Measuring Intergenerational Exposure to the U.S. Justice System:
Evidence from Longitudinal Links between Survey and Administrative Data
Keith Finlay

Michael Mueller-Smith†

Brittany Street‡§

U.S. Census Bureau

University of Michigan

University of Missouri

June 9, 2022
Abstract
Intergenerational exposure to the justice system is both a marker of vulnerability among
children and a measurement of the potential unintended externalities of crime policy in
the U.S. Estimating the size of this population has been hampered by inadequate data
resources, including the inability to (1) observe non-incarceration sources of exposure,
(2) follow children throughout their childhood, and (3) measure multiple adult influences in increasingly dynamic households. To overcome these challenges, we leverage
billions of restricted administrative and survey records linked with the Criminal Justice
Administrative Records System (CJARS). We find substantially larger prevalences of
intergenerational exposure to the criminal justice system than previously reported: 9%
of children born between 1999–2005 were intergenerationally exposed to prison, 18%
to a felony conviction, and 39% to any criminal charge; charge exposure rates reach as
high as 62% for Black children. We regress these newly quantified types of exposure
on measures of child well-being to gauge their importance and find that all types of
exposure (parent vs. non-parent, prison vs. charges, current vs. previous) are strongly
negatively correlated with development outcomes, suggesting substantially more U.S.
children are harmed by crime and criminal justice than previously thought.

Keywords: criminal justice, intergenerationality, economic mobility
JEL classification codes: K14, K42, J12, I32
U.S. Census Bureau, Washington, D.C. keith.ferguson.finlay@census.gov
Department of Economics, University of Michigan, mgms@umich.edu
‡
Department of Economics, University of Missouri streetb@missouri.edu.
§
Any views expressed are those of the authors and not necessarily those of the U.S. Census Bureau.
Results were approved for release by the Census Bureau’s Disclosure Review Board, authorization numbers CBDRB-FY22-ERD002-001, CBDRB-FY22-ERD002-003, and CBDRB-FY22-ERD002-009 (approved
11/3/2021, 02/15/2022, and 5/6/2022). We thank Anna Aizer, Carolina Arteaga, Hoyt Bleakley, Charlie
Brown, Sarah Burgard, Aaron Chalfin, Janet Currie, Jennifer Doleac, Bilge Erten, Paula Fomby, Katie
Genedak, Matthew Lindquist, Carla Medalia, Becky Pettit, Emily Owens, Steven Raphael, Ana Reynoso,
Kevin Schnepel, Jeffrey Smith, Mel Stephens, Cody Tuttle, and Bruce Western as well as conference/seminar
participants at the UM Population Studies Center, Southern Economic Association 2021 Conference, the
Virtual Economics of Crime seminar, American University, the UM School of Information, and the UNL
England-Clark Labor Conference for thoughtful comments. We thank the National Science Foundation, the
Bill & Melinda Gates Foundation, and Arnold Ventures for financial support.
†

1

Introduction

Both human capital investments and deprivation can have critical and dynamic impacts on
children throughout their lives (Currie and Almond 2011). Investments in the domains of
health (Black, Devereux, and Salvanes 2007; Campbell et al. 2014), education (Deming 2009;
Dynarski, Hyman, and Schanzenbach 2013), housing (Chetty, Hendren, and Katz 2016), and
financial well-being (Hoynes, Schanzenbach, and Almond 2016) made before children become
adults can influence a lifetime of outcomes, including educational attainment, employment,
earnings, and mortality. The implications of these findings for economic inequality, racial
disparities, and intergenerational mobility motivate a wide range of U.S. public policies
aiming to equalize opportunity regardless of one’s family background.
One area that has received growing attention is the influence of parental involvement in
the criminal justice system (e.g., National Research Council 2014; Wildeman 2009; Billings
2018; Norris, Pecenco, and Weaver 2021; Arteaga 2018). A half century of criminal justice
policy has expanded both the share and degree of contact that the U.S. population has
with the formal justice system, reflected in both the growing number of individuals with
criminal histories and dramatic expansions in imprisonment rates. Contractions in safety net
assistance and social support programs may also contribute to this pattern by increasing rates
of illicit activity in the population overall (Deshpande and Mueller-Smith 2022). Whether
justice involvement reflects household shocks to financial or emotional stability, or exposure
of children to undesirable living circumstances, the ramifications of this caseload growth
may potentially be felt for decades to come because of intergenerational spillovers within
households.
Precisely measuring intergenerational exposure to the justice system comes with several
challenging hurdles. Data limitations and a prior focus on incarceration have yielded almost exclusively intergenerational exposure estimates of parental incarceration in prison or
jail.1 However, many other forms of contact with the justice system exist, such as arrests,
charges, and convictions, which may also independently impact child development, whether
because of the justice system itself or the underlying criminal activity those events represent.
Likewise, existing estimates typically quantify contemporaneous intergenerational exposure
at a given point in time (e.g., Mumola 2000; Glaze and Maruschak 2008), which may be
insufficient if intergenerational exposure events produce long-term scarring effects.2 Finally,
1

For an example of prominent studies focusing on incarceration or jail, see Mumola (2000), Wildeman
(2009), Lee et al. (2015), Wildeman and L. H. Andersen (2015), Billings (2018), Norris, Pecenco, and Weaver
(2021), and Enns et al. (2019).
2
Studies that do seek to quantify the size of cumulative exposure, either apply strong assumptions to
aggregate data using a life tables methodology from demographic research (Wildeman 2009; Wildeman and
L. H. Andersen 2015) or use small longitudinal surveys like the newly fielded Family History of Incarceration

1

the literature predominately focuses on biological parents as the originating sources of intergenerational justice exposure, which overlooks a well established literature on changing
household structure in the U.S.,3 especially among racial and ethnic minority populations.4
In this paper, we take a new approach by leveraging billions of federal tax, household survey, and program participation records linked with Criminal Justice Administrative Records
System (CJARS; Finlay and Mueller-Smith 2020) data to quantify what share of recent
birth cohorts in the United States have ever experienced intergenerational exposure to multiple stages of the U.S. criminal justice system. We address three primary shortcomings in
prior estimates, accounting for: (1) multiple forms of exposure beyond just incarceration,
(2) cumulative exposure over childhood relative to point-in-time exposure given evidence of
long-term scarring from justice contact, and (3) sources of intergenerational exposure from
adults who are not biological parents (e.g., other adult caregivers or household members).
To accomplish this, we build national longitudinal relationship and residency crosswalks that
incorporate a range of novel data linkages: decennial census and American Community Survey household rosters, IRS Form 1040 tax returns, Social Security registration information,
and roster information from Housing and Urban Development, Indian Health Service, and
Medicare beneficiaries.
Starting with the most common measure used in the literature, we find that 0.8% of
children in 1999–2005 birth cohorts in our sample of states had a biological parent in prison
in any given year of childhood. The share of children exposed to the criminal justice system
in any given year increases when we expand our definition to other events, such as felony
convictions (0.9%), felony charges (1.2%), and any criminal charges (3.7%). The cumulative
share of children ever exposed to criminal justice events, which accounts for scarring effects
of exposure, dwarfs the point-in-time estimates with 3.6%, 9.2%, 11.4%, and 26% of children
exposed to incarcerations, felony convictions, felony charges, or any criminal charges of
biological parents during childhood. Even further, cumulative exposure estimates are 50%
to 140% larger once other potential caregivers are also considered: 8.8%, 18.3%, 21.4%, and
38.9%. Corresponding estimates for exposure of Black children through potential caregivers
Survey (Enns et al. 2019), which was conducted as part of the AmeriSpeak panel data collection effort
(N=4,041, 34% response rate). Smaller survey collections unfortunately lack the sample size and response
rates to precisely estimate the degree of intergenerational exposure in the population, much less demographic,
spatial, and temporal differences within the population.
3
For example, see Bumpass (1990), Cherlin and Furstenberg Jr (1992), Cherlin (2004), Andersson (2002),
Heuveline, Timberlake, and Furstenberg (2003), Curtin, Ventura, and Martinez (2014), Smock and Schwartz
(2020), Raley and Sweeney (2020), and Cavanagh and Fomby (2019).
4
See Cherlin and Furstenberg Jr (1992), Lichter et al. (1992), Bumpass and Lu (2000), Raley and Wildsmith (2004), Fomby and Cherlin (2007), McLanahan and Percheski (2008), Isen and Stevenson (2010),
Kreider and Ellis (2011), Raley, Sweeney, and Wondra (2015), and Parker, Sassler, and Tach (2021).

2

are 20%, 35%, 42%, and 62%.5 We observe a strong household income gradient with regard
to all types of exposure, although the benefit of household income at birth varies by racial
and ethnic background.
We investigate how these new measures correlate with several measures of child well-being
(e.g., household poverty, behind age-appropriate grade level, cognitive difficulty, teenage
parenthood, teenage criminal justice involvement, and death) after controlling for place of
birth, age, household income at birth, race and sex. We find that the estimated relationships
between exposure and child outcomes are often remarkably similar regardless of the type of
criminal justice exposure, the recency of the event, or whether a parent or other coresident
adult was the source of exposure. Heterogeneity analysis suggests differential sensitivity
to exposure by child’s gender and household income. Our results suggest that our new
broader measures of justice exposure are as important as the smaller, more narrowly defined
traditional measures from the literature, and may contribute meaningfully to propagating
economic inequality and racial inequities across generations in the U.S.
To our knowledge, we are the first to leverage U.S. administrative data to estimate: (1)
the prevalence of children’s exposure to a range of types of parental contact with the criminal
justice system, (2) cumulative exposure estimates over the duration of childhood, and (3)
the magnitude of exposure originating from adult household members who are not biological
parents.6 These expanded measures fundamentally redefine the scope of the spillover population in this literature, shifting the narrative from less that one-in-forty to almost one-in-two
minors in the U.S. Moreover, our newly documented relationships between intergenerational
exposure and child development highlight the potential unintended consequences of recent
crime and social policy onto the most vulnerable members of society, in ways that undermine
children realizing their full potential.

2

Why intergenerational exposure matters for children

Intergenerational exposure to the justice system simultaneously reflects two broad conceptual
influences on child development and well-being. First, adult involvement in the justice system
could be an indication that the household is actively in a moment of crisis (financial, health,
physical safety, or otherwise) that puts the child at risk. This reflects more the circumstances
that led to justice involvement in the first place, rather than the direct impact of the justice
system itself. But, in addition, a second channel arises since intergenerational exposure also
5

Estimates on the intensive margin of exposure provide further evidence of racial disparities. These are
discussed in Section 6.
6
For estimates of exposure using register data in other countries, see Wildeman and L. H. Andersen
(2015) using Danish registries and Hjalmarsson and Lindquist (2012, 2013) using Swedish registries.

3

represents the potential initiation of justice-based interventions to the adult that could have
ramifications for the entire household. Both of these channels are discussed in detail below.
Households under strain or in crisis. Being charged, convicted, or placed in correctional
supervision may indicate an unsafe or harmful environment for children in the household.
Criminal charges could reflect allegations of direct harm to the child, including: domestic
violence, abuse and neglect, sexual assault of a minor, or child pornography. Doyle and
Aizer (2018) provide a recent review of the literature, which finds that abuse, neglect, and
maltreatment is linked with future violence and criminal activity (Widom 1989; Currie and
Tekin 2012), impeded brain development (Medicine and Council 2014), and worsened education and earnings trajectories (Currie and Widom 2010). Together, these impacts are
estimated to generate substantial social costs (Fang et al. 2012; Peterson, Florence, and
Klevens 2018).
In addition, charges may reflect adult conduct, apart from the child, that still may put
the child at risk. These might include indications of substance abuse (possession of illicit
drugs, abuse of prescription medication, driving while intoxicated), acute financial hardship (burglary, fraud, prostitution, robbery, or theft), or emotional and mental instability
(disorderly conduct, violent offenses, intimate partner violence). Growing up with a parent
who struggles with substance abuse has associated with child behavioral problems (Chatterji
and Markowitz 2001), incidents of neglect and foster care (Cunningham and Finlay 2013),
and poorer labor market outcomes (Balsa 2008). Child poverty has been tied to impacts
on physical and mental health, human capital formation, youth delinquency, and economic
self-sufficiency (see National Academies of Sciences, Engineering, and Medicine (2019) for a
recent review of the literature). Intimate partner violence and household conflict have been
shown to worsen birth outcomes (Aizer 2010; Currie, Mueller-Smith, and Rossin-Slater 2022),
increase disruptive behavior (Carrell and Hoekstra 2010; Herrenkohl et al. 2008; Levendosky
et al. 2003), and even erode telomere length (Shalev et al. 2013).
Together, these scenarios capture a variety of serious and consequential experiences that
children may confront. While not necessarily a product of criminal justice policy (and in fact,
criminal courts may seek to minimize the potential harms of these situations), the justice
system provides a useful way to measure their prevalence in the population and to gauge the
effectiveness of broader safety net assistance programs to protect and provide for children in
the U.S.
Stress from criminal proceedings. The initiation of criminal charges might trigger
numerous factors that add and compound stress within the household, beyond what might
have existed prior to charges being filed. These include anxiety about the resolution of

4

the case and what potential sanctions might be applied, financial burdens associated with
fines and fees stemming from court charges and correctional supervision (Harris, Evans, and
Beckett 2010; Martin et al. 2018; Finlay et al. 2022), or internal strife if charges revealed
behavior that had otherwise been concealed from other household members (e.g., illicit drug
use).
Research has found that ambient stress levels can negatively impact children. From fetal
development (Aizer, Stroud, and Buka 2016; Persson and Rossin-Slater 2018) to elementary
(Sharkey et al. 2012) and high school (Sharkey 2010; Ang 2020) aged children, stress has been
shown to impede physical and cognitive development and worsen educational performance
(see Almond, Currie, and Duque (2018) for a review).
Ongoing financial security and future criminal activity. Research has also documented numerous mechanisms through which the justice system may interrupt labor market
activity, jeopardize financial security, and increase long-term criminality. Mechanisms include: pre-trial detention (Dobbie, Goldin, and Yang 2018), criminal convictions (Pager
2003; Agan and Starr 2017; Mueller-Smith and Schnepel 2021), and incarceration (MuellerSmith 2015). In fact, research indicates that criminal justice involvement is self-perpetuating
because it reduces one’s ability to engage in the legal labor market, which creates further incentives to continue or increase illicit activity (Mueller-Smith and Schnepel 2021; Deshpande
and Mueller-Smith 2022).
Financial resources have long been recognized as critical factors for child development.
Birth weight (Hoynes, Page, and Stevens 2011), academic performance (Dahl and Lochner
2012; T. N. Bond et al. 2021), mental health (Milligan and Stabile 2011), physical health
(Aizer, Stroud, and Buka 2016), and long-term self-sufficiency (Hoynes, Schanzenbach, and
Almond 2016) have all been shown to respond to changes in available household resources
during childhood. Consequently, justice contact may have long-term ramifications for the
household even after the initial circumstances that led to the criminal offense are resolved
due to the lasting scarring effects on work and recidivism.
Adult or child removal from the household. Finally, the allegations associated with
a criminal charge may be so severe that the composition of the household is fundamentally
altered. This may include the justice-involved individual exiting the household because
of incarceration, but could also reflect the dissolution of a romantic relationship due to
the enhanced stress in the household or the inability for the justice-involved individual to
financially provide for the family. Adult exit from the household could remove a negative
influence, jeopardize continuity of care as well as financial and emotional support, or both.
Research on the causal effect of parental incarceration on children has found mixed impacts

5

to date (Norris, Pecenco, and Weaver 2021; Arteaga 2018; Wildeman and S. H. Andersen
2017).
Likewise, it may be deemed that the household is no longer a safe environment for the
child. In this case, the child may be removed by child protective services and placed in
the foster care system, which research suggests can have important consequences for child
well-being (Doyle 2007; Gross and Baron 2022).

3

Previous research

3.1

Prior attempts to measure intergenerational exposure to criminal justice

The leading estimates of intergenerational exposure to the criminal justice system come predominantly from a select group of surveys and focus almost exclusively on incarceration.
Mumola (2000) and Glaze and Maruschak (2008) estimate that 2.1% of children in the U.S.
in 1999 and 2.3% in 2008 have a parent in prison using the 1997 and 2007 Survey of Inmates in State and Federal Correctional Facilities (SISFCF).7 However, these estimates are
built from surveys asking about the share of inmates with minor children, which rely on
assumptions about accurate sampling without non-response and attrition bias, non-overlap
of any minor child among inmates, and no influence of social desirability bias in inmate
responses. They also provide a point-in-time measure, which may be of limited value when
trying to assess what share of children are ever exposed to parental incarceration. Using
the same survey data, paired with aggregate caseload statistics and life tables methodology,
which relies on strong additional assumptions, Wildeman (2009) estimates a higher cumulative exposure to incarceration at age 14 for White (4%) and Black (25%) children born in
1990. Enns et al. (2019) also estimate cumulative exposure using the newly fielded Family
History of Incarceration Survey (FamHIS) a part of the AmeriSpeak panel, which directly
asks individuals if they have ever had a parent incarcerated in prison or jail. They find that
roughly 35% of adults aged 18–29 years in 2018 and 10% of Americans in their fifties report
ever having a parent incarcerated (prison or jail).8
Surveys can offer advantages in the form of detailed information on social networks often
omitted in administrative data, such as the nature of a caregiving relation (e.g., biological
parent, stepparent, aunt/uncle, and grandparent); however, they can suffer from reporting
biases, sample sizes, and often have poor coverage of the criminal justice population due
7

These statistics are inferred from reports that 55% of individuals in 1999 and 51.9% of individuals in
2007 in state prisons reported having a minor child and 63% and 62.9% of individuals in federal prisons in
1999 and 2007.
8
There is a similar body of evidence focusing on siblings and other members of an individual’s social
circle that have been to prison (Lee et al. 2015; Enns et al. 2019).

6

to low residential stability (Roman and Travis 2004), low educational attainment (Harlow
2003), and membership in racial and ethnic minority groups (Carson and Anderson 2016).
An emerging literature has sought to study intergenerational exposure to the criminal
justice system using administrative data. Benefits of this approach include population-level
measurement, without concerns about social desirability or attrition biases. Many of these
papers, however, are based in Sweden (Hjalmarsson and Lindquist 2012, 2013; Eriksson et
al. 2016) or Denmark (Wildeman and L. H. Andersen 2015), where integrated administrative
data systems to support research and statistical reporting are among the most advanced in
the world. At the same time, the informativeness of these findings for U.S. policy is limited,
given the vast differences in the operations of the respective criminal justice systems (Barclay
et al. 2003).
Two U.S.-focused studies, Norris, Pecenco, and Weaver (2021) and Billings (2018), examine the effects of parental criminal justice events on child outcomes using administrative
records in Ohio and North Carolina, respectively. Norris, Pecenco, and Weaver (2021) report
that 38.2% of defendants in Ohio are linked to children through birth certificates. Birth certificates capture biological parent links, but often have incomplete information, particularly
for the father. For example, maternal and paternal information are observed for 99.99%
and 88% of Ohio birth certificates (1972, 1984–2017) and both parents’ information is observed on 65% of Michigan birth certificates (1993–2006) (Norris, Pecenco, and Weaver 2021;
Almond and Rossin-Slater 2013). Billings (2018) links school-aged children to parents identified in educational records in North Carolina using address information on school, arrest,
and incarceration records and reports that 9.7% of unique children are exposed to a parental
arrest during the 1998/1999 to 2010/2011 school years, with a contemporaneous exposure
rate of 2% and 1% for parental arrests and incarcerations, respectively. As Sykes and Pettit (2014) point out, there is an added complexity to measuring parental contact with the
criminal justice system, using either survey responses or birth records, due to evolving household structures and multiple partnerships, particularly among those directly and indirectly
interacting with the criminal justice system.
3.2

Changing household structures and implications for measuring caregivers

Household formation and structure in the U.S. has undergone significant transformations
over the past half century,9 with important heterogeneity by race.10 First, there has been a
9
See Bumpass (1990), Cherlin and Furstenberg Jr (1992), Cherlin (2004), Andersson (2002), Heuveline,
Timberlake, and Furstenberg (2003), Curtin, Ventura, and Martinez (2014), Smock and Schwartz (2020),
Raley and Sweeney (2020), and Cavanagh and Fomby (2019).
10
Cherlin and Furstenberg Jr (1992), Lichter et al. (1992), Bumpass and Lu (2000), Raley and Wildsmith
(2004), Fomby and Cherlin (2007), McLanahan and Percheski (2008), Isen and Stevenson (2010), Kreider

7

notable decline in marriage rates. Between 1980 and 2012, the percentage of women aged
40–44 who had ever married decreased by 7.9 percentage points for White, non-Hispanic
women, 26.3 percentage points for Black, non-Hispanic women, and 10.6 percentage points
for Hispanic women (Raley, Sweeney, and Wondra 2015; Kreider and Ellis 2011). At the
same time, there has been a dramatic increase in the share of births for unmarried women,
increasing from less than 5% to roughly 40% from 1940 to 2010 (Curtin, Ventura, and Martinez 2014). The composition of unmarried births also appears to be changing with the
share of cohabiting partners among unmarried births increasing by 17 percentage points
over the first decade of the 2000s; this is echoed by the increasing share of unmarried births
with paternal information on birth records in Michigan between 1993 and 2006 (Almond
and Rossin-Slater 2013), reflecting shifts in the composition of unmarried births and societal norms and policies pertaining to reporting paternity information (Rossin-Slater 2017;
Massenkoff and Rose 2020).
There have also been significant increases in the number of divorces, partner separations,
and second families.11 For example, the ten-year marriage survival rate decreased by 8
percentage points for women married between 1990 and 1994 relative to two decades earlier
(Kreider and Ellis 2011). And married women are 10 percentage points more likely to be
on their second or subsequent marriage in 2013 relative to 1960 (Livingston 2014); however,
other forms of re-partnering, such as cohabitation, are more common in recent decades and
among women from racial and ethnic minorities. In fact, Kreider and Ellis (2011) estimate
that over 7% of children live with a cohabiting or legal stepparent.
Finally, multigenerational households in the U.S., with grandparents functioning as caregivers, are increasingly common.12 In 2012, 3.8% of adults 30 and over in the U.S. were
grandparents living with minor grandchildren and 38.8% of those grandparents have primary responsibility for the coresident grandchildren, with substantial heterogeneity by race
and ethnicity (Ellis and Simmons 2014).13
These changes have important implications for which caregivers are important to track
and how they can be observed. Declining marriage rates and increasing non-marital romantic
cohabitation heightens the need to consider relationships that are less legally formal, particularly in households with children from racial and ethnic minorities (Raley and Wildsmith
and Ellis (2011), Raley, Sweeney, and Wondra (2015), and Parker, Sassler, and Tach (2021).
11
See Sweeney (2010), Kreider and Ellis (2011), Isen and Stevenson (2010), Raley and Sweeney (2020),
Schoen and Standish (2001), Bramlett and Mosher (2001), and Teachman (2008).
12
See Landry-Meyer and Newman (2004), Hayslip and Kaminski (2005), Choi, Sprang, and Eslinger (2016),
and Williams (2011).
13
Corresponding estimates by race and ethnicity for the share of adults defined as grandparents and the
percentage with primary caregiver responsibility are: White (3.1%, 39.9%), Black (5.9%, 47.6%), American
Indian/Alaska Native (7.5%, 54.0%), Asian (5.9%, 15.3%), and Hispanic (7.2%, 30.9%).

8

2004). Increasing divorce rates and declining relationship stability motivate greater study of
intergenerational relationships that may no longer be coresident.
Increases in births to unmarried adults and decreases in the rate of paternity reporting
has lowered the share of men observed on birth record data. Similar dynamics yield declining rates of paternal coverage in IRS tax filings, since only legally married individuals can
jointly file and claim children. Snapshots of family composition at a given point in time, as
commonly observed in household survey collections like the decennial census, fail to capture
evolving household structure as romantic partners and other adult caregivers leave and join
the households of children over time.

4

Leveraging survey and administrative data to measure cohabitation, relationships, and justice contact

Currently, no single dataset in the U.S. captures all the potential intergenerational influences
a minor interacts with in their household over the course of their childhood. This project
brings together a number of restricted access administrative and survey datasets available
through the Census Bureau’s Data Linkage Infrastructure to address this problem. While
each individual dataset has its own limitations,14 together they provide an opportunity to
measure the population in unprecedented ways. Using their combined strength, we produce
new population-level residence and relationship crosswalks that identify where each individual in the U.S. lives in a given year, and with whom they share familial and coresidency
relationships. With intergenerational linkages identified, we leverage CJARS to track several
forms of exposure to the justice system. An overview of the data, linkage processes, sample
restrictions, and measurement concepts is provided below; detailed information on the construction and performance of our residency and relationship crosswalks, including successful
replications of recent fertility estimates from the National Center for Health Statistics and
National Vitality Statistics System, can be found in Appendix B.15
The residential crosswalk seeks to establish the best known address for every individual
in the U.S. on an annual basis. It incorporates both administrative sources like IRS tax
filings and household survey data like the Decennial Censuses.16 When multiple addresses
14
For example, information about dependents from tax returns is only available for individuals who file
a tax return. Similarly, public assistance caseload data is only available for low-income individuals who
participate in these programs. Decennial census data is comprehensive but only available every 10 years.
15
Person-level data are linked using the Census Bureau’s Protected Identification Key (PIK), which are
assigned to records using the Person Identification Validation System (PVS) (Wagner and Lane 2014). While
there is some non-random selection in PIK assignment (B. Bond et al. 2014), this project minimizes possible
linkage bias by combining data from many sources.
16
We harvest residential addresses from the following data source: decennial censuses (2000, 2010), Amer-

9

are identified for an individual in a given calendar year, priority is given first to Census
Bureau surveys, then to tax records, and then to public program data.
The residence crosswalk functions as the “backbone” of the relationship crosswalk. First,
for each year, all coresident individual pairs are identified.Where possible, these cohabiting
relationships are further delineated into specific relationship types based on information from
the 2000 and 2010 decennial censuses, American Community Survey, dependents listed on
IRS Form 1040 filings, HUD program data, and the Census Household Composition Key
(CHCK) file that is based on Social Security Administration SS-5 applications for Social
Security Numbers. Because of data limitations, we focus our analysis on the cohort of
children born between 1999 and 2005 to measure exposure to parental and other potential
caregiver criminal justice involvement (see Figure A1 for sample composition descriptives).17
Figure 1 summarizes the performance of the crosswalks for all children, and by racial subgroup. Overall, we are able to successfully link 97% of our focal children to female potential
adult caregivers and 95% to male potential adult caregivers; furthermore, we can identify
female biological parents for 90% of these children and male biological parents for 76%. We
also find that 4.7%, 22%, 29%, and 46% of children are observed with a step/adopted/foster
parent, extended family (grandparent/aunt/uncle), unclassified caregivers, and unclassified
cohabiting adults, respectively. In terms of the number of linked caregivers, we observe two
or more female (male) potential caregivers for 48.3% (46.1%) of children (see Figure A3),
which could be due to a variety of living circumstances: (1) parents with multiple romantic partners (due to divorce or separation) while raising their children, (2) households with
same-sex romantic partners, (3) multigenerational households, or (4) doubled-up households
where multiple families share the same accommodations. These estimates reflect how common it is currently for children in the U.S. to grow up with multiple adult influences in their
households beyond the traditional notion of the nuclear family with one mother and one
father. Further discussion of these results, particularly by racial subgroup, can be found in
Appendix B.
Adult experiences in the criminal justice system are measured using CJARS, which covers 23 states with over 175 million unique events spanning multiple procedural stages of
ican Community Survey (2001–2018), IRS Form 1040 tax filings (1969, 1974, 1979, 1984, 1989, 1994, 1995,
1998–2018 tax years), IRS Form 1040 electronic tax filings (2005, 2008–2012), Department of Housing and
Urban Development (HUD) program data (Longitudinal PIC/TRACS: 1995–2016, 2018; PIC: 2000–2014;
TRACS: 2000–2014), and county/state level information from Medicare (2000–2017 EBD) and Medicaid
(2000–2014 MSIS) enrollment databases, Indian Health Service (IHS) from 1999–2017, and the Master Address File-Auxiliary Reference File (MAF-ARF) (2000–2018).
17
The Decennial Census Digitization and Linkage project is an initiative to link microdata from the 1960–
1990 decennial censuses (Genadek and Alexander 2019). When these data become available at the Census
Bureau, we will be able to investigate how intergenerational exposure rates have changed over time.

10

the justice system (e.g., arrest, charge, conviction, incarceration, and/or parole) with some
jurisdictional data spanning back to the late 1970s. Because temporal and procedural coverage varies by jurisdiction, we restrict our analysis to children born in geographies covered
by CJARS in their year of birth through age 18. Sufficient coverage is available to study
charges, felony charges, and felony convictions in Arizona, Florida, Maryland, Michigan,
New Jersey, North Carolina, North Dakota, Oregon, Texas, and Wisconsin, which collectively cover 29.5% of the U.S. population in 2000, within the period when our focal children
were born. Likewise, sufficient statewide coverage is available to study prison incarceration
in Arizona, Florida, Michigan, Nebraska, North Carolina, Pennsylvania, Texas, Washington,
and Wisconsin, roughly 30.3% of the population.18,19 Even though CJARS does not have
complete national coverage, Appendix Figures B3 and B4 document how CJARS states do
not meaningfully differ from non-CJARS states in terms of crime and incarceration rates as
well as demographic and socioeconomic characteristics. Thus, we believe the estimates in
this paper generalize nationally.

5

Novel estimates of intergenerational exposure to criminal charges,
convictions, and incarceration

In this section, we report our overall findings on the extent of intergenerational exposure
of children to charges, convictions, and incarceration. To align with previous literature,
we begin by focusing on contemporaneous exposure rates stemming from justice-system
contact among biological parents. We then expand these definitions to account for cumulative
exposure over the duration of childhood. Finally, we incorporate other potential caregivers
in addition to biological parents as sources of potential intergenerational exposure to arrive
at our most comprehensive measures of the share of children in the U.S. who experience the
justice system secondhand through the adults in their households. Section 6 delves further
into these estimates, examining differences in exposure by child’s race, household income,
adult’s sex, and coresidency status at the time of exposure.
Contemporaneous exposure from biological parents. We first start with the most
common estimate of intergenerational exposure to the criminal justice system: child exposure
to a biological parent in prison at a point in time. Specifically, we measure the probability
that a minor child has a biological parent in prison in a given year using the following
18

For a full description of CJARS including a code book and location specific coverage, see Finlay and
Mueller-Smith (2020).
19
Jail records are currently not included systematically in CJARS.

11

equation:
Contemporaneous exposure =

q2005

by=1999

qNby qTby

CJ Exposureby,i,t
y=1999 Nby ◊ Tby

i=1

q2005

t=0

where by denotes the year of birth for a child, i references each child in the sample, Nby
reflects the total number of children born in birth year by, t refers to the age of a child, and
Ty denotes the number of years the child is in the sample – either until age 18 or the place
of birth is no longer covered by CJARS.20 CJ Exposureby,i,t will equal one if child i born in
year by had a biological parent (male or female) in prison when they were age t, and zero
otherwise.
In Figure 2A, we document that 0.8% of children in our sample have a biological parent
in prison in a given year during their childhood,21 with 0.3% having a parent enter prison
in a given year. A significantly larger share of children have biological parents face criminal
court proceedings, including felony convictions (0.9%), felony charges (1.2%), or any criminal
charges (3.7%).
Cumulative exposure from biological parents. If children experience long-term scarring from parental involvement in the justice system, it is insufficient to know what share
of children have intergenerational exposure in a given year. Instead, we need to identify
how many ever experience exposure over the course of their childhoods. To answer this,
we expand our previous measure to quantify the cumulative exposure to parental criminal
justice events. Formally, we calculate the following:
Cumulative exposure· =

q2005

by=1999

qNby

i=1 1

Ë1q

·
t=0 CJ Exposureby,i,t
q2005
by=1999 Nby

2

È

>0

where, the numerator sums over the total number of children with a given type of exposure
by age · and the denominator divides by the total number of children born in the birth
cohorts under consideration.
Figure 2B (solid lines) presents the share children ever exposed by age · , where · œ
[0, 18], to a biological parent in prison, convicted of a felony, charged with a felony, or
charged with any criminal offense. We find that 3.6% of children experience a biological
20

For example, children born in 2005 will only be in the sample 13 years by definition.
This number is lower than prior BJS estimates of 2.0% and 2.3% (Mumola 2000; Glaze and Maruschak
2008) for several reasons. First, the Survey of Inmates in Federal and State Facilities includes individuals in
state and federal prisons, where a larger share of federal prisoners report being a parent (≥ 60%). Second, the
survey asks about any minor children: biologic, step, or adopted. If we instead include potential caregivers,
we estimate 3.1% of children are exposed to prison, which is greater than the literature estimates as would
be expected given the more expansive definition of a potential caregiver relative to the survey’s question.
21

12

parent in prison by age 18, a 350% increase over the contemporaneous measure. Given that
prison spells typically occur over multiple years, this increase observed from contemporaneous
to cumulative exposure is large but relatively small in comparison with the other justice
exposure measures we consider. By age 18, 9.2% of children were exposed to a biological
parent’s felony conviction (900% increase), 11.4% of children were exposed to a biological
parent’s felony charge (854% increase), and 26.0% of children were exposed to a biological
parent’s (misdemeanor or felony) criminal charge (613% increase).22
Cumulative exposure over the age profile throughout childhood (ages 0 to 18 years old)
grows steadily. Because our focus is on extensive-margin exposure, the majority of first-time
exposure occurs by ages 5 to 7 years old. From that point forward, cumulative exposure
grows at a slower but roughly linear rate.
Cumulative exposure from all potential caregivers. Finally, we expand our measure of
exposure to all observed potential caregivers: biological parents, stepparents, adoptive parents, foster parents, unclassified caregivers, grandparents, aunts/uncles, non-familial adults
(cohabiting 2+ years) and unclassified adults (cohabiting 2+ years). To be conservative,
we do not include any criminal justice involvement from other potential caregivers prior to
cohabitation in our exposure measures. For example, if a stepparent has a felony conviction
when the child is 3, but does not coreside with the child until the age of 6, then the child is
not considered exposed to the event.23
In Figure 2C, we report that 8.8% of children are exposed to a potential caregiver in
prison by age 18, 18.3%, to a felony conviction, 21.4%, to a felony charge, and 38.9% to any
criminal charge. These estimates of intergenerational exposure to the criminal justice system
that incorporate other adult influences in the household are much larger than those that
restrict to just biological parents, with a 140% increase in exposure to prison, 99% increase
in felony convictions, 88% increase in felony charges, and 50% increase any criminal charges
(misdemeanor or felony). Relative to contemporaneous exposure from biological parents,
which has occupied most of the literature’s attention, these broadly defined cumulative
intergenerational exposure measures are 958%, 1881%, 1693%, and 964% higher for exposure
22

Enns et al. (2019) estimate that 20% of respondents report that a parent was in jail or prison for at
least one night; notably, these estimates have a steep age gradient with 34% of 18–29 year olds having a
parent incarcerated compared to roughly 10% of respondents in their fifties (see Figures 2 and 4 of their
paper). For children born in 1990, Wildeman (2009) and Wildeman and L. H. Andersen (2015) use life tables
and the Survey of Inmates of State and Federal Correctional Facilities to estimate exposure by age 14 for
children born in 1990; 25%-28% of Black children and 3.6%-4.2% of White children are exposed to parental
incarceration with 7.96% (0.58%) having paternal (maternal) exposure (see Table 3 of their paper).
23
This assumption is quite strong. While children may not have direct exposure to the event, they may
have indirect exposure to the justice system through ongoing community supervision requirements or direct
exposure to secondary effects of justice contact, such as diminished wages from criminal records.

13

to prison, felony convictions, felony charges, and any criminal charges, respectively.
Because we do not incorporate criminal justice contact among other potential caregivers
before they initiate cohabitation in the household, the growth rate in cumulative exposure
over ages 0 to 18 years old is more consistent year-over-year compared to exposure rates
focused just on biological parents (see Figure 2B dashed line). This reflects, in part, the fact
that other potential caregivers must first join the child’s household and then initiate contact
with the justice system before a child will be counted as “exposed.”
Intensive margin of exposure. Prior results have focused on the extensive margin of
exposure, referring to whether one or more events have happened in a child’s household
while they are minors. While this represents important new evidence, it only partially
characterizes the experience of children whose households are having repeated contact with
the criminal justice system over their childhood, which we refer to as the intensive margin
of exposure.
Figure 2D documents variation in the number of distinct events that children are exposed
to by activity type among those who are exposed at least once. For example, multiple felony
charges filed on the same date are considered one event, but felony charges filed within the
same year on different dates would be two distinct events. The marginal median exposed
child (bars denoted with stripes) lives in a household that faces 3 criminal charges, 2 felony
charges and convictions, 1 prison spell, and 4 years of adult incarceration.24 At the high
end, 21.2%, 7.9%, 5.1%, and 18.2% of exposed children live in a household with 10 or more
charges, felony charges, felony convictions, and years of adult incarceration, respectively.
This is important to keep in mind when interpreting Figure 2B, where we observe slowing
increases in extensive margin exposure starting around age 5. While a child only experiences
first-time exposure once, their household’s contact with the justice system likely continues
to deepen over the course of their childhoods.

6

Socioeconomic variation in exposure and discussion

In this section, we dig further into the findings of Section 5, disaggregating cumulative
exposure rates by age 18 from any potential adult caregiver by the child’s race and ethnicity,
family income at birth, and the potential caregiver’s sex. In addition, we present evidence
on the share of exposure that occurs after the adult in question has coresided with the child.
Exposure by child’s race and ethnicity. Stark divides emerge when disaggregating
exposure rates by the race and ethnicity of children. As seen in Figure 3A, 62% of Black,
24

The median for each type of exposure is conditional on being exposed to the specific event type.

14

non-Hispanic (referred to as Black for the remainder of Sections 6 and 7) children grow up in
a household where one or more potential caregivers are charged with either a misdemeanor or
felony criminal offense.25 American Indian/Alaska Native children have a similarly high rate
at 60%, and 45% of Hispanic children have a potential caregiver charged with a criminal
offense. White, non-Hispanic (referred to as White for the remainder of Sections 6 and
7) and Asian children have high but relatively lower rates of intergenerational exposure to
criminal charges at 32% and 17% respectively. Exposure rates decline for more serious forms
of criminal justice contact, but remain alarmingly high: 11–20% of Black, Hispanic, and
American Indian/Alaska Native children have a parent or other potential caregiver in prison
during their childhood; corresponding estimates for White and Asian children are 6% and
2% respectively.26,27
Racial and ethnic disparities increase with the seriousness of exposure type. The ratio of
exposure among children from racial and ethnic minorities28 relative to White children grows
by level of exposure: any criminal charges (1.40–1.90), felony charges (1.57–2.59), felony
convictions (1.63–2.53), and prison (1.86–3.23). Such disparities raise important questions
about the potential role the U.S. criminal justice system plays in propagating economic
inequality, racial inequities, and limiting social mobility across generations.
Important differences by race and ethnicity are also clear when examining the intensive
margin of exposure (see Figure A7). For example, the marginal median exposed Black child
grows up in a household with 6 criminal charges, 3 felony convictions, and 5 years of adult
incarceration. In fact, among Black children who are exposed to at least one criminal charge
in their household, 34% experience 10 or more criminal charges during their childhoods. In
contrast, the median White child grows up in a household with 3 criminal charges, 2 felony
convictions, and 4 years of adult incarceration; only 16% experience 10 or more charges
during their childhood conditional on being exposed.
Exposure by household income rank. Figure 3B documents changing exposure risk
25

Our data on misdemeanor and felony criminal charges do not include offenses classified as civil infractions
like many minor traffic offenses. Instead, these represent allegations that rise to the level of a court charge.
26
For comparison, Wildeman (2009) uses the Survey of Inmates of State and Federal Correctional Facilities,
the National Corrections Reporting Program and life-table methodology to estimate that 25–28% (13.8–15.2)
of Black children and 3.6–4.2% (2.2–2.4) of White children born in 1990 (1978) are exposed to a parental
incarceration by the age of 14 (see Table 3 of their paper).
27
See Appendix Figure A4 for additional results that compare differences in cumulative exposure from
just biological parents by child’s race. For example, the share of Black children exposed to an incarceration,
felony conviction, felony charge and any criminal charge increases by 55%, 100%, 118% and 180% respectively
when incorporating other potential caregivers into their potential sources of intergenerational exposure in
addition to biological parents, a higher increase compared to White children which reflects the important
differences in household structure by race as noted earlier in Figure 1.
28
In this context, we define racial and ethnic minorities to include Black, Hispanic, and American Indian/Alaska Native children.

15

over the household income distribution, as measured at birth and in the following 4 years.29
We observe a strong income gradient with regard to indirect criminal justice exposure by a
potential caregiver, which is consistent with prior work suggesting parental criminal justice
contact inhibits social mobility along a range of outcomes, including the child’s own likelihood
of adult incarceration (Chetty et al. 2018). Children born in households at the 10th percentile
of income experience exposure rates roughly 60–190% higher than children born at the
50th percentile of parental income, and 440–2950% higher than children born at the 90th
percentile of parental income. While any charges remains non-zero at slightly below 10
percentage points at the very top of the income distribution, no more than 0.3–3.2% of
children at or above the 90th income percentile experience felony charges, felony convictions,
or incarceration.
Figure A8 separates out the relationship between household income and criminal justice
exposure, by the child’s race and ethnicity.30 Across all types of exposure that we consider,
there remain stark exposure gaps by race conditional on income rank. Black and American Indian/Alaska Native children remain consistently 10–20 percentage points higher than
White children, conditional on income. The experience of Asian children is quite remarkable, where we observe consistently low rates of contact across the entire income distribution.
Finally, the relationship to household income varies by race. Hispanic children begin with
lower exposure rates than White children; however, this relationship is reversed by the 40th
percentile of household income.
Exposure by sex of potential caregiver. Figure 3C depicts the cumulative exposure
rates by age 18 by the sex of the potential caregiver in the household. The vast majority
(over four-fifths) of children with intergenerational exposure (at all levels of severity) observe
a male potential caregiver having contact with the justice system. Many children are also
exposed by female potential caregivers; in fact, across all types of measured exposure, 13–
18% of exposed children are exposed exclusively by female adults in their household. But,
with increasing severity of contact, the share exposed by both male and female potential
caregivers declines and the share exposed by exclusively male potential caregivers increases.
For instance, 80% of prison exposure exclusively comes from male potential caregivers, while
29

To construct this exercise, we linked children to tax filings for which they are claimed as a dependent
during the first 5 years of their lives. We average over total income and rank households within child birth
years. Exposure is then calculated within each individual percentile. Income is imputed to zero in years the
child is not claimed. Children never claimed or claimed on a tax filing with negative income in any year are
not included.
30
Note that White refers to White, non-Hispanic and Black refers to Black, non-Hispanic throughout
Sections 6 and 7.

16

only 53% of criminal charge exposure exclusively comes from male potential caregivers.31,32,33
Offense types. Adult criminal charges provide a window into the potential living circumstances of the most vulnerable children. Criminal charges include a range of risky and
dangerous behaviors, like violence and substance abuse, and may indicate material need
(e.g., theft, fraud), both of which are not conducive to child development.
In Figure 4A, we disaggregate the cumulative exposure rates by age 18 by the nature of the
criminal charge, including: violent, property, drug, driving under the influence (DUI), other
criminal traffic, and public order.34 Property offenses are the most commonly experienced
among children (29%), yet an astonishing 17% of children grow up in a household where an
adult faces violent criminal charges. In addition, 16% have adults in their household face
illicit drug charges at some point during their childhood.
Racial disparities previously discussed carry over to exposure by offense type as well
(Figure 4B). Roughly one-in-three Black and American Indian/Alaska Native children have
an adult in their household face violent criminal charges, while the corresponding estimate
for White children is only one-in-eight. Relative to the distribution of offense types for
exposed White children, exposed Black children are more likely to grow up in households
with drug charges and less likely to witness DUI offenses. Exposed Hispanic children are less
likely to experience other criminal traffic offenses. Finally, exposed American Indian/Alaska
Native children are less likely to have adults charged with property offenses but more likely
for public order offenses to have occurred.
Exposure to all offense types declines across the income distribution (Figure 4C). The
largest decline is observed for property offenses, which aligns with the fact that many offenses
categorized as property are financially motivated. These are followed closely by drug, violent,
and public order offenses. The offense type with the smallest decline across the income
distribution is DUI, making it relatively much more common among affluent households
with involvement in the justice system.
Incidence of exposure during recent coresidency with potential caregiver. A
concern that can be raised regarding the estimates presented so far relates to whether the
31

In Appendix Figure A4, we see relatively less sex differences in exposure for biological parents, although
the same qualitative pattern remains. This is likely due to differences in our ability to link children to their
biological parents by sex of the parent, which are not as stark for more inclusive definitions of adults who
contribute to intergenerational spillovers.
32
For comparison, Wildeman and L. H. Andersen (2015) uses the Survey of Inmates of State and Federal
Correctional Facilities and life-tables and estimates that 7.96% and 0.58% of children born in 1990 are
exposed to a paternal or maternal imprisonment by the age of 14 (see Table 3 of their paper).
33
Additional results by child race and potential caregiver sex in Appendix Figure A5.
34
See Appendix Table B4 for detailed information on the most commonly occurring offenses within each
of these broader categories.

17

observed caregiver is still in the child’s life in a meaningful way at the time of the criminal
justice exposure (Norris, Pecenco, and Weaver 2021). If it has been years since the adult
had any real connection with the child, should this be counted as a valid intergenerational
exposure?
To address this question, we document the cumulative exposure based on whether the
event happened when the adult in question was recently living in or out of the same home
as their child. Specifically, we measure coresidence between the potential caregiver and child
and consider the event to occur while the adult is “in the home” if they lived together in the
year of the event or either of the two years prior. Residence is observed quite well for the
population in Decennial Census years or if individuals file taxes, respond to the ACS, or are
enrolled in a public program35 in the relevant years.
We find that the share of children exposed to a potential caregiver in prison, receiving a felony conviction, felony charge, or any criminal charge decreases by 46%, 35%, 33%,
and 21%, respectively, when restricting to exclusively “recent coresidency” exposures (Figure 5A). The majority of cumulative exposure, however, is by adults recently coresiding in
the home. And in fact, the largest decrease in exposure based on the recent coresidency
restriction is for incarceration, which often occurs after several prior criminal justice interactions have occurred (e.g., prior arrests, convictions, and pre-trial detention) that might
cleave the individual from their family unit. Figure 5B shows the corresponding figure restricting to exposure originating from exclusively biological parents; here we see qualitatively
the same story, with the majority of exposed children experiencing at least one event with
recent coresidency across all exposure types. These results are remarkably consistent with
statistics from the Survey of Inmates: 36% of fathers and 59% of mothers cohabitate with
their minor children prior to incarceration (Mumola 2000). Consistent with our examination of the intensive margin of exposure, many children are also exposed to out-of-home
events, especially for more serious forms of exposure. This also reflects the evolving nature
of household composition for children growing up in the United States.
However, there are many channels through which exposure by a caregiver may indirectly
affect children, even if not currently or recently coresiding at the time of the criminal justice
event. Given the high divorce and separation rates in the U.S., it is very possible that
children still have significant contact with a biological parent that does not live in the home
(Cancian and Meyer 1998). Moreover, child support payments often follow parents regardless
of whether they coreside with their child or not (Amato and Gilbreth 1999). Thus, criminal
justice involvement may harm the finances of the child’s home even if the parent is not
35

This includes HUD, Medicare, and IHS. Medicaid records at the Census Bureau do not contain addresslevel information on enrollees.

18

currently living in the home.
For potential caregivers, note that exposure is not counted until we observe the caregiver
coresiding with the child. Thus, a stepparent or unclassified caregiver with a criminal justice
event prior to entering the household will not count as exposure during childhood to the
criminal justice system. We view this choice as conservative since pre-existing convictions,
for example, may continue to inhibit adult labor market success once cohabitation begins,
resulting in indirect impacts to the child.
Decomposing the determinants of exposure. To further explore the determinants
of the racial exposure gaps, we conduct a series of decomposition exercises in the style of
Blinder (1973) and Oaxaca (1973) which are presented in Figure 6.36 For each major type
of exposure (charges, felony charges, felony conviction, and incarceration), we evaluate the
role of observable differences between White children and Black, Hispanic, and American
Indian/Alaska Native children in explaining the overall gap. The exercises consider the
relative influence of county of birth, household income, and household structure (i.e., the
number of male and female, bio and non-bio adult links in the child’s life).
County of birth appears to play a minor role. For the White-Black comparison, if anything, the racial gap would be larger if White and Black children were equally distributed
across places of birth. Similar conclusions can be drawn from the White-American Indian/Alaska Native decomposition. The White-Hispanic gap, however, does appear to potentially partially attributable to difference in county of birth, explaining roughly 7% to 27%
of the raw gap.
When we add household income around the time of the child’s birth into the decomposition, a more substantial share of the raw gap across all minority groups is explained
by observable characteristics.37 Conditional on county of birth, household income explains
approximately 17% to 42% of the raw racial gap, depending on the specific type of exposure
and minority group under consideration.
The final observable trait we consider is household structure. The number of adults that
a child is linked with will obviously influence the likelihood of exposure to the justice system,
but at the same time, because this trait is realized over the course of their childhood, it may
be endogenous. For instance, if a child’s father is sent to prison, they may be more likely
to be linked with additional potential caregivers either because their mother moves in with
other family members or because she starts a new romantic relationship. For this reason,
36

To execute this exercise, we estimate models for both White and non-White groups separately, and
then take a population weighted average of the share of the explained variation stemming from observable
characteristics.
37
Household income is measured using the average AGI reported on the IRS Form 1040 that the child was
claimed on in their year of birth and subsequent four years.

19

these estimates should be evaluated with caution.
Household structure appears to play an important role in exposure to the justice system.
Conditional on county of birth and household income, realized household structure may
explain an additional 30% to 54% of the raw racial exposure gaps.38
After accounting for these three factors, almost the entire White-Hispanic gap is accounted for by observable characteristics. Likewise, about half to three quarters of the
White-American Indian/Alaska Native gap is explained by observable differences between
the racial groups. However, only about 50% to 60% of the White-Black gap is explained by
these characteristics, leaving a substantial portion of the differential unexplained. But, because realized family structure is endogenous, the shares explained by observable differences
are likely overstated for all groups.

7

Estimating the relationship between intergenerational exposure
and markers of childhood well-being

So far, we have documented profoundly high rates of intergenerational exposure to the U.S.
criminal justice system; for example, more than one in two Black children grow up in households where an adult has faced criminal charges during their childhood. But, how these
broad measures of intergenerational exposure actually impact child development and wellbeing remains unanswered.
It is entirely possible that less serious types of justice involvement (e.g., a criminal charge
versus prison), less meaningful adult-child relationships (e.g., an adult roommate versus a
biological parent), or less recent incidents of exposure are less impactful for children and/or
signal less about their environment. If true, the broad measures we introduce may not
be so dire. While we have documented new evidence on the reach of the justice system
within American households, researchers and policymakers would need to look further if the
ultimate goal is to understand and address the determinants of economic inequality and
racial inequities in the U.S.
However, evidence showing that these less studied and more common forms of intergenerational exposure to the justice system are harmful to children would raise serious concern.
Without remediation efforts, even a complete overhaul to American social policy to limit
intergenerational justice exposure going forward would only succeed in protecting future
generations; society would still have to bear the costs of the unrealized potential of, and
negative externalities from, current and former youth for decades to come.
38
Given our concerns regarding endogeneity, an alternative interpretation of these statistical relationships
is that the justice system has significant impacts on relationships, family stability, and living arrangements.

20

Establishing causal evidence uncontaminated by endogeneity bias on each of the dimensions of variation that we highlight (contemporaneous versus cumulative exposure, biological
parent versus all potential caregiver exposure, prison versus other justice events exposure) is
beyond the scope of this paper. Instead, we tackle a less perfect, but more realistic empirical
goal that we believe still yields valuable insight, noting that prevailing views of selection in
the literature would suggest that contemporaneous incarceration of biological parents should
be the most strongly correlated with negative outcomes for children.
We merge contemporaneous and cumulative exposure statuses over time to individual
observations from children (ages 0 to 18) in respondent households from the 2005 to 2018
waves of the American Community Survey (ACS) as well as to administrative data on teen
fertility (our relationship crosswalk), adult criminal justice contact through age 19 (CJARS),
and mortality through age 19 (Census Numident). We use the same exposure information
built using the microdata previously discussed and link at the individual-level to integrate
a range of well-being measures.39 We then examine the relative correlation between various
definitions of intergenerational exposure to the justice system and child outcomes. The
outcomes we consider include: (1) household-level variables—a child’s household poverty
status and whether a grandparent has primary responsibility for their care; (2) human capital
development measures—whether the child is behind in school given their age, has difficulty
concentrating, remembering, or making decisions, and whether they have dropped out of high
school; and, (3) other teenage outcomes—are they a teen parent, have they been criminally
charged in the adult system as a teenager, and did they die by age 19.
We estimate naive regressions without any controls showing the raw correlation between
each specific type of exposure and the outcome, as well as joint exposure specifications in
which we control for a third order polynomial in household income at birth, age-at-survey
fixed effects (“age ), fully saturated sex by race/ethnicity fixed effects (“race,sex ), survey-year
fixed effects (“t ), place-of-birth fixed effects (“geo ), and year-of-birth fixed effects (“by ), as
39

To avoid confusion, our exposure variables in this exercise do not reflect cohort-level exposure, but
instead whether that specific child respondent to the ACS had a parent or other potential caregiver involved
in the justice system in or before the year of their survey response. An adult in the household would
complete the ACS and respond to the questions for the child, including schooling information and cognitive
difficulty (see an example form here: https://www2.census.gov/programs-surveys/acs/methodology/questionnaires/
2020/quest20.pd). For teenage outcomes based on administrative data, we define contemporaneous exposure
as being exposed between the ages of 15 to 18, and prior exposure as ages 0 to 14 years old.

21

reflected in our estimating equation below.40
Outcomei,t = — bio,curr I(Bio-parent exposure)i,t + — other,curr I(Other adult exposure)i,t
Q

+ — bio,cum I a

t≠1
ÿ

· =by

Q

+ — other,cum I a

R

Bio-parent exposurei,· > 0b

t≠1
ÿ

· =by

R

Other adult exposurei,· > 0b

+ „1 HH Income + „1 HH Income2 + „1 HH Income3
+ “age + “race,sex + “t + “geo + “by + ‘i,t
To the extent that omitted variables bias contaminates our regression coefficients, one would
expect this to lead to relatively worse “impacts” associated with contemporaneous incarceration of biological parents. Children growing up in households with biological parents who
go to prison likely have more unobserved factors that inhibit their growth and development
compared to children with less serious, less direct, and less recent forms of exposure, for
example: their coresiding uncle who was once charged with possession of marijuana when
they were 2 years old.
Main results. Figure 7 plots the estimated coefficients between child outcomes and contemporaneous (in the survey year) and cumulative (prior to the survey year) criminal justice
exposure for each of the four event types (charge, felony charge, felony conviction, and prison)
for biological parents and other potential caregivers. The raw correlations without controls
are presented offset from the main results for comparison purposes.41
The estimated relationship between exposure and household poverty status (Panel A)
is remarkably similar after controlling for the observable characteristics described above.
Whether exposure reflects charges or incarceration, bio-parents or other household members,
or current or past events, the estimates consistently fall in the range of 60 to 90 percent of the
non-exposed child mean, with many of the coefficients being statistically indistinguishable.
These findings are consistent with the hypothesis that the negative economic impacts of
the justice system on families operate through the channel of criminal records more so than
incapacitation, which would mean that the impacts of the justice system are experienced
by a much larger swath of the population than generally recognized (Mueller-Smith and
Schnepel 2021). Whether grandparents are identified as the child’s primary caregiver (Panel
40

In addition, standard errors are clustered by the commuting zone/county of birth. Regressions are
weighted by the ACS-provided person weights.
41
Model estimates with controls can be found in Appendix Table C1. Raw correlations can be found in
Appendix Table D1.

22

B), however, does seem to be both: (1) more intimately connected to the most serious
forms of justice contact like felony convictions and incarceration, and (2) be most strongly
associated with the justice involvement of biological parents over other household members.
This clear pattern aligns with the legal processes in place for child welfare investigations,
child removal, and resulting kinship care placements.
The next three panels show a strikingly similar pattern of evidence regarding human capital formation during childhood. Whether considering being behind age-appropriate grade
level (Panel C), whether the child has shown signs of difficulty concentrating, remembering,
or making decisions (Panel D), or having dropped out of high school as a teenager (Panel
E), the strongest negative correlations are associated with cumulative rather than contemporaneous exposure to the justice system. Whether the source originated from a biological
parent or another adult in the household, or whether the type of exposure was incarceration
or something less serious, the estimated relationships are quite similar. This suggests that
some sort of social, emotional, or other scarring effects may be at work, and highlights the
importance of tracking cumulative exposure among children, which can be 10 times larger
than point-in-time estimates of contemporaneous exposure.
Teenage outcomes, that leverage administrative data on the full population,42 are the
final outcomes we consider.43 Teen parenthood (Panel F) follows a pattern most typically
associated with popular conceptions of the justice system: that current incarceration of a
family member has the largest impact on children. Interestingly, whether the source of
the exposure stems from a biological parent or another potential caregiver is statistically
indistinguishable, likely reflecting the departure many U.S. households have made from the
traditional nuclear family and the importance for children of non-biologically related adults
in their household.
The largest effects of exposure on being charged with an adult crime44 (Panel G) stem
from contemporaneous bio-parent justice involvement, regardless of the specific type of justice involvement. The estimated coefficients are roughly 3 times the size of the non-exposed
child mean. Other forms of exposure (other potential caregivers, prior exposures) remain
meaningfully high in the range of 1.5 to 2 times the non-exposed mean and are mostly
statistically indistinguishable.
Finally, teenage death (Panel H) follows a similar pattern to what we observed for educational outcomes. Cumulative exposure tends to have the largest and most precise effect
42

In this exercise, we exclude all observation that died prior to reaching age 12.
Model estimates with controls can be found in Appendix Table C12. Raw correlations can be found in
Appendix Table D2.
44
CJARS does not include data from the juvenile justice system, and so we are precluded from measuring
all criminal charges that might be occurring for these individuals.
43

23

on mortality, especially for the most serious forms of justice involvement. If this relationship
were purely mechanical (because the child had been killed by an adult household member),
the estimated relationship should principally load on contemporaneous exposure to felony
convictions and incarceration, which is not what we observe.45
Heterogeneity analysis. The impact of intergenerational exposure to the justice system
likely varies by a child’s background. In Figure 8, we further probe the relationship between exposure and child well-being through subgroup analysis along three dimensions: (1)
race/ethnicity, (2) sex, and (3) household income rank at birth. For the sake of simplicity,
we restrict ourselves to the broader form of exposure (any criminal charge) with control
variables for a select set of outcomes, but the full set of results for all potential specifications
can be found in Appendix Tables C1–C22.
In Figure 8A, we see what appears to be an inverse relationship between the underlying
non-exposed mean and the estimated coefficient on exposure of having a grandparent as a
primary caregiver. White and high-income children, who have the lowest rate of this outcome
in the non-exposed population, have the largest increase given some form of exposure in the
household. This suggests that there is heterogeneous capacity to leverage kinship networks
to remediate negative shocks arising from crime and justice involvement among adults in the
child’s life that reinforces economic inequality and racial inequities.
Figure 8B shows stark differences in human capital development among exposed children
by sex of the child. Boys appear to be significantly more sensitive to exposure, with estimated
coefficients more than double the size compared to those for girls. Given the disparity in
the originating source of exposure by the sex of the adult, it may be that boys are uniquely
impacted by the justice involvement or related events of their fathers and/or father-like
figures. More research is warranted to further probe this relationship.
Figures 8C and 8D continue this theme of differential effects by gender of the child.
Girls, and children from middle and low income families, are especially responsive on the
teen fertility outcome when after experience an exposure event. Boys, and children from
middle and low income families, are much more likely to be charged with adult criminal
charges after they experience an exposure event.
It is perhaps surprising that there is not clearer differentiation in the estimated effects of
exposure by racial and ethnic group. However, this should not be misconstrued as indicative
that the justice system does not potentially play a role in propagating racial inequities
across generations in U.S. As previously documented, there are stark differences in exposure
prevalences by child’s race, which itself will further racial gaps even with similar responses
45

Because we do not have access to cause of death, we unfortunately cannot differentiate between children
who were killed by members of their household versus other causes of death.

24

among children.

8

Conclusion

There is significant concern regarding, and a large body of evidence about, how childhood
conditions impact long-term outcomes. Given the wide reach of the criminal justice system
in the United States, the implications of the underlying illicit behavior for child safety and
household resources, and disproportionate impact among racial and ethnic minorities and
low-income communities (Shannon et al. 2017), intergenerational exposure to the criminal
justice system is likely an important factor in the development of many children in the U.S.
and contributes to the propagation of racial inequities.
However, the scope and role of intergenerational criminal justice exposure is largely unknown due to a multitude of data limitations in the United States. First, it is difficult to
measure and observe the relevant caregiver figures in a child’s life, particularly in light of
changing household structures over recent decades with substantial differences by racial and
ethnic groups. For example, paternal information is often left off of birth records and marriage has become less common among romantic partners in recent decades, which means they
do not file taxes and claim dependent children together. Second, criminal justice records are
not integrated across agencies, making it difficult to observe adult criminal justice contact
at the population level across geography and types of criminal justice involvement. To overcome these challenges, we have built residence and familial crosswalks within the Census
Bureau’s Data Linkage Infrastructure that leverage restricted-access microdata from Census
Bureau surveys, federal tax forms, and public program enrollment, and then link them to
the newly created Criminal Justice Administrative Records System (CJARS).
Using this new data infrastructure, built and linked at the person level, we are able
to produce novel measures of intergenerational exposure to charges, felony charges, felony
convictions, and incarcerations. For children born between 1999 and 2005, we find that 9% of
children have had an intergenerational exposure to prison, 18% have been exposed to a felony
conviction, and 39% have been exposed to any criminal charge over the course of childhood,
results that are much larger than previous estimates which primarily focused on parental
incarceration. Our estimates highlight, for the first time, the importance of other forms of
criminal justice exposure beyond incarceration, cumulative estimates that acknowledge the
potential for lagged and lasting impacts from crime and criminal justice contact, and the
importance of capturing diverse household structures.
We document important differences by race, which have significant implications for policies related to child well-being and persistent intergenerational inequalities. For example,

25

Black, non-Hispanic children have the highest rates of intergenerational exposure to prison
(20%), felony conviction (35%), felony charges (42%), and any criminal charge (62%). We
document similarly high rates for American Indian/Alaska Native children, a population
rarely studied, with corresponding estimates of 15%, 29%, 34%, and 60%. These estimates
stand in stark contrast to those of White and Asian children, who have the lowest rates of
intergenerational contact, with 6% and 2% of White and Asian children exposed to prison
and 32% and 17% of White and Asian children exposed to any charge during childhood,
respectively.
Finally, we document strong negative correlations between intergenerational exposure
to the criminal justice system and various measures of child development and well-being.
These findings support the conjecture that many forms of criminal justice contact beyond
contemporaneous exposure, exposure by a biological parent, and exposure to incarceration
have meaningful implications for childhood environments and outcomes. The degree and
expression of these negative outcomes varies by child’s gender and household’s income, with
more limited variation by child’s race (conditional on exposure). The results from these
exercises strengthen the case that our broader measures of intergenerational exposure to the
justice system are significant and consequential for the lives and outcomes of children. More
research is warranted to further probe the causality behind these estimated coefficients, but
given the scope of exposure that we identify, and the disparate racial nature of the exposure,
the justice system should become a first-order issue for those interested in economic inequality, intergenerational mobility, and racial inequities. Moreover, these results heighten the
need for policy accountability for factors that have contributed to such widespread intergenerational exposure to the justice system, and raise serious concern regarding the implications
for future generations given the extent of intergenerational exposure, especially when concentrated among children from racial and ethnic minorities.
As more data become available, future work should consider how exposure rates have
changed across temporal birth cohorts in the United States and between those born in different geographic jurisdictions. These are of particular interest given the changing landscape
of criminal justice policy over the last half century and vast differences in justice policy across
regions of the U.S. Such future work will be possible by leveraging the comprehensive and
growing geographic coverage of CJARS and expanding linkage opportunities with increasingly available historical data (e.g., past decennial censuses).

26

References
Agan, Amanda, and Sonja Starr. 2017. Ban the Box, Criminal Records, and Racial Discrimination: A Field Experiment. Quarterly Journal of Economics 133 (1): 191–235.
https://doi.org/10.1093/qje/qjx028.
Aizer, Anna. 2010. The Gender Wage Gap and Domestic Violence. American Economic
Review 100 (4): 1847–59. https://doi.org/10.1257/aer.100.4.1847.
Aizer, Anna, Laura Stroud, and Stephen Buka. 2016. Maternal Stress and Child Outcomes:
Evidence from Siblings. Journal of Human Resources 51 (3): 523–55. https://doi.org/
10.3368/jhr.51.3.0914-6664R.
Almond, Douglas, Janet Currie, and Valentina Duque. 2018. Childhood Circumstances and
Adult Outcomes: Act II. Journal of Economic Literature 56 (4): 1360–446. https://doi.
org/10.1257/jel.20171164.
Almond, Douglas, and Maya Rossin-Slater. 2013. Paternity Acknowledgment in 2 Million
Birth Records from Michigan. PLOS ONE 8 (7): 1–5. https://doi.org/10.1371/journal.
pone.0070042.
Amato, Paul R., and Joan G. Gilbreth. 1999. Nonresident Fathers and Children’s Well-Being:
A Meta-Analysis. Journal of Marriage and Family 61 (3): 557–73. https://doi.org/10.
2307/353560.
Andersson, Gunnar. 2002. Children’s Experience of Family Disruption and Family Formation: Evidence from 16 FFS Countries. Demographic Research 7 (7): 343–64. https :
//doi.org/10.4054/DemRes.2002.7.7.
Ang, Desmond. 2020. The Effects of Police Violence on Inner-City Students. Quarterly Journal of Economics 136 (1): 115–68. https://doi.org/10.1093/qje/qjaa027.
Arteaga, Carolina. 2018. The Cost of Bad Parents: Evidence from the Effects of Incarceration
on Children’s Education. Working paper. https : / / www . carolinaarteaga . com / s / jmp _
arteaga_nov2018-xask.pdf.
Balsa, Ana I. 2008. Parental Problem-drinking and Adult Children’s Labor Market Outcomes. Journal of Human Resources 43 (2): 454–86. https://doi.org/10.3368/jhr.43.2.
454.

27

Barclay, Gordon C., Cynthia Tavares, Sally Kenny, Arsalaan Siddique, and Emma Wilby.
2003. International Comparisons of Criminal Justice Statistics 2001. Home Office Statistical Bulletin 12 (3): 1–24. https://www.iprt.ie/site/assets/files/5897/crime_stats_
comparsions_uk_home_office.pdf.
Billings, Stephen B. 2018. Parental Arrest and Incarceration: How Does it Impact the Children? SSRN Working Paper 3034539. https://doi.org/10.2139/ssrn.3034539.
Black, Sandra E., Paul J. Devereux, and Kjell G. Salvanes. 2007. From the Cradle to the
Labor Market? The Effect of Birth Weight on Adult Outcomes. Quarterly Journal of
Economics 122 (1): 409–39. https://doi.org/10.1162/qjec.122.1.409.
Blinder, Alan S. 1973. Wage Discrimination: Reduced Form and Structural Estimates. Journal of Human Resources 8 (4): 436–55. https://doi.org/10.2307/144855.
Bond, Brittany, J. David Brown, Adela Luque, and Amy O’Hara. 2014. The Nature of
the Bias When Studying Only Linkable Person Records: Evidence from the American
Community Survey. Census Bureau CARRA Working Paper 2014-08, https : / / www .
census.gov/library/working-papers/2014/adrm/carra-wp-2014-08.html.
Bond, Timothy N., Jillian B. Carr, Analisa Packham, and Jonathan Smith. 2021. Hungry
for Success? SNAP Timing, High-Stakes Exam Performance, and College Attendance.
NBER Working Paper 28386. https://doi.org/10.3386/w28386.
Bramlett, Matthew D., and William D. Mosher. 2001. First Marriage Dissolution, Divorce,
and Remarriage: United States. Advance Data from Vital and Health Statistics 323:1–
20. https://www.cdc.gov/nchs/data/ad/ad323.pdf.
Bumpass, Larry, and Hsien-Hen Lu. 2000. Trends in Cohabitation and Implications for
Children’s Family Contexts in the United States. Population Studies 54 (1): 29–41.
https://doi.org/10.1080/713779060.
Bumpass, Larry L. 1990. What’s Happening to the Family? Interactions between Demographic and Institutional Change. Demography 27 (4): 483–98. https : / / doi . org / 10 .
2307/2061566.
Campbell, Frances, Gabriella Conti, James J. Heckman, Seong Hyeok Moon, Rodrigo Pinto,
Elizabeth Pungello, and Yi Pan. 2014. Early Childhood Investments Substantially Boost
Adult Health. Science 343 (6178): 1478–85. https://doi.org/10.1126/science.1248429.
Cancian, Maria, and Daniel R. Meyer. 1998. Who Gets Custody? Demography 35 (2): 147–
57. https://doi.org/10.2307/3004048.

28

Carrell, Scott E., and Mark L. Hoekstra. 2010. Externalities in the Classroom: How Children
Exposed to Domestic Violence Affect Everyone’s Kids. American Economic Journal:
Applied Economics 2 (1): 211–28. https://doi.org/10.1257/app.2.1.211.
Carson, E. Ann, and Elizabeth Anderson. 2016. Prisoners in 2015. Bureau of Justice Statistics
NCJ 250229. https://www.ncjrs.gov/App/Publications/abstract.aspx?ID=250229.
Cavanagh, Shannon E., and Paula Fomby. 2019. Family Instability in the Lives of American
Children. Annual Review of Sociology 45:493–513. https://doi.org/10.1146/annurevsoc-073018-022633.
Chatterji, Pinka, and Sara Markowitz. 2001. The Impact of Maternal Alcohol and Illicit
Drug Use on Children’s Behavior Problems: Evidence from the Children of the National
Longitudinal Survey Of Youth. Journal of Health Economics 20 (5): 703–31. https :
//doi.org/10.1016/S0167-6296(01)00090-X.
Cherlin, Andrew, and Frank F. Furstenberg Jr. 1992. The New American Grandparent: A
Place in the Family, A Life Apart. Cambridge, MA: Harvard University Press. https:
//www.hup.harvard.edu/catalog.php?isbn=9780674608382.
Cherlin, Andrew J. 2004. The Deinstitutionalization of American Marriage. Journal of Marriage and Family 66 (4): 848–61. https://doi.org/10.1111/j.0022-2445.2004.00058.x.
Chetty, Raj, John N. Friedman, Nathaniel Hendren, Maggie R. Jones, and Sonya R. Porter.
2018. The Opportunity Atlas: Mapping the Childhood Roots of Social Mobility. NBER
Working Paper 25147. https://doi.org/10.3386/w25147.
Chetty, Raj, Nathaniel Hendren, and Lawrence F. Katz. 2016. The Effects of Exposure to
Better Neighborhoods on Children: New Evidence from the Moving to Opportunity
Experiment. American Economic Review 106 (4): 855–902. https://doi.org/10.1257/
aer.20150572.
Choi, Jay, David Kilmer, Michael Mueller-Smith, and Sema Taheri. 2022. Hierarchical Approaches to Text-based Offense Classification. Working paper. http://sites.lsa.umich.
edu / mgms / wp - content / uploads / sites / 283 / 2022 / 01 / CJARS _ MFJ _ offense _ classification _
20220119.pdf.

Choi, Moon, Ginny Sprang, and Jessica G. Eslinger. 2016. Grandparents Raising Grandchildren. Family & Community Health 39 (2): 120–28. https://doi.org/10.1097/FCH.
0000000000000097.

29

Cohen, D’Vera, and Jeffrey Passel. 2018. A Record 64 Million Americans Live in Multigenerational Households. Washington, DC: Pew Research Center. https://www.pewresearch.
org/fact- tank/2018/04/05/a- record- 64- million- americans- live- in- multigenerationalhouseholds.

Cunningham, Scott, and Keith Finlay. 2013. Parental Substance Use and Foster Care: Evidence from two Methamphetamine Supply Shocks. Economic Inquiry 51 (1): 764–82.
https://doi.org/10.1111/j.1465-7295.2012.00481.x.
Currie, Janet, and Douglas Almond. 2011. Human Capital Development before Age Five.
In Handbook of Labor Economics, edited by David Card and Orley Ashenfelter, 4:1315–
486. Amsterdam: Elsevier. https://doi.org/10.1016/S0169-7218(11)02413-0.
Currie, Janet, Michael Mueller-Smith, and Maya Rossin-Slater. 2022. Violence While in
Utero: The Impact of Assaults During Pregnancy on Birth Outcomes. Review of Economics and Statistics, 1–16. https://doi.org/10.1162/rest_a_00965.
Currie, Janet, and Erdal Tekin. 2012. Understanding the Cycle: Childhood Maltreatment
and Future Crime. Journal of Human Resources 47 (2): 509–49. https://doi.org/10.
3368/jhr.47.2.509.
Currie, Janet, and Cathy Spatz Widom. 2010. Long-Term Consequences of Child Abuse
and Neglect on Adult Economic Well-Being. Child Maltreatment 15 (2): 111–20. https:
//doi.org/10.1177/1077559509355316.
Curtin, Sally C., Stephanie J. Ventura, and Gladys M. Martinez. 2014. Recent Declines in
Nonmarital Childbearing in the United States. National Center for Health Statistics
Data Brief 162. https://www.cdc.gov/nchs/products/databriefs/db162.htm.
Dahl, Gordon B., and Lance Lochner. 2012. The Impact of Family Income on Child Achievement: Evidence from the Earned Income Tax Credit. American Economic Review 102
(5): 1927–56. https://doi.org/10.1257/aer.102.5.1927.
Deming, David. 2009. Early Childhood Intervention and Life-Cycle Skill Development: Evidence from Head Start. American Economic Journal: Applied Economics 1 (3): 111–34.
https://doi.org/10.1257/app.1.3.111.
Deshpande, Manasi, and Michael G. Mueller-Smith. 2022. Does Welfare Prevent Crime? The
Criminal Justice Outcomes of Youth Removed from SSI. NBER Working Paper 29800.
https://doi.org/10.3386/w29800.

30

Dobbie, Will, Jacob Goldin, and Crystal S. Yang. 2018. The Effects of Pretrial Detention on
Conviction, Future Crime, and Employment: Evidence from Randomly Assigned Judges.
American Economic Review 108 (2): 201–40. https://doi.org/10.1257/aer.20161503.
Doyle, Jr., Joseph J. 2007. Child Protection and Child Outcomes: Measuring the Effects of
Foster Care. American Economic Review 97 (5): 1583–610. https://doi.org/10.1257/
aer.97.5.1583.
Doyle, Jr., Joseph J., and Anna Aizer. 2018. Economics of Child Protection: Maltreatment,
Foster Care, and Intimate Partner Violence. Annual Review of Economics 10 (1): 87–
108. https://doi.org/10.1146/annurev-economics-080217-053237.
Dynarski, Susan, Joshua Hyman, and Diane Whitmore Schanzenbach. 2013. Experimental
Evidence on the Effect of Childhood Investments on Postsecondary Attainment and
Degree Completion. Journal of Policy Analysis and Management 32 (4): 692–717. https:
//doi.org/10.1002/pam.21715.
Ellis, Renee R., and Tavia Simmons. 2014. Coresident Grandparents and Their Grandchildren: 2012. Census Bureau Report P20-576. https://www.census.gov/content/dam/Census/
library/publications/2014/demo/p20-576.pdf.
Enns, Peter K., Youngmin Yi, Megan Comfort, Alyssa W. Goldman, Hedwig Lee, Christopher Muller, Sara Wakefield, Emily A. Wang, and Christopher Wildeman. 2019. What
Percentage of Americans Have Ever Had a Family Member Incarcerated?: Evidence from
the Family History of Incarceration Survey (FamHIS). Socius: Sociological Research for
a Dynamic World 5:1–45. https://doi.org/10.1177/2378023119829332.
Eriksson, Karin Hederos, Randi Hjalmarsson, Matthew J. Lindquist, and Anna Sandberg.
2016. The Importance of Family Background and Neighborhood Effects as Determinants
of Crime. Journal of Population Economics 29 (1): 219–62. https://doi.org/10.1007/
s00148-015-0566-8.
Fang, Xiangming, Derek S. Brown, Curtis S. Florence, and James A. Mercy. 2012. The
Economic Burden of Child Maltreatment in the United States and Implications for
Prevention. Child Abuse & Neglect 36 (2): 156–65. https://doi.org/10.1016/j.chiabu.
2011.10.006.

31

Finlay, Keith, Matthew Gross, Elizabeth Luh, and Michael Mueller-Smith. 2022. The Impact
of Financial Sanctions in the U.S. Justice System: Regression Discontinuity Evidence
from Michigan’s Driver Responsibility Program. Unpublished manuscript. http://sites.
lsa . umich . edu / mgms / wp - content / uploads / sites / 283 / 2022 / 02 / CJARS _ FinSanc _ MI _ DRF _
20220225.pdf.

Finlay, Keith, and Michael Mueller-Smith. 2020. Criminal Justice Administrative Records
System (CJARS) [dataset]. Ann Arbor, MI: University of Michigan. https://cjars.isr.
umich.edu.
Fomby, Paula, and Andrew J. Cherlin. 2007. Family Instability and Child Well-Being. American Sociological Review 72 (2): 181–204. https://doi.org/10.1177/000312240707200203.
Genadek, Katie, Joshua Sanders, and Amanda Jean Stevenson. 2021. Measuring U.S. Fertility using Administrative Data from the Census Bureau. Census Bureau ADEP Working Paper 2021-02. https : / / www . census . gov / library / working - papers / 2021 / econ / adep measuring-us-fertility-workingpaper.html.
Genadek, Katie R., and J. Trent Alexander. 2019. The Decennial Census Digitization and
Linkage Project. Census Bureau ADEP Working Paper 2019-01. https://www.census.
gov/library/working-papers/2019/econ/adep-wp-dc-digitization-linkage.html.
Glaze, Lauren, and Laura Maruschak. 2008. Parents in Prison and Their Minor Children.
Bureau of Justice Statistics NCJ 222984. https://bjs.ojp.gov/content/pub/pdf/pptmc.pdf.
Gross, Max, and E. Jason Baron. 2022. Temporary Stays and Persistent Gains: The Causal
Effects of Foster Care. American Economic Journal: Applied Economics forthcoming.
https://doi.org/10.1257/app.20200204.
Harlow, Caroline Wolf. 2003. Education and Correctional Populations. Bureau of Justice
Statistics NCJ 195670. https://www.ncjrs.gov/App/Publications/abstract.aspx?ID=195670.
Harris, Alexes, Heather Evans, and Katherine Beckett. 2010. Drawing Blood from Stones:
Legal Debt and Social Inequality in the Contemporary United States. American Journal
of Sociology 115 (6): 1753–99. https://doi.org/10.1086/651940.
Hayslip, Jr., Bert, and Patricia L. Kaminski. 2005. Grandparents Raising Their Grandchildren: A Review of the Literature and Suggestions for Practice. Gerontologist 45 (2):
262–69. https://doi.org/10.1093/geront/45.2.262.

32

Herrenkohl, Todd I., Cynthia Sousa, Emiko A. Tajima, Roy C. Herrenkohl, and Carrie
A. Moylan. 2008. Intersection of Child Abuse and Children’s Exposure to Domestic Violence. Trauma, Violence, & Abuse 9 (2): 84–99. https : / / doi . org / 10 . 1177 /
1524838008314797.
Heuveline, Patrick, Jeffrey M. Timberlake, and Jr. Furstenberg Frank F. 2003. Shifting Childrearing to Single Mothers: Results from 17 Western Countries. Population and Development Review 29 (1): 47–71. https://doi.org/10.1111/j.1728-4457.2003.00047.x.
Hjalmarsson, Randi, and Matthew J. Lindquist. 2012. Like Godfather, Like Son: Exploring
the Intergenerational Nature of Crime. Journal of Human Resources 47 (2): 550–82.
https://doi.org/10.3368/jhr.47.2.550.
. 2013. The Origins of Intergenerational Associations in Crime: Lessons from Swedish
Adoption Data. Labour Economics 20:68–81. https://doi.org/10.1016/j.labeco.2012.
11.001.
Hoynes, Hilary, Marianne Page, and Ann Huff Stevens. 2011. Can Targeted Transfers Improve Birth Outcomes?: Evidence from the Introduction of the WIC Program. Journal
of Public Economics 95 (7): 813–27. https://doi.org/10.1016/j.jpubeco.2010.12.006.
Hoynes, Hilary, Diane Whitmore Schanzenbach, and Douglas Almond. 2016. Long-Run Impacts of Childhood Access to the Safety Net. American Economic Review 106 (4): 903–
34. https://doi.org/10.1257/aer.20130375.
Isen, Adam, and Betsey Stevenson. 2010. Women’s Education and Family Behavior: Trends
in Marriage, Divorce and Fertility. In Demography and the Economy, edited by John
B. Shoven, 107–40. University of Chicago Press. https : / / doi . org / 10 . 7208 / chicago /
9780226754758.001.0001.
Kreider, Rose Marie, and Renee Ellis. 2011. Number, Timing, and Duration of Marriages
and Divorces: 2009. Census Bureau Report P70-125. https://www.census.gov/library/
publications/2011/demo/p70-125.html.
Landry-Meyer, Laura, and Barbara M. Newman. 2004. An Exploration of the Grandparent
Caregiver Role. Journal of Family Issues 25 (8): 1005–25. https://doi.org/10.1177/
0192513X04265955.

33

Lee, Hedwig, Tyler McCormick, Margaret T. Hicken, and Christopher Wildeman. 2015.
Racial Inequalities in Connectedness to Imprisoned Individuals in the United States.
Du Bois Review: Social Science Research on Race 12 (2): 269–82. https://doi.org/10.
1017/S1742058X15000065.
Levendosky, Alytia A., Alissa C. Huth-Bocks, Deborah L. Shapiro, and Michael A. Semel.
2003. The Impact of Domestic Violence on the Maternal-Child Relationship and PreschoolAge Children’s Functioning. Journal of Family Psychology 17 (3): 275–87. https://doi.
org/10.1037/0893-3200.17.3.275.
Lichter, Daniel T., Diane K. McLaughlin, George Kephart, and David J. Landry. 1992. Race
and the Retreat from Marriage: A Shortage of Marriageable Men? American Sociological
Review 57 (6): 781–99. https://doi.org/10.2307/2096123.
Livingston, Gretchen. 2014. Four-in-Ten Couples are Saying ‘I Do,’ Again. Washington, DC:
Pew Research Center. https://www.pewresearch.org/social-trends/2014/11/14/four-inten-couples-are-saying-i-do-again.
Lofquist, Daphne. 2012. Multigenerational Households: 2009–2011. Census Bureau Report
ACSBR/11-03. https://www.census.gov/library/publications/2012/acs/acsbr11-03.html.
Luque, Adela, and Deborah Wagner. 2015. Assessing Coverage and Quality of the 2007
Prototype Census Kidlink Database. Census Bureau CARRA Working Paper 2015-07.
https://www.census.gov/library/working-papers/2015/adrm/carra-wp-2015-07.html.
Martin, Karin D., Bryan L. Sykes, Sarah Shannon, Frank Edwards, and Alexes Harris. 2018.
Monetary Sanctions: Legal Financial Obligations in US Systems of Justice. Annual Review of Criminology 1 (1): 471–95. https://doi.org/10.1146/annurev-criminol-032317091915.
Massenkoff, Maxim, and Evan K. Rose. 2020. Family Formation and Crime. Working paper.
http://maximmassenkoff.com/FamilyFormationAndCrime.pdf.
Mathews, T.J., and Brady E. Hamilton. 2016. Mean Age of Mothers Is on the Rise: United
States, 2000–2014. NCHS Data Brief Number 232. Hyattsville, MD: National Center for
Health Statistics. https://www.cdc.gov/nchs/products/databriefs/db232.htm.
. 2019. Total Fertility Rates by State and Race and Hispanic Origin: United States,
2017. National Vital Statistics Reports 68 (1): 1–10. https://stacks.cdc.gov/view/cdc/
61878.

34

McLanahan, Sara, and Christine Percheski. 2008. Family Structure and the Reproduction
of Inequalities. Annual Review of Sociology 34 (1): 257–76. https://doi.org/10.1146/
annurev.soc.34.040507.134549.
Medicine, Institute of, and National Research Council. 2014. New Directions in Child Abuse
and Neglect Research. Edited by Anne C. Petersen, Joshua Joseph, and Monica Feit.
Washington, DC: National Academies Press. https://doi.org/10.17226/18331.
Milligan, Kevin, and Mark Stabile. 2011. Do Child Tax Benefits Affect the Well-Being of Children? Evidence from Canadian Child Benefit Expansions. American Economic Journal:
Economic Policy 3 (3): 175–205. https://doi.org/10.1257/pol.3.3.175.
Mueller-Smith, Michael. 2015. The Criminal and Labor Market Impacts of Incarceration.
Unpublished manuscript. http://sites.lsa.umich.edu/mgms/wp- content/uploads/sites/
283/2015/09/incar.pdf.
Mueller-Smith, Michael, and Kevin T. Schnepel. 2021. Diversion in the Criminal Justice
System. Review of Economic Studies 88 (2): 883–936. https://doi.org/10.1093/restud/
rdaa030.
Mumola, Christopher J. 2000. Incarcerated Parents and Their Children. Bureau of Justice
Statistics NCJ 182335. https://bjs.ojp.gov/library/publications/incarcerated-parentsand-their-children.
National Academies of Sciences, Engineering, and Medicine. 2019. Consequences of Child
Poverty. In A Roadmap to Reducing Child Poverty, edited by Greg Duncan and Suzanne
Le Menestrel. Washington, DC: National Academies Press. https://doi.org/10.17226/
25246.
National Research Council. 2014. The Growth of Incarceration in the United States: Exploring
Causes and Consequences. Edited by Jeremy Travis, Bruce Western, and Steve Redburn.
Washington, DC: National Academies Press. https://doi.org/10.17226/18613.
Norris, Samuel, Matthew Pecenco, and Jeffrey Weaver. 2021. The Effects of Parental and
Sibling Incarceration: Evidence from Ohio. American Economic Review 111 (9): 2926–
63. https://doi.org/10.1257/aer.20190415.
Oaxaca, Ronald. 1973. Male-Female Wage Differentials in Urban Labor Markets. International Economic Review 14 (3): 693–709. https://doi.org/10.2307/2525981.
Pager, Devah. 2003. The Mark of a Criminal Record. American Journal of Sociology 108 (5):
937–75. https://doi.org/10.1086/374403.

35

Parker, Emily, Sharon Sassler, and Laura Tach. 2021. Fatherhood and Racial/Ethnic Differences in the Progression of Romantic Relationships. Journal of Marriage and Family
83 (4): 985–1003. https://doi.org/10.1111/jomf.12733.
Persson, Petra, and Maya Rossin-Slater. 2018. Family Ruptures, Stress, and the Mental
Health of the Next Generation. American Economic Review 108 (4-5): 1214–52. https:
//doi.org/10.1257/aer.20141406.
Peterson, Cora, Curtis Florence, and Joanne Klevens. 2018. The Economic Burden of Child
Maltreatment in the United States, 2015. Child Abuse & Neglect 86:178–83. https :
//doi.org/10.1016/j.chiabu.2018.09.018.
Raley, R. Kelly, and Megan M. Sweeney. 2020. Divorce, Repartnering, and Stepfamilies: A
Decade in Review. Journal of Marriage and Family 82 (1): 81–99. https://doi.org/10.
1111/jomf.12651.
Raley, R. Kelly, Megan M. Sweeney, and Danielle Wondra. 2015. The Growing Racial and
Ethnic Divide in U.S. Marriage Patterns. The Future of Children 25 (2): 89–109. https:
//doi.org/10.1353/foc.2015.0014.
Raley, R. Kelly, and Elizabeth Wildsmith. 2004. Cohabitation and Children’s Family Instability. Journal of Marriage and Family 66 (1): 210–19. https://doi.org/10.1111/j.00222445.2004.00014.x-i1.
Roman, Caterina Gouvis, and Jeremy Travis. 2004. Taking Stock: Housing, Homelessness,
and Prisoner Reentry. Washington, DC: Urban Institute. http://webarchive.urban.org/
UploadedPDF/411096_taking_stock.pdf.
Rossin-Slater, Maya. 2017. Signing Up New Fathers: Do Paternity Establishment Initiatives
Increase Marriage, Parental Investment, and Child Well-Being? American Economic
Journal: Applied Economics 9 (2): 93–130. https://doi.org/10.1257/app.20150314.
Ruggles, Steven, Sarah Flood, Sophia Foster, Ronald Goeken, Jose Pacas, Megan Schouweiler,
and Matthew Sobek. 2021. IPUMS USA: Version 11.0 [dataset]. Minneapolis, MN:
IPUMS, 2021. https://doi.org/10.18128/D010.V11.0.
Schoen, Robert, and Nicola Standish. 2001. The Retrenchment of Marriage: Results from
Marital Status Life Tables for the United States, 1995. Population and Development
Review 27 (3): 553–63. https://doi.org/10.1111/j.1728-4457.2001.00553.x.

36

Shalev, Idan, Terrie E. Moffitt, Karen Sugden, Brittany Williams, Renate M. Houts, Andrea
Danese, Jonathan Mill, Louise Arseneault, and Avshalom Caspi. 2013. Exposure to
Violence during Childhood Is Associated with Telomere Erosion from 5 to 10 Years of
Age: A Longitudinal Study. Molecular Psychiatry 18 (5): 576–81. https://doi.org/10.
1038/mp.2012.32.
Shannon, Sarah K. S., Christopher Uggen, Jason Schnittker, Melissa Thompson, Sara Wakefield, and Michael Massoglia. 2017. The Growth, Scope, and Spatial Distribution of People With Felony Records in the United States, 1948–2010. Demography 54 (5): 1795–
818. https://doi.org/10.1007/s13524-017-0611-1.
Sharkey, Patrick. 2010. The Acute Effect of Local Homicides on Children’s Cognitive Performance. Proceedings of the National Academy of Sciences 107 (26): 11733–38. https:
//doi.org/10.1073/pnas.1000690107.
Sharkey, Patrick T., Nicole Tirado-Strayer, Andrew V. Papachristos, and C. Cybele Raver.
2012. The Effect of Local Violence on Children’s Attention and Impulse Control. American Journal of Public Health 102 (12): 2287–93. https://doi.org/10.2105/ajph.2012.
300789.
Smock, Pamela J., and Christine R. Schwartz. 2020. The Demography of Families: A Review
of Patterns and Change. Journal of Marriage and Family 82 (1): 9–34. https://doi.org/
10.1111/jomf.12612.
Sweeney, Megan M. 2010. Remarriage and Stepfamilies: Strategic Sites for Family Scholarship in the 21st Century. Journal of Marriage and Family 72 (3): 667–84. https :
//doi.org/10.1111/j.1741-3737.2010.00724.x.
Sykes, Bryan L., and Becky Pettit. 2014. Mass Incarceration, Family Complexity, and the Reproduction of Childhood Disadvantage. ANNALS of the American Academy of Political
and Social Science 654 (1): 127–49. https://doi.org/10.1177/0002716214526345.
Teachman, Jay. 2008. Complex Life Course Patterns and the Risk of Divorce in Second
Marriages. Journal of Marriage and Family 70 (2): 294–305. https://doi.org/10.1111/
j.1741-3737.2008.00482.x.
Wagner, Deborah, and Mary Lane. 2014. The Person Identification Validation System (PVS):
Applying the Center for Administrative Records Research and Applications’ (CARRA)
Record Linkage Software. Census Bureau CARRA Working Paper 2014-01. https://
www.census.gov/library/working-papers/2014/adrm/carra-wp-2014-01.html.

37

Widom, Cathy Spatz. 1989. The Cycle of Violence. Science 244 (4901): 160–66. https://doi.
org/10.1126/science.2704995.
Wildeman, Christopher. 2009. Parental Imprisonment, the Prison Boom, and the Concentration of Childhood Disadvantage. Demography 46 (2): 265–80. https://doi.org/10.
1353/dem.0.0052.
Wildeman, Christopher, and Lars H. Andersen. 2015. Cumulative Risks of Paternal and
Maternal Incarceration in Denmark and the United States. Demographic Research 32
(57): 1567–80. https://doi.org/10.4054/DemRes.2015.32.57.
Wildeman, Christopher, and Signe Hald Andersen. 2017. Paternal Incarceration and Children’s Risk of Being Charged by Early Adulthood: Evidence from a Danish Policy Shock.
Criminology 55 (1): 32–58. https://doi.org/10.1111/1745-9125.12124.
Williams, Molly N. 2011. The Changing Roles of Grandparents Raising Grandchildren. Journal of Human Behavior in the Social Environment 21 (8): 948–62. https://doi.org/10.
1080/10911359.2011.588535.

38

Figures and tables
Figure 1: Share of children in 1999–2005 birth cohorts with observed relation types
A: Share of children with biological parent and other potential caregiver relationships
100%
95% 97%

90%

90%

100%

98% 98%

99% 99%

93% 94%

89% 89%

85%

80%
70%

94%

76%

92%

91%

87%

75%

60%

96%

98%

72%

61%
56%

50%
40%
30%
20%
10%
0%
All

White,
Non-Hispanic

Black,
Non-Hispanic

Hispanic

Asian

American Indian

bio-parent
•Female,
Female, bio-parent/other caregiver

■ Male,

bio-parent
Male,
bio-parent/other caregiver
□

□

B: Share of children with specific observed potential caregiver relationships
100%
90%
80%
70%

64%
52%

52%

50%

46%
40%

40%

38%

5%

32%

28%

22%

20%

41%

37%

29%

30%
10%

59%

58%

60%

17%

31%

22%
13%

6%

5%

White,
Non-Hispanic

Black,
Non-Hispanic

7%

7%

4%

0%
All

■ Step/Foster/Adoptive

Parent
Unclassified
caregiver
□

Hispanic

Asian

family
Unclassified
cohabiting adult
□

American Indian

□ Extended

Source: Calculations are based on the Census Numident, the Census BestRace files, CJARS, and the
CJARS relationship crosswalk.
Notes: This figure depicts the types of parental relationships identified for children in Census Numident born between 1999–2005 that are identified with a potential caregiver relationship at the national
level. Potential caregivers are defined as biological parents, stepparents, adopted parents, foster parents, unclassified caregivers, grandparents, aunts/uncles, non-familial adult (cohabiting 2+ years), and
unclassified adult (cohabiting 2+ years). All results were approved for release by the U.S. Census
Bureau, authorization number CBDRB-FY22-ERD002-001.

39

Figure 2: Exposure to the criminal justice system and comparison of measures

A: Contemporaneous exposure, biological par- B: Cumulative exposure, biological and other poents
tential caregivers parents
4%

40%

3.7%

35%
30%

3%

25%
20%

2%

15%

1.2%
0.9%

1%

0.8%

10%
5%

0.3%

0%

0%
Charge

Felony
Felony
Prison
Charge Conviction Entry

0

In Prison

3

12

15

18

Felony charge

Felony conviction

Bio-parents

With caregivers

In prison

D: Intensive margin of all-source, cumulative exposure
60%

45%
45%
40%
40%
35%
35%
30%
30%
25%
25%
20%
20%
15%
15%
10%
10%
5%
5%
0%
0%

9

Child Age
Charge

C: Comparison of exposure measures

6

38.9%

50%
40%

26.0%
21.4%

30%

18.3%

20%

11.4%
3.7%

9.2%

0.9%

Felony
Charge

•Cumulative

8.8%
3.6%

1.2%

Charge

■ Contemp.

Median intensive margin
experience conditional on
extensive margin exposure

Felony
1 Convinction

10%

0.8%

0%

In prison

1

□ Cumulative, w/ caregivers

■ Any

2 3 4 5 6 7 8 9 10+
Total Number of Exposure Events
Charge

□ Prison

Spells

■ Felony
□ Years

Charge

□ Felony

Conviction

of Prison

Source: Calculations are based on the Census Numident, CJARS, and the CJARS relations and residency crosswalks.
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. For Panels A, B, and C, the sample consists
of individuals in the Census Numident 1999–2005 birth cohorts in CJARS-covered geographies from birth until X, where X
represents years since birth (0–18) with the place of birth still covered or year 2018. CJARS court records cover AZ, FL, MD,
MI, NJ, NC, ND, OR, TX, and WI. CJARS incarceration records cover AZ, FL, MI, NE, NC, PA, TX, WA, and WI. Potential
caregivers are defined as biological parents, stepparents, adopted parents, foster parents, unclassified caregivers, grandparents,
aunts/uncles, non-familial adult (cohabiting 2+ years), and unclassified adult (cohabiting 2+ years). For Panel D, the sample
consists of individuals in the Census Numident 1999–2000 birth cohorts in CJARS-covered geographies from birth until age 18.
Distinct events are counted among children with any exposure. Thus, multiple charges filed on the same date are considered
one event, and similarly for the other types of criminal justice events. The number of events is truncated at 10+ events for
all events except prison spells, which are top coded at 8+. All results were approved for release by the U.S. Census Bureau,
authorization numbers CBDRB-FY22-ERD002-001, CBDRB-FY22-ERD002-003, and CBDRB-FY22-ERD002-009.

40

Figure 3: Heterogeneous cumulative exposure to the criminal justice system by all potential
caregivers by age 18: charge, felony charge, felony conviction, incarceration
A: By child’s race

B: By parents’ income rank

70%

70%
62%

60%

60%
50%

□
Charge

60%

Felony charge

Felony conviction

50%

45%

In prison

42%

40%

35%

32%

29%

30%
20%

25%
22%

20%

17%

16%
14%

10%
0%

40%

34%

15%

30%
20%

11%
6%

6%

White,
Black,
Non-Hispanic Non-Hispanic

■ Charge II Felony

Hispanic

charge

5%

10%
2%

Asian

Felony conviction

0%

American
Indian

□ In

0

20

40

60

80

100

Parent Household Income Rank

prison

C: By adults’ sex
40%

39%

Any adult exposure
Male adult exposure

33%

30%

Female adult exposure
21%

20%

18%

18%

18%

15%

10%

8%

9%
6%

8%
2%

0%

Charge

Felony charge

Felony conviction

In prison

Source: Calculations are based on the Census Numident, the Census BestRace files, CJARS, the CJARS
relations and residency crosswalks, and IRS Form 1040s (1999–2009 tax years).
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. The sample consists
of individuals in the Census Numident 1999–2000 birth cohorts in CJARS-covered geographies from
birth until age 18. Average exposure by age 18 is depicted for children across race (Panel A), income
percentile bins (Panel B), and adult potential caregivers’ sex (Panel C). Income percentile bins are
determined using the average adjusted gross income reported on IRS Form 1040s, in which the child is
claimed for the first five years. Children claimed on a form with negative AGI or never claimed in the
first five years are not included in the sample. All results were approved for release by the U.S. Census
Bureau, authorization numbers CBDRB-FY22-ERD002-001 and CBDRB-FY22-ERD002-003.

41

Figure 4: Cumulative exposure to criminal charges by offense type
A: Overall

30%

28.6%

25%

20%

18.1%

18.5%

Other Traffic

Public Order

17.2%
15.7%
15%
12.7%
10%

5%

0%
Violent

Property

Drug

DUI

B: By child’s race

C: By parents’ income rank

50%

50%

40%

40%

30%

30%

20%

20%

10%

10%

0%

0%
White, NonHispanic
■

Violent

Black, NonHispanic

• Property

II Drug

Hispanic
□

DUI

□

Asian
Other Traffic

1st
Quantile

American
Indian

Violent

Public Order

2nd
Quantile
Property

3rd
Quantile
Drug

DUI

4th
Quantile
Other Traffic

5th
Quantile
Public Order

Source: Calculations are based on the Census Numident, the Census BestRace files, CJARS, the CJARS
relations and residency crosswalks, and IRS Form 1040s (1999–2009 tax years).
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. The sample consists
of individuals in the Census Numident 1999–2000 birth cohorts in CJARS-covered geographies from
birth until age 18. Average exposure by age 18 to specific charge types is depicted for children overall
(Panel A), across race (Panel B), and across household income quantiles (Panel C). Income quantiles
are determined using the average adjusted gross income reported on IRS Form 1040s, in which the
child is claimed for the first five years. Children claimed on a form with negative AGI or never claimed
in the first five years are not included in the sample. All results were approved for release by the U.S.
Census Bureau, authorization number CBDRB-FY22-ERD002-009.

42

Figure 5: Cumulative current or recent coresidency exposure to the criminal justice system
A: All potential caregivers

45%
40%

Charge
Felony charge
Felony conviction
In prison

35%
30%

Share of exposed minors with:
In-home event Out-of-home event
79.2%
74.9%
67.4%
79.6%
64.8%
80.7%
54.2%
88.6%

------------- ---- --·-

25%
20%
15%
10%
5%
0%
0

3

6

9
Child Age

12

Felony
Felony
conviction
B: charge
All biological
parents
All exposures
In-home exposures

Charge
30%

15

18

In prison

Share of exposed minors with:

25%

Charge
Felony charge
Felony conviction
In prison

20%

In-home event
82.6%
69.3%
66.2%
50.4%

Out-of-home event
49.2%
54.5%
56.3%
67.9%

- -----·
..... . -·-

- - - - -

___ .-·-· ---·-

. . . .

•

15%

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

10%

. ... ... ·

5%
0%
0

3

6

9
Child Age

Charge

12

15

18

Felony charge
Felony conviction
In prison
All exposures
In-home exposures
Source: Calculations are based on the Census Numident, CJARS, and CJARS residence and relations
crosswalk.
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. The sample consists
of individuals in the Census Numident 1999–2005 birth cohorts in CJARS-covered geographies from
birth until X, where X represents years since birth (0–18) with the place of birth still covered or year
2018. In-home exposure is defined as exposure by an individual that was coresiding with the child in
the year of the event or in the preceding two years. All results were approved for release by the U.S.
Census Bureau, authorization numbers CBDRB-FY22-ERD002-001 and CBDRB-FY22-ERD002-003.

43

Figure 6: Oaxaca-Blinder decomposition of White-Minority gaps in intergenerational exposure
A: Charge

White - Black

White - Hispanic

B: Felony Charge

White - Black

White - American Indian

White - Hispanic

White - American Indian

0.05

0.05
-2%

0.00
2%

0.00

7%

-1%
2%

14%

-0.05
21%

-0.05

44%

15%
52%

28%

-0.10

41%

-0.10
107%

93%

-0.15

58%

-0.25

51%

-0.15

60%

-0.20
Share of raw gap
explained by
differences in
observable
characteristics

-0.20

74%

Share of raw gap
explained by
differences in
observable
characteristics

-0.25

-0.30

-0.30

-0.35
□ County

of birth

□+

Household income

iii +

Household structure

■ Raw

□ County

gap

of birth

□+

C: Felony Conviction
White - Black

White - Hispanic

Household income

■+

Household structure

■ Raw

gap

D: Incarceration

White - American Indian

White - Black

0.05

White - Hispanic

White - American Indian

0.02
-4%
-5%

0.00

0%

0.00
19%

13%

-0.05

27%

42%

31%

20%

-0.04

55%

-0.10

65%

-0.06

68%
114%

107%

53%

-0.08

52%
77%

-0.15

-0.20

0%

-0.02

-0.10
Share of raw gap
explained by
differences in
observable
characteristics

-0.12

Share of raw gap
explained by
differences in
observable
characteristics

-0.14
-0.25

-0.16
□ County

of birth

□+

Household income

+ Household structure

■ Raw

gap

□ County

of birth

□+

Household income

■+

Household structure

■ Raw

gap

Source: Calculations are based on the Census Numident, CJARS, CJARS residence and relations
crosswalk, and IRS Form 1040s.
Notes: The sample consists of children born in CJARS-covered states in 1999 and 2000 and remained
covered until age 18. The gap in cumulative exposure by a potential caregiver is measured between
minority and White children at the age of 18 for four types of criminal justice events: a criminal charge
(Panel A), felony charge (Panel B), felony conviction (Panel C), or incarceration (Panel D). Estimated
exposure is decomposed for each race using the county of birth, average Adjusted Gross Income (AGI)
reported on the IRS Form 1040 in the year of the child’s birth and next four years, and the number of
caregiver links. The population-weighted average of the two decompositions (White coefficients applied
to the minority population and minority coefficients applied to the White population) are reported for
the three observed characteristics.

44

Figure 7: Correlations between intergenerational exposure and child outcomes

I I

Non-exposed child mean: .144.

t !

I
I
I
I
f II
I

I

I~

I
~
;--g
--~
---1
2
2I

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

!

! !
.0154

~-----1---

I

I

~
2~

i

without controls

Other Caregiver
Prior Exposure

1

without controls

!
!!
t
!

I

I

f
----------

ii

~

!
I

~

with controls

I

! t !

g

~

f
I

t
0

□

y
Q

2

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

0

----------

I

beta
.015
.0148

~

f f

t+
~

\

without controls
\

.05

Other Caregiver
Contemporaneous

Bio-Parent
Prior Exposure

Non-exposed child mean: .036.

without controls

.01

t

! I I I rv i fl rf ~IJ yll

Bio-Parent
Contemporaneous

with controls

with controls

0

!

I

.0146

j j

.04

Other Caregiver
Prior Exposure

.03

Bio-Parent
Prior Exposure

.02

Other Caregiver
Contemporaneous

.04
.03
.02

Bio-Parent
Prior Exposure

D: Difficulty concentrating, remembering, or making decisions

~------------------ ---------- ~ --- ---- -Non-exposed child mean: .051.

.01

Other Caregiver
Contemporaneous

Non-exposed child mean: .028.

.0152

45

Bio-Parent
Contemporaneous

0

Bio-Parent
Contemporaneous

I

with controls

C: Behind age appropriate grade level

-.01

.2

Other Caregiver
Prior Exposure

.15

!

I

,!

Bio-Parent
Prior Exposure

0

.05

.1

.15

.2

.25

I

Other Caregiver
Contemporaneous

.1

Bio-Parent
Contemporaneous

B: Grandparent is primary caretaker

I

.05

.3

A: Household income below the poverty level

1

2
Any charge

■

Felony charge

sort

3
Felony conviction

4

• In prison

Other Caregiver
Prior Exposure

Figure 7: Correlations between intergenerational exposure and child outcomes (continued)
Bio-Parent
Contemporaneous

Other Caregiver
Contemporaneous

F: Teen parenthood

Bio-Parent
Prior Exposure

Other Caregiver
Prior Exposure

I

I I- §

T

,\

.

{

1

.J.

without controls

"

Bio-Parent
Contemporaneous

!
½

I

II

! t

!

Non-exposed child mean: .005.

L----------

Other Caregiver
Contemporaneous

I--I
I

V

!
½

! t

.002
Other Caregiver
Prior Exposure

.0015

Bio-Parent
Prior Exposure
without controls

l 2I I I I I
~
1
! !
g ~ 2
______
_J - ------- - - --------0

.0005

.001

1

with controls

0

beta
.015
.0148

£

I

y
2

£

I

£ £1

H: Death

.0146

Q

1

2
Any charge

■

Felony charge

-.0005

.0152

.04
.03
.02
.01

!!

child mean: .005.
-Non-exposed
-- -------

0

46

f ! t

with controls

I
I
I
I

with controls

Other Caregiver
Contemporaneous

! II II! II

without controls

t "'

£

Other Caregiver
Prior Exposure

Q
~
~
!:i
!:i
0
- - - - ------ -o-- - - - - - - - - -------- --

G: Charged with adult crime as teenager
Bio-Parent
Contemporaneous

Bio-Parent
Prior Exposure

0

I +.

..I

I ~
Iti .l ~t~ f1
.0154

.I ,I

-.01

0

.01

.02

------------------- ---------- ~--------Non-exposed child mean: .026.

.005 .01 .015 .02 .025 .03

E: High school dropout

sort

Bio-Parent
Contemporaneous

Other Caregiver
Contemporaneous

Bio-Parent
Prior Exposure

Other Caregiver
Prior Exposure

I

"1-lJ-- _1__1--t-t ---+
-¥ rtf-P
i ~ -4
f -

!I. f

Non-exposed child mean: .001.

t

~

3
Felony conviction

?

+

without controls

'"' I--... with controls

4

• In prison

Source: Estimates are based on the 2005–2018 American Community Survey (outcomes), the Census Numident (year of birth, sex, mortality), the Census BestRace Files
(race/ethnicity), CJARS (potential caregiver exposure and child adult charges) and the CJARS relationship crosswalk (identify potential caregivers and measure child fertility).
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. Estimates along with 95% confidence intervals are shown. All regressions with controls include
fixed effects for the county of birth, birth year, and race/ethnicity X gender along with a third-order polynomial for average adjusted gross income in the first five years of the
child’s life, as measured by IRS Form 1040. Standard errors are clustered by the commuting zone of birth. For Panels A through E, person weights provided by the American
Community Survey are used. The sample consists of individuals in the Census Numident 1999–2005 birth cohorts in CJARS-covered geographies from birth until X, where
X represents years since birth at the time of survey response (0–18) with the place of birth still covered or year 2018. Contemporaneous exposure to a charge is measured in
the year of the survey, and prior exposure is measured from birth until the year prior to the survey response. For Panels F through H, the sample consists of children born in
CJARS-covered states in 1999–2002, surviving until age 12. Contemporaneous exposure to any charge is defined for the teenage years (16–18) and prior exposure is from birth
through age 15. All results were approved for release by the U.S. Census Bureau, authorization number CBDRB-FY22-ERD002-009.

Figure 8: Charge exposure and child development coefficient heterogeneity, by race, sex, and household income
A: Grandparent is primary caretaker
Other Caregiver
Contemporaneous

Bio-Parent
Contemporaneous

Bio-Parent
Prior Exposure

,___

estimated coefficient

,r:::::::::---e

-

White, Non-Hisp.

-

non-exposed mean 95% C.I.

Black, Non-Hisp.
Hispanic

~

~

,:

>---❖

Female

>-----------3
1--------❖

l----0

>----------¢

Male

~

,______

==•

I

Asian
American Indian

Other Caregiver
Prior Exposure

>--------❖

I-------¢

>---,--

,_________

...__

>-------------<

>-----------9

0----------0

High Income

>----=O

Medium Income

~

>--------EO

Low Income

47

0

.05

>----------eQ

>----------,,¢

>-------=(>

.1

0

.15

.05

>-----------'OJ
~

t-----e<)

.1

.15

0

.05

.1

.15

0

.05

.1

.15

B: Difficulty concentrating, remembering, or making decisions

- -~
-

Bio-Parent
Contemporaneous

Other Caregiver
Contemporaneous

estimated coefficient

White, Non-Hisp.

non-exposed mean

Black, Non-Hisp.
Hispanic
Asian

,____

95% C.I.

>----¢

~

American Indian

--

>------¢

Female
High Income
Medium Income

◊-I

~

t-==O

~

-

~

.04

.06

0

.02

.04

.06

l----0

-

~

f-------B>
>---..e◊

.08

.02

.03

-

>--------o

~

>-,,<)

.08

-

-

--

~

◊

>-◊

E(>

.02

-

...._____

~

Other Caregiver
Prior Exposure

>----¢

f-0

Low Income
0

,_

>-¢

...._

Male

Bio-Parent
Prior Exposure

--

.04

.05

.06

>----------,-.¢

.07

0

.02

.04

.06

.08

Figure 8: Charge exposure and child development coefficient heterogeneity, by race, sex, and household income (continued)
C: Teen parenthood

Other Caregiver
Contemporaneous

Bio-Parent
Contemporaneous
estimated coefficient

White, Non-Hisp.

-

-

non-exposed mean 95% C.I.

Black, Non-Hisp.
Hispanic
Asian

~

American Indian

Bio-Parent
Prior Exposure

Other Caregiver
Prior Exposure

-

,__________

-

--------,______

~

f------------<>

.__

Male
Female
High Income

>---------=<>

Medium Income

t------------B>

~

Low Income

~

~

f----------COl

.005

.01

.015

.02

t------------ro

t----------'<>J

>-----------t()

0

1-----------iQJ

.025

0

.005

.01

.015

.02

.025

0

.005

.01

.015

.02

0

.005

.01

.015

.02

48

D: Charged with adult crime as teenager
Other Caregiver
Contemporaneous

Bio-Parent
Contemporaneous

-

estimated coefficient
~

White, Non-Hisp.

,'
non-exposed mean

"""!""!

95% C.I.

Black, Non-Hisp.
Hispanic

Bio-Parent
Prior Exposure

Other Caregiver
Prior Exposure

-

-

~

t..:1111

Asian
American Indian

-

Male
Female

~

-

~

I----------¢

-

High Income
Medium Income

>-------=<)

1-------9)

Low Income

~

0

.01

.02

.03

0

.005

.01

.015

.02

.025

0

.005

.01

.015

.02

0

.005

.01

.015

.02

Source: Estimates are based on the 2005–2018 American Community Survey (outcomes), the Census Numident (year of birth, sex, mortality), the Census BestRace Files
(race/ethnicity), CJARS (potential caregiver exposure and child adult charges) and the CJARS relationship crosswalk (identify potential caregivers and measure child fertility).
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. See Figure 7 notes for a description of outcomes, independent variables of interest, and the
regression specifications with controls. Outcome means for non-exposed children by subgroup are depicted by the black vertical line with point estimates added to the mean, shown
in diamonds, along with shaded 95% confidence intervals. All results were approved for release by the U.S. Census Bureau, authorization number CBDRB-FY22-ERD002-009.

Appendices
Appendix A

Supplementary results
Figure A1: Composition of sample

A: Birth year

B: Child’s race and sex

16%
14%

14%

14%

15%

15%

14%

14%

60%

60%

14%
51%
49%

50%

12%
10%

40%

8%
30%

6%
20%

4%

14%
11%

11%

10%

2%

3%

0%
1999

2000

2001

2002

2003

2004

1%

0%

2005

C: Average household AGI by percentile rank

White, Non- Black, NonHispanic
Hispanic

Hispanic

Asian

American
Indian

Other

Male

Female

D: Place of birth

$700,000
Washington

5%

$600,000

1%

Minnesota

3%

$500,000

Maine

North Dakota

Montana
Oregon
Idaho

Wisconsin

5%

South Dakota

$400,000

Nebraska

Pennsylvania

Iowa

10%

2%

Nevada

Ohio

Utah
California

Illinois

Indiana

Colorado
Kansas

$300,000

New York

Michigan

9%

Wyoming

5%

West
Virginia
Virginia

Missouri

Kentucky
North Carolina

8%

Tennessee

..

$200,000

$100,000

$-

----·····························

0

20

40

Arizona

■
■

6%

80

Arkansas

New Mexico

■

South Carolina

Mississippi

Alabama

Georgia

Texas

26%

···········

60

Oklahoma

Louisiana
Florida

Alaska

14%

100

Parent Household Income Rank

Powered by Bing
© GeoNames, Microsoft, TomTom

Source: Calculations are based on the Census Numident, the Census BestRace files, CJARS, and the CJARS
relationship crosswalk.
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. The sample consists of
individuals in the Census Numident 1999–2005 birth cohorts in CJARS-covered states during the year of
birth. Child’s year of birth, place of birth, and sex are measured using the Census Numident. Child’s race
is measured using the Census BestRace files. Average Adjusted Gross Income (AGI) is measured on the IRS
Form 1040 that the child is claimed on in their year of birth and the subsequent four years. AGI is reported
as zero if the child is not claimed. Children that are not claimed in their first five years of life or are ever
claimed on a form reporting negative AGI are dropped from the sample in Panel D. Panel D depicts average
AGI for children within percentile bins as rank-ordered within birth cohorts. All results were approved for
release by the U.S. Census Bureau, authorization number CBDRB-FY22-ERD002-009.

49

Figure A2: Life cycle fertility for females born in 1981 by race
A: Share with observed birth by age

90%
American Indian

80%
70%

Black, Non-Hispanic
Hispanic
White, Non-Hispanic

60%

Asian

50%
40%

Ov

era

ll
Average age at
first birth
Overall
25.0
White, Non-Hispanic
25.7
Black, Non-Hispanic
22.7
Hispanic
23.2
Asian
28.2
American Indian
22.9

30%
20%
10%
0%

≤ 14 15

Overall

16

17

18

19

20

White,
Non-Hispanic

21

22

23

24

25

26

Black,
Non-Hispanic

27

28

29

Hispanic

30

31

32

Asian

33

34

NVS
(2017)
24.9
25.9
22.3
22.2, 24.8
27.8
21.6
35

36

37

American Indian

B: Number of children
40%
35%

Overall
White, Non-Hispanic
Black, Non-Hispanic
Hispanic
Asian
American Indian

30%
25%

Average total
birth rate
1.67
1.66
1.74
1.82
1.39
2.05

NCHS
(2016)
1.77
1.67
1.83
2.01
-

5

6+

20%
15%
10%
5%
0%

0

1

■ Overall ■ White,

Non-Hispanic

2
II Black,

3

4

Ill Hispanic

Iii]

Asian

□

American Indian

Non-Hispanic

Source: Calculations are based on the Census Numident, the Census BestRace files, and the CJARS relationship crosswalk.
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. The sample consists of
females in the 1981 birth cohort in all states, as measured by the Census Numident. Race is measured using
the Census BestRace files. Individuals are linked to the CJARS family crosswalk to measure fertility and age
of first birth, defined as identifying a biological child and determining age at birth based on the year of birth
of the child. All results were approved for release by the U.S. Census Bureau, authorization number CBDRBFY22-ERD002-009. These estimates can be compared to national public statistics on fertility. For Panel A,
2016 National Center for Health Statistics for age at first birth in 2000: overall 24.9, White 25.9, Black 22.3,
Hispanic-Mexican 22.2 and Central/South American 24.8, Asian 27.8, and American Indian/Alaska Native
21.6 (Mathews and Hamilton 2016). For Panel B, National Vital Statistics 2017 report birth rates: overall
1.766, White 1.666, Black 1.825, and Hispanic 2.01 (Mathews and Hamilton 2019).

50

Figure A3: Distribution of total potential caregiver links identified per child

A: All children, all potential caregivers

B: By child’s race, all potential caregivers

50%

50%

45%

45%

40%

40%

37%

35%

35%

30%

30%

25%

25%

20%

20%

18%

9%

10%

7%

5%
0%

15%

13%

15%

3%

1%

10%
4%

3%

2%

8

9

5%

3%

0%

0

1

2

3

4

5

6

7

0

10+

1

2

3

Children
•White
Asian Children

C: All children, male potential caregivers
49%

5

6

Children
•Black
American Indian Children

D

50%

4

7

8

9

10+

Total Identified Potential Caregiver Links

Total Identified Potential Caregiver Links

D Hispanic Children

D

D: All children, female potential caregivers
49%

50%

40%

40%

30%

30%

25%

23%

20%

20%

12%

12%

10%

6%

5%

10%
5%

6%

5%

4

5+

3%

0%

0%
0

1

2

3

4

0

5+

1

2

3

Total Identified Female Potential Caregiver Links

Total Identified Male Potential Caregiver Links

Source: Calculations are based on the Census Numident, the Census BestRace files, and the CJARS relationship crosswalk.
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. The sample consists of
individuals in the Census Numident 1999–2005 birth cohorts in all states. Child’s race is measured using
the Census BestRace files. Potential caregivers are defined as biological parents, stepparents, adopted parents, foster parents, unclassified caregivers, grandparents, aunts/uncles, non-familial adult (cohabiting 2+
years), and unclassified adult (cohabiting 2+ years). All results were approved for release by the U.S. Census
Bureau, authorization numbers CBDRB-FY22-ERD002-001 and CBDRB-FY22-ERD002-003.

51

Figure A4: Heterogeneous cumulative exposure by biological parents
A: By child’s race

B: By biological parents’ sex

45%

30%

42%
40%

40%

26%

35%

Male bio-parent exposure

30%
25%

20%

Female bio-parent exposure

18%

26%
23%
21%

20%

9%

5%

6%

7%
4%

White,
Black,
Non-Hispanic Non-Hispanic

•Felony charge

9%
8%

10%

7%
3%

■ Charge

11%

10%

12%
10%
8%

13%

16%

15%
10%

15%

20%

16%

0%

Any bio-parent exposure

25%

3%

Hispanic

□ Felony

2%

4%

4%

3%

1%

1%

Asian

conviction

5%

5%
0%

American
Indian

□ In

Charge

Felony charge

Felony conviction

In prison

prison

Source: Calculations are based on the Census Numident, the Census BestRace files, CJARS, and the
CJARS relationship crosswalk.
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. The sample consists
of individuals in the Census Numident 1999–2005 birth cohorts in CJARS-covered geographies from
birth until X, where X represents years since birth (0–18) with the place of birth still covered or
year 2018. All results were approved for release by the U.S. Census Bureau, authorization number
CBDRB-FY22-ERD002-001.

Figure A5: Heterogeneous cumulative exposure by child race and adult sex
A: Male potential caregivers

B: Female potential caregivers

60%

60%
54%

52%

50%

50%
41%

40%
30%

40%

35%
30%

28%

20%

18%

0%

23%
20%

White,
Non-Hispanic

20%

6%

Black,
Non-Hispanic

Hispanic

30%

10%
4%

Asian

2%

19%

18%
14%

13%

11%
5%

35%

24%

15%

13%
11%

10%

29%

36%

15%
11%

13%
8%

5% 4%

4%
1%

0%

American
Indian

■ Charge ■ Felony charge □ Felony conviction □ In prison

White,
Non-Hispanic

■ Charge

Black,
Non-Hispanic

■ Felony

charge

6%

5%
2%

Hispanic

□ Felony

4%

1% 1%
0%
Asian

conviction

American
Indian

□ In

prison

Source: Calculations are based on the Census Numident, the Census BestRace files, CJARS, and the
CJARS relationship crosswalk.
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. The sample consists
of individuals in the Census Numident 1999–2005 birth cohorts in CJARS-covered geographies from
birth until X, where X represents years since birth (0–18) with the place of birth still covered or
year 2018. All results were approved for release by the U.S. Census Bureau, authorization number
CBDRB-FY22-ERD002-001.

52

Figure A6: Cumulative exposure to the criminal justice system, by child’s race
A: White, Non-Hispanic

B: Black, Non-Hispanic

60%

60%

50%

50%

40%

40%

30%

30%

20%

,,.,,,.-

-------------------------·

10%

,,

,,
-,,
-,,,,

---------------------

20%
10%

0%

0%
0

3

6

9
12
Child Age

15

18

0

3

C: Hispanic

6

9
12
Child Age

15

18

D: Asian

60%

40%
60%

50%

50%

40%

40%

30%

30%

20%

20%

10%

10%

0%

0%

35%
30%
25%
20%
15%
10%

0

3

6

9
12
Child Age

15

5%

18

r ·········
,.,_,r-crn- =-----= rs====-------

-

0

3

6

18

0

3

6

0%

E: American Indian/Alaska Native

1-

60%

Charge

50%

9
12
Child Age
9

Legend
Child Age

12

Felony charge

Felony conviction

Bio-parents

With caregivers

15
15

In prison

40%
30%
20%
10%
0%
0

3

6

9
12
Child Age

15

18

Source: Calculations are based on the Census Numident, the Census BestRace files, CJARS, and the CJARS relationship
crosswalk.
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. The sample consists of individuals in
the Census Numident 1999–2005 birth cohorts in CJARS-covered geographies from birth until X, where X represents
years since birth (0–18) with the place of birth still covered or year 2018. All results were approved for release by the
U.S. Census Bureau, authorization number CBDRB-FY22-ERD002-001.

53

18

Figure A7: Intensive margin of criminal justice exposure by parents and other potential
caregivers by age 18, by child’s race
A: White, non-Hispanic

B: Black, non-Hispanic

70%

70%

60%

60%

50%

50%

40%

40%

30%

30%

20%

20%

10%

10%

0%

0%
1

2

3
4
5
6
7
8
Total Number of Exposure Events

9

10+

Charge
■ Felony Charge
C: Hispanic
Felony Conviction
Prison Spells
Years of Prison

1

■ Any

70%

2

3
4
5
6
7
8
Total Number of Exposure Events

9

10+

9

10+

■ Any

Charge
■ Felony Charge
D: Asian
Felony Conviction
Prison Spells
Years of Prison

70%

60%

60%

70%

50%

50%

60%

40%

40%

50%

30%

30%

40%

20%

20%

30%

10%

10%

20%
0%

0%
1

2

3
4
5
6
7
8
Total Number of Exposure Events

9

10+

1

0%

■ Any

E:

10%

Charge
■ Felony Charge
Felony Conviction Prison Spells
Years of Prison
American
Indian/Alaska Native

1

2

3
4
5
6
7
8
Total Number of Exposure Events

■ Any

Charge
Felony Charge
3
4
5■ 6
7
8
Felony Conviction Prison Spells
Total Number of Exposure Events
Years of Prison
Legend

2

9

10+

Any Charge
Felony Charge
Felony
Conviction
□
□ Prison Spells
[] Years of Prison

70%
60%
50%
40%

I

30%
20%

Median intensive margin
experience conditional on
extensive margin exposure

10%
0%
1

2

3
4
5
6
7
8
Total Number of Exposure Events

9

10+

■ Any Charge
■ Felony Charge
Source: Calculations
are basedPrison
on the
Census Numident, the Census BestRace files, the CJARS relationship crosswalk,
Felony Conviction
Spells
and CJARS.Years of Prison
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. The sample consists of individuals in
the Census Numident 1999–2000 birth cohorts in CJARS-covered geographies from birth until age 18. Distinct events are
counted among children with any exposure. Thus, multiple charges filed on the same date are considered one event and
similarly for the other types of criminal justice events. The number of events are truncated at 10+ events for all events
except prison spells, which are top coded at 8+. Asian felony convictions and prison spells are top coded at 9+ and 5+
to preserve confidentiality, respectively. All results were approved for release by the U.S. Census Bureau, authorization
numbers CBDRB-FY22-ERD002-003 and CBDRB-FY22-ERD002-009.

54

Figure A8: Exposure heterogeneity by parent income rank, by child’s race
A: Any charge

B: Felony charge

60%
80%

White, Non-Hispanic
Black, Non-Hispanic
Hispanic
Asian
American
Indian _
-o-:_
_____

60%

45%

......

0 - -O--o--o---+

--

0----

0 - -0 - -<>--<>--<>--

20%

0

20

40

0----

15%

60

80

100

0%

0

20

40

Parent Household Income Rank

55

C: Felony conviction

80

100

White, Non-Hispanic
Black, Non-Hispanic
Hispanic
Asian
American Indian

•
23%

30%

15%

15%

8%

D: In prison

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

30%

45%

60

Parent Household Income Rank

60%

0%

Black, Non-Hispanic
Hispanic
Asian
-0- American Indian

•
0----

30%

40%

0%

• White, Non-Hispanic

0----

. .... .... .
•

• White, Non-Hispanic
• Black, Non-Hispanic

Hispanic
Asian
-0- American Indian

0%
0

20

40

60

Parent Household Income Rank

80

100

0

20

40
60
Parent Household Income Rank

80

100

Source: Calculations are based on the Census Numident, the Census BestRace files, the CJARS relationship crosswalk, CJARS, and IRS Form 1040s.
Notes: Estimates and sample sizes have been rounded to preserve confidentiality. The sample consists of individuals in the Census Numident born in 1999 and 2000
in CJARS-covered geographies from birth until age 18. Average exposure by age 18 is depicted for children across income percentile bins. Some bins, marked with
horizontal black lines, are wider to satisfy disclosure requirements. Income percentile bins are determined using the average adjusted gross income reported on IRS
Form 1040s, in which the child is claimed for the first five years. Children claimed on a form with negative AGI or never claimed in the first five years are not included
in the sample. All results were approved for release by the U.S. Census Bureau, authorization number CBDRB-FY22-ERD002-003.

Appendix B

Data appendix

Constructing the residency and relations crosswalks
The crosswalks are currently created for all individuals in the Census Numident with a valid
birth year and born between 1960 and 2018.46 The Census Numident is the “backbone” of
the residence crosswalks, setting the population and identifying date and place of birth.47
Address-level information is then harmonized for all subsequent years based on the 2000
and 2010 decennial censuses, American Community Survey (2001–2018),48 IRS Form 1040
tax filings (1969, 1974, 1979, 1984, 1989, 1994, 1995, 1998–2018 tax years),49,50 IRS Form
1040 electronic tax filings (2005, 2008–2012),51 Department of Housing and Urban Development (HUD) program data (Longitudinal PIC/TRACS: 1995–2016, 2018; PIC: 2000–2014;
TRACS: 2000–2014)52 and county-level information from Medicare (2000–2017 EBD) and
Medicaid (2000–2014 MSIS) enrollment databases,53 Indian Health Service (IHS) from 1999–
2017, and the MAF-ARF (2000–2018). Data are linked at the person-level using a Protected
Identification Key (PIK) created through the Census Bureau’s Person Identification Validation System (PVS).54 Similarly, addresses are assigned MAFIDs, a numeric key, to protect
PII. If more than one MAFID (i.e., address) is provided for an individual in a given year, the
following ranking is applied: decennial census, IRS Form 1040, IRS Form 1040 ELF, Amer46

The Census Numident is sourced from the Social Security Administration (SSA) Numident file, which
tracks all events related to Social Security Numbers (SSN) and Individual Taxpayer Identification Numbers
(ITINs) including applications, changes, and deaths. The Census Numident is a research file that de-identifies
the information by assigning a unique Protected Identification Key (PIK) for all SSNs and ITINs. For further
explanation of these files, see Genadek, Sanders, and Stevenson (2021).
47
We use a place of birth crosswalk which links unique place names and states to county and state FIPS
codes.
48
There are small samples during the survey trial years between 2001 and 2004, so while statistical information is not used for these years, respondents’ address-level information and household relationships are
used in the creation of these crosswalks.
49
Information varies by tax year and includes only the primary tax filer in 1969, primary and secondary
filer in 1974–1989 and both filers and four dependents for the remaining tax years.
50
IRS 1040s are based on tax years (TY). For example, the TY2009 1040s are filed in 2010 and thus
provide an accurate address for individuals in the filing year, which is 2010. This was confirmed by comparing
consistency with the 2010 Decennial Census.
51
The electronic tax files allow information for up to 20 dependents.
52
The longitudinal file is created by Census for research use combining information from the Public and
Indian Housing Information Center (PIC) and Tenant Rental Assistance Certification system (TRACS).
53
CMS MSIS only provides county-level geographic information.
54
PIKs are used to link data at the person level within the Census Bureau’s Data Linkage Infrastructure.
PIKs can be assigned deterministically using only SSN or probabilistically using names, dates of birth,
addresses, and other information as inputs into the Person Identification Validation System (PVS). For
further explanation of this process, see Wagner and Lane (2014).

56

ican Community Survey, CMS EDB, HUD Longitudinal PIC/TRACS, HUD PIC, HUD
TRACS, IHS, MAF-ARF, CMS MSIS.
The residence crosswalks are the basis of the familial crosswalks. First, for each year,
all coresidence pairs are created at a given address. Group quarters and addresses with
more than 20 individuals identified at the locations in a given year are suspect of not having familiar relations and thus not used to create pairwise relations among all cohabitants.
This should not impact individuals living in apartment buildings, since individual units are
assigned unique address identification numbers. Instead, examples of group quarters include
dormitory facilities, assisted living facilities, nursing homes, and prisons. Relationships are
enhanced above cohabitation based on information in the 2000 and 2010 decennial censuses,
American Community Survey (2001–2018), IRS Form 1040 tax filings (1969, 1974, 1979,
1984, 1989, 1994, 1995, 1998–2018 tax years),55,56,57 IRS Form 1040 electronic tax filings
(2005, 2008–2012),58 and HUD program data (1995–2018, 2018). The Decennial Censuses,
American Community Surveys, and HUD program data each provide relationships between
the household head and other household members, which is used to directly establish relationship types and infer the relationship between household members. Additionally, tax
filing and claiming behavior establish spousal relationships and dependents. Finally, we include the Census Household Composition Key (CHCK) which creates links between children
and parents based on information on birth certificates for children born between 1999–2018
(Luque and Wagner 2015).59 Children that do not have parental links established by the
CHCK file could be due to the father’s or mother’s information being left off of the birth certificate, inaccurate parent information, the parent not being assigned a PIK (SSN or ITIN),
or an inability to match the child-parent pair to the same address to confirm the link (Luque
and Wagner 2015; Genadek, Sanders, and Stevenson 2021; B. Bond et al. 2014). We confirm
that 93% of the biological relationships identified by our crosswalk are also observed in the
55
Information varies by tax year and includes only the primary tax filer in 1969, primary and secondary
filer in 1974–1989 and both filers and four dependents for the remaining tax years.
56
Again, IRS 1040s are based on tax years (TY). For example, the TY2009 1040s are filed in 2010 and thus
provide an accurate address for individuals in the filing year, which is 2010. Relations are also established
for the year 2010 when they are observed coresiding at a specific location to match the residence crosswalks,
although officially the relation existed in 2009.
57
Individuals in the 1040s receive a PIK deterministically from their SSN or ITIN. However, the address
information used to determine the associated MAFID is probabilistic. Thus, we are able to make some
relational pairs using filing information even if we can not match them to a specific MAFID in the residence
crosswalk.
58
The electronic tax files allow information for up to 20 dependents.
59
The Census Bureau uses parents’ names from birth certificates to probabilistically assign PIKs through
the PVS and the child’s SSN which uniquely determines the PIK to match to children in the Census Numident
ages 0–18 as of 2018 and 2019. Since parents’ SSNs are not available, the CHCK file requires that the childparent link be confirmed at the same address in the PVS reference file.

57

CHCK file. Additionally, 70% of all CHCK relations are confirmed as biological parents by
survey microdata or HUD program data, which increases to 97% once unclassified caregivers
(those observed claiming a child on a 1040 tax form accompanied by no other observable
information) are reclassified as biological parents.
Relationship types are established by combining the multitude of observations between
pairs across data sources and years into the following set:
Table B1: Relationship types in CJARS crosswalks
Relation Code
Relation Description
10
Cohabiting adults (=<13 years apart)
11
Spouse
12
Domestic partner
13
Romantic unmarried (e.g., boyfriend/girlfriend)
14
Unclassified romantic
15
Adult, non-romantic (e.g., roommate/boarder)
20
Cohabiting adult-minor (>13 years apart)
21
Bio parent - child
22
Adopted parent - child
23
Stepparent - child
24
Foster parent - child
25
Unclassified parent - child
26
Parent - child-in-law
27
Grandparent-grandchild
28
Aunt/uncle-niece/nephew
29
Non-familial adult - child
40
Cohabiting minors (=<13 years apart)
41
Bio-siblings
42
Adopted-siblings
43
Step-siblings
44
Foster-siblings
45
Unclassified siblings
45
Cousins
45
Second cousins
45
Siblings-in-law
Many relations can only be classified into the main category without the additional detailed information required to classify the relationship pair among the subcategories, for
several reasons. First, relationships in the decennial censuses, American Community Survey,
and HUD program data are all expressed in relation to the household head. Several assumptions are imposed to infer relations between other household members, but often there is

58

not enough information to classify a link beyond an unclassified parent-child link.60,61 Second, the American Community Survey between 2001–2007 and 2010 Decennial Census uses
broader relationship definitions.62 Finally, parental relations that are established only in the
tax records and not observed in the Census surveys or HUD program data (or are observed
with ambiguous relationships) can only be classified as a parent with no further information
to sub-classify into biological, step, adopted, foster, aunt/uncle, etc.63
A relation pair may be observed multiple times across source files and years. Relationship
types only need to be defined once in order to assign it to the cohabiting pair. This approach
is beneficial since individuals may be observed multiple years in the tax records, but without
detailed relational information and may be observed only once by the Decennial Census or
ACS.
Relationship types between pairs are sequentially established based on the strength of
the source information. For example, relations established in Census with the head of household directly define a relationship, while relations between other household members are
inferred. Figure B1 demonstrates the iterative process and assumptions used when defining
relationship types.
Performance of residential and relationship crosswalks
First, we benchmark our fertility statistics, as measured in the CJARS family crosswalk,
to published statistics in the 2016 National Center for Health Statistics and 2017 National
Vital Statistics System reports (Mathews and Hamilton 2016, 2019). In Figure A2 Panel A,
we show the cumulative distribution of age at first birth by race for females born in the U.S.
60

Some of the assumptions imposed to define parent-child relations (and vice versa): 1). if a household
head has a biological child, then the spouse to the household head is also linked as a biological parent, 2).
if a household head has a stepchild, then the spouse to the household head is assumed to be the biological
parent, 3). if a household head has an adopted child, then the spouse to the household head is assumed to
be an adopted parent, and 4). if a household head has a foster child, then the spouse to the household head
is assumed to be the foster parent as well.
61
Examples of inferred parent-child links where subclassification can not be ascertained: 1). a household
head linked to their parent is not further classified as biological, step, adopted, or foster 2). a spouse of a
household head linked to the household head’s mother/father-in-law is not further classified as biological,
step, adopted or foster 3). a sibling of the household head linked to the niece/nephew of a household head,
subject to age restrictions and other information if available 4). a child of a household head linked to a
grandchild of the household head, subject to age restrictions and other information if available.
62
The following relationship types to the household head were removed from the 2010 Decennial Census:
sibling-in-law, nephew/niece, uncle/aunt, cousin, grandparent, and foster child. The ACS did not offer more
detailed classifications for parent-child links (namely, biological, adopted, or step) or in-law links (parent or
child) until 2008.
63
A parent-child relationship is assumed if the age difference between the individuals is less than 45 years
and a grandparent-child relation if the age gap is 45 years or greater. However, it is possible for an aunt or
uncle to claim a child on tax records, although treating them as an unclassified parent in this circumstance
is likely permissible for our application.

59

in 1981, as measured by the Census Numident.64,65 The average age of first birth overall
is 25.01 based on our crosswalks, which is inline with an overall age of first birth in 2000
and 2014 of 24.9 and 26.3, respectively (Mathews and Hamilton 2016). Asian women have
the highest average age at first birth (our estimate 28.15; NCHS in 2000: 27.8), followed
by White (25.67; 25.9), Hispanic (23.23; Mexican 22.2, Central and South American 24.8),
American Indian/Alaska Native (22.85; 21.6), and Black (22.67; 22.3) women. In Panel B,
we show the number of births observed until the age of 37, the latest possible age for this
cohort. Again, our observed birth rates for women born in 1981 overall (1.673) are inline with
the reported birth rate by NVSS in (1.766). American Indian/Alaska Native women have
the highest number of observed births (our estimate: 2.047, NVSS not available), followed
by Hispanic (1.823; 2.01), Black (1.744; 1.825), White (1.655; 1.666), and Asian (1.39; not
available) women.
Next, we turn to our ability to measure caregiver links for children in our core sample.
Using the previously described crosswalks, we identify female (male) biological parents for
90% (76%) of children born between 1999 and 2005 (Figure 1A). When we relax the requirement for an explicitly defined relationship status and expand to include other types of adult
household members, we observe 97% and 95% of children are linked with one or more female
and male adult potential caregivers, respectively.66
There are important differences by race, due to both systematic differences in household
structure and in our ability to observe potential caregivers in administrative and survey
records. First, we see that only 55% of Black, non-Hispanic children are connected to a
biological male parent in the data. This could be due to either fathers being excluded from
birth records, not coresiding with children during household surveys, or not claiming their
children as dependents in tax filings. However, White, non-Hispanic and Black, non-Hispanic
children have very similar rates of being linked to a female biological parent (≥94%). Second,
Hispanic children are much less likely to be observed with a female biological parent (75%)
as well as male biological parents (60%), and just less than 90% are observed with a male
and female potential caregiver; this is likely the result of these individuals being less likely
to have a Social Security Number (SSN) or Individual Tax Identification Number (ITIN),
which is needed for individuals to receive a PIK and be linked across data sets within the
Census Bureau’s Data Linkage Infrastructure (B. Bond et al. 2014).
64
Due to data availability, we do not observe fertility beyond age 37 for this cohort and have more limited
ability to observe birth prior to age 18 since the CHCK file covers children starting in 1999.
65
Observed births are measured in our crosswalks as a reported biological child and age at birth is defined
using the year of birth for the child.
66
Figure A3 documents the distribution of number of links identified overall (Panel A) and by racial and
ethnic subgroup (Panel B).

60

Figure 1B documents the share of children born between 1999 and 2005 observed with
other types of intergenerational relationships in the household. We find that 4.7%, 22%,
29%, and 46.2% of children are observed with a step/adopted/foster parent, extended family
(grandparent/aunt/uncle), unclassified caregivers, and unclassified cohabiting adults, respectively.67 Unclassified caregivers, which occur when we observe an adult-child link resulting of
dependency claims in IRS tax records without any other information to pin down the nature
of their relationship, are more likely in intergenerational households,68 and among biological
parents that are left off of birth records (as proxied by the CHCK file). Importantly, all
these relations seem like important parental figures given the tax filing relationship even if
they are not included in the CHCK, which is our closest approximation to a birth record
database. Unclassified adults, in contrast, are other cohabiting relations that are either explicitly classified as non-familial in Census Bureau surveys or just adults that we observe
coresiding at the same address as the child (e.g., live-in boyfriends, roommates, etc); notably,
we require that these relations must coreside for 2 years or more in order to focus on those
with greater potential familial attachment.69
Again, there are important differences in the share of children that are observed with
various caregiver relations by race. Minority children are much more likely to have an extended family caregiver (Black, non-Hispanic 38%, Hispanic 28%, Asian 32%, and American
Indian/Alaska Native 31%) than their White, non-Hispanic counterparts (17%); this is consistent with previously documented differences in household structures by race and ethnicity
in the U.S. (e.g., Lofquist 2012; Cohen and Passel 2018).
Approximately 48.6% (49.3%) of children are linked with just one female (male) potential
caregiver, a grouping that combines biological parents and other potential adult caregivers
in their household. 24.6% (23.3%) of children are observed with two female (male) potential
caregivers, while 23.7% (22.8%) are linked with 3 or more female (male) potential caregivers.
67

Fewer children are observed with step/adopted/foster parents than what is observed in the SIPP
(Sweeney 2010; Kreider and Ellis 2011; Raley and Sweeney 2020); this is likely due to relationship misclassification into biological parents and unclassified caregivers due to relations being reported to the household
head in the Census surveys and HUD program data and a lack of relational information beyond claiming
behavior on tax forms. For example, if a household head has a biological child, then it is assumed they share
the biological child with their spouse. However, the relation could be a step or adopted parent. Similarly,
if a child is claimed by someone with an age gap less than 45 years and the relation is not observed in
the CHCK file, then the relation is considered an unclassified caregiver. However, the relation could be a
step/adopted/foster parent, an aunt or uncle, or a younger grandparent.
68
Household surveys enumerate relationships of all individuals in the household with respect to the head
of household. Relationships between other individuals must be inferred, which is increasingly complicated
in households that extend beyond nuclear families.
69
Utilizing a 2-year coresidency requirement also minimizes the influence of errors in the probabilistic
record linking process for address information that might lead children to be labeled as coresiding with
adults who in fact do not live at the same address.

61

A variety of living circumstances could give rise to more than one potential caregiver of the
same sex being linked with a child, for example: (1) parents with multiple romantic partners
(due to divorce or separation) while raising their children, (2) households with same-sex
romantic partners, (3) multigenerational households, or (4) doubled up households where
multiple families share the same accommodations. The high rate of children linked with 2
or more potential caregivers of the same sex reflects the experience of many children today
in the U.S. of growing up with multiple adult influences in their households beyond the
traditional nuclear family.
Figure B1: Sequential process to establish relations beyond cohabitation
I

I

I
+------

7 7
62

Figure B2: Construction of residential and relations crosswalks
Data

Data
Decennial
Census
2000, 2010
HUD TRACS
2000–2018
CMS EDB
2000–2019
MAF-ARF
2000–2019

IRS Form 1040
1969, 1974,
1979,
1984, 1989,
1994, 1995,
1998–2018

Residential
Crosswalk
1969–2018

ACS
2001–2018

DC
I
I
I

HUD PIC
2000–2018

CMS MSIS
2000–2014

HUD Longitudinal
PIC/TRACS
1995–2018

I
I

I

IRS Form 1040
1969, XXXX,
1999–2018

D
Clean

Clean

Merge

+

ACS
2001–2018

+

+

PIK-PIK-year relations from all source files
Unit of observation=PIK-PIK-year; years=1

Combine all associated MAFIDs across source files in
each year
Unit of observation=PIK-year-MAFID; years=1

Harmonize

+

+

Create relations based on panel with all source files
Unit of observation=PIK (in cohort)-PIK-year;
years=all

All associated MAFIDs are in long form, PIK-yearMAFID, rank MAFIDs, drop MAFIDs beyond two,
reshape wide, save mergedwide_US_year
Unit of observation=PIK-year; years=1
Compile

HUD Longitudinal
PIC/TRACS
1995–2018

Link to Numident year of birth, drop GQ observations, make all pairwise comparisons

Deduplicate, reshape IRS files, link to MAF-X for
state and county

Harmonize

Decennial
Census
2000, 2010

Census Household Composition Key
2018–2019

Indian Health
Service
2000–2019

i

Merge

I ID
101 I
Census
Numident
2019

Collapse

+

Unit of observation=PIK (in cohort)-year; years=all

+

i

Use Numident cohorts to build balanced panel linking in residences for available years
Unit of observation=PIK (in cohort)-year; years=all

Final relations crosswalks
Unit of observation=PIK (in cohort)-PIK-year;
years=all

+

Final residence crosswalks
Unit of observation=PIK (in cohort)-year; years=all

63

Figure B3: State comparisons, crime and criminal justice outcomes

B: Violent Crime Rates (UCR)

CJARS States
CJARS Pop. Wgt Mean

Non-CJARS States
Non-CJARS Pop. Wgt Mean

Avg Imprisonment Rate 2000-2018
200
400
600
800

C: Incarceration Rates (BJS)

0

500000
Avg Family Income 2006-2018
300000
400000

States (in ascending order)

200000

SD NY NH ND NJ MA VT PA ID ME WV CT VA KY WI IA RI WY MI MN IL MT CA NE IN MS CO MD DE OH AK NV KS UT MO OK NC AL OR AR GA FL TN TX LA NM HI SC WA AZ

Avg Violent Crime Rate 2000-2018
200
400
600

Avg Property Crime Rate 2000-2018
2000
2500
3000
3500
4000

800

A: Property Crime Rates (UCR)

ME VT NH ND UT WY ID VA KY RI MN HI WI SD CT OR IA NE MS WV NJ WA OH MT CO IN PA KS NC NY GA MA AL AZ TX OK CA IL MI MO AR DE MD FL LA NV SC AK NM TN

States (in ascending order)
CJARS States
CJARS Pop. Wgt Mean

Legend

Non-CJARS States
Non-CJARS Pop. Wgt Mean

NV UT ID NM AZ FL WA OR CO TX CA TN MI AL NJ IL WV WY MN GA MD WI OH NC AR LA ME MT DE KY IN KS NE MO SC OK MS NY VA NH HI IA PA CT AK ND VT MA SD RI

States (in ascending order)

1-"

0

CJARS States
CJARS Pop. Wgt Mean

0

Non-CJARS States
Non-CJARS Pop. Wgt Mean

ME MN RI MA NH ND UT VT NE WA NJ IA HI NY NM KS WV IL NC CT OR AK PA MT MD WI WY IN CA CO SD TN OH MI VA DE KY ID NV SC FL MO AR GA AZ AL TX OK MS LA

States (in ascending order)

Notes: DataCJARS
come
from the FederalNon-CJARS
BureauStates
of Investigation’s Uniform Crime Reporting (UCR) program
States
CJARS Pop.
Mean
Non-CJARS
Pop. Wgt Mean
and the Bureau
of Wgt
Justice
Statistics
National
Prisoner Statistics program. The rates per 100,000
residents have been averaged for each state over the period 2000–2018. Marker sizes are proportional
to each state’s population, averaged over the years 2000–2018.

64

Figure B4: State comparisons, socioeconomic outcomes

B: Percent Population Black

Avg Percent Black 2006-2018
10
20
30

0

0

40

Avg Percent White 2006-2018
60
80

40

100

A: Percent Population White

HI MS MD GA LA CA NY SC AL NJ NC VA DE NV AK IL NM TX FL AR CT TN OK AZ MI WA MA RI PA OH MO IN CO MN SD KS WI OR KY NE UT ND MT IA WY ID WV NH ME VT

CJARS States
CJARS Pop. Wgt Mean

Non-CJARS States
Non-CJARS Pop. Wgt Mean

25

Non-CJARS States
Non-CJARS Pop. Wgt Mean

500000

E: Family Total Income

200000

CJARS States
CJARS Pop. Wgt Mean

10

500000
Avg Family Income 2006-2018
300000
400000

0

States (in ascending order)

0

Avg Percent in Poverty 2006-2018
15
20

Avg Percent Not Citizen 2006-2018
5
10

15

D: Percent In Poverty

WV MT MS ME SD VT ND OH WY AL LA MO KY WI NH IA SC PA IN AR TN MI AK OK ID MN NE KS DE NC UT OR VA GA CO NM RI CT WA MD IL MA HI AZ FL NJ NY NV TX CA

Avg Family Income 2006-2018
300000
400000

~
States (in ascending order)

Non-CJARS States
Non-CJARS Pop. Wgt Mean

C: Percent Not A Citizen

200000

~

MT ID VT WY UT ME NH SD ND OR NM HI IA WV AK CO WA AZ NE MN CA WI KS RI MA OK KY NV IN CT PA MO TX OH NJ MI IL AR FL NY TN VA NC DE AL SC MD GA LA MS

States (in ascending order)
CJARS States
CJARS Pop. Wgt Mean

0

NH MD NJ HI AK CT MN UT WY VA CO MA WA VT WI ND DE NE IA NV KS IL ME PA ID RI OR CA SD MT FL IN NY MO OH MI AZ NC TX GA TN OK SC AL WV AR KY LA NM MS

States (in ascending order)
CJARS States
CJARS Pop. Wgt Mean

Legend

Non-CJARS States
Non-CJARS Pop. Wgt Mean

NV UT ID NM AZ FL WA OR CO TX CA TN MI AL NJ IL WV WY MN GA MD WI OH NC AR LA ME MT DE KY IN KS NE MO SC OK MS NY VA NH HI IA PA CT AK ND VT MA SD RI

States (in ascending order)
CJARS States

oc£f

CJARS Pop. Wgt
Mean
'----I
- "_
_
o_

Non-CJARS States
Non-CJARS _
Pop.
Wgt Mean
___J

~

~

0

NV UT ID NM AZ FL WA OR CO TX CA TN MI AL NJ IL WV WY MN GA MD WI OH NC AR LA ME MT DE KY IN KS NE MO SC OK MS NY VA NH HI IA PA CT AK ND VT MA SD RI

States (in ascending order)

Notes: DataCJARS
come
from the IPUMS
USA 2006–2018
ACS, with state averages reported and weighted by
States
Non-CJARS
States
CJARS Pop. Wgt Mean
Pop. Wgt
Mean (Ruggles et al. 2021). Marker sizes are proportional
the average population
during the Non-CJARS
same time
period
to each state’s population averaged over the years 2000–2018.

65

Table B2: Source files contributing to the residence crosswalk
Source

Years

MAF-X

2017

Variables

MAFID, state, county, group quarters flag
IRS Form 1040
1969
primary filer, MAFID, state
—
1974, 1979, 1984,1989 primary and secondary filer, MAFID,
state
—
1994, 1995, 1998–2019 primary and secondary filer, four dependents, MAFID, state
IRS Form 1040 ELF
2005, 2008–2012
primary and secondary filer, 20 dependents
Decennial Census
2000, 2010
household members, MAFID, state,
county, group quarters flag
ACS
2001–2004†
household members, state, county
—
2005–2018
household members, state, county,
MAFID, group quarters flag
HUD Longitudinal PIC/TRACS
1995–2016, 2018
household members of enrollees,
state, county, MAFID
HUD PIC
2000–2014
household members of enrollees,
state, county, MAFID
HUD TRACS
2000–2014
household members of enrollees,
state, county, MAFID
CMS EDB
2000–2019
enrollees, state, county, MAFID
CMS MSIS
2000–2014
enrollees, state, county
Indian Health Service
1999–2019
enrollees, state, county, MAFID
MAF-ARF
2000–2018
individuals with SSN or ITIN, state,
county, MAFID
Census Numident
2021Q1
individuals with SSN or ITIN, place
of birth (linked to state, county, and
commuting zone), date of birth, date
of death, sex, race
There are small samples during the ACS trial years between 2001 and 2004, so while statistical information is not
used for these years, respondents’ address-level information and household relationships are used in the creation
of these crosswalks.
†

66

Table B3: Source files contributing to the relations crosswalk
Source
Residential Crosswalk

IRS Form 1040
—
IRS Form 1040 ELF
Decennial Census
ACS
HUD Longitudinal PIC/TRACS
Census Household Comp. Key
Census Numident

Years

Variables

1969–2019

cohabiting pairs, except those in
group quarters or at address with
more than 20 individuals in a single
year
1974, 1979, 1984, 1989 primary and secondary filer
1994, 1995, 1998–2019 primary and secondary filer, four dependents
2005, 2008–2012
primary and secondary filer, 20 dependents
2000, 2010
household members and relation to
household head
2001–2018†
household members and relation to
household head
1995–2016, 2018
household members of enrollees and
relation to household head
2018, 2019
individuals with SSN between 0 and
18 years of age, child, mother, father
links
2021Q2
individuals with SSN or ITIN, place
of birth (linked to state, county, and
commuting zone), date of birth, date
of death, sex, race

There are small samples during the ACS trial years between 2001 and 2004, so while statistical information is not
used for these years, respondents’ address-level information and household relationships are used in the creation
of these crosswalks.
†

67

Table B4: Ten most common criminal charges within each offense category
Violent

Property

Drug

1200
1240
1180
1230
1990
1090
1220
1070
1010
1060

2040
2070
2010
2140
2050
2110
2060
2100
2120
2130

3150
3160
3250
3080
3110
3990
3140
3070
3030
3100

Aggravated assault
Extortion threat
Armed robbery
Simple assault
Other violent offense
Child molestation
Child abuse
Rape
Murder
Kidnapping

68

Driving under the influence
4020 Driving under the influence of alcohol
4010 Driving while intoxicated
4030 Driving under the influence of drugs

Forgery/fraud
Theft
Burglary
Criminal trespass
Grand theft
Destruction of property
Petty theft
Receiving stolen property
Hit and run driving, property damage
Unauthorized use of a vehicle

Public order

Possession/use of marijuana
Possession/use of unspecified drug
Drug paraphernalia
Distribution, drug unspecified
Possession/use of cocaine or crack
Other drug offense
Possession/use of controlled substance
Distribution of marijuana
Distribution of cocaine or crack
Possession of amphetamines

Criminal traffic

5130 Obstruction/resisting arrest
6010 Traffic offense, minor
5170 Disorderly conduct offense
5990 Public order offense, other
5180 Liquor law violation
5040 Weapons offense
5090 Other court offense
5080 Contempt of court/court order violation
5070 Probation violation
5150 Commercialized vice
5110 Offense against morals/decency
Source: Estimates calculated from CJARS court records held by the University of Michigan and not protected by 13 USC §9a.
Notes: Offense codes follow the classification schema outlined in Choi et al. (2022).