Content uploaded by Nathan W Link
Author content
All content in this area was uploaded by Nathan W Link on Oct 13, 2017
Content may be subject to copyright.
Longitudinal Associations among Child Support Debt,
Employment, and Recidivism after Prison
Nathan W. Link and Caterina G. Roman
Department of Criminal Justice, Temple University, Philadelphia, Pennsylvania, USA
ABSTRACT
Recently released prisoners in the United States are increasingly facing
the burden of financial debt associated with correctional supervision,
yet little research has pursued how—theoretically or empirically—the
burden of debt might affect life after prison. To address this gap,
we employ life course and strain perspectives and path analysis to
examine the impact of child support debt on employment and recidi-
vism, using longitudinal data from an evaluation of a prisoner reentry
program known as the Serious and Violent Offender Reentry Initiative.
Results indicate that having more debt has no effect on recidivism;
however, more debt was significantly associated with a decrease in
later legitimate employment. Implications for community reintegra-
tion and justice processing are discussed within the framework of past
and emerging work on legal financial obligations, employment, and
desistance from crime after prison.
KEYWORDS
Collateral consequences;
criminal justice debt; legal
financial obligations;
prisoner reintegration;
reentry
Each year roughly 640,000 inmates are released from state and federal prisons in the
United States (Carson and Golinelli 2013). Recidivism data from the U.S. Bureau of
Justice Statistics (BJS) show that over three-quarters of prisoners released from state
institutions will be rearrested within five years (Durose, Cooper, and Snyder 2014).
Some of the largest and most comprehensive studies to date on “what works”to lower
recidivism and increase employment have shown prisoner reentry interventions to be of
modest effectiveness (Lattimore and Visher 2013; Lattimore et al. 2012). This has led
some scholars to expand their focus from the much-studied domains of services for
employment, mental health, and housing, to include an examination of how correctional
policies might influence recidivism and employment. This focus on policy has engen-
dered discussion about how criminal justice–related and other financial obligations
might impact the community reintegration process as well as a prisoner’s readiness for
and access to services and supported reentry opportunities.
Released prisoners often face substantial financial burdens. Moreover, there is evidence that
the scale of debt in recent years among criminal justice populations is unprecedented (Bannon,
Nagrecha, and Diller 2010). A Maryland study found that, upon release, nearly two-thirds (62
percent) of respondents reported having legal/financial debt related to the criminal justice
system (Visher, La Vigne, and Travis 2004). The sources of this debt include—but are not
limited to—fines, court fees, treatment fees, law enforcement fees, and child support orders.
CONTACT Nathan W. Link Nathan.link@temple.edu Department of Criminal Justice, Temple University, 517
Gladfelter Hall, Philadelphia, PA 19122, USA.
THE SOCIOLOGICAL QUARTERLY
2017, VOL. 58, NO. 1, 140–160
http://dx.doi.org/10.1080/00380253.2016.1246892
© 2017 Midwest Sociological Society
Child support obligations are of particular significance for criminal justice populations
(Roman and Link 2015). According to the BJS, the majority of state and federal prisoners are
parents of children under the age of eighteen and their data suggest that 46 percent of fathers
with minor children are noncustodial parents (Glaze and Maruschak 2008). State-based
studies that link correctional data with child support enforcement data estimate that between
13 percent and 24 percent of released prisoners in the United States owe more than $400 per
month in child support (Griswold et al. 2004). Further, a 2004 study of fathers in the
correctional system in Massachusetts found that the median total for child support debt
across state and local prisoners was about $10,000 (Pearson 2004). The few other existing
studies attempting to quantify the child support–related debt of prisoners corroborate these
estimates (Griswold, Pearson, and Davis 2001; Ovwigho, Saunders, and Born 2005).
Contributing to the burdens rendered by child support debt is the fact that an order often
continues to accrue throughout a prisoner’s incarceration (Hairston 2002). Indeed, Pearson
(2004) found that virtually every prisoner with a child support order owed at least some
amount of “back due”support. Lattimore et al. (2012) found that 92 percent of participants
who entered prison with a child support order owed some amount of back due support.
Qualitative sociological research and anecdotal evidence suggest that child support
obligations and other correctional debt from fines and fees can create significant, collateral
barriers to successful reintegration (Richards and Jones 2004). Because returning prisoners
often have to pay large portions of their income to government agencies and/or the
mothers of their dependent children (up to 65 percent), scholars have hypothesized that
official child support policies reduce incentives to work (McLean and Thompson 2007)
and could push ex-prisoners into the underground economy. Furthermore, the conse-
quences of failure to pay legal financial obligations may be great. A Washington state
study found that one-fourth of returning prisoners who owed debt reported that an arrest
warrant had been issued because of failure to pay, and most were subsequently incarcer-
ated for nonpayment (Harris, Evans, and Beckett 2010).
Despite the increased role that child support systems have played in the lives of former
prisoners, no large-scale quantitative studies have examined how legal financial obligations
associated with child support influence key outcomes in the reintegration process. The
current work begins to address this empirical gap using longitudinal data from the
multistate Serious and Violent Offender Reentry Initiative (SVORI). Relying on panel
data, we build on the limited prior literature by examining whether child support obliga-
tions impact changes in employment and recidivism. We begin by reviewing the emerging
literature on legal financial obligations associated with criminal justice populations, with
particular emphasis on child support, and discuss how theory informs the relationship
between this parental responsibility and employment and recidivism after prison.
Background
The Intersection of Child Support Debt, Incarceration, and Prisoner Reintegration
In recent years there has been a dramatic increase in the application of legal financial
obligations to criminal justice defendants (Beckett and Harris 2011; Harris et al. 2010;
Levingston and Turetsky 2007). Under this umbrella are fines, restitution, child support
orders and arrears, and fees and costs. Fines are punitive and are applied during the court
THE SOCIOLOGICAL QUARTERLY 141
process. Restitution is reimbursement that defendants pay to victims for damage caused.
Child support orders are court-set payments a parent is required to pay to the custodian of
his/her dependent child(ren) on a regular basis. These orders often go unmodified while
the person is incarcerated, which results in child support arrears—the sum of outstanding
child support orders.
Regarding child support arrears, many former prisoners are confronted with large sums
upon reentry into the community (Cammett 2010;MincyandSorensen1998;Ovwigho,
Saunders, and Born 2005; Pearson 2004;RomanandLink2015; Sorensen 1997).This is largely
the result of three factors. First, the Bradley Amendment of President Clinton’sPersonal
Responsibility and Work Opportunity Reconciliation Act of 1996 (PRWORA, otherwise known
as “Welfare Reform”) stipulated that child support orders could not be modified retroactively
under any conditions. However, states were given wide discretion in determining whether and
how to modify child support orders prospectively. The relevant legal principle applied by the
states in this process is whether “substantial changes in earning capacity”have occurred.
Unemployment qualifies as one of these substantial changes, unless the unemployment is
considered “voluntary”such as quitting a job. An issue of much debate across the states is
whether incarceration should qualify as a form of voluntary unemployment. Whereas most
states have decided that incarceration is indeed a form of involuntary unemployment, many
have deemed that it is voluntary because the conduct that resulted in prison confinement was
voluntary (Cammett 2010; Miller and Mincy 2012; Pearson 2004). As such, prisoners in those
states cannot have their child support orders modified prospectively to reflect changes in their
earning capacity while they serve their sentences.
Second, even within the jurisdictions that do allow prospective child support modifica-
tion, few inmates actually have their orders modified (Roman and Link 2015). The reason
for this lack of modification is twofold: first, many inmates are simply not aware that their
orders can be modified, and second, even those who are aware may lack the resources and
legal knowledge required to initiate the child support modification process (Cammett
2010; Pearson 2004; Sorensen and Oliver 2002).
Third, in about half of the states, people who have outstanding child support
balances—including former prisoners—are taxed on their child support arrears.
Because these arrears are usually large and the taxes compound over time, already
large debt burdens often increase dramatically in the few years after release. In
California, using administrative data on obligors (i.e., noncustodial parents who owe
child support), Sorensen (2004) found that taxes levied specifically on these arrears
represented the largest contributor to escalating debt burdens.
After incarceration, former prisoners are responsible for repaying their debts, usually
through probation, parole, or child support enforcement offices. In addition to the
Massachusetts study cited earlier that showed median child support debt among prisoners
to be about $10,000, a Maryland study that examined the intersection of incarceration and
child support by choosing a random sample of noncustodial fathers with child support
orders found that of the subsample that was incarcerated at the time of the study, arrears
ranged from $552 to $70,305 (Ovwigho et al. 2005). For the formerly incarcerated
subsample, median arrears were $11,554, and ranged from $32 to $108,394.
There is a real need for children to receive financial support from their previously
incarcerated fathers. However, a large proportion of the returning prisoner population
consists of low income earners with little prospect of securing well-paying employment,
142 N. W. LINK AND C. G. ROMAN
so expecting large debt burdens accrued during incarceration to be paid in full upon release
has been described as unrealistic. As such, it is essential to strike an appropriate balance
between providing for dependent offspring and at the same time reducing the likelihood that
unmanageable orders will adversely affect the returning prisoner. Such potential negative
effects on the former prisoner include creating incentives to return to crime, implementing
punitive measures for nonpayment such as driver’s license suspension (which could hinder
employment opportunities), or returning the obligor to incarceration in the instance of late
or nonpayment (Bannon et al. 2010;Holzer,Offner, and Sorensen 2005).
Theoretical Support
Although the relationship between child support obligations and recidivism has received
little attention in the sociological or criminological literature, there are tenable theoretical
frameworks for why having child support obligations and related debt might influence
both recidivism and desistance from crime. Some theories lead to the suggestion that there
might be an inverse relationship between child support and recidivism, where the obliga-
tion can act as a protective factor against continued offending; other theories suggest that
it will increase the likelihood of continued offending.
Life course theories (Sampson and Laub 1995), which emphasize the factors implicated
in crime continuity and desistance beyond adolescence, offer a few relevant principles and
might suggest that legal financial obligations can encourage successful reintegration. By
design, child support systems legally bind former prisoners to their dependent families. By
virtue of having this formal arrangement in place, family relations can be enhanced by
giving the former prisoner the opportunity to foster increased parental involvement with
and attachment to family and their attendant needs (Laub and Sampson 2003). At a
minimum, having a child support order can serve as a perennial psychological reminder of
one’s parental role and responsibilities. Indeed, Seltzer, McLanahan, and Hanson (1998)
found that requiring parents to pay child support increased parental involvement between
the paying fathers and their dependent children. It is through this increased involvement
that family bonds can heal or be strengthened, which may in turn facilitate desistance. But
fewer than a handful of studies have addressed these linkages empirically. Former prison-
ers have offered historical narratives indicating that family and parenting responsibilities
acted as a “turning point”strong enough to redirect their lives in more promising
directions (Laub and Sampson 2003). Certainly, becoming a parent brings a host of new
responsibilities and changes to routine activities. In this context, former prisoners’short-
and long-term goals may shift in ways that address their immediate and long-term
parental responsibilities—goals that are compatible with the process of desistance from
crime, such as obtaining steady employment (Laub and Sampson 2003). Endeavoring to
find stable, legitimate employment not only allows former prisoners to financially support
their dependent family members, it also structures their routine activities in ways that are
likely less antisocial than before. Taken together, employment responsibilities and the day-
to-day demands of parental support can shape one’s routine so profoundly that the
opportunity to reoffend can be greatly minimized.
On the other hand, strain theory (Merton 1968) might suggest that being required to
pay what could amount to hefty child support payments back to the state could create or
exacerbate strain, during what is already known to be a transition marked by intense
THE SOCIOLOGICAL QUARTERLY 143
material and emotional hardship (Western et al. 2015). Growing child support debts, as a
financial strain (Agnew 2006), might serve as the impetus to “push”or motivate people to
offend, possibly in the form of revenue-generating or acquisitive crimes, especially if those
activities are perceived as being able to counteract or mitigate the discomfort or negative
feelings associated with the particular strain (Hairston 2002). With the threat of incarcera-
tion for debt nonpayment looming, returning to crime to make ends meet may appear to
be the rational option.
Within this context, more recent developments in strain theory suggest that certain
family forces may moderate or offset the negative, perhaps criminogenic, impact of debts
associated with the criminal justice system (Agnew 2006). If, for example, family support
can protect against reoffending behavior, and the driver of the offending behavior is
overwhelming financial strain, tangible support in the form of financial assistance will
theoretically reduce the harmful and criminogenic effects of strain. Viewed from another
angle, if a former prisoner receives family financial support, this may also reduce the
appeal of certain revenue-generating illegal behaviors. Though instrumental family sup-
port has been mentioned as consequential in recent works (Nagrecha, Katzenstein, and
Davis 2015; Visher, La Vigne, and Travis 2004), its role among former prisoners is
underdeveloped both theoretically and empirically. But, if in the eyes of former prisoners,
this financial strain is associated with their families, it could damage relationships further,
weakening the informal control of ties to family (Nagrecha et al. 2015).
This process might occur in one of two ways. First, families of former prisoners who
provide instrumental support—perhaps not financially secure themselves—may strug-
gle with supporting their recently released family member. In this instance, family
tension may arise if the family perceives that they have been given too much owner-
ship of their formerly incarcerated family member’s debt obligation. Second, for some
families, providing financial support to their recently released family member may not
be an option. In these cases, relations between the family and former prisoner may
deteriorate if the former prisoner perceives that his or her family cannot be relied on
or is not supportive during the precarious transition home. Such family tension and
conflict may reduce the family’s informal control capacity over their formerly incar-
cerated family member. In an Australian sample of released prisoners, Martire et al.
(2011) noted that 60 percent of their sample reported that their debt adversely affected
their relationship with their partner; the same percentage reported that it hurt their
family relationships.
A Pennsylvania study found that parole violators reported having a harder time making
ends meet because of debt than those who did not violate parole. However, there were no
significant differences in recidivism between those who had and those who did not have this
debt (Bucklen and Zajac 2009). Martire et al. (2011) reported descriptive statistics indicating
that debt associated with criminal justice was a perennial source of stress (64 percent
reported it as stressful). Thirteen percent of the sample cited debt as the motivating factor
for their last instrumental crime (Martire et al. 2011). Petersilia, Greenwood, and Lavin’s
(1977) well-known study of the criminal careers of habitual felons found that heavy debt
burdens were associated with a category of felons they termed “intensive”career criminals.
In addition, qualitative evidence has linked debt with instrumental crimes among property
offenders (Sutton 1995).
144 N. W. LINK AND C. G. ROMAN
With regard to employment, strain theory would suggest that former prisoners with debt
burdens might seek work in the underground economy. In fact, theorists from various dis-
ciplines suggest that rising child support obligations can lead to reductions in formal employ-
ment and labor force participation (Holzer et al. 2005; Miller and Mincy 2012;Pirog,Klotz,and
Byers 1998; Pirog and Ziol-Guest 2006). Pirog et al. (1998) demonstrated that child support
orders for economically disadvantaged fathers typically ranged from 20 percent to 35 percent of
their income. In addition, with payroll and other taxes factored in, this group’s marginal tax rate
can reach over 60 percent (Primus 2006). Should these fathers have outstanding payments,
states are within federal law to garnish up to 65 percent of their take-home pay. Because of these
stringent parameters, scholars have posited that noncustodial fathers facing these realities are
incentivized not to work. Rather, forgoing work in the mainstream economy for work in the
under-the-table economy—where their incomes will not be detected—may seem to be a
favorable alternative. By not working in the formal economy, the noncustodial parent can
earn money tax free and use more of it to support him- or herself and his or her family, while
steering clear of the threat of fiscal government intervention (McLean and Thompson 2007).
Legal financial debt might affect employment via other important structural barriers
common in the prisoner reintegration process. Studies have shown that having a criminal
record can make finding employment very difficult for former prisoners (Pager 2007).
Advocacy groups have attempted to reduce this barrier by expunging stale criminal
records. Scholars have contributed to this effort by showing that sufficiently old convic-
tions are not useful for predicting future criminality (Blumstein and Nakamura 2009). The
relevant issue here is that, in some jurisdictions, policy prohibits criminal record expun-
gement for former prisoners who still have outstanding child support debt (Vallas and
Patel 2012). In addition, in many jurisdictions the first penalty for nonpayment of child
support is driver’s license suspension (Bannon et al. 2010; Cammett 2010; Levingston and
Turetsky 2007). This could affect employment by severely limiting the pool of jobs feasible
for an individual without a license.
However, viewed through the lens of the life course paradigm, having a child support
obligation could be positively related to employment after prison, as former prisoners’goals
may shift in response to their mounting financial burdens. As others have noted (Visher,
Debus-Sherrill, and Yahner 2011), having financial obligations could be a motivating factor
to find more or better employment, as former prisoners who have debt need to earn more to
keep up with both debt payments and typical daily expenses. This goal-shifting dynamic in
response to family financial obligations may be especially pronounced in jurisdictions where
former prisoners are routinely reincarcerated solely for their failure to repay debts or for not
repaying them on time. As it stands, current theory on child support obligations and
employment and recidivism is ambiguous—suggesting that they could be associated with
both more and less employment and reoffending. Empirically, the role of these debt burdens
in the lives of former prisoners remains an unanswered question.
The Current Study
With the dramatic, national-level growth in the child support system and its strict
enforcement since the PRWORA, coupled with recent theorizing on how legal financial
obligations affect former prisoners as the backdrop (Bannon et al. 2010; Cammett 2010;
Harris et al. 2010), the current study leverages a multistate, longitudinal sample of
THE SOCIOLOGICAL QUARTERLY 145
returning U.S. prisoners to investigate the following research questions: (1) What is the
effect of increasing child support debt on employment and recidivism among returning
prisoners? (2) Do relationships between child support debt, employment, and recidivism
vary over a 15-month post-release period from prison? (3) Since instrumental support
theoretically can offset the burdens associated with large debt obligations, do higher levels
of family instrumental support reduce recidivism?
Data and Methods
Data
Data used in these analyses, made available through the Interuniversity Consortium of Political
and Social Research (ICPSR), are from 1,011 adult male prisoners with children under the age of
18 that were part of the evaluation of the Serious and Violent Offender Reentry Initiative
([SVORI] Lattimore and Visher 2013). Subjects involved in the study had extensive criminal
histories, substance abuse problems, low involvement in the legitimate labor market, and
generally high levels of needs across a range of domains (Lattimore et al. 2012). Twenty-five
percent of the subjects were in prison most recently for violent offenses, 17 percent for property
offenses, 34 percent for drug offenses, and 24 percent for other offenses.
The SVORI impact evaluation study focused on 16 programs from the following 12
states: Ohio, Indiana, Iowa, Kansas, Maine, Maryland, Missouri, Nevada, Oklahoma,
Pennsylvania, South Carolina, and Washington. The SVORI programs in Iowa and
Ohio randomly assigned participants to treatment and control groups. The other 10 states
did not assign to treatment randomly, but along dimensions such as age, criminal history,
and risk level (Lattimore and Steffey 2009). The reentry programs focused on improving
employment, health, education, housing, and criminal justice outcomes for returning
prisoners. Between 2004 and 2007, participants in the SVORI treatment group received
services designed to address these areas of need. Services began in prison and continued
through release into the community. The data set is appropriate for the present study
because it represents a rare, multistate opportunity to examine associations between child
support obligations, employment, and recidivism over time.
Respondents were interviewed at four time points: approximately 30 days prior to their
release from correctional institutions, and 3, 9, and 15 months post release. At 3 months,
60 percent (603) were successfully reinterviewed; 61 percent (616) were interviewed at 9
months; and 66 percent (672) at 15 months.
1
Subjects who were incarcerated at follow-up
were interviewed in jail or prison.
Endogenous Variables
Employment was measured as a binary (yes/no) variable at each wave indicating if the
respondent supported himself via a legitimate job since the last interview. Respondents were
coded as “1”if they reported legitimate employment in response to the question: “How did you
support yourself since the last interview?”This operationalization is in line with much of the
research on employment among criminal justice populations (Skardhamar and Savolainen
2014). Baseline employment (legitimate employment six months prior to the instant incarcera-
tion) was used as a control in the longitudinal models.
146 N. W. LINK AND C. G. ROMAN
Recidivism was operationalized as rearrest, which was a (yes/no) binary outcome
measured at 3, 9, and 15 months using official arrest data from the National Crime
Information Center (NCIC). These administrative data were collected by the SVORI
researchers and they contain arrests recorded by the Federal Bureau of Investigation.
The strength of this measure is that, unlike self-reported crime, which suffers from a
degree of attrition, this outcome has little missing data. Of the 1,011 subjects in our
sample, rearrest data are available for 951 respondents, or 96 percent. For those respon-
dents who had missing data on rearrest, official time-varying data on reincarceration were
inserted into the rearrest variable and used as a proxy measure. Thus, our rearrest
outcome variable contained no missing cases.
2
The subjects in the study were released
between 2004 and 2006, and the data on these rearrests were gathered in 2008 and 2009.
Exogenous Variables
The amount of child support debt owed by respondents was measured as an ordinal scale
at each wave. The baseline question asked about the six months prior to the current
incarceration; the follow-up items asked what was owed at the time of each interview. The
scale ranged from “0”to “less than $1,000”(coded 1), “$1,000 to $1,999”(coded 2),
“$2,000 to $2,999”(coded 3), “$3,000 to $3,999”(coded 4), “$4,000 to $4,999”(coded 5),
and to “more than $5,000”(coded 6).
Family instrumental (or “tangible”) support was included as a theoretically important
covariate (Visher et al. 2011) measured at each follow-up wave as the sum of five items
probing the degree to which family members provided support related to: housing,
transportation, employment, substance abuse treatment, and finances. The item asked
respondents to reflect on the time since incarceration (for Wave 2) or “since the last
interview”(for Waves 3 and 4). Responses ranged from “strongly disagree”(0) to “dis-
agree”(1) to “agree”(2) to “strongly agree”(3). Thus, the SVORI-created scale ranged
from 0 to 15 with higher values indicating greater support (Lattimore and Steffey 2009).
Cronbach’s alpha was high (.89) at each wave.
The type of offense for which the respondent was currently serving a sentence (i.e., the
index incarceration) was measured with dichotomous variables “property offense”and
“violent offense,”with “drug/other offense”as the reference category. Age at first arrest, a
measure often found in reentry evaluation studies (Lattimore and Visher 2013), was also
included as a covariate to control for criminal justice risk in the rearrest model.
Supervision status (“on supervision”) measured if the respondent was on probation or
parole (yes/no) at each subsequent interview and was used in the rearrest model. The
number of days the respondent spent in prison prior to release was included as a
continuous variable. To control for variation that might be due to SVORI participation,
we created a dichotomous indicator of whether the respondent was part of the treatment
condition (SVORI participation). “Job services”was measured at each wave with the item:
“Have you received any educational or employment services in prison/since release/in the
last six months?”Similar to studies controlling for opportunity by including “time on the
street”in the analysis, reincarcerated status was controlled with dummy indicators to
identify respondents who, according to official records, were incarcerated at each follow-
up interview point and were therefore unable to be employed or rearrested.
3
All reincar-
cerated subjects were interviewed.
THE SOCIOLOGICAL QUARTERLY 147
Because past research has shown that physical health is often a significant predictor for
obtaining and retaining a job (Visher et al. 2011), we included a measure of physical
health limitations as a predictor in the paths to employment outcomes. “Physical health
problems”reflects the following items: “Does your health now limit you in moderate
activities—such as moving a table or playing basketball—a lot (2), a little (1), or not at all
(0)?”and “Does your health now limit you a lot (2), a little (1), or not at all (0) when
climbing several flights of stairs?”Thus, the variable ranged from 0 to 4 with higher scores
indicating worse health. Cronbach’s alpha is .81.
Child support laws and policies vary by state, with some states having more stringent
laws than others. To address this interstate variation, we created a scale (State CS law)
that measures how strict the respondent’sstateisvis-à-vispoliciesregardingorder
modifications in the event of an incarceration, where 0 = incarceration is sufficient to
justify modification, 1 = incarceration is considered among other factors for modifica-
tion, and 2 = incarceration is insufficient to justify modification. As such, higher values
correspond to increasing strictness.
4
The following demographics were measured at baseline and considered time
invariant. Race was measured using the dummy variable “African American”with
“White/other”asthereferencecategory.BecausetheSVORIsurveyprotocolasked
the question: “Which of the following best describes you? Please check all that apply,”
this variable represents anyone who chose “African American”only. An education
indicator measured whether the respondent completed high school or received a GED
(high school/GED). Since financial status could affect how burdensome a debt obliga-
tion might be, a continuous hourly wage measure was included as an indicator of
income. We originally modeled this variable as time-varying, but since the results
were no different with the more complete baseline data, we opted to make hourly
wage time-invariant. In addition, baseline data might better capture ability to pay in
that it represents wages before leaving prison with a felony conviction. Married/
partner was measured as a dichotomous variable—wherethevalueof“1”indicates
that the person was married or had a steady partner. This variable was measured as
time-varying to account for respondents who might change their marital status after
release from prison. Age was measured as chronological age at release from the
instant incarceration, but was ultimately removed from the core models because of
collinearity issues with our key family support variable, though it was retained as a
predictor in the Heckman selection equation.
Analysis
Because we are using longitudinal data and the key outcomes of interest are dichot-
omous (employment and rearrest), generalized structural equation modeling (GSEM)
in Stata 13 was used to estimate a cross-lagged path model. The panel nature of these
data is leveraged for two related reasons. First, items for child support debt, employ-
ment, rearrest, family support, and other key variables are time-variant, thus we can
more accurately capture how levels of our key variables influence the outcomes vari-
ables over time in the reentry process. Second, panel data analysis allows for previous
levels of key variables to be incorporated in the analyses. In this way, relationships
examined no longer reflect the effect of Xon Y, but rather Xon later changes in Y.
148 N. W. LINK AND C. G. ROMAN
Whileitisimpossibletoprecludethefactthatsomethird,unidentified, time-varying
variable is causing these changes in Y, this method marks a strong improvement over
cross-sectional methods that suffer potential endogeneity and time-ordering issues
(Wooldridge 2010).
In the current analyses, employment and recidivism outcomes are treated as endogen-
ous—affecting each other over time. The core model indicating the theoretical paths for
the two key outcomes (without the covariates) and the main predictor (amount of child
support debt) is shown in Figure 1. Following Wilson’s(1997) deindustrialization model
whereby unemployment leads to changes in routine activities and increased criminal
behavior, we assess the impact that employment has on recidivism (job →rearrest).
Within each wave, the cross-sectional impact of being employed is assessed on the like-
lihood of rearrest (paths a1, a2, and a3) for each of the waves post-baseline.
Longitudinally, the impact of being employed at one time point is assessed on changes
in rearrest at later points (paths b1, b2, and b3). Conversely, the impact of being rearrested
on the likelihood of later changes in employment status is analyzed (paths c1, c2, and c3).
These paths are motivated by theoretical and empirical research emphasizing the powerful
and stigmatizing forces behind official arrest, and its social consequences, such as the
impact on one’s likelihood of securing legitimate employment (Uggen, Manza, and
Thompson 2006). Our preliminary analyses (not shown here) revealed that child support
debt having lagged impacts on rearrest and employment fit the data much better than
models with child support debt affecting outcomes in the same wave. The Bayesian
information criterion (BIC) difference between the lagged versus same-wave models was
15 (with the lagged model having a lower BIC), indicating that the lagged modeling
approach fit the data better (Raftery 1995). As such, paths f1 through f3 and paths g1
through g3 in Figure 1 illustrate the conceptual associations between child support and
employment and rearrest, respectively. The models also include stability coefficients—the
paths for the association between each outcome and itself across waves (paths d1 through
Employment Employment Employment
Re-arrestRe-arrestRe-arrest
Baseline Time 2 Time 3 Time 4
Amount
Child
Support
Amount
Child
Support
Amount
Child
Support
Employment
a1 a2 a3
b2 b3
c1 c2
f1 f2 f3
g1 g2 g3
b1
d1 d2 d3
e1 e2
Figure 1. Child support, employment, and recidivism after prison. Baseline interviews conducted
approximately 30 days prior to institutional release. Time 2, Time 3, and Time 4 interviews conducted
at 3, 9, and 15 months, respectively.
THE SOCIOLOGICAL QUARTERLY 149
d3 and e1 and e2). It is important to note that with regard to employment and reincar-
ceration, the SVORI protocol carefully asks respondents who were reincarcerated (at each
interview): “After you were released but before you were reincarcerated, how did you
support yourself?”This phrasing helps establish whether a respondent held any job in that
post–instant incarceration period but before he was reincarcerated for a violation or a new
offense.
In addition to the variables found in the conceptual model in Figure 1 (i.e., employ-
ment, rearrest, and child support debt), the model includes a set of covariates theoretically
and empirically grounded in the desistance and reentry literature (see Table 1). The
independent variables instrumental family support, marital status, supervision status, job
services, physical health problems, and reincarcerated status are time-varying; type of
offense, hourly wage, SVORI-group assignment, the control for state variation, and all
demographics are time-invariant.
Missing data due to subject attrition is addressed in two complementary ways. First, the
Heckman selection model was used to address sample selection bias due to attrition at
follow-up waves (3, 9, and 15 months). Distinct from the popular two-step Heckman
correction that models the selection equation using probit regression, employs the pre-
dicted values from this model to calculate the inverse mills ratio (IMR) for each case, and
includes the IMR as a control in an ordinary least squares (OLS) model, the Heckman
selection model in GSEM uses latent variables and probit regression only (Statacorp 2013).
In this way, it is able to model dichotomous outcomes, unlike the two-step Heckman
correction (Bushway, Johnson, and Slocum 2007). Age at release was used as an exclusion
restriction predicting selection, as it was shown not to be significantly related to our
endogenous variables of interest. Second, GSEM in Stata 13 uses equation-wise deletion
rather than listwise deletion, which does not automatically drop cases that have some
missing data. Instead, this approach uses all of the data available to it when estimating
parameters (StataCorp 2013). For example, a respondent who was interviewed at baseline
and nine months only would be included in the analyses relevant for those time points,
and dropped from the equations where he had missing values. In this way, GSEM was able
to use some data from all but one respondent from the sample (n= 1,010).
Results
Results from the cross-lagged path model indicated some support for the hypothesis
that having a higher child support order reduces the likelihood that a respondent will
have or obtain legitimate employment. As shown in Table 2, an increase in the amount
of child support debt reported at Wave 3 was associated with a 6 percent decrease in
the odds of reporting legitimate employment by Wave 4 (p<.05).Regardingmarital/
partner status, being married at baseline and at Wave 4 was associated with increases
in the likelihood of reporting being employed in the same wave (OR = 1.39 and 1.42, p
<.05,respectively).
Though the pathway from employment to recidivism is more studied, results here
showed that rearrest significantly predicted changes in later employment. The effect of
being arrested between release and the 3-month interview (Wave 2) was associated with a
39 percent reduction in the likelihood of reporting employment between the 3- and 9-
150 N. W. LINK AND C. G. ROMAN
month (Wave 3) interviews (p< .01). This effect was not significant at the following wave
(15 months), although the direction of the association was the same.
Other significant findings were that having a high school education or GED, and the
control for variation in child support laws across states predicted employment at Wave 2
(both positive; p< .05). Being African American (p< .05) and having greater physical
Table 1. Summary statistics for men (n= 1,011) in the SVORI sample.
Variables NMor % SD Range
Dependent Variables
Rearrest (1 = yes, 0 = no)
T2Rearrest 1,011 16% 0–1
T3Rearrest 1,011 32% 0–1
T4Rearrest 1,011 32% 0–1
Employment (1 = yes, 0 = no)
T2Employment 602 65% 0–1
T3Employment 588 70% 0–1
T4Employment 560 67% 0–1
Time-varying Covariates
Child Support (CS) Debt Amount
Baseline CS debt 1,011 1.21 2.25 0–6
T2Debt 603 1.48 2.42 0–6
T3Debt 616 1.47 2.38 0–6
T4Debt 672 1.57 2.46 0–6
Instrumental Family Support
T2FamilySupport 591 11.60 2.86 0–15
T3FamilySupport 572 11.17 3.00 0–15
T4FamilySupport 550 11.20 2.96 0–15
Marital Status/Steady Partner (1 = yes, 0 = no)
BaselineMarried 1,008 48% 0–1
T2Married 602 63% 0–1
T3Married 616 69% 0–1
T4Married 672 61% 0–1
Job Services (1 = yes, 0 = no)
T2JobServices 603 40% 0–1
T3JobServices 616 34% 0–1
T4JobServices 672 21% 0–1
Physical Health Problems (0–4)
T2PhysicalHealth 601 0.52 1.10 0–4
T3PhysicalHealth 616 0.58 1.12 0–4
T4PhysicalHealth 672 0.58 1.11 0–4
On Supervision (1 = yes, 0 = no)
T2Supervised 602 83% 0–1
T3Supervised 670 52% 0–1
T4Supervised 613 69% 0–1
Reincarcerated (1 = yes, 0 = no)
T2Reincarcerated 1,011 4% 0–1
T3Reincarcerated 1,011 16% 0–1
T4Reincarcerated 1,011 23% 0–1
Time Invariant Covariates
Age at release 1,011 29.68 6.44 18–73
African American 1,011 59% 0–1
White/other 1,011 41% 0–1
HS education (1 = yes, 0 = no) 1,011 59% 0–1
SVORI participation (1 = yes, 0 = no) 1,011 50% 0–1
Employed at baseline (1 = yes, 0 = no) 1,009 63% 0–1
Index offense-property 1,011 19% 0–1
Index offense-violent 1,011 37% 0–1
Index offense-drug/other 1,011 44% 0–1
Age at first Arrest 1,003 16.01 4.84 6–48
State CS law 1,011 1.29 .82 0–2
Hourly Wage 984 7.33 8.24 0–87.5
THE SOCIOLOGICAL QUARTERLY 151
health problems (p< .001) decreased the likelihood of reporting legal employment.
Coefficients for physical health problems and race more or less showed similar effects
through Waves 3 and 4. At Wave 4, SVORI participation was significantly associated with
reporting legal employment (OR = 1.35, p< .05). This last result mirrors findings from the
2004 evaluation of the SVORI that found participation in SVORI increased receipt of
employment-related services and was linked to better employment outcomes (Lattimore
et al. 2012). Instrumental family support did not have a statistically significant relationship
with the likelihood of reporting legal employment.
Table 2. Cross-lagged impacts of child support and covariates on employment outcomes (odds ratios
displayed), n= 1,010.
Baseline Wave 2 (3 mos.) Wave 3 (9 mos.) Wave 4 (15 mos.)
Employed OR SE OR SE OR SE OR SE
Amount child support debt 1.011 .025 1.007 .026 .940* .027
Prior rearrest .606** .117 .880 .151
Prior employment 1.461** .171 2.081*** .295 2.861*** .440
HS education/GED at baseline 1.060 .147 1.335* .153 1.264 .181 1.214 .179
Received job services .963 .112 1.078 .157 1.124 .187
African American .596*** .086 .752* .090 .702* .106 .868 .132
SVORI participant 1.196 .137 1.039 .145 1.345* .193
Physical health problems 1.027 .056 .797*** .039 .809*** .045 .871* .055
Married/partner 1.394* .187 1.088 .127 1.055 .164 1.422* .218
Family instrumental support .981 .020 1.017 .024 .958†.024
Property offense 1.228 .252 .834 .139 .905 .187 1.199 .258
Drug/other offense .808 .117 1.088 .133 1.064 .159 1.041 .159
State CS law 1.152* .082 1.181 .101 .954 .086
Reincarcerated 1.110 .251 1.288 .235 1.100 .218
Model log-likelihood df AIC BIC
–3,963.17 116 8,158 8,728
Note.†p< .10; *p< .05; **p< .01; ***p<.001, two-tailed tests.
Table 3. Cross-lagged impacts of child support and covariates on rearrest outcomes (odds ratios
displayed), n= 1,010.
Wave 2 (3 mos.) Wave 3 (9 mos.) Wave 4 (15 mos.)
Rearrest OR SE OR SE OR SE
Amount child support debt .888†.060 .943 .055 .960 .053
Current employment .378** .123 .655 .226 .390** .135
Prior employment .880 .641 1.203 .367 1.385 .473
HS education/GED .396** .111 .552* .155 1.036 .310
Age at first arrest 1.007 .031 .997 .030 1.002 .029
African American 1.493 .459 1.356 .408 1.509 .460
SVORI participant 1.147 .312 .999 .274 1.297 .370
On supervision 1.035 .351 .539* .152 1.408 .416
Married/partner .737 .205 .598†.183 1.047 .326
Family instrumental support .973 .046 .940 .040 .993 .047
Property offense .751 .315 1.542 .614 1.055 .412
Drug/other offense .776 .230 1.050 .317 .605 .194
State CS law .689* .111 .575** .097 .742 .134
Hourly wage .994 .024 1.019 .029 1.005 .017
Prior rearrest 1.761 .644 4.196*** 1.361
Reincarcerated 23.316*** 10.069 8.989*** 2.750 12.592*** 4.414
Model log-likelihood df AIC BIC
–3,963.17 116 8,158 8,728
Note.†p< .10; *p< .05; **p< .01; ***p<.001, two-tailed tests.
152 N. W. LINK AND C. G. ROMAN
Turning to the arrest outcomes presented in Table 3,findings at the 3-month interview
showed that increasing child support was associated with a reduction in rearrest, though
the effect slightly missed conventional significance levels (p= .06, two-tailed). The
coefficients for the following two time periods, though in the same direction, were not
significant.
Cross-sectional results revealed that reporting having legitimate employment signifi-
cantly reduced the odds of rearrest by 62 percent and 61 percent at the 3- and 15-month
interviews, respectively. The path was not significant (p= .15) for employment at 9
months on arrest at 9 months. Longitudinal analyses showed that employment did not
have lagged effects on rearrest.
Despite recent interest in the potential benefits of instrumental family support, this variable
showed no significant impacts at any wave on rearrest. Although not part of our key research
questions, there were a few covariates that significantly predicted rearrest. At Wave 2, these
were education (negative; p< .01) and our control for variation in child support laws across
states (negative; p< .05). At Wave 3, significant covariates were being under criminal justice
supervision (negative; p< .05), education (negative; p< .05), and the control for variation in
child support laws across states (negative; p< .001). By Wave 4, results showed that—other
than employment status—only prior rearrest and the control for reincarceration were sig-
nificant (positive; p< .001).
Discussion
In the context of unprecedented legal financial obligations that many individuals face once
released from prison (Bannon et al. 2010; Beckett and Harris 2011), and the life course
and strain perspectives suggesting that they matter, the analyses presented here sought to
assess whether child support debt in particular affects employment and recidivism among
former prisoners. Given the finding that higher child support debt significantly reduced
employment by the 15-month interview, it could be that as the reality of life sets in after
release from prison, the strain of a high child support burden begins to drive former
prisoners away from legitimate, on-the-books employment, and puts them on a pathway
to the underground economy. Much theorizing has assumed that official state child
support enforcement policies have created a perverse incentive structure that serves to
decrease the likelihood that those who owe child support will enter the legitimate labor
force. This position is based on the logical reasoning that, if the government can garnish a
majority of an obligor’s income (up to 65 percent)—which presumably the obligor does
not want—he or she will give up searching for employment. Alternatively, the obligor may
pursue work in the off-the-books economy where payments are made “under the table.”If
the government cannot detect any earnings, it cannot garnish any, and more of the former
prisoner’s earnings can be funneled to his or her family. Interestingly, this effect of debt on
legitimate employment emerged after one year out in the community, which could mean
that—among former prisoners—there is something about being in the community longer
that matters vis-à-vis the theory that high child support causes joblessness. Though it
remains an unanswered question, these results hint at several possible dynamics occurring
among debtors released from prison. First, the time-dependent association between the
family obligation and employment could mean that it takes several months for former
prisoners to internalize the large scale of their debts and the apparent futility of their
THE SOCIOLOGICAL QUARTERLY 153
repayment efforts. Intentions to keep up with payments may fade if the debt burdens
appear insurmountable. Further, the debt burden may have grown rapidly since release,
particularly for those who had their debts modified (i.e., debt did not accrue) while
incarcerated. In another scenario, upon release from prison, it likely takes state govern-
ments several months to begin the legal process of garnishing wages to repay outstanding
child support arrears or reimburse welfare benefits. If this is the case, the disincentive for
the former prisoner to find work in the mainstream economy—where wages are visible
and therefore can be garnished—would not arise for the several months before wage
garnishing begins.
In terms of recidivism, the theoretical literature supports both the possibility that debt
could be protective and, conversely, that it could encourage criminal activity. We found
no consistent support across the reentry period for either hypothesis. Nor did we find that
those respondents with lower baseline hourly wages were more susceptible to rearrest. An
important policy implication stemming from this finding is that despite recent valid
concern over the increasing size of child support arrears, it does not appear that judicial
actors should be overly concerned by the idea that child support–related financial obliga-
tions in particular might inspire illegal conduct that results in rearrest. There may well be
other adverse consequences of unending debt burdens for former prisoners that should
concern policymakers, but heightened recidivism may not be one.
In support of Wilson’s(1997) unemployment thesis, employment and rearrest from
the same waves were strongly negatively associated. However, since these cross-sec-
tional paths are subject to questions of causal directionality (employment affects
rearrest but rearrest also affects employment), we examined the lagged effects of both
on each other while controlling for prior employment and rearrest. Testing the long-
itudinal version of the Wilson (1997) thesis that lack of employment increases criminal
behavior, we found no significant relationship between employment at an earlier time
period and changes in later arrests. However, we did find support for the reverse:
arrests can decrease the likelihood of being employed later in time (while controlling
for reincarceration). An arrest by the 3-month interview was significantly and strongly
associated with a drop in reported employment at the 9-month interview. We couched
this finding using a stigma and labeling framework (Uggen et al. 2006); former inmates
who recidivate are marked and, consequently, are not attractive targets for hire.
Another possibility is that we may have uncovered a process whereby arrests caused
employers to terminate those arrested employees. However, if the latter is the case, one
might expect that termination to occur immediately, but the effect we found was lagged
over a period of months.
Granted the observation period was only 16 months, turning-point variables such as
employment and marriage (both were time-varying measures in our models) did not find
consistent, strong support. In the data, although employment and rearrest were strongly
correlated cross-sectionally, employment did not predict a change in later criminal
activity, suggesting that—in general—having a job or getting a new job did not redirect
the life course in a favorable direction for the men in this sample. The effect of having a
steady partner or being married on arrest was nonsignificant. The lack of findings for
marriage/partner and employment could be reflective of the SVORI sample being one that
consists mainly of violent and serious offenders. Sampson and Laub’s work has shown that
violent offenders tend to desist later than property offenders (Sampson and Laub 1995). In
154 N. W. LINK AND C. G. ROMAN
their study, the average age of desistance for violent offenders was 31; for property
offenders it was 26. The average age at release for the SVORI sample is 29. The lack of
significance for marriage could also be a measurement issue, as others have recently noted
(Bersani and Dipietro 2016), because the measures in this study did not capture the
quality of the job or marriage. As Laub and Sampson (2003) have shown, desistance is
more likely when attachment to a job or marriage is strong. In addition, the SVORI data
span less than two years—a small time period in the life course.
With respect to instrumental family support, we found no evidence that it impacted
recidivism or employment, except for a marginally significant effect on employment in
one wave only. Unlike emotional family support, instrumental support in this context may
not help to bring about changes in behavior among former prisoners. However, our
findings could also suggest that instrumental support can be hard to capture in survey
items, as this type of support surely manifests in varying ways.
The findings presented here are qualified by a few other limitations. First, our recidi-
vism dependent variable (i.e., official rearrest data) is not a perfect measure of recidivism,
as some reoffending certainly was not captured in this variable. However, its strength,
relative to the self-report measures of recidivism in these data, is the completeness of the
data. Further, official arrest dependent variables tend to be preferred in reentry research
(Lattimore and Visher 2013).
Second, as in most panel studies and all reentry studies, missing data due to subject
attrition at follow-up periods is an issue. We attempted to address this in two comple-
mentary ways. First, unlike longitudinal repeated measures analysis that requires complete
data at every time point, generalized structural equation modeling can use cases that have
some missing data at some waves. This method retains more information than casewise
deletion methods that drop entire cases if they have missing data at one or more waves.
Second, the Heckman weighting adjusts for sample nonrepresentativeness after the base-
line interview. We compared results from Heckman vs. non-Heckman models and found
them to be very similar, boosting confidence in the patterns uncovered. Further, this
finding echoes other scholars’work showing that various methods for addressing missing
data with the SVORI data set produce similar results (Taylor and Auerhahn 2013). While
these methods represent new and innovative approaches to address missing information
in panel data, we acknowledge that these methods do not provide a complete solution to
the perennial problem of the potential bias introduced by missing data in panel studies.
Third, because our measure of child support debt was the self-reported amount of a
debt obligation at each wave, it is subject to bias due to underreporting or simply due to
respondents not knowing the actual amount that they owe. The categorical grouping of
dollar amounts for the response values used in the SVORI interview protocol for the child
support debt item may minimize these potential reliability issues. With regard to possible
underreporting of child support debt, we have confidence in the frequencies reported
because data from other sources show similar percentages of inmates who have child
support obligations and related debt (Griswold et al. 2004; Visher et al. 2004).
Fourth, we were not able to completely account for interstate differences in our
analyses. Importantly, the child support law variable created to address state context
was significantly associated with employment at one wave and rearrest at two, indicating
real differences in the child support, employment, and rearrest landscape across states that
remain to be uncovered.
THE SOCIOLOGICAL QUARTERLY 155
Finally, the SVORI data—which focus on serious offenders—are not likely representa-
tive of the entire U.S. adult male prison population, and therefore the findings may not be
generalizable across all male state prisoners. However, because there are virtually no large
multistate data sets available to shed light on potential problems related to criminal justice
financial obligations and formerly incarcerated populations, the SVORI data set remains
an important avenue to explore how child support obligations and related debt affect
aspects of prisoner reentry.
Conclusions and Policy Implications
Prisoner reintegration research has traditionally focused on issues surrounding salient
domains of need such as housing, substance abuse, mental health, and employment, and
how these might hinder successful reintegration. As important as these investigations are,
their pursuit has relegated inquiry into “invisible barriers”—such as an exploration into
collateral issues such as fines, fees, child support, and other legal obligations—to the
criminological backburner. The current work, along with the recent work of others, has
attempted to bring these issues out of the shadows. Additional research to uncover the
theoretical mechanisms that link debt to negative outcomes—and any possible mediators
—in reentry is needed. Specifically, our findings that link high child support debt with
lower odds of legitimate employment highlight the need for research that assesses the
drivers of both legitimate and illegitimate work in the reentry period and how they might
vary over time.
Policy implications surrounding child support orders are debatable in the context of
fathers returning from incarceration—should they be “cracked down on”for the sake of
their family’sfinancial well-being, or should they be given a degree of leeway to mitigate
the stressful reentry experience? On one hand, enforcement mechanisms should be (and
are) in place to increase the likelihood that dependent children and their caretakers receive
critical financial support. This seems especially important in the instances where payments
go directly to the family rather than to the state as welfare reimbursement. On the other
hand, many fathers returning from incarceration are indigent, have little prospect of
obtaining well-paying employment, and already have sizable debts in place stemming
from fines, fees, and costs associated with the justice process. In this scenario, adding large
child support debt burdens to these cases and expecting payment is unrealistic, and likely
achieves little in terms of assisting dependent family members with their material needs.
In fact, strict enforcement on indigent fathers could undermine the very purpose of the
child support system by hurting the family if fathers’inability to make steady payments
results in their reincarceration. Of course, the former prisoners trying to reintegrate do
not benefit in this situation either. As an alternative angle, recent innovations in programs
and policies designed to increase fathers’ability to make payments may hold promise in
the realm of prisoner reintegration. These include offering work-oriented programs and
tax credits for child support payment to noncustodial parents, and such approaches have
shown to increase fathers’employment and child support compliance (Sorensen and
Lippold 2013). Further, recent work suggests that improving the quality of parents’
coparental relationships increases financial support provided by nonresident fathers
(Goldberg 2015). This family-strengthening intervention could offer several benefits to
former prisoners and their families.
156 N. W. LINK AND C. G. ROMAN
Our findings imply that cross-systems approaches or comprehensive reentry programs that
address debt in tandem with other aspects of reentry may be needed. On a larger scale, a
continued focus on child support and related debt remains important for criminal justice and
child welfare policy simply given the stark findings by other researchers that show a quarter of all
child support arrears owed to custodial parents are owed by individuals who were incarcerated
or previously incarcerated (Ovwigho et al. 2005). Given the extent of these consequential,
unpaid obligations among prisoners and former prisoners in the United States, policymakers
might begin to think differently about how to prioritize supports for returning prisoners and
their families.
Notes
1. Eighty-one percent completed at least one follow-up. T-tests compared those who completed all
interviews to those who did not and found no statistically significant differences between the
two groups (p> .05) along many dimensions, including age, race, education, marital status, type
of index offense, family instrumental support, employment status, SVORI treatment assign-
ment, and the presence of a child support obligation. This pattern echoes SVORI researchers’
findings that at the follow-up waves, neither treatment nor comparison groups were statistically
different on key characteristics from baseline (Lattimore et al. 2012).
2. We recognize the pros and cons of this strategy. While respondents who were reincarcerated
were almost certainly rearrested, there could exist respondents who were rearrested but not
reincarcerated. As such, we ran all analyses two ways: filling in reincarcerated as a proxy and
with those respondents dropped. The model results were very similar; hence, we report on
models using reincarceration to signify rearrest in cases missing rearrest information.
3. Due to concerns that this variable might be problematic for the rearrest outcome, models were
also run with this variable omitted. Results were not substantively different.
4. We did not control for each state in the model. Adding dummy controls for each state created
an “overparameterized”model with 77 new paths—11 states × 7 outcomes (one state would be
the reference category) that would not converge (Tanaka 1987). As such, we were forced to
exclude state controls from the final model but have included this variable to capture a degree
of variation across states.
Acknowledgments
We are grateful to the editorial office at The Sociological Quarterly and four anonymous reviewers
for their careful assessments of our article. In addition, we would like to thank Ralph B. Taylor for
his guidance and feedback on the analytical models. An earlier version of this article was presented
at the 2014 annual meeting of the American Society of Criminology.
Funding
This work was supported by Award No. 2012-IJ-CX-0012, awarded by the National Institute of
Justice, Office of Justice Programs, U.S. Department of Justice. The opinions, findings, conclusions,
and recommendations expressed here are those of the authors and do not necessarily reflect the
views of the Department of Justice.
Notes on Contributors
Nathan W. Link is a doctoral candidate at Temple University. His areas of research are corrections,
criminal justice policy, and criminological theory, and his recent work focuses on debt associated
THE SOCIOLOGICAL QUARTERLY 157
with criminal justice processing. His work is published in Justice Quarterly, Crime and Delinquency,
Criminal Justice and Behavior, Criminal Justice Policy Review, and Administration and Policy in
Mental Health and Mental Health Services Research. He can be reached at Nathan.link@temple.edu.
Caterina G. Roman is an associate professor in the Department of Criminal Justice at Temple
University. Her interests include the social networks of at-risk and gang youth; prisoner reentry;
gang violence; and the role of community organizations and institutions and other aspects of social
capital in crime prevention and neighborhood well-being. She can be reached at croman@temple.
edu.
References
Agnew, Robert. 2006. Pressured into Crime: An Overview of General Strain Theory. New York:
Oxford University Press.
Bannon, Alicia, Mitali Nagrecha, and Rebekah Diller. 2010. Criminal Justice Debt: A Barrier to
Reentry. New York: Brennan Center for Justice.
Beckett, Katherine, and Alexes Harris. 2011. “On Cash and Conviction.”Criminology and Public
Policy 10(3):509–37.
Bersani, Bianca E., and Stephanie M. Dipietro. 2016. “Marriage and Offending: Examining the
Significance of Marriage among the Children of Immigrants.”Sociological Quarterly 57(2):304–32.
Blumstein, Alfred, and Kiminori Nakamura. 2009. “Redemption in the Presence of Widespread
Criminal Background Checks.”Criminology 47(2):327–59.
Bucklen, Kristofer, and Gary Zajac. 2009. “But Some of Them Don’t Come back (to Prison!):
Resource Deprivation and Thinking Errors as Determinants of Parole Success and Failure.”
Prison Journal 89(3):239–64.
Bushway, Shawn, Brian D. Johnson, and Lee Ann Slocum. 2007. “Is the Magic Still There? The Use
of the Heckman Two-Step Correction for Selection Bias in Criminology.”Journal of Quantitative
Criminology 23(2):151–78.
Cammett, Ann. 2010. “Deadbeats, Deadbrokes, and Prisoners.”Georgetown Journal on Poverty Law
and Policy 18(2):127–68.
Carson, E. Ann, and Daniela Golinelli. 2013. “Prisoners in 2012.”Bureau of Justice Statistics.
Retrieved July 15, 2015 (http://www.bjs.gov/content/pub/pdf/p12tar9112.pdf).
Durose, Matthew R., Alexia D. Cooper, and Howard N. Snyder. 2014. “Recidivism of Prisoners
Released in 30 States in 2005: Patterns from 2005 to 2010.”Bureau of Justice Statistics. Retrieved
July 20, 2015 (http://www.bjs.gov/content/pub/pdf/rprts05p0510.pdf).
Glaze, Lauren, and Laura Maruschak. 2008. “Parents in Prison and Their Minor Children.”U.S.
Department of Justice, Office of Justice Programs. Retrieved August 1, 2015 (http://www.bjs.gov/
content/pub/pdf/pptmc.pdf).
Goldberg, Julia S. 2015. “Coparenting and Nonresident Fathers’Monetary Contributions to Their
Children.”Journal of Marriage and Family 77(3):612–27.
Griswold, Esther, Jessica Pearson, and Lanae Davis. 2001. Testing a Modification Process for
Incarcerated Parents. Denver: Center for Policy Research.
Griswold, Esther, Jessica Pearson, Nancy Thoennes, and Lanae Davis. 2004. Fathers in the Criminal
Justice System. Denver: Center for Policy Research.
Hairston, Creasie. 2002. “Fathers in Prison.”Marriage and Family Review 32(3–4):111–35.
Harris, Alexes, Heather Evans, and Katherine Beckett. 2010. “Drawing Blood from Stones: Legal
Debt and Social Inequality in the Contemporary United States.”American Journal of Sociology
115(6):1753–99.
Holzer, Harry, Paul Offner, and Elaine Sorensen. 2005. “Declining Employment among Young
Black Less-Educated Men: The Role of Incarceration and Child Support.”Journal of Policy
Analysis and Management 24(2):329–50.
Lattimore, Pamela K., Kelle Barrick, Alexandra Cowell, Debbie Dawes, Danielle Steffey, Stephen Tueller,
and Christy A. Visher. 2012. “Prisoner Reentry Services: What Worked for SVORI Evaluation
Participants.”Retrieved June 5, 2015 (https://www.ncjrs.gov/pdffiles1/nij/grants/238214.pdf).
158 N. W. LINK AND C. G. ROMAN
Lattimore, Pamela K., and Danielle M. Steffey. 2009. The Multi-Site Evaluation of SVORI:
Methodology and Analytic Approach. Research Triangle Park, NC.
Lattimore, Pamela K., and Christy A. Visher. 2013. “The Impact of Prison Reentry Services on Short-
Term Outcomes: Evidence from a Multisite Evaluation.”Evaluation Review 37(3–4):274–313.
Laub, John H., and Robert J. Sampson. 2003. Shared Beginnings, Divergent Lives: Delinquent Boys to
Age 70. Cambridge, MA: Harvard University Press.
Levingston, Kirsten, and Vicki Turetsky. 2007. “Debtor’s Prison—Prisoners’Accumulation of Debt
as a Barrier to Reentry.”Clearinghouse Review 41:187–97.
Martire, Kristy, Sandra Sunjic, Libby Topp, and Devon Indig. 2011. “Financial Sanctions and the
Justice System: Fine Debts among New South Wales Prisoners with a History of Problematic
Substance Use.”Australian and New Zealand Journal of Criminology 44(2):258–71.
McLean, Rachel L., and Michael D. Thompson. 2007. Repaying Debts. New York: Council of State
Governments Justice Center.
Merton, Robert K. 1968. Social Theory and Social Structure. New York: Simon and Schuster.
Miller, Daniel P., and Ronald B. Mincy. 2012. “Falling Further Behind? Child Support Arrears and
Fathers’Labor Force Participation.”Social Service Review 86(4):604–35.
Mincy, Ronald B., and Elaine J. Sorensen. 1998. “Deadbeats and Turnips in Child Support Reform.”
Journal of Policy Analysis and Management 17(1):44–51.
Nagrecha, Mitali, Mary F. Katzenstein, and Estelle Davis. 2015. When All Else Fails, Fining the
Family. New York: Center for Community Alternatives.
Ovwigho, Pamela, Correne Saunders, and Catherine Born. 2005. The Intersection of Incarceration
and Child Support: A Snapshot of Maryland’s Caseload. Baltimore: Family Welfare Research and
Training Group.
Pager, Devah. 2007. Marked: Race, Crime, and Finding Work in an Era of Mass Incarceration.
Chicago: University of Chicago Press.
Pearson, Jessica. 2004. “Building Debt While Doing Time: Child Support and Incarceration.”Judges’
Journal 43(4):5–12.
Petersilia, Joan, Peter W. Greenwood, and Marvin Lavin. 1977. “Criminal Careers of Habitual
Felons.”Rand 1–161. Retrieved July 1, 2015 (https://www.rand.org/content/dam/rand/pubs/
reports/2007/R2144.pdf).
Pirog, Maureen, Marilyn Klotz, and Katharine Byers. 1998. “Interstate Comparisons of Child
Support Orders Using State Guidelines.”Family Relations 47(3):289–95.
Pirog, Maureen A., and Kathleen M. Ziol-Guest. 2006. “Child Support Enforcement: Programs and
Policies, Impacts and Questions.”Journal of Policy Analysis and Management 25(4):943–90.
Primus, Wendell. 2006. “Improving Public Policies to Increase the Income and Employment of
Low-Income Nonresident Fathers.”Pp. 211–48 in Black Males Left Behind, edited by Ronald B.
Mincy. Washington, DC: Urban Institute.
Raftery, Adrian E. 1995. “Bayesian Model Selection in Social Research.”Sociological Methodology
25:111–63.
Richards, Stephen C., and Richard S. Jones. 2004. “Beating the Perpetual Incarceration Machine.”
Pp. 201–32 in After Crime and Punishment: Pathways to Offender Reintegration, edited by Shadd
Maruna and Ross Immarigeon. London: Willan.
Roman, Caterina G., and Nathan W. Link. 2015. “Community Reintegration among Prisoners with
Child Support Obligations: An Examination of Debt, Needs, and Service Receipt.”Criminal
Justice Policy Review. doi: 10.1177/0887403415611460.
Sampson, Robert J., and John H. Laub. 1995. Crime in the Making: Pathways and Turning Points
Through Life. Cambridge, MA: Harvard University Press.
Seltzer, Judith A., Sara S. McLanahan, and Thomas L. Hanson. 1998. “Will Child Support
Enforcement Increase Father-ChildContactandParentalConflict after Separation?”Pp.
157–90 in Fathers under Fire: The Revolution in Child Support Enforcement,editedby
Irwin Garfinkel, Sara S McLanahan, Daniel Meyer, andJudithASeltzer.NewYork:Russell
Sage.
Skardhamar, Torbjørn, and Jukka Savolainen. 2014. “Changes in Criminal Offending around the
Time of Job Entry: A Study of Employment and Desistance.”Criminology 52(2):263–91.
THE SOCIOLOGICAL QUARTERLY 159
Sorensen, Elaine. 1997. “A National Profile of Nonresident Fathers and Their Ability to Pay Child
Support.”Journal of Marriage and the Family 59(4):785–97.
Sorensen, Elaine. 2004. “Understanding How Child-Support Arrears Reached $18 Billion in
California.”American Economic Review 94(2):312–16.
Sorensen, Elaine, and Kyle Lippold. 2013. Strengthening Families through Stronger Fathers Initiative:
Summary of Impact Findings. Washington, DC: Urban Institute.
Sorensen, Elaine, and Helen Oliver. 2002. Policy Reforms Are Needed to Increase Child Support from
Poor Fathers. Washington, DC: Urban Institute.
Statacorp. 2013. “Stata Statistical Software: Release 13.”Statacorp LP.
Sutton, Mike. 1995. “Supply by Theft: Does the Market for Second-Hand Goods Play a Role in
Keeping Crime Figures High?”British Journal of Criminology 35(3):400–416.
Tanaka, Jeffrey S. 1987. “How Big Is Big Enough?: Sample Size and Goodness of Fit in Structural
Equation Models with Latent Variables.”Child Development (58):134–46.
Taylor, Caitlin J., and Kathleen Auerhahn. 2013. “Handling Missing Data in the SVORI Data Set:
Does Methodological Approach Matter?”Paper presented at the annual meeting of the American
Society of Criminology Annual Meeting, Chicago, November 14.
Uggen, Christopher, JeffManza, and Melissa Thompson. 2006. “Citizenship, Democracy, and the
Civic Reintegration of Criminal Offenders.”Annals of the American Academy of Political and
Social Science 605(1):281–310.
Vallas, Rebecca, and Roopal Patel. 2012. “Sentenced to a Life of Criminal Debt: A Barrier to Reentry
and Climbing out of Poverty.”Clearinghouse Review 46(3–4):131–41.
Visher, Christy, Sara Debus-Sherrill, and Jennifer Yahner. 2011. “Employment after Prison: A
Longitudinal Study of Former Prisoners.”Justice Quarterly 28(5):698–718.
Visher, Christy A., Nancy L. La Vigne, and Jeremy C. Travis. 2004. Returning Home: Understanding
the Challeneges of Prisoner Reentry. Washington, DC: Urban Institute.
Western, Bruce, Anthony Braga, Jaclyn Davis, and Catherine Sirois. 2015. “Stress and Hardship
after Prison.”American Journal of Sociology 120(5):1512–47.
Wilson, William J. 1997. When Work Disappears: The World of the New Urban Poor. New York:
Vintage Books.
Wooldridge, Jeffrey. 2010. Econometric Analysis of Cross Section and Panel Data. Cambridge, MA:
MIT Press.
160 N. W. LINK AND C. G. ROMAN