Parenthood Residency Status and Criminal Desistance across Neighborhood Contexts

Abstract

Research on desistance and parenthood has yielded mixed results bringing into question whether, and for whom, parenthood matters. Scholarship has not fully explored the importance of residency status or patterns of desistance across a full range of neighborhood contexts, nor examined distinctions between temporary and long-term desistance. Our study uses data from the National Longitudinal Study of Adolescent to Adult Health (Add Health) to examine the association between parenthood residence and criminal desistance across levels of adolescent neighborhood poverty. Our findings demonstrate that parenthood has different meanings for desistance, depending on its duration, residency status, and neighborhood context.

Introduction

Over the last twenty-five years scholars have focused greater attention on identifying life events and transitions in adulthood that may promote desistance (Farrington 2003). Sampson and Laub’s age-graded theory of informal social control (Laub and Sampson 1993, 2003; Sampson and Laub 1990, 1993), rooted in Elder’s (Elder Jr., Johnson, and Crosnoe 2003; Elder Jr. 1994) life course perspective and Hirschi’s (1969) social bond theory, has perhaps been most influential in guiding research on the factors that lead to criminal desistance in adulthood. A number of key life events have been identified as potential turning points in the process of desistance, including cohabitation and marriage (Forrest 2014; King, Massoglia, and Macmillan 2007; Sampson, Laub, and Wimer 2006), employment (Cernkovich and Giordano 2001; Horney, Osgood, and Marshall 1995; Piquero et al. 2002; Uggen 2000), and military service (Craig and Foster 2013; Sampson and Laub 1996).

Research focusing on the relationship between parenthood and desistance has received more limited attention and has yielded mixed empirical results. Although a number of studies have found that becoming a parent has little or no effect on criminal desistance (Blokland and Nieuwbeerta 2005; Giordano, Cernkovich, and Rudolph 2002; Mulvey and Aber 1988; Rand 1987), more recent research has found that becoming a parent promotes criminal desistance for particular subgroups, such as those who transition into parenthood in the context of a union (Savolainen 2009). A relatively large literature has focused on the link between parenthood and substance use with equally mixed findings. In this case, while some research finds that parenthood leads to reductions in substance use, especially when the parent resides with the child (Staff et al. 2010), other research (Thompson and Petrovic 2009) finds that having children is associated with an increase in use of illicit drugs. This mixture of findings is likely due to differences across selective samples (Giordano et al. 2002; Kreager, Matsueda, and Erosheva 2010; Savolainen 2009; Uggen 2000; Uggen and Kruttschnitt 1998), cross-sectional design (Giordano et al. 2002; Warr 1998), and variation in the measurement of both parenthood and criminal desistance.

These inconsistencies suggest a need for more research on whether, for whom, and for how long, parenthood is associated with desistance. Prior research has not fully explored, for example, whether parenthood produces enduring desistance, or whether its influence is only temporary. Prior research measures desistance through decreases in offending (Giordano et al. 2011; Kreager et al. 2010), changes in offending (Craig 2015), or the absence of offending at a single later period (Blokland and Nieuwbeerta 2005; Uggen 2000; Uggen and Kruttschnitt 1998). While useful in some respects, measuring desistance in these ways does not fully capture the temporality of desistance across the life course, or the categorical or discontinuous nature of complete desistance. A consideration of temporary versus long-term desistance may be particularly important in the case of parenting, as its demands change considerably over time.

Further, research on parenthood and criminal desistance also has not fully examined potential differences in the role of parenting across neighborhood contexts, with many studies focusing on the experiences of those within the most disadvantaged neighborhoods (Kreager et al. 2010). Variations across community contexts may be important with regard to the meaning of resident versus nonresident parenthood for desistance. For example, research strongly suggests that the neighborhoods in which one grows up vary in terms of the cultural heterogeneity of norms regarding the meaning of teenage sex, pregnancy, and relationship scripts (Harding 2007; Warner et al. 2011). These scripts likely also include norms and expectations regarding the roles of parents, both resident and nonresident, and hence may have implications of desistance. From a life course perspective, it is important to consider the timing and duration of neighborhood experiences (Wodtke, Harding, and Elwert 2011), including the neighborhoods in which one grew up and those in which respondents are currently living.

This study uses data from the National Longitudinal Study of Adolescent to Adult Health (Add Health) to examine the relationship between residency status of parents and desistance. Our measurement of desistance improves on prior studies by examining differences in temporary versus long-term desistance, thus providing a more complete understanding of the relationship between parenting, residency status, and desistance. Further, our study examines how neighborhood poverty contexts may moderate these associations.

Parenthood and criminal desistance

The relationship between parenthood and crime is theorized primarily through a social control perspective (Hirschi 1969; Laub and Sampson 1993, 2003; Sampson and Laub 1990, 1993). Reductions in crime from this perspective can stem from the strength or quality of the bond that the parent develops with his/her child; the stronger/deeper the bond, the greater the stake in conformity, which leads to an expectation of reduced or ceased involvement in crime. Parenthood also creates daily structure, provides meaning, and creates a change in identity that keeps an individual from criminal involvement. From a routine activities perspective (Cohen and Felson 1979; Osgood et al. 1996) as well as a social learning perspective (Warr 1998), parenthood reduces the amount of unstructured socializing with deviant peers that single persons may be more likely to have, and increases time at home due to new status obligations. Additionally, symbolic interactionist perspectives (Giordano et al. 2002; Maruna 2001) propose that parenthood may be linked to desistance when it produces changes in identity that disassociate parents from criminal involvement. From this perspective, parenthood produces a new identity that provides a new sense of purpose, a sort of master status that replaces the former criminal identity. Although these theories propose that parenthood should lead to reductions in criminal activity, empirical evidence regarding this relationship has been mixed.

Many previous studies have found null results, implying that parenthood does not reduce criminal behavior. In a study on marijuana use, Warr (1998) found that having children added nothing to the beneficial effect of marriage on time spent with delinquent friends. Laub and Sampson (2003) likewise found no added benefit of children for the Glueck men when it came to their criminal desistance. Other studies involving more serious offenses have also found no effect for parenthood (Blokland and Nieuwbeerta 2005). Giordano and colleagues (2002) found that attachment to children did not predict criminal involvement for those formerly incarcerated in Ohio, although some parents in the qualitative interviews described having children as a “hook for change” in desisting from crime.

Other research has found that the relationship between parenthood and crime varies across subgroups of the population. For example, Uggen and Kruttschnitt (1998) found that parenthood decreased mothers’ illegal earning activity, but it did not have an effect on fathers’, or either parent’s, arrest (for gender differences, see also Staff et al. 2010; Thompson and Petrovic 2009). Giordano et al. (2011) found mixed results regarding the role of the wantedness of pregnancy. They found that parents from lower socioeconomic status backgrounds and mothers who wanted their child experienced a decrease in criminal behavior, but this was not the case for other groups. In Savolainen’s (2009) study of convicted felons in Finland, he found that those in a union with children experienced a decrease in being reconvicted. The mixed findings of these studies suggest that researchers still cannot say conclusively if, or for whom, parenthood reduces crime.

Whether parenthood is associated with desistance may be related to the residency status of parents. From both a routine activities and social control perspective, we may expect that those who reside with their children have daily lives structured so that they spend more time with their children, and on average have greater emotional attachment (compared to nonresidents), thereby resulting in a greater reduction in criminal behavior. The few previous studies that exist on residency and desistance seem to support this expectation. For example, using data from Monitoring the Future, Staff et al. (2010) found that resident mothers experienced the largest reductions in substance use, but that parenthood was associated with desistance for all mothers regardless of co-residence. In contrast, fathers reduced substance use only when they lived with their child. Craig (2015) examined the relationship between parenthood residency and changes in criminal behavior using Add Health data and found that parents who reside with a biological child are more likely to experience decreases in criminal offending over time. This relationship was significant for white parents, but not for black or Hispanic parents.

Neighborhood contexts, relationship transitions, and crime

Associations between parenthood and criminal desistance could also be moderated by the neighborhood contexts in which individuals are raised. Research that examines neighborhoods over the life course has found that where an individual grows up may have long-term implications for a variety of outcomes (Galster et al. 2007). Adolescent neighborhood contexts have influence in the socialization of individuals through exposure to adult role models, institutions, and peer networks (Harding 2009; Jencks and Mayer 1990; Sampson and Wilson 1995; Wilson 1987), as well as exposure to resources and opportunities to experience transitions over the life course. This suggests that early neighborhood contexts may have long-term implications for social roles, such as parenthood, as well as other behavioral trajectories. Additionally, research has shown that individuals are likely to remain in the same neighborhood context of advantage or disadvantage over their life course, which illustrates the idea that individuals are “stuck in place” (Sharkey 2008). The respondents from Add Health have also been found to remain in similar neighborhood poverty contexts from adolescence through the transition into adulthood (Swisher, Kuhl, and Chavez 2013), providing further evidence that certain structural neighborhood contexts are relatively stable across developmental periods. This is not to say that adult neighborhood contexts are less meaningful in their own right, only that adolescents are less able to choose their neighborhoods and have more restrictions on social networks and role models outside of them, which perhaps implies a stronger link between youth neighborhood context and numerous behavioral outcomes than between adult neighborhoods and outcomes. In support of this contention, research on adult relationship transitions demonstrates a weak association between adult neighborhood socioeconomic context and outcomes. For example, South (2001) finds no direct association between neighborhood disadvantage and marital instability while South and Crowder (2000) identify a weaker effect of neighborhood socioeconomic status (SES) on the timing of first marriage for persons in their 30s compared to adolescents and younger adults.

Also with respect to relationship-related transitions and outcomes, prior research shows that individuals who are exposed to poverty in childhood have an increased likelihood of teenage pregnancy (Galster et al. 2007; Harding 2007; South and Baumer 2000; Wodtke 2013), and long-term exposure to childhood disadvantage increases the likelihood of experiencing nonmarital fertility (South and Crowder 2010). Recent research suggests that these differences reflect, in part, a greater heterogeneity of cultural norms among adolescents living in disadvantaged neighborhoods (Harding 2007). Harding (2007) finds that adolescents within disadvantaged neighborhoods express more heterogeneous views toward sex, romantic relationships, and pregnancy. Greater cultural heterogeneity, in turn, was found to loosen the connection between an individual’s own conceptions of relationship scripts and their actual behavior.

Later in life, these heterogeneous scripts, and a general lack of economic and social opportunities, are manifested in a lower likelihood of marriage in adulthood (Anderson 1999; Edin and Kefalas 2005; Edin and Nelson 2013). Nevertheless, research suggests that parenthood remains a meaningful marker of adult status across all neighborhood contexts. Ethnographic research has found that the transition into parenthood is transformative for both mothers and fathers in disadvantaged neighborhoods (Anderson 1999; Edin and Kefalas 2005; Edin and Nelson 2013). Individuals in disadvantaged neighborhoods are socialized from the time they are young to view parenthood as a crucial and attainable social role in their lives (Anderson 1999). Edin and Kefalas (2005) studied young single mothers in disadvantaged neighborhoods of Philadelphia and documented a positive transformative influence of parenthood, viewing motherhood as providing meaning, optimism and direction, and representing an essential part of their own identities.

Nonresident parenthood is also more common in disadvantaged contexts (Anderson 1999; Edin and Kefalas 2005; Edin and Nelson 2013), yet is still perceived as a life changing status. Edin and Nelson (2013) studied disadvantaged fathers in the city of Camden and found that fathers view the transition into fatherhood as a chance to improve themselves and their lives. However, the long-term transformative nature of fatherhood is diminished by the lack of opportunity for men who grow up in disadvantage to provide the full “respectability package” (Giordano et al. 2002) of marriage and economic stability. Regardless of these long-term changes, both resident and non-resident parenthood are viewed as meaningful in the context of disadvantage.

These differences across neighborhoods in the prevalence and potential meaning of parenthood, both resident and nonresident, suggest potential differences in the consequences of parenthood for desistance. Few studies of parenting and desistance have considered variations across neighborhood socioeconomic contexts, and those that have are limited to disadvantaged neighborhoods. For example, Kreager and colleagues (2010) found the transition to motherhood to be associated with decreased crime, alcohol, and marijuana use within disadvantaged areas of Denver, Colorado. As this was limited to a single city we do not know if this finding would generalize at the national level, nor whether it differs from the experiences of mothers in moderate-poverty or more advantaged communities.

Current study

In summary, the findings of previous research on the relationship between parenting and desistance are decidedly mixed, and limited by a variety of issues including: cross-sectional designs (Giordano et al. 2002; Graham and Bowling 1995; Warr 1998), geographically limited samples (Blokland and Nieuwbeerta 2005; Giordano, Cernkovich, and Rudolph 2002; Giordano et al. 2011; Kreager et al. 2010; Laub and Sampson 2003; Savolainen 2009), and selective samples of serious offenders (Blokland and Nieuwbeerta 2005; Giordano et al. 2002; Savolainen 2009; Uggen and Kruttschnitt 1998). Additionally, there are concerns regarding how to best conceptualize and measure parenthood and desistance. For example, studies have largely neglected to consider the resident status of parents (Craig 2015; Staff et al. 2010). The measurement of criminal desistance has also been somewhat problematic. For instance, there is a large focus on substance use in the literature (Staff et al. 2010; Thompson and Petrovic 2009; Warr 1998), and many studies have used formal measures of arrest or conviction which may miss involvement in crime that does not come to the attention of police (Blokland and Nieuwbeerta 2005; Savolainen 2009; Uggen and Kruttschnitt 1998). Most current desistance measures also fail to capture temporal nuances in desistance over time, such as the distinction between temporary and long-term desistance. Research that has examined the enduring effects of marriage has found that this particular transition has situational effects on criminal behavior, and that these effects tend to dissolve over time along with the unions (Bersani and Doherty 2013). It is not yet understood whether parenthood has a similarly fleeting relationship with crime.

The present study seeks to address these limitations by examining the relationship between parenthood residency status and criminal desistance. Using data from the National Longitudinal Study of Adolescent to Adult Health (Add Health), the current study contributes by comparing processes across levels of neighborhood poverty, using a nationally representative sample that includes both mothers and fathers, and by measuring desistance longitudinally to create classifications of both temporary and long-term cessation in criminal behavior.

In accordance with ethnographic research regarding parental residence (Anderson 1999; Edin and Kefalas 2005), as well as Sampson and Laub’s (1993) age-graded theory of informal social control and routine activities theory (Cohen and Felson 1979; Osgood et al. 1996), we expect that (H1) all parents will be more likely to temporarily desist from criminal behavior than nonparents, but that resident parents will be significantly more likely to experience long-term desistance from crime than nonresident parents. Transitioning to parenthood is transformative and increases bonds to conventional society, yet the routine activities of resident parents are likely to be more intensively affected. We also predict that resident parenthood will produce both temporary and long-term desistance for all individuals regardless of neighborhood poverty (H2), but that nonresident parenthood will be significantly related to temporary desistance for individuals from high-poverty neighborhoods especially (H3). As Edin’s scholarship (Edin and Kefalas 2005; Edin and Nelson 2013) illustrates, the portrait of two-resident parents is not common in neighborhoods of high disadvantage, and many fathers are nonresident. However, parenthood remains highly meaningful in impoverished contexts (Edin and Kefalas 2005). Thus, nonresident parenthood for individuals from neighborhoods of high poverty is likely to be related to criminal desistance due to the normative nature of nonresident parenthood and the meaning it provides for individuals from this particular neighborhood context. Nonresident parenthood, however, may not lead to long-term desistance due to lack of change in daily routines and the reduced potential for developing a strong bond between parent and child.

Method

Data and sample

The current research uses data from the National Longitudinal Study of Adolescent to Adult Health (Add Health). Add Health is a longitudinal study using a stratified, school-based, clustered sampling design to ensure the data are representative of the American adolescent school population (Harris et al. 2009). The first wave took place in the 1994–1995 school year. Roughly 90,000 students in grades 7 through 12 were surveyed, and from this group 20,745 were randomly chosen for in-home interviews of the adolescents and their parents. Wave two was conducted a year later in 1996. Wave III was conducted in 2002 (with respondents 18 to 26 years of age), and the most recent, fourth, wave took place in 2008 (ages 24 to 32). The fourth wave includes 15,701 of the original 20,745 respondents (75.7% retention rate). We use data from all four waves.

As our focus is criminal desistance, the sample is restricted to those involved in prior delinquent offending. The first and second interviews asked the respondents about their engagement in nonviolent and violent offending behaviors over the past 12 months, such as “sell[ing] marijuana or other drugs,” “driving a car without the owner’s permission,” “going into a house or building to steal something,” “you got into a physical fight,” “you shot or stabbed someone,” and “used or threatened to use a weapon to get something from someone.” Individuals who reported they had not engaged in any of these acts at wave I or II (and for whom, thus, desistance is not meaningful) were dropped, leaving 10,174 respondents who reported prior engagement in delinquency.

Our sample size is further reduced through listwise deletion of cases with missing data. Less than 1% of the sample (n = 84) is lost due to missing data on neighborhood poverty. We also drop cases missing on other covariates, losing 10 cases. One covariate, family socioeconomic status, is imputed using the ice command in Stata 13 to correct for the loss of cases from the sample as it has the most missing cases (n = 432). Our final analytic sample size is thus 10,080 respondents from the Add Health study.

Measures

Dependent variables

The dependent variable for the current research is criminal desistance; however, this variable is categorical, indicating different degrees of desistance in adulthood. During the third and fourth waves, respondents were asked about whether they had engaged in criminal behavior in the past 12 months. The criminal behavior measures change from the first and second waves to the third and fourth waves to reflect heterotypic continuity in offending over time, which more accurately captures patterns of long-term offending (Chen and Jaffee 2015). Criminal behavior includes both nonviolent and violent crimes such as “steal something worth more than $50,” “deliberately damage property that didn’t belong to you,” “buy, sell, or hold stolen property,” “you shot or stabbed someone,” “you go into a serious physical fight,” and “you pulled a knife or gun on someone.” Respondents who reported committing any criminal behavior in the third and fourth waves were categorized as criminally persistent and serve as the reference group for all models. If a respondent indicated that they had not engaged in any criminal behavior in the third wave, but reported criminal behavior in the fourth wave, then they fall into the temporary desistance category. Last, if a respondent reported that they had not engaged in criminal behavior at both the third and fourth waves, then they are placed in the long-term desistance category. Respondents who reported they had engaged in criminal behavior in the third wave but not in the fourth wave are also placed in the long-term desistance category. It is possible that this group of respondents will re-offend later, and thus not desist “long-term.” We ran supplemental analyses (not reported but available upon request) dropping respondents who reported offending in the third wave, but not in the fourth wave, and find that the results are substantively similar. Thus we keep this group of respondents in the long-term desistance category.

Focal independent variables

The first focal independent variable captures both the transition to parenthood and the residential status of parents from wave III. First, respondents were asked whether they have ever had a child. Those who indicated never having a child were coded as nonparents and serve as the reference category. Those who reported a child (parents) were then asked whether they currently resided with each of their reported children. Respondents who reported they did not reside with any of their children are coded as nonresident parents, while those who reported living with at least one of their children are coded as resident parents. All three groups are mutually exclusive dichotomous measures.

Another focal measure is adolescent neighborhood poverty, which comes from the wave I contextual database. We operationalize neighborhoods as Census tracts in our study, and this variable captures the proportion of households that live below the poverty line. In early models we use this continuous measure, but in order to test our hypotheses regarding poverty moderation (H2 and H3 above), we create three mutually exclusive categories of levels of neighborhood poverty. Neighborhoods with 10% or fewer households living below poverty are categorized as low poverty neighborhoods. Neighborhoods with between 11% and 20% of households living below poverty are categorized as moderate poverty neighborhoods. Last, the high poverty group represents neighborhoods with over 20% of households living below poverty. For theoretical reasons grounded in prior literature discussed earlier, we thought that neighborhoods in adolescence would be most important in setting the contextual norms for relationships and parenting, but we also considered the role of neighborhoods in early adulthood in order to gauge the more contemporary life course context. We thus constructed a similar measure of neighborhood poverty at wave III. In ancillary analyses (not shown but available upon request), adult neighborhood poverty appeared to play little role in moderating the relationships between parent residency status and patterns of desistance. Thus in the models to follow, all neighborhood measures and model stratifications are based on wave I.

Control variables

We control for several factors that prior research has indicated are associated with parenthood, residency status, and/or criminal desistance, in order to guard against spurious interpretations. Age is a continuous variable from wave IV and ranges from 24 to 32. Gender (wave I) is a dichotomous measure, with females coded as “1” and males coded as “0.” The race/ethnicity variables (wave I) consist of mutually exclusive dummy variables for Black, Hispanic, Other, and White, which serves as the reference category. Family socioeconomic status is constructed through reports at wave I about the respondent’s parent’s occupation and education levels (Ford, Bearman, and Moody 1999). Five parental education categories (1 = less than high school, 2 = high school degree, 3 = some college, 4 = college degree, 5 = graduate/professional degree) are combined with six occupation categories (0 = not in the labor force, 1 = unskilled laborer, 2 = skilled laborer, 3 = white-collar lower level, 4 = white-collar upper level, and 5 = professional) to yield an SES score for each parent from 1 to 10. Numerous studies have adopted this approach using Add Health (Haynie, Doogan, and Soller 2014; Kuhl, Warner, and Wilczak 2012). Family structure (wave I) is a dichotomous variable in which respondents who lived with two biological parents during adolescence are coded as “1” and those who lived in any other family structure are coded as “0.”

The last group of controls gauges additional emerging adulthood transitions at wave III. Education is constructed with four mutually exclusive dummy variables consisting of less than high school degree as the reference category, high school degree, associate’s degree, and bachelor’s degree or higher. Respondent’s reported hours worked per week is a continuous measure ranging from 0 to 90 hours. Last, we include three mutually exclusive dummy variables for married, cohabiting, and single (the reference category) to measure relationship status.

Analytic strategy

The dependent variable in this study is categorical to indicate different types of criminal desistance; therefore, we use multinomial logistic regression (comparing the reference group of persistent offenders to both temporary and full desisters) using mlogit in Stata 13.1. We begin by examining the association between parenthood residency status and criminal desistance. Table 2 addresses these associations in a series of models first including parenthood residency status, demographic controls, and neighborhood poverty (model 1), then additional adult transitions (model 2) that compare patterns for temporary and long-term desistance to persistence. We then address differences in these associations across levels of neighborhood poverty (Table 3); thus, models are stratified by low, medium, and high poverty. An analogous Chow test for maximum likelihood estimation determined that it is appropriate to stratify the sample for analyses. Additional correlation and VIF tests (correlation matrix and VIF test results available on request) find that the models do not suffer from multicollinearity issues. The highest VIF was 1.46 and no correlation between two covariates was higher than 0.328.

Table 2.

Multinomial Logistic Regressions for Criminal Desistance (N = 10,080).

	Temporary desistance vs. persistence				Long-term desistance vs. persistence
	Model 1		Model 2		Model 1		Model 2
Variable	Coef.	(S.E.)	Coef.	(S.E.)	Coef.	(S.E.)	Coef.	(S.E.)
Focal variable
Nonresident parent	0.546	(0.175)	0.637	(0.340)	−0.058	(0.257)	0.014	(0.257)
Resident parent	0.812***	(0.340)	0.770***	(0.186)	0.532***	(0.138)	0.467**	(0.141)
Demographic characteristics
Age at last interview	0.175***	(0.031)	0.137***	(0.034)	0.169***	(0.029)	0.130***	(0.031)
Female	0.948***	(0.111)	0.902***	(0.902)	1.098***	(0.096)	1.050***	(0.101)
Race/ethnicity
Black	0.019	(0.157)	0.062	(0.167)	−0.201	(0.136)	−0.163	(0.138)
Hispanic	0.143	(0.180)	0.160	(0.185)	−0.078	(0.146)	−0.067	(0.149)
Other	−0.012	(0.278)	−0.025	(0.275)	0.086	(0.237)	0.073	(0.233)
Family socioeconomic status	−0.025	(0.023)	−0.042	(0.024)	−0.001	(0.021)	−0.014	(0.022)
Two biological parent household	0.136	(0.114)	0.096	(0.114)	0.263**	(0.094)	0.232*	(0.096)
Adolescent neighborhood context
Adolescent neighborhood poverty	−0.047	(0.138)	−0.033	(0.140)	−0.159	(0.129)	−0.156	(0.131)
Emerging adulthood transitions
Education
High school degree			0.480*	(0.209)			0.164	(0.127)
Associate’s degree			0.870*	(0.034)			0.686**	(0.244)
Bachelor’s degree			0.983**	(0.323)			0.757**	(0.241)
Hours worked per week			−0.001	(0.003)			−0.002	(0.003)
Relationship status
Cohabiting			−0.018	(0.159)			−0.007	(0.140)
Married			0.448*	(0.172)			0.460**	(0.143)
Constant	−5.279***	(0.883)	−4.587***	(0.939)	−3.652***	(0.822)	−2.662**	(0.875)
Model F	13.14***		8.63***		13.14***		8.63***

^*p < .05;

^**p < .01;

^***p < .001

Table 3.

Multinomial logistic regression models for criminal desistance stratified by neighborhood poverty.

	Temporary desistance vs. persistence						Long-term desistance vs. persistence
	Low poverty		Moderate poverty		High poverty		Low poverty		Moderate poverty		High poverty
Variable	Coef.	(S.E.)	Coef.	(S.E.)	Coef.	(S.E.)	Coef.	(S.E.)	Coef.	(S.E.)	Coef.	(S.E.)
Focal variable
Nonresident parent	0.233	(0.280)	0.870	(0.657)	1.137	(0.644)	0.132	(0.216)	−0.657	(0.631)	0.453	(0.582)
Resident parent	0.916**	(0.517)	0.189	(0.415)	0.980**	(0.356)	0.412	(0.396)	0.218	(0.368)	0.754*	(0.298)
Demographic characteristics
Age at last interview	0.118**	(0.044)	0.138	(0.082)	0.163*	(0.068)	0.136**	(0.041)	0.060	(0.069)	0.161**	(0.055)
Female	1.040***	(0.149)	0.832**	(0.295)	0.632*	(0.245)	1.223***	(0.111)	0.850***	(0.234)	0.782**	(0.243)
Race/ethnicity
Black	0.182	(0.247)	−0.530	(0.376)	0.292	(0.278)	−0.141	(0.203)	−0.676*	(0.260)	0.185	(0.209)
Hispanic	0.178	(0.215)	0.059	(0.390)	0.278	(0.394)	−0.008	(0.189)	−0.216	(0.335)	0.011	(0.354)
Other	0.171	(0.304)	−0.777	(0.787)	−0.125	(0.749)	0.312	(0.230)	−0.625	(0.422)	−0.239	(0.578)
Family socioeconomic status	−0.067*	(0.027)	0.052	(0.055)	−0.059	(0.058)	−0.019	(0.027)	0.063	(0.046)	−0.071	(0.045)
Two biological parent household	0.074	(0.148)	0.312	(0.310)	0.034	(0.171)	0.178	(0.118)	0.136	(0.237)	0.034	(0.191)
Emerging adulthood transitions
Education
High school degree	0.442	(0.312)	0.914*	(0.378)	0.202	(0.330)	0.207	(0.213)	0.376	(0.237)	−0.116	(0.188)
Associate’s degree	1.036*	(0.424)	1.182	(0.744)	−0.282	(0.721)	0.814*	(0.320)	0.807	(0.527)	0.029	(0.633)
Bachelor’s degree	0.825	(0.449)	3.173***	(0.722)	1.209	(0.923)	0.549	(0.310)	2.622***	(0.645)	1.475	(0.799)
Hours worked per week	0.001	(0.004)	−0.005	(0.006)	−0.001	(0.006)	−0.001	(0.003)	−0.004	(0.005)	−0.001	(0.005)
Relationship status
Cohabiting	−0.157	(0.218)	0.213	(0.360)	0.260	(0.396)	−0.012	(0.174)	0.101	(0.326)	0.063	(0.377)
Married	0.629*	(0.279)	0.964*	(0.428)	−0.193	(0.328)	0.643*	(0.259)	0.599	(0.338)	0.136	(0.285)
Constant	−3.975**	(1.215)	−5.359*	(2.258)	−5.010*	(2.007)	−2.879*	(1.162)	−0.879	(1.968)	−3.425*	(1.646)
Model F	8.22***		3.25***		3.73***		8.22***		3.25***		3.73***
N of respondents	6,280		1,794		2,006		6,280		1,794		2,006

^*p < .05,

^**p < .01,

^***p < .001.

Results

Descriptive results

The weighted descriptive statistics for the sample are provided in Table 1. The majority of respondents fall into the long-term desistance category (71.9%), followed by temporary desistance (15.7%) and persistence (12.4%). The study sample largely consists of respondents who are not parents in the third wave of interviews (71.6% of the sample), with 22.4% of the sample reporting being a resident parent at the third wave and only 6% reporting being a nonresident parent. Respondents on average are around 28 years old in the last wave of interviews. A little under half of the respondents are female (48%), which is a close to even split across gender. A larger percentage of respondents identify themselves as white (67.7%), followed by 15.3% identifying as black, 11.9% identifying as Hispanic, and 5.1% identify themselves as belonging to some other racial group. The respondents, on average, come from a middle class family (average of parents’ SES is 5.510). Just over half of the respondents reported living in a two-biological parent household as adolescents (56.7%).

Table 1.

Weighted sample descriptive statistics.

Variables	Mean	(S.E.)	Range
Dependent variable
Criminal desistance
Persistence	0.124	–	0–1
Temporary desistance	0.157	–	0–1
Long-term desistance	0.719	–	0–1
Focal variable
Parenthood & residency
Nonparent	0.716	–	0–1
Resident parent	0.224	–	0–1
Nonresident parent	0.060	–	0–1
Demographic characteristics
Age at last interview	28.228	(0.117)	24–32
Female	0.480		0–1
Race/ethnicity
White	0.677	–	0–1
Black	0.153	–	0–1
Hispanic	0.119	–	0–1
Other	0.051	–	0–1
Family socioeconomic status	5.510	(0.114)	1–10
Two biological parent household	0.567	–	0–1
Emerging adulthood transitions
Education
Less than high school degree	0.113	–	0–1
High school degree	0.738	–	0–1
Associate’s degree	0.054	–	0–1
Bachelor’s degree	0.095	–	0–1
Hours worked per week	26.370	(0.512)	0–90
Relationship status
Single	0.702	–	0–1
Cohabiting	0.150	–	0–1
Married	0.147	–	0–1
Adolescent neighborhood poverty
Proportion of families below poverty-level	0.117	–	0–1
Neighborhood poverty
Low poverty	0.623	–	0–1
Moderate poverty	0.178	–	0–1
High poverty	0.199	–	0–1
N of respondents		10,080

Emerging adulthood transition variables capture education, employment, and relationship roles. By the third wave of interviews 11.3% of respondents had less than a high school degree, 73.8% of respondents had a high school degree, while 5.4% had received an associate’s degree and 9.5% of respondents had received a bachelor’s degree. The average number of hours worked per week is 26.370 in the sample. The majority of respondents are single in the third wave (70.2%), with the percentage of cohabiting (15%) and married (14.7%) respondents almost equal in the sample.

We also report the sample descriptive statistics for the adolescent neighborhood poverty measures. The percentage of households below poverty in the sample neighborhoods is 11.7%, which suggests that this sample comes, on average, from more advantageous neighborhood contexts. Based on our categorization of neighborhood poverty levels, 62.3% of respondents come from low-poverty neighborhoods, 17.8% come from moderate-poverty neighborhoods, and 19.9% of the sample come from high-poverty neighborhoods. The percentage of households in poverty is also, as suspected, relatively stable from adolescence to adulthood (results not shown in Table 1 but provided here for comparison). The percentage of families below poverty at the third wave of interviews is 11.1%. In terms of the categorical measure of poverty, there are slightly larger changes in representation over time: there is a slight decrease in the percentage of respondents who live in low-poverty neighborhoods (59.2%) and somewhat larger decrease in those who live in high-poverty neighborhoods (16.5%) at the third wave, while there is an offsetting increase in the percentage of respondents who live in moderate-poverty neighborhoods at the third wave (24.3%).

Multivariate results

Multinomial regression results for the full sample are presented in Table 2. Model 1 examines the relationship between parenthood and residency status and criminal desistance with the inclusion of demographic characteristics and adolescent neighborhood context. As we see, only resident parents have significantly higher levels of both temporary (left-side results) and long-term (right-side results) desistance compared to nonparents. Resident parents have 0.812 higher log odds of temporary desistance and 0.532 higher log odds of long-term desistance compared to nonparents. The continuous measure of neighborhood poverty during adolescence is not significant for either set of comparisons.

Model 2 in Table 2 includes adult transition variables. Marriage and education each have significant and positive effects on temporary and long-term desistance. However, the relationship between resident parenthood and temporary criminal desistance is not reduced in significance and is only reduced in magnitude by 5%, such that resident parents have 0.770 higher log-odds of temporary desistance. Similarly, the magnitude of the relationship between resident parenthood and long-term desistance is reduced in magnitude by 12%. Thus, even after controlling for emerging adulthood transitions, resident parents are found to have 0.467 higher log-odds of long-term desistance compared to nonparents in the full model.

To contextualize the association between parenthood and criminal desistance, we stratify the models by levels of neighborhood poverty during adolescence. We present the results of these stratified models for desistance in Table 3. All models here include the full set of covariates. The first column in each side of the table displays the results for those who come from low-poverty neighborhoods. The relationship between resident parenthood and temporary desistance is significant in this model, with resident parents in low-poverty neighborhoods having 0.916 higher log-odds of temporary desistance compared to nonparents. In both of the medium-poverty models, the relationship between parenthood residency status and desistance is not significant. In the high-poverty context, however, there is a significant relationship between resident parenthood and both temporary and long-term desistance. Resident parents have 0.980 higher log-odds of temporary desistance and 0.754 higher log-odds of long-term desistance when compared to nonparents. Contrary to our prediction, there are no significant associations between nonresident parenthood and desistance in any poverty context. We discuss the implications of these results in the concluding section.

Summary and discussion

The goal of the current study was to explore in greater depth the association between parenthood and criminal desistance. In order to build on prior literature, our study examined the influence of parents’ residency status, interacted with adolescent neighborhood poverty contexts, on temporary and long-term desistance. The first key finding pertains to residency status: not all parents (compared to nonparents) had greater likelihoods of desisting from offending—only resident parents did. This did not completely support our first hypothesis, which stated that all parents would be more likely to temporarily desist from offending in adulthood. However, the second part of this hypothesis (H1) regarding long-term desistance was supported, as resident parenthood was related to long-term desistance while nonresident parenthood was not. These findings lend support to theoretical arguments from the age-graded theory of social control and routine activities theories, in that time spent performing prosocial roles (in this case, parenthood) and developing meaningful responsibilities as caretakers may offer new stakes in conformity (Sampson and Laub 1993) and significantly interfere with time that would otherwise (as a nonresident parent) be available to commit crime (Osgood et al. 1996).

Our findings on desistance across neighborhood contexts also produced mixed support for our hypotheses. Yet these stratified models revealed interesting patterns, namely that it seems that adolescent residence in the extremes of neighborhood poverty (both high and low) alters the association between resident parenthood and desistance, if only temporarily in some cases. Demonstrating partial support for H2, results showed that for respondents who grew up in high-poverty neighborhoods, resident parenthood was related to both temporary and long-term desistance. Resident parenthood also produces temporary desistance for those who grew up in low-poverty neighborhoods. Going against expectations from H2 is the finding that resident parenthood was not related to long-term desistance from offending in low-poverty neighborhoods. Additionally, resident parenthood was not found to be related to any level of desistance in moderate-poverty neighborhoods. Nonresident parenthood was not related to criminal desistance at any level of neighborhood poverty, which does not support our third hypothesis (H3) that nonresident parenthood would be related to desistance in high poverty neighborhoods. Our findings pertaining to contextual influences of adolescent poverty indicate that future research regarding the nature of parenthood and crime should take into consideration not just the physical household residency status of those who transition to become parents, but also the long-reaching consequences of prior developmental contexts and the ways in which these contexts impart meaning to parenthood. It is also possible that current theoretical frameworks for explaining desistance (namely, age-graded informal social control, routine activities, and symbolic interactionism) are more complete when they consider neighborhood context. This is not to say that these frameworks ignore structural disadvantage, but it is not directly built into original statements. Informal social control theory is the result of a highly influential study [of men] who grew up many decades before the sample in Add Health, and thus many decades before contemporary demographic shifts in marriage, parenthood, and the most recent economic recession. How poverty might contextualize transitions to parenthood, and how this has implications for desistance, may differ in this current economic climate. Sampson and Laub (1997) certainly offer meaningful insight in their concept of “cumulative disadavantage” yet empirical studies on age-graded informal control almost entirely ignore low-poverty contexts. Future research may do well to directly examine the mechanisms that help account for why resident parenthood produces desistance only in high- and low-poverty contexts, but not for persons who grow up in moderate-poverty neighborhoods. It may be that the quality of bonds that develop between parents and children vary across neighborhood poverty context. We are unable to capture this possibility in our data here, but this is an important consideration for expanding informal control theory.

While some studies testing routine activities also examine these mechanisms within structurally disadvantaged contexts (e.g., Spano, Freilich, and Bolland 2008), this nonetheless leaves us unsure whether routine activities might help explain why resident parenthood matters at the extremes of poverty and not in moderate-poverty places. Similarly, other studies of desistance from more micro-level perspectives (such as the neo-Median perspective of Giordano, Schroeder, and Cernkovich (2007)) tend to focus on high-rate or serious offenders, which typically grew up in high-poverty places. Nevertheless, Giordano and colleagues (2007:1646) do make note of the fact that “poverty is systematically linked to the likelihood that individuals will be able to put together and maintain the type of respectability package” that is most relevant for desistance, and that “structurally based sources of disadvantage” help explain the link between emotional vulnerabilities to the desistance process (Giordano, Schroeder, and Cernkovich 2007:1651). Our findings that adolescent poverty interacts with parenthood status to influence desistance are thus in line with this argument, although we also find meaningful patterns in low-poverty contexts. We do not, however, directly test cognitive shifts. Future research using survey samples needs to include explicit measures of emotional changes that follow parenthood, as well as better measures of time use and routine activities following parenthood so that these theoretical mechanisms can be more thoroughly fleshed out. Scholarship also needs to include a wider range of structural contexts to examine whether these emotional shifts, qualitative bonds, or routine activities differ across the full spectrum from highly disadvantaged to highly advantaged samples.

There seem to be clear policy implications of our findings. Resident parents who grew up in impoverished contexts have higher odds of desisting from crime both temporarily and in the long term. Ethnographic research has found parenthood to be particularly salient for individuals in disadvantaged contexts (Anderson 1999; Edin and Kefalas 2001), yet scholarship also demonstrates that non-residency is more normative in such contexts. Thus, policies that aim to encourage families in high-poverty neighborhoods to live together should be encouraged. Yet historically, those policies have aimed to keep impoverished households separated. Federal policy for public and assisted housing, for example, has for decades required reporting of monthly income to calculate rent, which some scholars have argued creates incentives for mothers to remain single and “fathers to live elsewhere” (Katz et al. 2003:20). Considering the importance of residential parenthood in disadvantaged contexts that our findings reveal, federal and community guidelines for housing are distinctly out of touch with criminological theories that encourage routine prosocial activities for adults that can help not just develop, but maintain, stakes in conformity and thus cessation from criminal behavior. Moreover, our findings suggest that residential parenthood also plays a role in the cessation of offending even for individuals from low poverty contexts, at least in the short term.

Of course, although parenthood may be beneficial and rewarding, it is also stressful for parents (Nomaguchi and Milkie 2003). While this stress cuts across neighborhood poverty contexts, it may nonetheless be exacerbated in high-poverty contexts where families are struggling with resource deprivation. The stressors of parenting could in turn promote criminal behavior in some instances. Parents who grew up in more advantaged contexts may have normative expectations of placing their children in daycare while they work full-time jobs, yet those from disadvantaged contexts may have no such expectations, or may have made the transition to parenthood at younger ages when responsibilities are more difficult to manage. They may also not have made the “cognitive shift” to see themselves in a way that reduces prior bad behaviors (Giordano et al. 2002). Further, in neighborhoods with high levels of unemployment, parenting may become the master status of adults, which places even more pressure on parents. Thus, programs that offer nonjudgmental group support, that provide opportunities for parents to afford alternative child care and thus encourage full-time employment are necessary in highpoverty neighborhoods especially. These programs could be invaluable for encouraging role transformations to allow parents to identify not just as parents but as adults who perform multiple roles—and they could then come to reconfigure parenthood, employment, or other roles that are competing with former antisocial roles or behavioral repertoires (e.g., the partier, the drug dealer) within a “hierarchy of salience” (Stryker 2008).Thus, providing parents with psychological and material resource support can go a long way in helping them achieve an “identity altering experience with implications for criminal desistance” (Giordano et al. 2011:414).

Several limitations of the study should be noted. First, the study sample is school-based, which can bias the sample toward non-offenders. It is possible that individuals who are engaging in more criminal activities are not included in this sample due to missing school, being suspended or expelled, or having dropped out of school altogether. Second, while our measure of criminal desistance is an improvement over prior studies, it still is not able to perfectly capture desistance. To date, Add Health has followed its respondents only to young adulthood, and it is possible that those who currently fall into the long-term desistance group could go back to committing crime, or that those in the temporary group will be full desisters at the next follow-up interview. Wave V data should allow scholars to explore these possibilities. In addition, our measure here does not distinguish between violent and nonviolent desistance, which is likely to influence patterns. In supplemental analyses stratified into violent and nonviolent desistance (not shown), nonviolent crime was found to drive many of our present results, but due to lack of statistical power in subcategories for residency and poverty contexts, we needed to combine these offenders into a single group. Future studies with larger subsample sizes should aim to distinguish between these two types of desisters to further uncover the reasons why resident parenthood seems to play a smaller role for violent offenders than nonviolent offenders. Finally, our measure of neighborhood context takes only poverty into account, which is a crucial indicator of developmental structure that may matter, but future research should also consider other markers of structural context, such as racial/ethnic/immigrant composition, levels of educational attainment, and residential stability.

There are other avenues for future scholarship to pursue with regard to the relationship between parenthood and criminal desistance. Though we focused on stratifications by neighborhood poverty, future research should also consider gender-stratified models. This would help to more fully compare the results presented here to those of Kreager and colleagues (2010), who focused on motherhood in disadvantaged contexts. It is likely that fatherhood and motherhood carry different meanings that might produce different effects on desistance, as illustrated by ethnographic research (Anderson 1999; Edin and Kefalas 2005; Edin and Nelson 2013). Additionally, relationship context could moderate the association between parenthood and criminal desistance across neighborhood contexts. Research illustrates that romantic relationships are characteristically different in disadvantaged neighborhoods and this may change the association between parenthood and crime. While we do control for marriage and cohabitation in our models, it may be worthwhile to examine the combination of role transitions, to more fully gauge the entire “respectability package” that should theoretically offer the most benefit for desistance.

Future research should also further investigate the association between parenthood and temporary desistance to better understand why parenting does not have long-term benefits in neighborhoods of lower poverty. Research has found that other transitions, namely marriage, have a similarly “situational” effect whereby individuals who divorce do not experience the long-term benefits of marriage (Bersani and Doherty 2013). The mechanisms underlying the temporary benefits of life course transitions have not been fully explored yet. The temporary effect of the transition into parenthood on criminal desistance could be due to changes in routine activities when the child is younger and needs more care, or the initial joy of the transition into parenthood that might wear off as an individual ages. Additionally, the meaning of parenthood may be different in neighborhoods that are more advantaged and future studies should consider examining parenthood in contexts outside of disadvantage. What we do see here is that the link between parenthood and crime is rather complex, and scholarship moving forward should be mindful of these contingencies so as to more fully understand these associations, and the potential for desistance that parenthood can offer.

Biographies

JESSICA A. ZIEGLER is a Ph.D. student in the Department of Sociology at Bowling Green State University. Her research focuses on patterns of victimization over the life course and the relationship between life course transitions and criminal desistance with an emphasis on differences across neighborhood contexts.

DANIELLE C. KUHL is Associate Professor of Sociology at Bowling Green State University. Her current research focuses on violence, delinquency, and substance use within neighborhood and developmental contexts.

RAYMOND R. SWISHER is a Professor of Sociology at Bowling Green State University. His research focuses on risk factors in the lives of disadvantaged youth (neighborhood poverty, exposure to violence, parental incarceration), and their consequences for crime, violence, and psychological well-being (depression, survival expectations) across the life course.

JORGE M. CHAVEZ is an Associate Professor in the Department of Sociology at Bowling Green State University and a member of the Racial Democracy, Crime, and Justice Network. His research centers on integrating developmental, social, and environmental perspectives toward understanding problem behaviors, crime and violence, and mental health problems, with a focus on the intersection of race/ethnicity and social structural disadvantage. His recently published research has examined offending and locational attainment in young adulthood, immigration and violent crime, race/ethnicity and state level immigration policy, and the intergenerational transmission of crime and violence.