Influences of Neighborhood Context, Individual History and Parenting Behavior on Recidivism Among Juvenile Offenders

EMPIRICAL RESEARCH

Influences of Neighborhood Context, Individual Historyand Parenting Behavior on Recidivism Among JuvenileOffenders

Heidi E. Grunwald • Brian Lockwood •

Philip W. Harris • Jeremy Mennis

Received: 13 November 2009 / Accepted: 15 February 2010 / Published online: 4 March 2010

� Springer Science+Business Media, LLC 2010

Abstract This study examined the effects of neighbor-

hood context on juvenile recidivism to determine if neigh-

borhoods influence the likelihood of reoffending. Although

a large body of literature exists regarding the impact of

environmental factors on delinquency, very little is known

about the effects of these factors on juvenile recidivism. The

sample analyzed includes 7,061 delinquent male juveniles

committed to community-based programs in Philadelphia,

of which 74% are Black, 13% Hispanic, and 11% White.

Since sample youths were nested in neighborhoods, a hier-

archical generalized linear model was employed to predict

recidivism across three general categories of recidivism

offenses: drug, violent, and property. Results indicate that

predictors vary across the types of offenses and that drug

offending differs from property and violent offending.

Neighborhood-level factors were found to influence drug

offense recidivism, but were not significant predictors of

violent offenses, property offenses, or an aggregated recid-

ivism measure, despite contrary expectations. Implications

stemming from the finding that neighborhood context

influences only juvenile drug recidivism are discussed.

Keywords Juvenile recidivism � Community context �Neighborhood effects � Drug offending

While few empirical studies have examined neighborhood-

level predictors of juvenile recidivism, the effects of

environmental forces have played a leading role in the

development of criminological theory and juvenile justice

policy. The proliferation of juvenile courts during the early

twentieth century has been attributed to concern for

neighborhoods unable to prevent delinquency (Harris et al.

2000; Tanenhaus 2004). Shaw and McKay’s (1942) semi-

nal research on delinquency rates in Chicago concluded

that the spatial distribution of neighborhood characteristics

influences delinquency rates. Even today, consideration

of juveniles’ environments as they relate to delinquency

influences juvenile court decisions (Fader et al. 2001).

Reducing the likelihood that juvenile offenders will

commit future offenses is a primary goal of the juvenile

justice system. State-level juvenile recidivism rates as high

as 55% have been reported (Snyder and Sickmund 2006).

In 2003, the rate of juveniles in custody was 307 for every

100,000 juveniles, with more than 92,000 juveniles held in

public and private juvenile facilities, according to a 1-day

count (Snyder and Sickmund 2006: 201). This figure rep-

resents a 28% increase in juvenile confinement since 1991.

Moreover, Snyder and Sickmund (2006: 234) estimate the

juvenile reincarceration rate at 24%.

These recidivism and reincarceration rates are largely

attributed to individual and family factors, or to program

impact. Relatively little attention, however, has been given

to the environmental factors that increase or decrease the

H. E. Grunwald

Beasley School of Law, Temple University, Philadelphia,

PA 19122, USA

e-mail: [email protected]

B. Lockwood (&) � P. W. Harris

Department of Criminal Justice, Temple University,

Philadelphia, PA 19122, USA

e-mail: [email protected]

P. W. Harris

e-mail: [email protected]

J. Mennis

Department of Geography and Urban Studies, Temple

University, Philadelphia, PA 19122, USA

e-mail: [email protected]

123

J Youth Adolescence (2010) 39:1067–1079

DOI 10.1007/s10964-010-9518-5

likelihood of recidivism. Kubrin and Stewart (2006: 167),

in their study of the effects of neighborhood context on

adult recidivism, describe this important, yet overlooked

area of investigation when noting that:

Neighborhood context is fundamental to our under-

standing of why individuals offend, and potentially

even more important for understanding why former

offenders offend again, yet we know very little about

how the ecological characteristics of communities

influence the recidivism rates of this population.

The investigation of the effects of neighborhood effects on

reoffending is very new in the field of Criminology, with

several studies finding that space does influence the likeli-

hood of adult recidivism (Kubrin and Stewart 2006; Kubrin

et al. 2007; Mears et al. 2008). This conclusion is supported

by the even more sparse research on the effects of spatial

factors on juvenile recidivism (LeBaron 2002; Simmons

2001), and can be contrasted with a number of studies

examining ecological explanations of delinquency (Bursik

1988; Sampson and Groves 1989; Sampson et al. 1999).

This study explores the effects of environmental factors

on juvenile recidivism and their differential effects on types

of repeat offending. Preliminary hotspot analyses of our data

indicated that the spatial distribution of juvenile recidivism

differed by offense type: local spatial clusters of different

recidivism offense types (drug, property, and violent) were

located in clearly different areas of the city. This spatial

clustering by recidivism offense type indicates that an

investigation of individual- and neighborhood-level effects

on recidivism by offense type may add pertinent information

to the analysis. Earlier studies of delinquency have specified

the type of delinquency that certain neighborhood features

are apt to influence. Jacob (2006) found that residential

mobility is the best predictor of juvenile property crime

while the rate of lone-parent families is the best predictor of

violent crime. Sampson and Grove’s (1989) test of social

disorganization in Great Britain found that organizational

participation and local friendship groups were the strongest

predictors of burglary, while ethnic heterogeneity signifi-

cantly predicted only property crime. Family disruption was

found to predict violent crime and unsupervised peer groups

predicted both property and violent crime (Sampson and

Groves 1989). These findings support an investigation of the

impact of environmental factors on juvenile recidivism and,

further, of an investigation that distinguishes juvenile

recidivism by offense type.

Correlates of Juvenile Recidivism and Offending

Prior studies have uncovered a number of individual-level

factors that influence the likelihood that a juvenile will

re-offend. Juveniles at highest risk to offend are those who

have done so in the past (Cottle et al. 2001; Dembo et al.

1998). Other individual-level predictors of recidivism

include gender, race (Dembo et al. 1998), substance abuse

(Duncan et al. 1995; Elliott et al. 1985), early childhood

misbehavior (White et al. 1990), current age (Snyder and

Sickmund 2006), criminal history (Cottle et al. 2001), prior

out-of home placement (Myner et al. 1998), peer relations

(Akers 1985; Myner et al. 1998), mental health problems

(Huizinga et al. 2000; Pullmann et al. 2006), and family

problems (Wiebush et al. 1995).

Findings from these and other studies have been used to

construct risk assessment tools tasked with assigning levels

of risk of reoffending to juvenile offenders. Evaluations of

these tools have been mixed, with some studies concluding

that they provide little or no predictive ability, at least when

applied in different settings (Miller and Lin 2007; Schwalbe

et al. 2006). It is likely that the inability of risk assessment

tools to accurately predict juvenile re-offending is due to

the absence of spatial predictors (Webster et al. 2006: 12).

Some instruments consider environmental attributes, but

risk prediction instruments rarely take geographic or social

space into consideration, in spite of evidence suggesting the

importance of space when studying delinquency.

Our primary theoretical perspective is social disorgani-

zation theory. This theory provides an ideal framework for

explaining relationships between neighborhood processes

and juvenile recidivism rates. Socially disorganized neigh-

borhoods lack informal social controls which in turn

increases crime and delinquency in those neighborhoods

(Bursik 1988). The proliferation of such ecological con-

siderations within the field of criminology stems from the

work of Shaw and McKay (1942), who identified the effects

of environmental attributes such as poverty, ethnic hetero-

geneity, and residential mobility within neighborhoods on

rates of crime and delinquency. A number of subsequent

studies have operationalized these community-level con-

structs and confirmed these predictors of offending within

communities (Sampson and Groves 1989, Jacob 2006;

Obertwittler 2004; Osgood and Chambers 2000). A number

of studies have concluded that juvenile crime is dependent

on neighborhood processes, particularly where economic

disadvantage decreases collective efficacy, which, in turn,

increases delinquency rates (Bursik 1988; Sampson and

Groves 1989; Sampson et al. 1999). Drug and alcohol

availability (Freisthler et al. 2005; Herrenkohl et al. 2000),

the spatial concentration of juveniles with delinquent atti-

tudes (Oberwittler 2004), and number of ‘‘unconventional’’

friends (Rankin and Quane 2002) have also been identified

as neighborhood-level predictors of juvenile offending.

Other studies have integrated micro- and macro-level

explanations of delinquency and determined that cer-

tain individual attributes can increase an individual’s

1068 J Youth Adolescence (2010) 39:1067–1079

123

susceptibility to neighborhood forces. Cattarello (2000), for

example, found that delinquent peers mediate the effect of

social disorganization on delinquency: controlling for peer

associations rendered the effects of social disorganization

on delinquency insignificant. Similarly, a study by Wik-

strom and Loeber (2000) observed that the effects of

neighborhood socio-economic factors on serious offending

among juveniles is mediated by the risk-level and age of

juveniles; the effects of neighborhood context were limited

to juveniles with low and middle risk scores and to those

with a late adolescent onset of delinquency. Chung and

Steinberg (2006) found that the effect of economic disad-

vantage on persistent delinquency was mediated by par-

enting behavior and peer deviance.

Most research on specific types of juvenile delinquency

has done so for offenses related to drug sales and violence.

This is not surprising when considering the works of Wilson

(1996, 1987) and Anderson (1999, 1990) that describe the

structure of urban areas that contribute to increasing rates of

both types of offenses. They cite changes in economic

conditions, disruption within families, and pervading atti-

tudes that promote criminal activities as causal mechanisms

for increases in both drug selling and violent crime within

disadvantaged urban communities. Many subsequent stud-

ies have provided support for their finding that neighbor-

hood processes, most often relating to economic deprivation

and family organization, contribute to juvenile violence

(Liberman 2007; Osgood and Chambers 2000; Sampson

and Groves 1989). These works indicate that socially dis-

organized neighborhoods are more likely to experience

youth violence than are communities that do not exhibit

qualities of social disorganization.

More recent research has asked whether drug dealing

within neighborhoods can be attributed to neighborhood

processes. Little and Steinberg (2006) used an opportunity

framework to explain their findings that poor neighborhood

conditions and low neighborhood job opportunity influence

urban adolescent drug dealing. Their work draws in part on

literature that indicated a rising number of opportunities for

juveniles to sell drugs in urban areas. They conclude that

‘‘adolescents who sold the most drugs were more likely to

live in contexts characterized by high physical and social

disorder…’’ (Little and Steinberg 2006: 378). Martinez

et al. (2008) support Little and Steinberg’s (2006) finding

that drug crime is influenced by neighborhood-level indi-

cators of disadvantage. Additionally, they found that drug

activity increases violence within neighborhoods, net of

their measures of social disorganization. Their conclusion

that ‘‘traditional dimensions of social disorganization pre-

dict drug activity which, in turn, leads to higher levels of

criminal violence,’’ serves to tie drug and violent offending

together in disadvantaged neighborhoods (Martinez et al.

2008: 866). Further, research by Baumer and Gustafson

(2007) has linked neighborhood structure to rates of

instrumental crime such as drug trafficking.

Few studies have investigated differences in the corre-

lates of recidivism by offense type; rather, most research is

concerned with the presence or absence of re-offending in

general. Much of the small body of literature on specific

recidivism offense types concentrates on sex offenders

(Parks and Bard 2006; Rasmussen 1999) or violent

offenders (Howell 1995; Loeber et al. 1998). While many

studies of the effects of space highlight the importance of

community context on violent and drug offending, little

research has parceled out the predictors of recidivism based

on specific offense types. Research, however, suggests that

the disaggregation of recidivism offense type can serve to

unmask potentially varying effects of individual and

environmental predictors on recidivism. Accordingly, we

disaggregate our outcome measure to capture distinct types

of juvenile recidivism.

Hypotheses

Our sample of youths is likely to comprise the more serious

delinquency cases, since cases in which the disposition was

probation supervision were excluded. Thus the neighbor-

hoods in which most of our sample members reside are

likely to be disadvantaged. We expect, however, that

neighborhood-level variables will exert a significant influ-

ence on juvenile recidivism. Chung and Steinberg (2006),

for example, studied a sample of serious delinquent youths

from the same population and found that concentrated

poverty was associated with self-reported offending through

its effects on neighborhood disorganization and deviant

peers. We hypothesize that neighborhood indicators of

social disorganization found to predict delinquency will

continue to predict recidivism after controlling for indi-

vidual and family contexts. We also hypothesize that indi-

vidual and neighborhood predictors of juvenile recidivism

will vary depending on recidivism offense type. The liter-

ature reviewed above suggests that neighborhood processes

should have a significant influence on both drug and violent

crime. Relative to drug and violent re-offending, we expect

that property offense recidivism will be much less affected

by neighborhood attributes.

Methods

Data

Individual level data were taken from the Program

Development and Evaluation System (ProDES) database, a

population database of all juveniles committed by the

J Youth Adolescence (2010) 39:1067–1079 1069

123

Philadelphia Family Court to community and residential

programs between 1994 and 2004. Within ProDES, the

data were organized by program experience. That is, a

‘‘case’’ was created for a juvenile each time a decision was

made to commit a youth to a program or move the youth to

a different program. These data were collected to provide

program providers, court personnel, and funding agencies

with information on program outcomes. The data include

measures of family demographics, juvenile characteristics,

criminal history, current offense characteristics and peti-

tions for new offenses. All ProDES data used for this study

were collected by CJRC staff.

To test the impact of neighborhood-level attributes on

juvenile recidivism, any periods of residential program

participation were excluded. We included only periods

when neighborhood forces could directly affect recidivism.

Based on that criterion, 13,000 cases were selected who

entered community programs during the period 1996

through 2002–the years when the data were most complete.

The data set was further reduced by the removal of

females, since females made up a small proportion of

subjects (11%), their recidivism rates were less than half

that of males, and prior research has demonstrated marked

gender differences in the predictors of juvenile delinquency

and recidivism (Daigle et al. 2007; Funk 1999; Mazerolle

1998). These considerations reduced our sample for anal-

ysis to 10,971 male delinquents committed to community-

based programs by the Family Court. These 10,971 cases

included multiple records for the same youth, which rep-

resented each program for each youth. Approximately one-

third of the juveniles in the sample appear in the dataset

twice; approximately 300 juvenile offenders appear more

than twice. Since there were too few observations for a

longitudinal analysis, we selected the first community

program experience for each youth resulting in a sample of

7,061 male juveniles. Of these, 2,565, or 36%, were on

aftercare (or parole) status, meaning that their first com-

munity program experience followed a period of incar-

ceration. Since youths on aftercare are likely to differ from

other delinquents in terms of offense seriousness and/or

more troubled family environments (Fader et al. 2001), a

measure of aftercare status was included in the analysis.

The ProDES system includes home addresses for each

juvenile at the time of program participation. ArcView GIS

9.2 was used to geocode the home addresses of the juveniles

and assign these addresses to neighborhoods within Phila-

delphia. The spatial level of aggregation utilized in this

study includes neighborhood boundaries delineated by the

Philadelphia Health Management Corporation (PHMC),

which exhaustively partitions the city into 45 neighbor-

hood polygons. These neighborhood boundaries were con-

structed by researchers and officials familiar with the spatial

distribution of communities in Philadelphia. While larger

than the frequently-used Census geographies of block

groups and tracts, this level of aggregation has the advan-

tage of specifically representing the neighborhoods of

Philadelphia.

Data at the neighborhood level were taken from the

results of the PHMC’s Household Health Survey (HHS), a

biannual survey of residents from Philadelphia and sur-

rounding counties, and includes items related to health and

neighborhood perceptions. The 2000, 2002, and 2004 sur-

veys were used to match the time period of the ProDES

data. The PHMC surveyed 4,088 adults in Philadelphia in

2000, 4,133 in 2002, and 4,415 in 2004, with a mean

sample size of 4,212 over the three survey periods, and a

mean of 93.6 survey respondents per neighborhood. Since

the PHMC HHS also includes respondents from the sur-

rounding counties of Philadelphia, a weighting procedure

was undertaken so that the data represented the population

of the city (for a detailed description of this process, see

Garcia et al. 2007).

Measures

Recidivism

This study examines four juvenile recidivism outcomes: an

aggregate measure of any new offense and three measures

of specific re-offense types (property, drug, or violent). The

outcome measure representing drug recidivism captures

both drug selling and drug possession. Cases were followed

into the adult criminal justice system. The recidivism

measure was dichotomous indicating whether or not a

juvenile had a new petition filed at any time during par-

ticipation in the community-based program through

6-months following program discharge (0 = no new peti-

tion filed, 1 = new petition filed). Snyder and Sickmund

(2006) found that using referral to court to measure recid-

ivism produced an aggregate recidivism rate across several

states of 45%. Of our subjects, slightly more than 40% had a

new petition filed against them during the study period.

The follow-up period starts with program commitment

and ends 6 months following program discharge, during

which time the youths were exposed to home and neigh-

borhood influences. The average length of time per juvenile

spent in the community-based program was 203 days, or

approximately 7 months. With the 6-month post-discharge

period, the study includes an average follow-up period of

approximately 13 months. Preliminary models distin-

guishing between juveniles who recidivated during their

time in treatment and those who reoffended after treatment

indicate that the predictors for both outcomes behave

similarly.

1070 J Youth Adolescence (2010) 39:1067–1079

123

Individual-Level Predictors

A review of the literature and several preliminary analyses,

including binary logistic regression, CHAID (Chi-squared

Automatic Interaction Detector), and neural network anal-

yses, were conducted to reduce the number of predictors

from several hundred to the fourteen that were included in

this analysis. The final set of predictors includes demo-

graphic characteristics, family traits, current offense char-

acteristics, and criminal history. Demographic predictors

include age at first arrest and race, in the form of dummy

variables for white and Hispanic juveniles (black juveniles

are the reference category). Parental deviance was mea-

sured with items indicating drug abuse and criminal

behavior. Receiving public assistance is used to indicate

the family’s socio-economic status. A variable indicating

aftercare status is also included to identify youths returning

directly from residential facilities. Current offense predic-

tors classify the instant offense as violent, property, or sex-

related, based on the initial most serious charge. Lastly,

criminal history predictors indicate a prior drug crime,

violent crime, or out-of-home placement on the juveniles’

court records. All of the individual-level variables selected

were dichotomous (0 = no, 1 = yes), with the exception of

age at first arrest, which is continuous.

The few studies that have examined neighborhood

effects on recidivism have generally included measures of

economic disadvantage. This study also includes such a

measure in the form of concentrated disadvantage; addi-

tionally, we estimate the effects of a potentially protective

factor in the form of social capital. Data from the PHMC

HHS conducted in 2000, 2002, and 2004 were used to

construct indices representing concentrated disadvantage

and social capital.

Neighborhood Disadvantage

An index of concentrated disadvantage was created from a

linear combination of four HHS items: adults living below

the poverty line, unemployed, on welfare, and yearly

income. The first three items were converted into propor-

tions representing the percentage of adults per neighbor-

hood that live below the poverty line, are unemployed, and

are on welfare. The fourth item represented yearly income

on a scale of zero through seventeen, with seventeen rep-

resenting adults making more than $250,000 per year. The

values of this item were multiplied by -1 so that high

values would represent disadvantaged individuals, and thus

be consistent with the other items in the concentrated dis-

advantage index. This index demonstrates a high level of

internal reliability with a Cronbach’s alpha of 0.947 in

2000, 0.871 in 2002, and 0.845 in 2004. To further match

the data collection period of the ProDES data, and to

account for neighborhood change, the values of this index

for each of the 3 years was averaged so that a mean value

from 2000 to 2004 was created.

This operationalization of neighborhood disadvantage is

consistent with studies of the effects of neighborhood

context on crime (Baumer et al. 2003; Baumer 2002;

Morenoff et al. 2001; Sampson et al. 1997), with two

exceptions. The proportion of black residents was excluded

for two reasons: it decreased the reliability of the index

and, theoretically, race has been used to indicate financial

disadvantage, which is already well-represented in the

index.

Social Capital

An index representing social capital was also created from

the 2002 and 2004 HHS data. Prior to 2002, the survey did

not include these items. This index was constructed from

four items relating to perceptions of and levels of partici-

pation within their neighborhood: number of local groups

and organizations they participate in; a five-category rating

(ranging from ‘‘never’’ to ‘‘always’’) of how often neigh-

borhood residents are willing to help their neighbors with

basic tasks, such as picking up trash cans and shoveling

snow; and responses to two statements on a scale of ‘‘1’’ to

‘‘5’: ‘‘I feel that I belong and are part of my community’’

and ‘‘Most people in my neighborhood can be trusted.’’ This

index also demonstrates a high level of internal reliability,

with a Cronbach’s alpha of 0.785 in 2002 and 0.868 in 2004.

As with the concentrated disadvantage index, a mean value

of this item from 2002 and 2004 was used in the analysis.

The concept of social capital has been described by

some researchers as ‘‘varied, murky, and even circular’’

(Messner et al. 2004: 882). Several researchers, however,

have identified two key components of social capital: civic

engagement and trust (Kennedy et al. 1998; Messner et al.

2004; Rosenfeld et al. 2001). The social capital indicator in

the current study possesses both components. Additionally,

the components of this index closely resemble the items

used to conceptualize collective efficacy, an index intended

to represent favorable neighborhood processes that stand in

contrast to socially disorganized communities (Sampson

et al. 1997, 1999).

It is important here to note the importance of controlling

for neighborhood crime rates when estimating the effects

of neighborhood context on criminal activity. We found,

however, that neighborhood crime rates, including total

crime and crime by type, are highly correlated with our

social disorganization and social capital constructs. We

opted not to include neighborhood crime in our neighbor-

hood scale items, as crime and our items representing

social disorganization and social capital are distinct con-

structs. The inclusion of crime into these latent constructs,

J Youth Adolescence (2010) 39:1067–1079 1071

123

while statistically supported, would not be consistent with

the theoretical definitions of those measures. To be sure of

the similarities between crime and our level-2 constructs,

earlier models that included crime rate at level-2 support

the findings derived from the models included in this study.

While neighborhood crime is not included in the current

models, we believe that crime is accounted for by inclusion

of level-2 constructs that are highly correlated with crime.

Analytic Strategy

Univariate descriptive statistics were estimated to describe

the predictors and outcomes to be included in the analysis.

Bivariate correlations of the individual-level variables were

then analyzed to detect potential associations and multi-

collinearity. The main portion of this analysis includes the

estimation of hierarchical linear models (HLM) so that

potential neighborhood effects of juvenile recidivism can

be identified, while controlling for individual-level vari-

ables. Hierarchical models permit the simultaneous inclu-

sion of multiple levels of data and estimations of the

amount of variance in the outcome measure that each level

of data is responsible for. HLM was employed due to the

natural clustering of youth in neighborhoods.

In this analysis, juvenile offenders (n = 7,061) comprise

the level-1 units of analysis and neighborhoods (n = 45) the

level-2 units. For models estimating the effects of dichoto-

mous outcomes, it is suggested that an interval of plausible

values is calculated to determine if the outcome varies across

level-2 units of analysis (Raudenbush and Bryk 2002). A

significant interval of plausible values indicates variation in

the outcome between level-2 units of analysis and supports

the utility of HLM to analyze the data. These unconditional

models were followed by two-level, hierarchical, general-

ized linear models to estimate individual and neighborhood

effects on the odds of recidivism. We use a random intercept

model to predict all four outcomes.

Results

Univariate statistics for the outcomes and predictors are

shown in Table 1. An examination of the mean values of

the outcome measures indicates that 40% of the juvenile

offenders recidivated, 14% with a drug crime, 10 with a

violent crime, and 11% with a property offense. Although

drug offense recidivism was the most prevalent recidivism

offense type, the proportions of drug, violent, and property

reoffenders in the dataset are relatively similar. For race,

11% of the youths were white and 13% Hispanic. The

reference category of black juveniles comprised the largest

proportion of the population (74%). The average age of the

juveniles at the time of their first arrest was 14.2 years. As

with the recidivism offense types, there was little variation

between the proportion of juveniles who committed a

violent crime as their initial offense (36%) and those who

committed a property crime (32%). Few offenders com-

mitted a sex offense (6%). Descriptive statistics for the

court history predictors indicate that more of the juvenile

offenders have a prior drug offense on their criminal record

(33%) relative to prior violent offenses (20%). Few juve-

niles had been placed out of their home due to prior to their

instant offense (7.6%).

A comparison of these variables by recidivism type

indicates that juveniles in this population who recidivate

via drug crimes differ from juveniles who recidivated with

a violent or property offense (the supporting table can be

obtained from the author). Drug crime recidivists are more

likely to be Hispanic, have a prior drug arrest and have had

a prior out-of-home placement than are person or property

reoffenders. An examination of correlations and further

follow up of the tolerance statistics indicated no issues of

multicollinearity. Therefore, all the juvenile and family

variables were eligible for inclusion in the analysis. The

neighborhood-level items representing concentrated dis-

advantage and social capital were, not unexpectedly, found

to be highly correlated. As a result, these predictors were

entered into separate models.

The reliability of the neighborhood estimates for the

empty models without predictors is high. Once level-1

predictors are entered into the models, however, individual

reliability estimates for each subsequent model dropped to

approximately .15. This low reliability results in shrinkage

of the log odds of recidivism in the neighborhood estimates

back towards the grand mean log odds. There are two

structural neighborhood characteristics that can cause

shrinkage: (1) large intra-neighborhood variances (noisy

level-1 data) and/or (2) small sample sizes. A visual

inspection of the unadjusted recidivism rates across

neighborhoods shows proportions ranging from 0.3 to 0.7.

However, the adjusted estimates are shrunk to a plausible

interval of .38–.40 for all recidivism types combined,

indicating a significant amount of weighting back towards

the grand mean log odds across the neighborhoods. This is

indicative of the strong effects that the level-1 variables

have on the outcomes, relative to the effects of neighbor-

hood context. Based on the vast literature of individual

predictors of recidivism, this is not unexpected. Regardless,

the previously-mentioned studies of the effects of neigh-

borhood context on recidivism support the current inves-

tigation of space on juvenile recidivism (Kubrin et al.

2007; Kubrin and Stewart 2006; LeBaron 2002; Mears

et al. 2008; Simmons 2001).

Results of the eight hierarchical generalized linear models

are shown in Table 2. Odds ratios, with their subsequent

significance, are displayed. Most of the individual-level

1072 J Youth Adolescence (2010) 39:1067–1079

123

predictors in the models are dichotomous (indicating group

membership). A significant odds ratio of 1.5 for a predictor,

for example, indicates that, on average, the odds are 50%

higher of recidivating for youths with a ‘‘1’’ for that variable.

In contrast a significant odds ratio of 0.5, indicates that on

average, the odds of recidivating for youths with a ‘‘1’’ for

that variable are 50% lower. Age at first arrest and all five of

the neighborhood-level variables were standardized before

being entered into the models. As a result, their odds ratios

are interpreted differently. An odds ratio of 1.5 for one of

these variables indicates that, on average, a one standard

deviation increase will result in a 50% increased odds that a

juvenile will recidivate.

Each set of two models for the four outcome measures

will be discussed separately. There is very little variance in

the models measuring the same outcomes; the same pre-

dictors are found to be significant with only minor variation

in the effect sizes of predictors. Odds ratios for predictors

are listed first for the model with concentrated disadvan-

tage, followed by the model with social capital. A com-

parison of the models for each outcome follows.

Models 1 and 2: All Recidivism

Of the fourteen individual-level variables entered into the

models, eight are significant predictors of recidivism. Nei-

ther the race variable nor ‘‘age at first arrest’’ are significant

predictors of recidivism when all recidivism offense types

are combined. Juveniles designated as aftercare cases have

a relatively high odds ratio (OR = 1.49; 1.49, p \ .01),

indicating that juveniles on an aftercare status are more

likely (odds are nearly 50% higher) to reoffend than juve-

niles who were not. Juveniles who committed property

offenses (OR = 1.23; 1.23, p \ .05) are more likely (odds

are 23% higher) to recidivate than juveniles who do not

commit property offenses. All of the variables representing

Table 1 Descriptive statistics

of individual- and

neighborhood-level variables

Variable Metric N Mean SD

Outcome

Recidivism: All 0 = no, 1 = yes 7061 0.40 0.49

Recidivism: Drug crime 0 = no, 1 = yes 7061 0.14 0.35

Recidivism: Violent crime 0 = no, 1 = yes 7061 0.10 0.30

Recidivism: Property crime 0 = no, 1 = yes 7061 0.11 0.31

Level 1: Individual

Family receives public assistance 0 = no, 1 = yes 7061 0.31 0.46

Parental drug abuse 0 = no, 1 = yes 7061 0.20 0.40

Parental criminal history 0 = no, 1 = yes 7061 0.16 0.36

Age at 1st arrest (years) Continuous 7061 14.2 1.7

White 0 = no, 1 = yes 7061 0.11 0.32

Hispanic 0 = no, 1 = yes 7061 0.13 0.34

Aftercare case 0 = no, 1 = yes 7061 0.35 0.48

Prior drug offense 0 = no, 1 = yes 7061 0.33 0.47

Prior violent offense 0 = no, 1 = yes 7061 0.20 0.40

On probation at time of arrest 0 = no, 1 = yes 7061 0.10 0.30

Prior placement ever 0 = no, 1 = yes 7061 0.07 0.25

Sex offense 0 = no, 1 = yes 7061 0.06 0.24

Violent offense 0 = no, 1 = yes 7061 0.36 0.48

Property offense 0 = no, 1 = yes 7061 0.32 0.46

Level 2: Neighborhood

Concentrated disadvantage Scale item 45

Below poverty line Proportion 45 0.17 0.09

Unemployment Proportion 45 0.07 0.03

On welfare Proportion 45 0.04 0.03

Income -17–0 45 -9.21 1.71

Social capital Scale item 45

Belong 1–5 45 3.07 0.11

Trust 1–5 45 2.75 0.22

Help neighbors 1–5 45 3.45 0.20

Participate Continuous 45 0.75 0.17

J Youth Adolescence (2010) 39:1067–1079 1073

123

criminal history have a significant and positive relationship

with the likelihood of recidivating. Having committed a

prior drug offense (OR = 1.50; 1.51, p \ .01) and having a

prior out-of-home placement (any placement, dependent or

delinquent at any point in the youth’s life; OR = 1.48; 1.48,

p \ .01) are particularly strong predictors of juvenile

recidivism. The neighborhood concentrated disadvantage

index, however, is not significantly related to the aggregated

recidivism outcome. Similarly, the social capital predictor

is unrelated to recidivism in Model 2.

Models 3 and 4: Drug Recidivism

In models 3 and 4, six individual-level predictors are sig-

nificantly correlated with juveniles who commit drug

crimes as their recidivating offense. The attributes Hispanic

(OR = 1.33, p \ .05; 1.37, p \ .01) and being older at the

time of first arrest (OR = 1.14; 1.14, p \ .01) are both

positively related to drug recidivism. In contrast, white

juveniles, on average, are less likely than black youths to

recidivate through the commission of a drug crime

(OR = 0.65; 0.62, p \ .01), holding other variables in the

model constant. Juveniles on aftercare status are much more

likely to reoffend (OR = 1.68; 1.68, p \ .01) than their

non-placed counterparts. Of the current offense variables

representing the offenses that brought the juveniles into

court, only sex offense is significantly related to drug

recidivism (OR = 0.57; 0.57, p \ .01), and this relation-

ship is negative. Only one of the criminal history predictors

is significant for drug recidivism: prior drug offense

(OR = 2.81; 2.82, p \ .01) exerts the greatest influence of

all variables in any of the models. The effect of having a

prior drug offense in a juvenile’s criminal history is to

nearly triple the odds that the juvenile will reoffend with a

drug crime. At the neighborhood-level, both concentrated

disadvantage (OR = 1.19, p \ .01) and social capital

(OR = 0.89, p \ .05) are significant predictors of juvenile

drug recidivism. The odds ratios for both neighborhood

predictors are in the opposite but expected directions,

indicating that the likelihood of a juvenile offender recidi-

vating with a drug offense increases as the level of disad-

vantage in a neighborhood increases, and decreases as the

level of social capital increases.

Models 5 and 6: Violent Recidivism

Models 5 and 6 possess two significant demographic

variables. The first, Hispanic (OR = 0.68; 0.66, p \ .01),

Table 2 Odds ratios of individual- and neighborhood-level predictors of juvenile recidivism

Model 1 2 3 4 5 6 7 8

Recidivism type All All Drug Drug Violent Violent Property Property

Individual-level

Demographics

White 0.89 0.86 0.65** 0.62** 0.82 0.83 1.19 1.20

Hispanic 1.08 1.09 1.33* 1.37** 0.68** 0.66** 1.07 1.06

Age at 1st arrest 0.95 0.95 1.14** 1.14** 0.87** 0.90** 0.90** 0.90**

Family

Parental drug abuse 1.09 1.09 0.95 0.95 1.18 1.18 1.03 1.03

Parental criminality 1.13* 1.13* 1.08 1.09 1.27* 1.27* 1.04 1.04

On public assistance 1.10 1.11 1.10 1.12 0.97 0.96 1.12 1.12

Current offense

Aftercare case 1.49** 1.49** 1.68** 1.68** 1.03 1.03 1.15 1.15

Sex offense 0.53** 0.53** 0.57** 0.57** 0.62** 0.62* 0.53* 0.53*

Violent offense 1.01 1.01 0.88 0.87 1.25* 1.26* 1.05 1.05

Property offense 1.23* 1.23* 1.00 0.99 1.07 1.07 1.59** 1.59**

Criminal history

On probation 1.24* 1.24* 0.92 0.92 1.26 1.26 1.23 1.23

Prior drug offense 1.50** 1.51** 2.81** 2.82** 0.76* 0.76** 0.76* 0.76*

Prior violent offense 1.27** 1.27** 1.16 1.16 1.36** 1.36** 0.98 0.98

Prior placement 1.48** 1.48** 1.23 1.24 1.13 1.13 1.47** 1.47**

Neighborhood-level

Disadvantage 1.07 – 1.19** – 1.00 – 0.98 –

Social capital – 0.96 – 0.89* – 0.96 – 1.01

** p \ .01, * p \ .05

1074 J Youth Adolescence (2010) 39:1067–1079

123

indicates that Hispanic juveniles are less likely than black

youths to recidivate with a violent offense. Age at first

arrest is positively correlated with the target variable

(OR = 0.87; 0.90, p \ .01), signifying that juveniles who

were older at the time of their first offense are less likely to

recidivate with a violent offense. Only parental criminality

is significant among the family context variables

(OR = 1.27; 1.27, p \ .05): parental criminality is posi-

tively associated with violent re-offending. Sex offense is a

significant predictor as well (OR = 0.62; 0.62, p \ .05),

reducing the likelihood of recidivating via a violent offense

for juveniles previously convicted of sexual offenses. Two

criminal history variables exert significant influence on

violent recidivism, but in different ways. Having commit-

ted a prior drug crime significantly reduces the likelihood

that a juvenile will recidivate with a violent offense

(OR = 0.76; 0.76, p \ .05). In contrast, having committed

a prior violent offense increases the likelihood of reof-

fending with a similar violent offense (OR = 1.36; 1.36,

p \ .01). Neither concentrated disadvantage nor social

capital is a significant predictor of violent recidivism.

Models 7 and 8: Property Recidivism

As with Models 1 and 2, neither race variable is a signif-

icant predictor of property offense recidivism. Age at first

arrest (OR = 0.90; 0.90, p \ .01) is negatively correlated,

indicating that younger adolescents are more likely to

recidivate with a property crime. No family context vari-

ables are significant predictors of property recidivism. Sex

offense (OR = 0.53; 0.53, p \ .05), as with all other

models, is negatively correlated with property recidivism.

The effect of having committed a property offense as the

current offense is to increase the odds of reoffending with

a property offense by nearly 60% (OR = 1.59; 1.59,

p \ .01) over non-property offenders. Prior drug offenders

are less likely to reoffend with a property offense (OR =

0.76; 0.76, p \ .01), while juveniles who have been placed

out of their home in the past are nearly 50% more likely to

recidivate with a property offense (OR = 1.47; 1.47,

p \ .01) than are juveniles who have not been placed out of

their homes. As with violent offense recidivism in Models

5 and 6, the neighborhood context attributes are not sig-

nificantly correlated with property offense recidivism.

Discussion

This research is among the first to estimate neighborhood

effects on juvenile recidivism and is the first to do so while

disaggregating juvenile recidivism to measure specific

recidivism offense types. The analysis has uncovered sev-

eral findings that can be used to inform future research and

juvenile justice policy. Social context, in the form of

concentrated disadvantage and social capital, is a signifi-

cant predictor of a particular type of juvenile recidivism:

drug reoffending. The significant effects of environmental

variables representing community processes are supportive

of similar processes described by social disorganization

theory. The concentrated disadvantage predictor closely

approximates the poverty dimension of social disorgani-

zation theory, while the items used to create the social

capital index are similar to those used in the construction of

the collective efficacy, a concept representing socially-

shared processes that reduce social disorganization in

communities. As a result, this study can be said to have

found support for the effects of social disorganization on

juvenile drug recidivism. Consistent with social disorga-

nization theory, increased levels of concentrated disad-

vantage increase the likelihood of drug recidivism, while

increased levels of social capital decrease the likelihood of

drug recidivism.

Prior studies that have examined neighborhood corre-

lates of recidivism have not made this offense type dis-

tinction (Kubrin et al. 2007; Kubrin and Stewart 2006;

LeBaron 2002; Simmons 2001), but our findings regarding

drug offense recidivism are consistent with recent research

on drug selling among juvenile offenders (Little and

Steinberg 2006; Martinez et al. 2008). The specification of

these relationships to drug offending and the instrumental

nature of drug offending suggest that drug offending, and

possibly specialization in this type of crime, is supported

by some form of illicit organization.

Both Hagedorn (1994) and Decker et al. (2008) argue

that drug organizations are typically not highly organized.

However, Hagedorn notes that Hispanic kinship ties sustain

well developed drug organizations, whereas African

American and White drug sellers operate more indepen-

dently. This may help to explain why being Hispanic is a

predictor of this type of offending and reoffending. Using a

contingency theory framework, Hagedorn argues that

neighborhood conditions both block and create opportuni-

ties. Our findings are consistent with this view, both in

terms of ethnicity (Hispanic youths) and in terms of social

and economic disadvantage. In Philadelphia, the Hispanic

community is concentrated in one large area of North

Philadelphia with limited socioeconomic opportunities.

The combination of poverty, ethnic ties, opportunity, and

the economic benefits of drug selling identified in Hispanic

communities by this analysis is consistent with Little and

Steinberg’s (2006) interpretation of their findings, namely

that opportunity to engage in drug selling explains their

findings that poor neighborhood conditions and low

neighborhood job opportunity influence urban adolescent

drug dealing. We should also note that youths involved in

drug selling were older than the other youths in the other

J Youth Adolescence (2010) 39:1067–1079 1075

123

offense type categories when they first entered the juvenile

justice system, indicating that their participation in drug

selling was likely to have been less impulsive relative to

other recidivists.

Our findings also suggest that social disorganization

alone is insufficient to explain the involvement of delin-

quent youths in persistent drug offending. Youths engaged

in this behavior are likely to be residing in environments

characterized by ethnic ties, economic disadvantage, and

opportunities to become involved in an illicit, organized

business. Thus a social learning perspective, such as that

described by Akers (1985), should be useful for future

explorations of this pattern of behavior. It appears likely

that youths living in environments characterized by ethnic

homogeneity, poverty, and a well-organized drug-traffick-

ing business will be easily drawn into illicit activities that

are both encouraged by others and lucrative. The current

study did not include a measure of juvenile association

with deviant peers during the follow-up period to examine

the impact of such peers on the likelihood of recidivism,

but future work in this area would be well-served by testing

the effects of peer influence on juvenile recidivism.

Contrary to the expected findings, neighborhood disad-

vantage does not influence rates of violent recidivism

offenses by juveniles. Of all recidivism offense types, vio-

lent recidivism was expected to have the strongest rela-

tionship with neighborhood disadvantage, as many studies

have identified a link between areas of concentrated poverty

and violent crime (Anderson 1999, 1990; Liberman 2007;

Osgood and Chambers 2000; Sampson and Groves 1989;

Wilson 1996, 1987). The lack of significance of neighbor-

hood context for violent recidivism might again reflect the

greater impact of neighborhood processes on repeated

involvement in instrumental crimes that include drug sell-

ing and which are largely dependent on opportunity. It may

also be the case that social disorganization plays a critical

role in explaining delinquency in general but that once the

focus of analysis is narrowed to comparisons among

delinquent youths, social disorganization has little left to

explain. When we consider the relatively young age of these

youths at their first arrest, the significant impact of parental

criminality and their previous arrests for violent offenses,

we see a pattern that is more consistent with disorganization

at a more micro level, thus supporting the findings of Chung

and Steinberg (2006) regarding family dynamics and

delinquency among seriously delinquent youths. These

findings are consistent with those of Lattimore et al. (1995)

who found that parental criminality and family violence

were associated with violent offense recidivism among a

group of young parolees in California. Their findings, as

well as those of Widom (1989), suggest that both parental

criminality and parental abuse and neglect increased the

likelihood of an arrest for a violent offense.

There appears to be some consistency in offending

behavior among violent and property re-offenders, and

their relatively low age of first arrest, combined with

family problems ranging from parental criminality to abuse

and neglect, that suggests both impulsivity and a lack of

appropriate parental monitoring and supervision. It may be

that our distinctions between person and property offenders

are not warranted and that combining them would be more

useful. For example, Sampson and Groves (1989) found

that family disruption predicted violent crime. Family

disruption can be the result of system responses to abuse

and neglect, and it can result from a parent being incar-

cerated as a result of criminal prosecution.

The strongest individual-level predictors in our analysis

were derived from the juveniles’ criminal history, and

suggest specialization of juvenile drug offending. The

analysis found that juveniles with prior drug offenses are

significantly more likely to re-offend with a drug crime,

and are significantly less likely to re-offend with a violent

or property crime. The literature on offense specialization

has concluded that specialization among juveniles is

uncommon. Few offenders specialize at all, and those

individuals considered specialists generally age toward

specialization (Farrington et al. 1988; Piquero et al. 1999).

Note, however, that our drug offenders were likely to be

older at the time of their first arrest, suggesting a limited

impact of age on specialization. A recent study by Arm-

strong (2008) suggests that juvenile drug offenders are

more likely to specialize: age affects the specialization of

violent and property offending, but not drug offending.

Thus, it may be that the age effects on specialization found

in prior research do not have the same impact on drug

offense specialization (Armstrong 2008; Piquero et al.

1999).

Limitations

There are limitations of this study that should be men-

tioned. Regarding our neighborhood-level predictors, the

components of our concentrated disadvantage and social

capital indicators have been used extensively in prior

research estimating the effects of space on crime, but there

are countless other indicators that could be called upon to

determine whether the neighborhoods in this study influ-

ence juvenile recidivism rates. Ecological researchers have

called for an inclusion of community-level variables

beyond those that measure disadvantage; it would likely be

wise to heed this advice and consider the effects of vari-

ables that capture social interactions, protective factors,

and other neighborhood-level indicators that are not com-

monly found in studies of space and crime.

The measurement of social context should also be

considered in light of the arguments raised by Rankin and

1076 J Youth Adolescence (2010) 39:1067–1079

123

Quane (2002) and Chung and Steinberg (2006) regarding

the value of social attributes that mediate the relationships

between economic disadvantage and delinquency. We did

include several measures of parental behavior at the indi-

vidual level, and found that parental criminality was

associated with violent re-offending. Our knowledge about

parenting behavior, however, was limited to what appeared

in juvenile court records.

Our data suggest that some youths, particularly drug

offenders, specialize in the types of offenses they commit.

We recognize, however, that our data include only one

offense transition, and that our offense information is

derived from court data. Not only are many offenses not

known to the justice system, the criminal justice process

acts to siphon off both non-offenders (false positives) and

actual offenders (false negatives). Appropriate testing of a

hypothesis of specialization would require more than two

offenses and would ideally make use of self report offense

data. Although such a dataset may be difficult to obtain, it

would permit questions beyond the scope of this analysis to

be asked.

Considering the observed differences between juvenile

drug reoffenders and juveniles reoffending with violent and

property crimes, future research should consider the possi-

bility of different explanations of recidivism, depending on

the kinds of offending involved. For person and property

offending, it may be that individual, family, and peer

attributes are most critical to understanding subsequent

offending, while for drug selling, neighborhood context

should also be taken into account. Further disaggregating

recidivism to more detailed offense categories may also

reveal other patterns of predictors and, hence, more ways to

explain recidivism. For example, the theories called upon to

explain the neighborhood-level effects identified in our

analysis of juvenile drug recidivism suggest that drug sell-

ing is much more sensitive to community processes than is

drug using. Therefore, future research would be well-

advised to investigate what specific spatial factors influence

specific types of drug offending. Beyond drug offending,

disaggregating violent and property crime might also ben-

efit our understanding of the processes described in this

analysis. Auto theft, a property crime, may be classified as

an instrumental crime, but we have aggregated it with all

other property crimes, thus masking this possibility.

Conclusion

The spatial dependency of drug offending, combined with

the effects of different family predictors of violent and

property offense recidivism, imply the need for multiple

explanations of recidivism. Our findings indicate that no

single causal model of juvenile recidivism can effectively

explain all types of reoffending. Drug re-offenders, in

particular, appear likely to persist in drug offending. This

pattern of offense specialization is associated with high

levels of economic disadvantage and, in the case of Phil-

adelphia, social isolation of Latino communities (see also

Bourgois 2003, who describes a similar pattern in New

York’s South Bronx). This finding suggests that the juve-

nile justice system is unlikely to make headway through

punitive measures or by temporary removal of these youths

from their home environments. Neighborhood and family

contexts should be part of any strategy to reduce juvenile

recidivism.

References

Akers, R. L. (1985). Deviant behavior: A social learning approach(3rd ed.). Belmont, CA: Wadsworth.

Anderson, E. (1990). Streetwise: Race, class, and change in an urbancommunity. Chicago, IL: University of Chicago Press.

Anderson, E. (1999). Code of the street: Decency, violence, and themoral life of the inner city. New York: Norton.

Armstrong, T. A. (2008). Are trends in specialization across arrests

explained by changes in specialization occurring with age?

Justice Quarterly, 25, 201–222.

Baumer, E. P. (2002). Neighborhood disadvantage and police

notification by victims of violence. Criminology, 40, 579–616.

Baumer, E. P., & Gustafson, R. M. (2007). Social organization and

instrumental crime: Assessing the empirical validity of classic

and contemporary anomie theories. Criminology, 45, 617–664.

Baumer, E. P., Horney, J., Felson, R. B., & Lauritsen, J. L. (2003).

Neighborhood disadvantage and the nature of violence. Crim-inology, 41, 39–72.

Bourgois, P. (2003). In search of respect: Selling crack in el barrio.

New York: Cambridge University Press.

Bursik, R. J. (1988). Social disorganization and theories of crime and

delinquency: Problems and prospects. Criminology, 26, 519–552.

Cattarello, A. M. (2000). Community-level influences on individuals’

social bonds, peer associations, and delinquency: A multilevel

analysis. Justice Quarterly, 17, 33–60.

Chung, H. L., & Steinberg, L. (2006). Relations between neighbor-

hood factors, parenting behaviors, peer deviance, and delin-

quency among serious juvenile offenders. DevelopmentalPsychology, 42, 319–331.

Cottle, C. C., Lee, R. J., & Heilbrun, K. (2001). The prediction of

criminal recidivism in juveniles. Criminal Justice and Behavior,28, 367–394.

Daigle, L. E., Cullen, F. T., & Wright, J. P. (2007). Gender

differences in the predictors of juvenile delinquency. YouthViolence and Juvenile Justice, 5, 254–286.

Decker, S. H., Katz, C. M., & Webb, V. J. (2008). Understanding the

black box of gang organization: Implications for involvement in

violent crime, drug sales, and violent victimization. Crime andDelinquency, 54, 153–172.

Dembo, R., Schmeidler, J., Nini-Gough, B., Sue, C. C., Borden, P., &

Manning, D. (1998). Predictors of recidivism to a juvenile

assessment center: A three year study. Journal of Child andAdolescent Substance Abuse, 7, 57–77.

Duncan, R. D., Kennedy, W. A., & Patrick, C. J. (1995). Four-factor

model of recidivism in male juvenile offenders. Journal ofClinical Child Psychology, 24, 250–257.

J Youth Adolescence (2010) 39:1067–1079 1077

123

Elliott, D. S., Huizinga, D. H., & Ageton, S. S. (1985). Explainingdelinquency and drug use. Beverly Hills, California: Sage.

Fader, J. J., Harris, P. W., Jones, P. R., & Poulin, M. E. (2001).

Factors involved in decisions on commitment to delinquency

programs for first-time juvenile offenders. Justice Quarterly, 18,

323–341.

Farrington, D. P., Snyder, H. N., & Finnegan, T. A. (1988).

Specialization in juvenile court careers. Criminology, 26,

461–488.

Freisthler, B., Needell, B., & Gruenewald, P. J. (2005). Is the physical

availability of alcohol and illicit drugs related to neighborhood

rates of child maltreatment? Child Abuse and Neglect, 29, 1049–

1060.

Garcia, R. M., Taylor, R. B., & Lawton, B. A. (2007). Impacts of

violent crime and neighborhood structure on trusting your

neighbors. Justice Quarterly, 24, 679–704.

Hagedorn, J. M. (1994). Neighborhoods, markets and gang drug

organization. Journal of Research in Crime and Delinquency,31, 264–294.

Harris, P. W., Welsh, W. N., & Butler, F. (2000). A century of

juvenile justice. In G. LaFree (Ed.), Criminal justice 2000: Thenature of crime: Continuity and change. Washington, DC:

National Institute of Justice.

Herrenkohl, T. I., Maquin, E., Hill, K. G., Hawkins, J. D., Abbott,

R. D., & Catalano, R. F. (2000). Developmental risk factors for

youth violence. Journal of Adolescent Health, 26, 176–186.

Howell, J. C. (1995). Guide for implementing the comprehensivestrategy for serious, violent, and chronic juvenile offenders.

Rockville, MD: Office of Juvenile Justice and Delinquency

Prevention.

Huizinga, D., Loeber, R., Thornberry, T. P., & Cothern, L. (2000).

Co-occurrence of delinquency and other problem behaviors. In

Juvenile Justice Bulletin. Washington, DC: Office of Juvenile

Justice and Delinquency Prevention.

Jacob, J. C. (2006). Male and female youth crime in Canadian

communities: Assessing the applicability of social disorganiza-

tion theory. Canadian Journal of Criminology and CriminalJustice, 48, 31–60.

Kennedy, B. P., Kawachi, I., Prothrow-Stith, D., Lochner, K., &

Gupta, V. (1998). Social capital, income inequality, and firearm

violent crime. Social Science and Medicine, 47, 7–17.

Kubrin, C. E., Squires, G. D., & Stewart, E. A. (2007). Neighbor-

hoods, race, and recidivism: The community-reoffending nexus

and its implications for African-Americans. Race RelationsAbstracts, 32, 7–37.

Kubrin, C. E., & Stewart, E. A. (2006). Predicting who reoffends: The

neglected role of neighborhood context in recidivism studies.

Criminology, 44, 165–197.

Lattimore, P. K., Visher, C. A., & Linster, R. L. (1995). Predicting

arrest for violence among serious youthful offenders. Journal ofResearch in Crime and Delinquency, 32, 54–83.

LeBaron, J. (2002). Examining the relative influence of community

context on juvenile offender post-confinement recidivism. Doc-

toral Dissertation, Department of Criminal Justice, Rutgers

University, Newark, NJ.

Liberman, A. (2007). Adolescents, neighborhoods, and violence: Recentfindings from the project on human development in Chicagoneighborhoods. Washington, DC: National Institute of Justice.

Little, M., & Steinberg, L. (2006). Psychosocial correlates of

adolescent drug dealing in the inner city: Potential roles of

opportunity, conventional commitments, and maturity. Journalof Research in Crime and Delinquency, 43, 357–386.

Loeber, R., Farrington, D. P., & Waschbusch, D. A. (1998). Serious

and violent juvenile offenders. In R. Loeber & D. P. Farrington

(Eds.), Serious and violent juvenile offenders: Risk factors andsuccessful interventions. London: Sage Publications.

Martinez, R., Rosenfeld, R., & Mares, D. (2008). Social disorgani-

zation, drug market activity, and neighborhood violent crime.

Urban Affairs Review, 43, 846–874.

Mazerolle, P. (1998). Gender, general strain, and delinquency: An

empirical examination. Justice Quarterly, 15, 65–91.

Mears, D. P., Wang, X., Hay, C., & Bales, W. D. (2008). Social

ecology and recidivism: Implications for prisoner reentry.

Criminology, 46, 301–340.

Messner, S. F., Baumer, E. P., & Rosenfeld, R. (2004). Dimensions of

social capital and rates of criminal homicide. American Socio-logical Review, 69, 882–903.

Miller, J., & Lin, J. (2007). Applying a generic juvenile risk

assessment instrument to a local context. Crime and Delin-quency, 53, 552–580.

Morenoff, J. D., Sampson, R. J., & Raudenbush, S. W. (2001).

Neighborhood inequality, collective efficacy, and the spatial

dynamics of homicide. Criminology, 39, 517–560.

Myner, J., Santman, J., Cappelletty, G. G., & Perlmutter, B. F. (1998).

Variables related to recidivism among juvenile offenders.

International Journal of Offender Therapy and ComparativeCriminology, 42, 65–80.

Oberwittler, D. (2004). A multilevel analysis of neighbourhood

contextual effects on serious juvenile offending. EuropeanJournal of Criminology, 1, 201–235.

Osgood, D. W., & Chambers, J. M. (2000). Social disorganization

outside the metropolis: An analysis of rural youth violence.

Criminology, 38, 81–115.

Parks, G. A., & Bard, D. E. (2006). Risk factors for adolescent sex

offender recidivism: Evaluation of predictive factors and com-

parison of three groups based upon victim type. Sexual Abuse: AJournal of Research and Treatment, 18, 319–342.

Piquero, A. R., Paternoster, R., Mazerolle, P., Brame, R., & Dean,

C. W. (1999). Onset age and offense specialization. Journal ofResearch in Crime and Delinquency, 36, 275–299.

Pullmann, M. D., Kerbs, J., Koroloff, N., Veach-White, E., Gaylor,

R., & Sieler, D. (2006). Juvenile offenders with mental health

needs: Reducing recidivism using wraparound. Crime andDelinquency, 52, 375–397.

Rankin, B. H., & Quane, J. M. (2002). Social contexts and urban

adolescent outcomes: The interrelated effects of neighborhoods,

families, and peers on African-American youth. Social Prob-lems, 49, 79–100.

Rasmussen, L. A. (1999). Factors related to recidivism among

juvenile sexual offenders. Sexual Abuse: A Journal of Researchand Treatment, 11, 69–85.

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linearmodels: Applications and data analysis methods (2nd ed.).

Newbury Park, CA: Sage.

Rosenfeld, R., Messner, S. F., & Baumer, E. P. (2001). Social capital

and homicide. Social Forces, 80, 283–309.

Sampson, R. J., & Groves, W. B. (1989). Community structure and

crime: Testing social-disorganization theory. The AmericanJournal of Sociology, 94, 774–802.

Sampson, R. J., Morenoff, J. D., & Earls, F. (1999). Beyond social

capital: Spatial dynamics of collective efficacy for children.

American Sociological Review, 64, 633–660.

Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighbor-

hoods and violent crime: A multilevel study of collective

efficacy. Science, 277, 918–924.

Schwalbe, C. S., Fraser, M. W., Day, S. H., & Cooley, V. (2006).

Classifying juvenile offenders according to risk of recidivism.

Criminal Justice and Behavior, 33, 305–324.

Shaw, C. R., & McKay, H. D. (1942). Juvenile delinquency and urbanareas: A study of rates of delinquency in relation to differentialcharacteristics of local communities in American cities. Chi-

cago: University of Chicago Press.

1078 J Youth Adolescence (2010) 39:1067–1079

123

Simmons, R. D. (2001). Ecological factors that predict recidivism of

male juvenile offenders after release from institutional commit-

ment. Doctoral Dissertation, University of South Carolina,

Columbia, South Carolina.

Snyder, H. N., & Sickmund, M. (2006). Juvenile offenders andvictims: 2006 national report. Washington, DC: Office of

Juvenile Justice and Delinquency Prevention.

Tanenhaus, D. S. (2004). Juvenile justice in the making, studies incrime and public policy. New York: Oxford University Press.

Webster, C., MacDonald, R., & Simpson, M. (2006). Predicting

criminality? Risk factors, neighbourhood influence and desis-

tance. Youth Justice, 6, 7–22.

White, J. L., Moffitt, T. E., Earls, F., Robins, L., & Silva, P. A.

(1990). How early can we tell? Predictors of childhood conduct

disorder and adolescent delinquency. Criminology, 28, 507–527.

Widom, C. S. (1989). Child abuse, neglect and violent criminal

behavior. Criminology, 27, 251–271.

Wiebush, R. G., Baird, C., Krisberg, B., & Onek, D. (1995). Risk

assessment and classification for serious, violent, and chronic

juvenile offenders. In J. C. Howell, B. Krisberg, J. D. Hawkins,

& J. J. Wilson (Eds.), A sourcebook: Serious, violent, andchronic juvenile offenders. Beverly Hills, California: Sage.

Wikstrom, P. O. H., & Loeber, R. (2000). Do disadvantaged

neighborhoods cause well-adjusted children to become adoles-

cent delinquents? A study of male juvenile serious offending,

individual risk and protective factors, and neighborhood context.

Criminology, 38, 1109–1142.

Wilson, W. J. (1987). The Truly disadvantaged: The inner city, theunderclass, and public policy. Chicago, IL: University of

Chicago Press.

Wilson, W. J. (1996). When work disappears: The world of the newurban poor. New York: Random House.

Author Biographies

Heidi E. Grunwald is Deputy Program Director, Public Health Law

Research, Beasley School of Law, Temple University. Her current

research areas include research design and methodology, hierarchical

linear modeling, and causal analyses from observational data. Recent

publications have appeared in Research in Higher Education, Journal

of Higher Education, and Journal of Behavioral Health Services and

Research.

Brian Lockwood is a doctoral student in the Department of Criminal

Justice at Temple University. His recent research interests include

near—repeat offending the and social networks of juvenile offenders.

He has co-authored a book chapter that describes difficulties that can

arise when conducting spatial analyses.

Philip W. Harris is an Associate Professor in the Department of

Criminal Justice at Temple University. His research has focused

primarily on the areas of juvenile justice, juvenile correctional

strategies, and organizational and system development. Recent

publications have appeared in Criminology, Justice Quarterly,

Criminal Justice and Behavior, and Evaluation Review.

Jeremy Mennis is an Associate Professor in the Department of

Geography and Urban Studies at Temple University where he

specializes in Geographic Information Science. His recent research

has focused on modeling human-environment interaction for urban

public health and crime applications. Recent publications have

appeared in Annals of the Association of American Geographers,

Cartography and Geographic Information Science, and Drug and

Alcohol Dependence.

J Youth Adolescence (2010) 39:1067–1079 1079

123