Full text of Working Papers (Federal Reserve Bank of Chicago) : Female Offenders Use of Social Welfare Programs Before and After Jail and Prison : Does Prison Cause Welfare Dependency?, Working Paper 2006-13

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Federal Reserve Bank of Chicago

Female Offenders Use of Social
Welfare Programs Before and After
Jail and Prison: Does Prison Cause
Welfare Dependency?
Kristin F. Butcher and Robert J. LaLonde

WP 2006-13

November 2006

Female Offenders Use of Social Welfare Programs Before and After
Jail and Prison: Does Prison Cause Welfare Dependency?

Kristin F. Butcher
Federal Reserve Bank of Chicago and Wellesley College

Robert J. LaLonde
University of Chicago, NBER, and IZA

Abstract
Prior studies indicate that incarcerated women are among the most economically
disadvantaged populations in the U.S. An important difference between them and male
offenders is that these women are usually custodial parents. Therefore, the consequences
of incarceration for their well being are especially important because of its effect on
children. In this paper we focus on the links between incarceration and use of the social
welfare system. Is prison, for example associated with increased welfare dependency? To
better understand this relationship, we examine the temporal pattern of social welfare
receipt for 45,000 female offenders from Cook County, Illinois, the second most
populated county in the United States. We find that this group does in fact have high rates
of social welfare receipt, especially if they were incarcerated in state prison rather than in
county jail. But incarceration is associated with modestly lower rates of social welfare
receipt, especially for the less advantaged among the population of offenders. Further,
bans on TANF receipt for drug felons enacted as part ofwelfare reform have not
significantly affected this population's attachment to the social welfare system.
This research has been supported by grants from the Chicago Community Trust, The
Russell Sage Foundation, the Open Society Institute, and the Center for Human Potential
and Public Policy at the Irving B. Harris Graduate School ofPublic Policy Studies. We
thank Robert George and Allen Harden of the Chapin Hall Center for Children for
matching the incarceration and IDB data and many helpful comments. We have benefited
from comments by workshop participants at the Federal Reserve Bank of Chicago, the
University of Chicago, the Federal Reserve Bank of San Francisco, and UC-Berkeley.
We also are grateful to Steve Karr of the Illinois Department of Corrections, Dwayne
Peterson of the Cook County Department of Corrections and Vicky Nodal and Doris
Sutherland of the Illinois Department ofHuman Services for their assistance and helpful
discussions. The views expressed in this paper are those of the authors and do not reflect
those of any of these organizations or the Federal Reserve Bank of Chicago or the
Federal Reserve System.

I. Introduction

Incarcerated women are among the most economically disadvantaged populations
in the United States (Greenfeld and Snell, 2000). Many assume that these women's
interaction with the criminaljustice system serves to deepen that economic disadvantage
and increase welfare dependency. For example, a term of incarceration may hurt
women's employment prospects either through stigma or depreciation of skills.

Further,

although male prisoners are also usually poor, the consequences of both their economic
disadvantage and their incarceration may be different. Women prisoners are often
custodial parents at the time of their incarceration, while men are not (Mumola, 2000).
Department of Justice statistics indicates that on any given day there are nearly 150,000
women in U.S. jails or prisons and that these women are mothers to more than 250,000
children. Most of these children are less than ten years old (Greenfeld and Snell, 2000;
Mumola, 2000).
These women's role as caregivers, their prospects for economic self-sufficiency,
and their subsequent links to the social welfare system after leaving prison or jail are
important considerations when assessing the social consequences of incarcerating
women. If female incarcerations increase welfare dependency, this may constitute
another cost of prison to taxpayers and another of prison's effects on their children.
Accordingly, we examine how incarceration affects women's use of social welfare
programs, whether incarceration is associated with increased welfare dependency, and
how these assessments changed during the post-welfare reform environment when policy
makers limited some felons' access to welfare. 1

1

In August 1996, Congress passed and President Clinton signed the Personal Responsibility and Work
Opportunity Reconciliation Act (PRWORA). Among other things, this legislation ended the Aid to

2

To better understand the ties between incarceration and use of social welfare programs,
we examine the social welfare histories of more than 50,000 female offenders from Cook
County, Illinois. 2 We use a unique data set created from merged administrative records that
allows us to examine incarceration and social welfare use over a ten-year period. In particular,
we can examine social welfare use years prior to and years after a term of incarceration. We find
that incarceration does not increase welfare dependency, and indeed, over the long term, use of
social welfare programs after prison is about 6 percentage points, or 15 percent, lower.
Our results emphasize the importance of using data that allow one to examine women's
social welfare use over a long period of time. Women who will eventually go to prison have
higher rates of social welfare use than women who will only go to jail, so if one does not
adequately control for welfare use prior to the prison term, one may incorrectly interpret their
higher welfare use as an impact of prison. In addition, there is a pronounced temporal pattern to
welfare use around the time of a woman's first incarceration. Welfare use begins to drop several
months prior to her arrest, she is ineligible for social welfare during her period of imprisonment,
and then her welfare use begins to rebound after her release from prison, eventually rising to near
previous levels. This temporal pattern means that without a very long panel, it would be easy to
mistakenly conclude that prison increases welfare use. This is very similar to findings in the
literature on the impact of training programs on workers' subsequent wages: wages frequently
are aberrantly low in the months immediately preceding workers' participation in training

Families with Dependent Children program (AFDC) that had been first established as part of the Social
Security Act of 1935. It replaced this program with another cash assistance program entitled Temporary
Assistance to Needy Families (TANF).
2
According to the 2000 Census of Population and Housing, Cook County's population is 5.4 million. It is
the most populous county in Illinois and the second most populous county in the United States. Its largest
city, Chicago, has 2.9 million residents. The county's racial and ethnic composition is 26 percent black, 47
percent white, 21 percent Hispanic, and 5 percent Asian. Median household income was $45,239 or about 8
percent above the national median. The county's poverty rate stood at 12.3 percent about 1 percentage
point above the national rate. See U.S. Bureau of the Census web site www.census.gov.

3

programs, and the subsequent rebound may not be entirely due to the contents of the training
programs.
We investigate potential reasons for the drop in social welfare dependency for
formerly incarcerated women. Our analysis reveals considerable heterogeneity in the
effects of incarceration on social welfare usage. In general, we find that variables
associated with lower levels of life skills or functioning also are associated with larger
reductions in use of the social welfare system following incarceration. By contrast,
among the best-educated women in the sample, incarceration is associated with modestly
higher subsequent rates of social welfare receipt. Thus, the mechanism through which
prison leads to lower subsequent social welfare receipt is unlikely to be improvements in
these women's economic circumstances such that they are no longer eligible.
Finally, we also explore whether our findings are influenced by new rules
excluding some drug offenders from receipt of cash benefits through the Temporary
Assistance for Needy Families (TANF). In the Personal Responsibility and Work
Opportunities Reconciliation Act (PR WORA), Congress proscribed T ANF benefits for
drug felons for life, but gave states the option to opt out or modify this policy. Illinois
bans only serious drug felons from TANF benefits and they remain eligible for Food
Stamps. Although our analysis finds that TANF receipt rates of drug felons fell, we find
little evidence that this drop was the result of policy changes due to welfare reform.
Instead, this decline can be explained by the general decline in the Food Stamp and
T ANF caseloads experienced by all welfare recipients, including felons unaffected by the
rules proscribing TANF for serious drug felons. Once again, without a data set that
allowed one to construct a comparison group, and compare pre- and post- prison use of

4

these programs over a long period of time, it would be easy to conclude that welfare
reform had a bigger effect on drug offenders' use of social welfare programs than it likely
did.
We organize the remainder of this paper as follows. In section II, we summarize
the status of social welfare policy for felons. In section III, we discuss our database and
present some summary statistics and descriptive analyses. In section IV, we present a
statistical model and describe our comparison group strategies for identifying the causal
effect of jail or prison on social welfare receipt. Our empirical results and discussion
based on this model are found in section V. Some concluding remarks follow in section
VI.

II. Social Welfare Policy and Ex-offenders
When Congress enacted PRWORA in 1996, it banned individuals convicted of
drug felonies from receiving either TANF or Food Stamps for life. 3 Other felons and
felons arrested for drug-related offenses prior to August 22, 1996 remained eligible for
these benefits. However, Congress also allowed states to opt-out of this provision ofthe
law. Eight years after welfare reform, 17 states continue to deny both TANF and Food
Stamp benefits to felons convicted of drug-related offenses that occurred after August 21,
1996. In Illinois, as in 32 other states, the legislature modified or opted out of these bans
on T ANF and Food Stamp assistance.

Felons' Eligibility for Welfare in Illinois

3

PRWORA refers to Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (Public
Law 104-193). Temporary Assistance for Needy Families (TANF) replaced the former AFDC program,
which was part of the Social Security Act.

5

In 1997, Illinois limited lifetime welfare bans to individuals convicted of serious
drug-related crimes. 4 Felons convicted of a Class X or Class 1 drug law violations that
occurred after August 21, 1996 are ineligible for T ANF benefits. The ban has been
modified, but not lifted, for those convicted ofless serious drug-law violations. Drug
offenders who have been convicted of Class 2 through Class 4 felonies may not receive
T ANF benefits for 2 years after the date of their conviction. However, this 2 year ban
may be lifted when an offender is enrolled in or completes a drug treatment program or
participates in an aftercare program, such as Alcoholics Anonymous or while waiting for
a slot in a drug treatment program to become available (State of Illinois, 2003).
Under Illinois law all felons remain eligible for Food Stamps and all non-drugrelated felons remain eligible forT ANF. Further, despite any bans or limits on her own
participation, a woman's children remain eligible to receive TANF depending on their
mother's or any other household members' income and assets (State of Illinois, 2003).

Eligibility Criteria and Welfare Benefit Levels in Illinois
Many female parolees from lllinois prisons likely meet the income and asset
thresholds necessary to qualify for Food Stamps or TANF (Cho and LaLonde, 2005). For
Food Stamps, these thresholds require that eligible household units have less than $2,000
in "liquid" assets and to have less than $973 in monthly income. The monthly income
threshold rises when there are more adults in the household. When a household unit

4

Estimates from one study suggest that the Illinois ban on TANF benefits for serious drug felons affected
more than 10,000 women from Cook County during the 3 years following PRWORA's passage (Allard,
2002).

6

satisfies these and other eligibility criteria, monthly Food Stamp benefits range up to a
maximum of$371 for a unit ofthree. 5
The income and asset thresholds for T ANF eligibility in Illinois differ somewhat
from those used for Food Stamps. Family assets for a family of 3 must be less than
$3,000. But this amount excludes some durable goods such as one motor vehicle,
personnel furnishings and clothes. In addition, monthly family income, excluding that of
children, must be less than the Monthly Assistance Level plus $90. Once eligibi1ity is
established, additional earnings are disregarded at a rate of $2 for each $3 earned. T ANF
participants in Cook County receive $396 monthly if they are in a family of3, with one
adult head of household. A comparable child-only case would receive $107 monthly
(State of Illinois 2003).

III. The Sample of Incarcerated Women and their Social Welfare Histories
To construct the sample used in this study, the Chapin Hall Center for Children at
the University of Chicago matched administrative records from the Illinois Department of
Corrections (IDOC), and Cook County Department of Corrections (CCDOC), to the
Illinois Integrated Database (IDB) using probabilistic matching (Goerge, Van Voorhis,
and Lee, 1994). 6 The IDOC and CCDOC records from 1990 to June 30,2001 used in
this analysis consist of more than 52,000 women from Cook County. 7 The data in the

5

Under welfare refonn, when an applicant is childless, as might happen if an ex-offender has had her
parental rights terminated, she is eligible to receive food stamp benefits without working for up to three
months. The USDA allows states to exempt a significant percentage of such recipients and leaves it to state
policy makers to determine who in this group is to be exempted from this work requirement. During 2005,
monthly benefits for a household unit of one could range up to a maximum of $152.
6
The IDB on Child and Family Services is a multi-service integrated database constructed out of
administrative data gathered by public agencies serving children and families. Researchers at Chapin Hall
Center for Children at the University of Chicago have been working on the database since the mid-1980s
(Goerge, Van Voorhis and Lee, 1994).
7
The women in the sample may not be residents of Cook County, but instead they committed the offenses
that lead their incarcerations in Cook County. Our analysis of information available to us on women's street

7

IDB file comes mainly from two sources: the Illinois Department of Human Services
(DHS) and Illinois Department of Children and Family (DCFS) services. The DCFS
records provide information on whether a woman or her children were ever in the foster
care system or were ever involved in a substantiated charge of neglect or abuse. 8They
provide information on foster care spells from 1975 through 2001. The DHS records
provide information on the incidence and duration ofFood Stamps, AFDC/TANF, and
Medicaid spells covering the period from 1990 through 2001.
The match rate between the IDOC records from Cook County for women
committed to state prison and the IDB files was approximately 78 percent. The vast
majority of these women had at least one spell on Food Stamps. This high match rate is
consistent with the literature stating that female offenders consist oflargely economically
disadvantaged women. 9 The match rate among women who spent time in Cook County
jail, but not in state prison during the sample period was lower at about 50 percent. This
lower percentage suggests that women who have jail spells, but not prison spells, are less
economically disadvantaged than women who have prison spells. Overall, however, these
women, who constitute the population of incarcerated women in Cook County, have very
high rates of contact with the social or child welfare systems relative to the general
population, and even relative to the population of single women with minor children.

addresses at the time they entered Cook County jail, during 2001 and 2002, indicate that it is reasonable to
characterize our sample as consisting of Cook County residents.
8
Information on foster care spells is available from 1975 through 2001.
9

Below we examine the correlation between being on welfare and having been incarcerated. This
relationship may be biased toward zero to the extent that our measures of being on and off welfare are
misclassified. We have no evidence on its extent other than misclassification rates should vary across
people and not across time. Also we expect misclassification rates to be low among persons who we match
and higher among the non-matches.

8

In this study, we focus on participation in social welfare programs prior to and
subsequent to a woman's first prison spell between 1993 and 2001. We also examine this
same relationship for jail spells. 10 Although our incarceration data begin before 1993, we
limit the analysis to those who we first see in prison after January 1993, in order to ensure
that the spell we record as the first term of incarceration is indeed the first. Of particular
interest are short jail spells {lasting less than one month) which are the modal
incarceration spell in our data and for which incarceration would not necessarily be
associated with cessation in benefits. In some of our analyses, we treat woman who had
only one of these short spells during the time period studied as a comparison group.

11

Our social welfare records extend back at least three years prior to the first
incarceration spell in our sample. Our sample used in the analysis below consists of more
than 36,000 observations. This sample is smaller than the population of 52,000 women
that were incarcerated during this period, (i) because we exclude women who were
incarcerated in jail or prison during or after 1993, but who were admitted to prison for the
first time prior to that year, and (ii) because we exclude women with missing values for

w Individuals sentenced for felonies lasting more than one year usually are incarcerated in state prison; all
other felons are incarcerated in county jail. Time served is almost always less than the sentence. Median
time served in prison among women in our sample is about 9 months. Our sample of women incarcerated in
county jail includes women who were never convicted for a crime, but were detained in Cook County
facilities following their arrest.
11
Another reason for our separate interest in the association between incarceration in Cook County jail and
social welfare dependency is that when these women are in jail they usually are relatively close to their
communities and families. The jail campus is located within the city of Chicago and is accessible by public
transportation. By contrast, a woman in state prison will spend more time incarcerated and reside in a place
far removed from her community, family and other familiar social networks. Links to her family and
children while incarcerated may be a factor influencing social welfare usage by a woman when she is
released back into her community. The Cook County jail is located in the city of Chicago and is relatively
easily accessible by public transportation. By contrast, the three main prison facilities for women, Dwight,
Decatur, and Lincoln are located 74, 168, and 177 miles, respectively, from the Cook County jail and are
less accessible. In June 2005, the Greyhound web site quoted the daily round trip fare from Chicago to
Decatur at $69 per adult and $41 per child.

9

the demographic variables used in our analysis.

12

Finally, in order to make the statistical

analyses tractable, we take a 75 percent random sub sample of women who have jail
spells, but no prison spells. We retain the complete sample of women with prison
sentences. We create weights to adjust for this sampling.
We present the characteristics of female offenders from Cook County in Table 1.
The numbers in the table underscore the point made earlier that incarcerated women
constitute a disadvantaged population. As shown by the figures on the left side of the
table, these women are disproportionately high school dropouts. Nearly 60 percent of
them had received Food Stamps at least once between 1990 and 2001; about 47 percent
had received TANF/AFDC during the period. About 75 percent are African-American
even though African-Americans constitute only 25 percent of the county's population.
Further these women are mothers. The average number of children reported by these
women when they were admitted to jail or prison was 2.3. (Not shown in the table is that
more than 80 percent of these incarcerated women report that they had at least one child.)
Administrative records indicate that about 13 percent of these women had children in the
foster care system between 1975 and 2001. This percentage was higher among women
incarcerated in prison compared to women only incarcerated in jail. 13
Table 1 also indicates that most incarceration spe11s are very short. Many jail
spells last only a few days. Prison spells lasted on average about nine months, with a
median of 8 months. Statistics, not shown in the table, indicate that these women usually
committed drug-related offenses. About one-half of the women admitted to Illinois state
12

Our file of prison spells contains all admissions and exits starting July 1, 1989. We exclude from our
sample women who had a prison spell prior to January 1, 1993. Therefore, when we analyze the connection
between prison and social welfare spells, the prison spells are likely to be the first spell ever for these
women. We have no records on jail spells prior to October 1992.

10

prison during the 1990s had committed a drug law violation. This fact underscores the
potential for PRWORA and Illinois welfare policy toward drug offenders to have
important effects on social welfare receipt by female offenders. The potential importance
of welfare reform for drug offenders is magnified because women convicted of these
violations are more disadvantaged than other women. They are more likely to be high
school dropouts, to have had children in the foster care system, and to report having three
or more children. Among women admitted to county jail, the percentage arrested for drug
law violations is smaller but still substantial (about 30 percent).
The statistics on the right side of Table 1 present the sample means for the
approximately 5.1 million person-month observations covering the time period from
1990 to 2001. As shown by the first row, during any given month about 35 percent of
these women were either on Food Stamps, Medicaid, or TANF/AFDC. The percentage
on TANF/AFDC during any given month averaged approximately 20 percent. Being on
welfare during any given month was more common than being incarcerated even for this
sample ofwomen. Women in our sample were incarcerated in jail about 2.5 percent of
the time and were incarcerated in prison about 1.4 percent of the time. 14
Figures 1 through 4 show the temporal pattern of monthly rates of social welfare
receipt (Figure 1); AFDC/TANF receipt (Figure 2); Food Stamp receipt (Figure 3); and
Medicaid receipt (Figure 4) for our sample of female offenders. 15 We define monthly
social welfare receipt as a dummy variable indicating whether a woman received any
13

At any point in time approximately 0.7 percent of U.S. children are in foster care (Child Trends, 2005).
These percentages understate the likelihood of incarceration between 1990 and 2001, because our jail
incarceration data begins in January 1993. This suggests that scaling up the percentage of time in jail and
prison by about 25 percent.
14

11

combination of Food Stamps, Medicaid, or TANF/AFDC during the current month. The
patterns of receipt appear to mimic trends in social welfare receipt for the general
population. Among these women, monthly welfare receipt rates rise from about 35
percent in 1990 to about 41 percent by the mid-1990s, before falling steadily to a low of
about 28 percent by 2000. During the last year covered by the sample, social welfare
receipt rates rose slightly. 16 In our empirical work below, we control for this temporal
pattern in social welfare receipt rates by including dummy variables for year (and
calendar month) of the observation.
The next three figures (Figures 2 through 4) reveal that (1) the temporal pattern of
social welfare usage is driven by the Food Stamp and Medicaid receipt rates and (2) that
welfare reform as well as other forces likely had a marked impact on female offenders'
TANF/AFDC receipt rates. In 1990, about 27 percent of our sample of female offenders
received these cash benefits in any given month, but by 2000 this rate had fallen to only
five percent. Food Stamp and Medicaid receipt rates also fell during the period, from
about 36 percent in 1991 to about 23 percent in 2000. In contrast to the series for
TANFIAFDC, Food Stamp and Medicaid receipt rates started to rise again around 2000.

IV. A Statistical Model of Incarceration and Social Welfare Receipt

To investigate the relationship between incarceration and social welfare usage, we
specify a very flexible statistical model that can account for the variety of possible ways
that prison or jail could influence social welfare receipt. A priori, it is unclear how
incarceration will affect subsequent use of social welfare programs. If incarceration

15

These figures were created using the full sample of about 52,000 women who were incarcerated at least
once in jail or prison during the period, rather then the smaller sub-sample used for the regression analyses.
16
The data also suggest that during 1992 there was a temporary decline in receipt rates. As seen by
comparing Figures 2 through 4, this decline results from an anomaly in the Food Stamp and Medicaid data.

12

adversely affects offenders' employment prospects, it could increase the likelihood that
these women subsequently receive social welfare benefits. Alternatively, if incarceration
increases the chances that an individual will be incarcerated in the future, incarceration
could be associated with reduced use of social welfare programs.
Other possibilities also imply that incarceration is associated with lower
participation rates in social welfare programs. First, incarceration may be an indicator of
declining life skills. These declining life skills may be associated with either increasing or
declining welfare take-up rates prior to incarceration and this trend could continue even
after parole. The possibility that these women's life skills are declining prior to prison is
compelling in this study, because of evidence of high rates of substance abuse
Second, as already discussed above, Illinois law, after the implementation of
welfare reform, denies T ANF benefits to serious drug felons and limits it to less serious
drug felons. This policy suggests that the temporal pattern of receipt may differ for
different types of felons, and before and after the implementation ofwelfare reform.
Further, this policy also could adversely affect non-drug related felons' chances of
receiving welfare benefits to the extent that there is confusion about the rules among
potential applicants, community organizations, and caseworkers. Finally, incarceration
could also be associated with the loss of parental rights among custodial parents. Such a
loss could be associated with a decline in social welfare benefits, especially
TANF/AFDC benefits.
To sort out the foregoing issues, we focus on the temporal pattern of social
welfare usage around the time of the first jail and prison spells observed during our
sample period. We model the temporal pattern of social welfare receipt by examining the

13

relationship between the numbers of the months that the current month is from the month
of a woman's first incarceration and a dummy variable indicating whether she received
social welfare benefits during the current month. We present estimates from a statistical
model that has the following form:
(1)

In (1 ), we define Yit as a dummy variable equal to one if a woman received Food Stamps,
Medicaid, or TANF/AFDC during the (calendar) month t. The variable Xit denotes a
vector of observed characteristics described above in the text.
The term Yt in ( 1) denotes time-effects that account for the effect of changing
statewide economic conditions and policies on rate of social welfare receipt by female
offenders. 18 The term Eit in (1) denotes unobserved characteristics. We assume that the
time-varying component of the error is independently distributed across individuals. Our
standard error estimates are "robust standard errors" that take into account that these
unobservable characteristics are not identically distributed across individuals and time
periods.
The term o(~; Si, Zi) in (1) denotes the effect of prison on social welfare receipt.
We allow this effect to vary according to the number of months,

t',

between the current

month and the entry and exit months from prison. We assume that the prison effect is
time-invariant. Thus, we assume that there are no cohort effects of prison among these
women. We also allow these effects to vary by time served in prison and by a limited
vector of individual characteristics.
17

This model is borrowed from the program evaluation literature and is described in Jacobson, LaLonde,
and Sullivan ( 1993) and in Heckman, LaLonde, and Smith, ( 1999).

14

The term Rit(1:; Si, Zi) in (1) is a vector of dummy variables that denotes the
current month in terms of months relative to the month of incarceration. In our empirical
work, we explicitly allow for the possibility prison affects social welfare receipt rates
prior to entering prison. This specification attempts to account for the possibility that
women who are going to go to prison soon have had a particularly bad time and that
either their level of functioning or their fortunes are likely to improve regardless of
whether they go to prison. Our empirical evidence reported below appears to be
consistent with this view.
To implement this specification, we include in our model a vector of 17 dummy
variables that indicate whether the current calendar month is a given number of months
prior to a prison spell. We allow separate monthly effects of prison on social welfare
receipt for the first 13 months prior to prison. 19 We then control for these effects during
pre-prison months 14 through 29 using a step function consisting of 4 dummy variables.
By including these variables in the model, we measure the effect of prison on social
welfare receipt relative to social welfare receipt rates prior to the 29th month before
incarceration. We assume that incarceration does not affect social welfare receipt rates
more than 29 months prior to incarceration. We test this assumption with our data. We
also include three dummy variables to control for incarceration in the current month. We
control for whether the woman is in prison for her first prison term, whether she is in jail
for a prior or subsequent offense, and whether she is in prison on for a subsequent term.

18

We model these time effects with 11 dummy variables denoting the month of the year, and 12 dummy
variables denoting the year of the observation.
19
Note that while we refer to these as months prior to prison, the first month prior to prison is the month in
which the woman enters prison. The fraction of the month actually spent in prison will depend on whether
the woman enters early or late in the month.

15

We model the impact of prison on social welfare receipt after incarceration using
15 parameters. The first 13 parameters are dummy variables indicating whether the

°

current month is a given number of months after exit from prison. 2 For example, one of
the dummy variables is equal to 1 if the current month is six months after the month in
which the woman was released from prison.
We intend the last two post-incarceration parameters to provide a parsimonious
summary ofthe temporal pattern of social welfare receipt during the months beyond the
first year out of prison. These two parameters are associated with (I) a dummy variable
indicating whether the current month is 14 or more months after the month a woman left
prison and (ii) the inverse of the number of months (starting with the 13th month) after a
woman left prison as follows:
(2) Yo*l('t>13) + Y1*(ll('t -13), if't >13.
This specification produces a straightforward estimate of the long-term effect of
incarceration on social welfare receipt. As the number of months since exit from prison
increases, ll't-13 approaches zero. Therefore, the long-term effect of incarceration on
social welfare receipt is given by y0 • The effect during the 14th month after exiting prison
.

IS

.
b
21
given y Yo+ Yl·

The terms S; and Z; in Rit('t; Si, Zi) denote time served and observed
demographic characteristics, respectively. In the empirical work below, we focus on
interactions between a dummy variable indicating whether the current month is a post20

The first month after prison is actually the month of prison exit. The fraction of the month spent in
rrison will depend on whether the woman is released early or late in the month.
1
See Jacobson, LaLonde, and Sullivan (2005) for an example of this specification in the program
evaluation literature. We also experimented with using a linear trend in (2) instead of the inverse ofk-13.
This linear specification did not affect inferences about the short and medium term impact of incarceration,
but did affect inferences about its long-term effects.

16

incarceration month and, whether the woman's time served on her sentence was greater
than the median (8 months), the woman's highest grade completed, whether we ever
observe a drug law drug-law violation in the woman's records, and whether we ever
observe that the woman had one or more children in foster care. 22
Consequences of State Dependence for Incarceration Effects
Besides our desire to specify a flexible statistical model to account for a variety of
potential factors influencing the pattern of social welfare receipt, another rationale for our
model is to account for state dependence in welfare transitions. State dependence implies
that the probability a woman receives welfare during one month depends on whether she
received welfare during the previous month. We find strong evidence in our data of state
dependence in the transition rates on and off welfare. The percentage of women in our
sample who at some point go to prison, who are off welfare during month t-1 who
received welfare during month t is about 2.6 percent. By contrast, the percentage of
women on welfare during month t-1 who also received it during the following month is
95.3 percent. 23
One implication of this state dependence is that the impact of prison on welfare
receipt should vary with time since incarceration. When women are released from prison
they are off welfare. Therefore their welfare receipt rates are below expected levels, but
this gap should diminish with time as they transition back on to these programs. This
change occurs even if prison has no effect on the underlying transition rates onto or off of

22

This amounts to including additional variables into the model that are interactions between a dummy
variable indicating that the current month is a post-incarceration month and variables in Zi.

23

These figures include transitions during the months leading up to and immediately after prison. The
corresponding figures for women in the short jail sample are 1.5 and 97.1 percent, respectively. The figures

17

welfare. But this change-the rising welfare receipt rates following release from prison-is nonetheless an effect of having been incarcerated. This part of the effect of
incarceration on social welfare receipt is mechanical and it should dissipate with time.
But our estimates of the longer term effects of incarceration should be unaffected
by these post-prison adjustments. These longer term effects, if they exist, should result
from changes in welfare transition rates. The idea that in the presence of state
dependence, the short and long-term effects of incarceration might differ underscores the
importance of specifying a statistical model that allows the effect of prison to vary with
time since incarceration. 24
Constructing Comparison Groups

To estimate (1) we compare social welfare receipt among women incarcerated in state
prison to observationally similar women who were incarcerated only in county jail.
Ideally, our comparison group would include women who are identical to the women
who are incarcerated in state prison, but whose welfare receipt is not interrupted by a
period of incarceration. In our empirical work, we examined the sensitivity of our results
to the following samples ofwomen with jail spells:
(i)

women whose only incarcerations are jail spells;

(ii)

women whose only incarceration is a single jail spell;

(iii)

women who have only one jail spell that lasted less than 30 days.

Women in these three comparison groups were incarcerated at approximately the same
time as women who served time in state prison and were similarly involved with the

for women who eventually go to prison during the period more than 28 months prior to a woman's first
spell are 2.4 and 94.0 percent, respectively.
4
This point is discussed in more detail in Appendix A.

~rison

18

criminal justice system. However, these offenders are likely to have been arrested less
often and to have committed less serious offenses than women incarcerated in prison.
Associations that we might find between incarceration and social welfare receipt
could result from differences in unobserved attributes that determine both the seriousness
of a woman's criminal behavior and her social welfare receipt rate. If this concern is
important, we expect that our results should vary depending on which comparison group
we choose. The first of these comparison groups includes more serious offenders than
does the last. By contrast, women in the last group-who we call the short jail sample-include some who were arrested and held in county jail, but later released after the
charges were dropped. Moreover, because jail spells were so short we do not expect the
incarcerations of women in the third group to cause breaks in their welfare spells.

IV. Empirical Findings
Incarceration in County Jail, State Prison and Social Welfare Participation

In Figure 5, we present the mean rate of social welfare receipt relative to the
month of incarceration for three groups of incarcerated women: Women who (i) only had

jail spells (this group includes the next); (ii) only had one short jail spell and (iii) ever had
a state prison spell. We do not control for any demographic or time varying variables.
Each point on the graph, except the last, indicates the mean rate of social welfare receipt
in that month; the last point, after the shaded vertical bar, is the mean rate of social
welfare receipt for all subsequent months.
As shown by the figure, women incarcerated in state prison have social welfare
receipt rates that averaged nearly 40 percent 13 months prior to their first full-month in
prison. These rates declined during the pre-prison period and drop sharply during the two

19

months prior to their first full-month in prison. After exiting prison, women's social
welfare receipt rates rise for six months and return to levels approaching those that we
observed one year prior to the year that they entered prison.
The pre-jail social welfare receipt rates for two groups of jail only women were
about three percentage points lower than those of women incarcerated in prison. This
difference likely results because these women were likely less economically
disadvantaged than women who were incarcerated in prison. As expected, the pattern of
welfare receipt among women that had only a single short jail spell appears unaffected by
their incarcerations, because they were incarcerated for less than one month.
The information in Figure 5 suggests that prison is not associated with increased
welfare dependency. Further as indicated in the previous section where we discussed
state dependence in social welfare transitions, we observe an adjustment period after
prison. During the first 5 months after their paroles, these women's welfare receipt rates
rise rapidly. After this point, their welfare receipt rates have returned to close their preprison levels.
The figure also indicates that in the longer term, rates of welfare receipt appear to
be somewhat below pre-prison levels. The last point in the figure, the mean rate for all
subsequent months, starting 14 months after the month these women exited prison, is
about 30 percent. 25 This percentage is lower than the rates of social welfare receipt at the
end of the first year following incarceration and results in part because of the lower later
rates of social welfare receipt later in the sample period. (See Figures 1 through 4),
emphasizing the need to examine these patterns controlling for year effects.

20

Regression-Adjusted Estimates of the Effect of Prison
In Figure 6, we present the estimates based on equation (1) of the differences
between the rates of social welfare receipt of women incarcerated in prison and of women
with short jail spells, or 8('t), starting with the 291h month prior to the month they enter
prison or jail. We condition on the observed characteristics including: age, age 2 , race and
ethnicity, highest grade completed, any drug law violations, any foster care record, time
served on first prison greater than the median, controls for current incarceration26 , and
month and year effects.

27

The year effects help account for the fall in social welfare

receipt rates during the latter half of the sample period. The two lines in the figure are
associated with (i) the model estimated without individual fixed effects (the solid line)
and (ii) the model estimated with individual fixed effects (the line with squares). 28
We identify the "effects" of incarceration, 8('t), because we assume that the
regression-adjusted rates of social welfare receipt are the same for both the short jail and
prison groups prior to the 29th month. The figure indicates that this assumption is
approximately consistent with the data. During the period that falls 25 to 29 months prior
to the month of incarceration for the women who goes to prison there were only slight
differences between their regression-adjusted rates of social welfare receipt.

25

In Figure 5 as well as the remaining figures in the paper, the vertical bar serves to indicate that month 14
refers to all subsequent months pooled starting with the 14th month following the month these women were
paroled from prison.
26
Current incarceration controls include three separate indicator variables if the individual is currently in
jail, if the individual is currently in prison for her first prison term, and if the individual is currently in
~rison on a second or higher prison term.
7
Age is the only time-varying individual characteristic here. We experimented with including the cube of
age to account for nonlinearities in the relationship between age and social welfare receipt, but it was not
statistically significant. We intend our controls for age to take into account any life cycle effects on social
welfare receipt.
28
Estimates of the other coefficients in the model are available from the authors upon request.

21

Our empirical results indicate that starting with the 24th month prior to
incarceration, the rates of social welfare receipt of prison group slowly diverges. By the
8TH

month prior to incarceration, women who will go to prison have social welfare

receipt rates that are about 5 percentage points less than observationally similar women
who have a single short jail spell. At this point the rate of divergence increases. By the
second month prior to incarceration the two groups' social welfare receipt rates differ by
about 10 percentage points. We have explored whether these patterns in pre-prison
welfare receipt can be explained by jail spells that are not associated with the arrests that
lead to women's imprisonments. These estimates include controls for jail spells for
separate arrests from those leading to the current prison term, and for subsequent prison
and jail terms. However, leaving these controls out of our model does not significantly
change the pattern shown in Figure 6. 29
After paroling from prison, our estimates reveal the same post-prison patterns that
we inferred from the unadjusted mean rates of social welfare receipt in Figure 5. The first
five months after prison is a period of adjustment during which women's social welfare
receipt rates rapidly approach their expected levels based on their pre-prison histories and
the social welfare receipt rates ofwomen with single short jail spells. After this point,
post-prison welfare receipt rates remain about 4 to 5 percentage points below expected
levels. Over the longer term, the rates of social welfare receipt tend to fall somewhat. As
indicated by the last value in Figure 6, by the 25th month following these women's

29

Women who go to prison are substantially more likely than is typical for them at other times to be
incarcerated in jail in the months just prior to prison for unrelated offenses. But, we find including controls
for these jail terms, as we do in the models that underlie Figure 6, the decline in social welfare use prior to
prison remains. Recall that most jail spells in our full sample are short, many lasting less than 30 days. So
by themselves we would not expect such spells to affect rates of social welfare receipt.

22

paroles from prison, their rate of social welfare receipt has fallen to about 7 percentage
points (i.e., estimates of 30 in (2)) below expected levels. 30
Comparisons between the OLS and fixed-effect (FE) estimates in Figure 6
indicate that the FE estimates are about 2 percentage points lower (in absolute value). But
the FE estimates also were about 2 percentage points lower two years prior to prison.
Therefore, we conclude that both sets of estimates generate the same conclusion that
prison is associated with much lower rates of social welfare receipt during the first 4 or 5
months after prison, about a 5 percentage point lower rate by the end of the first year after
paroling from prison, and somewhat larger impacts in the long-term. Importantly, both
sets of results do not support the contention that incarceration in prison or involvement
with the criminal justice system is associated with greater welfare dependence. Without a
long panel of data, however, this would be difficult to detect given that the pre-prison
drop in social welfare receipt begins as much as two years prior to a woman's prison
term.
It is unclear how to interpret the long-term decline in social welfare receipt. It

could signify increased self-sufficiency, reduced life functioning or the possibility that
women who enter prison sometimes lose their parental rights, which could diminish the
likelihood that they receive social welfare benefits. Note that while recidivism might
lower welfare rates after the first prison term, we control for subsequent prison terms
here, and thus recidivism should not be driving the post-prison pattern depicted in Figure
6.
30

Because we include time-varying age effects and time effects, this pattern does not obviously result from
life cycle, macro-economic, and policy effects. In particular, this pattern does not result because the postprison period tends to occur later during the sample period when social welfare receipt rates are relatively

23

The Characteristics ofPrisoners and the Effects of Prison on Social Welfare Receipt

We explore whether our findings for the effect of prison on social welfare receipt
vary according to a woman's personal characteristics. Four variables of special interest
were as follows:
(i)

whether a woman's time served in prison was above or below the median
time served (which was 8 months);

(ii)

whether she ever had an offense that involved a drug law violation;

(iii)

her highest grade completed;

(iv)

and whether she has ever had a child in the foster care system.

We view these variables as a proxy for women's life skills and functioning. They likely
are associated with a woman's competence as a parent, in the labor market, and in being
able to complete the application process and collect welfare benefits.
In table 2 we present the results for whether post-prison welfare use varies with
individual characteristics. The first and second columns present the results for OLS and
fixed effect estimation. Columns 3 and 4 allow the pre-prison period to vary by these
characteristics as well. 31 This is to take into account the decrease in welfare receipt prior
to the first prison term that we observe in figures 5 and 6. Ifbetter educated women, for
example, do not decrease their welfare usage as much in the months leading up to their
first prison term, and we fail to take that into account, we might find that their post-prison

low. We find that when we remove the time effects, the temporal pattern is similar to that depicted in
Figure 5.
31
In particular, we allow the level of usage to vary by these characteristics in the 29 months prior to the
woman's first prison term. We also allow for a linear trend in usage that varies with these characteristics
in the 29 months leading up to the first prison term. The maintained assumption is that welfare receipt does
not vary with these characteristics in the 30th , or greater, months prior to prison.

24

welfare usage is higher than other women's simply because we did not correctly specify
their pre-prison usage.
Our results indicate that time served in prison does not appear to affect our
estimates of the effect of prison on social welfare use. As shown by the first row of Table
2, 32 both the OLS and fixed effect estimates indicate that women who served more than
the median prison spell had post-prison welfare receipt rates that were essentially the
same as those their counterparts who served shorter spells. If prison had an adverse
impact on women's social welfare receipt, then one would expect that a larger "dose" of
prison, as measured by the length of time spent there, would have a bigger adverse effect.
We see no effect of the size of the "dose." This finding is robust to allowing for preprison differences in welfare receipt for women who will serve longer than average
sentences.
As shown by the second row of the table, women who at some point have a drug
law violation have lower post-prison use of social welfare programs. Post-prison social
welfare use among such former prisoners was an additional 3 to 7 percentage point below
that of other observationally similar female felons.
One explanation for this finding is that women who are drug law violators also are
likely to report substance abuse problems when they enter prison. 33 These problems may
be associated with declines in life skills and such a decline might contribute to their
inability to qualify for benefits and so such women appear to become less connected with
32

We present estimates of the main effects of these variables in appendix table B. We observe that older
and better-educated women have lower rates of social welfare usage. Whereas African-American women,
women who have had children in the foster care system, and women who have ever violated a drug law
have higher rates of social welfare usage.

25

the social welfare system after prison. Another possibility that we explore in detail below
is whether these declines in social welfare receipt after prison is the result of policy
changes under PRWORA.
Turning to the other estimates in the table, it appears that after leaving prison, the
decline in social welfare use is somewhat greater among women whose characteristics are
associated with poor life skills. For example, less schooling is associated with lower
welfare rates after prison. As shown the first column of the table, each grade completed
is associated with about a 1. 7 percentage point increase in post-prison rates of welfare
receipt; the fixed-effect estimates are about one-half this magnitude. So we expect an exprisoner with 14 years of schooling to have social welfare receipt rates after prison that
are about 5 percentage points larger than an observationally similar ex-prisoner with 9
years of schooling. These results suggest that prison is associated with increased welfare
dependency only among the most educated former prisoners in our sample.
The results for women who ever had a child placed in the state foster care system
(FC women) are consistent with our view that women with declining life skills are likely
to have lower than expected use of social welfare programs after prison. But, we find that
these results are sensitive to the underlying econometric model and to the comparison
group used in the analysis. As shown by column 1 of the table, such women have sharply
declining rates of social welfare use after prison. The much smaller estimate associated
with the fixed-effect estimator indicates that these women are different from other felons.
One way that we find that they are different is that during the 29 months prior to their
incarcerations, their rates of social welfare use declined more sharply than it did for other
33

For women who serve in prisons, we have information both on their offenses and self-reported drug
abuse and addiction. For women who only have jail sentences, we only have information on their offense

26

felons. These sharper rates of decline are not controlled for with the FE estimator in
column 2
In column 3 ofTable 3, we account for differences in the decline of pre-prison
social welfare use among women who were drug felons, who had children ever placed
into the foster care system, and by schooling attainment. These estimates are indeed
smaller in magnitude, indicating that the OLS estimates in column 1 are biased
downward, because they do not take into account the variation in women's pre-prison
decline in social welfare use. As shown by column 4 of the table, now the OLS and fixed
effect estimates for the highest grade completed and ever foster care variables are about
the same.
The results for women who ever had children in foster care remain striking and
suggest sharp declines in use of social welfare programs after prison for them. But we
find that these results are very fragile and depend on the comparison group used in our
analysis. We compared the prison women to their counterparts who also had histories of
children in foster care, but only had one short jail spell. After re-estimating our model
using this sub sample, we find that after paroling from prison, social welfare use by these
two groups of incarcerated women are comparable, suggesting that time in prison does
not affect the post-prison use of social welfare programs differently for the FC women.
The foregoing finding suggests that our econometric model is misspecified for the
FC women. As shown by Panel B of Table 2, among women in the "No FC" group, the
post-prison interaction effects for time served, drug law violations, and highest grade
completed are the same as before (compare column 3 in Panel B to column 4 in Panel
A.). But the estimated coefficients associated with these interactions are different for the

type.

27

FC women. Although the coefficients are imprecisely estimated, they suggest that these
variables may influence post-prison social welfare use differently for these women than
they do for women who never had a child in the foster care system. The links between
prison, loss of parental rights, and use of the social welfare system is a topic that we will
address in future research.
Did PRWORA Affect Receipt ofSocial Welfare Benefits by Ex-Prisoners?

The lower post-prison rates of social welfare use by drug law violators suggest
that welfare reform might have affected these women's access to cash benefits. As
discussed above in Section II, all felons in Illinois, including drug felons, remained
eligible for Food Stamps. But serious drug felons face a lifetime ban from TANF, and
less serious drug felons may face a ban on benefits for up to two years. Here we ask
whether the temporal patterns of welfare receipt between the pre- and post-PRWORA
periods changed in a manner that is consistent with these policy changes.
To address this question, we first plot the monthly receipt rates of AFDC/TANF
relative to the dates of entry and exit from a woman's first prison spell for four groups of
women with prison records defined based on the date of their arrests and whether they
served time for drug-law violations .34 The four groups are as follows:
(i)

women who served prison time for a drug offense for which they were
first arrested prior to August 1996;

(ii)

women who served prison time for a drug offense who were arrested
after that date;

34

We do not have the date of arrest in our data. But we approximate this date with the date on which they
were incarcerated in county jail for the offense for which they were subsequently incarcerated in prison.

28

(iii)

women who served prison time for a non-drug-law violation for which
they were arrested prior to August 1996;

(iv)

women who served prison time for a non-drug law violation for which
they were arrested after August 1996.

In principle, only the second groups' eligibility for TANF was affected by PRWORA.
As shown by Figure 7, average TANFIAFDC receipt rates for all four groups fell
during the months leading up to women's incarcerations in prison and rise again
afterwards, but they do not return to their pre-prison levels. Among women arrested prior
to PRWORA for drug-law violations (the line with the diamonds), the monthly
TANFI AFDC receipt rates were about 30 percent one year prior to going to prison for the
first time. After this the date temporal pattern ofTANFI AFDC receipt follows the "V"
shape observed above for social welfare receipt overall. 35 After exiting prison, these
women's TANFIAFDC receipt rate rose steadily before leveling off at about 20 percent.
Among these pre-PRWORA drug law violators, the rate ofTAJ\TFIAFDC receipt is about
10 percent points lower one year after exiting prison than it was one year prior to entering
pnson.
Ifwe view the foregoing temporal pattern ofTANFIAFDC receipt as the baseline
pattern, we would expect that the gap between the receipt rates of the pre-PRWORA and
post-PRWORA drug law violators to widen during the post-prison period. To be sure, as
shown by (the solid line with the triangles in) Figure 7, the TANFIAFDC receipt rates
for the cohort of drug felons potentially affected by the PRWORA ban have been

35
The period 0 in Figure 7 encompasses all full months in which a woman was incarcerated in prison. In
principle, a woman is ineligible for welfare benefits while incarcerated so we expect receipt rates to be very
close to zero. Deviations from zero may result from administrative errors, or errors in matching prison and
welfare records.

29

consistently lower than those of their pre-PRWORA counterparts. However, the gap
between pre- and post- PRWORA drug felons' post-prison receipt rates did not widen
and in fact became somewhat smaller compared with the gap during their pre-prison
periods. During the year prior to prison the TANF/AFDC receipt rates of these two
cohorts differed by about 10 percentage points. During the year after prison, this gap also
was about 10 percentage points. If welfare reform operated as designed, we would expect
this post-prison gap between the pre- and post-PRWORA cohorts to have become
larger. 36 The last dots on right side ofFigure 7 show the average TANF/AFDC receipt
rates for all months 14 and beyond the months these women left prison. They indicate
that over the long-term, this gap became smaller.
The pattern observed in Figure 7 for post and pre-PRWORA drug felons is
supported by the temporal patterns ofTANFIAFDC receipt for non-drug law violators.
As shown by the two lines with squares and crosses in Figure 7, women incarcerated
prior to welfare reform for non-drug law violations had higher rates ofTANF/AFDC
receipt both before and after prison compared to women who were incarcerated for the
same kinds of offenses after PRWORA. Contrary to expectations, the difference between
TANF/AFDC receipt rates ofthe pre- and post-PRWORA non-drug law violators became
larger during the post-prison period. Significantly, the pattern ofTANF/AFDC receipt
among post-PRWORA non-drug felons is nearly identical to that for drug felons. By
contrast, during the pre-PRWORA period, the welfare receipt rates of non-drug felons
were consistently lower than those of the drug felons.
36

Given that we are studying each woman's first prison spell, the lower TANF receipt for the later cohort
of drug felons is not likely the result of some of them being banned from TANF because of a prior drug
offense. We believe the most likely explanation for the pattern observed in these data is that the later cohort

30

As shown by Figure 8, women in the four offender groups, described above who
spent less than one month in jail do not have "V" shaped T ANF? AFDC receipt rates like
the women who had been incarcerated in prison. Although all four groups of these short
jail women show a downward trend in their receipt rates, including women arrested for
drug-related offenses after PRWORA, there is no break in this trend around the time that
they are incarcerated in jail. In the short jail spell sample, we do not observe any
tendency for the welfare receipt rates to fall among post-PRWORA drug arrestees
relative to other women.
Turning from TANF/AFDC receipt to Food Stamp receipt, we observe that the
temporal patterns of Food Stamp receipt rates corroborate our interpretation of the
TANF/AFDC results. As shown by Figure 9, Food Stamp receipt rates are highest for
drug felons during the pre-PRWORA period. These rates decline during the months
leading up to prison and rise during the months after prison before leveling off at a rate
that is about 5 percentage points below pre-prison levels.
Among post-PRWORA drug-law violators, Food Stamp receipt rates start from a
lower level, follow a similar pattern, but return close to their pre-prison levels during the
post-prison period. The gap between the Food Stamp receipt rates of the pre- and postPROWRA drug law violators' is about the same as the gap between these two groups'
TANF/AFDC receipt rates. As we observed with TANF/AFDC receipt, we find that both
prior to and after PRWORA, the Food Stamp receipt rates of the drug and the non-drug
felons converge during the post-prison period.

of drug felons had lower rates ofTANF receipt because, as shown above in Figure 4, TANF caseloads in
general were low and falling during this period.

31

The foregoing discussion is based on the unadjusted mean receipt rates of
TANF/AFDC and Food Stamp benefits. To identify the effect ofPRWORA on
TANF/AFDC receipt, controlling for differences in characteristics and time periods, we
use a "triple differences-in-differences" specification. We compare the post-prison TANF
use of women arrested for drug law violations after welfare reform was enacted to other
felons. The term of interest then, is the coefficient on the triple interaction: whether the
woman was in prison for a drug law violation, whether her arrest was after PRWORA
was enacted, whether the current period is a post-prison month.
Besides this interaction term, we also add the following "main" and "interaction"
effects to our model and re-estimate equation (1):
(i)

whether she was arrested for a felony after PRWORA's enactment;

(ii)

whether the current period is a post prison period and the women was
arrested for a felony after PRWORA's enactment;

(iii)

whether a woman's incarceration was due to a drug law violation;

(iv)

whether the woman was incarcerated for a drug law violation and she was
arrested for a felony after PRWORA's enactment;

(v)

whether the current period is a post prison period and the women was
incarcerated for a drug law violation.

Our regression-adjusted results are consistent with our inferences from Figures 7
through 9 and reveal little evidence of substantially lower social welfare receipt rates
after prison among women who were incarcerated for drug-law violations after August
1996 compared (i) to women who committed similar crimes prior to that date and (ii) to
women who were incarcerated for non-drug felonies. As shown by the last row of Table

32

3, women affected by the lifetime or limited TANF bans received cash assistance at the
same rate as other felons after they were paroled from prison. The estimated standard
error associated the point estimate of -0.001 in the table suggest that the data is
consistent with views that TANF participation rates of drug felons rose or fell by as much
as 3 to 4 percentage points. In column 2, we observe that these conclusions are not
affected by adding controls for individual fixed-effects.
The remaining estimates in Table 3 suggest an explanation as to why some
observers have contended that PRWORA severely limited drug felons' access to TANF
and ultimately to other social welfare benefits even though we find no evidence of such
effects, at least in Illinois. The first two estimated coefficients in column 1 indicate that
women arrested after PRWORA had participation rates in TANF that were about 2
percentage points above their expected levels both before and after their incarcerations.
The coefficient estimate of 0.005 indicates that after prison this group's use ofTANF
rose by about !12 of a percentage point,
The next two estimated coefficients coeffienents in Table 3 indicate that drug
felons had TANF participation rates that average 5.4 percentage points above non-drug
felons throughout the sample period. But among drug felons arrested after PRWORA
this gap was only about 1.5 percentage points (i.e. 0.054- 0.039). This estimate implies
that during both the pre- and post-prison periods TANF use by post-PRWORA drug
felons was lower than it had been among earlier cohorts of drug felons. However, in the
fifth row in the table indicates throughout the sample period TANF use during the postprison period by drug felons was about 5 percentage points lower than is was for nondrug felons.

33

Therefore, taken together the foregoing estimates indicate that T ANF use among
drug felons after prison was substantially below expected levels based on their
TANF/AFDC use prior to entering prison and on TANF/AFDC receipt by drug felons
arrested prior to PRWORA. But this result is not caused by PRWORA's specific bans for
drug law violators. Instead, it appears that drug felons' post-prison AFDC/TANF
participation rates have always fallen more relative to their pre-prison levels compared
with non-drug felons. And further after PRWORA, TANF use among drug felons was
below pre-PRWORA levels even prior to going to prison. 37
As we explained above, because the PRWORA bans in Illinois apply only to
T ANF, but not to Food Stamps. Interestingly, the pattern of coefficients in in Table 3 is
nearly identical for Food Stamps and T ANF receipt rates. This corroborates our
contention that whatever is driving post-prison TANF receipt rates for female exoffenders, it is unlikely to be the specific bans on serious drug offenders receipt of cash
assistance.
Results not shown in Table 3 indicate that our findings also hold when we
estimate the short- and long-term effects ofPRWORA separately. In the short-termdefined here as the six months after leaving prison -- women covered by either the
lifetime or limited T ANF bans had modestly lower rates ofTANF receipt. But over the
long-term their relative rates ofTANF receipt were at expected levels.
Our analysis indicates that post-PRWORA drug felons in Illinois have not
decreased their use ofTANF due to policy changes intended to limit their access to

37

We checked to see whether this finding might result because after PRWORA, but prior to their first
incarcerations in prison, these women had been incarcerated in jail for a drug offense that was unrelated to
the arrest that led to their imprisonments. But this possibility does not account for our results the results
in table 3 control for imprisonment in jail and on a subsequent prison spell.

34

TANF. 38 One explanation for findings is that ex-prisoners and/or officials may be
confused about welfare policy believing that it limits access to TANF benefits to all
felons not just drug felons, and perhaps that the ban applies to Food Stamps as well as to
T ANF. However, rows l and 2 of Table 3, are not consistent with this interpretation as
they suggest a slight rise above expected levels in TANF and Food Stamp receipt rates
among non-drug felons during the post-PRWORA period.
The temporal pattern ofTANF/AFDC receipt observed in Figure 7 among women
potentially affected by the PRWORA bans also is consistent with the pattern of
TANF/AFDC receipt among low income women around the time ofwelfare reform.
During the pre-prison period, pre-PRWORA drug law violators had higher receipt rates,
because they went to prison when the TANF/AFDC caseload was relatively high. PostPRWORA drug violators have lower TANF receipt rates compared with their pre-TANF
counterparts largely due to the sharp decline in the T ANF case load in Illinois since
1997. 39

Finally, although it is true that the population of incarcerated women consists of a
disproportionately large fraction of drug law violators, in Illinois only about 15 percent of
them are classified as serious Class 1 or Class X drug felons. 40 The vast majority of

38

We discount the possibility that women who become ineligible forT ANF but remain eligible for Food
Stamps, now have a greater incentive to apply for and receive such benefits. In the past, women who
receive T ANF or AFDC benefits almost always receive Food Stamps. So in practice a policy banning drug
felons from T ANF, while still allowing them the receive Food Stamps, should not cause their Food Stamp
receipt rates to rise.
39
Statistics from the Illinois Department of Human Services indicated that the numbers of families
receiving AFDC/T ANF fell by nearly 80 percent from 198,923 per month in 1997 to 41,625 in August
2005. (ACF, 2004; IDHS, 2005)
40
About 327 women fall into the category of Class 1 or Class X drug felons arrested after PRWORA took
effect. This small number of serious drug law violators when compared to the entire T ANF case load raises
the question about whether it is worthwhile to enforce the PRWORA policy on drug offenders. One
practical problem faced by DHS is that any file that IDOC provided it with a listing of serious drug felons
does not contain an individual identifier that is both verified and common to both agencies. Although IDOC
files contain most female prisoners' Social Security numbers and names, these data are not verified by the

35

female drug offenders are covered by the limited TANF ban. However the legislature
provided these women with a variety of ways to have these bans waived so that in
practice few Illinois drug felons' eligibility for TANF appears to be affected by
PRWORA. In this light our finding that PRWORA appears to have had no effect on drug
felons' use ofTANF in Illinois is less surprising.
V. Conclusions
In this paper, we have used an unusual administrative data set to examine how
incarceration in prison affects women's subsequent receipt of social welfare benefits.
Our sample consists of women from Cook County, Illinois who were incarcerated in state
prison between 1993 and 2001 and comparison group ofwomen who were incarcerated
in the Cook County jail, with a particular focus on women who had one and only one jail
spell lasting less than a month. These data were matched to state administrative records
on AFDC/TANF receipt, Food Stamp receipt, and participation in Medicaid. The main
contribution of this paper is that our unique data allow us to examine women's interaction
with the social welfare system for a long period of time both before and after their
incarceration. This longer horizon allows us to dispute some claims about the likely
effects of prison on welfare dependency, while simultaneously helping to explain why
those conclusions made sense given the available data. For example, many believe that a
prison term will increase women's welfare dependency. Until now, the available data
would have supported such a claim -women who go to prison have very high rates of

agency. This information is self-reported by the inmate at the time she is admitted to prison. Instead,
IDOC's files include a verified identification number that is associated with a prisoner's fingerprint. DHS
files do not include this fingerprint ID. Because of reporting errors and missing Social Security numbers in
IDOC files, in practice it is not a straightforward task for DHS to verify without error whether an applicant
for TANF has been banned from TANF for life or faces a limited ban. Nor is it clear whether the applicant
herself would know which category she fell into-limited ban or lifetime ban-even if she truthfully reported
that she had a previous drug conviction.

36

welfare use, and they are higher six months after release from prison than they were
immediately before hand. Only with our long panel can we observe that this is because
a) women who go to prison have higher rates of welfare receipt than other wise similar
women in all periods, and b) women's welfare receipt drops precipitously in the months
just prior to their prison terms. Thus, a simple pre- and post-prison comparison will
inflate the impact of prison on welfare receipt.
Our analysis indicates that incarceration in prison is associated lower levels of
welfare dependency. During the first 4 to 5 months after parole, women's social welfare
receipt rates are far below expected levels. After one year they approach their expected
levels based on their own rates of receipt prior to prison and the receipt rates ofwomen
with a single short jail spell. Over the longer term, we estimate that having been in prison
is associated with a decline in receipt of social welfare benefits by about 6 percentage
points or about 15 percent below expected levels.
One question raised by our study is whether the reduced use of social welfare
programs among former prisoners who were more disadvantaged to begin with is an
indication of greater levels of long-term economic deprivation following prison or
whether it reflects a trend toward greater self-sufficiency. 41 Our analysis controls for
recidivism which could cause welfare use to decline in the years after a woman's first
prison term.
Finally, we find that the specific bans on drug offenders' receipt of cash
assistance that were introduced as part ofPRWORA cannot explain reduced social
41

One possibility is these women are more likely employed, though the patterns observed here work
against that contention. Cho and LaLonde (2005) use data from Illinois covering 1995 to 2001 to study the
links between incarceration and employment. They find that prison appears to be associated with higher

37

welfare usage among former state prisoners after their release from prison compared with
women with more modest contacts with the criminal justice system and poor women
more generally. However, our data can explain why the belief that these bans had a large
effect on female ex-offenders' access to cash assistance is widespread. Individuals
convicted of drug felonies after the PRWORA bans came into effect have lower TANF
receipt rates. However, our long panel demonstrates that this is a cohort effect- they
have lower receipt of cash assistance in all years (even those prior to TANF), and they
have lower rates of cash receipt prior to their first conviction. Furthermore, drug felons
convicted after the PRWORA bans on cash assistance also have lower receipt rates of
Food Stamps, a program not included in the bans in Illinois. Although there are a variety
of alternative explanations for these findings, they appear to reflect the general decline in
caseloads rather than the effect of changes in drug offenders' eligibility for cash
assistance.

employment rates during the year following exit, but over the longer term post and pre prison employment
rates are similar.

38

References

ACF (Administration for Children and Families, Department of Health and Human
Services) (2004) "Aid to Families with Dependent Children/Temporary Assistance for
Needy Families, Average Monthly Number ofFamilies and Recipients, 1997," Office of
Family Assistance, www.acf.dhhs.gov/programs/ofa/caseload/1997/FYCY97 .htm
Allard, Patricia (2002)."Life Sentences: Denying Welfare Benefits To Women Convicted
Of Drug Offenses," Washington D.C.: The Sentencing Project (February)
Bane, Mary Jo, and David Ellwood (1983). "The Dynamics ofDependence: The Routes
to Self Sufficiency," Prepared for the U.S. Department of Health and Human Services,
Office of the Assistant Secretary for Planning and Evaluation. Cambridge, MA: Urban
Systems Research and Engineering, Inc.
Chay, Kenneth, Hilary Hoynes, and Dean Hyslop (1999). "A Non-experimental Analysis
of"True" State Dependence in Monthly Welfare Participation Sequences," Working
Paper, October 1999.
Child Trends (2005). http://www.childtrendsdatabank.org/indicators/12FosterCare.cfm
Freeman, Richard (1999). "The Economics of Crime." Handbook of Labor Economics,
Orley Ashenfelter and David Card, eds., Amsterdam: Elsevier.
Goerge, Robert; Van Voorhis, John and Lee, Bong Joo (1994). Illinois Longitudinal and
Relational Child and Family Research Database. Social Science Computer Review, 12:3,
351-365.
Greenfeld, Lawrence, and Snell, Tracy (2000). Women Offenders. Bureau of Justice
Statistics Special Report. U.S. Department of Justice, Washington, D.C. NCJ 175688
Jacobson, Louis, Robert LaLonde and D. Sullivan "Earnings Losses of Displaced
Workers" American Economic Review (September 1993): 685-709.
Heckman, James and Burton Singer (1984). "Econometric Duration Analysis," Journal of
Econometrics, 24, 63 - 132.
Heckman, James, Robert LaLonde, and Jeffrey Smith ''The Economics and Econometrics
of Active Labor Market Policies," in The Handbook of Labor Economics, eds. Orley
Ashenfelter and David Card, Amsterdam: North-Holland, 1999.
Holzer, Harry and Robert LaLonde (2000). "Job Change and Job Stability Among Less
Skilled Young Workers," in Finding Jobs: Work and Welfare Reform, David Card and
Rebecca Blank, eds., New York: Russell Sage Foundation.

39

Hausman, J.A., Jason Abrevaya, and F.M. Scott-Morton (1998). "Misclassification of the
Dependent Variable in a Discrete-response Setting," Journal of Econometrics 87: 239269.
IDES (Illinois Department ofHuman Services) (2005). "DHS TANF Statistics,"
http://www .dhs.state.il. us/ts/fss/tanfstatistics.asp
Lancaster, Tony (1990). The Econometric Analysis of Transition Data. Cambridge:
Cambridge University Press.
Moffitt, Robert, (2001). "Economic Effects ofMeans-Tested Transfers in the U.S.,"
Prepared for the 2001 NBER Tax Policy and the Economy Conference, October 30,
2001.
Moffitt, Robert (1992). "The Incentive Effects ofthe U.S. Welfare System: A Review,"
Journal of Economic Literature, 30(March): 1-61.
Mumola, Christopher (2000). Incarcerated Parents and Their Children. Bureau of Justice
Statistics Special Report. U.S. Department of Justice: Washington, D.C. NCJ 182335.
Pavetti, LaDonna, Michelle Derr, Getchen Kirby, Robert Wood, and Melissa Clark
(2004). The Use ofTANF Work-Oriented Sanctions in Illinois, New Jersey, and South
Carolina, Final Report April30, 2004 Washington D.C.: Mathematica Policy Research
Corp. HHS-1 00-01-0011.
The Sentencing Project (2004). "Life Sentences:Denying Welfare Benefits To Women
Convicted ofDrug Offenses, State modifications updated February 2004," Fact Sheet,
State of Illinois (2003). "Plan for Temporary Assistance for Needy Families
October 1, 2003- September 30, 2005," October 1, 2003.

40

Appendix A

In the presence of state dependence, the short and long-term effects of
incarceration should differ. To see this point, consider the following model that
determines whether a woman receives food stamps during any given month:

where FSi,t is a dummy variable equal to one, if a woman received Food Stamps in month
t ; and Dit is a dummy variable equal to one if a women has previously been to prison.
The parameters of this model determine the two transition rates from Food Stamp receipt
to non-receipt and from non-receipt to Food Stamp receipt. For example, the probability
of receiving Food Stamps in month t, given that a woman was offFood Stamps in month
t-1 is given by a 0 for women without a prison record.
When a woman is admitted to prison the process that determines whether she
receives Food Stamps is interrupted. Individuals in prison are ineligible to receive
benefits and prison officials regularly provide lists of new prison admissions to state
welfare authorities. Therefore, we expect that when a woman exits prison during months,
the Food Stamp indicator variable equals zero (i.e., FSi,s = 0). The statistical model
described above indicates that the probability that these women will receive Food Stamps
during months+ 1 is given by ao + Ot. By contrast, the rate of Food Stamp receipt during
that month for a comparison group is given by the steady state rate of a 0/( 1-a. 1). Because
a 1 is larger than a 0 , and assuming the value of &1 is not too large, we expect the rate of
Food Stamp receipt by recently released offenders will be less than that rate for nonoffenders. This difference, however, is our estimate ofthe effect of prison on food stamp
receipt 8,""1 = E(FSi;r=I,t I Di,t=O,t = 1) E(FSi,t I Di,t = 0, for all t) = (ao +&I) (a.o/(1-a.I)).
Notice that if prison has no effect on welfare transition rates (i.e. = 0), we expect the ·
effect of prison on social welfare receipt, during the first full month out of prison to be
negative.
During the months following their release from prison the fraction of woman
receiving food stamps will rise relative to rate for the comparison group. In our statistical
model (1 ), we intend that the parameter o,(Si, ~) to capture these mechanical effects of
prison during the 't months after prison. Because of state dependence, the effect of prison
on the rate ofFood Stamp receipt will rise with time since exit ('t>O) as more women
enter Food Stamps and stay on the program. This result is not a general result, of course,
and depends on the values of the parameters o 1 and &2 and whether they are time varying
in 't. The key point is that the process described by (3) implies that the impact of
incarceration on offenders use of social welfare programs should be time varying.

41

42

Figures 1-4
Fraction of Former Female State Prisoners Ever Receiving Social Welfare Benefits
between January 1, 1990 and June 30, 2001
Figure 2
Receive AFDC/T ANF

Figure 1
Ever Receive any Welfare
.47
.27

.45

.24
.42
~
c:
~

=

.,

1-

~

.39

=
.,

"'"'

"'

c:

0

<

-<=

=

"'

=

.36

"'
c:

.15

.2

.2

0

0

::"

.18

=

c:

u_

.21

<!l

~

.33

u..

'l

.00
.05~

.27

19roffi1

.12

I

1991m9

I

1993m5

19!im1

I

1900m9

I

19!&15

Yrsa'ld~h

~1 2001~7

I

1900m1

"'
...
"'

"'

1900m9

19!&15

mlm1 2001m7

.32

:::IE

=
=

.29

=

.!il

c:

0

.!2

~

u

~

.35

u

.32

u_

LL.

2()&1

:s!

0
0

"'

19!&5

.36

.35

~

=
=

2001~7

1~9

.41

E

...

1900m1

.43

.36

(/)

19!t3m5

Yrsa'ld~

.4

"'
<>-

1wlm9

u_

.26

.29

.23

.26

.2

.23

1900m1

1991m9

1993m5

1995m1

1900m9

199&n5

mlm1 2001m7

YI!Ha'ld~h

Figure 3
Received Food Stamps

1900m1

1991m9

1993m5

1995m1

YI!Ha'ld~

Figure 4
Received Medicaid
43

Figure 5: Fraction Receiving Any Transfers in Months Before and After 1st Incarceration (Jail
or Prison)
0.5

®
~.

~ 0.45

.
u

!i

'6
:I<

0.4

~

:!{

.
Ul

E 0.35

-g

~

0.3

(1_

u:
z

~
g
...

0.25 c-------0.2

<(

g>
·:; 0.15

-~

~

0.1

~

~----------~---

----

....

I

if!
\
v

I

)

0.05

0

-

~
Ji

1

~;

-~
lJ

l~

0

~~~~~~~p~~~~~~,~~~~~~~~~~v~~

Months before and After 1st Incarceration

•

--Only 1 Short Jail Spell
---Jail Only
--1st Prison

Figure 6: The Impact of Incarceration in State Prison on Women's Social Welfare Receipt

3

5

7

Time In Months Prior to or After Prison

45

9

11

13 15 17

19 21

23

Figure 7: Fraction of Felons Receiving AFDCITANF in months Before and After 1st Prison
Term, by Arrest Date pre-post August 1996
0.5
0.45
0.4
0.35

.....

z
<(
t::
u

0.3

............

~--~---

<(

c::: 0.25

0

c:::

.2
t)

~

:--Drug Felony-prior to ban
: ----- Other-prior to ban
--...-Drug Felony-after ban
Other-after ban

Q

.....

0.2

!!
.....

~~~~~~~~~~~~~~,~~~~~~~~~0~~~

Months Before and AFter 1st Prison Term

46

Figure 8: Fracton Receiving AFDCfTANF in months Before and After Incarceration in Only Jail
Spell, By Offense Category and Arrest pre-post August 1996
0.5
0.45
0.4
~ 0.35

5
.r
'§
~

0.3

0.25

':;

......

~·
I

...

--

0.2

0

J

,.;

;,;..

0.15

---

~

~"'.

. ""'"'~"""··;< ·'-'

0.1

1\
....... •
"
\I
I

0.05
0
-13-12-11 -10 -9

..a

-7 -6 -5 -4 -3 -2 -1

0

1

2

3

4

5

6

7

8

Months Before and After Only Jail Spell (Jail Spell< 30 days)

47

9

10 11 12 13 14

__.Drugs-prior to ban
--Other-prior to ban
--Drugs-after ban
~'!c~ Other-after ban

Figure 9: Fraction Receiving Food Stamps In Months Before and After 1st Prison Term, by
Arrest pre-post August 1996

~~~~~~~~~~~~~~,~~~~~~~~~0~~~

Months Before and After 1st Prison Term

48

Table 1: Means For Sample of Incarcerated Women From Cook County, Dlinois
(Standard Deviations)
Person Records
Person-Month Records: Monthly Average
35.7
ReceiveAnyWelfare
0.346
Age on July 1, 200 I
(9.01)
(FoodStamps,Medicaid
(0.476)
or Cash Grant)
0.301
0.756
Receive Food Stamps
Fraction African
(0.459)
(0.430)
American
Receive Medicaid
Fraction White
0.180
0.303
(0.459)
(0.384)
Receive Cash Grant
0.204
Fraction Hispanic
0.060
(0.403)
(0.237)
0.025
0.004
Fraction Other Race
In Jail
(0.155)
(0.062)
Highest Grade
11.6
In Prison
0.014
(1.74)
(0.119)
Completed
Age
30.00
Nwnber of Children
2.3
(1.94)
(9.62)
0.581
Ever Receive Food
(0.493)
Stamps
0.572
Ever Receive Medicaid
(0.495)
Ever Receive
0.461
(0.498)
AFDC/TANF
Ever Receive FS, Med
0.600
or AFDC/TANF
(0.490)
0.133
Ever Have a Child in
(0.340)
Foster Care
Ever have Drug Law
0.319
Violation1
(0.466)
0.962
Ever have a Jail Term
(0.190)
Ever have a Prison
0.120
(0.324)
Term2
0.414
1 Short Jail Spell(< I
(0.493)
month)
2.00
Average Nwnber of
(1.72)
Criminal Justice Spells
1.28
Average Months Served
4.68
Notes: The data are linked administrative records for women with Cook County Jail records from January
1990 through June 200 I. Administrative come from: for Cook County Department of Corrections, Illinois
Department of Corrections, illinois Department of Children and Family Services, and Department of
Hwnan Services. We begin with data on over 50,000 women. This sample includes only those women
v.rhose first criminal justice spell in our data began after January 1993, to help ensure that the spell we
designate as the first criminal justice spell is really the first. The last month included here is June 200 I.
Our working sample is a 7 5% random subset of the women with "Jail Only" records, and all the women
v.rho ever had a prison record. Weights are used to adjust the samples. See text for additional information.
t.'Ever Have Drug Violation" indicates that we see either a jail or a prison term for a drug law violation at
some point. 2 Criminal justice spells can be "Jail Only," "Combined Jail/Prison," or "Prison Only." The last
is rare, and we have combined the "Jail/Prison" and "Prison only" spells here. The data are in month-person
form. If one has any jail time in a month, for example, one is coded as having jail= I for that month. The
minimwn value for time served is zero, v.rhich would include spells lasting less than one month duration.

49

Table 2
Differences in Effect of Prison on Welfare Receipt,
by Characteristic of Woman and Offense Category

Panel A: Full Sam(!le
No Controls for
Pre-Prison Effects
FE
OLS
(2)
(1)
0.008
-0.014
(0.009)
(0.011)

Controls for
Pre-Prison Effects
OLS
FE
(3)
(4}
0.004
0.009
(0.011)
(0.009)

Post * Drug Law Violation

-0.072
(0.011)

-0.044
(0.009)

-0.028
(0.012)

-0.053
(0.010)

Post * Highest Grade Completed

0.017
(0.003)

0.008
(0.002)

0.011
(0.003)

0.010
(0.003)

Post* Ever Child in Foster Care

-0.162
(0.012)

-0.068
(0.011)

-0.107
(0.013)

-0.096
(0.012)

Post * Time Served > 8 months

Panel B: Se(!arate Estimates By Whether Women Ever Had a Child in Foster Care
No Controls for PreControls for PrePrison Effects
Prison Effects
NoFC
FC
NoFC
FC
(4)
(2)
(3)
(1)
0.008
0.020
0.008
0.020
Post * Time Served > 8 months
(0.020)
(0.010)
(0.010)
(0.020)
Post * Drug Law Violation

-0.055
(0.010)

-0.002
(0.021)

-0.063
(0.011)

-0.012
(0.023)

Post * Highest Grade Completed

0.008
(0.003)

0.005
(0.006)

0.011
(0.003)

0.005
(0.006)

Fixed Effects

Yes

Yes

Yes

Yes

Notes: The sample has 20169 women (2803491 observations) and includes women with one and only 1 jail
spell lasting less than a month from a random subset of all women with only jail spells, and all the women
who had a prison spell. The estimates in the table give the effect of the indicated characteristic on the longterm effect of prison on social welfare receipt based on equation ( 1) in the text. The OLS estimate of -.072
implies that women with drug law violations have post-prison welfare receipt rates that are about 7
percentage points less than observationally similar women with person-related or property-related offenses.
The term "Controls for Pre-Prison Effects" indicates that the model includes separate effects for the level
and change in social welfare receipt during the 29 months prior to prison for each characteristic. "FE"
refers to fixed effect estimates. FC refers to the subsample of short jail and prison women who ever had a
child in foster care. Other variables included in the model include race/ethnicity, time varying age, highest
grade completed, and number of children.

50

Table 3
Impact of PRWORA on Drug Felons Monthly Participation Rates in
TANF and Food Stamps

Variable
Arrested After PRWORA?

Monthly Participation Rate in:
TANF/AFDC
Food Stamps
(1)
(2)
(3)
(4)
0.020
0.034
(.005)
(.005)

Post-Prison* Arr. After PRWORA? 0.005
(.012)

0.013
(.013)

0.009
(.015)

0.054
(.010)

0.060
(.010)

Arr. AfterPRWORA* Drug Felon? -0.039
(.012)

0.018
(.013)

-0.041
(.013)

Drug Felon?

-0.051
(.011)

-0.039
(.010)

-0.054
(0.012)

-0.040
(.011)

Post-Prison* Arr. After PRWORA* -0.001
(.016)
Drug Felon?

-0.011
(.016)

-0.001
(.019)

-0.008
(.017)

Fixed Effects

Yes

No

Yes

Post-Prison*Drug Felon?

No

Notes: Coefficients in row 1 are estimates ofTANF and Food Stamp participation of non-drug felons
relative to the baseline participation rates of non-drug felons arrested before PRWORA was enacted in
August 1996. The post-prison interaction term in row 2 is an estimate ofhow non-drug felons' welfare use
changed during the post-prison period if they were arrested after welfare reform. In row 3, we control for
whether the woman was a drug felon; and in row 4 whether the drug felon was arrested after welfare reform
was enacted. The coefficients estimates in row 5 indicate the average difference between post-prison
period welfare use by drug felons compared with non-drug felons. The coefficient estimates are for the
"triple differences" term is in row 6. The model also includes controls whether a women is currently in
prison for her first prison spell, in jail, or in prison for a subsequent offense or a parole violation. Year and
month effects also are controlled for to account the effects of the economy and policy on average
participation rates. Other variables included in the model are described in the text. The numbers in
parentheses are the standard errors. The standard errors for the fixed effect estimates are unadjusted.
Source: Authors' calculations from the merged IDOC, CCDOC, and IDB data base described in the text.

51

Appendix Table A
Committing Offenses for IDOC Female Inmates from Cook County,
Illinois: 1993-2002.
(Percentage of Inmates by Cate2ory with Indicated Criminal Offense)
Education

Race

Offense Type
Cook
Person
County · Crimes
Crimes Other
Property
Crimes
Retail Theft
< $150
Retail Theft
> $150
Fraud&
Forgery
Controlled
Sub
Manufacture/
Delivery
[ Controlled
. Sub
, Possession
· Cannabis
1

Age

Children

Under 30

3 or more

Total

Black

White

<
High
School

14.3%

14.0%

14.3%

15.3%

13.5%

11.3%

20.8%

11.8%

8.9%. 16.9%

9.7%

9.5%

12.9%

11.1%

8.5%

High
School
/GED

Beyon
dHigh
School

i

10.1%
10.3%

10.3%

11.3%

8.5%

12.5%

13.9%

7.6%

10.1%

6.0%

6.4%

4.8%

5.2%

7.6%

7.2%

3.9%

6.1%

5.2%

5.0%

7.6%

2.9%

6.8%

12.9%

3.5%

4.6%
i

29.2%

32.0%

11.0%

33.2%

24.8%

19.0%

31.4%

32.5%

21.4%

20.7%

26.6%

21.8%

21.4%

19.8%

18.4%

23.2%

0.7%

0.6%

0.7%

0.6%

0.8%

1.0%

1.0%

0.7%

1.9%

1.3%

5.1%

2.0%

2.2%

1.0%

1.4%

1.8%

0.8%

0.7%

1.7%

0.6%

1.0%

1.1%

0.9%

0.6%

100%

100%

100%

100%

100%

1

i

Sex Crimes
Other Crimes
Total

i

100%

100% 1100%

i

Appendix Table B
Coefficient Estimates for Other Variables in Figure 6 and Table 2
Control
Variables

Figure 6
OLS

Figure 6
Fixed
Effects

Table 2
OLS
Panel A,
col. (1)

Table 2
Never Had
Child in FC
Fixed Effects
Panel B, col. 1
0.004
(0.001)
-0.00009
(0.00002)

Table 2
Ever Had Child
inFC
Fixed Effects
Panel B, col. 2
-0.020
(0.005)
0.00007
(0.00008)

-0.006
-0.006
0.003
(0.001)
(0.001)
(0.001)
0.00005
-0.00009 0.00005
Age Squared
(0.00002) (0.00002) (0.00002)
0.208
0.208
African-Amer.
(0.006)
(0.006)
0.065
0.065
Hispanic
(0.010)
(0.010)
-0.036
-0.035
Other Race
(0.023)
(0.023)
0.018
0.018
# of Children
(0.002)
(0.002)
0.010
0.009
Time Served
(0.008)
(0.008)
> 8 months
-0.017
-0.016
Highest Grade
(0.001)
(0.001)
Completed
0.178
0.199
Ever Any Kids
(0.007)
(0.007)
in Foster Care
0.056
Ever Any Drug 0.050
(0.006)
(0.006)
Law Violation
0.0840
0.0862
R-squared
2803491
2803491
Number of Obs. 2803491
2451265
352226
Notes: See table 1 for a description of the data. Data are person-month records from
January 1990 to June 2001. We exclude those whose first prison term is before January
1993, in order to ensure that the 1st prison term we observe is the actual first prison term.
Linear probability models; the standard errors are clustered on the individual to account
for multiple observations per individual. The dependent variable is equal to 1 if the
individual received food stamps, Medicaid, or a cash grant in a given month. These
regressions examine welfare participation in the months before and after the person's first
prison term (either a prison "only" term or a jail/prison term combined). The regressions
also include a constant, a dummy for each month, a dummy for each year, 17 dummy
variables controlling for 29 months prior to the first full month in prison, and 14 dummies
indicating the number of months after the last full month of the first prison term, a
dummy for people with highest grade completed greater than or equal to 18, and a
dummy variable if a given month falls during the first prison term, during a prior or
subsequent jail term, or a subsequent prison term. See Figure 6 for how welfare use
changes in the years surrounding the first prison spell. See table 2 for how welfare use in
the post prison period differs by individual characteristics.
Age

54

Working Paper Series
A series of research studies on regional economic issues relating to the Seventh Federal
Reserve District, and on financial and economic topics.
A Proposal for Efficiently Resolving Out-of-the-Money Swap Positions
at Large Insolvent Banks
George G. Kaufman

WP-03-01

Depositor Liquidity and Loss-Sharing in Bank Failure Resolutions
George G. Kaufman

WP-03-02

Subordinated Debt and Prompt Corrective Regulatory Action
Douglas D. Evanoff and Larry D. Wall

WP-03-03

When is Inter-Transaction Time Informative?
Craig Furfine

WP-03-04

Tenure Choice with Location Selection: The Case of Hispanic Neighborhoods
in Chicago
Maude Toussaint-Comeau and Sherrie L.W. Rhine

WP-03-05

Distinguishing Limited Commitment from Moral Hazard in Models of
Growth with Inequality*
Anna L. Paulson and Robert Townsend

WP-03-06

Resolving Large Complex Financial Organizations
Robert R. Bliss

WP-03-07

The Case of the Missing Productivity Growth:
Or, Does information technology explain why productivity accelerated in the United States
but not the United Kingdom?
Susanto Basu, John G. Fernald, Nicholas Oulton and Sylaja Srinivasan

WP-03-08

Inside-Outside Money Competition
Ramon Marimon, Juan Pablo Nicolini and Pedro Teles

WP-03-09

The Importance of Check-Cashing Businesses to the Unbanked: Racial/Ethnic Differences
William H. Greene, Sherrie L.W. Rhine and Maude Toussaint-Comeau

WP-03-10

A Firm’s First Year
Jaap H. Abbring and Jeffrey R. Campbell

WP-03-11

Market Size Matters
Jeffrey R. Campbell and Hugo A. Hopenhayn

WP-03-12

The Cost of Business Cycles under Endogenous Growth
Gadi Barlevy

WP-03-13

The Past, Present, and Probable Future for Community Banks
Robert DeYoung, William C. Hunter and Gregory F. Udell

WP-03-14

1

Working Paper Series (continued)
Measuring Productivity Growth in Asia: Do Market Imperfections Matter?
John Fernald and Brent Neiman

WP-03-15

Revised Estimates of Intergenerational Income Mobility in the United States
Bhashkar Mazumder

WP-03-16

Product Market Evidence on the Employment Effects of the Minimum Wage
Daniel Aaronson and Eric French

WP-03-17

Estimating Models of On-the-Job Search using Record Statistics
Gadi Barlevy

WP-03-18

Banking Market Conditions and Deposit Interest Rates
Richard J. Rosen

WP-03-19

Creating a National State Rainy Day Fund: A Modest Proposal to Improve Future
State Fiscal Performance
Richard Mattoon

WP-03-20

Managerial Incentive and Financial Contagion
Sujit Chakravorti and Subir Lall

WP-03-21

Women and the Phillips Curve: Do Women’s and Men’s Labor Market Outcomes
Differentially Affect Real Wage Growth and Inflation?
Katharine Anderson, Lisa Barrow and Kristin F. Butcher

WP-03-22

Evaluating the Calvo Model of Sticky Prices
Martin Eichenbaum and Jonas D.M. Fisher

WP-03-23

The Growing Importance of Family and Community: An Analysis of Changes in the
Sibling Correlation in Earnings
Bhashkar Mazumder and David I. Levine

WP-03-24

Should We Teach Old Dogs New Tricks? The Impact of Community College Retraining
on Older Displaced Workers
Louis Jacobson, Robert J. LaLonde and Daniel Sullivan

WP-03-25

Trade Deflection and Trade Depression
Chad P. Brown and Meredith A. Crowley

WP-03-26

China and Emerging Asia: Comrades or Competitors?
Alan G. Ahearne, John G. Fernald, Prakash Loungani and John W. Schindler

WP-03-27

International Business Cycles Under Fixed and Flexible Exchange Rate Regimes
Michael A. Kouparitsas

WP-03-28

Firing Costs and Business Cycle Fluctuations
Marcelo Veracierto

WP-03-29

Spatial Organization of Firms
Yukako Ono

WP-03-30

Government Equity and Money: John Law’s System in 1720 France
François R. Velde

WP-03-31

2

Working Paper Series (continued)
Deregulation and the Relationship Between Bank CEO
Compensation and Risk-Taking
Elijah Brewer III, William Curt Hunter and William E. Jackson III

WP-03-32

Compatibility and Pricing with Indirect Network Effects: Evidence from ATMs
Christopher R. Knittel and Victor Stango

WP-03-33

Self-Employment as an Alternative to Unemployment
Ellen R. Rissman

WP-03-34

Where the Headquarters are – Evidence from Large Public Companies 1990-2000
Tyler Diacon and Thomas H. Klier

WP-03-35

Standing Facilities and Interbank Borrowing: Evidence from the Federal Reserve’s
New Discount Window
Craig Furfine

WP-04-01

Netting, Financial Contracts, and Banks: The Economic Implications
William J. Bergman, Robert R. Bliss, Christian A. Johnson and George G. Kaufman

WP-04-02

Real Effects of Bank Competition
Nicola Cetorelli

WP-04-03

Finance as a Barrier To Entry: Bank Competition and Industry Structure in
Local U.S. Markets?
Nicola Cetorelli and Philip E. Strahan

WP-04-04

The Dynamics of Work and Debt
Jeffrey R. Campbell and Zvi Hercowitz

WP-04-05

Fiscal Policy in the Aftermath of 9/11
Jonas Fisher and Martin Eichenbaum

WP-04-06

Merger Momentum and Investor Sentiment: The Stock Market Reaction
To Merger Announcements
Richard J. Rosen

WP-04-07

Earnings Inequality and the Business Cycle
Gadi Barlevy and Daniel Tsiddon

WP-04-08

Platform Competition in Two-Sided Markets: The Case of Payment Networks
Sujit Chakravorti and Roberto Roson

WP-04-09

Nominal Debt as a Burden on Monetary Policy
Javier Díaz-Giménez, Giorgia Giovannetti, Ramon Marimon, and Pedro Teles

WP-04-10

On the Timing of Innovation in Stochastic Schumpeterian Growth Models
Gadi Barlevy

WP-04-11

Policy Externalities: How US Antidumping Affects Japanese Exports to the EU
Chad P. Bown and Meredith A. Crowley

WP-04-12

Sibling Similarities, Differences and Economic Inequality
Bhashkar Mazumder

WP-04-13

3

Working Paper Series (continued)
Determinants of Business Cycle Comovement: A Robust Analysis
Marianne Baxter and Michael A. Kouparitsas

WP-04-14

The Occupational Assimilation of Hispanics in the U.S.: Evidence from Panel Data
Maude Toussaint-Comeau

WP-04-15

Reading, Writing, and Raisinets1: Are School Finances Contributing to Children’s Obesity?
Patricia M. Anderson and Kristin F. Butcher

WP-04-16

Learning by Observing: Information Spillovers in the Execution and Valuation
of Commercial Bank M&As
Gayle DeLong and Robert DeYoung

WP-04-17

Prospects for Immigrant-Native Wealth Assimilation:
Evidence from Financial Market Participation
Una Okonkwo Osili and Anna Paulson

WP-04-18

Individuals and Institutions: Evidence from International Migrants in the U.S.
Una Okonkwo Osili and Anna Paulson

WP-04-19

Are Technology Improvements Contractionary?
Susanto Basu, John Fernald and Miles Kimball

WP-04-20

The Minimum Wage, Restaurant Prices and Labor Market Structure
Daniel Aaronson, Eric French and James MacDonald

WP-04-21

Betcha can’t acquire just one: merger programs and compensation
Richard J. Rosen

WP-04-22

Not Working: Demographic Changes, Policy Changes,
and the Distribution of Weeks (Not) Worked
Lisa Barrow and Kristin F. Butcher

WP-04-23

The Role of Collateralized Household Debt in Macroeconomic Stabilization
Jeffrey R. Campbell and Zvi Hercowitz

WP-04-24

Advertising and Pricing at Multiple-Output Firms: Evidence from U.S. Thrift Institutions
Robert DeYoung and Evren Örs

WP-04-25

Monetary Policy with State Contingent Interest Rates
Bernardino Adão, Isabel Correia and Pedro Teles

WP-04-26

Comparing location decisions of domestic and foreign auto supplier plants
Thomas Klier, Paul Ma and Daniel P. McMillen

WP-04-27

China’s export growth and US trade policy
Chad P. Bown and Meredith A. Crowley

WP-04-28

Where do manufacturing firms locate their Headquarters?
J. Vernon Henderson and Yukako Ono

WP-04-29

Monetary Policy with Single Instrument Feedback Rules
Bernardino Adão, Isabel Correia and Pedro Teles

WP-04-30

4

Working Paper Series (continued)
Firm-Specific Capital, Nominal Rigidities and the Business Cycle
David Altig, Lawrence J. Christiano, Martin Eichenbaum and Jesper Linde

WP-05-01

Do Returns to Schooling Differ by Race and Ethnicity?
Lisa Barrow and Cecilia Elena Rouse

WP-05-02

Derivatives and Systemic Risk: Netting, Collateral, and Closeout
Robert R. Bliss and George G. Kaufman

WP-05-03

Risk Overhang and Loan Portfolio Decisions
Robert DeYoung, Anne Gron and Andrew Winton

WP-05-04

Characterizations in a random record model with a non-identically distributed initial record
Gadi Barlevy and H. N. Nagaraja

WP-05-05

Price discovery in a market under stress: the U.S. Treasury market in fall 1998
Craig H. Furfine and Eli M. Remolona

WP-05-06

Politics and Efficiency of Separating Capital and Ordinary Government Budgets
Marco Bassetto with Thomas J. Sargent

WP-05-07

Rigid Prices: Evidence from U.S. Scanner Data
Jeffrey R. Campbell and Benjamin Eden

WP-05-08

Entrepreneurship, Frictions, and Wealth
Marco Cagetti and Mariacristina De Nardi

WP-05-09

Wealth inequality: data and models
Marco Cagetti and Mariacristina De Nardi

WP-05-10

What Determines Bilateral Trade Flows?
Marianne Baxter and Michael A. Kouparitsas

WP-05-11

Intergenerational Economic Mobility in the U.S., 1940 to 2000
Daniel Aaronson and Bhashkar Mazumder

WP-05-12

Differential Mortality, Uncertain Medical Expenses, and the Saving of Elderly Singles
Mariacristina De Nardi, Eric French, and John Bailey Jones

WP-05-13

Fixed Term Employment Contracts in an Equilibrium Search Model
Fernando Alvarez and Marcelo Veracierto

WP-05-14

Causality, Causality, Causality: The View of Education Inputs and Outputs from Economics
Lisa Barrow and Cecilia Elena Rouse

WP-05-15

5

Working Paper Series (continued)
Competition in Large Markets
Jeffrey R. Campbell

WP-05-16

Why Do Firms Go Public? Evidence from the Banking Industry
Richard J. Rosen, Scott B. Smart and Chad J. Zutter

WP-05-17

Clustering of Auto Supplier Plants in the U.S.: GMM Spatial Logit for Large Samples
Thomas Klier and Daniel P. McMillen

WP-05-18

Why are Immigrants’ Incarceration Rates So Low?
Evidence on Selective Immigration, Deterrence, and Deportation
Kristin F. Butcher and Anne Morrison Piehl

WP-05-19

Constructing the Chicago Fed Income Based Economic Index – Consumer Price Index:
Inflation Experiences by Demographic Group: 1983-2005
Leslie McGranahan and Anna Paulson

WP-05-20

Universal Access, Cost Recovery, and Payment Services
Sujit Chakravorti, Jeffery W. Gunther, and Robert R. Moore

WP-05-21

Supplier Switching and Outsourcing
Yukako Ono and Victor Stango

WP-05-22

Do Enclaves Matter in Immigrants’ Self-Employment Decision?
Maude Toussaint-Comeau

WP-05-23

The Changing Pattern of Wage Growth for Low Skilled Workers
Eric French, Bhashkar Mazumder and Christopher Taber

WP-05-24

U.S. Corporate and Bank Insolvency Regimes: An Economic Comparison and Evaluation
Robert R. Bliss and George G. Kaufman

WP-06-01

Redistribution, Taxes, and the Median Voter
Marco Bassetto and Jess Benhabib

WP-06-02

Identification of Search Models with Initial Condition Problems
Gadi Barlevy and H. N. Nagaraja

WP-06-03

Tax Riots
Marco Bassetto and Christopher Phelan

WP-06-04

The Tradeoff between Mortgage Prepayments and Tax-Deferred Retirement Savings
Gene Amromin, Jennifer Huang,and Clemens Sialm

WP-06-05

Why are safeguards needed in a trade agreement?
Meredith A. Crowley

WP-06-06

6

Working Paper Series (continued)
Taxation, Entrepreneurship, and Wealth
Marco Cagetti and Mariacristina De Nardi

WP-06-07

A New Social Compact: How University Engagement Can Fuel Innovation
Laura Melle, Larry Isaak, and Richard Mattoon

WP-06-08

Mergers and Risk
Craig H. Furfine and Richard J. Rosen

WP-06-09

Two Flaws in Business Cycle Accounting
Lawrence J. Christiano and Joshua M. Davis

WP-06-10

Do Consumers Choose the Right Credit Contracts?
Sumit Agarwal, Souphala Chomsisengphet, Chunlin Liu, and Nicholas S. Souleles

WP-06-11

Chronicles of a Deflation Unforetold
François R. Velde

WP-06-12

Female Offenders Use of Social Welfare Programs Before and After Jail and Prison:
Does Prison Cause Welfare Dependency?
Kristin F. Butcher and Robert J. LaLonde

WP-06-13

7