The More You Have, The More You Lose: Criminal Justice Involvement, Ascribed Socioeconomic Status, and Achieved SES

Abstract

In the present study, we examine the relationship between involvement in the criminal justice system and achieved socioeconomic status (SES), as well as the moderating effect of ascribed SES. Using data from the National Longitudinal Study of Adolescent to Adult Health, we find a nonlinear relationship between criminal justice involvement and achieved SES, such that deeper involvement leads to increasingly negative consequences on achieved SES. Furthermore, those coming from the highest socioeconomic backgrounds are not “protected” from the deleterious consequences of system involvement, but instead experience the greatest declines in achieved SES relative to where they started. In contrast, the effect of criminal justice involvement for those from below average ascribed SES is not significant. Our findings reinforce how normal such experiences are for people with the fewest resources, and also how system involvement inevitably destroys human capital, undermines future life chances, and ultimately promotes a “rabble” class.

The growth of the criminal justice system (CJS hereafter) in the United States has led to an overabun-dance of damaging individual and societal collateral consequences, ranging from economic hardship to health and familial problems (Massoglia 2008; Western 2002, 2006; Wildeman 2009; Wildeman and Western 2010). Moreover, mass incarceration has exacerbated the lives of poor, male minorities, as these individuals are most destined for incarceration (Sampson and Loeffler 2010; Tonry 2011; Western 2006; Western and Wildeman 2009). And while incarceration has been credited by some with contributing to the decline in violence across the U.S. (Levitt 2004), society no longer sees the same benefits from the system (Clear 2007), as harsh criminal justice practices continue to preserve inequality by targeting the poorest of the poor in society, an underclass that John Irwin (1985) refers to as, “the rabble.”

A key way in which inequality is perpetuated lies in the effect that a criminal record has on hindering one’s attained socioeconomic status (SES hereafter) (Western 2002; Western and Beckett 1999). Moreover, preexisting disadvantage combined with the consequences associated with criminal justice involvement compounds the limited socioeconomic opportunities available to released offenders, as such experiences prolong a stable life course of disadvantage (Sampson and Loeffler 2010). Yet, few studies have extensively considered the depth to which an individual must be involved in the system before experiencing the negative effects on socioeconomic opportunities. Furthermore, prior research provides little insight into the ways in which ascribed socioeconomic factors may moderate the relationship between system involvement and later achieved SES. Involvement in the CJS may have only some consequences for those from advantaged backgrounds as they possess greater social capital to overcome blocked opportunities and have more resources with which to avoid the imposition of negative labels (Turk 1969), or, alternatively, these experiences may induce downward social mobility and ultimately lead to more detrimental outcomes relative to those of lower socioeconomic standing.

Using data from the National Longitudinal Study of Adolescent to Adult Health (Add Health), we examine multiple questions about the association between involvement in the CJS and later socioeconomic attainment. First, are all forms of involvement in the system harmful to future socioeconomic prospects or only the most serious and prolonged—i.e., incarceration? In the present study, we examine how prior criminal justice experiences ranging from just having been stopped by the police all the way to being incarcerated shape social mobility. Second, are the deleterious effects of system involvement the same for people of different socioeconomic backgrounds? Does coming from a family with greater educational and economic resources help to buffer against the “mark of a criminal record” (Pager 2003), or does having more to lose just mean that you end up losing more? In this study, we examine how ascribed SES moderates the relationship between system involvement and achieved SES. Our findings buttress the need for policies that proactively try to keep people from being pulled deeper into the CJS, and reinforce just how readily system involvement can destroy human capital, undermine future life chances, and ultimately promote the rabble transition.

LITERATURE REVIEW

Collateral Consequences and “The Rabble”

As the size, intrusiveness, and punitiveness of the CJS has grown over the past forty years, researchers have become increasingly concerned about what social and economic consequences exist for those who have a criminal record (Apel and Sweeten 2010; Geller et al. 2012; Miller and Barnes 2015; Pager 2003; Pager, Western, and Bonikowski 2009; Swisher and Roettger 2012; Uggen et al. 2014; Western 2002, 2006; Wildeman 2009, 2014; Wildeman and Western 2010). For instance, there is a robust consensus that incarceration has detrimental familial consequences (Geller et al. 2012; Wildeman 2009; Wildeman and Western 2010). Aside from the evidence of incarceration becoming a normative life course transition for young minority males with economic disadvantages (Sampson and Loeffler 2010; Western 2006), removing individuals from society in this manner also means removing them from their families and ultimately placing successive generations at risk for increased aggression (Geller et al. 2012), depression, delinquency (Swisher and Roettger 2012), homelessness (Wildeman 2014), and hampered educational attainment (Miller and Barnes 2015).

John Irwin (1985) explains how (system involvement), “…tends to maintain people in a rabble status or convert them to it” (p. 45). Those referred to as the rabble are individuals detached from society through involvement in the system while also acquiring the perception of being, “…irksome, offense, threatening, capable of arousal, even protoeveolutionary” (Irwin 1985:2). Thus, aside from the explicit collateral consequences through means such as damaging one’s family life, those exposed to the CJS are conditioned to see themselves as unworthy of acceptance from conventional society.

Other latent consequences associated with criminal justice involvement arise in the loss of property, particularly as deeper stages of the system (i.e., incarceration) leave individuals susceptible to falling behind on obligations such as payments for personal property (Irwin 1985). Moreover, upon incarceration, individuals are immediately separated from their social circle and inevitably fail to keep up with any employment or educational responsibilities. As such, the incapacity to perform societal duties and the absence of human agency becomes an abrupt realization upon entry into the CJS (see Emirbayer and Mische’s (1998) conceptualization of the temporally embedded process of human agency).

Criminal Justice Involvement and SES

The consequences after release become increasingly apparent when assessing measures of one’s socioeconomic achievements (Apel and Sweeten 2010; Pager 2003; Pager et al. 2009; Uggen 2000; Western 2002). While employment (Uggen 2000) and education (Blomberg et al. 2011) have been shown to promote desistance, a fundamental problem exists in that a criminal record hinders the likelihood of obtaining employment and pursuing education (Uggen and Stewart 2015). Devah Pager’s (2003) experimental work exposes the ways in which criminal records affect employment, but also reveals that building rapport with employers can alleviate the detrimental consequences associated with a record (Pager et al. 2009). Unfortunately, this opportunity only arises when an interview actually occurs, and having a criminal record minimizes such opportunities. Furthermore, prior convictions are not the sole contributor to reduced employment chances. A record of arrest without conviction can burden socioeconomic achievements, as Christopher Uggen and associates (2014) demonstrate when sending pairs of applicants out to apply for entry-level positions (with one individual having a disorderly conduct arrest record that did not lead to a conviction). Results show that callback rates are lower for those with an arrest on their record compared to the control group (Uggen et al. 2014).

Few studies have examined the association between criminal justice involvement and achieved education; more often considered are studies on the association between system involvement and dropping out of high school. Using data from the Rochester Youth Development Study (RYDS), Giza Lopes and colleagues (2012) show arrest during adolescence is associated with a reduced likelihood of obtaining a high school degree, and also increases the likelihood of being on welfare. Moreover, Gary Sweeten’s (2006) study using the National Longitudinal Survey of Youth (NLSY97) indicates that a first-time arrest with a court appearance has a deleterious effect on high school graduation compared to an arrest without a court appearance, which has only a minimal effect.

Christopher Uggen and Robert Stewart (2015) explain how system involvement “piles on” the collateral consequences associated with a deviant label by explicitly blocking educational opportunities. With admission forms increasingly requiring applicants to disclose their criminal past, the likelihood of being disqualified from enrollment has never been greater. Moreover, scholars also postulate that such increases in the surveillance of criminal records result in individuals avoiding civic institutions like post-secondary education (Brayne 2014). Consequently, system involvement appears to undermine educational attainment, which further damages employment opportunities given such a strong association between education and occupational attainment (Grubb 1999), and that post-secondary education is increasingly a requirement among employee prospects (Center on Education and the Workforce 2013).

A study by Michael Massoglia, Glenn Firebaugh, and Cody Warner (2012) also examines the consequences of criminal justice involvement on achieved SES, but in the form of neighborhood attainment. Using data from the NLSY97, the authors find that while whites reside in better neighborhoods compared to blacks after release, whites actually experience the greatest loss in neighborhood attainment when compared to their residence prior to incarceration (Massoglia et al. 2012). These findings convey the phenomenon that those with “more to lose” are affected by their criminal past more than those not as advantaged. Moreover, these results are contrary to the presumed assumption that advantaged individuals may be able to avoid the system, and instead buttress claims by John Irwin (1985) that those new to the rabble class will experience an unavoidable loss in conventional sensibilities.

Selection and Variation in the Criminal Justice Research

A host of studies indicate detrimental effects of criminal justice involvement on numerous outcomes throughout the life course, though the issue remains as to whether these consequences are robust and not a product of predisposed selection into the system. That is, the observed relationship between system involvement and low-attained SES may be spurious considering the population most at risk for such experiences are of low SES prior to involvement in the system (Sampson and Loeffler 2010; Tonry 2011). Illustrations of this are in the effect that criminal justice involvement has on objective and psychological achievements in disadvantaged contexts, whereby those with limited life chances or where arrest is normalized notice no repercussion from their criminal mark (Bernburg and Krohn 2003; Hirschfield 2008). Moreover, Gary Western (2002) shows incarceration reduces wage growth by roughly 30%, using data from the NLSY, but also observes differences in wages prior to incarceration between individuals who experienced incarceration and those who do not. Similarly, Robert Apel and Gary Sweeten (2010) find minimal differences in achieved SES by incarceration when employing propensity score models and fixed effects regression, but acknowledge differences at the bivariate level in wages prior to incarceration, as well.

To summarize, the consequences of involvement in the CJS materialize in familial hardships, as well as damages to occupational, educational, and residential attainments. Also, the consequences on achieved socioeconomic outcomes can surface from the mark on one’s record and the blocked opportunities associated with exposure to the system (Bernburg and Krohn 2003; Lopes et al. 2012; Sampson and Laub 1997; Wiley, Slocum, and Esbensen 2013). Moreover, it is possible that those who are seen as least susceptible to be involved in the CJS may actually find themselves losing more than those not as advantaged.

THEORY

System Involvement and the Life Course

Robert Sampson and John Laub (1993, 2005) identify employment as a turning point in the criminal career, thus creating the opportunity to “knife off” a criminal past and progress in the desistance process (Crutchfield and Pitchford 1997; Uggen 1999, 2000; Wadsworth 2006). Prior research also posits education as a potential turning point (Blomberg et al. 2011), with even greater effects of education on desistance among those who are especially delinquent (Ford and Schroeder 2011). And while the life course perspective adequately recognizes variation in the consequences of criminal justice involvement, the potential moderating effects of ascribed SES further complicate the system involvement-achieved SES relationship.

On one hand, involvement with the CJS may have no bearing on later socioeconomic attainments for those from lower social class as the likelihood for potential turning points is already improbable. This scarcity of turning points derives from the cumulative process of disadvantage unraveling over the life course, where Robert Sampson and John Laub (1997) believe a delinquent past (i.e., a criminal record) for those of low social standing leaves individuals with minimal options other than a life of crime. That is, a disadvantaged socioeconomic background increases the potential for involvement in the system, bolsters the challenges associated with a deviant label, and further perpetuates the inability to acquire pro-social bonds (e.g., attained education or occupation). Moreover, system involvement to the point of incarceration damages human capital as one’s ability to develop skills for future socioeconomic endeavors (through means of work experience and education) are delayed during incapacitation.

Conversely, Jeffrey Reiman and Paul Leighton’s (2010) perception of cumulative advantage in the CJS, whereby those deemed economically advantaged are weaned out of the system in early stages, suggests that socioeconomically privileged individuals are less prone to deal with collateral consequences (Savelsberg 1992). However, by exposing these individuals to the correctional system and stigmatizing them to the point of becoming vulnerable to such consequences, it is possible those of higher socioeconomic backgrounds who get involved in the CJS may find the rabble treatment especially daunting. Michael Massoglia and associates (2012) make similar speculations, specifically regarding whether the result of “losing more” after incarceration is present among those convicted of a crime who do not experience incarceration. The abrupt removal of individuals from society via incarceration suggests an immediate loss in status upon entering a correctional facility, but earlier stages in the system do not pose such drastic consequences.

Rabble Preparation

The most plausible explanation for potential consequences at earlier stages of the CJS acknowledges the effects of social mobility, whereby one’s attained social class differs from their origin class (either upwardly or downwardly). As Pitirim Sorokin (1927, 1959) argues, social mobility (regardless of direction upward or downward) leads to the exposure of new class norms as well as a violation of previous norms, thus increasing stress and conflict between new and old norms. Moreover, scholars suggest that downward social mobility may be even more stressful due to the loss of social prestige and capital from such transitions (Houle 2011; Newman 1988).

The problem of downward mobility due to criminal justice involvement also parallels John Irwin’s (1985) early intuitions of the justice system preparing individuals for the rabble life through psychological, cultural, and social transformations. The loss of conventional sensibilities and the acquisition of the rabble mentality (e.g., maintaining vigilance against figures of authority) conditions individuals to internalize an inferior position in society. And although the experience of involvement in the CJS is undoubtedly reflective of underprivileged circumstances, system involvement for those from poor backgrounds may reflect social stability. That is, the so-called “preparation for the rabble life” will likely be minimal for those from lower social standing who enter the CJS as the norms of the rabble class may closely resemble their ascribed social norms. Conversely, any exposure to the system for those from a higher social class will be both strikingly different from the norms of their ascribed class, and also indicative of downward social mobility. Moreover, these experiences might also result in a loss of human agency considering that fruitful socioeconomic endeavors across the life course are now obstructed by a criminal mark (Emirbayer and Mische 1998).

THE PRESENT STUDY

Drawing on theories of social mobility and the preparation for rabble life, this study considers several research questions. First, after taking into account preexisting characteristics related to both criminal justice involvement and SES attainment in adulthood, what is the relationship between involvement in the CJS and achieved SES? Furthermore, does it matter how deeply involved a person becomes in the system? For example, is there a difference between arrest and incarceration? We expect that involvement in the CJS will negatively affect achieved SES. However, given the consideration of the depth into the justice system with a dynamic measure ranging from no involvement to incarceration, we expect the relationship between depth of involvement and achieved SES to be nonlinear and reflective of an accumulating, accelerative decline on achieved SES as one penetrates deeper into the system.

Next, how does ascribed SES moderate the relationship between CJS involvement and achieved SES? For reasons outlined above, we hypothesize that the consequences of involvement in the CJS on achieved SES will be more detrimental for those coming from more advantaged backgrounds. Much of this expectation is attributed to the presumed damaging transition into the rabble class experienced by those of affluent backgrounds prior to criminal justice involvement. At the same time, it is plausible that ascribed SES will not influence the relationship between criminal justice involvement and achieved SES. That is, the null effects of system involvement by ascribed SES may be indicative of social stability with regard to the rabble transition, whereby moving into deeper into the system for those of low ascribed SES is not reflective of a change in class norms.

DATA AND MEASURES

Data

We use data from The National Longitudinal Study of Adolescent to Adult Health (Add Health). Add Health is a study of adolescents in the United States (grades 7–12 in 1994–95) followed into adulthood with the last data collection (Wave IV) in 2007–08, when respondents are largely between 25 and 32 years of age. Data were collected from adolescents and their peers, as well as their parents, siblings, and school administrators. Of the 20,745 respondents from Wave I, 14,738 (71%) were re-interviewed one year later (Wave II). Follow-up interviews for Wave III and Wave IV consist of 15,197 (73%) and 15,701 (76%) of the participants from Wave I, respectively.

For this study, we use data from Wave I, III, and IV, specifically data from 15,517 respondents who were not missing on our dependent variable. Missing data on the majority of our measures are minimal, with only 4% missing on family SES, and at most 1% missing on all other measures. Exceptions are in our measures of religious youth group participation at Wave I and involvement in the CJS, with roughly 14% and 18% missing, respectively. To ensure robustness of our findings, we address this issue in several ways. First, we implement listwise deletion, yielding a sample of 10,207, and then run our analyses. Next, we impute the mean and mode of respective continuous and dichotomous measures with missing data (which retains our sample of 15,517 respondents) and re-analyze our models. We also examine models employing involvement in the CJS at Wave IV to minimize missing data on our focal measure. In all cases, results yield similar findings to our presented results with minimal changes in coefficients (results available upon request). Thus, for the analyses shown here, we use multiple imputation via the SAS 9.3. PROC MI procedure to retain the most powerful sample of 15,517 respondents with the most precise imputation of missing data.

Dependent Variable

Achieved Socioeconomic Status (SES)

Ranging from 1 to 10, respondents’ SES (Achieved SES) at Wave IV is based on the sum of the respondent’s achieved education (ranging from 1 to 5) and occupational status (ranging from 0 to 5). Education is coded as 1 for those with less than a high school degree, 2 for those with a high school degree, 3 for those with some college, 4 for those who have completed a four-year degree, and 5 for those who completed at least some graduate school. For the occupational status measure, 0 is coded for individuals who are unemployed, while the range from 1 to 5 is based on the Standard Occupational Classification (SOC), which consists of a hierarchical list of hundreds of occupational statuses by the U.S. Bureau of Labor Statistics (For a detailed list of the occupational classification, see http://www.bls.gov/soc/).

The hierarchical construction of the occupation measure collapses these occupational statuses into 5 categories, ranging from material moving occupations up to managerial positions. This approach is guided by Carol Ford, Peter Bearman, and James Moody (1999), who constructed the original measure of family SES at Wave I, which is indicative of one’s family SES (Ascribed SES) used in these analyses. The only difference between the measure of ascribed SES in adolescence and respondents’ achieved SES in adulthood is that the family indicator from Wave I draws on the maximum value of either parents’ level of SES (i.e., each parents’ SES is calculated separately, with the highest value representing the respondent’s ascribed SES), which allows for a greater likelihood of higher ascribed statuses given the possibility of pooling the maximum score from two parents. Carol Ford and associates’ (1999) measure of ascribed SES has been widely used by scholars analyzing Add Health (e.g., Kuhl, Warner and Wilczak 2012; Swisher and Roettger 2012), and our identical operationalization of achieved SES in the present study further improves the sensitivity of our analyses as ascribed SES can be thought of as a stability coefficient, thus accounting for stable differences between respondents.

Independent Variables

Involvement in the Criminal Justice System (CJS)

Involvement in the CJS (System) is measured by creating dichotomous indicators reflecting the respondent’s experiences in the system. Indicators of system involvement are gathered from Wave III, when respondents are predominantly between the ages of 18 to 26 years, an approach consistent with other scholars examining involvement in the CJS using Add Health (e.g., Brayne 2014; Weaver and Lerman 2010). A measure indicating that a respondent had police contact is based on responses to the question, “How many times have you been stopped or detained by the police for questioning about your activities? Don’t count minor traffic violations”. Respondents who report this occurring once or more than once are coded as 1. Next, respondents who report being arrested are coded as 1 and are then asked about whether their first arrest, most recent arrest, or any other arrests ever led to a conviction in adult court. Respondents who report ever being convicted are coded as 1. Finally, those who report being convicted are asked about whether their first conviction, most recent conviction, or any other convictions ever led to a sentence of probation, jail, or prison. Respondents who report ever being sentenced to jail or prison are coded as 1 while probation is coded as 0. These variables (stopped, arrested, convicted, and incarcerated) are then used to create a continuous measure reflecting a count of the dichotomous indicators, ranging from 0 (uninvolved) to 4 (incarcerated). Multivariate analyses also include a quadratic term to capture any nonlinear effects (System2).

While some might argue that treating involvement in the CJS as a continuous measure as opposed to a series of dichotomous indicators is problematic due to its assumption of interval properties (i.e., equal unit impact for each successive increase in the measure), sensitivity analyses confirm the appropriateness of the approach employed in the present study. For instance, measures of discriminatory power are nearly identical between the unconstrained model (i.e., a model regressing achieved SES on a series of dichotomous measures indicating involvement in the CJS) and the quadratic model from the zero-order level up to the full model. Also, the unconstrained model reveals, as hypothesized, a series of increasingly negative estimates, which is consistent with the negative departure from linearity in the quadratic model (results available upon request). Thus, given the similarities and theoretical consistency in both approaches, a continuous measure of involvement in the CJS exhibits the most parsimonious approach.

Ascribed Socioeconomic Status

As previously discussed in the construction of the dependent measure, ascribed SES is created in a similar manner and based on the maximum value of the sum of either parents’ achieved education (ranging from 1 to 5) and the maximum value of their occupational status (ranging from 0 to 5). Specifically, the parent with the highest value in combined education and occupational status is used as the value of the respondent’s family SES.

Demographic Measures

Demographic controls (measured at Wave I) included in the multivariate analyses are a dichotomous indicator of sex (with males coded as 1 and females coded as 0), four measures of race, with black, Hispanic, and other race as the focus (reference being white), and a continuous measure of age (as well as a quadratic term).

Covariates Addressing Selection

Aside from ascribed SES, we include a host of controls suspected to render the hypothesized association between system involvement and achieved SES spurious. First, we include a dichotomous indicator of respondents living with their biological parents, which is measured at Wave I. We also include an indicator of involvement in delinquency, which is measured at Wave I based on nine self-reported indicators. The items are measured by asking respondents how often they did any of the following in the last 12 months: deliberately damage property that didn’t belong to you; steal something worth more than $50; go into a house or building to steal something; use or threaten to use a weapon to get something from someone; sell marijuana or other drugs; steal something worth less than $50; take part in a physical fight where a group of your friends was against another group; get into a serious fight; and hurt someone badly enough in a physical fight that he or she needed care from a doctor or nurse. Each of these items is coded a 0 if the event never happened, 1 if the event happened 1 or 2 times, 2 if the event happened 3 or 4 times, and 3 if the event happened 5 or more times. A sum of all nine responses is used to create the final delinquency variable. Given that each of the nine items ranges from 0 to 3, the final count variable has a possible minimum value of 0 and a possible maximum value of 27 (α = .80).

We include a 3-item measure of education-related self-control, previously employed by Kevin Beaver, Marie Ratchford, and Christopher Ferguson (2009). Three items ask respondents how often since school started this year (ranging from never to every day) they had trouble: getting along with your teachers; paying attention in school; and getting your homework done. The items are then averaged to create a scale in which higher values are indicative of low educational self-control (α = 0.68). To account for involvement in pro-social activities during adolescence, indicators of religious youth group participation (Wave I) are measured based on responses to the following: “Many churches, synagogues, and other places of worship have special activities for teenagers—such as youth groups, Bible classes, or choir. In the past 12 months, how often did you attend such youth activities?” Dichotomous measures indicative of participating in these activities at least once a week; once a month; or rarely are included in our multivariate analyses (referencing no participation in such activities). Moreover, we include a measure of delinquent peers at Wave I in a similar manner as John Wright and associates (2008). Respondents are asked how many of their three closest friends smoked at least one cigarette a day, drank alcohol at least once a month, and smoked pot at least once a month. A sum of all three responses (ranging from 0 to 9) is used to create this measure (α = .76).

Mediating Mechanisms

The relationship between involvement with the CJS and achieved SES might also be indirect, or mediated, due to several covariates. For instance, marriage is associated with higher SES [Current Population Survey (CPS 2015)], yet those involved in crime—or the CJS for that matter—are also among those least likely to transition into marriage (King, Massoglia, and Macmillan 2007). Thus, we include familial variables from Wave IV to account for such transitions, including: being married with and without children; cohabiting with and without children; and being single with and without children (referencing married with children).

Furthermore, from the perspective of social mobility and exposure to the rabble class, criminal justice involvement is expected to damage self-efficacy, a characteristic which is shown to also affect achieved SES (Phan 2012). Therefore, we include an indicator of one’s self-efficacy, which is measured at Wave IV by taking the average of a five-item Likert scale of the following four questions: in the last 30 days, how often have you felt that you were unable to control the important things in your life (reversed coded); how often have you felt confident in your ability to handle personal problems; how often have you felt that things were going your way; and how often have you felt that difficulties were piling up so high that you could not overcome them (reverse coded) (α = .72). We also include a measure of one’s perceived SES from Wave IV in the form of the MacArthur scale, which encompasses a 10-step ladder (with 1 being low SES and 10 being high SES), where respondents place themselves on a step depending on where they believe they fall in comparison to others (Giatti et al. 2012). Table 1 provides descriptive statistics for all measures in the study.

Table 1.

Descriptive Statistics

Variable	Mean	Std. Dev.	Min.	Max
Dependent Variable
Achieved SES	5.90	2.09	1	10
Independent Variables
System	0.37	0.83	0	4
System2	0.84	2.40	0	16
Ascribed SES	6.18	2.63	1	10
Additional Covariates
Male	0.47	-	0	1
White	0.55	-	0	1
Black	0.22	-	0	1
Hispanic	0.16	-	0	1
Other Race	0.08	-	0	1
Age	29.11	1.75	25	34
Age2	850.25	101.61	625	1156
Lives w/ Bio. Parents	0.54	-	0	1
Delinquency	1.86	3.05	0	27
Low Self-Control	1.09	0.79	0	4
No Youth Group Participation	0.44	-	0	1
Youth Group Weekly	0.25	-	0	1
Youth Group Monthly	0.16	-	0	1
Youth Group Rarely	0.15	-	0	1
Delinquent Peers	2.52	2.64	0	9
Married w/ Kid	0.29	-	0	1
Married No Kid	0.11	-	0	1
Cohab. w/ Kid	0.08	-	0	1
Cohab. No Kid	0.10	-	0	1
Single w/ Kid	0.10	-	0	1
Single No Kid	0.32	-	0	1
Efficacy	3.79	0.74	1	5
Perceived SES	5.02	1.74	1	10
N: 15,517

PLAN OF ANALYSIS

Our analyses are presented in the following manner. First, we examine the bivariate associations between all covariates of interest and deepest involvement in the CJS. We then analyze the relationship between system involvement and achieved SES in our multivariate analyses through a series of six sequential models. We begin with the zero-order model followed by the inclusion of demographic controls. Next we control for ascribed SES (as well as residing with biological parents at Wave I) to account for stable differences between respondents and also measure how much of the system involvement-achieved SES relationship is explained by these measures. Then we include our measures geared towards accounting for selection, followed by the remaining covariates that plausibly mediate the focal relationship. After controlling for all of these factors, we examine how ascribed SES moderates the relationship between system involvement and achieved SES. Finally, we implement inverse probability of treatment weights (IPTW) to our model examining interactions to account further for confounding bias and test the robustness of our findings.

RESULTS

Bivariate Relationships

Table 2 presents a description of respondents’ characteristics partitioned by their deepest involvement in the CJS. For instance, those who experience incarceration report the lowest average achieved SES (4.07) and also come from families with the lowest average SES in the sample (5.02). Additionally, 89% of those incarcerated are male while roughly 50% are either black or Hispanic. Finally, as expected, those incarcerated report the highest average levels of delinquency (5.94), delinquent peers (4.84), low self-control (1.63), while also reporting the lowest average of efficacy (3.67) and perceived SES (4.38) compared to respondents experiencing less or no involvement with the CJS.

Table 2.

Average ‘Profile’ of Respondent involved in the CJS

Variable	Uninvolved (11,781)	Stopped (2,083)	Arrested (1,033)	Convicted (510)	Incarcerated (110)
Achieved SES	6.02	5.85	5.25	5.01	4.07
Ascribed SES	6.16	6.34	6.24	6.13	5.02
Male	0.40	0.61	0.75	0.82	0.89
White	0.54	0.56	0.54	0.67	0.46
Black	0.22	0.20	0.22	0.15	0.25
Hispanic	0.16	0.16	0.16	0.13	0.23
Other Race	0.08	0.07	0.07	0.06	0.05
Age	29.14	29.94	29.84	29.28	29.51
Lives w/ Bio. Parents	0.55	0.54	0.45	0.53	0.43
Delinquency	1.45	2.32	3.74	4.54	5.94
Low Self-Control	1.03	1.19	1.40	1.54	1.63
No youth Group Participation	0.42	0.46	0.51	0.53	0.53
Youth Group Weekly	0.26	0.23	0.18	0.15	0.15
Youth Group Monthly	0.17	0.16	0.16	0.16	0.17
Youth Group Rarely	0.15	0.15	0.16	0.17	0.14
Delinquent Peers	2.34	2.60	3.45	4.01	4.84
Married w/Kid	0.31	0.23	0.18	0.22	0.24
Married No Kid	0.12	0.10	0.08	0.07	0.06
Cohab. w/ Kid	0.08	0.08	0.11	0.12	0.16
Cohab. No Kid	0.09	0.14	0.12	0.13	0.05
Single w/ Kid	0.11	0.07	0.06	0.05	0.03
Single No Kid	0.30	0.38	0.44	0.42	0.45
Efficacy	3.81	3.75	3.74	3.68	3.67
Perceived SES	5.06	5.03	4.82	4.71	4.38

We perform an analysis of variance (ANOVA), using Tukey’s multiple comparison of means, to examine the relationships between ascribed SES, achieved SES, and CJS involvement. Our results reveal that, compared to respondents with no prior involvement in the CJS, being arrested, convicted, and incarcerated is associated with 0.77, 1.01, and 1.95 points lower on levels of achieved SES, respectively (p < .05). Moreover, those experiencing incarceration are from families that are roughly one point lower on the ascribed SES measure than respondents with no involvement (p < .05). This also implies that respondents involved with the CJS only at earlier stages are from higher ascribed socioeconomic backgrounds, a finding that is consistent with our hypotheses.

Multivariate Models

Given the normal distribution of the dependent variable (skew = .09), ordinary least squares (OLS) regression appears to be sufficient for the multivariate analyses. However, to ensure the most sensitive significant tests in the models, we first employ the Breusch-Pagan test (Breusch and Pagan 1979) that considers the null hypothesis that the errors in the OLS regressions are homoscedastic. After regressing the squared residuals from each respective OLS model on the independent variables, a significant F-test for all of the models leads to rejecting the null hypothesis and concluding that heteroscedasticity is present and the standard errors are underestimated (resulting in less sensitive significant tests). In order to account for these issues, we use weighted least squares (WLS) regression by constructing weights from the fitted values after regressing the log-transformed (squared) residuals on each respective model (see DeMaris 2004 for a detailed description of ensuring positive weights).

Table 3 shows the results from the six nested models, starting with Model 1 in which achieved SES is regressed on the linear and quadratic measures of involvement in the CJS. As indicated in the bivariate analyses, the effect of involvement in the justice system is nonlinear and in the expected direction. We interpret the criminal justice involvement measure with the partial derivate in mind, β₁ + 2β₂X, where β₁ is the linear estimate and β₂ is the departure from linearity. And while the linear component is not significant, the departure from linearity is highly significant (p < .001) and negative, indicative of a segment I curve (DeMaris 2004). The pattern persists when controlling for demographic factors in Model 2. Although again not significant, it is worth noting that the linear term associated with involvement in the CJS increases by roughly 44% (from .05 to .09). This is especially important given the interpretation of system involvement via the partial derivative. That is, an increasingly positive linear coefficient creates a more gradual curve in the partial derivative of system involvement, which we attribute to our model accounting for at least some of the selection into the CJS with demographic controls. Figure 1 displays the nonlinear trajectory of involvement in the CJS, net of all covariates in Model 2.

Table 3.

WLS Regression of Achieved SES on Criminal Justice Involvement^‡

Variable	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
	Estimate	Estimate	Estimate	Estimate	Estimate	Estimate
System	.05(.10)	.09(.06)	−.02(.06)	.08(.06)	.05(.06)	.06(.06)
System2	−.14(.03)***	−.12(.02)***	−.07(.02)***	−.07(.02)***	−.06(.02)**	−.07(.02)***
Male		−.67(.04)***	−.73(.03)***	−.68(.03)***	−.83(.03)***	−.83(.03)***
Black		−.29(.04)***	−.10(.04)*	−.18(.04)***	−.03(.04)	−.03(.04)
Hispanic		−.35(.04)***	.12(.04)**	.07(.04)†	.04(.04)	.04(.04)
Other Race		.41(.07)***	.34(.06)***	.28(.06)***	.22(.06)***	.22(.06)***
Agec		−.02(.0l)*	.0l(.0l)	.04(.0l)***	.03(.0l)**	.03(.0l)***
Age2c		−.03(.0l)***	−.02(<.0l)***	−.03(.0l)***	−.02(<.0l)***	−.02(<.0l)***
Live w/ Bio. Parents			.4l(.03)***	.33(.03)***	.2l(.03)***	.21(.03)***
Ascribed SESc			.25(.0l)***	.25(.0l)***	.19(.0l)***	.19(.0l)***
Delinquency				−.03(.0l)***	−.02(.0l)***	−.02(.0l)***
Low Self-Control				−.20(.02)***	−.13(.02)***	−.14(.02)***
Youth Group Weekly				<.01(.05)	.02(.05)	.02(.04)
Youth Group Monthly				.04(.05)	.05(.04)	.04(.04)
Youth Group Rarely				.08(.05)†	.09(.04)*	.09(.04)*
Delinquent Peers				−.08(.0l)***	−.05(.0l)***	−.05(.0l)***
Married No Kid					.72(.05)***	.72(.05)***
Cohab. w/ Kid					−.28(.05)***	−.29(.05)***
Cohab. No Kid					.43(.05)***	.41(.05)***
Single w/ Kid					−.14(.05)**	−.14(.05)**
Single No Kid					.33(.04)***	.33(.04)***
Efficacy					.13(.02)***	.13(.02)***
Perceived SESc					.26(.0l)***	.26(.0l)***
SES*System						.07(.02)**
SES*System2						−.02(.0l)**
Intercept	6.00(.02)***	6.45(.03)***	6.12(.03)***	6.62(.05)***	5.85(.09)***	5.86(.09)***
${\text{R}}_{\text{WLS}}^2$	0.02	0.06	0.18	0.20	0.28	0.28
						N:15,517

^†p < .1,

^*p < .05,

^**p < .01,

^***p < .001

^cCentered

^‡Standard Errors in Parentheses

Figure 1.

Partial Derivative of Involvement in the CJS on Achieved SES

As expected in Model 3, ascribed SES positively affects achieved SES (p < .001). Moreover, the departure from linearity in system involvement is reduced from −.12 to −.07 (a reduction of roughly 42%), although the coefficient itself remains significant. This suggests that some of the apparent effect of involvement in the CJS on achieved SES is spurious. For example, having higher ascribed SES both reduces the likelihood of involvement in the system while also leading to higher relative achieved SES in later life.

Model 4 includes the measures further addressing issues of selection, and although delinquency, low self-control, and delinquent peers have a significantly negative association with achieved SES, the quadratic term associated with involvement in the justice system remains strong and negative. Model 5 then includes of all the remaining covariates presumed to mediate the system involvement-achieved SES relationship, and again, the significantly negative departure from linearity of involvement in the CJS persists, although again slightly attenuated. The final model, Model 6, adds interaction terms of ascribed SES with the linear and quadratic indicators of involvement in the CJS. As expected, the deleterious effects of being involved in the CJS on achieved SES are greater for those coming from higher socioeconomic backgrounds based on the negative coefficient associated with the interaction term. Targeted centering is applied to test the association between system involvement and varying levels of ascribed SES (i.e., average ascribed SES, as well as one standard deviation above and below the mean). Results shown in Figure 2 suggest that those coming from high socioeconomic backgrounds experience the greatest negative consequences on achieved SES while the ramifications are less severe for those from more disadvantaged backgrounds. Furthermore, the effect of criminal justice involvement for those from below average ascribed SES is not significant, providing support for the null effects of social stability in the rabble transition.

Figure 2.

Partial Derivative of CJS Involvement on Achieved SES by Ascribed Status

Aside from examining the association between involvement in the CJS and achieved SES at multiple ascribed levels, one might presume that those with “more to lose” will experience the greatest loss but still achieve more than their lower SES counterparts. To investigate these inquiries, Figure 3 displays the predicted level of achieved SES by each targeted level of ascribed SES. Predicted levels of achieved SES are based on models stratified by below average, average, and above average ascribed SES, controlling for all covariates from Model 5 in Table 3. The criminal justice measure is now estimated with the unit impact in mind rather than the partial derivative, such that the effect of involvement in the CJS is calculated as: β₁ + β₂ + 2β ₂x, where β₁ is the linear estimate and β₂ is the departure from linearity. As expected, those coming from the highest socioeconomic backgrounds are also predicted to achieve the highest SES, particularly at minimal involvement in the CJS. However, as all groups get further into the CJS, the levels of achieved SES begin to converge, such that by the time one experiences incarceration, the predicted achieved SES are similar across all levels of ascribed SES.

Figure 3.

Predicted Achieved SES by Involvement in the CJS and Ascribed SES

Sensitivity Analyses

To further test the robustness of our findings, we implement a second set of analyses with an even greater emphasis on selection into the CJS. Specifically, we employ analyses utilizing the inverse probability of treatment weights (IPTW) (Lanehart et al. 2012). As means for a dichotomous “treatment,” we collapse our system involvement measure to indicate those who experience either a conviction or incarceration, referencing any system involvement equal to or below experiencing an arrest (which we call, mark on record). We then predict the probability of experiencing at least a conviction via logistic regression, with all covariates from Model 4 in Table 3, as these are the measures presumed to render the hypothesized relationship between system involvement and achieved SES spurious. Models predicting achieved SES are then re-analyzed with the dichotomous measure of experiencing at least a conviction substituted for our continuous measure of system involvement (Table 4), with respondents weighted on the inverse of the probability of being convicted or incarcerated. A common check for balance in the covariates when employing IPTW is a comparison of the standardized differences in the observed means (or proportions for dichotomous measures) for the treated and control groups (see Austin 2011), and mean differences in these analyses are well below the generally accepted threshold of .25 (Vaughan et al. 2015), with all but two differences (i.e., white and other race) at or below the more sensitive threshold of .10 (Austin 2011).

Table 4.

Achieved SES and System Mark via Inverse Probability of Treatment Weights (IPTW)^‡

Variable	Model 1	Model 2
	Estimate	Estimate
Mark on Record	−.43	−.43
	(.03)*	(.03)*
Ascribed SESc	.23	.26
	(.01)*	(.01)*
Mark*Ascribed SESc		−.06
		(.01)*
Intercept	6.67	6.69
	(.05)*	(.05)*
${\text{R}}_{\text{adj.}}^2$	0.22	0.22
		N: 15,517

^*p < .001

Note: Models include controls from Model 4 in Table 3

^cCentered

^‡Standard Errors in Parentheses

As expected in Model 1 of Table 4, the coefficient for ascribed SES remains positive and significant while having at least a conviction is negatively associated with achieved SES (p < .001). Moreover, the interaction in Model 2 is again significant and negative, suggesting a greater drop in achieved SES among those of higher ascribed SES. The results (depicted in Figure 4) are consistent with the analyses from Table 3, whereby those from the highest socioeconomic backgrounds experience the greatest loss in achieved SES, followed by those of average and one standard deviation below average ascribed SES, respectively.

Figure 4.

Effect of Conviction or Incarceration on Achieved SES by Ascribed SES

DISCUSSION

The present study reinforces research exposing the negative relationship between involvement in the CJS and achieved SES. Overall, our results show that deeper involvement in the system leads to increasingly negative socioeconomic achievements, net of factors presumed to confound the association as well as mediating mechanisms. Moreover, we find support for our hypothesis that criminal justice involvement among those with high ascribed SES promotes downward social mobility, which bolsters the consequences on later attainments. Furthermore, null effects for respondents with low ascribed SES are consistent with our expectation of such experiences reflecting social stability.

Our findings have several implications for research on the collateral consequences associated with the CJS. For example, by examining the system as a dynamic process as opposed to a single dichotomous indicator, we find the deleterious effects associated with system involvement occur as early as arrest and increase exponentially as one penetrates deeper into the system. Moreover, the effect of system involvement is robust even after accounting for a host of other selection factors as well as analyses further accounting for confounding bias through inverse probability weights.

This study also considers what role ascribed SES plays in the relationship between system involvement and achieved SES. We find that ascribed SES is associated with minimal involvement in the system and also associated with higher achieved SES, evidence which is consistent with our hypothesis. Both John Irwin (1985) and Jeffrey Reiman and Paul Leighton (2010) posit that individuals from affluent backgrounds are least likely to experience deeper involvement in the system, as their resources work to reduce the likelihood of being pulled into the deepest stages of the system (i.e., incarceration). For example, such “protection” could be in the form of superior legal representation compared to those of low ascribed SES that minimizes any chance of being incarcerated. Similarly, factors associated with involvement in the CJS—e.g., more delinquency, delinquent peers, low self-control—are also most pronounced among those with lower ascribed SES. Thus, the extent to which ascribed SES protects individuals from system involvement may simply be a product of other factors shown to be correlated with involvement in the system, thus reducing the effect of ascribed SES. Yet, our analyses do not show support for such claims, as the effect of ascribed SES remains nearly identical after controlling for delinquency, low self-control, pro-social behaviors, and delinquent peers (see Model 4 in Table 3). Moreover, the effect of system involvement is also not attenuated from the inclusion of these factors, rather it appears that ascribed SES is the dominant factor confounding the relationship between the CJS and achieved SES. To further validate such findings, we ran analyses controlling for delinquency, low self-control, pro-social behaviors, and delinquent peers without ascribed SES in the model, only to find that changes in the estimates associated with system involvement were minute compared to changes after solely introducing ascribed SES in the analyses.

Such findings motivate our investigation into how ascribed SES moderates the relationship between system involvement and achieved SES. Specifically, our results show that the effect of involvement in the CJS varies in magnitude by one’s socioeconomic background. Those from high SES backgrounds experience the greatest decrease in achieved SES after involvement in the CJS, followed by individuals from average and low SES backgrounds, respectively. These findings replicate a similar pattern that prior research has shown regarding neighborhood attainment post-incarceration (e.g., Massoglia et al. 2012), and our results also show that the phenomenon of having “more to lose” occurs well before individuals are incarcerated. Thus, particularly for those coming from high ascribed SES, involvement in the CJS appears to destroy human capital, pull everyone down to a poor economic standing, and ultimately transitions all released offenders into the rabble.

Yet, while predicted levels of achieved SES from the multivariate analyses convey a convergence in achievements (See Figure 3), whereby those who experience incarceration exhibit similar levels of achieved SES regardless of ascribed SES, we interpret the predicted achieved SES values cautiously. Instead, the observed levels of achieved SES in these data are similar to the findings by Michael Massoglia and associates (2012), as those from high ascribed SES who experience incarceration still achieve roughly .8 points higher on the dependent measure compared to those from average ascribed SES, and over 1 point higher than those from low ascribed SES (results available upon request).

Considering the findings from the perspective of having “more to lose,” Pitirim Sorokin’s (1927, 1959) ideas of social mobility best explain this phenomenon, but coupling these explanations with John Irwin’s (1985) ideas of the rabble transition is an even stronger claim. Criminal justice involvement for those of affluent backgrounds may exemplify a change in norms, and likely be indicative of a more stressful experience as individuals are, “prepared for the rabble class” (Irwin 1985). It is also plausible that the experience of getting involved in the CJS is seen as a greater disappointment for those coming from high socioeconomic backgrounds, which in turn increases the stress associated with system involvement as well as the internalization of a criminal label (Lemert 1951). Yet, while we include self-efficacy and perceived SES as controls expected to mediate the system involvement-achieved SES relationship, no such finding is shown here. Thus, future research would benefit greatly from understanding what other social-psychological consequences stem from criminal justice involvement as means for further validity in the rabble transition.

We ensure the robustness of our findings with several sensitivity analyses. For instance, we use data for system involvement from both Wave III and Wave IV of Add Health to examine changes in involvement in the CJS across the life course, and find results to be similar to our findings shown here That is, deeper involvement in the CJS between waves is negatively associated with achieved SES, with effects stronger among respondents from higher ascribed SES. Moreover, we examine interactions with ascribed SES and criminal justice involvement while stratifying by race, and results are consistent with our presented findings as well as prior research (e.g., Massoglia et al. 2012). Specifically, significant interactions between ascribed SES and criminal justice involvement are present for whites and blacks, but the deleterious effects of the system are significant for those from average or high ascribed SES among whites and only significant among blacks from high ascribed SES (with non-significant effects of system involvement for blacks from average and low ascribed SES, respectively). We attribute this to the established racial inequalities that exist with involvement in the CJS (e.g., Western 2006). Thus, it is conceivable that blacks from average ascribed SES are more exposed to the rabble class compared to whites from averaged ascribed SES. As such, blacks likely do not experience a significant transition to the rabble class unless they are from exceptionally advantaged SES.

As shown, we employ sensitivity analyses utilizing regression models weighting for the inverse probability of selection into deeper stages in the CJS, and again find similar results. Yet, this should further motivate research to continue the investigation into what unobserved factors contribute to the “more to lose” hypothesis. Specifically, what is it about those from affluent backgrounds who prevail against the odds and still get involved in the CJS? Our results show protective effects of coming from an advantaged social standing on minimizing system involvement, but are those who still find a way to get involved such a select group of individuals that even their socioeconomic advantage cannot protect them from these collateral consequences? Thus, is this pattern of “more to lose” that scholars consistently support simply a product of individuals most destined for the CJS who just happen to be from a high social class? Our results exhibit how detrimental criminal justice involvement can be on one’s future, but these questions are valuable considerations for future studies.

LIMITATIONS

As with all studies, our analyses are not without research limitations. First, one of the major limitations of Add Health is that it is a school-based sample, whereby those who are most at-risk to engage in criminal behavior (i.e., get involved in the CJS) are likely not present in the analyses as they may have dropped out of school. Next, achieved SES only captures respondents’ occupation and education, whereas ascribed SES is comprised of the maximum value from the respondent’s parent(s). The measure of achieved SES would benefit from incorporating a potential partner’s occupation and education to gain a better understanding of how the CJS affects later attainments.

Also, the present analyses do not consider the type of crime that an individual commits among those convicted; the focus here is solely how far one gets into the CJS. With that in mind, future research would greatly benefit from considering how ascribed SES moderates the system involvement-achieved SES relationship depending on the type of crime one was committing with more appropriate data capable of addressing these inquiries (we examine such relationship among those convicted of a crime in these data and find no significant interaction). Similarly, these analyses focus solely on a conviction for something other than a traffic violation, which means misdemeanors are weighted the same as a felony. However, given the significant findings present in this study, it is likely that the separate effects (particularly with felonies) would yield an even stronger relationship. Additionally, for those who experience incarceration, an analysis of how long one was incarcerated would enhance these analyses as the length of time spent in a correctional institution can certainly render consequences on achieved SES, particularly through means of destroying human capital and delaying educational and occupation progression. And while we treat a sentence to probation as ceasing one’s involvement in the system at “conviction,” research should further examine this sentence as increased supervision in the form of probation can inevitability affect later involvement in the system (e.g., Blumstein and Beck 2005; Lin, Grattet, and Petersilia 2010). Finally, we do not parse out the number of police contacts, arrests, or convictions occurring prior to incarceration (i.e., multiple experiences in the early stages of the CJS will likely increase the probability of an incarceration sentence). Yet, we believe that factors such as this are beyond the scope of these data. For instance, of the 110 respondents who receive a sentence of incarceration, 101 are incarcerated after their first conviction while the remaining 9 receive an incarceration sentence for a subsequent conviction. Taken as a whole, similar analyses examining these specific criminal justice factors would be more appropriate with a sample of incarcerated respondents.

CONCLUSION

Consequences of involvement in the CJS are so extensive that these issues are nothing short of a societal problem. Our findings show that downward social mobility stemming from involvement in the CJS occurs as early as arrest and exponentially increases as one gets deeper into the system. Furthermore, results show that even those who are seen as “protected” from the system due to socioeconomic advantages from their family ultimately attain lower levels of SES upon criminal justice involvement. Aside from the explicit negative consequences from system involvement, regardless of one’s ascribed social class, these findings exacerbate the issue of inequality in society. The null effects of system involvement for those from low socioeconomic backgrounds further validates just how substantial the gap between social classes has grown. That is, to say that some individuals are no worse off after involvement in the CJS in achieved SES is frightening given the well-established unfavorable consequences of the system. And to find that exposure to the system pulls even those with greater socioeconomic resources down to the rabble further exemplifies how policy needs to address its correctional practices. It is hoped that these findings will continue to motivate change in this era of mass incarceration.