Winship 1984).

This article examines the historical relationship between penal confinement and military employment. The degree to which these institutions intersect and the way that rapid and simultaneous policy shifts during the 1960s and 1970s have interacted to affect the risk of incarceration among particular sociodemographic groups have garnered limited scholarly attention. This neglect persists despite the fact that the state establishes admissions criteria for both institutions, and the carceral system and the armed forces have been separately identified as deeply intertwined with the labor market for young men of color (Pettit and Western 2004; Western and Pettit 2005). We join Pierre Bourdieu in invoking “hyperbolic doubt” (1994, 1)—a rethinking of state institutional functions that appear distinct from one another, as well as the synergistic ways in which those institutions may operate in concert to influence and obscure patterns of social inequality. The correctional system and the military are institutional expressions of the state’s coercive power and symbolic violence (Bourdieu 1994, 4–5), representing the polarities of the internal (penal system) and external (military) application of state force. To comprehend the state, we must understand the interrelated functioning of its bureaucratic institutions, as well as the synchronistic consequences of its actions.

We draw on a wealth of data from various sources to examine how the demographic composition of these social institutions have changed in the wake of policies enacted during the mid- to late twentieth century. We contribute to the literature on social inequality and to the body of work on the consequences of military service by assessing the changing risks of incarceration among veterans. Our reasoning is that, as a larger share of men with minimal formal education were excluded from the armed forces, the incarceration patterns among veterans and nonveterans should diverge, particularly among young, black high school dropouts. Specifically, we investigate whether and to what extent changes in the racial, educational, and population distributions of active duty military personnel are associated with racial and educational inequality in incarceration. Using standardization and decomposition techniques (Kitagwa 1955; Preston, Heuveline, and Guillot 2001) and other statistical methods, our analysis quantifies levels of incarceration associated with shifts in the race and class distribution of military enlistment across generations. In doing so, we demonstrate the demographic consequences of institutional castling—the shifting prominence of state institutions in the lives of particular demographic groups following changes in social policies or judicial decisions, allowing more prominent, protective, and integrative institutions to exchange their risk of exposure with less active and more punitive agencies across generations.

LABOR MARKET STRUCTURE, THE MILITARY, AND PENAL SYSTEM EXPOSURE

Much has been written about the connection between crime, incarceration, and labor-force participation. Marxian approaches understand the prison system as a means of managing surplus labor (Spitzer 1975), as the connection between labor market structure, crime, and the criminal justice system can be complex and multilayered (Cantor and Land 1985; Bellair and Roscigno 2000; Phillips and Land 2012; Wadsworth 2004). Research suggests that the structure of the labor market (that is, the relative concentration of primary and secondary sector jobs), and not the overall supply of jobs, drives rates of imprisonment (Sutton 2004). Similarly, scholarship is mixed on whether crime rates are more responsive to the quantity of jobs (relative to the population) or to the quality of jobs available (Doeringer and Piore 1971; Crutchfield 1989, 2014). The labor stratification perspective that Robert Crutchfield (1989, 2014) advances finds that communities plagued by higher levels of crime experience labor market instability and possess a concentration of secondary sector jobs. This relationship may be particularly salient for young adult men, who make up roughly 90 percent of new military accessions (Krivo and Peterson 2004; Office of the Under Secretary of Defense, Personnel and Readiness 2017).

Additionally, workers who have jobs with characteristics that mark them as being in the primary sector, such as anticipating that their current job will last for a longer period of time (Crutchfield and Pitchford 1997) or reporting receiving benefits and emotional rewards from their employment (Wadsworth 2006), are less likely to engage in criminal activity.[1] Work in the military and in the civilian sector of the labor market are known to operate as turning points from crime, delinquency, and social disadvantage (Shattuck 1945; Mattick 1960; Dressler 1946; Elder 1986, 1999; Sampson and Laub 1993; Uggen 2000), particularly for those who secure high-quality jobs that provide them with a level of professional satisfaction (Uggen 1999; Van der Geest, Bijleveld, and Blokland 2011). The deterrent effect of work may also operate via an emotional dimension because formerly incarcerated men who are employed report less criminal activity when they feel committed to their jobs, regardless of job characteristics (Apel and Horney 2017).

A more limited body of research investigates the link between service in the armed forces and subsequent criminality. Although veterans have lower criminal justice involvement overall relative to nonveterans (Teachman and Tedrow 2016), a history of military employment does appear to increase violent offending (Bouffard 2005; Crutchfield and Pitchford 1997; Wadsworth 2006). However, the circumstances of each tour of duty, including the era of service and the selectivity of enlistment, shape that association (Bouffard 2014; Culp et al. 2013), given that mental health problems and substance abuse could prevail after discharge (Erickson et al. 2008; Tanielian and Jaycox 2008; Tsai et al. 2013).

Despite the benefits and costs associated with military employment, disadvantaged men who aspired to enlist saw their enlistment opportunities diminished twice. The first time was in 1980 and 1981, when the previously mis-normed (that is, inaccurately calibrated) Armed Forces Qualifying Test (AFQT)—the standardized entrance examination used to both establish eligibility for enlistment and to assign new service personnel to a military occupational specialty—was replaced (U.S. Congress 1989). Wide racial disparities emerged in AFQT scores (Kilburn, Hanser, and Klerman 1998), and large percentages of black men who would have previously qualified for the armed forces no longer met the minimum requirements (Angrist 1993). In the wake of this error, the Department of Defense struggled to find what it termed “highly qualified” applicants—those possessing both a diploma (rather than a GED or no certification) and an AFQT score above the median (Office of the Under Secretary of Personnel and Readiness 2000). By 1985, for example, only in the Air Force did a majority of enlistees clear these criteria (Office of the Undersecretary of Defense, Personnel and Readiness 2017, table D-9).

The second diminished opportunity for military employment occurred a decade later, when concerns over the capabilities of those in uniform converged with moves toward privatization, resulting in a “drawdown” of military personnel in the early 1990s (Lytell et al. 2015). The armed forces increased enlistment criteria in anticipation of military downsizing, virtually eliminating opportunities for enlistment among young men who lacked either a high school diploma (or GED) and a standardized test score in the middle range (Boesel 1992). The number of new accessions shrank rapidly, from 313,777 in 1986 to 205,501 in 1991, and down further to 160,511 by 2010 (Office of the Undersecretary of Defense, Personnel and Readiness 2017, table D-8). The total number of active duty enlisted personnel declined by roughly seven hundred thousand during the same period (Office of the Undersecretary of Defense, Personnel and Readiness 2017, table D-11). Joshua Angrist (1993, 2) finds that changes in minimum enlistment criteria in the early decades of the All-Volunteer Force “had an adverse effect on service opportunities for minority applicants.” We suggest that the shrinking availability of primary sector employment offered by the U.S. military and the changes in their enlistment criteria may have altered the demographic composition of both the armed forces and the penal system via a set of social policy changes during the mid- to late twentieth century.

INSTITUTIONAL CASTLING: RACE, LABOR STRATIFICATION, AND THE MILITARY AND PENAL SYSTEMS

We explore the demographic consequences of what we term institutional castling—the shifting prominence of state institutions in the lives of particular demographic groups following a change in social policy or judicial decisions, allowing more prominent, protective, and integrative institutions to exchange their risk of exposure with less active and more punitive agencies across generations.[2] Institutional castling may be a result of either disparate treatment (intentionality), disparate impact (unintentional consequences), or unconscious bias following the enactment of a particular social policy or judicial decision. State priorities may be revealed through legislation and budget allocations; the joint effect of new and large federal resource allocations and the creation of new social policies that upend and reconstitute the hierarchal stratification of institutions in the lives of particular demographic groups facilitates the castling of government agencies tasked with both integrating the American underclass into normative-based conceptions of society and punishing the particular types of behaviors associated with poverty and inequality. There have been two moments in American history when the military and penal systems have castled: the Civil War and, in analyses that follow, the mid- to late twentieth century.

Institutional Castling in U.S. History

A longer read of historical labor stratification points to an unassailable truth: prior to deindustrialization, penal growth, a drawdown in the size of the armed forces, and an increase in the selectivity of military enlistees during the close of the twentieth century, the first instance of institutional castling (between the military and carceral systems) took place during the era of slavery.[3] In Black Reconstruction in America, W.E.B. DuBois (1998) highlights the ways in which shifting Northern interpretations of two key policies led to the institutional castling of the military and criminal justice system for slaves and freedmen, given that fugitives from slavery and free Negroes were precluded from joining the Union Army. First, the Fugitive Slave Act of 1850 decreed that people who escaped slavery had to be returned to their masters, and the Dred Scott v. Sandford decision of 1857 ruled that the U.S. Constitution did not apply to blacks, whether enslaved or free.[4] Despite these laws and judicial decisions, Union Army generals routinely employed fugitive and freed slaves as “soldiers, spies, servants, military laborers, and laborers on plantations” (DuBois 1998, 66). Although initially supportive of excluding fugitive and enslaved men from the Union Army, President Lincoln “faced the truth front forward … the Negro was to be allowed to fight or the draft itself would not bring enough white men into the army to keep up the war” (82). Thus, the initial disparate treatment of slaves and free men under the Fugitive Slave Act and the Dred Scott decision were set aside to ensure that the Union Army had sufficient soldiers and resources throughout the war, providing the first instance of institutional castling where “fugitives became organized and formed a great labor force for the army” (65).

Second, the military policy to integrate fugitive and freed slaves into the Union Army also produced a set of disparate impacts that segregated formerly enslaved blacks and their families from white Union servicemembers, despite fighting alongside each other. For instance, Du-Bois (1998, 70) writes that “There were new and strange problems of the social contract. The white soldiers, for the most part, were opposed to serving Negroes in any manner, and were even unwilling to guard the camps where they were segregated or protect them against violence.” The disparate impact of integrating freed slaves into the military but holding them at arm’s length, facilitated a de facto system of racial segregation that would engender the disbandment of blacks and freed slaves from specific units of the armed forces after the Civil War ended, recastling the institutions of the military and penal system during Reconstruction in Southern states where “black codes” emerged to control the labor and criminalize the behavior of former slaves and African Americans (Roediger 2014).

Institutional Castling During the Mid- to Late Twentieth Century

We argue that the mid- to late twentieth century was another turning point in the American lifecycle of institutional castling between the military and penal spheres for young, undereducated black men. Race-neutral social policies enacted by Congress and state agencies since the late 1960s may have shifted institutional castling from being predicated on (mostly) disparate treatment during the Civil War to new forms of normalization that are rooted in disparate impacts and unconscious biases.

On February 27, 1877, Congress passed an act whereby “no minor under the age of sixteen years, no insane or intoxicated person, no deserter from the service, and no person who has been convicted of a felony shall be enlisted or mustered into the military service.”[5] This law remained unchanged until 1968, when Congress amended and expanded it to specify, in 10 USC§504(a), that “No person who is insane, intoxicated, or a deserter from an armed force, or who has been convicted of a felony, may be enlisted in any armed force. However, the Secretary [of Defense] concerned may authorize exceptions, in meritorious cases, for the enlistment of deserters and persons convicted of felonies.”[6] The amendment is important because it carves out exceptions in the U.S. Code allowing for the possibility of felony and conduct waivers during the enlistment process. Between 1948 and 1968, the percentage of non–African Americans with a felony record had nearly doubled (from 1 to 2 percent); among African Americans, the percentage grew from 5 to 8 percent over the same period, with men bearing the brunt of this institutional growth (Shannon et al. 2017). Yet contemporary research shows that the military personnel granted these felony waivers do not reflect the racial and educational distribution of people with a felony record (Lundquist, Pager, and Strader 2018; Travis, Western, and Redburn 2014).

The same year that the 1877 law was amended to create 10 USC§504(a), to carve out provisions for felony and conduct waivers that would subsequently disproportionately favor whites and those with a higher education in enlistment decisions, Congress passed the Omnibus Crime Control and Safe Streets Act of 1968,[7] which established the Law Enforcement Assistance Administration (LEAA), now known as the Office of Justice Programs, which allocated $100 million in block grants to states, with half of the money going to local law enforcement (Office of Justice Programs 1996). The LEAA also provided resources to develop alternative sanctions for the punishment of young offenders and to deal with rioting and organized crime (Office of Justice Programs 1996). The Safe Streets Act also expanded the use of surveillance at the local level and provided for closer training between local law enforcement and the FBI (Office of Justice Programs 1996).

The coterminous creation of 10 USC§504(a), aimed at excluding people with felony records from the military unless waived, and the enactment of crime control legislation (and its authorized funding apparatus) were not accidental; these decisions represent intentional actions to reconstitute the functioning of particular American institutions.[8] Jonathan Simon (2007, 75) observes that “it is not just the scope of this wave of lawmaking [starting in the late 1960s] that makes it impressive, it is also the coherence of this body of law as reflecting a vision of how institutions govern through crime.” Similarly, Elizabeth Hinton (2016, 340) notes that “Questions of intent, or the degree to which federal policymakers foresaw the consequences of the choices they made with respect to urban social programs in black communities, are only relevant to a certain extent. The issue is to uncover the series of decisions that made contemporary mass incarceration possible in order to discover our own actual history.” Thus the functioning of state institutions since the late 1960s and their demographic reconstitutions must be understood through the refracted prism of governing through crime and social control following the civil rights movement (Garland 2001b; Simon 2007).

It is our contention that race-neutral policies such as 10 USC§504(a) may have had a disparate impact on the ability of African American men to enlist in the military—a trend that was exacerbated by other policy changes that increased the educational and testing requirements for enlistment during military downsizing. Despite moments in American history when felons have been allowed to enlist, particularly during World Wars I and II, the men who served underwent extensive assessments and reviews by parole boards, Selective Service personnel, and psychologists, and the military considered their enlistment an experiment (Shattuck 1945; Mattick 1960; Dressler 1946). Even now, felons may obtain conduct and felony waivers for military employment; however, research shows that these waivers are selectively administered across military branches and vary inversely with civilian labor market trends, disproportionately favoring whites, men, and individuals with a high school degree or more, much to the exclusion of blacks and high school dropouts (see Lundquist, Pager, and Strader 2018).

These changes may have jointly cleaved off potential turning points in the life course of young, undereducated black men, allowing the military and penal institutions to castle in ways that would elevate the risk of criminal justice contact for African American men during labor market restructuring. The combination of increasing requirements for eligibility and the exclusion of those with prior criminal records intersect to deny large segments of the young, black male population access to the protective or corrective benefits of military employment. These federal codes and legislation were intentionally amended, enacted, and funded, regardless of whether the resulting biases were unconscious, unintended, or deliberate. Local, state, and federal governments began to govern through crime in ways that may have portended the castling of major institutions in the lives of specific disadvantaged men. The generational shift in these modalities of governance, in relation to the military and penal institutions, may mean that the same demographic group will have radically different experiences with these social institutions over time. Our conceptual framework and empirical inquiry seeks to examine whether the military is protected from demographic groups with the highest risks of incarceration, given their lower rates of felony waiver issuance (Lundquist, Pager, and Strader 2018).

To underscore the plausibility of our inquiry into the changing composition of the civilian and military labor markets, figure 1 displays employment trends since the early 1970s for young, non-Hispanic white and black men who did and did not graduate from high school. As panel A illustrates, the percentage of young men on active duty has declined since the 1970s, particularly among black men without a secondary education. Blacks with less than a high school education saw their relative employment presence in the armed forces eliminated by 1986, but white men who dropped out of high school maintained a small presence. At the same time, civilian labor-force statistics, presented in panel B, also underscore the growing disadvantage of young, African American men with less than a high school education. At the beginning of the series, educational and racial parity in employment rates was higher. Yet, deindustrialization (Wilson 1987) and penal expansion (Western and Pettit 2005; Western 2002, 2006; Pettit 2012) led to sharply declining employment rates among black men with little formal education.

Figure 1.

Trends in Employment Status for U.S. Men Age Twenty to Thirty-Four, by Race and Educational Attainment, 1971–2018

Source: Authors’ calculations based on the Annual Social and Economic Supplement of the Current Population Survey.

Panels C and D of figure 1 convey complementary stories about labor inactivity for young, undereducated men. Panel C shows trends in the unemployment rate by race and education level. Young black men who have dropped out of high school have the highest unemployment rates, whereas white men with a high school diploma have the lowest unemployment rates. African American men with a secondary education and white men who have dropped out of high school tend to have similar unemployment rates.

Panel D displays the percentage of each demographic group that has given up its search for work. Black men who did not complete high school have the highest rates of removal from the labor force, reaching almost 50 percent by the close of the first decade of millennium, a rate double that of white high school dropouts. Again, black high school graduates and white dropouts generally have similar levels of employment hopelessness, though the trends vary in the last two decades of our time series.

Historically, men without a high school diploma had more military enlistment and labor market opportunities than they do today (Carnevale, Smith, and Strohl 2010). The mass unemployment and joblessness of undereducated men may have differentially exposed them to the criminal justice system during penal expansion and military downsizing. We therefore begin our investigation into contemporary institutional castling by documenting veteran and nonveteran incarceration rates over time by race and educational attainment. Our analyses focus on a “downstream” measure—changes in the incarceration rates of veterans and nonveterans over time. Next, we decompose the change in the incarceration rate that is due to changes in veteran educational attainment over time, allowing for a closer inspection into how military enlistment policies may have severed a conduit to upward mobility for severely disadvantaged men. Then, we examine how decades of penal expansion are associated with compositional changes (in both size and educational quality) in military employment across generations. Finally, we illustrate how contemporary institutional castling between the military and penal institutions severely disadvantaged young, undereducated black men.

DATA AND CODING

To understand how the protective role of military service castled with the carceral institutions of punishment for young, African American men, we begin by compiling a unique dataset to analyze immersion in the military and penal systems. We use twelve waves of data from twin sources collected by the U.S. Census Bureau and distributed by the Bureau of Justice Statistics. The Survey of Inmates in Local Jails series has periodically collected nationally representative data on individuals held in local correctional facilities across the span of four decades. Respondents include those being held pretrial, those serving local sentences, and those awaiting transfer into the custody of another correctional facility. The survey was fielded in 1972, 1978, 1983, 1989, 1996, and 2002. Although prior research on veteran incarceration has excluded inmates in local custody when investigating differences between veterans and nonveterans (Culp et al. 2013; Greenberg and Rosenheck 2007), including data on inmates in local jails has increased in importance, as extended periods of pretrial detention, prison overcrowding, and other policy shifts (such as AB109 in California) increasingly mean that inmates will spend at least a fraction of their sentenced time in local and county jails (Turney and Connor 2019).

We also include the Survey of Inmates in State and Federal Correctional Facilities. In 1974, 1979, and 1986, this survey was conducted only with individuals at state correctional facilities. In 1991, it was administered to inmates in both state and federal prisons. In 1997 and again in 2004, the state and federal surveys were combined. These surveys were designed to be nationally representative of the inmate population in state and federal facilities. Although several studies have used data from the Survey of Inmates in State and Federal Facilities since the mid-1980s (Culp et al. 2013; Greenberg and Rosenheck 2012), no study systematically uses all inmate data since 1972 to present national estimates of veteran incarceration rates by race and education. However, Sanjiv Gupta and Jennifer Lundquist (2012) use aggregate data to show how the fraction of service personnel and inmates has stabilized since the mid-1990s. In total, the fifteen facility-year inmate series covers 121,554 respondents. Each survey wave contains information on whether the respondent has ever been or is currently employed by the armed forces.

To allow a comparison between the incarcerated and noninstitutionalized populations, we also use the March Current Population Survey (also known as the March CPS) for the years 1972 through 2012, inclusive. It is collected by the Census Bureau and the Bureau of Labor Statistics, surveys approximately fifty to sixty thousand noninstitutionalized respondents in settled households, and collects data on a variety of socioeconomic and demographic characteristics, including race and veteran status. In all, our analysis draws on almost 6.9 million records over the forty-one years in this study. All analyses use sample weights for national representation.

METHODS

The rate at which veterans are siphoned into the criminal justice system is critical to understanding the changing sociodemographic distribution of each institution over time, and how the military’s capacity to protect at-risk men from criminal justice contact may have shifted after periods of increased enlistment criteria. We follow methods outlined in previous studies to generate race and class estimates of incarceration (Pettit and Western 2004; Western 2006; Pettit, Sykes, and Western 2009; Pettit 2012; Ewert, Sykes, and Pettit 2014; Sykes and Pettit 2014). We begin by constructing annual estimates of grouped incarceration rates for each year beginning in 1972. We calculate the cross-classified proportion of inmates in each demographic group—race, sex, age, educational level, and veteran status—within facility type, and use survey weights to obtain nationally representative estimates. We then linearly interpolate weighted group means between survey years and through 2012 to obtain a complete time series. Finally, we apply the weighted proportion for each group to annual correctional counts within facility type, as reported by the Bureau of Justice Statistics, and sum the total number of inmates within each group across different facility types (jails and prisons).

Data from the Current Population Survey and the Survey of Inmates have been analyzed systematically to ensure that our estimates of veteran and nonveteran incarceration rates do not reflect the sample selectivity associated with household-based surveys. The weighted counts derived from the Survey of Inmates and the Bureau of Justice Statistics are used in the numerator of the incarceration rate. To avoid sample selection bias reported in Pettit (2012) and other studies, denominators for the incarceration rate are obtained by combining noninstitutionalized population counts (from CPS) and inmate counts (used in the numerator).

We use a standard demographic technique to understand racial and educational inequality in military enlistment and incarceration. Evelyn Kitagawa’s (1955) standardization and decomposition method allows us to examine how the changing size of the veteran population and the increased educational selectivity for enlistment converge to explain differences in the veteran and nonveteran incarceration rates for particular demographic groups. Essentially, the decomposition quantifies how much of the difference in incarceration rates, between veterans and nonveterans, is explained by the compositional differences in the (racial and educational) distribution of veterans and nonveterans over time, net of the underlying population incarceration rate across demographic groups.

CPS population counts (c) and the incarceration rate (M) for veterans (v) and nonveterans (nv) during year (t) can be decomposed into two parts: the contribution of compositional differences in the veteran population (the first term) and the contribution of differences in the incarceration rate (the second term), as presented in equation (1).

The first term (before the plus sign) measures the relative difference in the composition of the veteran and nonveteran population, weighted by the average incarceration rate. The second term (after the plus sign) captures the difference in the incarceration rate schedule (when the two groups are incarcerated), weighted by the average population size of veterans and nonveterans. In the decomposition, we hold constant age and examine men age twenty through thirty-four because the turning points literature focuses on young men during ages at risk of military employment, criminal delinquency, and incarceration (Elder 1986; Sampson and Laub 1993, 1996).

The decomposition captures the average cohort effect in compositional and rate differences across periods for a fixed age group (such as men age twenty through thirty-four). We perform race-specific decompositions between veterans and nonveterans to uncover the patterns of educational inequality in enlistment and incarceration for white and black males. We apply this technique twice across each level of educational attainment (less than high school and high school graduates who have not completed any college coursework).

Second, we explore how demographic change in military employment across generations is associated with penal expansion using ordinary least squares regression analysis. We construct a panel dataset—using aggregated individual observations—and estimate a model of how logged adjusted incarceration rates (which includes inmates into the population denominator of the rate) for r race (white or black), e education group (dropout or high school graduate), and v veteran status (veteran or nonveteran) during year t are associated with a vector of X demographic variables, as displayed in equation (2).

We also include a vector of lagged military employment rates from two decades before (D) to estimate how the compositional shift to a smaller, more educated military affected incarceration rates across generations. A two-decade lag was selected to provide a few years for the implementation of military enlistment policies and to measure the association across a complete generation (eighteen years). Year fixed-effects (λ) are included in the model to capture period-specific events that are not directly measured in the regression equation (such as crime rates, changes in the economy, educational expansion, military actions, and other national conditions that are associated with that year, relative to other years in the data). Standard errors have been clustered on year. Because equation (3) is a log-level regression model, to calculate the percentage change in the adjusted incarceration rate, the coefficients for β and γ must be retransformed, such that %Δy = 100 * (e – 1) If the military provided protection against incarceration for specific demographic groups, we expect that the coefficients for β and γ will be negative.

FINDINGS

Figure 2 presents the prevalence of incarceration and veteran status among the American adult population. In 1972, nearly 308,000 Americans (or 150 per hundred thousand) were behind bars. By 2012, more than 2.2 million men and women were in prison or jail, with the incarceration rate peaking at 764 inmates per hundred thousand in 2008 (Glaze and Kaeble 2014; West and Sabol 2009). Current Population Survey data indicate that in 1972 more than twenty-five million U.S. veterans, roughly one in eight American adults and nearly one in four men, had been employed by the military. Although the number of active duty personnel increased slightly during the 1970s, the deceleration in enlistment after the implementation of an All-Volunteer Force in 1973 was rapid. Increased population growth, higher death rates for veterans of earlier service periods, and a smaller active duty population with longer average terms of service have substantially decreased the rate of veterans returning to civilian life over the last four decades. These trends suggest the relative risk of an American adult participating in either the armed forces or the penal system have become chiral (the mirror image of these trends), with the rate of change (slope) remaining relatively constant for both trends, despite differences in their absolute levels. Figure 2 highlights contemporary institutional castling.

Figure 2.

Veteran and Incarceration Rates for U.S. Adults Age Eighteen to Sixty-Four, 1972–2012

Source: Authors’ calculations from the Surveys of Inmates, the Bureau of Justice Statistics population counts, and Current Population Survey data.

Note: Rate is per one hundred thousand.

Trends in figure 2 suggest a compositional change among the veteran and incarcerated population by race and education. Much research shows the lifetime risk of incarceration has changed for white and black men with low levels of education born during the mid- to late twentieth century (Pettit and Western, 2004; Pettit et al. 2009; Heckman and LaFontaine 2010; Pettit, 2012; Ewert, Sykes, and Pettit 2014; Neal and Rick 2014). Figure 3 shows the percentage of men age eighteen through sixty-four who are veterans, by race and education. In 1972, almost 43 percent of white male dropouts were veterans. By 2012, that number had declined to 4.3 percent—a 90 percent reduction. Among black men who did not complete high school, 28.7 percent were veterans in 1972, but only 5.5 percent were by the close of 2012, a reduction of 80.8 percent.

Figure 3.

Veteran Percentage of U.S. Men Age Eighteen to Sixty-Four, by Race and Education, 1972–2012

Source: Authors’ calculations from the Surveys of Inmates, the Bureau of Justice Statistics population counts, and Current Population Survey data.

For more educated men, the percentage of veterans in the population is higher and the decline occurs more slowly. For white men who completed high school, 55.4 percent were veterans in 1972 versus 13.9 percent in 2012. For blacks with a high school diploma, the percentage declines from 44.8 percent to 13.4 percent. Interestingly, figure 3 shows approximate racial parity in the proportion of high school dropouts and high school graduates with and without prior military employment by 2012. Such convergence for both racial groups across levels of education may be the result of both the downsizing of the military and a shift toward a more educated personnel over time.

Figure 4 displays veteran and nonveteran incarceration rates by race for men age eighteen through sixty-four between 1972 and 2012. In the earliest year (1972), these men were born between 1908 and 1954, and would have been at the highest risk of serving in World War II, Korea, and Vietnam. Their counterparts in 2012 would have been born between 1948 and 1994 and were most likely to have served in Vietnam, during the early years of the All-Volunteer Force, or in recent U.S. engagements in the Middle East. Incarceration rates for black men have soared since 1972 and the gap in incarceration rates by veteran status has increased. In 1972, black nonveterans were incarcerated at a rate of 1,054 per hundred thousand. By 2012, that number had more than tripled, to 3,252 per hundred thousand. Similarly, the veteran incarceration rate among black men quintupled over this period, increasing from 490 in 1972 to 2,054 in 2012. This difference between groups suggests that the gap in incarceration rates by veteran status was 564 per hundred thousand among black men in 1972. By 2012, the veteran status gap had nearly doubled, to 1,198. Among black men, across all four decades of our data, being a veteran clearly served as protection against involvement with the criminal justice system.[9]

Figure 4.

Incarceration Rates by Race and Veteran Status for U.S. Men Age Eighteen to Sixty-Four, 1972–2012

Source: Authors’ calculations from the Surveys of Inmates, the Bureau of Justice Statistics population counts, and Current Population Survey data.

White men, however, had lower rates of incarceration for both veterans and nonveterans. The rate increased more gradually for white men than it did for blacks, and the gap between veterans and nonveterans is much smaller. For instance, white nonveteran incarceration rates increased from 130 per hundred thousand in 1972 to 442 in 2012. Comparatively, veteran incarceration rates among white men rose by a factor of 6.7, from 87 to 584 between 1972 and 2012. We estimate that, among whites, veteran incarceration rates now outpace those for nonveterans for the first time in the All-Volunteer era, and have done so since 2007. This finding suggests that veteran status operates as a protective shield against involvement with the criminal justice system for black men but not for white men.

Next, we present results from the decomposition of the intersection between the shifting educational distribution and the changing rate of incarceration between veterans and nonveterans in the American population. Our presentation in figure 3 of the changing concentration of veterans, by race and education, among the male working-age population, and the percentage of black and white men who are incarcerated, by veteran status, depicted in figure 4, are animated by research that documents how penal growth obscures the measurement of social inequality by excluding inmates from household-based surveys. This exclusion has led scholars to adjust measures of social inequality to include institutionalized populations.[10] Because data from the CPS are used to construct our rates, the trends we present for veteran status, educational attainment, and incarceration are adjusted for the exclusion of inmates from household surveys, allowing for more precise estimates of within-group educational differences for veterans and between-group incarceration rate differences among veterans and nonveterans.[11] Our focus here is identifying the unique contributions of each component in explaining racial and educational differences in incarceration. Because the goal of this decomposition is to explain the unique contribution of each factor to the observed change or difference, the effects should sum to 100 percent, and in most cases, both numbers are positive.

Samuel Preston, Patrick Heuveline, and Michelle Guillot (2001, 29) argue that in many applications, however, one factor—either differences in compositional effects or in rate schedules—“will account for more than 100% of the original difference,” particularly when “two factors work in opposite directions and there is no reason to believe such phenomena should operate in concert.” This implicitly means that one factor will be negative.[12] We contend that such processes may be at work (as displayed in figure 2) for young, African American men with low levels of education, who do not meet military enlistment criteria, face difficulties in the civilian labor market, and experience an increased risk of criminal justice contact. If this is true (as shown in figures 1, 3, and 4), then among black men only, we expect to find a positive value in excess of 100 percent for the veteran compositional difference, rendering the incarceration rate schedule for black men largely negative to offset the difference between these two components of change.

Table 1 shows that for white high school dropouts, changes in the characteristics of veterans over time account for 85.9 percent of the changes in institutional composition. Incarceration rates for poorly educated whites explain 14.1 percent of the difference. For black dropouts, however, the role of the military is more pronounced: the educational distribution of veterans accounts for 135.6 percent of the compositional change in institutions. Because the difference in the incarceration rate underexplains 35.6 percent of the change, this finding suggests that many of the poorly educated black men who have been incarcerated may have been at risk for enlistment in the military as a competing institution to criminal justice contact. These estimates—+14.1 percent (for white dropouts) and −35.6 percent (for black dropouts)—are closely aligned with research that estimates contemporary civilian incarceration rates for white and black dropouts to be 12 percent and 37.2 percent, respectively (Pettit, Sykes, and Western 2009, 13; Pettit 2012, 15). Findings from the decomposition support our contention that the negative incarceration effect in table 1 for black high school dropouts points to the differential enlistment patterns for severely disadvantaged men over time. Put simply, military enlistment may have been a protective factor for poorly educated blacks against rising levels of incarceration given that this group experienced significantly more positive compositional effects in enlistment than their white counterparts.

Table 1.

Decomposition of the Differences in U.S. Veteran Composition and Adjusted Incarceration Rates for Men Age Twenty to Thirty-Four, by Education and Race, 1972–2012

	Less Than High School		High School
	Non-HispanicWhite	Non-HispanicBlack	Non-HispanicWhite	Non-HispanicBlack
Veteran compositional difference	85.9	135.6	85.2	102.1
Incarceration rate difference	14.1	−35.6	14.8	−2.1

Source: Authors’ calculations from the Surveys of Inmates, the Bureau of Justice Statistics population counts, and Current Population Survey data.

Blacks who completed high school, however, were not significantly harmed by these compositional shifts, as indicated by numerical estimates in the last column of table 1. Because the new enlistment criteria did not exclude them, black men who had graduated from high school remained able to use the armed forces as a protective institution against the criminal justice system.[13]

Finally, table 2 presents estimates from our log-level regression model examining how veteran status is associated with incarceration for young black and white men between 1972 and 2012. Not surprisingly, the baseline model (model 1) shows that incarceration rates for black men are three times (301.1% = 100 * (e1.389 – 1)) higher than the incarceration rates of white men. Similarly, young men who have dropped out of high school have incarceration rates 2.3 times (230.7% = 100 * (e1.196 – 1)) larger than men with a high school diploma.

Table 2.

Estimates from a Model Predicting the Natural Log of the Adjusted Incarceration Rate for U.S. Men Age Twenty to Thirty-Four, by Race, Education, Veteran Status, and Enlistment Rates, 1972–2016

	Baseline(1)	M1 +Veteran(2)	M2 +Interactions(3)	M3 + %Employed inMilitary(4)	M4 +EducationInteraction(5)
Non-Hispanic black	1.389***(0.0361)	1.389***(0.0361)	1.744***(0.0249)	1.646***(0.0360)	1.654***(0.0372)
Less than high school (LTHS)	1.196***(0.0517)	1.196***(0.0518)	1.465***(0.0647)	1.654***(0.0945)	1.755***(0.104)
Veteran		−0.0247(0.0384)	0.332***(0.0407)	0.392***(0.0413)	0.392***(0.0414)
Non-Hispanic black x LTHS			−0.213***(0.0285)	−0.105*(0.0450)	−0.162*(0.0643)
Non-Hispanic black x veteran			−0.388***(0.0487)	−0.376***(0.0615)	−0.376***(0.0616)
LTHS x veteran			−0.218***(0.0594)	−0.248**(0.0741)	−0.248**(0.0742)
Non-Hispanic black x LTHS x veteran			−0.215+(0.121)	−0.370*(0.143)	−0.370*(0.143)
Percent employed by military twenty years ago				0.0699*(0.0288)	0.0573+(0.0330)
LTHS x percent employed in military twenty years ago					−0.241*(0.102)
Constant	5.192***(0.0228)	5.205***(0.0382)	4.946***(0.0380)	5.184***(0.104)	5.298***(0.111)
Observations	360	360	360	280	280
R2	0.913	0.913	0.935	0.917	0.919

Source: Authors’ compilation of data based on the Surveys of Inmates, the Bureau of Justice Statistics, the March Current Population Survey, and the Annual Social and Economic Supplement of the Current Population Survey.

Note: Authors’ calculations from a OLS regression model predicting the natural log of the adjusted incarceration rate. All models include year fixed effects (to control for period-specific events such as crime rates, changes in the economy, educational expansion, military actions, and other national conditions associated with that year, relative to other years in the data), and standard errors have been clustered on year. Non-Hispanic whites, those with a high school diploma, and nonveterans are the reference groups. All coefficients must be retransformed to estimate percentage changes (%Δy = 100*(e–1)).

p < .1

p < .05

p < .01

p < .001

Veteran status is not associated with incarceration (model 2); however, on fully interacting veteran status with race and education (model 3), the main effect of veteran status is positively associated with incarceration rates for white men. Model 3 also shows that black veterans and veterans with low levels of education have incarceration rates 19.6 to 32.2 percent lower than their reference groups. This association is so strong that the three-way interaction involving race, education, and veteran status increases by 72.1 percent in logged-units, going from a coefficient of −0.215 (in model 3) to −0.370 (in model 4), when military employment rates from a generation ago are controlled (model 4). The same interaction for black veterans with less than a high school education shows that their incarceration rates were 30.9 percent (= 100 * (e0.370 – 1)) lower than their same-race counterparts with similar formal education who were not veterans, when the percentage of military personnel from a generation ago is included in the model, pointing to both the changing size and evolving composition of military enlistees. The increased educational requirements for employment in the military are evident in model 5: a 1 percentage point increase in the employment of military servicemembers with less than a high school education a generation ago is associated with a 21.4 percent (= 100 * (e0.241 – 1)) reduction in incarceration rates.

Figure 5 displays the predicted incarceration trends from model 5 of table 2, with shaded 95 percent confidence intervals. Of particular importance is that changes in enlistment selectivity over time did not produce statistically significant differences between veterans and nonveterans who were white dropouts or black high school graduates. However, among white high school graduates, incarceration rates for veterans were higher than for nonveterans for nearly twenty years (from 1987 through 2006). Among young black men who dropped out of high school, differences in incarceration rates among veterans and nonveterans prior to the mid-1970s were not measurable. Yet, beginning around 1977, nearly four years after the transition to the All-Volunteer Force and almost a decade after 10 USC§504(a)—incarceration rates increased for black nonveterans who dropped out of high school. This trend diverges significantly from the trend associated with black veterans who dropped out of high school, confirming results from the decomposition of incarceration by race and education (presented in table 1). These findings speak to the role of contemporary institutional castling following a series of policy changes that, within a decade, would cleave off military enlistment as a potential turning point in the lives of young, undereducated black men.

Figure 5.

Fitted Linear Model of Incarceration Rates by Veteran Status Across Race and Educational Levels, 1972–2012

Source: Authors’ calculations from the Surveys of Inmates, the Bureau of Justice Statistics population counts, and Current Population Survey data.

CONCLUSIONS AND IMPLICATIONS

In the opening line of Rethinking the State, Bourdieu (1994, 1) argues that “to endeavor to think the state is to take the risk of taking over (or being taken over by) a thought of the state, i.e. of applying to the state categories of thought produced and guaranteed by the state and hence to misrecognize its most profound truth.” The misrecognition of truth occurs when social scientists do not employ a radical rethinking of the state through hyperbolic doubt—striving “to question all the presuppositions and preconstructions inscribed in the reality under analysis as well as in the very thoughts of the analyst” (1). Hyperbolic doubt facilitates the rethinking and reimagining of social problems through a critical examination of the state and its processes, functions, conventions, and teachings. Although Bourdieu’s case examined the role of state in matters of orthography (that is, correct spelling), he states that social artifacts in one realm under examination are “the product of a work of normalization and codification, quite analogous to that which the state effects concurrently in other realms of social life” (2).

In this article, we have sought to employ Bourdieuan hyperbolic doubt by reexamining the presuppositions and preconstructions of racial and educational inequalities in incarceration by exploring an often underexamined explanation for the disproportionate representation of young, undereducated men in the correctional system: the role of the military industrial complex. Growth in the penal system and diminished opportunities for enlistment among the disadvantaged during the late twentieth century have profoundly changed the contemporary demographic composition of American penal and military institutions.

In the early 1970s, increasing punishment and inequality occurred in tandem with a decline in military employment rates and a shifting educational composition of the armed forces. At the same time, an increasing share of veterans reentering civilian life meant elevated risks for criminal justice contact, depending on their race and education. Spending time in the armed forces can provide a critical opportunity for positive redirection in the life course of economically disadvantaged (Elder 1999) or delinquent (Sampson and Laub 1996) young people. If adolescents and young adults are to benefit from this institutionally driven course correction, however, they must first apply and be accepted for admittance. Yet, for the first time, the 1973 transition to the All-Volunteer Force positioned the U.S. armed forces as an active competitor in the entry-level labor market, even in the presence of shifting restrictions on enlistment among those with prior criminal justice contact (10 USC§504(a)). Black men who were admitted to the armed forces in the wake of the All-Volunteer Force tended to be positively, educationally selected relative to other blacks (Angrist and Krueger 1994; Elder 1987; Moskos and Butler 1996; Fernandez 1996; Teachman, Call, and Segal 1993; Mare and Winship 1984), and whites who found themselves in uniform typically had lower social class origins than other whites (Appy 1993).

Indeed, until recently, the Department of Defense was the nation’s largest employer as well as the largest employer of black high school graduates (Lundquist, Pager, and Strader 2018; Segal and Segal 2004). In the early years of the All-Volunteer Force, the low pay scale and limited opportunities for educational advancement effectively made the military an employer of last resort for young men (and some women) who had limited alternative educational or occupational opportunities. During the All-Volunteer Force era, the average length of enlistment more than tripled, from fewer than two years to more than six (Morin 2011), and the share of blacks among those in uniform grew, meaning that blacks were overrepresented in the armed forces for the first time in U.S. history (Nalty 1986; Segal and Segal 2004).

Rising levels of educational attainment among the U.S. population during the closing decades of the twentieth century meant that the military was able to institute and maintain high school graduation or GED certification as a nearly universal requirement for enlistment (Asch, Hosek, and Warner 2001; Day and Bauman 2000). In 1984, the armed forces instituted the Montgomery GI Bill and other benefits to provide college assistance for a limited number of veterans (Fredland et al. 1996; Thirtle 2001), a program retroactively made available to virtually all veterans who have served since 2001 (Steele, Salcedo, and Coley 2010).[14] Potential college funding, as recruitment and enlistment tools, allows the armed forces to continue to target young adults who have the aptitude to perform academically but cannot independently fund higher education—and perhaps also lack the sociocultural resources to identify alternate funding for postsecondary education (Bachman, Freedman-Doan, and O’Malley 2001; Houle 2013; Thirtle 2001).

A practical consequence of this selected recruitment has been the partial and institutional reshuffling of young, undereducated men from the military to the penal sphere by institutional castling. We show that the generational divide in who was employed by the military is related to incarceration rates two decades later, given that a 1 percentage point increase in the employment rate of military servicemembers with less than a high school education a generation ago is associated with a 21.4 percent reduction in incarceration rates. Military employment reduces incarceration rates among black and undereducated men for two reasons. First, blacks have longer military careers, on average, than whites (Moskos and Butler 1996), providing stable primary sector employment during their prime criminogenic years (Office of the Deputy Assistant Secretary of Defense, Military Community and Family Policy 2013). Longer tenure in the military also insulates men of color from the vicissitudes of the civilian labor market.

A second reason the military’s protective effects are concentrated among blacks and undereducated men is due to perceptions about military employment; blacks and whites appear to have different orientations to military service. Whereas black men are drawn to military occupational specialties that provide a high degree of convergence with the civilian labor market, white men are overrepresented in combat arms specialties, which have fewer analogous occupations outside the military (Gifford 2005; MacLean and Parsons 2010) and that place them at risk for incarceration if exposure to violence, and its physical and psychological consequences, elevates the risk of subsequent criminal offending. Blacks in the military are also significantly more likely than whites to take advantage of training and educational opportunities (Barley 1998) and are more likely to apply for promotion than to voluntarily exit the armed forces under its “up or out” structure (Moskos and Butler 1996; Asch, Miller, and Malchiodi 2012). In short, the military historically provided a path for upward mobility for disadvantaged men during distinct phases of military staffing, even though contemporary research shows that the positive benefits of enlistment have largely vanished (Bailey and Sykes 2018).

The conscription of fugitive slaves and freedmen during the Civil War set the stage for institutional castling (and recastling) in the lives of young, undereducated black men across periods of American history. Yet, contemporary institutional castling is not localized solely to the military and penal systems, palimpsests of a particular epoch, or a particular race or gender; the durability of institutional castling can be observed for other demographic groups in research on institutions (directly and indirectly) tied to the penal system and labor market. For instance, Bernard Harcourt (2006, 2011a, 2011b) shows that the deinstitutionalization of mental hospitals and asylums coincided with the growth of the penal system, suggesting that medical and penal institutions castled in the lives of disadvantaged people suffering from various health and mental afflictions. Similarly, Loïc Wacquant (2001) argues that the ghetto and penal system share a symbiotic relationship (they castled) in ways that make the ghetto look more like a prison and the prison look more like a ghetto, drawing attention to the chiral relationship between residential location and systems of surveillance and punishment for poor communities of color. Wacquant (2010) also illustrates how welfare reform at the close of the twentieth century castled the institutions of the social safety net (welfare) and the low-wage labor market (workfare) for poor, disadvantaged women, many of whom were African American and Latina. Subsequent to the period of institutional castling between welfare and workfare, the specter of penality and punishment was always lurking beneath the surface, giving rise to what Wacquant (2010) termed prisonfare, given that welfare fraud investigations and prosecutions rose after the passage of the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (Gustfason 2011). Thus, contemporary institutional castling in the lives of disadvantaged men and women has now become a normative feature of the American underclass and can be readily observed in a variety of social contexts. Military enlistment and penal confinement are but one configuration.

In sum, given the high rates of enlistment and veteran status among African American men, our findings suggest that ignoring differences in veteran status and military employment rates conceals a key aspect of racial inequality in the criminal justice system across generations. Bruce Western and his colleagues (Western and Beckett 1999; Western and Pettit 2005) show that ignoring the effects of incarceration—and specifically the omission of incarcerated men from statistics on employment and income—distorts our understanding of the true nature of racial inequality in labor market outcomes. We concur, and show that the military—as an independent, bureaucratic institution—has remarkable protective effects against the penal institution for men with low levels of education, in general, and for African American men, in particular. Bourdieu (1994, 4) powerfully wrote that “the instituted institution makes us forget that it issues out of a long series of acts of institution (in the active sense) and hence has all the appearances of the natural.” Secondary educational attainment and nonfelony status for military employment were not historically natural for servicemembers of the armed forces, particularly during World Wars I and II (Dressler 1946; Mattick 1960; Shattuck 1945). The expansion of the criminal justice apparatus under the Safe Streets Act, the simultaneous barring of felons from military enlistment under 10 USC§504(a), and the subsequent requirements of a high school diploma or GED during military downsizing meant that young men with criminal records or those who did not graduate from high school were at an increased risk of civilian labor market exclusion (during periods of deindustrialization) and correctional custody (during periods of criminal justice expansion). The castling of these two institutions in the lives of disadvantaged men requires hyperbolic doubt to reimagine how state power is exercised to shape (and reshape) the composition of the armed forces, the penal system, and the civilian labor market synchronistically.