The Role of Marriage and Military Service on Reoffending: Race, “The Respectability Package,” and the Desistance Process

Abstract

We build on prior research examining military involvement and criminal involvement by investigating the importance of acquiring the more complete “respectability package” that includes marriage as well as military experience and variations among White and Black respondents. Using data from Waves I and IV of the National Longitudinal Study of Adolescent to Adult Health (Add Health; n = 5,801), analyses use logistic regression models to assess associations of military service, marriage, and race with odds of reoffending among White and Black young adults who reported offending at Wave I. Military involvement was associated with lower odds of offending for Black respondents only, while marriage was associated with decreased odds of reoffending across both groups. Among Black respondents, analyses also highlighted the importance of acquiring both components of the respectability package (military service and marriage) in the context of today’s all-volunteer force in reducing criminal involvement.

The relationship between age and crime, referred to as the “age–crime curve,” has been described as one of the “brute facts” of criminology (Hirschi & Gottfredson, 1983). Accordingly, there is little controversy around the notion that offending behaviors emerge and rise dramatically during the adolescent years and are followed by a rapid decline as youth transition from adolescence to adulthood (Sweeten et al., 2013). This decline in offending is theorized to reflect individuals’ inclination to disengage from crime as they transition into adult roles such as employment and marriage (Barr & Simons, 2015; Giordano et al., 2002; King et al., 2007) and is referred to as desistance. Consistent with Laub and Sampson’s (2003) line of theorizing, we conceptualize desistance as a process, rather than a discrete event. Thinking of desistance as a process suggests the presence of transformation, but such transformations do not necessarily preclude individuals from experiencing a reversal of fortunes in the future and resuming criminal behaviors. Thus, even among those who report a protracted period of refraining from crime, the possibility that they may resume offending at some future time point remains. Nevertheless, a lengthy suspension in offending is strongly suggestive of the individual orienting themselves toward a more prosocial path (Giordano et al., 2002).

Sampson and Laub (1993), in developing an age-graded theory of informal social control, argued that these adult roles operate as sources of social bonds and informal social control and explain changes in criminal involvement through the life course even after accounting for underlying predispositions toward crime. Sampson and Laub also noted benefits of other adult experiences such as military service, particularly in connection with their sample of men who came of age in the 1940s, yet fewer investigations have explored the role of military service within a contemporary context. American military veterans make up approximately 7% of the total U.S. population with years of service spanning from World War II to the recent military intervention in Iraq and Syria against the Islamic State (Bialik, 2017).

We center on the specific experience of military service and marriage as anchors for making behavioral transformations and determine how race affects the transformation process. This may be particularly salient within the context of the military which may increase the odds for Black Americans to acquire both components of the respectability package (marriage and employment) relative to their odds in the civilian world (Craig & Connell, 2015; Lundquist, 2004, 2008). Ongoing concerns regarding Black enlistments in the all-volunteer force (AVF) era further reinforce the need to understand the potential benefits of military service to racial and ethnic minorities (Armor & Gilroy, 2010). Furthermore, post-service civilian opportunities such as postsecondary education and increased earnings suggest that benefits of service in the AVF continue to accrue following discharge (Brown & Routon, 2016).

In the current study, we drew on data from Waves I and IV of the National Longitudinal Study of Adolescent to Adult Health (Add Health) to examine the associations of race, marriage, and past or present military service with reoffending among those who reported offending at Wave I. We also accounted for whether respondents were active duty or employed in the civilian labor market, as veterans may be more vulnerable relative to their civilian and active military counterparts (Teachman, 2009). Finally, we also considered whether job commitment and relationship quality strengthen the associations between military service and marriage on level of offending.

Background

Theorizing Desistance

Sampson and Laub’s (1993) age-graded theory of social control posits that transitions into adult social roles, or “turning points” such as marriage, full-time employment, and military service, provide opportunities for acquiring social capital which is derived through strong social bonds with others. Investment in these social relationships builds stakes in conformity that individuals do not wish to jeopardize and thus promotes movement into a more prosocial lifestyle. This suggests that prior to the turning point, offending was largely stable and subsequently declined following the turning point and enmeshment in the new prosocial environment.

In contrast, Giordano et al.’s (2002) theory of cognitive transformation suggests that individuals must be oriented toward prosocial change, but access to adult role transitions such as marriage, employment, and military service may serve as a catalyst or “hook for change” that facilitates the transition toward a new, more prosocial orientation. In the context of the AVF, individuals who join the military are voluntarily committing themselves to enter into a long-term enforceable contract (Teachman, 2007). Thus, entry into the military may signal an active choice on the part of the actor to chart a new course away from their deviant past.

An often overlooked aspect of theories that posit a significant role for subjective changes ( see, e.g., Maruna & Roy, 2007; Paternoster & Bushway, 2009) is their attention to the intersection between agency and social structural opportunities and constraints. Individuals must not only be open and committed to change, but structural opportunities that provide access to hooks for change must also be present. Structural disadvantage in the form of racially segregated neighborhoods characterized by high levels of unemployment provide few opportunities for building a high-quality respectability package that includes both a stable family life and meaningful employment (W. J. Wilson, 1997). Consequently, life in the military as a potential hook may be particularly salient for this population.

Military Service

Within the civilian context, the segregation and economic marginalization of racial minorities has persisted (Kochhar & Fry, 2014; Pfeffer et al., 2013). Thus, for many young Black men and women, enlisting in the military has often been seen as offering a “second chance” for personal development and educational opportunities often denied to them in the civilian sector (Lowe et al., 2006). The military has been argued to serve as a “bridging environment” that allows for disadvantaged groups to gain access to occupational opportunities otherwise denied (Bouffard, 2005). Other scholars (e.g., Teachman, 2007) have drawn on Goffman’s (1961) notion of the “total institution” that buffers racial and ethnic minorities against the institutionalized racism present in civilian labor markets. Some empirical evidence has been supportive of this notion.

Craig and Foster (2013), using data from the Add Health, for example, found that while the military was associated with reduced offending for women, similar associations were not present for men. These analyses, however, did not decompose military and unemployment effects by race and ethnicity. Subsequent analyses by Craig and Connell (2015) revealed military involvement was associated with desistance among non-White respondents, but not White respondents. MacLean and Elder (2007) also noted the importance of historical context. While evidence suggests that service during World War II served as a positive turning point that promoted desistance (Elder & Shanahan, 2006; Sampson & Laub, 1993), service in the AVF era has been associated with increases in offending among those with delinquent histories (Bouffard, 2005). Bouffard’s analyses which drew on the National Longitudinal Survey of Youth, 1979 (NLSY79), however, also suggested that military service did offer some protection against offending among Black respondents. Based on these limited and mixed findings, we hypothesized the following:

Hypothesis 1a:

Past or current military involvement will be associated with lower odds of reoffending.

Hypothesis 1b:

Associations between past or current military involvement and reoffending will be stronger for Blacks compared to Whites.

Employment, whether military or civilian, has been theorized to influence desistance through the acquisition of social capital that builds stakes in conformity (Laub & Sampson, 2003). Subsequent work has qualified this thesis, however, noting that employment is more likely to facilitate desistance efforts when the actor is already oriented to committing toward a more prosocial path (Bushway & Reuter, 1997; Giordano et al., 2002). Thus, one’s commitment toward employment may be more meaningful as a predictor of offending than whether one is simply employed or not. A study by Huiras et al. (2000) found workplace misconduct was lowest when employees had greater commitment to the job, thereby providing some empirical support for this idea. Consequently, we hypothesized the following:

Hypothesis 1c:

High job commitment will strengthen the association between military involvement and reoffending.

Marriage

Derived from their analyses of the longitudinal data collected by Sheldon and Eleanor Glueck (1950), Sampson and Laub’s theoretical work posits that marriage also operates as a turning point that facilitates desistance through the acquisition of social capital and a restructuring of routine activities (Laub & Sampson, 2003; Sampson & Laub, 1993). They further theorize that marriage will operate most effectively in the context of strong marital bonds. Thus, high-quality marriages characterized by high levels of warmth, love, closeness, and trust will be more meaningful in connection with desistance efforts relative to those that are lacking in these traits. Marital partners are also viewed as operating as agents of informal social control that restructure routine activities away from antisocial peers toward more prosocial activities and relationships. Thus, we hypothesized the following:

Hypothesis 2a:

Married respondents will have lower odds of reoffending compared to those in cohabiting or dating relationships.

Hypothesis 2b:

Relationship quality will interact with marriage to strengthen the association between marriage and reoffending.

To date, only a few studies have explicitly tested whether associations between marriage and offending vary across racial and ethnic groups. Doherty and Ensminger (2013) examined marriage as a turning point among a cohort of economically disadvantaged African Americans. Results indicated a stronger association between marriage and reduced offending for men relative to women. Using the public data from the Add Health, Craig (2014) found that those who reported ever being married were marginally less likely to offend relative to those who had never married. Upon decomposing effects by race, this association was not observed among Black respondents. In contrast, Bersani and DiPietro (2014) using data from the NLSY97 found that Black men enjoyed a stronger “marriage effect” on offending relative to White men, with no difference observed for White men. Given these mixed findings, we had no a priori expectations with regard to associations between marriage and desistance varying by race.

“The Respectability Package”

Previous findings from Giordano et al. (2002) suggest the importance of considering the joint effects of marriage and employment. That is, those who are able to access both elements have increased odds of desistance relative to those who are lacking one or both elements of the respectability package. Additionally, their findings based on a highly delinquent sample indicated that a relatively small proportion of their sample was able to access this package (8%) and it accounted for relatively little of the variability among their previously institutionalized sample. Yet, there is reason to believe that across a wider cross section of the population, military service is linked to marriage, particularly among Black Americans.

During the era of conscription, marriage was largely discouraged (Lundquist, 2004). The emergence of the AVF, however, has promoted family-friendly policies that are viewed as facilitating recruitment and retention efforts. Consequently, military enlistment may promote earlier entries into marriage relative to the civilian context where educational pursuits may take priority. Furthermore, although Black Americans having considerably lower odds of entering marriage relative to their White counterparts (Kuo & Raley, 2016), these disparities are often absent within military samples.

In analyses using the NLSY79, Teachman (2007) found that military service was associated with an increased probability of marriage. While there was evidence of selectivity, selection effects were strongest among White personnel, suggesting that for racial and ethnic minorities, marriage is more likely an outcome of military involvement rather than a predictor for entry into the military. Lundquist (2004) also found that marriage rates of those in the military were higher relative to those in the civilian sector. Additionally, Black–White differences in marital rates observed in the civilian subsample were not observed in the military subsample. These selection processes suggest that military involvement may not only reduce offending directly but also indirectly by providing a pathway to marriage, particularly for Black young adults regardless of whether they make a career out of the military or return to civilian life. Thus, any military involvement, past or present, may combine with marriage to promote desistance. This relationship may take one of two forms. The first is an additive effect. That is, military involvement and marriage independently contribute to desistance efforts. Although each is beneficial in its own right, the presence of both translates to higher odds of desistance. The second is an interactive effect, whereby the presence of one amplifies the effect of the other. That is, marriage may be particularly beneficial within the military context or vice versa. We hypothesized the following regarding the combined effects of military involvement, current employment, and marriage:

Hypothesis 3a:

Military involvement (past or present) and marriage will each contribute to higher odds of reoffending independent of the other (additive effect).

Hypothesis 3b:

Marriage and military involvement will interact to strengthen their respective associations with reoffending (interactive effect).

Method

Data

We drew on the restricted data from Wave I and Wave IV of the National Longitudinal Study of Adolescent and Adult Health. The Add Health is a large, nationally representative longitudinal study of adolescents first interviewed in1994 and 1995. The initial core study employed a two-stage stratified design that included 132 schools at the first stage. At the second stage, approximately 90,000 students in Grades 7–12 completed the in-school survey. An estimated 20,000 students were selected to complete the in-home questionnaire and to be followed longitudinally. The Wave IV in-home interviews were conducted in 2008 when respondents were 24–32 years old. The study was able to interview 80.3% of eligible respondents at Wave IV. Interviews were conducted using a 90-min computer-assisted personal interview/computer-assisted self-interview instrument. For more information on the Add Health study design, see Harris (2005).

The analytic sample consisted of respondents who completed both Wave I and Wave IV; reported engaging in at least one delinquent act at Wave I; were in a marital, cohabiting, or dating relationship at Wave IV; identified as non-Hispanic White or non-Hispanic Black; and had valid Wave IV weights (n = 6,031). Key challenges associated with assessing desistance are establishing the time order of variables and ruling out potential selection effects (Skardhamar et al., 2015). Due to the limited number of data points and inconsistencies in measures between waves, a time series or fixed effects analysis was not feasible given our variables of interest. To help address these limitations, we eliminated respondents who first enlisted in the military prior to 1996 or after 2007 (n = 41). Thus, we could be sure that self-reports of offending at Wave I occurred prior to military enlistment and Wave IV reports of offending followed military enlistment. Similarly, respondents who had continuously been in the same relationship since Wave I were also eliminated (n = 90).

Analyses were performed to compare results with and without multiple imputation and no substantive differences were observed. Given the relatively small amount of missing data (<2%) and the lack of observed differences, listwise deletion offered the most parsimonious solution for handling missing data. Following the removal of respondents with any missing data (n = 99), we arrived at our final analytical sample (n = 5,801).

Measures

Reoffending.

Respondents were asked at Wave IV if they had engaged in an array of criminal behaviors over the past 12 months. An 11-item mean scale was constructed from these items (e.g., property damage, burglary, fighting). Items included both violent and non-violent offenses. Table S1 (see Online Appendix) rank ordered items by prevalence with “getting into a serious physical fight” as the most common offense (7.3%) and “use someone else’s credit card, bank card, or automatic teller card without their permission or knowledge” as the least common (0.6%). Response categories were 0 for never, 1 for 1 or 2 times, 2 for 3 or 4 times, and 3 for 5 or more times. McDonald’s omega for this scale was approximately .74.

The high number of zeros in the adult offending measure ruled out the possibility of employing ordinary least squares as this would violate the assumption of normality. Tobit regressions are often used for censored data such as ours and this possibility was explored. The distribution of the errors, however, did not meet the requirements of normality, indicating the presence of heteroscedasticity and would have resulted in biased estimates (T. Wilson et al., 2018). Consequently, we chose to employ a dichotomous measure for level of offending at Wave IV that was coded as 1 for those respondents who reported at least one offense at Wave IV (21.7%) and 0 for those respondents who scored 0 on adult offending (78.3%).

Military/employment context.

Ever in the military was assessed using a binary variable that included those with past or present military involvement. Given that ever in the military may include both active and veteran members, we control for current employment status. This was assessed using three binary variables to differentiate respondents’ employment at Wave IV—active military, employed full time in the civilian labor market, and employed part time in the civilian labor market or unemployed. Part time or unemployed was used as the reference category. Job commitment was measured using a single item which asked respondents “Which one of the following best describes your primary job?” Respondents who answered “it is part of my long-term career or work goals” or “it is preparation for my long-term career or work goals” were coded as 1 and 0 otherwise.

Romantic context.

Preliminary analyses revealed no differentiation between cohabiting and dating couples on offending at Wave IV. Consequently, we created a binary variable to distinguish between married and nonmarried individuals. Relationship quality was measured using a 7-item mean scale that was centered at the mean. Respondents were asked how much they agreed or disagreed with statements such as “we enjoy doing even ordinary, day-to-day things together” and “I am satisfied with the way we handle our problems and disagreements” (see Online Appendix Table S2). The McDonald’s omega for this scale is approximately .89.

Race.

Due to the low number of respondents reporting military involvement, we limited our analyses to those respondents who identified as either White or Black. Binary variables were created for each with White used as the reference.

Sociodemographic variables.

A number of additional sociodemographic variables were also included as controls. Female is coded as 1 for female respondents and 0 for male respondents. Age measured in years is centered at 24, the youngest age observed at Wave IV. Educational attainment is assessed at Wave IV using four dummy variables—less than 12 years, high school graduate, some college, and college graduate. Financial hardship is a 6-item summed scale measured at Wave IV. Respondents were asked whether there was a time in the past 12 months they had experienced difficulty in paying rent, telephone service or utilities, utility disconnection, eviction, or food insecurity. McDonald’s omega for this scale is approximately .74. This measure was quite skewed, given that most respondents did not report financial hardship. Thus, we transformed the variable by taking the natural logarithmic of the scale and centered it at the mean of the transformed variable. Religiosity was assessed using a single item that asked respondents how often they attended religious services in the past 12 months. Responses ranged from 0 for never to 5 for more than once a week. This variable was also centered at the mean.

Early risk factors.

Childhood victimization was assessed using a binary measure that relied on retrospective reports of verbal, physical, and sexual abuse during childhood. Respondents who reported experiencing any of these were coded as 1 and 0 otherwise. Despite our restricting the analytic sample to those who reported delinquent behavior at Wave I, we included a continuous measure of adolescent delinquency to control for the propensity to offend. Adolescent delinquency was measured using items and procedures similar to those for the desistance measure. Response categories were 0 for never, 1 for 1 or 2 times, 2 for 3 or 4 times, and 3 for 5 or more times. The 11-item mean scale demonstrated good internal reliability with a McDonald’s omega of .77. Due to the skewed nature of this measure, we employed a natural logarithmic transformation and centered it at the mean. Items are listed in Table S3 (see Online Appendix) and ranked by prevalence. Similar to reoffending, the most prevalent item was “get into a serious physical fight (55.4%) and the least prevalent item was “use or threaten to use a weapon to get something from someone” (6.7%). Mean scores of adolescent delinquency for military and civilian respondents were not significantly different.

Analytic Strategy

Analyses included bivariate analyses decomposed by race and multivariate analyses. Multivariate analyses are comprised of a series of logistic regressions. Models are labeled and presented in the same order as our hypotheses described above. All analyses were conducted using Stata 14 (StataCorp, 2015) and were weighted and adjusted for the complex survey design (for sampling details, see Harris, 2005).

Results

Table 1 provides descriptive statistics for the full sample and by race and highlight structural differences between the White and Black samples. Race differences were assessed using the adjusted Wald test in Stata. Black respondents reported higher rates of offending (27.3%) compared to White respondents (21.0%). No race differences were observed in military involvement or in active military enlistment. With respect to civilian labor force participation, however, a lower proportion of Black respondents (63.0%) reported being employed full time compared to White respondents (71.0%). Rates of high levels of job commitment among Black respondents (55.9%) were also lower compared to Whites respondents (65.0%). Race differences also emerged with respect to marriage and relationship quality, with Black respondents reporting lower rates of marriage (25.9%) and relationships quality (3.89) compared to White respondents (44.3% and 4.04, respectively). Within our analytic sample, Black respondents also were more likely to be female (47.7% vs. 40.5%), slightly older (28.3 years vs. 28.1 years), reported lower rates of college completion (19.1% vs. 28.0%), experienced more financial hardship (0.44 vs. 0.27), and attended religious services more frequently (1.92 vs. 1.24) compared to Whites.

Table 1.

Weighted Means and Percentages for Variables by Race.

	Total		White (80.2%		Black (19.8%)		Adjusted Wald F
	Mean/%	SE	Mean/%	SE	Mean/%	SE
Desistance at Wave IV	22.2%		21.0%		27.3%		6.72*
Ever in the military	7.5%		7.2%		8.4%		1.09
Current employment status
Active military	1.4%		1.4%		1.4%		0.01
Full-time civilian	69.4%		71.0%		63.0%		9.08**
(Part-time or unemployed civilian)	29.2%		27.6%		35.6%
High job commitment	63.2%		65.0%		55.9%		17.46***
Married	40.6%		44.3%		25.9%		67.12***
Relationship quality (range = 1–5)	4.01	(.014)	4.04	(.016)	3.89	(.026)	21.78***
Female	41.9%		40.5%		47.7%		12.07***
Age (range = 24–34)	28.18	(.032)	28.15	(.037)	28.30	(.061)	0.46
Educational attainment
Less than 12 years	11.1%		10.2%		14.6%		4.17*
(High school graduate/GED)	18.6%		17.9%		21.4%
Some college	44.1%		43.9%		44.9%		0.15
College graduate	26.3%		28.0%		19.1%		8.14**
Financial hardship (logged; range = 0–1.95)	0.31	(.009)	0.27	(.010)	0.44	(.021)	37.41***
Religiosity (range = 0–5)	1.38	(.026)	1.24	(.029)	1.92	(.054)	78.42***
Childhood victimization	54.5%		54.4%		54.8%		0.03
Adolescent delinquency (logged; range = 0.09–1.39)	0.29	(.004)	0.287	(.004)	0.285	(.008)	0.03

Source. National Longitudinal Study of Adolescent and Adult Health.

Note. n = 5,801.

GED = General Education Diploma.

^*p < .05,

^**p < .01,

^***p < .001.

Table 2 presents the results of the logistic regressions. Model 1 assessed the influence of the military/employment context on desistance. Although ever in the military was associated with lower odds of offending (odds ratio [OR] = 0.636, SE = .120) in support of Hypothesis 1a, no differences were observed for current employment status. High job commitment, however, was associated with a decrease in the odds of offending (OR = 0.776, SE = .068). Additionally, Black respondents had higher odds (OR = 1.497, SE = .158) of offending relative to White respondents. Women had lower odds of offending (OR = 0.391, SE = .039), and the odds of offending decreased with age (OR = 0.852, SE = .023) and religious service attendance (OR = 0.898, SE = .023). Financial hardship (OR = 2.071, SE = .165), childhood victimization (OR = 1.542, SE = .154), and adolescent delinquency (OR = 3.722, SE = .656) were all associated with increased odds of offending.

Table 2.

Weighted Logistic Regression of Level of Offending at Wave IV Among Youth Reporting Delinquency at Wave I.

	Model 1		Model 2		Model 3
	Hypothesis 1a		Hypothesis 1b		Hypothesis 1c
	OR	SE	OR	SE	OR	SE
Intercept	0.576**	.111	0.568**	.109	0.576**	.111
Ever in the military	0.636*	.120	0.760	.159	0.638	.226
Ever in the Military × Black	—	—	0.442*	.176	—	—
Ever in the Military × High Job Commitment	—	—	—	—	0.994	.404
Current employment status
Active military	0.618	.266	0.609	.268	0.617	.309
Full-time civilian	1.054	.105	1.051	.104	1.054	.105
High job commitment	0.776**	.068	0.777**	.068	0.776**	.073
Married	—	—	—	—	—	—
Relationship qualitya	—	—	—	—	—	—
Black	1.497***	.158	1.582***	.168	1.497***	.158
Female	0.391***	.039	0.390***	.039	0.391***	.039
Ageb	0.852***	.023	0.852***	.023	0.852***	.023
Educational attainment
Less than 12 years	0.869	.147	0.868	.147	0.869	.147
Some college	1.131	.122	1.134	.121	1.131	.122
College graduate	1.012	.135	1.017	.136	1.012	.135
Financial hardshipa	2.071***	.165	2.073***	.165	2.071***	.165
Religiositya	0.898***	.023	0.897***	.024	0.898***	.024
Childhood victimization	1.542***	.154	1.539***	.154	1.542***	.154
Adolescent delinquencya	3.722***	.656	3.700***	.650	3.722***	.658
F statistic	28.70***		27.41***		26.79***

	Model 4		Model 5		Model 6		Model 7
	Hypothesis 2a		Hypothesis 2b		Hypothesis 3a		Hypothesis 3b
	OR	SE	OR	SE	OR	SE	OR	SE
Intercept	0.607**	.098	0.607**	.098	0.638*	.125	0.643*	.125
Ever in the military	—	—	—	—	0.789	.169	0.694	.168
Ever in the Military × Black	—	—	—	—	0.434*	.179	0.462	.197
Ever in the Military × married	—	—	—	—	—	—	1.369	.477
Current employment status
Active military	—	—	—	—	0.813	.372	0.745	.335
Full-time civilian	—	—	—	—	1.102	.112	1.102	.112
High job commitment	—	—	—	—	0.824*	.075	0.822*	.075
Married	0.514***	.054	0.516***	.054	0.529***	.057	0.517***	.057
Married × Relationship Quality	—	—	1.071	.126	—	—	—	—
Relationship qualitya	0.843**	.053	0.821*	.066	0.845**	.053	0.845**	.053
Black	1.283*	.141	1.282*	.141	1.369**	.147	1.365**	.146
Female	0.413***	.040	0.414***	.040	1.369**	.147	0.395***	.040
Ageb	0.869***	.023	0.869***	.023	0.869***	.023	0.870***	.023
Educational attainment
Less than 12 years	0.902	.162	0.899	.162	0.883	.157	0.881	.157
Some college	1.097	.122	1.093	.123	1.138	.125	1.137	.125
College graduate	0.944	.130	0.941	.130	0.981	.139	0.979	.138
Financial hardshipa	2.040***	.172	2.044***	.171	2.034***	.173	2.034***	.173
Religiositya	0.937*	.024	0.936*	.024	0.937*	.024	0.936*	.024
Childhood victimization	1.496***	.151	1.495***	.151	1.503***	.152	1.504***	.152
Adolescent delinquencya	3.421	.610	3.420***	.611	3.421***	.610	3.426***	.613
F statistic	29.92***		27.96***		23.48***		22.56***

Source. National Longitudinal Study of Adolescent and Adult Health.

Note. n = 5,801. Reference categories are part-time employment or unemployed, low job commitment, cohabiting or dating, White, male, high school graduate/GED, and no childhood victimization.

^aCentered at the mean.

^bCentered at 24 years.

^*p < .05.

^**p < .01.

^***p < .001.

Model 2 tested Hypothesis 1b by introducing the interaction between ever in the military and Black. This interaction was significant showing that the association between ever in the military and desistance varied by race. In Model 2, the inclusion of the interaction term meant the OR for ever in the military was the result for Whites only and indicated no association between military involvement and desistance. By switching out the reference categories for race, we were able to obtain the OR for ever in the military for Blacks (OR = 0.336, SE = .120) which was significant at p < .01. Thus, Hypothesis 1b was also supported. Model 3 tested whether the association between ever in the military and desistance was conditional on high job commitment. This interaction was not significant; therefore, Hypothesis 1c was not supported.

Model 4 examined the associations between the romantic context variables and desistance. Marriage (OR = 0.514, SE = .054) and relationship quality (OR = 0.843, SE = .053) were both associated with decreased odds of offending, thereby supporting Hypothesis 2a. Additionally, the romantic context attenuated much of the effect for race on desistance. Examination of the covariates revealed that marriage and relationship quality each contributed to this attenuation and thus accounted for much of the Black–White gap in the odds of offending. Model 5 tested whether relationship quality strengthened the association between marriage and offending. This interaction was not significant, thus Hypothesis 2b was not supported.

Model 6 is a full model that tested the additive effects of military involvement and marriage on desistance. Having already established the race-specific association of ever in the military and desistance, we also included the ever in the Military Black interaction term. Findings showed that similar to Model 2, ever in the military was not significant for Whites, but when the reference category was switched, it remained significant for Blacks (OR = 0.342, SE = .127) at p < .01. Marriage also remained significant (OR = 0.529, SE = .057). Together these results supported Hypothesis 3a that military involvement and marriage each contributed to the odds of offending. Finally, we tested our interactive model, Hypothesis 3b, but the interaction between ever in the military and marriage was not significant.

To aid in the interpretation of the results, predicted probabilities were calculated based on results from Model 6 and graphed (Figure 1). Among White respondents, married individuals have lower odds of offending compared to unmarried individuals irrespective of military involvement. No differences were noted between White respondents based on military involvement. The pattern for Black respondents differed in that both military involvement and marriage decreased the probability of offending independent of the other. Thus, Black respondents with both components of the respectability package had a lower probability of offending compared to those with only one or neither component. Nevertheless, even those with only one of the protective measures (marriage or military) had a lower probability of offending compared to those who had neither.

Figure 1.

Predicted probability of offending by military status, marital status, and race.

Discussion

In the current study, we found evidence to support four of seven of our hypotheses. First, military involvement was associated with lower odds of offending. Furthermore, our hypothesis that the association between military involvement and level of offending would vary by race was also supported. Specifically, the protective benefits of military involvement on offending were not present for White respondents. Past or current enlistment in the military was uniquely protective for Black respondents even after controlling for marriage. In addition to a steady paycheck, subsidized housing, health insurance, pension benefits, and greater job security found in the military may be particularly meaningful for Black respondents who feel they have few opportunities to gain access to these benefits in the civilian labor market (Lundquist, 2004; Teachman, 2007). Moreover, the military may also serve to disrupt routine activities associated with peer influences. Entry into the military may be a means of “knifing off’ from deviant peers that would be more difficult to do in the context of full-time employment in one’s neighborhood of origin (Pyrooz, Decker, & Webb, 2010). Changes in the geographic environment correspond to changes in the social milieu which may help initiate the development by distancing oneself from a past identity (Maruna & Roy, 2007).

While the literature on work and recidivism has produced mixed results, a few consistent themes have emerged. Employment effects are likely to be age-graded, contact with the criminal justice system weakens future employment opportunities, and employment quality may be more relevant than the simple presence or absence of a job (Uggen & Wakefield, 2008). Within the current study, we found support for the importance of job quality in that job commitment was consistently associated with offending across racial and ethnic groups and military and employment statuses. This is consistent with findings from Huiras et al. (2000), who found that individuals who viewed their work as part of their long-term career goals were less likely to engage in workplace misconduct relative to those who viewed their employment as a “survival job.” In the current analyses, however, high job commitment did not moderate the association of military involvement on desistance. Contrary to Laub and Sampson’s (2003) line of theorizing, job commitment appeared to exert its own independent influence on patterns of offending. Furthermore, no significant differences were observed for current employment status. This pattern of findings suggests similar levels of offending for both active duty personnel and veterans. Prior work by Culp et al. (2013) found that AVF veterans had higher arrest rates compared to draft era veterans. They noted, however, that a number of factors could account for these differences including the trend toward mass incarceration that emerged following transition to the AVF era.

Prior work by Laub et al. (1998) highlighted the gradual effects of marriage on desistance, suggesting that as the relationship matures, the strength and quality of the bond increases. In support of our hypothesized relationship, we found a robust association between marriage and decreased odds of offending. This association remained net of military involvement, current employment, and high job commitment and was similar for both Black and White respondents. Similar to findings for high job commitment, however, we found that while relationship quality was associated with decreased odds of offending, it did not moderate the effects of marriage on offending. Instead, both marriage and relationship quality each contributed independently to the level of offending. It may be that marriage and relationship quality influence the continuity of offending through different mechanisms. Consistent with a social control perspective, marriage may help to reshape routine activities that make offending more difficult. Relationship quality, on the other hand, may operate less directly by instilling positive feelings of the self that are in turn supported by and reflected back by the partner (Giordano et al., 2007). Consistent with symbolic interactionism and differential association, this results in definitions of crime that are less conducive to offending.

Finally, with respect to testing the joint effects of military involvement and marriage, our analyses supported our hypothesis for an additive effect, rather than interactive for Black respondents. Among Whites, married individuals demonstrated a lower probability of offending, regardless of whether there was any past or present military involvement. Among Blacks, however, benefits appeared to accrue in an additive fashion such that those with either marriage or military experience had a lower probability of offending, relative to those with neither, and those with the full respectability package (military involvement and marriage) had the lowest probability of offending.

We concur with Bouffard (2005) and others (Craig & Connell, 2015; Lundquist, 2004; Teachman, 2007), who have pointed to the military as a bridging environment that provides individuals with structural opportunities less readily available to them in the civilian labor market. Such opportunities not only facilitate efforts to refrain from offending through the acquisition of financial and social capital accrued through one’s employer but also provide access to additional sources of social capital such as marriage. Prior work examining levels of commitment to American ideals of economic success show little evidence that Black Americans are any less committed to achieving the American dream relative to White Americans (Cernkovich et al., 2000). On the contrary, evidence suggests a greater willingness to work hard and make sacrifices in order to achieve aspirational goals. Thus, structural constraints that systematically limit access and opportunities for segments of the American population to accrue the kind of social capital that has been shown to be associated with reduced offending creates for some a perpetuating cycle of disadvantage. As previously noted by Bersani and DiPietro (2014), such patterns have potentially profound implications for any individual agentic efforts toward charting a new course away from crime.

Another possibility is that the military may provide access to mentors who serve not only as an important source of social support but also as models of prosocial behavior. This would be consistent with the line of theorizing put forth by Farrall et al. (2010) who suggest that friends and family serve to enable individual efforts to negotiate structural constraints that facilitate offending. Additionally, as further noted by Maruna and Roy (2007), knifing off is more likely to promote desistance when it is accompanied by new scripts for behavior. Thus, desistance through military involvement, even in the context of the reserves, may reflect socialization processes associated with the acquisition of new behavioral scripts and definitions. Greater attention to how professional socialization processes across an array of occupations relate to criminal offending would be a welcome addition to the literature.

The current study is not without its limitations. First, due to the small number respondents with military involvement, our analyses were limited to White and Black respondents. Nor, were we able to decompose results by gender. Second, we cannot make any causal claims given that are analyses were not able to fully account for the sequencing of marriage, military, and offending and possible selection effects. Although the military does allow for felony waivers to those with past offending records, the issuance of these waivers has historically been higher during times of war (Boucai, 2007). Given that nearly three quarters of our sample entered the military prior to 9/11, it is possible that some of the observed association reflects selection processes. This may help explain why contrary to some prior studies (see Baktir et al., 2018) we do not find evidence of military involvement as an aggravating factor on offending. Yet, post hoc analyses revealed that Black respondents entering the military post-9/11 had lower odds of offending relative to those who entered the military pre-9/11. This could reflect a different kind of selection process, whereby the military’s use of a “whole-person screening” has selected exceptional applicants not only less likely to offend but more likely to be successful (Lundquist et al., 2018). This finding should be interpreted with caution, however, due to the small sample size of Black military personnel who continued to offend (n = 18). Finally, our dichotomous measure of desistance does not fully capture the more processual nature of desistance. Initial analyses conducted using a Tobit model produced substantively similar results but was rejected in favor of the logistic regression analyses presented here due to the presence of heteroscedasticity.

Our study limitations notwithstanding, these findings have implications for promoting desistance, particularly as it relates to racial and ethnic minority populations. Given the consistent effects of marriage on desistance, marriage promotion policies may seem intuitively attractive, but we would urge caution before embracing such directives. Despite massive welfare reform policies aimed at reducing poverty and single motherhood (i.e., marriage promotion policies), rates of marriage have continued to decline, while poverty rates have soared (Acs, 2007). Furthermore, evidence indicates that although rates of marriage continue to be lower among marginalized groups, intentions to marry remain high across racial, ethnic, and socioeconomic categories, (Kuo & Raley, 2016; Raley et al., 2015).

Instead, our results suggest for many Black Americans, enlistment in today’s AVF provides a pathway for acquiring the full respectability package that consists of both employment and marriage which helps facilitate individuals’ efforts to refrain from offending. Today’s AVF presents itself as a compelling case that merits further study of how to address social and economic inequalities that inhibit marriage and by extension, desistance from criminal offending. Today’s AVF has refashioned itself from an institution that actively discouraged marriage as reflected in the old Army adage, “If the Army wanted you to have a wife, it would have issued you one,” to one that more actively embraces family-friendly policies (Bouffard, 2005; Lundquist, 2004; Teachman, 2007). This stands in stark contrast to criminal justice policies that have effectively disrupted the lives of many American families, particularly families of color, through mass incarceration (Western & Wildeman, 2009). Policy efforts aimed at identifying specific areas of strength in the military and replicating them in other institutional settings are worth pursuing.

Author Biographies

Wendi L. Johnson is an assistant professor in the Department of Sociology, Anthropology, Social Work, and Criminal Justice at Oakland University. Her research focuses on micro-sociological processes including human development, interpersonal relationships, and intrapersonal change in risk behaviors and covers substantive areas relating to parent-child relationships, intimate partner violence, and criminal desistance.

Peggy C. Giordano, PhD, is a distinguished research professor in the Department of Sociology at Bowling Green State university. Her research focuses on qualities and dynamics of close relationships during adolescence and young adulthood, and the influence of those close ties on a range of developmental outcomes, including juvenile delinquency and intimate partner violence.