Beruflich Dokumente
Kultur Dokumente
A research note
Chris Gibson
a,
*, John Wright
b
a
Department of Criminal Justice, University of Nebraska at Omaha, Durham Science Center 208, 60th and Dodge Street,
Omaha, NE 68182-0149, USA
b
Division of Criminal Justice, University of Cincinnati, 600 Dyer Hall ML 0389, Cincinnati, OH 45221-0389, USA
Abstract
Since the publication of Gottfredson and Hirschi [A General Theory of Crime. Stanford, CA: Stanford
Univ. Press, 1990], a large amount of research has shown a link between low self-control and delinquency.
Some research has revealed that low self-control has not been able to account for the strong effects of peer
delinquency on delinquency. Criminological literature has, until recently, neglected the interactional
relationship between low self-control and delinquent peers in predicting delinquency. This study used a
sample of employed high school seniors to assess the interaction between low self-control and coworker
delinquency on occupational delinquency. Regression analyses indicated that the interaction term was a strong
predictor of occupational delinquency, even after controlling for several established predictors of delinquency.
D 2001 Elsevier Science Ltd. All rights reserved.
Introduction
Since the publication of Gottfredson and Hir-
schis (1990) General Theory, a large body of
research has examined the effects of low self-
control on offending behaviors and deviant acts
(Arneklev, Grasmick, Tittle, & Bursik, 1993; Bur-
ton, Evans, Cullen, Olivares, & Dunaway, 1999;
Evans, Cullen, Burton, Dunaway, & Benson, 1997;
Gibbs & Giever, 1995; Keane, Maxim, & Teevan,
1993; Piquero & Tibbetts, 1996; Polakowski, 1994;
Wright & Cullen, 2000). These studies, overall,
have generated moderate support for the hypothesis
that low self-control is significantly related to
offending and analogous behaviors.
Tests of self-control theory reveal that low self-
control has indirect and direct effects on drunk driving
and intentions to drink and drive (Keane et al., 1993;
Piquero & Tibbetts, 1996), on self-reported juvenile
delinquency (Wood, Pfefferbaum, & Arneklev, 1993),
on adult criminal and analogous behaviors (Arneklev
et al., 1993; Burton et al., 1999; Grasmick, Tittle,
Bursik, & Arneklev, 1993), on negative social con-
sequences (i.e., quality of friendships, quality of family
relationships, attachment to church, delinquent peers)
(Evans et al., 1997), on excessive alcohol consumption
and class cutting (Gibbs & Giever, 1995), and on
courtship aggression (Sellers, 1999). Other studies
have also shown that the interaction between low
self-control and opportunity to offend has significant
effects on crime and delinquency (Burton, Cullen,
Evans, Alarid, & Dunaway, 1998; Grasmick et al.,
1993; LaGrange & Silverman, 1999).
While these studies are important and implicate
the role of self-control in a wide range of problem
0047-2352/01/$ see front matter D 2001 Elsevier Science Ltd. All rights reserved.
PII: S0047- 2352( 01) 00111- 8
* Corresponding author. Tel.: +1-402-554-3104; fax:
+1-402-554-2610.
E-mail address: gibsonuno@cs.com (C. Gibson).
Journal of Criminal Justice 29 (2001) 483492
behaviors and social outcomes, research is only now
beginning to understand the complexity of this
relationship. Individuals low in self-control likely
face a range of interpersonal relationships and sit-
uations where their personality differences interact
with the expectations and social boundaries estab-
lished by context of the interaction. The effects of
low self-control may interact with the social
setting in important ways, either by reducing the
potentially deleterious effects associated with low
self-control, or by exacerbating them. In a recent test
of low self-control theory, Evans et al. (1997) found
that delinquent peer influences remained a substan-
tial and significant predictor of criminal behavior
even after the effects of low self-control were
removed. Their findings led them to conclude that
self-control and social learning theory may be
related in complex, mutually reinforcing ways.
The tendency of persons with low self-control to
engage in criminal and analogous behaviors, state
Evans et al. (1997, p. 494), can be exacerbated, or
strengthened, by exposure to criminal associates and
criminal values.
Recognizing the potential interactive effects
between measures of individual differences and the
social setting, Wright and his colleagues (1998) have
recently investigated a variable effects model of
criminal behavior. They suggest that the impact of
explanatory variables, specifically social learning and
social bonding variables, are strongest for persons that
have a predisposition towards crime such as those with
low self-control. They argue that sociological corre-
lates of crime have smaller effects on individuals that
do not have individual criminal propensities. In a test
of their proposed model, Wright et al. (1998) found
that the learning variables (i.e., delinquent associates)
that exerted positive effects on crime did so most
strongly for individuals with criminal propensities.
Although few studies have investigated inter-
action effects of low self-control and social variables
predicting offending behaviors, there is reason to
believe that these effects will manifest across social
contexts such as the work environment. In a recent
study of occupational delinquency among high
school students, Wright and Cullen (2000) found that
occupational delinquency was affected both by
underlying criminal propensities and by exposure to
delinquent coworkers on the job. They found empir-
ical support for an interaction effect between prior
delinquency and delinquency of coworkers that, in
turn, amplified involvement in occupational delin-
quency. It is likely that delinquent youths select
themselves into poor work environments where they
come into contact with fellow delinquents, which
increases delinquent behavior within the workplace
(Wright & Cullen, 2000).
Occupational delinquency
Modern youths are sophisticated economic actors,
often with fairly extensive employment histories
established prior to graduating from high school
(Wright, Cullen, & Williams, 1997). Even though
working appears to be a common experience among
in-school youths, numerous studies have found that
certain individual differences differentiate youths who
work from those who work extensively (Bachman &
Schulenberg, 1993; Cullen, Williams, & Wright,
1997; Elliott & Wofford, 1991; Ruggiero, Green-
berger, & Steinberg, 1982; Ruhm, 1995; Steinberg
& Dornbusch, 1991; Wright et al., 1997). Individual
differences such as early school performance difficult-
ies, early drug-use, and early delinquent behavior
account for part of the correlation between the average
number of hours worked per week and a youths
misbehavior (Bachman & Schulenberg, 1993). Since
delinquent youth most likely self-select themselves
into premature work roles at a higher rate than con-
forming youth points to the possibility that the ado-
lescent workplace is an important domain of behavior
that mixes youths with varying levels of criminal
propensity. This possibility was examined by Rug-
giero et al. (1982), who suggested that occupational
deviance can be attributed to environmental aspects
of the job, specifically coworker occupational delin-
quency and personal characteristics of the individual
worker. Using a sample of high school students from
Orange County, California, Ruggiero and colleagues
found personal characteristics of workers and work
environment factors may often reinforce each other,
which may produce deviant occupational behavior.
The adolescent workplace, while understudied,
appears to be a potentially important context for
youth (mis)behavior. While all youths do not par-
ticipate in work-related deviance, some are more
likely than others, some jobs are more likely to
provide opportunities for delinquency; and some
characteristics of persons and jobs, coupled together,
generate more occupational deviance than either
alone (Greenberger & Steinberger, 1980). Youth
employment thus draws attention to the possible
interactive effects between low self-control and var-
iables from other criminological theories.
Current study
This study attempted to build on Wright and
Cullens (2000) investigation. The present study
extended their research by assessing the interaction
effect of low self-control and coworker delinquency in
predicting occupational delinquency. This study
addressed shortcomings in the research that has been
conducted on the general theory in two different
C. Gibson, J. Wright / Journal of Criminal Justice 29 (2001) 483492 484
ways. First, given Gottfredson and Hirschis (1990)
claim that their theory predicts crime across social
contexts, low self-control should exert a significant,
independent effect on occupational delinquency
(Wright & Cullen, 2000). Second, this study also
assesses the interaction between low self-control and
delinquent coworkers on levels of occupational delin-
quency. Several scholars suggest that individuals with
low self-control and who work in an environment with
delinquent coworkers should be significantly more
likely to be involved in occupational delinquency,
even after controlling for competing theoretical vari-
ables such as family cohesiveness or school commit-
ment (Greenberger & Steinberger, 1980; Wright &
Cullen, 2000).
Methods
Sample
Data for the current study came from the Tri-
Cities Adolescent Employment Survey (TCAEP).
The TCAEP is a cross-sectional self-report survey
that was conducted in eight high schools located in
northeast Tennessee. The inclusion of high schools
was dependent upon whether school officials
allowed permission for their students to participate
in the survey during the school day. Due to restric-
tions placed on the investigators by the high
schools, only individuals who had reached their
senior year in high school were allowed to be
surveyed (N= 436); therefore, surveys were admin-
istered in homerooms or in classes required of all
seniors. All seniors in attendance on the day the
survey was administered were allotted a one-hour
time span to complete the survey. The survey was
completely voluntary and confidential.
High schools included in the sample encom-
passed a wide range of students of varying social
and economic backgrounds. Youths in the sample
reported an average household size of four, 65
percent reported living with both parents, 89 percent
of the sample was White, and 47 percent were male.
Given that the current study specifically investigated
occupational delinquency, a subsample of respond-
ents was extracted of only those who had been
employed during high school (n = 296). Univariate
statistics indicated that respondents in the current
study worked at jobs in a similar proportion as to
seniors nationally (see Steinberg & Caufman, 1994).
Given that the sample was of convenience (or
nonrandom) and was restricted to a specific geo-
graphical region, appropriate caution should be used
in generalizing the results to the population of
adolescents at-large.
Dependent variable
Occupational delinquency
Involvement in occupational delinquency was
measured by a nine-item scale, developed and used
by Ruggiero et al. (1982), that assessed the degree of
participation in a range of misbehaviors while at
work, such as taking things from an employer or
employee, giving away goods or services without
permission, or lying to the employer to get or keep a
job (see Appendix A). Responses for each item
ranged on a three-point scale from zero (never)
through two (often). To compute a general occu-
pational delinquency scale, responses were summed
across items where higher scores indicated increased
occupational delinquent involvement. An estimated
reliability coefficient (a=.79) gave support to the
internal consistency of the occupational delinquency
scale (see Table 1 for descriptive statistics).
1
Independent variables
Low self-control
Self-control was measured using a twelve-item
scale that was a modified version of that developed
by Grasmick et al. (1993) (see Appendix B). Several
studies have revealed that this scale is a psychometri-
cally appropriate measure of self-control (Burton et
al., 1999; Piquero & Tibbetts, 1996). Responses to
items ranged on a four-point Likert-type scale from
one (disagree strongly) to four (agree strongly).
Higher scores on this scale indicated lower self-
control. An estimated reliability coefficient (a=.75)
revealed that the measure of self-control was intern-
ally consistent. Furthermore, similar to findings gen-
erated from other studies (Burton et al., 1999; Piquero
& Tibbetts, 1996), principle components factor ana-
Table 1
Descriptive statistics for all variables (n =296)
Variable Mean S.D.
Occupational delinquency 1.52 2.23
Low self-control 26.59 5.00
Coworker delinquency 3.10 3.23
Low self-control
Coworker delinquency
3.05 26.72
Time spent studying 4.81 4.82
Grade point average 2.86 0.76
Goals/aspirations 10.22 1.58
Family cohesiveness 78.57 15.42
Gender 0.55 0.50
Race 0.01 0.27
Family structure 0.32 0.47
Household size 3.81 1.15
C. Gibson, J. Wright / Journal of Criminal Justice 29 (2001) 483492 485
lysis indicated that the self-control measure formed a
unidimensional construct.
Coworker delinquency
Similar to the occupational delinquency scale,
occupational delinquency of coworkers was meas-
ured by a nine-item scale that asked respondents the
extent to which their coworkers were involved in
misbehavior at work. It should be noted that while the
coworker delinquency scale has been used in pre-
vious research, second hand accounts of others
misbehavior in the workplace suffer from many of the
same validity problems found in traditional measures
of delinquent peers. The scale, which was composed
of item responses on a three-point scale ranging from
zero (never) through two (often), had an acceptable
level of reliability (a=.81) and statistical validity.
Future research should, undoubtedly, use better, more
sophisticated measures of coworker misbehavior.
School commitment
School commitment was assessed by two separate
measures: (a) time spent studying and (b) grade point
average. Time spent studying was measured by a
two-item scale that asked respondents the number of
hours, on average, they spent studying over the
weekdays and weekends, which had a reliability
coefficient of .66. Grade point average was measured
using a single-item that asked respondents about the
average grade they received in their classes. Item
responses ranged from one (Mostly Fs) to five
(Mostly As).
Goals and aspirations
A three-item scale was used to measure the
perceived importance of current and speculative
events in the youths lives. Respondents were asked
how important it was to them to be able to rely on
their parents, go to college, and do well in school.
Item responses ranged on a four-point Likert scale
from one (disagree strongly) to four (agree strongly).
Responses were summed across items to create a
scale, which had a reliability coefficient of .66.
Higher scores indicated increased importance of
goals and aspirations.
Family cohesiveness
The twenty-four-item family cohesiveness scale,
derived from the National Youth Survey, was
employed to measure the perceived relationships
youth had with their mothers and fathers. The scale
was composed of four dimensions that included
parental reliability, parental supervision, parental con-
flict, and parental communication (see Appendix C).
Responses to items ranged on a five-point Likert type
scale from one (never) to five (always). To compute
the family cohesiveness scale, responses were
summed across items where higher scores indicated
stronger family cohesiveness. An estimated reliability
coefficient (a=.87) generated support for the internal
consistency of the scale.
Demographics
Four demographic characteristics were included as
controls in our study: (a) gender, (b) race, (c) family
structure, and (d) household size. Gender was coded
as zero (male) or one (female). Due to limited
frequencies in other categories, race was dichotom-
ized and coded as zero (White) or one (other),
whereas, minorities accounted for 7 percent of the
sample. Family structure was a dichotomized meas-
ure that indicated whether or not respondents resided
with both parents; therefore, family structure was
coded as zero (intact) or one (nonintact). Finally,
household size captured the number of people living
in the residence of the youth. Due to the large number
of missing cases, which would have decreased the
sample size, family income was excluded from the
analysis. Subsequent regression analyses that were
not reported indicated that family income, once
entered into the model, did not have a significant
effect on occupational delinquency.
Interaction term: Low self-control Coworker
delinquency
The product term between low self-control and
occupational delinquency of coworkers was of specific
importance to this study; therefore, it was important to
explain how the interaction was created. The compon-
ent factors of the interaction were mean centered prior
to multiplication in order to minimize the nonessential
ill-conditioning that would most likely cause multi-
collinearity in a product term (Aiken & West, 1991).
When mean centering a variable, it is necessary to
subtract the mean score of the variable from the actual
variable before multiplying the two component terms.
This procedure typically reduces collinearity concerns
and reveals low correlations between the interaction
term and the independent component parts of the
interaction term (Cronbach, 1987).
Research strategy
The analysis presented in this article investigated
whether individuals with low self-control and who
work in environments with delinquent coworkers
were more likely to be involved in occupational
delinquency. First, bivariate correlations were exam-
ined to investigate multicollinearity among the prod-
uct term and its independent components, and to also
C. Gibson, J. Wright / Journal of Criminal Justice 29 (2001) 483492 486
investigate whether there was multicollinearity
among other independent variables. Second, ordinary
least squares (OLS) regression was used to examine
the effect of the interaction term on occupational
delinquency in three models: (1) controlling for the
independent components of the interaction term, (2)
controlling for demographic factors, and (3) control-
ling for other competing theoretical variables.
Results
Table 2 shows the zero-order correlations. A two-
tailed test of significance at the .05 level was the
criteria for all analyses. The bivariate assessment of
variables, variance inflation factors, and condition
number tests revealed that there were no signs of
multicollinearity among the interaction term and the
independent components of the interaction term.
Additionally, it seems that there were no signs of
collinearity among other independent variables. In
general, the bivariate relationships were significant
and in the predicted direction.
Table 3 shows three separate OLS models predict-
ing occupational delinquency. Model one shows that
coworker delinquency exerted a positive and signific-
ant effect on occupational delinquency (b=.47), indic-
ating that youth in an environment with more
delinquent coworkers were significantly more likely
to be involved in occupational delinquency. Similarly,
low self-control generated a positive and significant
effect on occupational delinquency (b=.18), indic-
ating that individuals with low self-control were
significantly more likely to be involved in occu-
pational delinquency. The interaction effect between
low self-control and coworker delinquency, as sug-
gested by Evans et al. (1997), and hypothesized by
Wright, Moffitt, and Caspi (1998), had a positive and
significant effect on occupational delinquency. The
model accounted for 39 percent of the variance in
occupational delinquency.
To illustrate the effect of the interaction, a bar graph
was created showing different mean levels of occu-
pational delinquency (see Fig. 1). Following Piquero
and Tibbetts (1999) and Raine, Brennan, and Mednick
(1994), a four-category variable was created that
indexed the presence of low self-control and coworker
delinquency. The measures of low self-control and
coworker delinquency were dichotomized based on
their mean scores. Respectively, scores below the
mean indicate higher self-control and minimal cow-
orker involvement in delinquency, while scores
above the mean indicate lower self-control and more
coworker involvement in delinquency. Next, a four-
category variable was created and was coded
according to the following classifications: zero if
youth had high self-control and low number of co-
workers who were delinquent (n =103), one if youths
had low self-control and a low number of coworkers
who were delinquent (n = 73), two if youths had
higher self-control and a high number of coworkers
who were delinquent (n = 60), and three if youths
had low self-control and worked in an environment
with high coworker delinquency (n = 110). The final
category, indicating low self-control and higher
coworker involvement in delinquency, is where the
highest risk for occupational delinquency should be
found (Evans et al., 1997; Wright et al., 1998;
Wright & Cullen, 2000).
The results are depicted in Fig. 1. The y-axis
represents the mean level of occupational delin-
quency, while the x-axis indicates the categories of
the interaction. As can be seen, those youth in the last
category had the highest mean level of occupational
Table 2
Intercorrelations between all variables (n = 296)
Variable X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11
X1. Low self-control
X2. Coworker delinquency .32*
X3. Low self-control
Coworker delinquency
.35* .05
X4. Time spent studying .25* .05 .02
X5. Grade point average .18* .07 .07 .22 *
X6. Goals/aspirations .34* .08 .07 .30 * .23*
X7. Family cohesiveness .32* .16 * .16* .01 .08 .32*
X8. Gender .13* .00 .04 .28 * .11 .26* .04
X9. Race .04 .03 .08 .05 .12* .04 .03 .02
X10. Family structure .03 .11 .05 .01 .06 .09 .37* .08 .11
X11. Household size .07 .01 .05 .05 .04 .07 .17* .01 .04 .27*
* P< .05.
C. Gibson, J. Wright / Journal of Criminal Justice 29 (2001) 483492 487
delinquency (mean = 2.82), indicating that youth with
low self-control and who worked in an environment
where coworkers were more delinquent had a sub-
stantially higher mean score on occupational delin-
quent involvement.
Turning now to Model 2 of Table 3, demographic
controls were included to investigate whether or not
the interaction would retain its significance. The
interaction between low self-control and coworker
delinquency remains the second strongest predictor
of occupational delinquency (b =.23). Although the
independent effect of coworker delinquency had the
strongest effect on occupational delinquency
(b =.49), the independent effect of low self-control
exerted a positive and significant effect on occu-
pational delinquency (b =.13), indicating that youth
with low self-control were significantly more likely
to be involved in occupational delinquency. Gender
(b = .19) and race (b =.09) both had significant
effects on occupational delinquency. Model 2
accounted for 44 percent of the variance in occu-
pational delinquency.
Finally, in Model 3 of Table 3, controls for
competing theoretical variables that have been estab-
lished as known correlates of delinquency were
introduced, such as school commitment and family
cohesiveness. Coworker delinquency still remained
as having the strongest overall effect on occupational
delinquency (b =.54). The interaction between low
self-control and coworker delinquency, after control-
Table 3
Estimated standardized regression coefficients predicting occupational delinquency
Model 1 (n = 346) Model 2 (n = 333) Model 3 (n = 296)
Variable b t b t b t
Low
self-control
.18 3.96* .13 2.71* .09 1.79
Coworker
delinquency
.47 10.87* .49 11.45* .54 12.09*
Low
self-control
Coworker
delinquency
.21 4.63* .23 5.18* .21 4.45*
Time spent
studying
.06 1.29
Grade point
average
.15 3.18*
Goals/
aspirations
.02 0.39
Family
cohesiveness
.03 0.69
Gender .19 4.40* .16 3.45*
Race .09 2.25* .10 2.23*
Family
structure
.04 0.82 .03 0.69
Household
size
.01 0.27 .01 0.28
Constant 3.139* 0.85 0.18
R
2
.39 .44 .48
F 71.35 36.33 23.16
df 345 332 295
Dashes indicate that parameters were not estimated.
* P< .05.