week 5 paper

Feminist Criminology 2018, Vol. 13(2) 182 –204

© The Author(s) 2016 Reprints and permissions:

sagepub.com/journalsPermissions.nav DOI: 10.1177/1557085116654566

journals.sagepub.com/home/fcx

Article

Gender Differences on the Road to Redemption

Gina Curcio1, April Pattavina1, and William Fisher1

Abstract Redemption research examines how much time must pass after a criminal offense before an offender is considered “redeemed.” This study adds to redemption research by using a nationally representative sample from the United States to determine whether years to redemption found in prior research replicate and will be the first to determine whether there are gender differences. We also explore factors that influence who makes it to the redemption point. Findings reveal that while men reach the redemption point after 10 years, women reach the redemption point after 4 years. Policy implications of these findings are discussed.

Keywords gender, desistance, social policy, redemption research, collateral consequences

Introduction

Criminologists have recently begun to examine the long-term risk of recidivism, in what is referred to as redemption research (Blumstein & Nakamura, 2009; Bushway, Nieuwbeerta, & Blokland, 2011; Kurlychek, Brame, & Bushway, 2006, 2007; Soothill & Francis, 2009).1 This body of work attempts to determine when the risk of recidi- vism for an individual with a criminal record becomes equal to the risk of a first offense for a nonoffender. In statistical terms, redemption research attempts to calcu- late how much time (usually in years) needs to pass for individuals with a criminal record to have the same likelihood of recidivism as a comparison group committing a crime for the first time. The number of years that an offender has to remain crime free

1University of Massachusetts Lowell, USA

Corresponding Author: Gina Curcio, University of Massachusetts Lowell, One University Avenue, Lowell, MA 01854, USA. Email: Gina_Curcio@uml.edu

654566 FCXXXX10.1177/1557085116654566Feminist CriminologyCurcio et al. research-article2016

Curcio et al. 183

before their risk of reoffending matches the risk of a nonoffender group offending for the first time is termed the “point of redemption.”

Redemption studies were motivated in part by research consistently showing that it is often very difficult for those with a criminal record to secure gainful employment (Holzer, Raphael, & Stoll, 2003; Pager, 2003; Uggen & Staff, 2004). Although innova- tive, this body of work has been limited to studies involving males. Given that the findings of redemption research can be used to motivate policies that limit the use of criminal records in hiring decisions, it is important to consider how redemption points for women may differ from their male counterparts.

In this study, we add to the small but developing body of redemption research using a recent, nationally representative sample of individuals born in the United States. Our analysis will be among the first to examine gender differences in time to redemption (also referred to as points of redemption). In the sections that follow, we first review existing redemption research literature and discuss the relevance of considering gen- der differences in this context. Second, we present the research questions along with data and methods used for the analysis. Third, we present the findings followed by a discussion of important implications for policy and future research.

Literature Review

Redemption research is concerned with making an important connection between the theory of desistance, which seeks to explain when and why people stop committing crime, and policy and practice where the stigma of prior formal justice system involve- ment may negatively affect future opportunities for employment. Prior research con- sistently shows that many employers today are unwilling to hire individuals with a criminal record even if they surpass the job qualifications for a position (Laub & Sampson, 2003; Pager, 2003; Uggen & Staff, 2004). By understanding how much time must pass before offenders in a cohort have the same likelihood of criminal involve- ment as a comparison group without initial involvement, policy makers and employers can be better informed about how much time must pass before offenders can be con- sidered “redeemed.”

The general approach is to quantify years to redemption (hereafter redemption points) for individuals with a criminal record. This type of research requires longitudi- nal analyses, and there have been several noteworthy studies published on the subject. Kurlychek et al. (2006, 2007) utilized the Second Philadelphia Birth Cohort Study and the Racine Birth Cohort Study to calculate points of redemption and determined that the risk of recidivism for those who have not been arrested in 6 to 7 years begins to approximate the risk of offending in the general population. Their work was the first to show empirically that individuals with a criminal record do eventually look like nonoffenders.

Using a larger, more recent sample of offenders from the New York criminal history repository, Blumstein and Nakamura (2009) examined redemption points over a 27-year period and concluded that redemption points vary based on the age at first arrest and the offense type. One important disadvantage of using criminal history data

184 Feminist Criminology 13(2)

is the lack of a nonoffender comparison group. For their comparison group, the researchers created a group from the general population of New York at the same time period and of the same age as the baseline offender group. Overall, they concluded that redemption occurs after 4 to 5 years of being crime free for property offenders and between 7 and 8 years for violent offenders.

Bushway et al. (2011) used court records from the Netherlands to calculate points of redemption and were the first to utilize a matched control group of nonoffenders to create a relevant comparison group. They concluded that redemption points vary depending on age of first conviction and the number of prior convictions. According to their results, redemption occurs between 2 and 6 years for older offenders with no prior convictions and after 10 years for younger offenders with prior convictions. Soothill and Francis (2009) used conviction data on youthful offenders (under age 20) and reported redemption points between 10 and 15 years.

The findings of existing redemption research do provide evidence that the risk of reoffending does significantly decrease after a certain number of years of remaining crime free. Furthermore, this body of work indicates that many factors can influence time to redemption. Although redemption research has made important contributions to the criminological literature, findings have been limited to studies involving males.

Consideration of gender differences in offending and desistance is important because over the past few decades the rate of women becoming involved in the crimi- nal justice system has increased significantly (Belknap, 2007). According to Frost, Greene, and Pranis (2006) in 1977, the U.S. imprisoned 11,212 women and by 2004, that number had ballooned to 96,125, a 757% increase. The rise in female arrests and incarceration has particular implications for children because an estimated 66% to 80% of incarcerated women are the primary caregiver of their minor children prior to arrest (Snell & Morton, 1994). Women also experience significant levels of recidi- vism. According to recent Bureau of Justice Statistics report, within 3 years of release from prison, 69% of male offenders and 58% of female offenders had been rearrested at least once. Five years after release from prison, 77% of male offenders and 68% of female offenders had been rearrested at least once (Durose, Cooper, & Snyder, 2014). While more recent research indicates that recidivism rates can vary by jurisdiction (Pew Research Center, 2011) and by recidivism measure, the high rates of continued involvement in the justice system speak to the importance of identifying factors that influence desistance and exploring the possibility of gender differences in the predic- tion of desistance.

A gender-based orientation to redemption adds to the existent desistance literature by examining not only whether female offenders are more or less likely to be redeemed than their male counterparts but also whether they are likely to have lower redemption points. We turn to the literature on gender differences in offending to guide our consid- eration of gender differences in points of redemption. Research on gender differences in offending indicates that female offenders have less extensive criminal records than their male counterparts and are much less likely than male offenders to have been convicted of a violent crime (Belknap, 2007). Although women are less likely than men to be convicted of a violent crime, the number of women going to prison for a

Curcio et al. 185

violent offense increased 83% between 1991 and 2011 (Carson & Golinelli, 2013), reflecting what some have described as policy changes that expanded the definition of violence to include minor acts violence which women are more likely to commit (Schwartz, 2013). It is still more likely that the types of offenses most often associated with female offending are nonviolent in nature, including drug crimes, property crimes, and sex work (Mallicoat & Ireland, 2014). Prior redemption research has shown that nonviolent offenders and those with less extensive criminal records have lower points of redemption (Blumstein & Nakamura, 2009; Bushway et al., 2011; Kurlychek et al., 2007). Thus, the available research on gender differences in the nature and extent of offending suggests that female offenders will likely have lower points of redemption than their male counterparts.

Evolving out of feminist pathways, research that examines unique avenues that women take into crime (Belknap, 2007; Chesney-Lind, 1997; Daly, 1998; Owen, 1998) is a growing body of desistance research that focuses on why women desist from crime (Cobbina, 2010; Huebner, DeJong, & Cobbina, 2010) and how factors that affect desistance can be similar and different for men and women (Giordano, Cernkovich, & Rudolph, 2002). For example, there is some evidence to suggest that having dependent children may have particularly strong and positive effects on female desistance because children facilitate attachments and enhance one’s stake in confor- mity (Kreager, Matsueda, & Erosheva, 2010) and can help female offenders develop a prosocial self-image (Giordano et al., 2002).

Education has been identified as an important factor in promoting crime desistance for both men and women (Holtfreter, Reisig, & Morash, 2004; Ulmer, 2001), and some suggest that educational achievement is particularly important for female offend- ers for several reasons, including finding gainful employment not only because an education builds job skills but also because it supports the development of life skills and increases self-esteem (Blitz, 2006; Case & Fasenfest, 2004). Case and Fasenfest (2004) identified poor self-esteem as one of the most difficult challenges facing female offenders. Similarly, Salisbury and Van Voorhis (2009) found that female probationers with stronger educational backgrounds tended to have more self-efficacy and were less likely to reoffend.

Evidence related to factors influencing female desistance reveals some mixed find- ings. For example, in their study of female offenders, Huebner et al. (2010) examined factors that influence success on parole and found that women who were drug depen- dent, had less education, and have more extensive criminal histories were more likely to fail on parole. Interestingly, while their study also included dependent children as a predictor of recidivism, it was negatively associated with failure at the bivariate level but not significant at the multivariate level. Giordano and colleagues (2002) found that level of attachment to a marital partner and job stability were not significantly related to likelihood of desistance for male or female respondents. Instead, they reported that women were more likely to describe their children or religious transformations as reasons for their desistance.

Furthermore, given the newness of redemption research, scholars have noted that more studies are needed to determine whether these redemption points “hold up”

186 Feminist Criminology 13(2)

across different time periods and locations (Kurlychek et al., 2006). Most redemption research has used data that are between 50 and 60 years old. The most current data used in redemption studies are more than 30 years old (Bushway et al., 2011; Soothill & Francis, 2009). The present study will seek to address many of the limitations of prior redemption research by using a more recent, nationally representative sample from the United States to see whether points of redemption are consistent with what prior studies have found and more specifically to examine whether there are gender differences in points of redemption.

Data and Method

Data for the current study are from the 1997 National Longitudinal Survey of Youth (NLSY97). The NLSY97 is an ongoing study compiled by U.S. Bureau of Labor Statistics (BLS) and is designed to gather information on the labor market activities and other significant life events of several groups of men and women in the United States. The NLSY97 consists of a nationally representative sample of 8,984 youths who were 12 to 16 years old at the time of the first interview in 1997 and between 27 and 31 years old at the time of the last available round of data collection in Round 15 as of November 2014. Youths are interviewed on an annual basis. The sample of respondents was selected using a multistage cluster sampling design of 75,291 house- holds in the United States (Moore, Pedlow, Krishnamurty, & Wolter, 2000).2

In addition to the longitudinal design of the NLSY97, there are several other impor- tant advantages of using this data set for the current study. First, the survey includes both males and females, which allows for comparison across gender. Second, unlike most longitudinal studies, the NLSY97 attempts to follow up with respondents even if they have moved to another city or state or are incarcerated, which minimizes possible attrition. When the NLSY97 does catch up with missing respondents, they fill in the missing data from previous years. Based on the consideration of interview dates for the respondents, the attrition rate does not exceed 14%. Third, as the NLSY97 is a nationally representative sample of individuals born between 1980 and 1984, the results of the study can be generalized to all individuals born between 1980 and 1984 in the United States (Moore et al., 2000). Fourth, this data set includes a comparison group of nonoffenders, which is ideal for redemption research.

During the first wave of the NLSY97, respondents were between the ages of 12 and 16 years as of December 31, 1996. Because our primary interest is in comparing haz- ard rates of 18-, 19-, and 20-year-old individuals who have had at least one arrest to those who have had no arrests at all, a subsample of respondents who were 18 to 20 years old in Round 4 in 2000 was created. This round of data was chosen as a starting point because it is during this round that we can maximize the number of individuals in the sample while still maintaining the maximum follow-up period of 11 years. Following the lead of Kurlychek et al. (2006), two baseline comparison groups will be created for the present analysis: (a) age 18-, 19-, and 20-year-old nonoffenders and (b) age 18-, 19-, and 20-year-old offenders with one or more arrests.3 The final sample includes 4,663 respondents.

Curcio et al. 187

Analytic Approach

The first part of our study involves identifying the year (or point) of redemption for the entire population and then separately by gender. Many researchers recommend that feminist criminologists utilize split sample analyses by gender to determine whether independent variables suggested by mainstream criminological theories exert unique influences on males and females rather than simply using gender as a control variable (Huebner et al., 2010; Wattanaporn & Holtfreter, 2014). We rely on arrests as our mea- sure of formal justice system involvement. Convictions were also considered for the analysis as the NLSY97 includes this information, but the data set only provides the year of conviction rather than the date, and thus, the conviction could have been for a crime that happened years prior. The NLSY97 data set does provide the month and year of arrest, and having the date is important for redemption research because we are modeling time to rearrest (or arrest for the nonoffending group).

The analysis involved modeling time to arrest (or rearrest) using survival analy- sis. An advantage of survival models over models for binary outcomes is that varia- tion in the time to an event is modeled, rather than just whether the event occurred. These models estimate hazard rates which have two foci: (a) whether the “event of interest” has occurred or not, which in the present analyses is whether the respon- dent is arrested (or rearrested for the baseline offender groups); and (b) the trend in conditional likelihood of arrest, which in the present analyses is years to arrest (or rearrest for the baseline offender groups; Miller, n.d.). To examine the research questions, we modeled the hazard probabilities across follow-up years using a dis- crete survival model (Singer & Willett, 2003). The model predicts arrest (or rearrest for baseline offender groups) over the follow-up period of 11 years from 2001 through 2011.

Our goal is to establish how many years of nonoffending it takes before a baseline offender begins to equal baseline nonoffenders in terms of their offending rate. The analysis examines the time to arrest between respondents with no arrests at age 18, 19, and 20 years and rearrest for those with at least one arrest at age 18, 19, or 20 years. Like prior redemption research, we must decide when the hazards will be “equal.” Bushway et al. (2011) examined three methods of convergence, including comparing the upper bound, the point estimate, and the lower bound of the offender hazard to the upper bound of the nonoffender hazard. Borrowing one of the methods from Bushway et al., our strategy argues that the two groups are “equal” the year in which the lower bound of the offender group confidence interval crosses the upper bound of the nonof- fender confidence interval. This is considered to be a conservative approach according to Bushway and colleagues. We first examine the redemption results for the entire sample and compare our results with prior studies. We then compare separate results for males and females.

In the survival analysis (N = 4,663), about 51% of the sample is male and about 51% of the sample is White. About 35.5% of the sample was 18 years old, 34% of the sample was 19 years old, and 30% of the sample was 20 years old. Approximately 15.8% of the sample was classified as baseline offenders, and about 23.3% of the

188 Feminist Criminology 13(2)

sample offended (or reoffended for the baseline offender groups) during the observa- tion period from 2001 through 2011.

The point of redemption is estimated for the population, but only some people will make it to the point of redemption. A secondary goal of our study is to determine whether individual characteristics predict who will make it to the point of redemption and whether they vary by gender. After determining the time to redemption, logistic regression analysis is used to explore factors that may predict who makes it to the point of redemption without an arrest. This analysis is limited to those who had at least one arrest at baseline (n = 737). The dependent variable is whether or not a person is rearrested by the point of redemption (coded as 0 = no, 1 = yes).

Our independent variables include factors considered in prior redemption research along with those that may offer some insight into redemption outcomes for those who come in contact with the criminal justice system in emerging adulthood. Prior research shows that the point of redemption is longer for those with multiple arrests at baseline (Bushway et al., 2011). To account for the possibility that those with multiple arrests are less likely to make it to redemption, we included a mea- sure that indicates whether the baseline offender was arrested for one offense only or had two or more arrests at baseline (0 = one arrest only, 1 = two or more arrests at baseline). We are also interested in exploring the role that life course milestones may have in facilitating redemption for emerging adults. A dichotomous variable indicates whether the respondent had at least one dependent child during the time to redemption (0 = no children by the point of redemption, 1 = at least one child by the point of redemption).4

Education is included as an independent variable because of the possible connec- tions to both redemption and employment readiness. A significant body of research has found that postsecondary education is effective in reducing subsequent involve- ment in criminal activity by enhancing employability, improving self-esteem, and encouraging personal growth (Batiuk, Lahm, McKeever, Wilcox, & Wilcox, 2005; Stevens & Ward, 1997; Vacca, 2004), yet no research has examined whether there are gender differences in how education may affect redemption. Two education level vari- ables were included in the model as dummy variables, including one for those having a high school diploma and one for college degree, each coded as (0 = no, 1 = yes). No high school diploma is the reference category.5 Demographic variables include gender (0 = female, 1 = male), race (0 = people of Color,6 1 = White), and age at the start of the observation period measured as a continuous variable.7 Checks for multicollinear- ity revealed no correlations greater than .2.

We also considered the type of crime at the baseline for respondents.8 The majority of offenses for first arrest were for relatively minor offenses, including disorderly conduct, reckless driving, driving under the influence (DUI), other major trafficking offenses, or drug offenses (81%). Only a small percentage of crimes were for serious property (14%) and serious violent (5%) offenses. Very few NLSY97 respondents (1%) were incarcerated during the observation period. Preliminary analysis indicated that crime type was not a significant predictor of redemption and was thus excluded from the multivariate models.

Curcio et al. 189

Table 1 presents descriptive statistics for all variables used in the logistic regression analyses (n = 737). In the year 2000, when the baseline offender and nonoffender groups were created, about 35.5% of the sample was 18 years old, 34% of the sample was 19 years old, and 30% of the sample was 20 years old. Approximately 74% of the sample is male, and about 49% of the sample is White.

Results

Redemption Point Analysis

The hazard model presented in Figure 1 is estimating “years to arrest” (or rearrest for the baseline offender groups). Arrest rates were regarded as equal when the lower bound of the offender group confidence interval crosses the upper bound of the nonof- fender group confidence interval (Bushway et al., 2011). Overall, the mean number of years to arrest (or rearrest for the baseline offender groups) was 9.418 years, and the median number of years to arrest was 11 years. The minimum number of years to arrest or rearrest was 0, and the maximum number of years to rearrest was 11 years (January 2001-December 2011).

We first compared hazard rates of all individuals with at least one arrest at age 18, 19, and 20 years to individuals with no arrests at age 18, 19, and 20 years to determine statistically when the risk of recidivism for individuals with one arrest begins to resemble the risk of a nonoffender offending for the first time. This analysis reveals how much time an individual with at least one arrest has to remain crime free before their risk for reoffending matches that of a nonoffender. In other words, how old does a criminal record have to be in order for prospective employers to be confident that an individual with an old criminal record does not pose a higher risk of offending than an individual with no record at all?

Figure 1 presents the arrest hazard rates with upper and lower bounds of the confi- dence intervals for the full sample (males and females) for both baseline offender and

Table 1. Descriptive Statistics for All Variables Used in Logistic Regression (n = 737).

Variable %

Male 74.65 White 49.30 Age 18 38.10 Age 19 33.40 Age 20 28.50 2+ arrests baseline 47.60 High school diploma 55.80 College degree 14.40 Dependent child 40 Redeemed 51.80 Not redeemed 48.20

190 Feminist Criminology 13(2)

nonoffender groups. Recall that our convergence strategy is to compare the lower bound of the offender group confidence interval to the upper bound of the nonoffender confidence interval. There are important differences in the hazard rates for new offenses in the early years of the follow-up period for the baseline offender group as compared with the baseline nonoffender group. At t = 1 year, the lower bound of the confidence interval for the baseline offender group is 15%, which implies that about 15% of individuals at risk of being arrested for the first time at the start of the observa- tion period actually are arrested. At t = 1 year, the upper bound of the confidence interval for the baseline nonoffender group is 5%.

At t = 2 years, the lower bound of the confidence interval for the baseline offender group and the upper bound of the confidence interval for the baseline nonoffender group decline to 10% and 3%, respectively. At t = 7 years, the lower bound of the confidence interval for the baseline offender group and the upper bound of the confi- dence interval for the baseline nonoffender group become virtually indistinguishable at rates of 3% and 1%, respectively. At t = 10 years, the lower bound of the baseline offender group becomes completely indistinguishable from the upper bound of the confidence interval for the baseline nonoffender group at 1%.

The findings of the first part of the hazard rate analysis are consistent with findings of prior redemption research. While the hazard rates for the offending baseline groups do not actually cross the hazard rates for the nonoffending group until 10 years, the differ- ences in the hazard rates between the two baseline groups after 7 years are very small.

Males

Figure 2 presents the arrest hazards with confidence intervals for the males only for both baseline groups over the observation period between 2001 through 2011. At t = 1 year, the lower bound of the confidence interval for the males-only baseline offender

Figure 1. Hazard probabilities—All respondents (N = 4,663).

Curcio et al. 191

group is 17%, which implies that about 17% of individuals at risk of being arrested for the first time at the start of the observation period actually are arrested. At t = 1 year, the upper bound of the confidence interval for the males-only baseline nonoffender group is 8%.

At t = 4 years, the lower bound of the confidence interval for the males-only base- line offender group and the upper bound of the confidence interval for the males-only baseline nonoffender groups go down to 6% and 3%, respectively. At t = 7 years, the lower bound of the confidence interval for the males-only baseline offender group and the upper bound of the males-only baseline nonoffender group go down to 2.5% and 1.7%, respectively. At t = 10 years, the lower bound of the confidence interval for the males-only baseline offender group and the upper bound of the confidence interval for the males-only baseline nonoffender group become completely indistinguishable at 1%. Similar to the results of the hazard rate analysis for the full sample, male baseline offenders are “redeemed” after 10 years of remaining crime free.

Females

Figure 3 presents the arrest hazards with confidence intervals for the females only for both baseline groups over the observation period between 2001 through 2011. At t = 1 year, the lower bound of the confidence interval for the females-only baseline offender group is 7%, which implies that about 7% of individuals at risk of being arrested for the first time at the start of the observation period actually are arrested. At t = 1 year, the upper bound of the confidence interval for the females-only baseline nonoffender group is 3%. At t = 4 years, the lower bound of the confidence interval for the females- only baseline offender group and the upper bound of the females-only baseline nonof- fender group go down to an identical 1%.

Figure 2. Hazard probabilities—Males only (n = 2,352).

192 Feminist Criminology 13(2)

When comparing the hazard rates across gender, the findings showed that there are differences in time to redemption for male and female offenders. The findings of the present study indicate that women reach the point of redemption sooner than their male counterparts. In fact, while the results show that men reach the point of redemp- tion after 10 years, women with a criminal record reach the point of redemption after only 4 years. These results demonstrate the importance of considering males and females separately in redemption research.

Factors Predicting Redemption

As the hazard rate analyses showed that hazard rates for the three baseline groups did appear to cross, we can create a redemption variable to use for redemption that we can predict with respondent characteristics using logistic regression. The logistic regres- sion analyses allow us to predict who “makes it” to the “point of redemption” (“redeemed” coded as 1) and who does not (“not redeemed” coded as 0). The first logistic regression model will include the entire sample of baseline offenders (n = 737). Then, separate logistic regression models will be estimated for males and females. About 79% of female baseline offenders made it to the 4-year redemption mark, whereas 47% of male baseline offenders made it to the 10-year redemption mark. Logistic regression results are presented in terms of odds ratios (ORs).9

The results of the logistic regression analysis for the full sample (Model 1), males only (Model 2), and females only (Model 3) are presented in Table 2. Not surprisingly, baseline offenders who had two or more arrests at baseline were less likely to be redeemed than baseline offenders who had one arrest only. Baseline offenders with two or more arrests were 54% less likely than baseline offenders with one arrest only to be redeemed. Male baseline offenders are about 50% less likely to be redeemed than

Figure 3. Hazard probabilities—Females only (n = 2,311).

193

T a b

le 2

. Lo

gi st

ic R

eg re

ss io

n Pr

ed ic

ti ng

R ed

em pt

io n.

M o de

l 1 : A

ll

(r ed

em pt

io n

po in

t =

1 0

ye ar

s) M

o de

l 2 : M

al es

o nl

y

(r ed

em pt

io n

po in

t =

1 0

ye ar

s) M

o de

l 3 : F

em al

es o

nl y

(r ed

em pt

io n

po in

t =

4 y

ea rs

)

V ar

ia bl

e O

R β

SE O

R β

SE O

R β

SE

H ig

h sc

ho o l d

ip lo

m a

1. 06

7 0.

06 4

0. 17

8 0.

89 2

− 0.

11 4

0. 20

1 1.

09 0.

08 6

0. 42

7 C

o lle

ge d

eg re

e 2.

65 8*

** 0.

97 8

0. 27

8 2.

16 *

0. 77

0. 32

4 2.

16 4

0. 77

2 0.

65 2

2+ a

rr es

ts b

as el

in e

0. 45

6* **

− 0.

78 5

0. 15

9 0.

50 1*

** −

0. 69

1 0.

18 2

0. 42

9* −

0. 84

7 0.

37 6

W hi

te 1.

20 7

0. 18

9 0.

16 2

1. 49

4* 0.

40 1

0. 18

4 1.

16 3

0. 15

1 0.

38 4

A ge

a t

ba se

lin e

1. 25

3* 0.

22 5

0. 09

8 1.

22 8

0. 20

5 0.

11 1

1. 04

3 0.

04 2

0. 23

9 D

ep en

de nt

c hi

ld 0.

84 4

− 0.

16 9

0. 16

6 0.

88 5

− 0.

12 2

0. 18

6 0.

49 −

0. 71

3 0.

39 8

M al

e 0.

50 2*

** −

0. 69

0. 19

C

o ns

ta nt

0. 03

2 −

3. 45

2 1.

84 7

0. 02

2 −

3. 81

2. 11

2 2.

74 1.

00 8

4. 47

N ot

e. O

R =

o dd

s ra

ti o . F

o r

M o de

l 1 , n

= 7

37 ; f

o r

M o de

l 2 , n

= 5

50 ; f

o r

M o de

l 3 , n

= 1

87 .

*p ≤

.0 5.

* **

p ≤

.0 01

.

194 Feminist Criminology 13(2)

their female counterparts, which indicates that female baseline offenders would have higher odds of being redeemed than their male counterparts if the redemption point was extended to 10 years. Baseline offenders who were older at the time of their first arrest were more likely to be redeemed than baseline offenders who were younger at the time of their first arrest. Those who had a college degree were 66% more likely to be redeemed than baseline offenders without a high school diploma. The other vari- ables in the model were not significant.

Separate logistic regression models were also run for males and females to examine how these independent variables may differentially affect whether male and female baseline offenders are redeemed. For females, the redemption point is 4 years, and for males, the redemption point is 10 years. For the males, baseline offenders with two or more arrests at baseline were 49% less likely to be redeemed than baseline offenders with only one arrest. White offenders are 49% more likely to be redeemed than minor- ity offenders. Having a college degree increased the likelihood of redemption by 16%. None of the other variables are significant.

For the female sample, the only significant predictor of redemption is having more than one arrest at baseline. Female offenders with two or more arrests are 57% less likely to be redeemed than baseline offenders with one arrest only when other factors are controlled. No other variables were significant. Differences between male and female coefficients were tested (Paternoster, Brame, Mazerolle, & Piquero, 1998), and results determined that none of the male and female coefficients differed significantly. It is possible that these findings are related to the differences in time to redemption between males and females. For example, education was found to be more beneficial for men than women, but this may be because women are likely to be redeemed within 4 years. If the effects of education are cumulative over time, then they may be more likely to accrue to males who are redeemed by the 10-year mark. Our findings under- score the importance of examining time to redemption separately for males and females, and we urge further research on this topic.

Discussion

The analyses indicate that after a certain number of years of remaining crime free, individuals with an “old” criminal record do eventually begin to resemble nonoffend- ers with regard to their risk for rearrest. In the hazard rate analyses for the full sample (which included male and female baseline offenders), the point of redemption was 10 years. The results of the separate hazard rate analyses for males and females indicate that female baseline offenders are “redeemed” after about 4 years of remaining crime free, whereas male baseline offenders are “redeemed” after about 10 years of remain- ing crime free. These results are consistent with prior redemption research on young offenders. Soothill and Francis (2009) determined redemption points to be between 10 and 15 years. Similarly, Bushway et al. (2011) concluded that redemption points vary depending on age of first conviction and the number of prior convictions and deter- mined that redemption occurs between 2 and 6 years for older offenders with no priors and after 10 years for younger offenders with prior convictions.

Curcio et al. 195

Multivariate analyses were used to determine who is likely to make it to the redemp- tion point. The results of the logistic regression models indicate that certain factors do affect whether or not a former offender will reach redemption. For the full model, having more than one arrest at baseline had a negative impact on redemption. These same results presented in the male- and female-only models. This finding is consistent with other studies that consider multiple offenses at baseline (Bushway et al., 2011) and was the only variable that showed a consistent influence across gender.

We also explored the role that certain life milestones, including having children and education, may have in facilitating redemption and, by extension, may signal a readi- ness for employment. Interestingly, having dependent children and education were not significantly related to redemption for females once other factors were controlled. These findings suggest that for women experiencing an arrest in emerging adulthood, redemption comes rather quickly at 4 years and appears to be motivated by factors other than education and having dependent children, such as stable relationships, employment, and, more broadly, increased human, social, and state capital (Giordano et al., 2002; Holtfreter et al., 2004; Salisbury & Van Voorhis, 2009) .10 These forms of capital are important across gender but research shows that female offenders tend to face more issues related to these forms of capital than their male counterparts (Heimer, 2000).

For males, having a college degree was positively associated with redemption and may therefore signal a readiness for employment. However, for both males and females, those who experienced more than one arrest at baseline were significantly less likely to make it to redemption. This finding suggests that for this group, adher- ence to criminal lifestyles may be more firmly entrenched by this time in their lives.

The large discrepancy between males and females in the time to redemption has significant implications not only for considering what role a prior criminal record should have on employment considerations for men and women but also for under- standing the age–crime relationship to redemption more generally. As we discussed earlier, the literature on the relationship between having children and desistance from crime is inconsistent. Although there is research to suggest that having children pro- motes desistance (Giordano et al., 2002; Graham & Bowling, 1995; Kreager et al., 2010), others describe the relationship between having children and desistance as quite complex. For example, in Michalsen’s (2011) study, respondents reported that children not only facilitated prosocial bonds but also acted as stressors, indicating that future research may uncover a complicated relationship between having children and desistance. Furthermore, recent studies that examine the impact of children on desis- tance note the importance of examining the quality of the relationship or bond with their children as well as an indicator of custody status (Huebner et al., 2010). In the present study, at the bivariate level, our results revealed that having dependent chil- dren was significantly and negatively associated with desistance but was reduced to insignificance when other factors were taken into account. Recall that the women in this sample were measured for redemption after 4 years, or between the ages of 22 and 24, which is still considered in emerging adulthood by some. The men in our study were much older before reaching redemption.

196 Feminist Criminology 13(2)

In this study, 39% of the women had children before the 4-year redemption mark. Indeed, these women and their children are likely to be relatively young by the 4-year mark. Our results could indicate that some of the same factors that lead women to become justice involved in early adulthood are similar to those that preclude forming the types of bonds with their own children that could motivate desistance by the 4-year mark. Such a mechanism may be even more salient for those with multiple arrests at baseline which may explain the dominance of this variable in the multivariate model. A Vera Institute of Justice study found that a substantial portion of mothers of foster care children have a history of arrests and convictions and that mothers of foster care children were more likely to be arrested in the 18 months after placement than in the equivalent period before the placement, suggesting that children are removed in the midst of a downward spiral in the mother’s life that continues after removal (Ehrensaft, Khashu, Ross, & Wamsley, 2003). While this possibility is speculative, studies on this topic should seek to move beyond indicators that simply capture whether or not justice-involved persons have children to include those that measure how involved people are in caring for their children, to consider whether that relationship changes over time as both parents and children age, and consider how removing a child from the home may affect desistance for both men and women. Future research in this area would expand our understanding of desistance by seeking to unpack how having chil- dren may influence men’s and women’s involvement with the justice system over the life course. As Giordano et al. (2002) noted, even a high level of attachment to one’s children may not on its own constitute a powerful motive for desistance unless accom- panied by cognitive transformations from accepting or internalizing a new self-image.

Limitations and Suggestions for Future Research

There are several limitations and caveats to this study that must be discussed. First, the NLSY97 is self-report data and is subject to the limitations typically associated with self-report data, including honesty and accurate memory of the respondents.11 Second, arrest data rather than conviction data were used to examine time to redemption because arrest dates are available in the NLSY97 data set. This is a limitation because many criminologists consider arrests to be a better indicator of police behavior rather than offender behavior (Sherman, 1980). Considering police discretion, some research suggests that demographic factors, such as gender and race, may affect arrest out- comes (Alpert, Dunham, Stroshine, Bennett, & MacDonald, 2006; Kirk, 2008; Sherman, 1980; Worden, 1989). It is important to note that even though arrest mea- sures may have biases, for this redemption study, they are relevant because they are formal measures of justice system involvement that may become permanent of an individual’s record and subsequently used to make judgments about employability. We do, however, recommend that conviction data be used in future research because they are more likely to be used by employers in hiring decisions than arrest data.

Third, it is also possible that there are other important factors that would help to explain gender differences in redemption paths that were not included in the survival

Curcio et al. 197

analysis. We were constrained by the small sample size for female offenders, and con- ducting subgroup analyses by crime type or race would risk introducing unreliability in the estimates. For example, Bushway et al. (2011) suggested that crime type may influence the redemption point. While the vast majority of offenders in our sample were arrested for minor crimes, future research should continue to examine how crime type may influence the redemption point for offenders. In addition, our sample was necessarily limited to respondents in emerging adulthood. It is possible that other offending age groups may show different trajectories and should be considered in future research.

Researchers also noted the importance of exploring within-gender differences, including by pathways into crime (Wattanaporn & Holtfreter, 2014) and subgroup analyses by race (Huebner et al., 2010). For example, Potter (2015) argued that peo- ple’s lived experiences are affected by their gender and racial identities and that these identities, and others, interact with one another to affect one’s experiences and how they are perceived by others. Thus, it is imperative that future redemption research explore the intersectionality of race and gender by examining whether points of redemption and factors that affect who makes it to the point of redemption vary for women of different races as well as men of different races.

Fourth, future research should continue to explore the possibility that different offender groups exhibit different patterns of desistance, especially over the long term. For example, the traditional definition of desistance found in the criminology litera- ture is that an offender instantaneously and permanently desists from crime (Blumstein, Cohen, Roth, & Visher, 1986). A more recent definition of desistance allows the offending rate to vary over a criminal career, emphasizing gradual behavior change (Laub & Sampson, 2003). Some studies have investigated the possibility of different trajectories in crime desistance for different offenders. For example, using an 18-year follow-up study of convicted felony offenders in Essex County, New Jersey, Kurlychek, Bushway, and Brame (2012) studied the nature of desistance and recidivism and found that high-risk offenders reoffend rather quickly and low-risk offenders fail much more slowly or not at all. If this is the case, a declining hazard rate may not be an indication of desistance but instead may capture a strong grouping process where those with varying propensities to commit crime reoffend at different time points. In addition, the effects that were found in the logistic regression analysis may, in the same way as a declining hazard rate, be spurious.

Finally, other potentially relevant factors that change for better or worse over time, such as quality and stability of adult relationships and gainful employment, were not possible to capture in our analyses but should be considered in future redemption research. Future research in this area should continue to employ large-scale represen- tative samples to examine how factors such as having dependent children (including custodial children and noncustodial children), the quality of bond or relationship with their children, offense type, age at first conviction, race and ethnicity, along with other transitional circumstances suggested in life course research (Laub & Sampson, 2003) such as being married and employment histories, influence points of redemption and how these factors may influence males and females differently. Furthermore, future

198 Feminist Criminology 13(2)

redemption studies should continue to examine milestones that research shows are particularly important for female desistance, such as stable relationships and, more broadly, increased human, social, and state capital (Giordano et al., 2002; Holtfreter et al., 2004; Salisbury & Van Voorhis, 2009).

Policy Implications

Prior research has noted the importance of studies that utilize split sample analyses to determine whether independent variables suggested by predominant criminological theories exert unique influences on males and females and how the findings of these studies can be instrumental in informing gender-responsive policies and practices (Huebner et al., 2010; Wattanaporn & Holtfreter, 2014). Redemption research is important not only for furthering our understanding of processes related to desistance but also for informing policies relevant to the employment of persons with a history of involvement with the criminal justice system. It is often considerably difficult for indi- viduals with a criminal record to find gainful employment. Prior research consistently shows that many employers today are unwilling to hire individuals with a criminal record even if they surpass the job qualifications for a position (Pager, 2003; Sampson & Laub, 2003; Uggen & Staff, 2004). Similarly, research consistently shows that peo- ple with a criminal record experience more difficulty in securing and maintaining employment than any other disadvantaged group (minorities, welfare recipients, ille- gal aliens; Holzer et al., 2003).

Given what we know about the importance of gainful employment in promoting crime desistance, it is ultimately in the best interest of ex-offenders and society to reduce barriers to employment for males and females with a criminal record, espe- cially those individuals who made a mistake in the distant past but are now trying to live a lawful, productive life.

Understanding desistance from a redemption perspective can offer useful insight to policy makers that must decide on practices regarding criminal record checks by employers. For example, the findings of redemption research have significant implica- tions to better inform states on appropriate sealing waiting periods and/or offer guide- lines to employers about the diminished value of old criminal records as a measure of future risk of offending (Blumstein & Nakamura, 2009).

Furthermore, for those making employment decisions concerning women with a criminal record, our study shows redemption comes much sooner than for men, and therefore policies related to using criminal records in employment decisions should be considered separately for men and women. In fact, prior redemption researchers con- tend that a “one-size-fits-all” approach to sealing or expunging criminal records is not recommended (Bushway et al., 2011). Instead, policies on sealing and expunging criminal records should take into consideration offense type (Blumstein & Nakamura, 2009; Kurlychek et al., 2007), age at first offense (Blumstein & Nakamura, 2009; Bushway et al., 2011), prior criminal record (Bushway et al., 2011), and, from the present study, gender. Not doing so will unnecessarily disadvantage property offend- ers, offenders with less extensive records, older offenders, and women not just in terms

Curcio et al. 199

of employment but also in terms of access to government benefits, housing, and stu- dent loans in some states for some offenses (Jacobs & Crepet, 2006).

In many states, people with felony drug convictions are banned for life from receiv- ing certain types of government assistance, and in many cases, these bans dispropor- tionately affect female offenders. For example, in 2009, 85.9% of adult Temporary Assistance for Needy Families (TANF) recipients were women; women are also about twice as likely as men to receive food stamp benefits at some point in their lives (Morin, 2013). Although a growing number of states “opt out” of this ban and allow those with drug convictions to receive assistance, as many as 180,000 women in the United States are subject to the lifetime ban on TANF (The Sentencing Project, 2011). Given that bans on TANF benefits and other government assistance programs have already had dramatic consequences for female offenders and their children, policy makers and legislators in states that maintain these bans should “opt out” of these bans to give female ex-offenders the resources they need to desist from crime and be pro- ductive, law-abiding members of society (Allard, 2002). If states are not willing to completely “opt out” of these bans, they should at least consider modifying the ban and limit ineligibility to a short, fixed period of time or by creating possibilities to regain eligibility (e.g., by participating in a substance abuse treatment program) rather than instituting lifetime bans (The Sentencing Project, 2013).

Finally, it is important to highlight that the present study utilized the most conserva- tive approach to calculating redemption points. Arrests were also used as a measure of redemption, which are much more common than convictions. Thus, policy implications of future studies that utilize less conservative approaches and convictions as opposed to arrests may likely find even lower points of redemption and even more dramatic differ- ences in points of redemption for males and females to better inform policy.

Conclusion

In the present study, we added to the small but growing body of redemption research using a recent, nationally representative sample of individuals born in the United States to calculate redemption points and be the first to examine possible gender dif- ferences in points of redemption. In addition, we also explored possible factors that affect whether baseline offenders made it to the point of redemption, including multi- ple arrests at baseline, education variables, and whether the respondent had dependent children and whether these factors differ for males and females.

Overall, this study does highlight the importance of considering males and females separately in redemption research. As noted previously, redemption research is still in the early stages. The results of this study are consistent with prior literature and further suggest that gender does matter on the road to redemption. Females get there much sooner, but more research is needed to understand why. When examining factors that affect whether a former offender will make it to the point of redemption, the multivari- ate analysis showed that the only variable that showed consistent influence across gender was having multiple arrests at baseline, and this was consistent with other redemption studies that consider multiple offenses at baseline (Bushway et al., 2011).

200 Feminist Criminology 13(2)

Redemption research is important not only for furthering our understanding of pro- cesses related to desistance but also for informing policies and laws relevant to the collateral consequences of a criminal record.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

1. The term redemption was first applied to this question by Blumstein and Nakamura (2009). 2. The 1997 National Longitudinal Survey of Youth (NLSY97) is a longitudinal data set with

14 rounds of follow-up data (1997-2011) available as of November 2014, making it very suitable to examine cause-and-effect relationships between factors such as criminal records and subsequent employment outcomes. Besides labor market variables (employment, unemployment, wages, fringe benefits, etc.), the NLSY97 contains detailed information on a wide variety of subjects, including demographic information, education, vocational train- ing, household, neighborhood and geographical variables, family background variables, criminal or delinquent activity, and crime victimization.

3. As noted previously, prior redemption research has shown that points of redemption can vary greatly based on the number of prior arrests or convictions at baseline. In a prelimi- nary analysis, three baseline groups were created, including (a) age 18-, 19-, and 20-year- old nonoffenders; (b) age 18-, 19-, and 20-year-old offenders with one arrest only; and (c) age 18-, 19-, and 20-year-old offenders with two or more arrests. However, there were too few cases when running separate analyses for males and females.

4. The NLSY97 collects yearly data on the number of biological children living with the respondent as well as yearly data on the number of any biological children of the respon- dent (whether they reside with the respondent or not). For the present study, any biological children of the respondent (whether they reside with the respondent or not) were included.

5. Education was over the course of the study. Given that it takes time to complete college degrees, attending school over the course of the study suggests a commitment to pursuing goals which may influence whether or not a person makes it to redemption. Because we do not have a measure of school enrollment for each year, we calculated education measures for the entire study period.

6. Recently, scholars have noted that researchers should refrain from using the term “non- White” as it has negative connotations and instead should use “people of Color,” as it is the most current accepted term (see Potter, 2015).

7. It is also important to note that the respondent’s census region was initially also included as a control variable in preliminary analyses. The NLSY97 includes four census regions, including Northeast, North Central, South, and West. Dummy census region variables for the Northeast, the North Central, and the West with the South region as the reference cat- egory were included in preliminary analyses and were not significant in any of the models so they were removed from the analyses.

Curcio et al. 201

8. The respondent’s first offense type was included as a control variable in preliminary analyses. The NLSY97 asks respondents whether they have been convicted of the following offenses: assault (includes simple assault, battery, rape, and aggravated assault), robbery, burglary/ breaking and entering, theft (includes auto theft, larceny, and shoplifting), destruction of property (includes vandalism, arson, and malicious destruction of property), other property offenses (fencing, receiving stolen property, and possession or sale of stolen property), posses- sion of illicit drugs, sale of illicit drugs, major traffic offense (includes driving under the influ- ence [DUI], reckless driving, and driving without a license), and public order offense (includes disorderly conduct, drinking or purchasing alcohol underage, and sex offenses). These crime types were collapsed into the following four categories: violent crimes, property crimes, drug crimes, and other offenses (combined public order offenses and major traffic offenses and other offense categories). Dummy variables were created for violent crime, property crime, and drug crime, and the reference category used in the analysis is other offenses. None of these crime-type variables were significant in any of the models so they were removed from the logistic regression analyses. Small sample size precluded crime-type redemption trajectories.

9. When conducting separate hazard rate analyses by gender, the results show that male base- line offenders reach the point of redemption after 10 years, whereas female baseline offend- ers reach the point of redemption after only 4 years. Therefore, the dependent variable was measured differently for the three separate binary logistic regression models for the full sample and the separate binary logistic regression models for males and females separately. For the full sample and the males-only model, the dependent variable was the following: 0 = not redeemed (baseline offender with years to rearrest <10) and 1 = redeemed (baseline offender with years to rearrest ≥10). For the females-only model, the dependent variable was the following: 0 = not redeemed (baseline offender with years to rearrest <4) and 1 = redeemed (baseline offender with years to rearrest ≥4).

10. Portes (1998) defines social capital as “the ability to secure benefits by virtue of member- ship in social networks or other social structures” (e.g. access to employment). Relatedly, human capital refers to the internal knowledge, skills and capabilities that allow individu- als to act in new ways (Coleman, 1988). State capital refers to state-sponsored programs and services, such as services related to education, healthcare, housing and job training (Holtfreter, 2004).

11. All follow-up interviews with NLSY97 respondents are conducted using a computer- assisted personal interview (CAPI), and during sensitive portions of the interview, the respondents enter their answers directly into the laptop rather than interacting with the interviewer, which allows respondents to be more assured that answers to sensitive ques- tions will remain private. This self-administered portion, called audio computer-assisted self-administered interview (ACASI), includes an audio option so that respondents can listen to the questions and answers being read via headphone if they prefer. Furthermore, when a respondent is reluctant or unable to be interviewed in person due to their location, interviews are conducted by phone (U.S. Department of Labor, 2014).

References

Allard, P. (2002). Life sentences: Denying welfare benefits to women convicted of drug offenses. Washington, DC: The Sentencing Project.

Alpert, G., Dunham, R., Stroshine, M., Bennett, K., & MacDonald, J. (2006). Police officers’ decision making and discretion: Forming suspicion and making a stop. Washington, DC: U.S. Department of Justice.

202 Feminist Criminology 13(2)

Batiuk, M., Lahm, K., McKeever, M., Wilcox, N., & Wilcox, P. (2005). Disentangling the effects of correctional education: Are current policies misguided? An event history analy- sis. Criminal Justice, 5, 55-74.

Belknap, J. (2007). The invisible woman: Gender, crime, and justice. Belmont, CA: Cengage Learning.

Blitz, C. (2006). Predictors of stable employment among female inmates in New Jersey: Implications for successful reintegration. Journal of Offender Rehabilitation, 43, 1-22.

Blumstein, A., Cohen, J., Roth, J., & Visher, C. (Eds.). (1986). Criminal careers and career criminals (Vols. 1 and 2). Washington, DC: National Academy Press.

Blumstein, A., & Nakamura, K. (2009). Redemption in the presence of widespread criminal background checks. Criminology, 47, 327-360.

Bushway, S., Nieuwbeerta, P., & Blokland, A. (2011). The predictive value of criminal back- ground checks: Do age and criminal history affect time to redemption? Criminology, 49, 27-60.

Carson, E. A., & Golinelli, D. (2013). Prisoners in 2012: Trends in admissions and releases, 1991-2012 (Special report). Washington, DC: Bureau of Justice Statistics, U.S. Department of Justice.

Case, P., & Fasenfest, D. (2004). Expectations for opportunities following prison education: A discussion of race and gender. Journal of Correctional Education, 55, 24-39.

Chesney-Lind, M. (1997). The female offender: Girls, women and crime. Thousand Oaks, CA: Sage.

Cobbina, J. E. (2010). Reintegration success and failure: Factors impacting reintegration among incarcerated and formerly incarcerated women. Journal of Offender Rehabilitation, 49, 210-232.

Coleman, J.S. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94, 95-121.

Daly, K. (1998). Gender, crime and criminology. In M. Tonry (Ed.), The Oxford handbook of crime and justice (pp. 85-108). Oxford, UK: Oxford University Press.

Durose, M., Cooper, A., & Snyder, H. (2014). Recidivism of prisoners released in 30 states in 2005: Patterns from 2005 to 2010 (Special report). Washington, DC: Bureau of Justice Statistics, U.S. Department of Justice.

Ehrensaft, M., Khashu, A., Ross, T., & Wamsley, M. (2003). Patterns of criminal conviction and incarceration among mothers of children in foster care in New York City. New York, NY: Vera Institute of Justice.

Frost, N., Greene, J., & Pranis, K. (2006). Hard hit: The growth in the imprisonment of women, 1977-2004. Institute on Women and Criminal Justice. Retrieved from http://csdp.org/ research/HardHitReport4.pdf

Giordano, P. C., Cernkovich, S. A., & Rudolph, J. L. (2002). Gender, crime, and desistance: Toward a theory of cognitive transformation. American Journal of Sociology, 107, 990-1064.

Graham, J., & Bowling, B. (1995). Young people and crime. London, England: Home Office.

Heimer, K. (2000). Changes in the gender gap in crime and women’s economic marginalization. Criminal Justice, 1, 427-483.

Holtfreter, K., Reisig, M. D., & Morash, M. (2004). Poverty, state capital, and recidivism among women offenders. Criminology & Public Policy, 3, 185-208.

Holzer, H., Raphael, J., & Stoll, M. (2003). Employment barriers facing ex-offenders. New York, NY: The Urban Institute.

Curcio et al. 203

Huebner, B. M., DeJong, C., & Cobbina, J. E. (2010). Women coming home: Long-term pat- terns of recidivism. Justice Quarterly, 27, 225-254.

Jacobs, J., & Crepet, T. (2006). The expanding scope, use, and availability of criminal records. Legislation and Public Policy, 11, 177-213.

Kirk, D. (2008). The neighborhood context of racial and ethnic disparities in arrests. Demography, 45, 55-77.

Kreager, D. A., Matsueda, R. L., & Erosheva, E. A. (2010). Motherhood and criminal desistance in disadvantaged neighborhoods. Criminology, 48, 221-258.

Kurlychek, M., Brame, R., & Bushway, S. (2006). Scarlet letters and recidivism: Does an old criminal record predict future offending? Criminology & Public Policy, 5, 483-522.

Kurlychek, M., Brame, R., & Bushway, S. (2007). Enduring risk? Old criminal records and prediction of future criminal involvement. Crime & Delinquency, 53, 64-83.

Kurlychek, M., Bushway, S., & Brame, R. (2012). Long-term crime desistance and recidivism patterns—Evidence from the Essex County convicted felon study. Criminology, 50, 71-103.

Laub, J., & Sampson, R. (2003). Shared beginnings, divergent lives: Delinquent boys to age 70. Cambridge, MA: Harvard University Press.

Mallicoat, S., & Ireland, C. (2014). Women and crime: The essentials. Los Angeles, CA: Sage. Michalsen, V. (2011). Mothering as a life course transition: Do women go straight for their

children? Journal of Offender Rehabilitation, 50, 349-366. Miller, J. (n.d.). Writing about hazard analysis. New Brunswick, NJ: Rutgers University

Institute for Health, Health Care Policy and Aging Research. Moore, W., Pedlow, S., Krishnamurty, P., & Wolter, K. (2000). National Longitudinal Survey

of Youth 1997 (NLSY97) (Technical sampling report). Chicago, IL: National Opinion Research Center.

Morin, R. (2013). The politics and demographics of food stamp recipients. Pew Research Center. Retrieved from http://www.pewresearch.org/fact-tank/2013/07/12/the-politics- and-demographics-of-food-stamp-recipients/

Owen, B. (1998). In the mix: Struggle and survival in a women’s prison. Albany: State University of New York Press.

Pager, D. (2003). The mark of a criminal record. American Journal of Sociology, 108, 937-975. Paternoster, R., Brame, R., Mazerolle, P., & Piquero, A. (1998). Using the correct statistical test

for the equality of regression coefficients. Criminology, 36, 859-866. Pew Research Center. (2011). State of recidivism: The revolving door of America’s prisons

(NCJ Publication No. 234180). Washington, DC: Author. Potter, H. (2015). Intersectionality and criminology: Disrupting and revolutionizing studies of

crime. New York, NY: Routledge. Portes, A. (1998). Social capital: Its origins and application in modern sociology. Annual

Review of Sociology, 24, 1-24. Salisbury, E. J., & Van Voorhis, P. (2009). Gendered pathways: A quantitative investigation of

women probationers’ paths to incarceration. Criminal Justice and Behavior, 36, 541-566. Sampson, R., & Laub, J. (2003). Life-course desisters? Trajectories of crime among delinquent

boys followed to age 70. Criminology, 41, 319-339. Schwartz, J. (2013). A “new” female offender or increasing social control of women’s behav-

ior? Cross-national evidence. Feminist Studies, 39, 790-821. The Sentencing Project. (2011). A lifetime of punishment: The impact of the felony drug ban on

welfare benefits. Retrieved from www.sentencingproject.org/doc/publications/cc_A%20 Lifetime%20of%20Punishment.pdf

204 Feminist Criminology 13(2)

The Sentencing Project. (2013). Incarceration. Washington, DC: Author. Sherman, L. (1980). Causes of police behavior: The current state of qualitative research. Journal

of Research in Crime & Delinquency. 17, 69-100. Singer, J., & Willett, J. (2003). Applied longitudinal data analysis: Modeling change and event

occurrence. New York, NY: Oxford University Press. Snell, T., & Morton, D. (1994). Women in prison (Special report). Washington, DC: Bureau of

Justice Statistics, U.S. Department of Justice. Soothill, K., & Francis, B. (2009). When do ex-offenders become like non-offenders? The

Howard Journal of Criminal Justice, 48, 373-387. Stevens, D. J., & Ward, C. S. (1997). College education and recidivism: Educating criminals is

meritorious. Journal of Correctional Education, 48, 106-111. Uggen, C., & Staff, J. (2004). Work as a turning point for criminal offenders. In J. L. Krienert &

M. Fleisher (Eds.), Crime and employment: Critical issues in crime reduction for correc- tions (pp. 141-166). Walnut Creek, CA: AltaMira Press.

Ulmer, J. T. (2001). Intermediate sanctions: A comparative analysis of the probability and severity of recidivism. Sociological Inquiry, 71, 164-193.

U.S. Department of Labor, Bureau of Labor Statistics. (2014). National Longitudinal Survey of Youth 1997 cohort, 1997-2011 (rounds 1-15). Chicago, IL: National Opinion Research Center, University of Chicago (Producer). Columbus: Center for Human Resource Research, The Ohio State University (Distributor).

Vacca, J. S. (2004). Educated prisoners are less likely to return to prison. Journal of Correctional Education, 55, 297-305.

Wattanaporn, K. A., & Holtfreter, K. (2014). The impact of feminist pathways research on gender-responsive policy and practice. Feminist Criminology, 9, 191-207.

Worden, R. (1989). Situational and attitudinal explanations of police behavior: A theoretical appraisal and empirical assessment. Law & Society Review, 23, 667-711.

Author Biographies

Gina Curcio, PhD, is a recent graduate of the PhD Program in Criminology and Justice Studies at the University of Massachusetts Lowell. She has taught courses in corrections, criminal investigations, gender and crime, race and crime, and critical issues in criminal justice. Her previous work experience includes working as a correctional officer and as a domestic violence advocate. Her primary research interests include offender reentry, criminal records and policy reforms, and gender-responsive programs for female offenders.

April Pattavina, PhD, is a senior research scientist at the Wellesley Centers for Women at Wellesley College and an associate professor in the School of Criminology and Justice Studies at the University of Massachusetts Lowell. Her research areas include the theoretical, policy, and social implications of correctional practices and justice system responses to women who are victims of violent crime. Her articles have appeared in journals such as the Journal of Criminal Law and Criminology, Violence Against Women, Crime & Delinquency, and Medicine, Science and the Law.

William Fisher, PhD, is a professor in the School of Criminology and Justice Studies at the University of Massachusetts Lowell. His research interests include mental health factors in the criminal justice system and factors affecting recidivism. He has published in Crime & Delinquency and numerous journals focused on law and mental health.