Content uploaded by Jón Gunnar Bernburg
Author content
All content in this area was uploaded by Jón Gunnar Bernburg on Oct 19, 2017
Content may be subject to copyright.
LABELING, LIFE CHANCES, AND ADULT
CRIME: THE DIRECT AND INDIRECT
EFFECTS
OF
OFFICIAL INTERVENTION
IN ADOLESCENCE ON CRIME
IN EARLY ADULTHOOD*
JON
GUNNAR
BERNBURG
University
of
Iceland and Icelandic Research Council
MARVIN
D.
KROHN
University at Albany
Scholars have recently revitalized labeling theory as a developmental
theory
of
structural disadvantage. According to this approach, official
intervention increases the probability
of
involvement in subsequent
delinquency and deviance because intervention triggers exclusionary
processes that have negative consequences for conventional opportuni-
ties. The theory predicts that official intervention
in
adolescence
increases involvement in crime in early adulthood due to the negative
effect
of
intervention
on
educational attainment and employment.
Using panel data on urban males that span early adolescence through
early adulthood, we find considerable support for this revised labeling
approach. Official intervention in youth has a significant, positive
effect
on
crime in early adulthood, and this effect is partly mediated by
life chances such as educational achievement and employment.
KEYWORDS Labeling, life chances,
early
adult crime, official
intervention.
The labeling perspective
of
deviant behavior has been the subject
of
considerable debate among students
of
crime and deviant behavior. This
perspective argues that official intervention can be a stepping stone in the
*This article was prepared under Grant 86-JN-CX-0007 from the Office
of
Juvenile Justice and Delinquency Prevention, U.S. Department
of
Justice; Grant
5
R01
DA05512 from the National Institute on Drug Abuse; and Grant SBR09123299 from
the National Science Foundation. Work
on
this project was
also
aided by grants to the
Center for Social and Demographic Analysis at the University at Albany
from
NICHD
(P30 HD32041) and NSF (SBR- 9512290). Points
of
view or opinions in this document
are those
of
the authors and do not necessarily represent the official position
or
policies
of
the funding agencies. We thank Alan Lizotte, Steve Messner, Carolyn Smith, and
Terry Thornberry for comments and Pam Porter for editorial suggestions. Send
correspondence to Marvin Krohn, University at Albany, Department
of
Sociology,
1400
Washington Avenue, Albany, NY 12222 (e-mail: m.krohn@albany.edu).
CRIMINOLOGY VOLUME
41
NUMBER
4
2003
1287
1288
BERNBURGANDKROHN
development of a delinquent career. In the late seventies and early eight-
ies, critics argued that the labeling approach as originally presented
(Becker, 1963; Lemert, 1967) was vague and ambiguous and failed to pro-
vide empirically testable propositions. Moreover, research findings failed
to provide evidence consistent with the theory (Hirschi, 1980; Tittle, 1980).
However, recent work suggests that a rejection of the labeling approach
may have been unjustified. Efforts to modify labeling theory by explicat-
ing the social processes that translate deviant labeling into
a
deviant career
or “secondary deviance” and by providing empirically testable proposi-
tions regarding the consequences
of
deviant labeling (see Paternoster and
Iovanni, 1989, for
a
review of this issue) have provided theoretical clarity.
These theoretical developments have underscored the developmental
nature of labeling theory. Sampson and Laub (1997:138) have character-
ized labeling theory as “truly developmental in nature, because
of
its
explicit emphasis on processes over time” (see Loeber and LeBlanc, 1990).
Deviant labeling, official labeling in particular, is seen
as
a transitional
event that can substantially alter the life course by reducing opportunities
for a conventional life (Becker, 1963; Link, 1982; Link et al., 1989).
Thus,
labeling is seen as being indirectly related to subsequent behavior through
its negative impact on conventional opportunities. Sampson and Laub
(1997) suggest that labeling is one factor that leads to “cumulative disad-
vantage” in future life chances and, thereby, increases the probability of
involvement in delinquency and deviance during adulthood.
Recent reviews of labeling research have concluded that there is limited
research on mediational processes and the long-term effects of labeling
(Paternoster and Iovanni, 1989; Sampson and Laub, 1997). Moreover, the
methodological rigor
of
prior research on the consequences of official
labeling has been questioned, rendering an appreciable part of existing
evidence open for reevaluation. The present study uses panel data that
span early adolescence through early adulthood to explore the importance
of labeling in the development of structural disadvantage. Our main pur-
pose is to examine whether official intervention in adolescence increases
involvement in crime in early adulthood due to the negative effect
of
inter-
vention on educational attainment and employment.
DEVIANT LABELING AND
STRUCTURED OPPORTUNITIES
Common to the classic labeling theories (Becker, 1963; Lemert, 1967) is
the view that deviant labeling can have a profound, detrimental impact on
the person’s social standing and may thus be a crucial step in building a
stable pattern of deviant behavior. Due to the widespread cultural
imagery attached to deviant statuses, people tend to assume that deviants
LABELING, LIFE CHANCES,
AND ADULT
CRIME
1289
possess undesirable traits allegedly associated with their status. In
Lemert’s (1967:17) famous conceptualization
of
secondary deviance, devi-
ant behavior can be a “means of defense, attack, or adaptation to the overt
and covert problems created by the societal reaction to primary
deviation.”
Although often referred to as a single theory, within the labeling per-
spective, there are arguments that differ in important respects. This fact
has become increasingly apparent in recent theoretical work where schol-
ars have explicated the processes through which deviant labeling may
influence involvement in subsequent deviant behavior (Link, 1982; Link et
al., 1989; Matsueda, 1992; Paternoster and Iovanni, 1989; Sampson and
Laub, 1997). In
a
recent review
of
the literature, Liska and Messner
(1999:118-125) identlfy two major theoretical perspectives. First, deviant
labeling may influence subsequent deviance by altering the person’s self-
concept. This approach, which has been recently developed in the work of
Matsueda (1992), highlights the role of the self and the dynamics of sym-
bolic interaction; deviance amplification occurs when the labeled person
conforms to the stereotypical expectations of others. The second
approach focuses on the more tangible (social structural) aspects of social
exclusion; deviance is stabilized due to blocked access to structured oppor-
tunities and conventional others.
In the present paper, we focus on the latter perspective and empirically
examine the mediating role
of
structured opportunities. This perspective
has received attention recently in developmental criminology, specifically
in Sampson and Laub’s (1993, 1997) life-course approach. These authors
build on the systematic effort of Link and his colleagues
to
emphasize the
social structural implications
of
labeling theory (Link, 1982; Link
et
al.,
1989). Sampson and Laub describe how official intervention during ado-
lescence may negatively impact future life chances and, therefore, increase
the likelihood of later involvement in delinquency and deviance. This
approach views public labeling as
a
transitional event (where the deviant
label is not necessarily a permanent social status) that tends to push young
people on a trajectory of structural disadvantage and involvement in devi-
ance and crime. A key notion is that the application of a deviant label,
which is most successfully (but not only) achieved by official intervention
(Becker, 1963), during a crucial period in the life course tends to lead to
marginalization from conventionally structured opportunities, particularly
as
these are shaped by education and employment. In turn, due to these
problems, the likelihood of subsequent deviance increases. The exclusion-
ary and stigmatizing effect
of
deviant labeling may be an important
explanatory factor as to why some individuals continue to deviate later in
life (Sampson and Laub, 1997:147-148):
Cumulative disadvantage is generated most explicitly by the negative
1290
BERNBURGANDKROHN
structural consequences
of
criminal offending and official sanctions
for life chances. The theory specifically suggests a ‘‘snowball” effect
-
that adolescent delinquency and its negative consequences (e.g.,
arrest, official labeling, incarceration) increasingly “mortgage” one’s
future, especially later
life
chances molded by schooling and employ-
ment.
.
. .
The theoretical perspective in turn points to a possible indi-
rect effect of delinquency and official sanctioning
in
generating future
crime.
Official intervention may negatively affect educational attainment by
triggering stigma and exclusion in school. Bodwitch (1993) has found that
students defined as having a delinquent character by school officials are
subject to harsher disciplinary procedures, such as temporary suspension,
transfer to another school, or even expulsion. Also, incarceration can
directly impede educational opportunities. Educational attainment, in
turn, shapes employment opportunities in adulthood. In addition, an offi-
cial label may directly impede employment opportunities. First, many
employers may avoid hiring known delinquents (Schwartz and Skolnick,
1962), and second, individuals who have experienced official intervention
may expect and fear rejection from conventional others, including employ-
ers, and thus be less likely to apply for good jobs (Link, 1982). Blocked
educational and employment opportunities weaken the “social and institu-
tional bonds linking adults to society” (Sampson and Laub, 1997:144).
Over time, then, the social marginalization caused by the stigma attached
to the deviant label raises the likelihood
of
subsequent, even more stable,
involvement in deviant activity.
STRUCTURAL LOCATION AND LABELING
Clarifications and extensions
of
labeling theory have emphasized the
contingent nature of this theory. Official intervention, such
as
being
arrested, convicted,
or
sent to a mental hospital, does not automatically
lead
to
deviant labeling (Paternoster and Iovanni, 1989). Structural loca-
tion, such as race or social class, may provide people with differential
means to resist deviant labeling in the face of official intervention.
From a life-course perspective, Sampson and Laub (1997) argue that
disadvantaged structural location should facilitate labeling effects. “Defi-
cits and disadvantages pile up faster” among the disadvantaged (p. 153).
Due
to
higher stakes in conformity and continuity in social resources over
time, high structural location should diminish the effect of labeling. We
also suggest that deviant labeling of disadvantaged youths who are
processed by the police and the juvenile justice system is enhanced by the
negative stereotypes that are already associated with these youths in the
mainstream culture (see Gans, 1995). Conversely, other scholars have
LABELING, LIFE CHANCES, AND ADULT CRIME
1291
argued that higher status people may be more vulnerable to labeling than
are the disadvantaged because they have more to lose (Ageton and Elliott,
1974;
Jensen,
1972).
Although official intervention may, indeed, impact
higher status people more than it does lower status people, the argument
of Sampson and Laub
(1997)
concerning the impact
of
labeling on conven-
tional opportunities are compelling and more consistent with arguments
relating to the structural effects of labeling. Hence, we hypothesize that
the impact of official intervention on educational and employment oppor-
tunities and on early adult criminality will be stronger among the more
disadvantaged offenders. Specifically, we examine whether race and pov-
erty status moderate the effect of official intervention on educational and
employment opportunities.
The theoretical model is depicted in Figure
1.
Official intervention is
hypothesized to have a negative effect on educational attainment (Path a).
In turn, reduced educational attainment should increase unemployment in
early adulthood (Path
b).
In addition, official intervention may carry a
social stigma that blocks employment opportunities due to direct discrimi-
nation (Schwartz and Skolnick,
1962)
or expectation of rejection (Link,
1982),
independent of the effect such intervention has on educational
attainment. Hence, official intervention should have a significant effect on
unemployment, while controlling for educational attainment (Path c).
Both educational attainment and unemployment should have direct effects
on early adult crime (Paths d and e, respectively). Official intervention
should also have a direct positive effect on adult crime (Path f). However,
educational attainment and unemployment should mediate this effect on
adult crime (Path ad and Path ce, respectively). Finally, the theory implies
interaction effects; structural disadvantage should enhance the effect of
official intervention on educational attainment, unemployment, and adult
crime (Paths g, h, and i, respectively).
PRIOR
RESEARCH
Although numerous studies have examined various consequences of
official labeling (see Bernburg,
2002;
Palarma
et
al.,
1986;
and Paternoster
and Iovanni,
1989,
for reviews), methodological problems
of
prior research
limit what we can conclude from them. Specifically, much prior research
suffers from the following deficiencies:
(1)
many studies include only indi-
viduals who have experienced intervention,
(2)
most studies are either
cross-sectional or entail relatively short follow-up periods,
(3)
most studies
do not investigate intervening processes that may mediate the relationship
between official labeling and subsequent delinquent behavior, and
(4)
few
studies examine whether the relationship between labeling and both struc-
tural mediators and subsequent delinquency is contingent on structural
1292
BERNBURG AND
KROHN
Official
Intervention
in
Adolescence
Figure
1.
Hypothesized Effects of Official Intervention in
Adolescence on Crime in Early Adulthood
Crime in
Early
Adulthood
location. Our review of prior research is organized around these four
deficiencies.
SAMPLE SELECTION
Paternoster and Iovanni (1989) conclude their review of the labeling
literature by noting that the research has often been based
on
samples of
individuals drawn from police records and similar nonrandom sources
(e.g., Horowitz and Wasserman, 1979; Sampson and Laub, 1993; Smith and
Paternoster, 1990; Zhang and Messner, 1994). This strategy is problematic
for two reasons. First, comparisons are limited to a focus on the effects of
different sanctions (juvenile court appearance, probation, or incarcera-
tion), suggesting that the relative harshness or seriousness of the sanction
is related to the probability of subsequent delinquent behavior. The
researcher is thus limited to examining the relative, rather than the abso-
lute, effect of labeling. There is no comparison between those who have
not experienced official intervention and those who have. This is particu-
larly problematic considering that labeling theorists emphasize the impor-
tance of the initial labeling event as a dramatic event, a transition to a new
status (Becker, 1963; Lemert, 1967), as follows (Paternoster and Iovanni,
1989:385):
When one takes for study a group which appears at the end of a long
series of discretionary decisions, it is reasonable that the labeling pro-
cess has run its course by that time. Having already experienced a
repudiation of character and an exclusion from the normal routines
of
life, the “hard-core” offenders may be immune to additional labeling
effects.
Consistent with this view, the relatively few studies that have used samples
from the general population tend
to
support hypotheses derived from
LABELING, LIFE CHANCES, AND ADULT CRIME
1293
labeling theory (Farrington,
1977;
Hagan and Palloni,
1990;
Palarma et al.,
1986;
Ray and Downs,
1986;
Thomas and Bishop,
1984),
whereas studies
using samples of offenders have produced mixed results (e.g., Horowitz
and Wasserman,
1979;
Klein,
1974;
McEachern,
1968;
Smith and Paternos-
ter,
1990).
Paternoster and Iovanni
(1989)
suggest that an additional problem with
using samples of offenders is that too often the impact of the police is
overlooked. Most offenders who experience any official intervention will
have had an encounter with the police. Paternoster and Iovanni
(1989:383)
suggest:
Because only a small proportion of deviants undergo these exper-
iences, it may be more germane from a theoretical and a policy stand-
point to examine the consequences of an experience common to a
larger number of juveniles, such as encounters with the police.
TIMEFRAME OF THE STUDY
Although panel data with long follow-up periods are required to study
the long-term effects of labeling on structured opportunities, research on
labeling effects has mostly been based on cross-sectional data and panel
studies entailing short follow-up periods that “rarely include the develop-
mental transition from adolescence to adulthood” (Sampson and Laub,
1997:139-141;
examples include Ageton and Elliott,
1974;
Horowitz and
Wasserman,
1979;
Ray and Downs,
1986;
Smith and Paternoster,
1990;
and
Thomas and Bishop,
1984).
A strategy of examining short-term rather
than long-term effects is particularly problematic when emphasizing the
mediating effects of structural constraints. Labeling may occur early in the
life course, but impaired life chances may
slowly
unfold in the years to
come. Official intervention during adolescence may have an observable
impact as late as in early adulthood when individuals become fully
affected by the deviant label’s impact on life chances through diminished
chances of education and stable employment.
Although studies examining the long-term effects of official intervention
are limited, there is some evidence that supports this line of inquiry.
Hagan and Palloni
(1990)
found a significant effect
of
conviction in adoles-
cence on delinquent behavior in early adulthood, net of delinquency in
adolescence and numerous control variables. Sampson and Laub
(1993)
show a significant amplification effect
of
length of incarceration on adult
criminal behavior. However, the representativeness of these findings is
limited. The study by Hagan and Palloni is limited to British working-class
males, whereas Sampson and Laub limited their analysis of the effect of
official intervention to individuals incarcerated during adolescence.
1294
BERNBURGANDKROHN
INTERVENING PROCESSES
Labeling theory argues that deviant labeling sets in motion various, spe-
cific social processes that impact subsequent deviant behavior. However,
research has rarely examined the role of intervening processes in translat-
ing labeling into subsequent deviance. Although there is evidence sug-
gesting that official labeling negatively affects opportunities for
employment and education (Bodwitch, 1993; Freeman, 1991; Hagan, 1991;
Link, 1982; Schwartz and Skolnick, 1962; Sullivan, 1989), researchers have
failed to examine whether conventional opportunities mediate the rela-
tionship between official labeling and subsequent delinquent behavior.
Paternoster and Iovanni conclude that “by failing to consider the requisite
intervening effects, the bulk of these studies do not constitute a valid test
of labeling theory” (1989:384).
Although research is limited on this point, recent studies provide evi-
dence consistent with an intervening role of educational attainment and
employment stability in the labeling process. Sampson and Laub’s (1993)
study found that employment stability in adulthood mediates the relation-
ship between length of incarceration before age 17 and criminal behavior
in adulthood (controlling for prior delinquency and other important fac-
tors). Although not providing a formal test of mediational processes,
Sampson and Laub found that the effect of incarceration on subsequent
criminal behavior drops below the significance level when controlling for
job stability in adulthood.
There is some indirect evidence suggesting that official labeling may
have a negative impact on educational attainment. Hagan (1991) found
that police contact in adolescence negatively impacts occupational status
in early adulthood, partly through educational attainment. However, this
study is limited in that it controls for delinquent preferences rather than
actual delinquent activities in adolescence. In studying the effect of ado-
lescent delinquency on adult outcomes, Tanner et al. (1999) found that
educational attainment partly mediates the relationship between police
contact in adolescence and unemployment in adulthood, while controlling
for delinquent behavior in adolescence. Finally, Menard and Morse
(1984) provide some evidence supporting the notion that negative social
labeling by teachers and by significant others leads to school alienation.
CONTINGENT RELATIONSHIPS
There has been limited research on how structural location, particularly
race and social class, conditions the effect of labeling on subsequent life-
course opportunities and delinquency. Existing research has generated
inconsistent results. Hagan and Palloni (1990) found that being convicted
of a crime had greater positive effects
OR
subsequent delinquency for sons
LABELING, LIFE CHANCES, AND ADULT CRIME
1295
of parents who had also been convicted. They interpret this finding by
suggesting that official intervention when combined with a stigmatized sta-
tus (parental conviction) generates a stronger labeling effect. There is
some experimental evidence suggesting that arrest for domestic violence
increases the likelihood of subsequent violence only among individuals
who are unemployed (Berk et al., 1992; Sherman and Smith, 1992). Con-
versely, Ageton and Elliott (1974) found a weak but statistically significant
relationship between police contact and subsequent delinquent “orienta-
tions” for white juveniles but not for blacks (see also Jensen, 1972).
PRESENT
STUDY
The present study addresses hypotheses derived from structural labeling
theory with panel data on a stratified random sample of males living in
Rochester, New York. Data are available from when these youths were
on average 13.5 years old to when they were 22 years old. The panel
design, combined with the richness
of
the measurement space, allows us to
contribute to the literature on labeling theory in three significant ways.
First, we examine the effect of both police intervention and juvenile justice
intervention on subsequent early adulthood crime using a random sample
from a population of adolescents. Second, we examine the long-term
effect of official intervention during adolescence on young adult criminal-
ity and test whether educational attainment and periods of nonemploy-
ment mediate this effect. Finally, we test whether the effect of official
labeling on life chances and future crime is contingent on race and impov-
erished family background.
METHOD
The analysis is conducted with data from the Rochester Youth Develop-
ment Study (RYDS), a multiwave panel study of the development of
delinquent behavior among adolescents and young adults. This panel is
based on an initial sample
of
1,OOO
students selected from the seventh and
eighth grades
of
the public schools in Rochester, New York, during the
1987-1988 academic year. Interviews were conducted at 6-month intervals
over a 4%-year period (Waves
1
through 9) with each adolescent and his or
her parent or primary caretaker. After a 2%-year gap, adolescents and
parents were interviewed once a year for the next
3
years (Waves
10
through 12). All interviews were conducted in private settings; most were
face-to-face, but in later waves some long distance interviews were com-
pleted by telephone. Chronic truants and students who had left the Roch-
ester schools were interviewed at their homes, as were most parents. Data
on subjects were also collected from school, police, courts, and social ser-
vice agencies.
1296
BERNBURG AND
KROHN
SAMPLE
The Rochester Youth Developmental Study was designed to oversample
youth at high risk for serious delinquency and drug use because the base
rates for these behaviors are relatively low (Elliott et al., 1989).
To
accom-
plish this while being able to generalize the findings to a population of
urban adolescents, the following strategy was used. The target population
was limited to seventh- and eighth-grade students in the public schools of
Rochester, New York, a city that has a diverse population and a relatively
high crime rate.
The sample was then stratified on two dimensions. First, males were
oversampled
(75%
versus
25%)
because they are more likely than females
to be chronic offenders and to engage in serious delinquency (Blumstein
et al., 1986). Second, students from high crime areas of the city were over-
sampled on the premise that subjects residing in high crime areas are at
greater risk for offending.
To
identify high crime areas, each census tract
in Rochester was assigned a resident arrest rate reflecting the proportion
of the tract’s population arrested by the Rochester police in 1986.
Because the true probability of each adolescent being selected is known,
the sample can be weighted to represent all seventh and eighth graders in
the Rochester Public Schools. We weight the sample using the procedure
suggested by Kish (196577-9) in the analysis to follow.
There are
1,000
adolescents in the base panel. At Wave 12, 846 individ-
uals remained in the study. This represents a retention rate of 85%. Com-
paring the characteristics of respondents who are included in the present
analysis with the total sample indicates that attrition did not bias the sam-
ple (Krohn and Thornberry, 1999). We use only the male respondents in
the present study, a total
of
605
males.1 A deletion of cases due to missing
values results in 529 valid cases. Deleting the cases with missing values
does not change the demographic composition of our sample? The pre-
sent analysis is based on data covering a nine-year time period from when
1.
In the RYDS, questions tapping the homemaker role were not included.
Therefore, we are unable to measure nonemployment among females.
2.
There are no significant differences in the demographics between the cases
that remain in the analysis and those lost due to incomplete data
(p
>
.05;
two-tailed).
Neither is there a significant difference in our adolescent delinquency measure between
the two groups
(p
>
.05;
two-tailed).
To
determine
if
our results are affected by the
loss
of these cases, we replicated the analyses presented by using a Markov Chain Monte
Carlo multiple imputation procedure, which is in an experimental stage in
SAS (SAS
Institute,
2001;
option “PROC
MI”
and “PROC
MIANALIZE”
in
SAS;
see Schafer,
1997).
This procedure produces and combines several different sets
of
missing values
that reflect the uncertainty about the predictions
of
the unknown missing values. Using
this experimental procedure, we retain
582
of
the total sample
of
605
cases. The results
are substantively similar to those presented in this paper.
LABELING,
LIFE
CHANCES, AND ADULT
CRIME
1297
the subjects were approximately 13 years old until they were about
22
years old.
MEASURES
OFFICIAL INTERVENTION
Using police records, we construct a dummy variable labeled
police
intervention,
with
“1”
equal to having a recorded arrest or police contact in
Waves
1
through
7,
or approximately
in
the period between the ages of
13.5 to 16.5, and
“0”
equal to having neither arrest nor police contact in
this period.3 We do not have reliable official records about intervention
by the juvenile justice system during this period. However, we do have
self-reported data on involvement with the juvenile justice system for seri-
ous
forms of violence, property offenses, and drug use.
If
the subject indi-
cated that he had contact with the police, he was also asked if he had
further juvenile justice system involvement (put on probation, sent to a
correctional center, referred to community service, put in detention,
brought to court, or referred to a treatment program in Waves
1
through
7).
At each wave, the subject was asked only about the most serious inter-
vention; therefore, we only know
if
the subject had been through the juve-
nile justice system at least once. Thus, a second dummy variable,
juvenile
justice intervention,
was constructed with
“1”
equal to some involvement
with the juvenile justice system and
“0”
equal to no involvement. Because
these two measures are expected to overlap, we use them as alternative
measures of official intervention.
CRIMINAL BEHAVIOR
IN
EARLY
ADULTHOOD
Wave 10 and Wave
12
interviews included self-report inventories from
which information on offending over the past year was elicited. We com-
pute three measures of adult criminal behavior from the items in this
inventory.
To
measure involvement in criminal behavior at ages 19-20
(Wave lo), we use both a serious crime index (consisting of seven non-
overlapping items, including robbery, gang fights, attacks with a weapon,
breaking and entering, theft
of
$50-$100, theft of more than $100, and car
theft) and the number
of
drug sales in which subjects engaged. The seri-
ous crime measure was used in order to tap the type of street crime that
would be expected among this population as a result of official interven-
tion. Given the theoretical argument that intervention inhibits conven-
tional opportunities,
we
also include drug sales to provide a measure
of
crime that could be assumed to be economically motivated. At ages 21-22
3.
We have replicated the analyses reported
in
the present paper by using police
contact and police arrest separately. This strategy produces a simi!ar pattern
of
results.
1298
BERNBURGANDKROHN
(Wave 12), there is not sufficient variation on the serious crime index, and
therefore, we use
a
general crime index consisting of 32 nonoverlapping
items ranging in ocgree of seriousness from minor crimes such as vandal-
ism to major crimes such as robbery and assault. In addition, we use a
measure of drug sales similar to that from Wave
10.
We discuss statistical
issues regarding the dispersion of these variables in the Results section.
INTERVENING
VARIABLES
Information about educational attainment and employment stability is
obtained from the subject interviews. In New York State, a person is not
allowed to attend high school on a regular basis after the age of 20.
Hence, those subjects who report graduating from high school have done
so
no later than the time that we measure periods
of
nonemployment and
adult crime.4
A
dummy variable measures educational attainment, with
“1”
equal
to
graduating from high school and
“0”
equal to no high school
graduation (includes
GED).
Another education variable that indicates
whether the subject attended school in Waves
8
and 9
(1
=
yes;
0
=
no) is
also used. Periods
of
nonemployment are measured by the proportion of
months that the subjects reported being nonemployed in a given period.
We define an individual as nonemployed during a month that he is not
fully or partly employed, in school (college
or
high school),
or
serving in
the military. We take the square root of nonemployment to adjust for
skewness in the distribution (see Rummel, 1970:280-6). We use two mea-
sures of nonemployment. When
we
treat it as a dependent variable (Table
3)
we measure it for ages 19-22. However, when we examine the inter-
vening effect of nonemployment (on crime measured in Wave 12), we
measure
it
at Waves
10
and
11
to assure proper temporal order.
CONTROL VARIABLES
In the analyses to follow, we control for a number
of
variables that are
potentially related to official intervention, educational attainment, periods
of nonemployment, and adult crime. Specifically, we control for adoles-
cent serious delinquent behavior, race/ethnicity, parental poverty, and aca-
demic aptitude.
We measure adolescent serious delinquency with a self-report measure
of cumulative serious delinquent activities in Waves
2
through 7,
or
approximately between the ages of
14
and
16.
The measure consists of
seven nonoverlapping items, including robbery, gang fights, attack with a
4. Seven males graduated from high school after Wave
10
(which ends at about
age
20).
We have replicated the analyses (not shown) reported in Tables
3,
4,
and
5
coding these individuals as not having graduated. The results are substantively identical
to
those presented here.
LABELING, LIFE CHANCES, AND ADULT CRIME
1299
weapon, breaking and entering, theft of $50-$100, theft of more than
$100,
and car theft. A square root transformation
of
this variable is conducted
to adjust for skewness in the distribution (see Rummel, 1970).
Information about race/ethnicity was obtained in the first wave of inter-
views. Two dichotomous variables represent the racial/ethnic categories
African American and Hispanic, with white serving as the reference cate-
gory (coded
“0”).
Poverty status is measured by a dummy variable with
“1”
equal to the
household having an income below the poverty line and
“0”
equal
to
income above the poverty level. Information about the income
of
the
household is obtained directly in interviews with the primary caretakers of
the subjects in the first wave (when subjects were about 13.5 years of age).
If income information was not available from Wave 1, we used data from
Wave 2
or
3.
We measure subject’s academic aptitude with the percentile ranking on
the California Achievement Test in math in 1987 (when subjects are about
12 years of age). Thus, this measure serves as an indicator
of
academic
ability in the year preceding the first wave of interviews.5 Table
1
shows
the mean and standard deviation
of
the variables after weighting the sam-
ple. Experiencing juvenile justice intervention and particularly police
intervention is quite common among males in our sample. Among the 529
males included in the analyses, 39% had a record of police intervention
and 12% reported juvenile justice intervention between Waves
1
and 7.
RESULTS
Figure
1
will serve to organize the presentation of results, and the rele-
vant paths will be referred to in reporting the findings. We first examine
the effects of official intervention on educational attainment and employ-
ment. We then examine whether these measures
of
life chances mediate
the effect
of
official intervention on adult crime.
DOES OFFICIAL INTERVENTION INFLUENCE LIFE CHANCES?
In Table 2, we examine the effect of official intervention on educational
attainment (Path a in Figure 1). In Models
1
and 2, we present results
from a logistic regression of high school graduation by age 22 on official
intervention and the control variables. As predicted, both types of official
intervention in adolescence significantly reduce the odds
of
graduating
5.
In
six
cases, we substituted missing values on the math exam with percentile
ranking on the California reading exam. Excluding those
six
cases produces identical
findings
to
those reported here
(N
=
523).
We replicated the results (not shown) using
the reading exam instead of the math exam. This analysis also produced identical
results to those reported here.
1300
BERNBURG AND
KROHN
Table
1.
Variable Descriptives
(N
=
529)
Police intervention, Waves
1-7
(ages
13.5-16.5)
yes
=1,
Juvenile justice intervention, Waves
1-7
(ages
13.5-16.5)
Incidence
of
serious crime, Wave
10
(ages
19-20)
Incidence
of
general crime, Wave
12
(ages
21-22)
Incidence
of
drug selling, Wave
10
(ages
19-20)
Incidence of drug selling, Wave
12
(ages
21-22)
High school graduation
=
1,
otherwise
=
0
Attended school, Waves
8-9
(ages
17-17.5)
yes
=
1,
Nonemployment between nineteenth and twenty-second
Nonemployment in Waves
10
and
llb
Cumulative incidence
of
serious adolescent delinquency,
African American
=
1,
otherwise
=
0
Hispanic
=
1,
otherwise
=
0
Parental poverty
=
1,
otherwise
=
0
Math score percentile, age
12
(California Achievement
no
=
0
yes
=
1,
no
=
0
no
=
Oa
birthdaysb
Waves
2-7
(ages
14-16.5)b
Test)
Mean S.D.
.39
.12
.36
54.52
11.36
33.97
.53
.86
.32
.25
.88
-55
-16
-26
-60
.49
.32
3.76
214.86
65.64
219.27
SO
.35
.32
.32
1.62
.50
.37
.43
.27
NOTES:
Descriptives are weighted. Age indicates mean age unless otherwise
a
Subjects who left school by Wave
7
were deleted prior to calculating descriptive
specified.
statistics for this variable
(N
=
429).
Variable has been transformed by a square root transformation.
from high school, net of other variables in the model. The estimated odds
ratios indicate that official intervention decreases the odds in favor of high
school graduation by more than
70%.
The analysis in Models
1
and
2,
however, does not temporally separate
official intervention and subsequent educational disruption. In Models
3
and
4,
we examine whether official intervention (in Waves
1
through
7)
reduces the odds of attending school in a subsequent period (in Waves
8
and
9).6
To
ensure a clear temporal design, males who have left school by
Wave
7
(by about age
16.5)
are dropped from the analysis. The findings
6.
The RYDS data provide information on official intervention at six-month
intervals from age
13.5
(Wave
1)
to age
16.5
(Wave
7)
and on staying in school
to
age
17.5
(Wave
9).
We have separated intervention and staying in school temporally for a
total
of
six pairs
of
successive wavespans (not shown), regressing staying in school at the
LABELING, LIFE CHANCES, AND ADULT CRIME
1301
Table
2.
Logistic Regression
of
High School Graduation
and Attended School in Waves
8-9
Model
1 2 3
4
High School Attended School
Graduation in Waves
8-9
--
Independent Variables
Police Intervention
.26**
-
Juvenile Justice Intervention
-
.28**
-
.48*
.26**
-
Serious Adolescent
Delinquency
Parental Poverty
Math Score
African American
Hispanic
.65** .65** .75
.80
.58**
.49** 39
.80
l.Ol** 1.01** 1.02** 1.01**
1.79** 1.27 1.32 .89
.64 .61 1.16 1.14
Likelihood Ratio
141.0 112.9 31.1 15.0
d.
f.
6 6 6 6
N
529 529 429 429
NOTES:
Odds ratios are reported. In Models
3
and
4,
all
subjects who left school
by Wave
7
are deleted from the analysis.
*
p
<
.05;
**
p
<
.01
(one-tailed).
support the causal direction suggested by labeling theory. Experiencing
official intervention in adolescence is significantly associated with reduced
odds in favor of staying in school in a subsequent period.
We created product terms (table not presented) to examine whether the
effects
of
official intervention on educational attainment are dependent on
race and poverty status (Path g). The interaction terms (race
x
official
intervention, impoverished family background
x
official intervention)
were not significant
(p
>
.05;
one-tailed test). Thus, our findings do not
support the hypothesis that the effect
of
official intervention on educa-
tional attainment is contingent on structural location.
latter wavespan (Time
k)
on a variable coded
“1”
if the subject has experienced inter-
vention at any prior wave (Time
k
-
I).
For example, we looked at any reports of
official contact in Waves
1
through
3
and then at whether they stayed in school in
Waves
4
through 9, then any contact in Waves
1
through
4
and staying in school in
Waves
5
through 9, and
so
on.
To
ensure a clear temporal design, youths who leave
school during wavespan
k
-
I
were dropped from the analyses (minimum
N
=
429). The
findings from this analysis are similar to those reported in Table 2. Official intervention
during the
Kh
-
1
wavespan significantly reduces the odds in favor
of
staying in school
during the subsequent
Kh
wavespan. See Bernburg
(2002)
for an analysis
of
this issue.
1302
BERNBURG
AND
KROHN
In Table
3,
we assess the effect of official intervention on periods of
nonemployment in early adulthood (Paths ab and c).'
To
see the total
effect of official intervention on nonemployment, we first regress nonem-
ployment on intervention without including high school graduation (Mod-
els
1
and 3, respectively). As predicted, both police intervention and
juvenile justice intervention are positively and significantly related to peri-
ods of nonemployment in adulthood (Path c).*
Educational attainment may mediate some
of
the effect of official inter-
vention on nonemployment (Path ab). We examine this possibility in
Models
2
and
4
by adding high school graduation. Educational attainment
has a sizeable and significant negative effect on nonemployment and
reduces the regression coefficient for police intervention by 33% (from
.15
to
.lo)
and the coefficient for juvenile justice intervention by
40%
(from
.10 to .06). Sobel's (1982) test of the significance of indirect effects indi-
cates that educational attainment significantly mediates part of the effect
of both police intervention and juvenile justice intervention on nonem-
ployment (t
=
3.94 and 2.76, respectively).9 However, the effect of police
intervention
on
nonemployment in early adulthood remains statistically
significant and positive (Path c).
We created product terms (table not presented) to examine
if
the effect
of official labeling
on
nonemployment is contingent
on
race and impover-
ished family background (Path h). The product terms were not statisti-
cally significant.
LABELING, LIFE CHANCES, AND ADULT CRIME
Does official intervention affect involvement in serious criminal activity
in early adulthood?
As
our dependent variable in this part of the analysis
is measured as event counts, the use
of
classic linear modeling is problem-
atic (Beck and Tolnay, 1995; King, 1988). To deal with this issue, we use a
7.
White's (1980) test for heteroscedasticity indicates that some of the models in
Table
3
violate the assumption of homoscedasticity.
To
adjust for heteroscedasticity in
the error variances, we follow White (1980) and obtain the standard errors from the
consistent covariance matrix.
Forty-two males reported that they had been in a correctional facility for at
least a month during the period when nonemployment was measured.
As
an individual
may be defined as nonemployed due
to
incarceration, we have replicated the analysis
in
Tables
3,6,
and
7
(not shown), controlling for the number
of
months subjects were in
a
correctional facility. The results from this analysis are substantively identical to those
reported here.
9. Sobel (1982) derives the following formula to obtain the standard error for an
indirect Path
xz:
SQRT
(z2S,'
+
2s:
+
S,'S,'),
where
x
is the unstandardized effect
of
the independent variable on the mediator and
z
the effect
of
the mediator on the
dependent variable.
8.
LABELING,
LIFE
CHANCES,
AND
ADULT CRIME 1303
Table
3.
OLS
Regression
of
Nonemployment
in
Early
Adulthood (ages
19-22)
Independent Variables
Police Intervention
Juvenile Justice Intervention
Serious Adolescent Delinquency
Parental Poverty
Math Score
African American
Hispanic
High School Graduation
Intercept
R2
Model
1
2
3
4
_________-
-
.06
-.oo
(.05)
(-01)
(-04)
(.001)
(-03)
(-05)
(*03)
.09**
-.001**
.16**
.09
-.19**
.36
.21 .27
.18
.25
NOTES:
N
=
529.
Unstandardized coefficients are reported. Standard errors are
in parentheses. (The standard errors are obtained from the consistent covariance
matrix to adjust
for
heteroscedasticity.)
*
p
<
.05;
**
p
<
.01
(one-tailed).
modified Poisson regression procedure.10 Tables
4
and
5
present a series
of Poisson regressions where the dependent variable is the number of seri-
ous crimes and drug sales in which males engaged at ages
19-20.
In Tables
6
and
7,
we
use
general crime and drug selling at ages
21-22
as the depen-
dent variables.
10. The standard Poisson regression assumes that the conditional mean of the
dependent variable is equal
to
its variance, an assumption that often does not hold
empirically for count variables, and does not seem to hold for our dependent variables
(see Hagan and Palloni, 1990, Sampson and Laub, 1993). The modified Poisson regres-
sion relaxes this restriction by adding an unknown error term to
the
Poisson regression
(Beck and Tolnay, 1995). This unknown parameter and the effect parameters are then
estimated using quasi-maximum likelihood.
As
our dependent variables display
overdispersion, this method produces larger standard errors than the standard Poisson
regression.
1304
BERNBURGANDKROHN
We first consider the effects of police intervention and juvenile justice
intervention on involvement in serious crime at ages 19-20 (Table
4).
Model
1
shows the effect
of
police intervention on serious crime, without
controlling for educational attainment. As predicted, police intervention
is significantly associated with increased serious crime in early adulthood,
net of the control variables (Path f). The odds ratio for this effect indi-
cates that police intervention in youth increases the predicted number of
crime events at ages 19-20 by a factor of 1.63 (e.")."
Labeling theory emphasizes the indirect effects of criminal labeling on
subsequent deviant behavior. We have hypothesized that educational
attainment should mediate some
of
the effects of official intervention on
subsequent crime (Path ad). Consistent with this hypothesis, adding high
school graduation to the equation (Model 2) produces a substantial drop
in the regression coefficient for the effect of police intervention on early
adult crime, about a
40%
drop (compare Models
1
and 2). Moreover,
educational attainment has significant, negative effects on adult crime
(Path d). Educational attainment significantly mediates a part of the
effect
of
police intervention on crime at ages 19-20 (t
=
2.40).12
Models
3
and
4
present the same procedure for juvenile justice interven-
tion. The results in Model
3
show that juvenile justice intervention has
significant effects on crime at ages 19-20 (Path f), increasing the predicted
number of serious criminal events by a factor of 5.31 Adding edu-
cational attainment to the equation produces an
8%
drop in the regression
coefficient for the effect of juvenile justice intervention on crime (compare
Models 3 and
4).
The indirect effect of juvenile justice intervention on
crime through educational attainment (Path ad) is statistically significant
(t
=
1.67; one-tailed).
We have hypothesized that the effect of official intervention in youth on
early adult crime is contingent on structural location (Path i). As disad-
vantage cumulates more rapidly among the already disadvantaged, and
because powerless groups may be relatively less able to resist public label-
ing when official intervention occurs, official intervention should have a
stronger effect on subsequent deviance among minorities and people with
impoverished backgrounds (Sampson and Laub, 1997). We assess this
hypothesis by creating product terms, which
we
estimate simultaneously to
test for the presence
of
an interaction effect between official intervention
11.
Replicating this analysis using the general crime index instead of the serious
crime index produces similar results.
12.
It
is possible that in some cases subjects have left high school prior
to
official
intervention.
To
ensure that the results concerning the mediating role of educational
attainment are
robust
to
this possibility, we have added a dummy variable
for
leaving
school by Wave
7
(analysis not shown). This procedure does not change the results
reported in Tables
4
through
7.
LABELING, LIFE CHANCES, AND ADULT
CRIME
1305
Table
4.
Modified
Poisson
Regression of Serious Crime at
Ages
19-20
(Wave
10)
Model
1
2
3
4
5
6
------
Independent Variables
Police Intervention
.49* .30
-
-
-.41
-
Juvenile Justice
-
-
1.67** 1.53**
-
-.36
(54)
(.26)
(.26)
Intervention
(.26) (.26) (.71)
Delinquency
(.W
(.W
(.W
(.05)
(.MI
(35)
Parental Poverty
.63**
.60**
.63** .60** -.87
-.04
(.24) (.23) (.23) (.23)
(.a)
(.37)
Math Score
.002
.005
.008
.009 ,002
.001
(.005) (.005)
(.005)
(.005)
(.005)
(.005)
African American
.54 .62* .55
.58*
.41 .19
(.33) (.33) (.32) (.32) (.42) (.36)
Hispanic
-.42 -.38 -.35 -.35 -.25 .12
(.a)
(.47) (51) (.47)
(.50) (.49)
Graduation
(.29) (.28) (.29) (.28)
Serious Adolescent
.24** .22** .15** .14** .24** .15**
-
-.58**
-.81** -.63**
High School
-
-.76**
Interaction Terms
Official Intervention
-
-
-
-
.41 1.86**
Official Intervention
-
- - -
2.01** 1.24**
x
African American
(.58) (.72)
x
Parental Poverty
(.70) (.49)
Intercept
-2.28 -2.06 -2.73 -2.53 -1.93 -2.47
Deviance
1,263 1,245 1,185 1,175 1,235 1,148
d.
f.
522 521 522 521 519 519
NOTES:
N
=
529.
Unstandardized coefficients are reported. Modified standard errors are
in parentheses. The standard errors from the Poisson regression are multiplied by a scale
parameter, which is estimated by the square root
of
the deviance divided by the degrees of
freedom.
*
p
<
.05;
**
p
<
.01
(one-tailed).
and structural location. In terms of race/ethnicity, we estimate a product
term for African Americans only, collapsing whites and Hispanics into a
single reference category.13
In Model
5,
the interaction effect between police intervention and
impoverished status of family is statistically significant. The interaction
term is positive, which means that the positive effect of police intervention
on crime is stronger among males who have impoverished family back-
grounds. The effects of police intervention do not differ significantly by
13.
Due to collinearity problems, we cannot estimate product terms for all three
raciavethnic groups
in
our data.
Due
to the unique societal status of African Ameri-
cans, it seems most appropriate to estimate effects
for
this group.
1306
BERNBURG AND KROHN
race/ethnicity. In Model 6, both product terms are statistically significant.
Juvenile justice intervention has stronger effects on adult crime among
African Americans as well as among males from impoverished back-
grounds. In sum, the results suggest that official intervention has stronger
crime amplification effects among the disadvantaged.
In Table
5,
we replicate the analysis in Table
4
by using the number of
drug sales at ages 19-20 as the dependent variable.
As
before, both police
intervention and juvenile justice intervention have significant, positive
effects on drug selling (Path
f;
Models
1
and
3,
respectively). Moreover,
educational attainment accounts for more than
50%
of
the effect of police
intervention (compare Models
1
and 2) and about 20% of the effect of
juvenile justice intervention (compare Models
3
and
4).
As
predicted,
educational attainment significantly mediates a part
of
the effect
of
police
intervention and juvenile justice intervention on drug selling (Path ad;
t
=
3.08
and 2.26, respectively).
In Models
5
and 6, we examine the interaction effects (Path i). Only
one of the product terms is statistically significant in the direction pre-
dicted.14 Thus, the results in Model
6
indicate that the effect of juvenile
justice intervention on drug selling is stronger among those from impover-
ished family backgrounds.
Theorists have highlighted the role of reduced employment opportuni-
ties in mediating the effects of labeling on adulthood crime (Sampson and
Laub, 1993, 1997). In Tables
6
and
7,
we
further examine the mediating
role
of
nonemployment (Path ce).ls To ensure a temporal separation
between nonemployment and subsequent crime, we use a measure of gen-
eral crime at ages 21-22 (Wave 12) and
of
nonemployment at ages 19-21
(in Waves
10-11).
As
there was insufficient variation on the serious crime
14. The interaction term for African American status in Model
5
is significant in
the opposite direction to what we have predicted. This finding indicates that police
intervention actually has weaker effects on drug selling at ages 19-20 among African
Americans. We should recognize that there are ambiguities in the literature regarding
how race moderates the effects
of
official labeling. For example, Ageton and Elliott
(1974) found significant labeling effects for whites only, whereas Palarma et al.
(1986)
report the opposite finding. Our data do not allow
us
to examine the potential reasons
why, for this type of crime, African-American status decreases the impact
of
police
intervention
on
subsequent crime. It may be because
of
the different way in which the
event
is
perceived among individuals in areas where these activities are more prevalent.
Again, this is an issue to be examined in future research.
As individuals are defined as nonemployed during periods in which they are
not in college, it is possible that some
of
the observed effects
of
nonemployment on
subsequent crime could be confounded with the effects
of
not being in college. We
have replicated the analysis in Tables
6
and 7 (not shown) controlling for a dummy
variable indicating whether subjects were in college during any semester that falls in the
Wave 10 to Wave 11 period. The results from that analysis are similar to those reported
here.
15.
LABELING, LIFE CHANCES, AND ADULT
CRIME
1307
Table
5.
Modified Poisson Regression
of
Drug Selling at
Ages
19-20
(Wave
10)
Independent Variables
Police Intervention
Juvenile Justice
Intervention
Serious Adolescent
Delinquency
Parental Poverty
Math Score
African American
Hispanic
High School
Graduation
Interaction Terms
Official Intervention
x
African American
Official Intervention
x
Parental poverty
Intercept
Deviance
d.
f.
Model
2 3 4
---
5
6
1.51**
C45)
-
.19**
(.W
.52
(.35)
.023
(.005)
1.78**
(.42)
.46
(.38)
-.go**
(3)
-1.75
-.12
-.47
32,071
519
(.48)
(.W
NOTES:
N
=
529.
Unstandardized coefficients are reported. Modified standard errors are
in parentheses. The standard errors from the Poisson regression are multiplied by
a
scale
parameter, which is estimated by the square root of the deviance divided by the degrees of
freedom.
*
p
<
.05;
**
p
<
.01
(one-tailed).
measure at ages
21-22
(there are only
16
cases in the nonzero category),
we replace this measure with a measure of general crime in Models
1
through
4
for this age span.16
Models
1
and
3
in
Table
6
show the effects of police intervention and
juvenile justice intervention, respectively, on crime at ages
21-22.
Police
and juvenile justice intervention have significant, positive effects on early
16.
When we did the analysis using the serious crime index at ages
21-22,
we
obtained findings similar to those reported in Table
6.
However, coefficients
for
some
of
the control variables in
our
model were unstable due to
an
insufficient number
of
cases.
1308
BERNBURGANDKROHN
adult crime, further demonstrating the long-term impact of formal crimi-
nal intervention during adolescence on adult crime (Path
f).
Educational
attainment is not statistically significant, however, even without control-
ling for nonemployment (results not shown). When both educational
attainment and nonemployment are taken into account in Models
2
and
4,
nonemployment but not educational attainment is significantly related to
subsequent crime (Path e).
Table
6.
Modified Poisson Regression
of
General Crime
at
Ages
21-22
(Wave 12)
Independent Variables
Police Intervention
Juvenile Justice
Intervention
Serious Adolescent
Delinquency
Parental Poverty
Math Score
African American
Hispanic
High School
Graduation
Nonemployment
(ages
19-21)
Interaction Terms
Official Intervention
x
African American
Official Intervention
x
Parental Poverty
Intercept
Deviance
d. f.
522 520 522 520
518
518
NOTES:
N
=
529.
Unstandardized coefficients are reported. Modified standard errors are
in parentheses. The standard errors from the Poisson regression are multiplied by a scale
parameter, which is estimated by the square root
of
the deviance divided by the degrees
of
freedom.
*
p
<
.05;
**
p
<
.01
(one-tailed).
As
predicted, periods of nonemployment significantly mediate the effect
of police intervention on crime at ages
21-22
(t
=
2.50),
producing a
38%
LABELING, LIFE CHANCES, AND ADULT CRIME
1309
drop in the effect of police intervention (Path ce; compare Models
1
and
2).17 Nonemployment produces
a
10%
drop in the effect
of
juvenile jus-
tice intervention (Models
3
and
4),
but this mediation effect is not signifi-
cant (t
=
1.60).
Finally, product terms are added in Models
5
and
6.
Only one product
term is significant (see Model
6).
As predicted, the results indicate that
the positive effect
of
juvenile justice intervention
on
subsequent crime is
stronger for African Americans than for whites and Hispanics (Path i).
The effect of official intervention on crime at ages 21-22, however, is not
significantly enhanced by impoverished family background.
In Table
7,
we replicate this procedure for drug selling at ages 21-22.
Again, police and juvenile justice interventions are significantly and posi-
tively associated with drug selling (Path
f;
Models
1
and
3,
respectively).
Adding educational attainment and periods of nonemployment to the
equations in Models 2 and
4
reduces the effects of official intervention on
adult crime, as expected. The intervening variables account for about
40%
of the effect of police intervention and about
9%
of the effect of juvenile
justice intervention. Nonemployment has a significant and positive effect
on subsequent involvement in drug sales (Path e) and significantly medi-
ates the effects of police intervention on subsequent drug selling (Path ce;
t
=
3.45).
Nonemployment, however, does not significantly mediate the
effects of juvenile justice intervention on drug selling. Although educa-
tional attainment has a substantially significant effect
on
drug selling (not
shown), nonemployment fully accounts for these effects. Finally, one of
the product terms in Models
5
and
6
is statistically significant. Again, the
effect of juvenile justice intervention
on
adult drug selling is stronger
among African Americans.
To
summarize our findings, we combine the results from equations pre-
dicting high school graduation and periods of nonemployment with those
predicting early adult crime in path models (Figures
2
through
5).
The
figures provide an overview of the paths depicting direct and indirect
effects. Overall, the overview shows that our results provide consistent
support for the hypotheses presented in Figure
1.
Official intervention in
adolescence has positive indirect effects on adult crime through reduced
educational attainment and nonemployment across alternative measures
of intervention and adult crime. High school graduation is indirectly
related to the measures of adult crime through nonemployment, whereas
nonemployment
is
directly related to these outcome measures.
17. We replicated the analysis in Table
3
by regressing nonemployment at ages
19-21 (Waves 10
to
11). The coefficients and standard errors
from
this analysis were
used along with the coefficients and the standard errors in Tables
6
and 7 in testing the
significance
of
the mediating effects
of
nonemployment (Sobel, 1982).
1310
BERNBURGANDKROHN
Table
7.
Modified Poisson Regression of Drug Selling
at
Ages 21-22
(Wave
12)
Independent Variables
Police Intervention
Juvenile Justice
Intervention
Serious Adolescent
Delinquency
Parental Poverty
Math Score
African American
Hispanic
High School
Graduation
Nonemployment
(ages
19-21)
Interaction Terms
Official intervention
x
African American
Official Intervention
x
Parental Poverty
Intercept
Deviance
d. f.
1
2
3 4
Model
JLL
520
5
1.26**
(.63)
-
.06
-.54
(.05)
(52)
(.004)
.008
3.55**
2.58**
(1.01)
(.97)
(.27)
(.35)
-.24
.17**
-1.09
(.66)
(58)
.63
-1.14
74,816
518
6
-
.46
-.02
.01
,009
2.79**
2.91**
(.74)
(.W
(.29)
(.004)
(.92)
(.94)
(.26)
(.34)
-.09
.18**
1.21*
-.20
-.91
(.74)
(.45)
71,161
518
NOTES:
N
=
529.
Unstandardized coefficients are reported. Modified standard errors are
in parentheses.
The
standard
errors
from the Poisson regression are multiplied by a scale
parameter, which is estimated by the square root
of
the deviance divided by the degrees
of
freedom.
*
p
c
.05;
**
p
c
.01
(one-tailed).
DISCUSSION
We examined whether official labeling increases the probability of
involvement in subsequent crime and deviance by triggering processes that
have negative consequences for conventional opportunities. We hypothe-
sized that official intervention in adolescence increases involvement in
crime in early adulthood due
to
the negative effect of intervention on edu-
cational attainment and employment. Using panel design data that spans
early adolescence through early adulthood, our findings lend considerable
LABELING, LIFE CHANCES, AND ADULT
CRIME
1311
Figure
2.
Effects of Police Intervention on General Crime
(Unstandardized Coefficients)
Police
Intervention
-.17"
General
Crime
Ages
21-22
Police
Intervention
*
p
<
.05;
**
p
<
.01
(one-tailed).
-.17"
1.26"
Dw
b
Selling
Ages
2
1-22
support to the structural implications of the labeling approach (Becker,
1963; Link, 1982; Link et al., 1989; Sampson and Laub, 1997).
The present analysis goes further than most prior studies by temporally
Figure
3.
Effects
of
Police Intervention on Drug Selling
(Unstandardized Coefficients)
*
p
<
.05;
**
p
c
.01
(one-tailed).
separating official intervention from subsequent life-course outcomes and
criminal behavior. We find that official intervention affects educational
attainment by decreasing the odds that those labeled will graduate from
high school. In turn, educational attainment has a direct effect on employ-
ment and mediates the long-term effect of official intervention on adult
crime. Further research is needed to determine the processes by which
1312
BERNBURGANDKROHN
Juvenile
Justice
A
Figure
4.
Effects of Juvenile Justice Intervention on
General Crime (Unstandardized
Coefficients)
-.19** .02 General
F
Crime
Juvenile
Justice
Intervention
A
*
p
<
.05;
**
p
<
.01
(one-tailed).
-.19"
.46
Dw
b
Selling
Ages 21-22
official labeling affects educational attainment. However, Bodwitch
(1993)
provides some evidence that may account for these effects. In an
ethnographic study
of
high school students, Bodwitch found that students
1.21'
Figure
5.
Effects of Juvenile Justice Intervention on Drug
Selling (Unstandardized Coefficients)
v
Nonemplo yment
Ages 19-2
1
*
p
<
.05;
**
p
<
.01
(one-tailed).
who are defined as troublemakers are disciplined by school officials in
ways that conspire to push students out of school. One
way
in which stu-
dents acquire the troublemaker label may be involvement with juvenile
LABELING,
LIFE
CHANCES, AND ADULT CRIME
1313
justice authorities.18 Although the current study cannot address the spe-
cific processes involved, results do confirm a link among official interven-
tion, educational attainment and subsequent employment problems, and
criminal involvement.
We have replicated results from other studies (Freeman, 1991; Schwartz
and Skolnick, 1962; Western and Beckett, 1999) indicating that official
intervention during adolescence negatively influences employment in
young adulthood. Moreover, we go beyond prior research by showing that
periods of nonemployment, along with educational attainment, partially
mediate the effect of official intervention on adult involvement in crime.
Our findings thus underscore the role
of
socioeconomic life-course out-
comes in transforming official intervention into subsequent criminal
involvement (Sampson and Laub, 1993, 1997).
Our results are consistent with the hypothesis that official intervention
during adolescence influences criminal involvement as late as early adult-
hood when individuals become fully affected by blocked life chances
shaped by education and employment. But, in addition to the indirect
effects that both intervention by the police and the juvenile justice system
have
on
adult crime, official intervention also directly influences subse-
quent criminality. The fact that official intervention has a direct effect on
adult crime even after controlling for educational attainment, employ-
ment, and a number
of
other potential covariates, including adolescent
delinquent behavior, may indicate the influential role that such an experi-
ence has on the lives
of
those
so
labeled, especially African-American
males. It also suggests that we have not measured other intervening
processes that may be at work. For example, labeling theorists have
emphasized the importance of the development
of
a deviant identity
(Lemert, 1967; Matsueda, 1992), blocked access to conventional others
(Becker, 1963; Link et al., 1989; Sampson and Laub, 1997; Zhang and
Messner, 1994), and association with unconventional others (Adams, 1996;
Becker, 1963; Bernburg
et
al., 2002). Future research should examine the
role of these processes in translating the impact of official labeling on sub-
sequent deviance.
Some theorists have argued that disadvantaged structural location
should facilitate labeling effects (Sampson and Laub, 1997). Other
researchers suggest that labeling has more of an effect on people
of
18.
We have examined this possibility in a supplementary analysis (not shown).
Subjects were asked if school officials were told about their involvement with the
police/justice system. In an analysis deleting
27
males who had dropped out of school
by age
15,
we found that males who report school notification by age
15
are significantly
less likely
to
graduate from high school, net
of
controls
(p
<
.01,
one-tailed test). See
Bernburg
(2002)
for further analysis of this issue.
1314
BERNBURG AND
KROHN
advantaged location because these people have higher stakes in maintain-
ing a nondeviant status (Ageton and Elliott,
1974).
To
date, there has
been limited resel::ch on this issue. Although our findings are not conclu-
sive on this point, they lend more consistent support to the former
approach. The effect
of
police and juvenile justice intervention on early
adult crime is significantly stronger among males with impoverished back-
grounds. Also, the effect of juvenile justice intervention on adult crime is
significantly stronger among African Americans. It may be that a higher
structural location provides people with the necessary resources and com-
mitment to conventional pursuits to resist deviant labeling in the face of
official intervention (Sampson and Laub,
1997).
Moreover, deviant label-
ing may be more easily triggered when impoverished youths and African-
American youths are processed by the police and the juvenile justice sys-
tem, since negative stereotypes are already associated with these groups in
the mainstream culture (see Gans,
1995).
This study contributes to our understanding of the impact of official
intervention on the life course of individuals by addressing some of the
weaknesses of past research efforts. Specifically, we examined the impact
of labeling among a sample of males drawn from a general urban popula-
tion. The longitudinal panel design enables
us
to temporally distinguish
the effects of official intervention on adult crime and to examine the inter-
vening processes of educational attainment and employment. We include
the most commonly experienced official intervention, police intervention,
which has frequently been overlooked (Paternoster and Iovanni,
1989),
and we examine whether the effect
of
official intervention is contingent on
social structural factors. One important limitation of the study is that it
includes only males, a limitation that must be addressed in future research.
Overall, we have found that hypotheses from the labeling approach con-
cerning the effect that official intervention has on important dimensions of
life chances and subsequent criminal behavior are supported. The findings
attest to the viability
of
the labeling approach for explaining secondary
deviance and point to the processes that may account for the impact of
official intervention on subsequent criminal behavior.
REFERENCES
Adams, Mike
S.
1996
Labeling and differential association: Towards a general social learning the-
ory
of
crime and deviance. American Journal
of
Criminal Justice
20
149-1 64.
Ageton, Suzanne and Delbert Elliott
1974
The effect
of
legal processing on delinquent orientations. Social Problems
22:87-100.
LABELING, LIFE CHANCES, AND ADULT CRIME
1315
Beck,
E.
M. and Stewart
E.
Tolnay
1995 Analyzing historical count data: Poisson and negative binomial regression
models. Historical Methods 28125-131.
Outsiders: Studies in the Sociology
of
Deviance. New York: Free Press.
The deterrent effect of arrest in incidents
of
domestic violence: A Bayesian
analysis
of
four field experiments. American Sociological Review
Becker, Howard
Berk, Richard A., Alec Campbell, Ruth Klap, and Bruce Western
1963
1992
57698-708,
Bernburg, J6n Gunnar
2002 State reaction, life-outcomes, and structural disadvantage: A panel study of
the impact
of
formal criminal labeling on the transition to adulthood. Ph.D.
Dissertation, Department of Sociology, University at Albany.
Formal criminal intervention, deviant networks, and subsequent delin-
quency: A test of structural labeling theory. Presented at the Annual Meet-
ing
of
the American Society
of
Criminology, Chicago,
111.
Criminal Careers and Career Criminals. Washington, D.C.: National Acad-
emy Press.
Getting rid
of
troublemakers: High school disciplinary procedures and the
production
of
dropouts. Social Problems 40493-509.
Multiple Problem Youth: Delinquency, Substance Use, and Mental Health
Problems. New York: Springer-Verlag.
The effects of public labelling. British Journal of Criminology 17:112-125.
Bernburg, J6n Gunnar, Craig Rivera, and Marvin D. Krohn
2002
Blumstein, Alfred, Jacqueline Cohen, Jeffrey A. Roth, and Christy A. Visher
1986
Bodwitch, Christine
1993
Elliott, Delbert
S.,
David
S.
Huizinga, and Scott Menard
1989
Farrington, David P.
Freeman, Richard
1977
1991 Crime and Employment of Disadvantaged Youth. Cambridge, Mass.:
Harvard University, National Bureau of Economic Research.
The War Against the Poor: The Underclass and Antipoverty Policy. New
York: Basic Books.
Destiny and drift: Subcultural preferences, status attainment, and the risk
and rewards of youth. American Sociological Review 56567-582.
The social reproduction
of
a criminal class in working-class London,
1950-1980. American Journal
of
Sociology 96:265-299.
Labelling theory and juvenile delinquency: An assessment of the evidence.
In Walter Gove (ed.), The Labelling of Deviance: Evaluating a Perspective.
2d
ed. New York: Wiley.
Gans, Herbert
J.
1995
Hagan, John
1991
Hagan, John and Alberto Palloni
1990
Hirschi, Travis
1980
1316
BERNBURGANDKROHN
Horowitz, Allan and Michael Wasserman
1979
The effect
of
social control
on
delinquent behavior: A longitudinal test.
Sociological Focus
1253-70.
Delinquency and adolescents’ self-conceptions: A study
of
the personal rele-
vance
of
infraction. Social Problems
2084-102.
Jensen, Gary
F.
1972
King, Gary
1988
Statistical models for political science event counts: Bias in conventional
procedures and evidence for the exponential Poisson regression model.
American Journal of Political Science
325338463.
Survey Sampling. New York: Wiley.
Labeling, deterrence and recidivism: A study
of
police dispositions of juve-
nile offenders. Social Problems
22:292-303.
Retention
of
minority populations in panel studies
of
drug use. Drugs and
Society
14:185-208.
Human Deviance, Social Problems and Social Control. Englewood Cliffs,
N.J.: Prentice-Hall.
Mental patient status, work, and income: An examination
of
the effects
of
a
psychiatric label. American Sociological Review
47:202-215.
Link, Bruce, Francis
T.
Cullen, Elmer Struening, Patrick E. Shrout, and Bruce P.
A modified labeling theory approach to mental disorders: An empirical
assessment. American Sociological Review
54:400-423.
Perspectives
on
Crime and Deviance. Englewood Cliffs, N.J.: Prentice
Hall.
Toward a developmental criminology. In Michael Tonry and Norval Morris
(eds.), Crime and Justice,
Vol. 7.
Chicago: University
of
Chicago Press.
Reflected appraisal, parental labeling, and delinquency: Specifying a
sym-
bolic interactionist theory. American Journal of Sociology
97:1577-1611.
The juvenile probation system. American Behavioral Scientist
11:1-45.
A structuralist critique
of
the 1.Q.-Delinquency Hypothesis: Theory and evi-
dence. American Journal
of
Sociology
891347-1378.
The
effect
of
police and mental health intervention
on
juvenile deviance:
Specifying contingencies in the impact
of
formal reaction. Journal
of
Health
and Social Behavior
27:90-105.
Kish, Leslie
Klein, Malcolm
W.
1965
1974
Krohn, Marvin D. and Terence
P.
Thornberry
1999
Lemert, Edwin
1967
Link, Bruce
1982
Dohrenwend
1989
Liska, Allen
E.
and Steven
F.
Messner
Loeber, Rolf and Marc LeBlanc
1999
1990
Matsueda. Ross L.
1992
McEachern, Alexander W.
Menard, Scott and Barbara Morse
1968
1984
Palarma, Frances, Francis
T.
Cullen, and Joanne C. Gersten
1986
LABELING, LIFE CHANCES, AND ADULT CRIME
1317
Paternoster, Raymond and LeeAnn Iovanni
1989
The labeling perspective and delinquency: An elaboration of the theory and
assessment of the evidence. Justice Quarterly
6359-394.
An empirical test of labeling theory using longitudinal data. Journal of
Research in Crime and Delinquency
23:169-194.
Applied Factor Analysis. Evanston, 111.: Northwestern University Press.
SAWSTAT Software: Changes and Enhancements, Release
8.2.
Cary, N.C.:
SAS Institute Inc.
Crime in the Making: Pathways and Turning Points Through Life. Cam-
bridge, Mass.: Harvard University Press.
A life-course theory of cumulative disadvantage and the stability of delin-
quency. In Terence P. Thornberry (ed.), Developmental Theories of Crime
and Delinquency. New Brunswick, N.J.: Transaction.
Analysis
of
Incomplete Multivariate Data. New York: Chapman and Hall.
Two studies of legal stigma. Social Problems
10133-143.
Crime, punishment, and stake in conformity: Legal and informal control of
domestic violence. American Sociological Review
57680490.
Formal processing and future delinquency: Deviance amplification as selec-
tion artifact. Law and Society Review
24:1109-1131.
Asymptotic confidence intervals for indirect effects in structural equation
models. Sociological Methodology
13:290-312.
Getting Paid: Youth Crime and Work in the Inner City. Ithaca, N.Y.: Cor-
nell University Press.
What happened to yesterday’s rebels? Longitudinal effects of youth delin-
quency on education and employment. Social Problems
46250-274.
The effect of formal and informal sanctions on delinquency:
A
longitudinal
comparison of labeling and deterrence theories. The Journal of Criminal
Law
&
Criminology
75:1222-1245.
Labeling and crime: An empirical evaluation. In Walter Gove (ed.), The
Labelling
of
Deviance: Evaluating a Perspective.
2d
ed. New York: Wiley.
How unregulated is the US. labor market? The penal system as a labor
market institution. American Journal
of
Sociology
104:1030-1060.
Ray, Melvin C. and William Downs
1986
Rummel, Rudolph J.
SAS
Institute Inc.
1970
2001
Sampson, Robert J. and John H. Laub
1993
1997
Schafer, Joseph L.
Schwartz, Richard D. and Jerome H. Skolnick
Sherman, Lawrence W. and Douglas A. Smith
1997
1962
1992
Smith, Douglas A. and Raymond Paternoster
1990
Sobel, Michael
E.
1982
Sullivan, Mercer
L.
1989
Tanner, Julian, Scott Davies, and Bill O’Grady
1999
Thomas, Charles W. and Donna M. Bishop
1984
Tittle, Charles
1980
Western, Bruce and Katherine Beckett
1999
1318
BERNBURGANDKROHN
White, Halbert
L.
1980
A
heteroskedasticity-consistent
covariance matrix estimator and a direct
test of heteroskedasticity. Econometrica
482317-838.
The
severity
of
official punishment
for
delinquency and change in interper-
sonal relations in Chinese society. Journal
of
Research in Crime and Delin-
quency
31:416-433.
Zhang, Lening and Steven
F.
Messner
1994
Jon Gunnar Bernburg teaches Sociology at the University
of
Iceland and
is
currently
a Fellow
of
the Icelandic Research Council. His current research uses both quantitative
and qualitative methods to examine labeling processes. Recent publications include
“Anomie, Crime and Social Change:
A
Theoretical Examination
of
Institutional-ho-
mie Theory”
(British Journal
of
Criminology,
2002)
and, with Thorolfur Thorlindsson,
“Routine Activities in Social Context: A Closer
Look
at the Role
of
Opportunity in
Deviant Behavior”
(Justice Quarterly,
2001).
Marvin
D.
Krohn is Professor
of
Sociology at the University at Albany with a joint
appointment in the School
of
Criminal Justice. In addition, he
is
a Co-principal Investi-
gator
of
the Rochester Youth Development Study. His research focuses on theoretical
explanations
of
juvenile delinquency and adolescent drug use with special attention to
social network and interactional theories.
He
has recently edited (with Terence P.
Thornberry)
Taking Stock
of
Delinquency:
An
Overview
of
Findings from Contempo-
rary Longitudinal Studies,
a
volume on longitudinal panel studies
of
delinquency, and is
a co-author (with Terence
P.
Thornberry, Alan
J.
Lizotte, Carolyn
A.
Smith, and
Kimberly Tobin)
of
Gangs and Delinquency in Developmental Perspective.
