See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/272413343

Juvenile Incarceration, Human Capital and Future Crime: Evidence from

Randomly-Assigned Judges

Article in Quarterly Journal of Economics · April 2015

DOI: 10.1093/qje/qjv003

CITATIONS READS
345 2,899

2 authors, including:

Anna Aizer
Brown University
57 PUBLICATIONS 3,229 CITATIONS

SEE PROFILE

All content following this page was uploaded by Anna Aizer on 04 May 2016.

The user has requested enhancement of the downloaded file.

 Juvenile Incarceration, Human Capital and Future Crime:
Evidence from Randomly-Assigned Judges
Anna Aizer∗ and Joseph J. Doyle, Jr. †‡

January 16, 2015

Abstract

Over 130,000 juveniles are detained in the U.S. each year with 70,000 in detention on any
given day, yet little is known whether such a penalty deters future crime or interrupts social
and human capital formation in a way that increases the likelihood of later criminal behavior.
This paper uses the incarceration tendency of randomly-assigned judges as an instrumental
variable to estimate causal effects of juvenile incarceration on high school completion and
adult recidivism. Estimates based on over 35,000 juvenile offenders over a ten-year period
from a large urban county in the U.S. suggest that juvenile incarceration results in substan-
tially lower high-school completion rates and higher adult incarceration rates, including for
violent crimes. In an attempt to understand the large effects, we found that incarceration for
this population could be very disruptive, greatly reducing the likelihood of ever returning to
school and, for those who do return, significantly increasing the likelihood of being classified
as having an emotional or behavioral disorder.

∗
Brown University and NBER. email: anna [email protected]
†
MIT and NBER. email: [email protected].
‡
We would like to thank David Autor, Julie Biehl, Janet Currie, Pedro Dal Bo, Alison Flaum, Lawrence Grazian,
Lawrence Katz, Jens Ludwig, Derek Neal, Steven Raphael, Roberto Rigobon, Thomas Stoker, Tavneet Suri, Heidi
Williams and seminar participants at Aarhus University, Harvard University, Institute for Research on Poverty, Mid-
west Economics Association, MIT, NBER Childrens/Labor Studies Summer Institute, Tulane University, and the
University of Maryland. We would like to acknowledge the Chapin Hall Center for Children at the University of
Chicago for the creation of the Integrated Database on Child and Family Programs in Illinois (IDB) that was used in
this study. All findings, interpretations and conclusions based on the use of the IDB are solely our responsibility and
do not necessarily represent the views of the Chapin Hall Center for Children.
I Introduction

The U.S. has the highest incarceration rate of any OECD country - with rates triple that of the

next highest country (Walmsley, 2013). The high rate of incarceration in the U.S. cannot be

explained by higher rates of crime. Since 1990, U.S. crime rates have fallen each year, while

incarceration rates have doubled to the point where over 2.2 million adults were incarcerated in

the U.S., and an additional 4.8 million were under supervision of correctional systems in 2011

(Glaze and Parks, 2012).1 Thus, what distinguishes the U.S. is the punitiveness of its criminal

justice policies: the ratio of those incarcerated to those convicted is 70% higher in the U.S. than

the next highest country (Civitas, 2012). Such punitive policies are extremely costly: federal,

state, and local expenditures on corrections currently exceed $82 billion annually, with the direct

expenditures on the wider justice system totaling over $250 billion (Kennelman, 2012).

The high rate of incarceration in the U.S also extends to juveniles. In 2010, the stock of juve-

nile detainees stood at 70,792, a rate of 2.3 per 1,000 aged 10-19 (OJJDP, 2011). Including those

under correctional supervision, the U.S. has a juvenile corrections rate that is five times higher

than the next highest country, South Africa (Hazel, 2008). Despite the high rate of juvenile incar-

ceration in the U.S., little is known about its impact on juveniles. In a life-cycle context, incarcer-

ation during adolescence may interrupt human and social capital accumulation at a critical time

leading to reduced future wages in the legal sector and greater criminal activity. More generally,

interventions during childhood are thought to have greater impacts compared to interventions

for young adults due to propagation effects (see, for example, Cunha et al., 2006), and criminal

activity is a particularly important context to consider such effects due to the negative exter-

nalities associated with it.2 This paper aims to estimate causal effects of juvenile incarceration

on human capital accumulation, as measured by high school completion, and recidivism as an

adult.

The existing research on the impacts of incarceration on future outcomes has focused largely
1
See Raphael and Stoll (2013) and Neal and Rick (2014) for detailed discussions of why incarceration rates have
increased so much over this period.
2
When considering the determinants of criminal activity dominated by young adults, large effects of juvenile
interventions are plausible. See, for example, Currie and Tekin (2006).

1
on adults and on such outcomes as employment, earnings and recidivism. The main challenge

inherent in estimating the causal impact of incarceration (for both adults and juveniles) is to

control or otherwise account for the influence of individual characteristics that may jointly in-

fluence incarceration and future human capital accumulation, criminal activity and labor mar-

ket outcomes. These characteristics include greater socio-economic disadvantage, lower levels

of cognitive achievement and less self control.

The previous work on recidivism conducted by criminologists yields mixed results.3 Mean-

while, the literature on labor market outcomes generally suggests that incarceration has a small

causal impact on the labor market earnings and employment of adult men (Western, Kling and

Weiman (2001)).4 These studies attempt to address the potential endogeneity of incarceration by

controlling for a limited set of observable characteristics (Freeman, 1992; Western and Beckett,

1999), or they use panel datasets that enable one to compare earnings before and after a spell

of incarceration (Lott, 1992a, 1992b; Waldfogel, 1994; Grogger, 1995). The fixed effect approach,

however, cannot be used to study the impact of juvenile incarceration, as juveniles have not

yet entered the labor market. Moreover, this approach assumes that the timing of incarceration

is exogenous, and that it is not correlated with changing life circumstances that might also af-

fect labor market outcomes. A shock to labor market productivity, for example, could lead to

criminal behavior rather than the opposite.

Another approach, first employed by Kling (2006), is to instrument for sentence length using

an index of each judge’s sentencing severity. Kling (2006) finds that incarceration has small pos-

itive effects on employment that fade over time. Di Tella and Schargrodsky (2013) and Green

and Winik (2010) have also used this strategy to estimate the impact of incarceration on recidi-

vism with differing results. This approach implicitly controls for all unobservables (fixed and

changing) that might bias estimates because judges are randomly assigned to cases and is the
3
Some work finds that incarceration increases recidivism (Spohn and Holleran, 2002; Bernburg, Krohn and Rivera,
2006), others find that it has no effect (Gottfredson, 1999; Smith and Akers, 1993), and still other work finds that it re-
duces recidivism (Murray and Cox, 1979 and Brennan and Mednick, 1994). Kerley et al. (2004) find that incarceration
is negatively correlated with income, especially for those incarcerated earlier in life.
4
Rehabilitation programs could improve labor market outcomes (Landerso (2012)), although prisoners in the U.S.
have access to few rehabilitative services. For example, according to a 2012 GAO report, 31,000 prisoners are enrolled
in drug rehabilitation programs, while another 51,000 remain on waiting lists.

2
approach that we take in this paper.5

Much less is known about the consequences of incarcerating juveniles on future outcomes.

The handful of studies that examine the effect of juvenile criminal activity on education and

labor market outcomes generally find a negative relationship. Most of the existing studies at-

tempt to identify the causal link by controlling for observed individual characteristics (De Li,

1999; Tanner et al., 1999; Sweeten, 2004). More sophisticated studies of this type also control for

interactions with the criminal justice system as well as unobserved household fixed characteris-

tics (Hjalmarsson, 2008).

A second complicating factor is that effects for juveniles on the margin of juvenile incar-

ceration may differ from the average juvenile, and it is the former group that is most likely

to be affected by policy changes. Perhaps the most convincing evidence to date comes from

a regression-discontinuity design using sentencing rules to identify the impact of juvenile in-

carceration on recidivism in Washington state (Hjalmarsson, 2009). At the margins where the

sentencing becomes more severe, juveniles just above the thresholds were found to be less likely

to recidivate as a youth.

Our estimation strategy addresses these complicating factors. We exploit plausibly exoge-

nous variation in juvenile detention stemming from the random assignment of cases to judges

who vary in their sentencing. To illustrate, consider two juveniles randomly assigned to two

different judges with differing incarceration tendencies. With random assignment, differences

in incarceration between juveniles are attributed to the effect of the judge and not individual

characteristics of the juvenile or the case, as are differences in outcomes. With this strategy we

address the issue of negative selection into juvenile incarceration and estimate effects for those

at the margin of incarceration where the judge assignment matters for the incarceration deci-

sion.6 But unlike previous work, we use this strategy in a context of juvenile offending where
5
A somewhat related paper by Drago, Galbiati and Vertova (2009) exploits a natural experiment induced by an
Italian clemency bill that increased sentence length associated with future crime for some former prisoners but not
others. The authors find that the increase in expected sentence length exerted a strong deterrent effect on future
crime, but that the deterrent effect decreased with length of time previously incarcerated.
6
Chang and Schoar (2008) and Dobbie and Song (2013) employ a similar strategy using judges assigned to
bankruptcy cases, Maestas, Mullen and Strand (2013) use disability examiner propensities to approve disability
claims, and Doyle (2008) uses case worker propensities to place children in foster care.

3
human capital accumulation may still be in its formative stages, and thus the long-term effects

may well be greatest.

To carry out this strategy, we employ a unique source of linked administrative data for over

35,000 juveniles over 10 years who came before a juvenile court in Chicago, Illinois. These data

were linked to both public school data for the same city and adult incarceration data for the

same state to investigate effects of juvenile incarceration on high school completion and adult

imprisonment.

We find that assignment to a judge with a high incarceration rate in other cases leads to a

significantly lower likelihood of high-school completion and a significantly higher likelihood of

incarceration as an adult, including incarceration for violent crimes. Under the stronger assump-

tions necessary to use this rate as an instrumental variable, juvenile incarceration is estimated

to decrease high school graduation by 13 percentage points and increase adult incarceration by

23 percentage points. In comparison, in OLS regressions with minimal controls, those incar-

cerated as a juvenile are 39 percentage points less likely to graduate from high school and are

41 percentage points more likely to have entered adult prison by age 25 compared with other

public-school students from the same neighborhood. Though the instrumental variable results

are considerably smaller than the OLS results with minimal controls, the differences remain

large and suggest substantial negative effects of juvenile incarceration on long-term outcomes.

The main IV estimates and subgroup analyses suggest that marginal cases are at particularly

low (high) risk of high school completion (adult incarceration) as a result of juvenile custody.

Indeed, the effect sizes are larger for juveniles whose observable characteristics suggest that

they are less likely to be incarcerated as a juvenile. The results are also consistent with the idea

that the timing of incarceration matters: the strongest results are for juveniles aged 15 and 16

– a critical period of adolescence when incarceration is most likely to end one’s high school

education.

Finally, we explore the potential mechanisms behind the estimated negative effects. We find

that although incarceration of these juveniles is intended to be short in duration (one to two

months), it can be very disruptive. Once incarcerated, juveniles are unlikely to ever return to

4
school. However, conditional on returning to school they are not more likely to be transferred

to an alternative (potentially inferior) school, nor are they more likely to be classified as special

education students. Interestingly, they are more likely to be classified for special education

services due to behavioral/emotional disorders rather than a cognitive disability.

Our results have important implications for policies related to juvenile incarceration, such

as the adoption of alternatives to incarceration in juvenile courts across the U.S.. They also

have implications for recent changes in education policy that have placed more police officers

in U.S. schools. This increase has lead to an increase in juveniles being arrested, often for less

serious crimes.7 Our results suggest that more research needs to be done evaluating whether

and to what extent this has led to an increase in juvenile incarceration in order to better inform

policy-makers’ decision to expand such a program.

The rest of the paper is organized as follows: in Section II we provide background infor-

mation on the juvenile justice system and judge assignment in our context; in Section III we

describe the data; in Section IV we describe the empirical strategy; Section V presents the re-

sults; and Section VI offers interpretation and conclusions.

II Background

II.A The Juvenile Justice System & Judge Assignment

In Chicago, juvenile offenders of minor crimes are often dealt with directly by the police. Only

after a number of smaller infractions, or a major infraction, will a child enter the juvenile court

system.8

When juveniles are charged with a crime in juvenile court, they are assigned to a calendar

which corresponds to the youth’s neighborhood of residence. Calendars generally have one or

two judges who usually preside over cases assigned to them.9 Furthermore, approximately one-
7
In 1975, 1% of U.S. schools had police, increasing to 22% in 1997, and 40% by 2007 (Na and Gottfredson, 2011).
8
Every juvenile arrest is reviewed two times before proceeding to juvenile court: first, by the police and a second
time by the prosecutor’s office. At each review the juvenile’s case can be disposed. Only those cases not dismissed
by the police or the prosecutor proceed to juvenile court.
9
The district attorney is also assigned by calendar so that within a calendar the characteristics of the district

5
fifth of hearings are presided over by judges that cover the calendar when the main judge(s) are

not available. These judges are known as ”swing judges”.

Within a calendar, the judge assignment is a function of the sequence with which cases hap-

pen to enter into the system and the judge availability that is set in advance. In particular, there

does not appear to be scope for influencing the first judge seen. It is at the first court hearing,

for example, that juveniles meet their public defenders (who are also assigned based on day

of hearing) and learn who the judge will be. Conversations with court administrators confirm

that these assignments are effectively random and that there is no way to influence the judge

assigned to the case. As a partial check on this important assumption of random assignment,

we test the relationship between observable characteristics and judge assignment below.

One exception to calendar assignment based on residence of the juvenile is youths charged

with a weapons offense. Over our time period, these youths can be assigned to a separate calen-

dar that oversees such offenses, but assignment to a judge within the ”weapons” calendar is still

based on the sequence of court cases being heard.10 We account for this differential treatment of

weapons charges in our analysis, as described in the section on empirical strategy.

In terms of sentencing, nearly all cases, 96%, that come before the court are found guilty

(typically by plea) of the charges (Peters et al., 2002). As a result, a judge’s main influence on

the case is whether the juvenile is placed on probation or detained and then placed on proba-

tion. For the first cases we consider here, custody is nearly always in the Cook County Juvenile

Temporary Detention Center, which is available for children aged 10-16 – the ages applicable

for juvenile offenses in Illinois.11 These sentences are indeterminate in length, but typically last

1 to 2 months including pre-trial detention. We do not observe length of time incarcerated in

our data. As a result, our analysis considers the effects of a typical stay in incarceration (approx-

imately 42 days in our data), but not whether and to what extent length of time incarcerated
attorney is constant across the judges.
10
We attempted to learn more about the weapons court, but since it no longer exists, administrators could only
confirm its previous existence, but could not provide more detail.
11
Juveniles may also be sentenced to a juvenile facility run by the Illinois Department of Corrections where typical
stays are between 6 months and 2 years, which we also regard as incarcerated in our treatment variable. Only 0.6% of
the cases in our analysis sample are found in the Department of Corrections facility within a year of the first hearing,
however.

6
matters. However, we will describe analyses using a proxy for length of stay in the robustness

discussion.

The alternative for juveniles in our sample who receive a sentence of incarceration followed

by probation is simply probation. The conditions for successful probation are universal and

include: attending school (which we will consider directly below), not associating with known

criminals, and not using illegal drugs. Once the juvenile is on probation, the judge no longer

has any contact with the juvenile.12

It is important to note that the juvenile incarceration rate in this state is similar to the aver-

age for the US as a whole.13 This increases the likelihood that the results apply more broadly

compared to a situation where the state was an outlier in terms of incarceration rates.

II.B Mechanisms: How Juvenile Incarceration Can Affect Outcomes

Juvenile incarceration can affect high school completion and future criminal activity through

two potential channels: changing the skills or actions of the individual juvenile (a behavioral

channel) or changing the ways in which institutions regard and treat him (a deviant labeling

channel). With respect to the former, incarceration can negatively affect child mental health,

leading to behavioral problems in school and at home (Kashani et al, 1980; Forrest et al, 2000).

Incarceration can also encourage the accumulation of ”criminal capital” (Bayer, Hjalmarsson

and Pozen, 20110) and hinder the accumulation of social capital (Ganovetter,1995). Disrupting

school attendance, even if only for one month, can also have the effect of increasing the cost

of going to school (if ”catching up” on lost schoolwork is costly), leading to drop out. Alter-

natively, incarceration could have a positive effect on human capital accumulation and future

crime by either reducing the uncertainty regarding the cost of jail so that juveniles who are de-

tained adjust (upwards) the cost of spending time in detention, or by reducing truancy since the
12
One issue is that judges could sentence juveniles to electronic monitoring or home curfews (considered alter-
natives to detention). However, these alternatives were not introduced until 1995 and not widely used until much
later. Our sample is comprised primarily of youth who came before the juvenile court prior to adoption of these
alternatives.
13
Juvenile incarceration rates per 100,000 range from 53 to 440 across the 50 U.S. states with an average 225 (Office
of Juvenile Justice and Delinquency Prevention, 2011). In Illinois , the rate (178) is similar to the average for the US,
suggesting that the state is not an outlier in its juvenile incarceration tendencies.

7
detention facility includes a school administered by Chicago Public Schools.

With respect to the deviant labeling channel, schools, for example, may be unwilling to al-

low the juvenile to re-enroll in school once released, forcing the juvenile to enroll in another,

potentially inferior school. Even if re-enrolled in their original school, schools could treat ju-

veniles differently, such as classifying them as special-education students. The criminal justice

system might also regard them differently - police may be more likely to suspect and arrest them

and, conditional on conviction, judges may be more likely to sentence them more harshly, as-

suming their juvenile records haven’t been expunged.14 Finally, employers may be less likely to

hire them (see Bernburg and Krohn, 2003), thereby increasing the likelihood of future criminal

activity.

In our empirical work, we begin by estimating the overall effect of incarceration as a juvenile

on high school completion and adult incarceration. We follow this with an exploration of the

potential mechanisms underlying the estimated effect. To do so we begin with an examination

of whether juveniles who are incarcerated are less likely to return to school. Among those who

do return to school, we examine whether they are more likely to transfer schools upon release,

consistent with schools being unwilling to re-enroll juveniles who have been incarcerated. We

follow this with an exploration of whether students who are incarcerated and return to school

(any school) are more likely to be classified as a special-education student and the nature of the

disability.15 Finally, we examine the impact of juvenile incarceration on different types of crimes.

If the effects remain for types of crime that would result in adult incarceration regardless of the

characteristics of the offender (eg, homicide), that would be consistent with a change in criminal

activity on the part of the juvenile, not simply a change in the way the criminal justice system

treats the offender.

14
Juvenile records can be expunged but it is not automatic and must be requested. Moreover, certain conditions
must be met, namely lack of future criminal conviction for five years. We spoke with juvenile defense attorneys in
Cook County to learn more about the role of juvenile detention in adult court sentencing decisions. While adult
court judges do learn about previous convictions and may learn about previous incarceration in a Department of
Corrections facility (0.6% of our cases), they are highly unlikely to learn about detention in the Juvenile Temporary
Detention Facility (99.4% of our cases), suggesting that this is an unlikely channel.
15
There are no data on suspensions or other disciplinary actions during most of the period of time our sample was
school-age.

8
III Data Description

III.A Data Sources

The data come from three primary sources: Chicago Public Schools Student Database (1990-

2006), the Juvenile Court of Cook County Delinquency Database (1990-2006), and the Illinois

Department of Corrections Adult Admissions and Exits Database (1993-2008). The data were

linked using identifiers including name, date of birth, and address information, by the Chapin

Hall Center for Children, a child welfare research institute – and a leader in administrative-data

linkage – located at the University of Chicago (Goerge, Van Voorhis, and Lee, 1994).

The CPS data come from a system that characterizes each child by his or her age, race, sex,

birth year, measures of special education needs, as well as the U.S. Census tract of residence.

We linked the tract information to 2000 U.S. Census data describing the fraction of families in

poverty. We aggregated each student’s residence to one of 76 long-standing neighborhoods in

Chicago, 67 of which are included in our analysis dataset.16 Results controlling for the tract itself

will be reported in the robustness section.

The raw Juvenile Court data are at the hearing level. These data include the date, a judge

identifier, the offense, and the disposition: probation or detention followed by probation. Un-

fortunately, the length of time in a juvenile facility is not part of the disposition – rather, the

sentences tend to be indeterminate subject to future hearings.

The Illinois Department of Corrections data describe each adult prisoner’s spell and allow

us to observe whether or not these juveniles are found in adult prison in Illinois later in life.

Further, the data list the offense for which the individuals are incarcerated, and we test the

effects of juvenile incarceration on adult incarceration for different types of offenses.

16
On average, a community comprises 14 Census tracts. We use the definitions of community as defined by the
University of Chicago and which can be found here: http://www.lib.uchicago.edu/e/collections/maps/ssrc/

9
III.B Sample Construction

Table AI describes the sample construction.17 We begin with the sample of students in the eighth

grade in the Chicago Public School (CPS) system. One of our main outcomes of interest is adult

incarceration by age 25, and to measure this outcome without censoring, we restrict the sample

to those who are at least 25 by 2008 – the last year of our incarceration data, corresponding to

the cohort born between 1971 and 1983. This also ensures that we do not have censoring with

respect to the high school graduation outcome. Two percent of the data were excluded due to

missing U.S. Census tract information or a small number of recorded tracts that were not linked

to a Chicago community, resulting in 440,797 children.

Of these, 41,764 (9.5%) of these students came before the juvenile court system during our

timeframe (1990-2006). We focus on the juvenile’s first case in our data. We excluded a small

number of observations where the case was transferred to an adult court, where the recorded

age was miscoded (i.e., not between 10 and 16), as well as cases assigned to judges with fewer

than 10 cases. Finally, the baseline regressions employ fixed effects defined at the community

x year x weapons offense level (for reasons explained in the empirical strategy below), and we

drop cases where these cells have fewer than 10 observations. This results in 37,692 observations

in the juvenile court data.

III.C Sample Description

Table I reports sample means for the entire Chicago Public School sample and the juvenile court

sample. For the latter, we further divide the sample into those incarcerated as a juvenile, and

those not incarcerated. The only characteristic along which all three groups appear similar is

birth year, with most of the mass in the 1974-1982 birth cohorts. The samples differ consid-

erably along all other dimensions. The juvenile court sample is more likely to be male and

African American, more likely to be special-education students in eighth grade, and live in

higher poverty neighborhoods.18 They are also less likely to graduate high school and more
17
All appendix material is in the on-line appendix
18
The poverty rate was filled in with the average of the contiguous Census tract poverty rates for 6 observation in
court sample and 23 observations in full CPS sample when that information was suppressed by the Census.

10
likely to be incarcerated by age 25. The graduation rate for the full sample is only 40%, defining

transfers as not graduating from high school.19 These comparisons suggest that limiting the

analysis sample to the juvenile court sample and comparing outcomes for those incarcerated

with those convicted but not incarcerated is likely to reduce the bias associated with negative

selection into incarceration on underlying characteristics. However, it will not likely eliminate

the bias - the samples still differ on key characteristics that are correlated with the outcomes.

Those incarcerated are more likely to be male, slightly more (less) likely to be African Ameri-

can (Hispanic), and much more likely to be special-education students in eighth grade. Those

incarcerated are also slightly younger at the time of the first offense–a predictor of adult incar-

ceration. With an average birth year of 1978 and an average age at the first offense of 14-15 years

old, the typical case occurs in 1992-1993 in this analysis sample.

One drawback of the data is that they include only school completion (incarceration) out-

comes in the same city (state) as the juvenile court. If individuals move away, we do not observe

their high school completion or their recidivism. Regarding high school completion, among ju-

veniles charged with a crime, 3.4% transfer to private school and 10% transfer out of the district,

suggesting that we can accurately measure high school completion for the vast majority of juve-

niles. For the main specification, we code this 13.4% of the sample as non-graduates. Another

18% of the sample transfer from the Chicago Public Schools to an adult correctional facility with-

out completing high school. These individuals are also coded as non-high school graduates.20

In the robustness section we consider these transfers directly. Regarding our measure of adult

recidivism, data from the 2000 Census show that among those born in Illinois between 1970 and

1982, by the year 2000 (when they range in age from 18 to 30), three quarters remain in Illinois,

and the rate of migration is lower for those with less education. We anticipate little bias to be

introduced by this form of sample selection.

19
In Illinois during this period the school leaving age was 16. This age has been raised to 17.
20
These individuals can earn a GED in prison, but we do not have that information. Even if they did complete a
GED, a GED confers much lower wages than a high school diploma.

11
IV Empirical Framework

IV.A Set Up

For juvenile i, consider a model that relates an outcome such as adult recidivism, Yi , to an

indicator that the juvenile was incarcerated at some point during his youth, JIi :

Yi = β0 + β1 JIi + β2 Xi + i (1)

where Xi is a vector of control variables and i is the error term.

Any assessment of the impact of juvenile incarceration on high school completion and adult

incarceration must address the problem posed by the positive correlation between juvenile in-

carceration and factors such as severity of the crime, criminal history and characteristics of the

juvenile that are also likely to be correlated with the outcomes.

In our analysis, we take several steps to address this. Specifically, we present several differ-

ent specifications that incrementally control for confounding factors so that we can observe the

extent to which omitted variables may be driving the observed correlations between juvenile in-

carceration and the outcomes. Initially, we compare juveniles incarcerated with other children

in the public school system from the same neighborhood. We then present specifications that

1) add controls for multiple demographic characteristics including race, sex, birth year, share in

poverty in the Census tract of residence, and an indicator for special- education status in eighth

grade, 2) employ propensity score techniques using these same geographic and demographic

controls in an attempt to further control for omitted variables, and 3) limit the analysis to all

juveniles charged with a crime and brought before the juvenile court, though not necessarily in-

carcerated, further controlling the age at the time of the offense (instead of birth year), the type

of crime (10 categories) and a risk assessment index which is a checklist of criteria that is applied

by the Department of Probation to rate each juvenile for specific detention-related risks.21

Despite the inclusion of an increasingly comprehensive set of controls, there may still be
21
The scale ranges from 1 to 15 with a higher number indicating greater risk and therefore stronger recommen-
dation for detention. We calculated the index from the charge information. In the models with the charge category
indicators, this index serves to further control for the severity of the charge among those with ”other offenses”.

12
unobservable characteristics of either the crime or the juvenile that are correlated with both the

probability of juvenile incarceration and future outcomes. In the case of high school completion,

it’s most likely that these unobservable characteristics are negatively correlated with juvenile

incarceration, biasing OLS estimates of the impact of JI downward, and in the case of adult

incarceration, its most likely that the unobservable characteristics are positively correlated with

JI, which would bias OLS estimates of the impact of JI upward.

In addition, the effects of juvenile incarceration are likely to be heterogeneous, and we could

augment the above model to allow for a random coefficient on juvenile incarceration, which

would allow the effects to vary by juvenile. A concern in estimating such models is a correlated

random coefficient (Bjorklund and Moffitt, 1987), where the placement into custody may be

related to the effect on adult incarceration. That is, judges choose the sentence, and if they

tailor sentences with the idea of deterring future criminal activity, then a selection bias could

understate the causal effect of juvenile incarceration for cases on the margin of commitment:

those cases most likely affected by policy.

Our main empirical strategy uses a measure of the tendency of a randomly-assigned judge

to order a juvenile be placed in custody, Z, as an instrument for juvenile incarceration. Essen-

tially, we compare high school completion and adult incarceration rates for juveniles assigned

to judges that have different propensities to incarcerate, and interpret any difference as a causal

effect of the change in incarceration associated with the difference in these propensities. These

can be considered marginal cases where the judges may disagree about the custody decision, a

margin of particular policy relevance. In the next subsection, we describe how we calculate the

instrument in greater detail.

IV.B Instrumental Variable Calculation

For each juvenile we assign an instrument that corresponds to the ”incarceration propensity” of

the initial judge in the juvenile’s first case. The instrument, which is defined for each juvenile i

assigned to judge j(i) is simply a leave-out mean:

13
  
  nj(i) −1
1 X
˜ k
Zj(i) =  JI
nj(i) − 1
k6=i

Here, nj(i) is the total number of cases seen by judge j; k indexes the juvenile case seen
˜ is equal to 1 if the juvenile was incarcerated during the juvenile’s first
by judge j where JI

case. Thus the instrument is the judge’s incarceration rate among first cases based on all the

judge’s other cases. Algebraically, this is the judge fixed effect in a model of custody in the

initial case estimated in a ”leave-out” regression estimated over all years. This measure ties the

decision making of the judge in the first case more directly to the first cases we consider here.

The resulting two-stage least squares estimator is a Jackknife Instrumental Variables Estimator

(JIVE), which is recommended for models when the number of instruments (the judge fixed

effects) is likely to increase with sample size (Stock, Wright, and Yogo, 2002; Kolesar et al.,

2011).

In both the first and second stages of the IV regressions, we also include a vector of commu-

nity x weapons-offense x year fixed effects. Recall that judge assignment is based on community

and, during part of our time period, whether there was a weapons charge. Including this fixed

effect thus effectively limits the comparison to juveniles at risk of being assigned to the same

set of judges. With the inclusion of these controls, we can interpret the within-cell variation in

the instrument, Zj(i) , as variation in the propensity of a randomly assigned judge to incarcerate

a juvenile relative to the other juvenile cases seen from the same neighborhood and with either

a weapon or non- weapon offense in the same year. Note that the instrumental variable calcu-

lation is not conditional on characteristics of the juvenile or the crime in order to allow a direct

examination of the sensitivity of the results with and without controls.

IV.C Judge variation

Our analysis dataset includes 62 judges. The average number of initial cases per judge is 607.

More than one judge can hear each juvenile’s case over time, and the instrument is based on the

14
incarceration propensity of the first judge assigned for the juvenile’s first offense.22 If the initial

judge is missing in the data as it is in 17.8% of the cases, we assign the juvenile to the second

judge of record. While the potential for another, later, judge to make the incarceration decision

may lead to a weaker estimated relationship between the first judge’s propensity to incarcerate

(the instrument) and an individual juvenile’s incarceration status, the focus on the first judge

has the advantage of not capturing any (potential) non-random changing of judges.

The initial-case incarceration propensity has a mean of 0.097 with a standard deviation of

0.039. Results will be shown with alternative measures of the instrument as checks on robust-

ness as well. Variation in the instrument can also be seen in Figure I, where we present the

distribution of the instrument defined two ways. First, as the leave-out mean of the probability

of incarceration for each judge (denoted “raw” in the figure). Second, as the residual from a

regression in which we include controls shown in Table I, including indicators for each year of

age at the time of the offense (rather than birth year) and community x weapons offense x year

indicators. The residualized measure represents the variation in the instrument that we use for

identification and suggests substantial variation even with a full set of controls. In particular, the

raw measure ranges from approximately 4% to 21%, and the residualized measure still shows

substantial variation: ranging from 6% to 18%.

This variation comes from two sources: variation among ”regular” (i.e., non-swing) judges

assigned to the same calendar (roughly 80% of cases are seen by a regular judge) and variation

from swing judges who oversee the remaining 20% of the cases.23
22
35% of the initial cases have the same initial and final judge across all of the hearings. Over the course of the
criminal proceedings, which often involve multiple hearings, the judge may change either temporarily or perma-
nently.
23
We define regular judges as those who see at least 75% of their cases in the given calendar x year and swing
judges as those who see fewer than 75% of their cases in a given calendar x year.

15
V Results

V.A Instrument Validity

While we cannot directly test the exclusion restriction, we can provide evidence consistent with

the condition being met. First, we have confirmed with court personnel that judges are assigned

in a way that leads to a “natural randomization” of cases to judges: cases are assigned to cal-

endars based on the juvenile’s residence and within calendars, judges cannot influence which

cases they hear. Second, we can partially test this empirically in the data by examining whether

the characteristics of juveniles and their cases differ by judge. We do this by testing whether

the characteristics of juveniles differ based on whether they are assigned to a judge with either

a high, medium or low propensity to incarcerate (defined by bottom, middle or top tercile of

the distribution of propensity to incarcerate) relative to other judges in the same community x

weapons-offense x year cell. The results (Table II) show that judges with high, medium and low

propensities to incarcerate are assigned juveniles that are extremely similar in terms of their gen-

der, race, and special education needs, poverty rate of his Census tract, and age at the time of the

offense, despite significant differences in the incarceration rate used to define the categories.24

We also calculate a single measure - the propensity to be incarcerated based on the above observ-

able juvenile characteristics and report the average propensity for lenient, moderate and strict

judges. The propensity is the same (0.22) regardless of judge type.25 Table II reports results for

exogenous variables, whereas the results below will be shown with and without controls for

the potentially endogenous control variables determined by the court (ie, charges). Despite not

finding a relationship between observable juvenile characteristics and the judge’s propensity to

incarcerate, we offer a final piece of evidence which is to present and compare results when we
24
An F-test of joint significance for whether these control variables predict that the judge is in the top tercile yields
a p-value of 0.15. When we regress our (continuous) instrument on all of the controls, however, the controls are
jointly significant at the 1% level. That said, the coefficient values themselves are very small: the two variables that
are individually significant include special education status and the age = 10 indicator. Special-education status
is associated with an increase in the instrument of 0.0017 (compared to a mean of 0.097). The age=10 indicator is
associated with a 0.0088 reduction in the instrument (compared to the Age = 16 indicator, the excluded category).
25
The propensity is predictive of incarceration, however. The predicted propensity is 0.272 for juveniles who were
incarcerated, and 0.214 for those who were not incarcerated.

16
control for case characteristics and models when we do not.26

We will interpret the IV results as local-average treatment effects: average effects for cases

where the judge assignment matters for the incarceration decision. This requires a monotonicity

assumption: assignment to a strict judge need not increase the likelihood of incarceration for

each type of offender. This assumption is stronger in this setting (Imbens and Angrist, 1994).

For example, a judge might treat cases involving drugs relatively harshly, but theft/larceny

relatively leniently. We consider this concern more fully in our robustness checks.

V.B First Stage: Judge Assignment and Juvenile Incarceration

To consider the first-stage relationship between initial-judge assignment and whether the ju-

venile is ever incarcerated as a juvenile (JI), we estimate the following equation for juvenile i

assigned to judge j(i) in community x weapon-offense x year cell c(i) using a linear probability

model:

JIi = α0 + α1 Zj(i) + α2 Xi + δc(i) + νi

The vector Xi represents demographic controls and court measures described above, as well

as an indicator that the judge identifier at the first hearing is missing). Similar results are found

for both the first stage and the instrumental variable results when probit models are used, which

is unsurprising given that the outcome variables are relatively far from zero. Zj(i) refers to the

judge’s incarceration rate among juveniles’ initial cases. The mean initial judge custody rate is

0.09, whereas the mean of the dependent variable in this first-stage model – an indicator that

the juvenile was ever-incarcerated – is 0.23. All standard errors are clustered at the community

level.

The results of the first stage (Table III) show that the judge’s incarceration rate is highly pre-

dictive of whether an individual will ever be incarcerated as a juvenile. Including additional

26
Another concern would be that judges may affect juveniles in other ways besides the likelihood of juvenile
incarceration. Again, conversations with court personnel suggest that this is not the case. Moreover, judges who
are more likely to incarcerate are not more likely to incarcerate for a longer period of time, conditional on any
incarceration.

17
controls in columns 2 and 3 does not change the estimated effect of being assigned to a strict

judge in one’s first court appearance, consistent with the randomness of judge assignment. Col-

umn (3) which includes the full set of controls, reports a coefficient 1.06. The coefficient is not

statistically significantly different from 1, meaning that if a juvenile is assigned to a judge that is

10% more likely to incarcerate other juveniles in their initial case, he is 10% more likely to be in-

carcerated as well.27 In particular, the estimate suggests that a two standard deviation increase

in the judge incarceration rate would imply an increase in the likelihood of juvenile incarcera-

tion of 8.5 percentage points – or 37% of the mean rate of juvenile incarceration. All first-stage

estimates are precise, with t statistics around 11.

V.C Juvenile Incarceration and High School Completion

We estimate the impact of incarceration at any time as a juvenile on the probability of graduating

from high school according to the equation below that echoes (1) above:

Yi = β0 + β1 JIi + β2 Xi + ηc(i) + i

where Yi is an indicator for whether juvenile i in community x weapons-offense x year cell

c(i) graduated from high school, and JIi is an indicator for whether juvenile i was ever incarcer-

ated as a juvenile. We present both OLS regression results and results in which we instrument

for JIi using the judge incarceration rate, Zj(i) . As with the first stage, we present results both

with and without controls (Xi ). When we report results for the full Chicago Public School sam-

ple, the year-of-offense and weapons-offense components of the fixed effects do not apply to

those not part of the juvenile justice system. As a result, those models include community fixed

effects and birth-cohort indicators instead.

Table IV reports the results. The table is organized such that with each column we further

control for potential omitted variables so that we can learn about the source(s) and size of any
27
A coefficient greater than one is possible because the incarceration rate (Zj(i) ) applies to whether the juvenile
was incarcerated in his first case, whereas the endogenous variable for which we instrument is whether the juvenile
was ever incarcerated as a youth - for his first case or any subsequent cases.

18
bias. In the first three columns, the sample includes all children in the Chicago Public Schools.

Therefore in the first three specifications we are comparing the high school completion rates of

children incarcerated as juveniles to a control group from the same community that includes

two groups: those without any juvenile court involvement and those with juvenile court in-

volvement but who were not incarcerated as juveniles. In the first column which includes only

community fixed effects as controls, we observe a strong negative relationship: children in-

carcerated as juveniles are 39 percentage points less likely to complete high school than other

children from their neighborhood. In column 2 we include the following demographic controls:

sex, race/ethnicity, share below poverty in Census tract, year of birth fixed effects, and an indi-

cator for special education status in eighth grade. When we do, the coefficient estimate falls by

almost a fourth from -0.39 to -0.29, which is still very large given an average rate of high school

completion among this sample of 43%.28

We also present propensity score estimates to determine whether this method can further

limit the amount of omitted variable bias. We predict the probability of juvenile incarceration

using a probit regression with the demographic characteristics listed above as well as commu-

nity indicators and estimate the relationship between juvenile incarceration and high school

completion using inverse-propensity score weighting. The result (column 3) is an estimate of

the impact of incarceration on high school completion that is the same as the result obtained

when we excluded most of the controls, suggesting that this method does not effectively reduce

omitted variable bias in this particular context.29

In the next two columns (columns 4 and 5), we limit our sample to children with a crimi-

nal case in juvenile court. By using this subsample, we are limiting our comparison or control

group to juveniles charged with a crime in court but not incarcerated. We argue that this sam-

ple restriction is likely to further reduce potential omitted variable bias, as this sub-sample is
28
As noted above, those that do not graduate include those who have transferred out of Chicago Public Schools
and it’s possible that they may have graduated from another school, though we do not observe this. We investigate
sensitivity to removing those that transfer or recoding them as graduates as robustness checks.
29
The propensity score estimates are based on a slightly smaller sample due to the fact that we were unable to
calculate a propensity based on a probit regression for a small subset of the sample for whom the probit perfectly
predicted failure/success.

19
much more disadvantaged than the general CPS sample and therefore more similar to the sam-

ple of juveniles who are incarcerated (see Table I). Moreover, this limits the control group to

those at risk of incarceration. Our OLS estimate in column 4, which includes only community

x weapons-offense x year-of-offense fixed effects, supports this: the coefficient on juvenile in-

carceration falls to -0.088 when we restrict the sample in this way, although this is still large

compared to the mean graduation rate in the sample of 9.9%. Adding additional controls for

the demographic characteristics listed above and the characteristics of the case (type of charge,

etc.) in column 5 reduces the OLS estimate only slightly to -0.073. This suggests that either we

have adequately addressed most of the potential bias from omitted variables with our sample

selection and set of controls, or that the only way to improve upon these estimates is to employ

an identification strategy that exploits plausibly exogenous variation in juvenile incarceration.

Our final set of estimates does just that by instrumenting for juvenile incarceration using

the propensity of an individual’s randomly assigned judge to incarcerate. The instrumental-

variable point estimates, -0.108 (column 6) excluding controls and -0.125 including controls

(column 7), are much smaller than the OLS estimates based on the entire sample of children

(columns 1-2), but larger than the OLS estimates based on the subsample of children with a ju-

venile court case (columns 4-5). The standard errors are larger as well; however, the estimates

are not statistically-significantly different from the OLS estimates based on the juvenile court

sample.30

Reduced-form estimates (Table AII) are very similar to the IV estimates, consistent with the

strong relationship between the propensity of the assigned judge to incarcerate and one’s own

incarceration, as captured by the first stage coefficient of 1.06. These results imply that within

the range of incarceration rates in our data described in Figure I, moving from the least strict to

most strict judge - an increase of 12 percentage points - increases the probability of high school

drop out by 1.6 percentage points, or 16% of the mean.

To the extent that cases are randomly assigned and the main effect of the judge on juveniles is
30
We cluster at the level of the community in all specifications. Clustering at the level of the judge (of which there
are 62) yields a standard error of 0.065 in the 2SLS with full controls, column 7.

20
whether or not they are incarcerated, we can interpret the IV estimates to suggest that juveniles

on the margin of incarceration – compliers where the judge assignment induces a change in the

incarceration decision – are 12.5 percentage points less likely to complete high school: essentially

all students on this margin who enter juvenile incarceration do not graduate.31 Taken at face

value, the OLS and IV point estimates suggest that the children on this margin may experience

larger effects of juvenile incarceration on high school completion than the average incarcerated

juvenile. That is, many juveniles may experience little causal effect of juvenile incarceration on

their high school completion – those with minor offenses are at lower risk of not completing

high school, or those charged with very serious crimes and certain incarceration may be at such

a disadvantage at school that high school completion is already extremely unlikely – whereas

the marginal cases may be particularly affected by incarceration. We explore heterogeneity in

the treatment effects across different types of cases to explore this possibility below.

Moreover, the treatment of interest is binary: an indicator if the juvenile were ever incar-

cerated. The instrumental-variable estimate extrapolates the change in the propensity to be

incarcerated to a change in the indicator for incarceration from zero to one. This extrapolation

can lead to large point estimates, as well as larger standard errors. It is worth reiterating that the

range of variation in the instrument (and subsequently in the propensity to be incarcerated) is

only 12 percentage points, and so any relationship between the instrument and the unobserved

propensity to graduate high school will be magnified. In the end, we regard the point estimate as

evidence of large effects of juvenile incarceration on high school completion for marginal cases

but recognize that the larger standard errors suggests caution in the interpretation, especially in

comparison to the magnitude of the OLS estimates.

Our finding of a strong negative impact of juvenile incarceration on this measure of human

capital accumulation suggests that we may find negative effects on adult recidivism as well,

which we explore in the next section.

31
With 23% ever placed in custody, we can calculate the weighted average of those placed in detention and those
not placed that results in the overall mean graduation rate of 9.9%: 0.23(X - 0.125) + 0.77(X) = 0.099, implying that X =
12.7% graduate among those not placed in detention and X - 0.125 = 0.2% graduate among those placed in detention.

21
V.D From Juvenile Incarceration to Adult Incarceration

We analyze the impact of juvenile incarceration on the probability of adult incarceration in the

state of Illinois using the same empirical specifications as above. We define adult incarceration

by whether an individual was present at any point by the age of 25 in an adult correctional facil-

ity anywhere in the state. Moreover, since we observe the types of crimes for which individuals

are assigned to adult correctional facilities, we can define adult recidivism by type or severity of

the adult crime.

Table V reports results for any adult incarceration, regardless of crime type. The adult im-

prisonment rate, defined this way, is 6.7% in the larger CPS sample. The OLS results show

a strong relationship between juvenile incarceration and adult incarceration: those who were

in juvenile detention are 41 percentage points more likely than other children residing in the

same community to be found in an adult correctional facility by age 25 (column 1). Adding

demographic controls reduces this relationship to 35 percentage points (column 2), and inverse

propensity score weighting reduces the estimated effect further still to 22 percentage points,

(column 3).

When we limit the control group to those who came before the juvenile court but were not

committed and include controls for demographic characteristics and the type and severity of

the juvenile crime (column 5), the estimated effect falls to 16 percentage points. Note that the

average adult incarceration rate for this group is considerably higher (32.7%) so that the esti-

mate represents an increase in adult recidivism associated with juvenile incarceration of 49 %

compared to the mean.

The instrumental-variable point estimates with and without controls in columns 6 and 7

(0.26 and 0.23, respectively) are similar to each other but slightly larger than the most restrictive

OLS estimates for adult recidivism.32 However, the loss of precision in the IV estimates means

that they are not statistically-significantly different from these OLS estimates and both can be
32
While the point estimate declines somewhat with the addition of controls, the difference is not statistically sig-
nificant. Further, if the decline suggested that ”strict” judges hear ”tougher” cases, then we would expect a similar
change in magnitude when considering high-school completion. Instead, the magnitude increased when we added
controls to the model for high-school completion. Together, this suggests that any differences in the types of juveniles
who go before stricter judges are not systematically related to the outcomes.

22
characterized as large.33 Moreover, the reduced-form estimates suggest that in practice, the

direct effect of the judge assignment is more moderate in size. Moving from the least to the

most strict judge increases the probability of incarceration as an adult by 3 percentage points, or

9% of the the mean (Table AII).

Overall, these estimates suggest that of the two potential effects of juvenile incarceration on

future criminal activity (deterrence of future criminal activity vs. reductions in human capital

accumulation, social capital and networks, or other factors such as deviant labeling), the latter

dominates.34

We also estimate the impact of juvenile incarceration on adult recidivism by crime type,

given that some types of crime generate larger welfare costs. Specifically, we estimate the impact

of juvenile incarceration on adult recidivism for four types of crime: homicide, violent crime,

property crime and drug crimes. These categories are not exclusive and an individual might

have been incarcerated for more than one type of crime by age 25.35 For each crime type, we

present three sets of results: OLS based on the full CPS, OLS based on the juvenile subsample

and IV based on the juvenile subsample. The results (Table VI) show that in the OLS for the full

CPS sample, those who are incarcerated as juveniles are much more likely to have recidivated for

each of the four types of crime. Limiting the sample to those with a juvenile court case reduces

the estimates considerably though they are still large: those incarcerated are 2.1 percentage

points more likely to be incarcerated for a homicide as an adult (mean = 4%), 6.1 percentage

points more likely to be incarcerated for violent crime (mean = 12%), 4.7 percentage points more

likely to be incarcerated for property crime (mean = 6%) and 7.8 percentage points more likely

to be incarcerated for a drug offense (mean =18%).

The IV estimates are larger, increasing to 3.5 percentage points for homicide (though not

statistically significant), 15 for a violent crime, 14 for property crime, and 10 percentage points
33
We cluster at the level of the community in all regressions. When we cluster at the level of the judge in the 2SLS
regression with full controls, column 7 of Table V, the standard error increases to 0.084, still highly significant.
34
We considered employment and earnings as well, although it is more difficult to link juvenile cases to wage
report data in Illinois that do not include the date of birth. While we find negative point estimates of the effects of
juvenile incarceration on employment, the standard errors are not precise.
35
We present additional results disaggregating the crime types to specific crimes and again find substantial in-
creases across a wide range of offenses (Table AVIII).

23
for drug-related crimes. It is important to note that even though the point estimates more than

double in some cases, the standard errors also increase substantially compared to the OLS stan-

dard errors. The results broken down by type suggest that children incarcerated as juveniles are

not only more likely to recidivate as adults, but that the recidivism is for types of crime that are

both serious and costly.

A potential explanation is that judges may be more likely to incarcerate adults if they have

been incarcerated as juveniles, as juvenile records are not routinely expunged in Cook County.

However, the large estimated effect for adult recidivism for homicide and violent crimes - crimes

for which incarceration is nearly certain, regardless of juvenile incarceration, suggest that this

is not driving the results. In addition, nearly all of the cases that come before the juvenile court

result in a conviction that may affect later sentencing in an adult court; we are considering the

additional effect of juvenile incarceration.36

These results are considerably larger than effects generally estimated for adults, which would

be consistent with juvenile incarceration occurring during a particularly sensitive time in the life

cycle when human and social capital are forming. They are consistent with estimates in Hjal-

marsson (2008) based on survey data and including very detailed characteristics of the juveniles,

including interactions with the criminal justice system. They are not consistent with Hjalmars-

son (2009) who uses a sentencing index with a cutoff score for incarceration in a regression

discontinuity design to identify the impact of juvenile incarceration on juvenile recidivism and

finds that juvenile incarceration reduces juvenile recidivism. One possible explanation for the

different results is that we consider different margins (cases where judges may disagree about

the incarceration decision vs cases near the cutoff). Another is that Hjalmarsson (2009) consid-

ered juvenile recidivism, which does not include serious offenses committed as a juvenile and

transferred to adult courts, while this paper considers adult recidivism and we show below that

many juveniles are transferring out of high school and into an adult correctional facility.37
36
As noted above, our discussions with court officials suggest that stays in the Temporary Detention Facility are
unlikely to be featured in adult proceedings above and beyond the juvenile conviction.
37
That said, when we consider juvenile recidivism our context, we find that our findings are more robust to time-
frame and potential censoring: incarceration within one year of the first hearing is associated with a greater likeli-
hood of re-appearing before the juvenile court in a subsequent case.

24
V.E Heterogeneous Treatment Effects Across Observable Characteristics

In this section we explore potential heterogeneity in the treatment effects. We present OLS and

2SLS estimates stratified by observable child characteristics (Table VII). Differences in the IV re-

sults are suggestive of differential impacts of incarceration on the propensity to complete high

school and adult recidivism. Given the data requirements of the approach, however, differ-

ences across subgroups are rarely statistically significantly different and should be regarded as

suggestive only.

When we characterize juveniles by type of their first offense (violent vs. non-violent), the

OLS estimates of the impact of juvenile incarceration on high school completion are similar for

the two types, but when we instrument, the negative impact of incarceration increases in mag-

nitude for the non-violent for whom the IV estimate is roughly double the estimate based on the

whole sample (Table VII panels 1 and 2). In contrast, the IV estimate of high school completion

for juveniles accused of a violent crime are much smaller in magnitude and insignificant. One

interpretation of these results is that the effects of juvenile incarceration on high school com-

pletion are larger for those at the margin of incarceration in contrast to those most surely to be

incarcerated. This is consistent with results in which we split the sample based on juveniles’

predicted probability of juvenile incarceration (estimated by a probit with the full set of con-

trols.) The negative effect of incarceration on high school graduation is much greater for those

with a lower propensity of juvenile incarceration in both the OLS and IV settings (Panels 3 and

4 of Table VII).

With respect to the adult incarceration effects, we also find larger effects of juvenile incarcer-

ation on recidivism (for any crime and for a violent crime) for those with a lower propensity of

juvenile incarceration, similar to the high school completion results. However, juveniles charged

with a violent first crime are more likely to recidivate for any crime, and much more likely to

recidivate for a violent crime relative to those charged with a non-violent crime, contrary to the

high school completion results.

The impact of incarceration on high school completion and adult recidivism also varies with

25
juvenile characteristics such as age (Table VII panels 5 and 6).38 The overall effects are largely

coming from juveniles aged 15-16, perhaps because the incarceration occurs during a point in

the life cycle when dropping out of school is possible. Meanwhile, the impact of incarceration

is qualitatively similar for those with and without special-education needs.39

That stronger estimated effects of juvenile incarceration on high school completion for some

groups are not necessarily accompanied by stronger effects on adult incarceration suggests that

the impact of juvenile incarceration on adult incarceration is not working entirely through the

negative impact on high school completion. This is not surprising, as we expect incarceration

to affect a juvenile in many ways, including impacts on social capital and networks or ”deviant

labeling”, in addition to any effect on high school graduation. Still, to gauge the potential mag-

nitude of the high-school completion channel, consider that Lochner and Moretti (2004) found

that among African Americans, high school completion results in an 8 percentage-point decline

in the likelihood of being in jail as an adult (the point estimates for whites are lower and less

precise, but not significantly different from the estimates for blacks). Based on this, we calcu-

late that of the 20 percentage point increase in adult incarceration, only 5% comes from the 13

percentage point decrease in high school completion.40

One caveat is that Lochner and Moretti (2004) base their analysis on the 1960, 70 and 80 Cen-

suses. Since then, the labor market return to high school completion has increased significantly.

Between 1980 and 2000, Deschenes (2006) estimates that the causal return to a year of single

year of schooling increased by as much as 40%. As such, it is likely that the causal impact of

education on crime has likewise increased over this period which would result in a larger role

for high school completion in explaining the impact of juvenile incarceration on adult crime. In
38
We stratify by gender and race as well. Of the 37,692 juveniles in the sample, less than 6000 are female and the
results for females, while large in magnitude with respect to high school completion in particular, are very imprecise.
With respect to race, the main results are similar to those found for African Americans, the point estimates for
high school graduation are larger in magnitude for white and Hispanic juveniles. For adult incarceration, the point
estimate is particularly large (and imprecise) for Hispanic juveniles (Table AV).
39
In Appendix Table VII we present results of regressions for additional subsamples defined by gender and race.
The results show that the main results stem from male offenders, whereas the results for female offenders (a much
smaller subset of the data) are noisier.
40
There is a related literature on the relationship between school days and criminal activity among juveniles (see
Jacob and Lefgren, 2003; Luallen, 2006) in which they explore the incapacitating effect of school attendance on juve-
nile criminal activity.

26
any event, the results suggest that for juveniles on the margin of incarceration, such detention

appears to negatively affect the human and social capital formation in more ways than we can

measure through high school completion and adult incarceration.

In summary, the results suggest that across different groups of children, juvenile incarcera-

tion is associated with lower high school completion and higher adult recidivism. In general,

the high school completion results show large differences across type, with greater effects com-

ing from those less likely to be incarcerated as a juvenile. The adult recidivism results do not

display the same pattern, but show large effects across types.

V.F Exploring Potential Mechanisms

To further examine the potential mechanisms behind our results (behavioral changes or deviant

labeling), we consider a number of additional analyses.

First, we examine the extent to which juveniles incarcerated for even a relatively short period

of time (one to two months) ever return to school. We find that 62% of all children with a spell in

the school located in the detention facility never return to a CPS school. Table VIII presents the

results of more formal OLS and IV analyses that show that juvenile incarceration significantly

reduces the likelihood that the student is observed in school one year after the initial hearing

(-0.215 compared to a mean of 0.67). This can be interpreted as either a change in the behavior

of the juvenile for whom catching up represents a significant barrier to return, or an act on the

part of the school actively discouraging a previously incarcerated student from returning.

To explore this further, we examine whether students who do return to a CPS school are

more likely to transfer to another school, which we interpret as evidence that the schools are

treating incarcerated juveniles differently. In our data, only 28 % of juveniles who leave the

school in the detention facility returned to the same CPS school in which they were enrolled

prior to incarceration, while 10% transferred to another school within the CPS. We also consid-

ered whether incarceration increases the likelihood of transferring to another CPS school after

the initial hearing (outside of the school located in the detention facility or in the Cook County

Jail). This analysis requires us to limit our sample to the 18,195 juveniles whom we observe in

27
high school prior to juvenile incarceration, reducing our power considerably. When we do, in

the OLS regressions with a full set of controls we find that juvenile incarceration is positively but

only slightly related to high school transfer in the years after the initial hearing (0.055 relative

to a mean of 0.242), but when we instrument for juvenile incarceration, the effect is negative al-

though very imprecise (Table VIII).41 Given the imprecision it is difficult to make strong claims,

but we did not find evidence consistent with juvenile incarceration leading to high school trans-

fers. When we consider transfers to an adult correctional facility, juvenile incarceration is found

to lead to significant increases in this outcome.

In a second analysis we examine whether juveniles who spend time incarcerated are more

likely to be classified as a special education student upon release. For this analysis we must

limit our sample to those for whom we can observe special-education status in the year after

the initial hearing (79 % of our sample, n=29,794). There is no relationship between juvenile

incarceration and whether one is designated special education (Table VIII), but interestingly,

incarceration is associated with a change in the source of the disability. Those incarcerated are

more likely to be characterized as having an emotional or behavioral disorder and equally less

likely to be characterized as having a learning disability. The estimated effects are larger (in

absolute magnitude) in the IV regressions, though standard errors also increase (the estimates

remain significant at conventional levels). Again, the re-classification could be due to changes in

behavior on the part of the juvenile or a change in the way that the school system labels students

who have been incarcerated.42

Another test involves looking at the impact of juvenile incarceration on crimes for which it is

reasonable to assume that being labeled as having spent time incarcerated as a juvenile should

have little impact on arrest or incarceration: homicide and other violent crimes. We found (Table
41
We also examined transfers to ”Alternative High Schools” outside of the criminal justice system, but only 1.5%
are found to do so and we do not find a statistically-significant relationship between juvenile incarceration and this
(relatively rare) outcome.
42
As noted previously, a limitation of the database is that data on suspensions or disciplinary infractions, as well
as truancy, do not extend back to our time period. Nevertheless, outcomes based on these measures would also
reflect actions by the juvenile and school. Test score data are available, but assessing the impact on test scores is
complicated by the fact that so many of the juveniles never return to school, and many of these juveniles managed
not to take an exam in years when they are in school. Those with a panel of test scores constitute a very selected
sample of juveniles.

28
VI) that juvenile incarceration does result in greater likelihood of being incarcerated as an adult

for these serious crimes, consistent with changes in criminal behavior upon release.

In sum, these results are consistent with both changes in juvenile behavior and changes in

institutions’ treatment of juveniles explaining the large effects of juvenile incarceration on high

school completion and recidivism.

V.G Additional Tests of Robustness

A concern when using judge fixed effects as instruments is that the monotonicity assumption

may fail.43 For example, some judges could be particularly strict for only a subset of offenses,

such as violent crimes, and these judges could be relatively lenient for, say, property crimes.

To investigate this possibility, we categorized the juvenile offenses into four mutually exclusive

groups violent, property, drug, and other. We found that judges who are strict for violent crimes

tend to be strict for other offense types as well.44 However, we still re-calculate the instrument

for each judge x offense type, thereby relaxing the monotonicity assumption. The results (Table

AIV) show similar impacts for high-school graduation and adult incarceration. We take this

as strong evidence that this potential failure of the monotonicity assumption is not driving the

main results. We also find similar results when we allow the incarceration rate to vary within

judge but across cases with African-American and non-African-American defendants.

As a second robustness check, we allow the fixed effects within which juveniles are com-

pared to vary. Specifically, we include fixed effects defined at the level of the community, com-

munity x year, community x weapon, Census tract, tract x year, tract x weapon, and finally tract
43
Juvenile incarceration is monotonically increasing in the leave-out mean of the judge’s incarceration rate, which
provides some evidence that the monotonicity assumption may be satisfied. Further, we investigated whether treat-
ment effects differed across judges in an effort to estimate marginal treatment effects (Heckman and Vytlacil, 2005,
Doyle 2008). We found that these estimates were too imprecise to explore variation across judges. The point es-
timates suggested that the negative effects for high-school completion and adult incarceration were more likely to
stem from judges with higher incarceration rates, where the marginal juveniles are likely to be cases with unobserv-
able characteristics associated with a lower likelihood of incarceration (compared to the margins relevant among the
more lenient judges).
44
The relationship is not 1-1, however, which is why it is useful to estimate effects using the re-calculated instru-
ment. In particular, in a regression of the judge’s violent-crime incarceration rate on the judge’s property-crime
incarceration rate within the usual fixed-effect cells, we find a coefficient of 0.84 (s.e.=0.10), for drug crimes the
coefficient is 0.68 (s.e.=0.11) and for other crimes the coefficient is 0.64 (s.e. = 0.09).

29
x weapon x year (Table AV). The results are remarkably stable across these different types of

fixed effects.

In a third set of robustness checks, we change how we treat transfers in our high school grad-

uation analysis. Previously we define graduation only if the records in the public school data

indicate graduation with certainty and those who transferred (16% of the sample) were consid-

ered non-graduates. In Table AVI, Panel A, we simply remove from the sample all transfers for

whom high school completion is not known. The results are similar though slightly larger than

the main set of estimates. In Panel B, we keep the full sample but define as the outcome an indi-

cator equal to 1 if the student is one of these 16% of transfers. The estimates are positive but not

significant. Last, we explore whether juvenile incarceration affects transfers to a private school

or to another school outside of Chicago, which may result in better educational outcomes. We

find that juvenile incarceration reduces the likelihood of transferring in this way, although the

estimates are not statistically significant.45

We also consider the robustness of the results to changes in calculation of our instrument.

We drop cases with a missing judge at the first hearing, trim the instrument of extreme values,

and estimate the instrument using a probit (Table AVII), and the results are unchanged. We also

explore the extent to which the effects are driven by ”swing judges” who preside over one fifth

of our cases. When we exclude swing judges, the results are unchanged.

V.H Length of Stay

Last, we do not observe length of stay in the detention facility, but we do observe a proxy: the

length of time spent in the high school located in the facility. This is just a proxy because stu-

dents could drop out of high school, and they do not appear in our data during the summer.

Table AVIII shows that length of stay is not related to our instrument: ”strict” judges in terms of

incarceration rates are not associated with longer stays, conditional on entering detention. The

sign of the estimate suggests shorter stays for those assigned to strict judges, consistent with
45
In 2SLS models with full controls, the estimated effect of juvenile incarceration on transfer to a private school is
-0.019 (s.e. 0.026); for transfer outside of Chicago is -0.070 (s.e. 0.036)

30
these judges incarcerating less serious offenders on average, but the magnitude is small.46 Fur-

ther, the leave-out mean of the judge’s length of stay is not predictive of a juvenile’s length of

stay when we control for the judge’s incarceration rate. Not surprisingly, then, when we esti-

mate the impact of length of stay (including zeros) on our outcomes, with the instrument being

the judge’s leave-out mean length of stay in the year after initial hearings, we find estimates that

are similar to moving from 0 days to the mean number of days in the facility (Table AVIII).

We considered one final specification - dividing cases into short vs. long stays (defined by

the median in our sample) and then estimating the effect of a short or long stay (compared to

no stay) on these outcomes. For this specification, our instrumental variables are the judge’s

leave-out means of short and long stays. While the point estimates for the recidivism outcome

suggest larger effects for longer stays, statistically the estimated effects of the two stay lengths

are the same due to large standard errors. We conclude that we do not have the power to detect

whether length of stay matters in these data, and we interpret our main results as the impacts of

a typical stay in incarceration (averaging 42 days according to our proxy) on future outcomes.

Future work exploring the impact of length of incarceration as a juvenile on future outcomes is

needed.

VI Conclusions

Juvenile incarceration is expensive, with expenditures on juvenile corrections totaling $6 billion

annually in the US, and the average annual (direct) cost of a incarcerating a juvenile topping

$88,000 (Mendel, 2011). If juvenile incarceration either enhanced human capital accumulation or

deterred future crime and incarceration, a tradeoff could be considered. Rather, we find that for

juveniles on the margin of incarceration, such detention leads to both a decrease in high school

completion and an increase in adult incarceration. In exploring the mechanisms behind these

effects, we find that once incarcerated, a juvenile is unlikely to ever return to school, suggesting

that even relatively short periods of incarceration can be very disruptive and have severe long-
46
While imprecise, the point estimate suggests that a 2 standard deviation increase in our instrument (0.08) is
associated with 2 fewer days in detention, compared to a mean of 100 days.

31
term consequences for this population. Moreover, for those who do return to school, they are

more likely to be classified as having a disability due to a social or behavioral disorder, likely

reducing the probability of graduation even among those who do return to school and possibly

increasing the probability of future criminal behavior.

Our results imply that it may be welfare enhancing to use alternatives to juvenile incarcer-

ation. Illinois has more recently adopted an array of such policies, including electronic mon-

itoring and well-enforced curfews that serve as substitutes for juvenile incarceration. These

substitutes have been growing in popularity across the U.S. Our results suggest that their con-

tinued expansion have the potential to increase high school graduation rates and reduce the

likelihood of adult crime. In addition to reducing juvenile incarceration, policies that address

the low rates at which juveniles return to school upon release by providing additional support

and resources for these at-risk juveniles may also be effective in reducing the negative impact of

incarceration on human capital accumulation and other outcomes.

In contrast to the increasing adoption of alternatives to incarceration that reduce juvenile

incarceration, many states have adopted policies of increasing police presence in schools which

has led to an increase in juvenile arrests for relatively mild infractions. If this leads to an increase

in juvenile detention, which seems likely, then the continued expansion of this policy has the

potential to reduce high school graduation rates for those directly affected.

To consider the full set of costs and benefits of policies affecting juvenile arrest and incarcer-

ation, one must also consider the potential reduction in crime due to the incapacitation effect

of incarceration as well as the deterrent effects of strict punishment on the criminal activity of

other youths. Regarding incapacitation, to the extent that alternatives such as strict curfews

or electronic monitoring also serve to incapacitate, this should be less of a concern. Regarding

deterrence, evidence suggests that juveniles’ criminal propensity is particularly inelastic with

respect to penalties (Lee and McCrary, 2006), which implies that this may be of second order

importance compared to the large decrease in high school completion and increase in adult in-

carceration found here. If this is the case, then the results suggest that a continued move toward

less restrictive juvenile sentencing would increase human capital accumulation and lower the

32
propensity of these juveniles to become incarcerated as adults without an increase in juvenile

crime.

33
VII References

Bayer, Patrick, Randi Hjalmarsson, and David Pozen, “Building Criminal Capital Behind

Bars: Peer Effects in Juvenile Corrections,” Quarterly Journal of Economics, 124(2009), 105-47.

Becker, Gary, “Crime and Punishment: An Economic Approach,” Journal of Political Economy,

76 (1968), 169-217.

Bernburg, Jon Gunnar, Marvin D. Krohn, and Craig J. Rivera, “Official Labeling, Criminal

Embeddedness and Subsequent Delinquency: A Longitudinal Test of Labeling Theory,” Journal

of Research in Crime and Delinquency, 43(2006), 67- 88.

Bernburg, Jon Gunnar and Marvin D. Krohn, “Labeling, Life Chances, and Adult Crime:

The Direct and Indirect Effects of Official Intervention in Adolescence on Crime in Early Adult-

hood,” Criminology, 41(2003), 1287-1318.

Bjorklund, Anders and Robert Moffitt, ”The Estimation of Wage Gains and Welfare Gains in

Self-Selection Models,” The Review of Economics and Statistics, 69(1987), 42-49.

Brennan, Patricia A. and Sarnoff A. Mednick, ”Learning Theory Approach to the Deterrence

of Criminal Recidivism,” Journal of Abnormal Psychology, 103(1994), 430-440.

Cameron, Stephen and James J. Heckman, ”Determinants of Young Male Schooling and

Training Choices,” NBER Working Paper No. 4327, 1993.

Chang, Tom, and Antoinette Schoar, “Judge Specific Differences in Chapter 11 and Firm

Outcomes,” MIT Working Paper, 2008.

Civitas, “Comparisons of Crime in OECD Countries,” 2012. Accessed at http://www.

civitas.org.uk/crime/crime_stats_oecdjan2012.pdf.

Cook, Philip and Songman Kang, “Birthdays, Schooling and Crime: New Evidence on the

Crime-Dropout Nexus,” NBER Working Paper No. 18791, 2013.

Cunha, Flavio, James J. Heckman, and Lance Lochner, ”Interpreting the Evidence on Life

Cycle Skill Formation,” Handbook of the Economics of Education, Elsevier (2006).

Currie, Janet, and Erdal Tekin, “Does Child Abuse Lead to Crime?” NBER Working Paper

No. 12171, 2006.

34
 De Li, Spencer, “Legal Sanctions and Youths’ Status Achievement: A Longitudinal Study,”

Justice Quarterly, 16 (1999), 377-401.

Deschenes, Olivier, “Unobserved Ability, Comparative Advantage, and the Rising Return to

Ability in the US,” Working Paper, 2006.

Di Tella, Rafael and Ernesto Schargrodsky,“Criminal Recidivism after Prison and Electronic

Monitoring,” Journal of Political Economy 121(2013), 28-73.

Dobbie, William and Jae Song, “Debt Relief and Debtor Outcomes: Measuring The Effects of

Consumer Bankruptcy Protection,” Harvard University mimeo, 2013.

Doyle, Joseph,“Child Protection and Adult Crime,” Journal of Political Economy, 116(2008),

746-770.

Drago, Francesco, ROberto Galbiati, and Pietro Vertova, “The Deterrent Effects of Prison:

Evidence from a Natural Experiment,” Journal of Political Economy, 117(2009), 254-270.

Forrest, C.B., Tambor, E., Riley, A.W., Ensminger, M.E. and Starfield, B. , “The Health Profile

of Incarcerated Male Youths,” Pediatrics, 105(2000), 286-291.

Freeman, Richard B., “Crime and the Employment of Disadvantaged Youths,” in Urban Labor

Markets and Job Opportunities G. Peterson and W. Vroman, eds. (Washington DC: The Urban

Institutes Press, 1992).

–, “Why do so Many Young American Men Commit Crimes and What Might We Do About

It?” Journal of Economic Perspectives, 10(1996), 25-42.

General Accounting Office, Growing Inmate Crowding Negatively Affects Inmates Staff and In-

frastructure, 2012.

Glaze, Lauren E. and Erika Parks, ”Correctional Populations in the United States, 2011,”

Bureau of Justice Statistics Bulletin NCJ 239972, 2012.

Glaeser, Edward and Bruce Sacerdote, “Why is There More Crime in Cities?” Journal of

Political Economy, 107(1999), s225-s258.

Goerge, R. M., Van Voorhis, J. and B.J. Lee, “Illinois’s Longitudinal and Relational Child and

Family Research Database,” Social Science Computer Review, 12(1994), 351-65.

Gottfredson, Don, Effects of Judges’ Sentencing Decisions on Criminal Careers Washington, D.C.

35
National Institute of Justice: Research in Brief, 1999.

Granovetter, Mark, Getting a Job: A Study of Contacts and Careers, (Chicago: University of

Chicago Press, 1995).

Green, Donald and Daniel Winik, “Using Random Judge Assignments to Estimate the Effects

of Incarceration and Probation on Recidivism among Drug Offenders,” Criminology, 48(2010),

10-23.

Grogger, Jeffrey, ”The Effect of Arrests on the Employment and Earnings of Young Men,”

Quarterly Journal of Economics, 110(1995), 51-72.

Hazel, Neal, Cross-National Comparison of Youth Justice, (London: Youth Justice Board for

England and Wales, 2008).

Heckman James J. and Edward Vytlacil, “Structural Equations, Treatment Effects, and Econo-

metric Policy Evaluation,” Econometrica, 73(2005), 669-738.

Hjalmarsson, Randi, “Criminal Justice Involvement and High School Completion,” Journal

of Urban Economics, 63 (2008), 613-630.

–, “Juvenile Jails: A Path to the Straight and Narrow or Hardened Criminality?” Journal of

Law and Economics, 52(2009), 779-809.

Imbens, Guido W. and Joshua D. Angrist, “Identification and Estimation of Local Average

Treatment Effects,” Econometrica, 62(1994), 467-475.

Jacob, Brian and Lars Lefgren, “Are Idle Hands the Devil’s Workshop? Incapacitation, Con-

centration, and Juvenile Crime,” American Economic Review, 93(2003), 1560 -1577.

Kashani, J.H., Manning, G.W., McKnew D.H., Cytryn, L., Simonds, J.F. and Wooderson, P.C.,

“Depression Among Incarcerated Delinquents,”Psychiatry Resources, 3(1980), 185-191.

Kennelman, Tracey, Justice Expenditure and Employment Extracts 2009 (preliminary), Bureau of

Justice Statistics, NCJ 237913, 2012.

Kerley, Kent R., Benson, Michael L., Lee, Matthew R., and Francis Cullen, “Race, Criminal

Justice Contact, and Adult Position in the Social Stratification System,” Social Problems, 51(2004),

549568.

Kling, Jeffrey, “Incarceration Length, Employment, and Earnings,” American Economic Re-

36
view, 96(2006), 863-876.

Kling, Jeffrey and John Tyler, “Prison-Based Education and Reentry into the Mainstream

Labor Market,” in Barriers to Reentry? The Labor Market for Released Prisoners in Post-Industrial

America, Shawn Bushway, Michael Stoll, and David Weiman eds. (New York: Russell Sage

Foundation Press, 2007).

Kolesar, Michal, Raj Chetty, John Friedman, Edward Glaeser and Guido Imbens, ”Identifica-

tion and Inference with Many Invalid Instruments,” NBER Working Paper No. 17519, 2011.

Landerso, Rasmus, ”Does Incarceration Length Affect Labor Market Outcomes for Violent

Offenders?” The Rockwool Foundation Research Unit. Study Paper Number 39, 2012.

Lee, David and Justin McCrary,“Crime Punishment and Myopia,” NBER Working Paper No.

11491, 2005

Levitt, Steven, “Juvenile Crime and Punishment,” The Journal of Political Economy, 106(1998),

1156-1185.

Levitt, Steven, “Understanding why Crime Fell in the 1990s: Four Factors that Explain the

Decline and Six that do Not,” The Journal of Economic Perspectives, 18(2004), 163-190.

Lochner, Lance and Enrico Moretti, “The Effect of Education on Crime: Evidence from Prison

Inmates, Arrests and Self-Reports,” American Economic Review 94(2004), 155-189.

Lochner, Lance,“Education, Work, and Crime: A Human Capital Approach,” International

Economic Review, 45(2004), 811843.

Lott, John, “An Attempt at Measuring the Total Monetary Penalty from Drug Convictions:

the Importance of an Individual’s Reputation,” Journal of Legal Studies, 21 (1992a), 159-187.

–,“Do We Punish High Income Criminals too Heavily?” Economic Inquiry, 30 (1992b), 583-

608.

Luallen, Jeremy, “School’s Out...Forever: A Study of Juvenile Crime, At-Risk Youths and

Teacher Strikes,” Journal of Urban Economics, 59(2006),75-103.

Machin, Stephen, Oliver Marie, and Suncica Vujic,“The Crime Reducing Effect of Educa-

tion,” Economic Journal, 121(2011), 463484.

Maestas, Nicole, Kathleen Mullen, and Alexander Strand, ”Does Disability Insurance Re-

37
ceipt Discourage Work? Using Examiner Assignment to Estimate Causal Effects of SSDI Re-

ceipt,” American Economic Review, 103(2013), 1797-1829.

Mendel, Richard A., No Place for Kids: The Case for Reducing Juvenile Incarceration, (Baltimore:

The Annie E. Casey Foundation, 2011).

Murray, Charles and Louis Cox, Beyond Probation: Juvenile Corrections and the Chronic Delin-

quent, (Beverly Hills: Sage Publications, 1979).

Na, Chongmin and Denise C. Gottfredson, “Police Officers in Schools: Effects on School

Crime and the Processing of Offending Behaviors,” Justice Quarterly, 30(2011), 619-650.

Neal, Derek and Armin Rick, “The Prison Boom and the Lack of Black Progress after Smith

and Welch,” NBER Working Paper No. 20283, 2014.

Office of Juvenile Justice and Delinquency Prevention, Census of Juveniles in Residential Place-

ment 2010, Washington, D.C.: OJJDP, 2014.

Peters, Clark, Haywood, Thomas, Lee, Bong Joo, and Ada Skyles, Case Processing and Services

to Children in the Juvenile Justice Division of the Cook County Juvenile Court Chicago: Chapin Hall,

2002.

Raphael, Steven and Michael Stoll, Why Are So Many Americans in Prison? (New York:

Russell Sage Foundation, 2013).

Sampson, Robert and John Laub, “A Life Course Theory of Cumulative Disadvantage and

the Stability of Delinquency,” Advances in Criminological Theory, 7 (1997), 133-161.

Sweeten, Gary, “Who Will Graduate? Disruption of High School Education by Arrest and

Court Involvement,” Justice Quarterly, 23 (2006), 462-480.

Smith, Linda and Ronald Akers, “A comparison of recidivism of Florida’s Community Con-

trol and Prison: A five-year Survival Analysis.” Journal of Research in Crime and Delinquency,

30(1993): 267-292.

Spohn, Cassia, and David Holleran, “The Effects of Imprisonment on Recidivism Rates of

Felony Offenders: A Focus on Drug Offenders,” Criminology, 40 (2002), 329-357.

Stock, James, Jonathan Wright, and Motochiro Yogo, ”A Survey of Weak Instruments and

Weak Identification in Generalized Method of Moments,” Journal of Business and Economic Statis-

38
tics, 20(2002), 518-529.

Tanner, Julian, Scott Davies, and Bill O’Grady,“Whatever Happened to Yesterday’s Rebels?

Longitudinal Effects of Youth Delinquency on Education and Employment,” Social Problems, 46

(1999), 250-274.

Waldfogel, Joel,“The Effect of Criminal Conviction on Income and the Trust Reposes in the

Workmen,” Journal of Human Resources, 29 (1994) 62-81.

Walmsley, Roy, “World Prison Population List (tenth edition),” International Centre for Prison

Studies, London, United Kingdom, 2013.

Western, Bruce and Katherine Beckett, ”How Unregulated is the U.S. Labor Market? The

Penal System as a Labor Market Institution,” American Journal of Sociology, 104(1999), 1030-60.

Western, Bruce, Jeffrey Kling, and David Weiman,“The Labor Market Consequences of In-

carceration,” Crime and Delinquency, 47(2001), 410-427.

Wilson, David B, Catherine Gallagher, and Doris MacKenzie, “A Meta Analysis of Corrections-

Based Education, Vocation and Work Programs for Adult Offenders,” Journal of Research in Crime

and Delinquency, 37(2000),347-368.

39
 Table I: Sample Means

Full CPS Sample Juvenile Court Sample

Not Incarcerated Incarcerated
(1) (2) (3)
Juvenile Characteristics
Incarcerated as a Juvenile 0.021 0.00 1.00

Male 0.51 0.81 0.93

African American 0.55 0.75 0.77
Hispanic 0.27 0.17 0.15
White 0.14 0.07 0.07
Other 0.04 0.01 0.01
Special Education in Eighth Grade 0.12 0.22 0.31
Birth year 1977 1978 1978
US Census Tract Poverty Rate 0.221 0.28 0.29
Age at Offense N/A 14.81 14.27

Charges
Aggravated Assault N/A 0.13 0.11
Burglary 0.11 0.13
Drug Law Violation 0.20 0.20
Larceny Theft 0.05 0.03
Car Theft 0.11 0.11
Robbery 0.06 0.07
Simple Assault 0.10 0.06
Vandalism 0.05 0.04
Weapons Offense 0.11 0.17
Other Offense 0.09 0.08

Outcomes:
Graduated High School 0.40 0.12 0.03
Incarcerated as an Adult by Age 25 0.064 0.28 0.49

Sample Size 440797 29141 8551

Notes: This table reports summary statistics for the Chicago Public School Student Database (1990-2006) for students in a
CPS school in 8th grade and at least 25 years old in 2008; columns (1) and (2) restrict this sample to those linked to the
Juvenile Court of Cook County Delinquency Database, including cases from 1990-2000, as described in the text.
 Figure I:
Distribution of Z: Judge Incarceration Rate
Residualized Raw
12000

10000

8000

6000

4000

2000

0
0 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 0.19 0.2 0.21 0.22 0.23 0.24
Notes : These histograms display the distribution of the leave-one-out mean incarceration rate for the first judge in the first case in the linked
Chicago Public School - Juvenile Court of Cook County data including cases from 1990-2000 as described in the text. The residualized measure
was calculated from a regression model with full controls listed in Table 1, including indicators for each year of age at the time of the offense and
community x weapons offense x year indicators.
 Table II: Instrument vs. Juvenile Characteristics

Z Distribution:
Middle vs. Bottom Top vs. Bottom
Bottom Tercile Middle Tercile Top Tercile p-value p-value

Z: First Judge's Leave-out Mean

Incarceration Rate in first cases 0.062 0.094 0.147 (0.000) (0.000)

Juvenile Characteristics:
Male 0.827 0.830 0.833 (0.561) (0.311)
African American 0.724 0.737 0.742 (0.096) (0.249)
Hispanic 0.189 0.176 0.172 (0.061) (0.272)
White 0.078 0.079 0.078 (0.833) (0.957)
Other race/ethnicity 0.009 0.008 0.007 (0.352) (0.345)
Special education 0.241 0.237 0.252 (0.549) (0.130)
US Census Tract Poverty Rate 0.264 0.265 0.265 (0.572) (0.696)
Age at offense 14.8 14.8 14.8 (0.437) (0.434)

P(Juvenile Incarceration | X) 0.219 0.221 0.220 (0.251) (0.516)

Observations 37692
Notes : This table reports summary statistics for the linked Chicago Public School - Juvenile Court of Cook County data including cases
from 1990-2000 as described in the text. p-values reported in parentheses were calculated from separate regression models of each
characteristic on indicators that the judge's incarceration rate (Z) was in the middle or top tercile along with community x weapon x year
fixed effects using standard errors clustered at the community level. P(Juvenile Incarceration | X) is a predicted propensity for ever being
incarcerated as a juvenile using a probit model that employs the juvenile characteristics listed above, including indicators for each year of
age at the time of the offense.
 Table III: First Stage

Dependent Variable: Juvenile Incarceration

Model: OLS
(1) (2) (3)
First Judge's Leave-out Mean Incarceration Rate 1.103 1.082 1.060
among First Cases (0.102) (0.095) (0.097)

Demographic controls No Yes Yes

Court controls No No Yes

Observations 37692
Mean of Dependent Variable 0.227
Notes : This table reports the first-stage relationship between juvenile incarceration and the instrument:
the judge's incarceration rate using the linked Chicago Public School - Juvenile Court of Cook County
data including cases from 1990-2000 as described in the text.. All models include community x weapons-
offense x year-of-offense fixed effects. Demographic controls include indicators for 4 age-at-offense
categories, 4 race/ethnicity categories, sex, special education status, and the 2000 US Census tract family
poverty rate. Court controls include 9 offense categories, indictors for 7 risk-assessment index categories,
and whether the first judge assigned was missing. Standard errors are reported in the parentheses and are
clustered at the community level.

r
 Table IV: Juvenile Incarceration & High-School Graduation

Dependent Variable: Graduated High School

Full CPS Sample Juvenile Court Sample

Inverse Propensity
OLS OLS OLS OLS 2SLS 2SLS
Model: Score Weighting
(1) (2) (3) (4) (5) (6) (7)
Juvenile Incarceration -0.389 -0.292 -0.391 -0.088 -0.073 -0.108 -0.125
(0.0066) (0.0065) (0.0055) (0.0043) (0.0041) (0.044) (0.043)
Demographic controls No Yes Yes No Yes No Yes
Court controls N/A N/A N/A No Yes No Yes
Observations 440797 440797 420033 37692
Mean of Dependent Variable 0.428 0.428 0.433 0.099
Notes : This table reports the relationship between juvenile incarceration and graduation from Chicago Public Schools. Columns (1)-(3)
include all students in Chicago Public Schools in 8th grade during 1990-2006 and at least 25 by 2008. Columns (1) and (2) include
community fixed effects, while Column (2) also includes indicators for race, sex, special education status, each year of birth, and the 2000 US
Census tract family poverty rate. Column (3) used the same controls and community indicators to calculate the propensity score using a
probit model, estimated on a subsample where probit estimation is possible (where there is variation in juvenile incarceration within cells).
Columns (4)-(7) use the linked Chicago Public School - Juvenile Court of Cook County data including cases from 1990-2000 as described in
the text. These models include community x weapons-offense x year-of-offense fixed effects. Demographic controls include those listed for
Column (2). Court controls include 9 offense categories, indictors for 7 risk-assessment index categories, and whether the first judge assigned
was missing. Standard errors are reported in the parentheses and are clustered at the community level. The propensity score standard errors
were calculated using 200 bootstrap replications.
 Table V: Juvenile Incarceration & Adult Crime

Dependent Variable: Entered Adult Prison by Age 25

Full CPS Sample Juvenile Court Sample

Inverse
Propensity
OLS OLS OLS OLS 2SLS 2SLS
Score
Weighting
(1) (2) (3) (4) (5) (6) (7)
Juvenile Incarceration 0.407 0.350 0.219 0.200 0.155 0.260 0.234
(0.0082) (0.0064) (0.013) (0.0072) (0.0073) (0.073) (0.076)
Demographic controls No Yes Yes No Yes No Yes
Court controls N/A N/A N/A No Yes No Yes
Observations 440797 440797 420033 37692
Mean of Dependent Variable 0.067 0.067 0.057 0.327
Notes : This table reports the relationship between juvenile incarceration and imprisonment in an adult facility by the age of 25. Columns (1)-(3)
include all students in Chicago Public Schools in 8th grade during 1990-2006 and at least 25 by 2008. Columns (1) and (2) include community
fixed effects, while Column (2) also includes indicators for race, sex, special education status, each year of birth, and the 2000 US Census tract
family poverty rate. Column (3) used the same controls and community indicators to calculate the propensity score using a probit model,
estimated on a subsample where probit estimation is possible (where there is variation in juvenile incarceration within cells). Columns (4)-(7)
use the linked Chicago Public School - Juvenile Court of Cook County - Illinois Department of Corrections data including juvenile cases from
1990-2000 as described in the text. These models include community x weapons-offense x year-of-offense fixed effects. Demographic controls
include those listed for Column (2). Court controls include 9 offense categories, indictors for 7 risk-assessment index categories, and whether the
first judge assigned was missing. Standard errors are reported in the parentheses and are clustered at the community level. The propensity score
standard errors were calculated using 200 bootstrap replications.
 Table VI: Juvenile Incarceration & Adult Crime Type

Dependent Variable: Entered Adult Prison by Age 25 for Crime Type:

Homicide Violent
OLS OLS 2SLS OLS OLS 2SLS
(1) (2) (3) (4) (5) (6)
Juvenile Incarceration 0.051 0.021 0.035 0.138 0.061 0.149
(0.0031) (0.0030) (0.030) (0.0046) (0.0050) (0.041)

Sample Full CPS Juvenile Court Juvenile Court Full CPS Juvenile Court Juvenile Court

Mean of Dep. Var.: JI=0 0.008 0.043 0.043 0.024 0.121 0.121
Observations 440797 37692 37692 440797 37692 37692

Property Drug
OLS OLS 2SLS OLS OLS 2SLS
(1) (2) (3) (4) (5) (6)
Juvenile Incarceration 0.079 0.047 0.142 0.183 0.078 0.097
(0.0040) (0.0038) (0.044) (0.011) (0.0068) (0.052)

Sample Full CPS Juvenile Court Juvenile Court Full CPS Juvenile Court Juvenile Court

Mean of Dep. Var. 0.013 0.060 0.060 0.034 0.176 0.176

Observations 440797 37692 37692 440797 37692 37692
Notes: This table reports the relationship between juvenile incarceration and imprisonment in an adult facility by the age of 25 for
particular types of offenses. The outcomes are not mutually exclusive. All models include full controls listed in Table V. Full CPS models
include community fixed effects. Juvenile Court models include community x weapons offense x year-of-offense fixed effects. Standard
errors are reported in the parentheses and are clustered at the community level.
 Table VII: Effects of Juvenile Incarceration By Case & Child Types

Entered Adult Prison by Age 25 for

Graduated High School Entered Adult Prison by Age 25
Dependent Variable: Violent Offense
OLS 2SLS OLS 2SLS OLS 2SLS
Juvenile Offense: Violent (1) (2) (3) (4) (5) (6)
Juvenile Incarceration -0.080 -0.046 0.140 0.276 0.055 0.219
(0.006) (0.071) (0.010) (0.109) (0.008) (0.080)
Mean of Dependent Variable 0.118 0.118 0.295 0.295 0.121 0.121
Observations 15561
Juvenile Offense: Non-violent
Juvenile Incarceration -0.067 -0.155 0.165 0.200 0.065 0.109
(0.005) (0.042) (0.010) (0.108) (0.006) (0.058)
Mean of Dependent Variable 0.085 0.085 0.349 0.349 0.122 0.122
Observations 22131
Propensity of Juvenile Incarceration <= Median
Juvenile Incarceration -0.080 -0.206 0.116 0.410 0.053 0.211
(0.007) (0.075) (0.013) (0.092) (0.0079) (0.063)
Mean of Dependent Variable 0.125 0.125 0.246 0.246 0.090 0.090
Observations 18,846
Propensity of Juvenile Incarceration > Median
Juvenile Incarceration -0.068 -0.051 0.172 0.056 0.064 0.081
(0.005) (0.054) (0.008) (0.114) (0.0062) (0.058)
Mean of Dependent Variable 0.073 0.073 0.407 0.407 0.152 0.152
Observations 18,846
Age = 13 or 14
Juvenile Incarceration -0.070 -0.096 0.174 -0.189 0.066 0.015
(0.006) (0.075) (0.012) (0.140) (0.009) (0.079)
Mean of Dependent Variable 0.082 0.082 0.343 0.343 0.134 0.134
Observations 11404
Age = 15 or 16
Juvenile Incarceration -0.072 -0.150 0.132 0.435 0.050 0.224
(0.005) (0.056) (0.010) (0.098) (0.006) (0.064)
Mean of Dependent Variable 0.109 0.109 0.314 0.314 0.112 0.112
Observations 23734
Special Education
Juvenile Incarceration -0.055 -0.090 0.181 0.170 0.081 0.169
(0.005) (0.055) (0.012) (0.125) (0.009) (0.098)
Mean of Dependent Variable 0.072 0.072 0.400 0.400 0.159 0.159
Observations 8999
Not Special Education
Juvenile Incarceration -0.079 -0.114 0.146 0.229 0.053 0.129
(0.005) (0.055) (0.008) (0.103) (0.005) (0.061)
Mean of Dependent Variable 0.108 0.108 0.303 0.303 0.110 0.110
Observations 28693
Notes: This table reports results for different subgroups of juveniles. All models include community x weapons offense x year-of-offense fixed effects and full controls as listed in Table
V. Standard errors are reported in the parentheses and are clustered at the community level.
 Table VIII: Intermediate Schooling Outcomes

A. Outcome: High School Transfers

Ever Present in CPS school Transferred to another CPS
Ultimate Transfer: Adult
at least 1 Year after Initial High School in Years After
Correctional Facility
Dependent Variable: Hearing Hearing
OLS 2SLS OLS 2SLS OLS 2SLS
(1) (2) (3) (4) (5) (6)
Juvenile Incarceration -0.025 -0.215 0.055 -0.115 0.127 0.243
(0.0063) (0.069) (0.010) (0.243) (0.006) (0.060)

Mean of Dependent Variable 0.666 0.242 0.175

Observations 37692 18,195 37692

B. Outcome: Special Education Status

Special Education Type Observed in Years after Initial Hearing:
Emotional/Behavioral
Dependent Variable: Any Disorder Learning Disability
OLS 2SLS OLS 2SLS OLS 2SLS
(1) (2) (3) (4) (5) (6)
Juvenile Incarceration -0.024 -0.003 0.027 0.133 -0.040 -0.097
(0.004) (0.037) (0.003) (0.043) (0.004) (0.039)

Mean of Dependent Variable 0.193 0.082 0.085

Observations 29,794
Notes : This table reports the relationship between juvenile incarceration and intermediate schooling outcomes. In Panel A, "ultimate
transfer" means this is the last known whereabouts by CPS; the "transferred high school" sample is restricted to students who were in high
school at the time of the first juvenile-court hearing; and "Ever present in a CPS school at least 1 year after initial hearing" excludes schools
within the detention facility or the Cook County Jail. Panel B is restricted to students where the special education status is observed in years
following the hearing. Any Special Education includes categories not included in the two main categories reported separately. All models
include community x weapons offense x year-of-offense fixed effects and full controls as listed in Table V. Standard errors are reported in
the parentheses and are clustered at the community level.
Juvenile Incarceration, Human Capital and Future Crime:
Evidence from Randomly-Assigned Judges
Anna Aizer∗ and Joseph J. Doyle, Jr. †

January 16, 2015

SUPPLEMENTAL ONLINE APPENDIX

∗
Brown University and NBER. email: anna [email protected]
†
MIT and NBER. email: [email protected].
 Appendix Table AI: Data Sources & Sample Construction

Data Sources:
Chicago Public Schools Student Database 1990-2006
Juvenile Court of Cook County Delinquency Database 1990-2006
Illinois Department of Corrections Adult Admissions and Exits Database 1993-2008

Data are linked together as part of the Chapin Hall Illinois Integrated Database using name, date of birth, and address.

All students in CPS schools in 8th grade:

born between 1971-1983 (at least 25 in 2008)
left CPS schools after 1989 450317

Non-missing US Census tract information 441055

Non-missing Chicago community information 440797

Linked to Juvenile Court of Cook County cases: 41764

Case not transferred to adult court 41757
10 <= recorded age <= 16 40563
Assigned to judges with >=10 cases 40346
Community x weapon x year cell >= 10 cases 37692
Notes Chicago Public Schools (CPS) Student Database 1990-2006 linked to Juvenile Court of Cook County Juvenile
Court of Cook County Delinquency Database 1990-2006 as part of the Chapin Hall Center for Children Illinois
Integrated Database. Resulting dataset includes juvenile cases from 1990-2000.
 Table AII: Reduced Form Estimates

Entered Adult
Graduated High Entered Adult Prison by Age 25
Dependent Variable: School Prison by Age 25 for Violent Offense
OLS OLS OLS
(1) (2) (3)
First Judge's Leave-out Mean Incarceration Rate among First Cases -0.132 0.248 0.158
(0.048) (0.082) (0.046)

Mean of Dependent Variable 0.099 0.327 0.121

Observations 37692
Notes : This table reports the reduced-form relationship between the main outcomes and the instrument: the judge's incarceration rate using
the linked Chicago Public School - Juvenile Court of Cook County - Illinois Department of Corrections data including cases from 1990-
2000 as described in the text. All models include community x weapons offense x year-of-offense fixed effects and full controls as listed in
Table V. Standard errors are reported in the parentheses and are clustered at the community level.
 Table AIII: Juvenile Incarceration & Other Adult Crime Types by Age 25

Dependent Variable: Weapon Offense Robbery

OLS 2SLS OLS 2SLS
(1) (2) (3) (4)
Juvenile Incarceration 0.039 -0.005 0.025 0.065
(0.004) (0.026) (0.003) (0.035)

Mean of Dependent Variable 0.0555 0.0490

Dependent Variable: Assault Motor Vehicle Theft

OLS 2SLS OLS 2SLS
(5) (6) (7) (8)
Juvenile Incarceration 0.018 0.068 0.020 0.066
(0.002) (0.025) (0.002) (0.025)

Mean of Dep. Var. 0.0243 0.0241

Dependent Variable: Burglary

OLS 2SLS
(9) (10)
Juvenile Incarceration 0.020 0.061
(0.003) (0.029)

Mean of Dependent Variable 0.0238

Observations 37692
Notes: This table reports the relationship between juvenile incarceration and imprisonment in an adult facility by
the age of 25 for the five most common offense categories, not including homicide which is shown in Table 6.
The outcomes are not mutually exclusive. Models include community x weapons offense x year-of-offense fixed
effects and full controls as listed in Table V. Standard errors are reported in the parentheses and are clustered at
the community level.
 Table AIV: Alternative Instrument Calculation

A. Models with Community x Subgroup x Year Fixed Effects (FE)

Instrument Calculated by Judge within Cells defined by:

4 Offense Categories African American Juvenile (0/1)

Imprisoned Imprisoned
by Age 25 for by Age 25 for
Dependent Graduated Imprisoned Violent Graduated Imprisoned Violent
Variable: High School by Age 25 Offense High School by Age 25 Offense
2SLS 2SLS 2SLS 2SLS 2SLS 2SLS
(1) (2) (3) (4) (5) (6)
Juvenile -0.104 0.236 0.143 -0.129 0.227 0.121
Incarceration (0.046) (0.076) (0.058) (0.040) (0.073) (0.049)

Mean of Dep.
Variable 0.097 0.333 0.124 0.099 0.326 0.121
Observations 34,872 38,680

B. Models with Community x Subgroup FE & Separate Year FE

Instrument Calculated by Judge within Cells defined by:

4 Offense Categories African American Juvenile (0/1)

Imprisoned Imprisoned
by Age 25 for by Age 25 for
Dependent Graduated Imprisoned Violent Graduated Imprisoned Violent
Variable: High School by Age 25 Offense High School by Age 25 Offense
2SLS 2SLS 2SLS 2SLS 2SLS 2SLS
(1) (2) (3) (4) (5) (6)

Juvenile -0.129 0.162 0.132 -0.132 0.171 0.117

Incarceration (0.042) (0.057) (0.047) (0.039) (0.062) (0.042)

Mean of Dep.
Variable 0.099 0.322 0.120 0.100 0.322 0.120
Observations 40,079 40,079 40,079 40,163 40,163 40,163
Notes : This table reports results for different instrument calculations and fixed effects. Columns (1)-(3)
include community x offense x year-of-offense fixed effects, where the four offense categories are violent,
property, drug, and other. Columns (4)-(6) include community x African-American x year-of-offense fixed
effects, and all models include full controls as listed in Table V. Standard errors are reported in the
parentheses and are clustered at the community level. Sample sizes vary because each model is estimated
using a sample restricted to cells that comprise the fixed effects to at least 10 observations, and judges with at
least 10 observations in those cells.
 Table AV: 2SLS Models with Alternative Fixed Effects

Community FE Community x Year FE Community x Weapon FE Community x Weapon x Year FE

Imprisoned Imprisoned Imprisoned Imprisoned
Graduated by Age 25 Graduated by Age 25 Graduated by Age 25 Graduated by Age 25
High Imprisoned for Violent High Imprisoned for Violent High Imprisoned for Violent High Imprisoned for Violent
Dependent Variable: School by Age 25 Offense School by Age 25 Offense School by Age 25 Offense School by Age 25 Offense
2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
-0.110 0.168 0.134 -0.116 0.238 0.133 -0.111 0.167 0.135 -0.133 0.224 0.149
Juvenile Incarceration
(0.036) (0.062) (0.036) (0.040) (0.072) (0.041) (0.037) (0.063) (0.036) (0.043) (0.075) (0.041)

Mean of Dependent Variable 0.100 0.321 0.120 0.099 0.325 0.121 0.100 0.321 0.120 0.099 0.327 0.121
Observations 40,339 39,382 40,296 37692

Tract FE Tract x Year FE Tract x Weapon FE Tract x Weapon x Year FE

Imprisoned Imprisoned Imprisoned Imprisoned
Graduated by Age 25 Graduated by Age 25 Graduated by Age 25 Graduated by Age 25
High Imprisoned for Violent High Imprisoned for Violent High Imprisoned for Violent High Imprisoned for Violent
Dependent Variable: School by Age 25 Offense School by Age 25 Offense School by Age 25 Offense School by Age 25 Offense
2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS 2SLS
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
-0.102 0.155 0.134 -0.108 0.271 0.173 -0.124 0.144 0.131 -0.178 0.223 0.155
Juvenile Incarceration
(0.038) (0.066) (0.037) (0.049) (0.092) (0.052) (0.041) (0.071) (0.041) (0.054) (0.100) (0.058)

Mean of Dependent Variable 0.100 0.323 0.120 0.097 0.351 0.130 0.099 0.324 0.120 0.095 0.353 0.129
Observations 39561 20983 37592 15242
Notes : This table reports results for different fixed effects. All models include full controls listed in Table V. Sample sizes vary because each model is estimated using a sample
restricted to cells that comprise the fixed effects to at least 10 observations. Standard errors are reported in parentheses, clustered at the community level.
 Table AVI: Robustness Checks for Transfers from CPS

A. Restricted to Students Found to Graduate OR Dropout OR Transfer to Adult Prison

Dependent Variable: Graduated High School

OLS OLS 2SLS 2SLS
(1) (2) (3) (4)
Juvenile Incarceration -0.104 -0.086 -0.138 -0.156
(0.005) (0.005) (0.057) (0.057)
Full controls No Yes No Yes
Mean of Dependent Variable 0.118
Observations 31652

B. Transfers
Do not observe outcomes listed in Panel A: graduation, dropout, or
Dependent Variable: transfer to correctional facility

OLS OLS 2SLS 2SLS

(1) (2) (3) (4)
Juvenile Incarceration 0.020 0.012 0.068 0.076
(0.006) (0.005) (0.049) (0.050)
Full controls No Yes No Yes
Mean of Dependent Variable 0.160
Observations 37692
Notes: This table reports robustness checks for transfers. All models include community x weapon offense x year
fixed effects and full controls as listed in Table V. The sample size in Panel A is smaller because the sample is
restricted to students where the outcomes of graduation, drop out, and transfer to an adult correctional facility, are
observed. Panel B tests whether incarceration is related to the ability to observe those outcomes.
 Table AVII: Additional Robustness Checks and Heterogeneity

Dependent Variable: Graduated High School Entered Adult Prison by Age 25 Entered Adult Prison by Age 25 for
Violent Offense
First judge not missing OLS 2SLS OLS 2SLS OLS 2SLS
Juvenile Incarceration -0.073 -0.137 0.153 0.235 0.064 0.160
(0.005) (0.041) (0.007) (0.077) (0.005) (0.043)
Mean of Dependent Variable 0.098 0.098 0.330 0.330 0.125 0.125
Observations 29239 29239 29239 29239 29239 29239
Trim 1% extremes of Z
Juvenile Incarceration -0.072 -0.126 0.154 0.254 0.061 0.161
(0.004) (0.041) (0.007) (0.075) (0.005) (0.041)
Mean of Dependent Variable 0.098 0.098 0.328 0.328 0.122 0.122
Observations 36,802 36,802 36,802 36,802 36,802 36,802
Two-step probit IV
Juvenile Incarceration -0.103 0.254 0.157
(0.021) (0.090) (0.056)
Mean of Dependent Variable 0.103 0.328 0.125
Observations 36328 37516 36563
African American
Juvenile Incarceration -0.072 -0.063 0.147 0.192 0.054 0.141
(0.005) (0.201) (0.009) (0.090) (0.005) (0.415)
Mean of Dependent Variable 0.099 0.099 0.363 0.363 0.130 0.130
Observations 28524 28524 28524 28524 28524 28524
Hispanic
Juvenile Incarceration -0.077 -0.179 0.169 0.379 0.087 -0.005
(0.007) (0.144) (0.015) (0.191) (0.011) (0.137)
Mean of Dependent Variable 0.099 0.099 0.230 0.230 0.107 0.107
Observations 6192 6192 6192 6192 6192 6192
White
Juvenile Incarceration -0.054 -0.352 0.184 0.151 0.083 0.225
(0.012) (0.142) (0.028) (0.190) (0.019) (0.145)
Mean of Dependent Variable 0.087 0.087 0.186 0.186 0.071 0.071
Observations 2686 2686 2686 2686 2686 2686
Male
Juvenile Incarceration -0.070 -0.094 0.158 0.238 0.064 0.162
(0.004) (0.038) (0.008) (0.084) (0.005) (0.044)
Mean of Dependent Variable 0.088 0.088 0.378 0.378 0.141 0.141
Observations 31702 31702 31702 31702 31702 31702
Female
Juvenile Incarceration -0.094 -0.673 0.082 0.008 0.016 -0.042
(0.015) (0.342) (0.023) (0.215) (0.010) (0.127)
Mean of Dependent Variable 0.159 0.159 0.053 0.053 0.017 0.017
Observations 5990 5990 5990 5990 5990 5990
Notes: This table reports robustness tests as well as results for different subgroups of juveniles. All models include community x weapons offense x year-of-offense fixed effects and full
controls as listed in Table V. Standard errors are reported in the parentheses and are clustered at the community level. The probit model results are marginal effects calculated from a
model that includes the residual from an OLS regression of juvenile detention on the instrument and full controls; both steps also include community x weapons offense x year indicators,
and the standard errors have not been corrected for the additional noise introduced by the generated regressor.
View publication stats

Table AVIII: Proxy for Length of Stay in Juvenile Incarceration

A. Relationship between Judges and Length of Stay

Dependent Variable: Proxy for Length of Stay in Juvenile Incarceration (100s of Days)

Conditional on Length of Stay>0 Including Zero Length of Stay

Model: OLS OLS
(1) (2)
First Judge's Leave-out Mean Incarceration Rate among -0.296 0.315
First Cases (0.451) (0.121)
First Judge's Leave-out Mean Length of Stay within 1 0.258
Year of Initial Case (0.140)

Observations 7990 37692

Mean of Dependent Variable 1.00 0.213

B. Explanatory Variable: Length of Stay Proxy rather than Juvenile Incarceration

Dependent Variable: HS Grad Incarcerated by Age 25

Model: 2SLS 2SLS
(1) (2)

Proxy for Days in Incarceration (including zeros) (100s) -0.234 0.440

(0.187) (0.255)

Leave-out mean LOS within 1

Instrument year of hearing
First-stage F-statistic 7.78

Observations 37692
Mean Proxy Number of Days in Incarceration (100s) 1.00
Mean of Dependent Variable 0.099
Notes : This table presents results using a proxy for length of stay in juvenile incarceration: days in the school located in the
detention facility. Panel A reports first-stage like models, where the dependent variable is the number of days in this school. The
explanatory variables are the instrument used in the main text: the judge's leave-out mean incarceration rate in first cases, and the
judge's leave-out mean of the length of stay in the school within one year first-cases' initial hearings. Panel B presents IV
results, where the instrument is the judge's leave-out mean of the length of stay in the school within one year first-cases' initial
hearings, including zeros. All models include community x weapons offense x year fixed effects and full controls as listed in
Table V. Standard errors are reported in the parentheses and are clustered at the community level.