NALL Working Paper
#05-1999
Whatever Happened to Yesterday’s Rebels?
Longitudinal Effects of Youth Delinquency on Education and Employment
Julian Tanner
University of Toronto at Scarborough
Scott Davies
McMaster University
Bill O’Grady
University of Guelph
Abstract
This paper examines whether and how teen delinquency
is consequential for a variety of educational and employment outcomes. From the
National Longitudinal Survey of Youth we measure five forms of delinquency from
1979 when respondents were 14-17 years old, and investigate whether they predict
five different outcomes when those individuals were aged 25-30. We measure
delinquency as the prevalence of skipping school, drug use, violent behavior,
engaging in property crime, and contact with the criminal justice system. Using
a variety of regression models, we explore whether delinquency has negative
zero-order effects, and negative partial effects net of standard status
attainment variables. We find that all types of delinquency have consistently
significant and negative impacts on educational attainment among both males and
females, net of status attainment variables. Delinquency has also a fairly
consistent impact on male occupational outcomes, but has weaker effects on
female occupational outcomes. Overall, the data suggests that delinquency
has autonomous and negative effects on later life chances. We discuss
these findings in light of links between Status Attainment models and theories
of crime and delinquency.
What happens in young adulthood to those who were
rebellious, delinquent, or resistant during their youth? Does teenage
delinquency tend to spoil adult outcomes, or alternatively, is it a mere passing
phase with few long term consequences? And what do these outcomes tell us about
the meaning of youthful deviance? In this paper we pursue empirically these
questions by focusing upon educational and occupational attainments, and explore
whether there are direct or indirect links between juvenile misconduct,
educational success, and adult occupational attainment.
Status Attainment and Delinquency: Good and Bad Capital Investments
To explain school and occupational achievement,
status attainment models originally employed a host of individual background,
social psychological, and educational variables (see Sewell and Hauser, 1980).
More recent research has shown how, in addition, disciplined work habits,
cultural capital, social capital and parental involvement in their children’s
education all aid attainment net of social background and measured school
ability (e.g. Farkas et al.,1990; Schneider and Coleman, 1993; Aschaffenburg and
Maas, 1997).
The metaphor of “capital” speaks to a process
whereby various forms of cultural and social activity in one’s early years
later pay off by bettering one’s life chances, and helping to smooth the
transition from school to the labor market. Conversely, other forms of
cultural activity do the opposite: they can depreciate such investments.
Delinquency is one such potential source of capital disinvestment.
Life-style choices and indiscretions that range from tardiness and indiscipline,
to delinquency or criminality, we argue, can harm attainments and create
cumulative, long term disadvantage. Moreover, they can have this effect
independent of one’s resources, economic or cultural. A recent study by
Alexander, Entwisle and Horsey (1997), for instance, found that poor
attitudes and undisciplined behavior among children in the first grade had a
strong impact on eventual school success, net of family background and measures
of academic standing and ability (ie. grades and test scores).
Such a finding can be construed as offering a mirror-image
of what Status Attainment researchers have discovered. A key contribution
of American applications of Bourdieu’s (1984) ideas about the import of
cultural capital is to show that its effects upon achievement cannot be reduced
to class origins, as measured by parental socioeconomic background (DiMaggio,
1982; Aschaffenburg and Maas, 1997). Similarly, the future capital
disinvestments associated with delinquency may also be independent of the
present economic and cultural endowments of one’s parents. We aim to test if
delinquency has a long and deleterious reach. Our premise is that the scope of
educational and occupational attainment models can be enhanced by looking at the
barriers in adolescence to attainments in adulthood. Research on social
stratification and crime and deviance has been, until recently, relatively
uncoupled (Hagan, 1991; but see also Crutchfield and Pitchford, 1997).
Typically, Status Attainment research focuses upon the conformist endeavors of
adults, whereas students of delinquency concentrate on the miscreant
activities of youth. Our task in this paper is to help further bridge these two
research worlds by incorporating measures of teenage delinquency into attainment
models.
Sociological Images of the Long Term Impact of Delinquency:
Classic sociological images of the long term impact of
teenage delinquency differ sharply. On the one hand, Subcultural
theorists, beginning with Albert Cohen (1955), anticipate that deviance will
eventually deliver rebellious (male) youth to the rough-end of the labor market.
Paul Willis (1977) and Jay Macleod (1995) are the principle contemporary
suppliers of this image of clear and predictable pathways from resistance in
school to working-class jobs. Working-class boys get working-class jobs - the
all-important sub-title of Learning to Labour - because their celebratory
rejection of middle class schooling, their repertoire of minor delinquency, and
their masculinist bravado, all disqualify them for anything other than low-skill
manual occupations. It is a fate, however, the “Lads” welcome as
confirmation of their working class status and identity. MacLeod’s
ethnography, Ain’t No Making It, travels similar terrain in an American
context.
On the other hand, there is a tradition of
sociological criminology which downplays the significance of teenage deviance,
and counters what it sees as an over-deterministic portrayal of delinquency. The
principal architect of this view, David Matza (1961), confronted
Subcultural theory by emphasizing the essential similarities between most
deviant behavior and conformity. By doing so, Matza downgraded the
long-term significance of milder forms of teenage rule breaking. Indeed, he took
his argument further and proposed that mainstream teenage culture functions as a
safety valve for most pleasure-seeking adolescents by insulating them from more
serious, and consequential, delinquency. Further, he accused the original
Subculturalists of, variously, explaining more delinquency that actually occurs,
exaggerating the class basis of delinquency, and of offering few clues as to why
rebellion fades away as adulthood approaches. Willis and the newer
“Resistance” school have been similarly accused of this type of
over-determinism (Tanner, 1996; Davies, 1995)
These diverging views of the severity of teenage
deviance highlight the need to unravel the long-term effects of delinquency with
systematic research. Influential as Willis, Macleod and Matza have become, they
derive their alluring images from very small, unrepresentative samples, and in
the case of Willis, without benefit of a follow-up study to verify the original
argument.1 Whether or not high school delinquents fare poorly in
school, and are ultimately relegated to low-status jobs or unemployment is an
open empirical proposition that can only be tested with longitudinal data.
Further, there is the issue of gender. Early studies
all but ignored female deviance, or treated it as a sexual, rather than a
law and order, problem (Murdock, 1982; Frith, 1985). Accordingly, the classic
images of rebellious or resistant youth rarely includes females (ie. those
offered by Cohen, Matza, Willis and Macleod). An overweening concern with
male rebellion and its consequences has forced the study of female deviance and
its impact upon jobs to the sidelines. Our study compensates for this by
including an examination of the effects of delinquency on jobs and education for
both males and females. It thus stands apart from other studies, including
some of the most prominent longitudinal ones (described further below), that
concentrate exclusively upon male behavior and experiences (Sampson and Laub,
1993; Hagan, 1993; Freedman, 1992).
However, what might be expected from a separate
analysis of males and female trajectories is a relatively open question. On the
one hand, it can be argued that females will be less seriously damaged by
earlier delinquent experiences than males. The recent, increasingly
gender-sensitive, literature tells us that most of the important differences
between female and male offenders relate to incidence and prevalence.
Adolescent girls are less frequent offenders than boys; they more to likely
desist from delinquent activity at an earlier age (Chesney-Lynn and Sheldon,
1992), and they are more likely to be deterred from further wrong-doing as a
result of their initial contacts with police (Keane, Gillis, and Hagan, 1989).
Similarly, female labor force participation rates have been traditionally lower
than those of males - a reflection of the fact that female destinations are less
intimately bound to employment outside the home than is the case for males. If
delinquency is thus a less regular occurrence among adolescent females, and if
paid employment outside the home has been until recently less central to the
adult goals of young women, then it would be unreasonable to expect their lives
to be as strongly governed by these activities. Furthermore, since women -
no matter how well-endowed with human, cultural or social capital - enter a
narrower range of jobs than men - it is correspondingly less likely that the
cultural deficit of delinquency will leave an equivalent indentation upon female
careers. As a result, we might expect female life-course trajectories to be less
significantly marred by delinquency than those of males.
On the other hand, a case can also be made for the
opposite outcome: that even though girls are less immersed with delinquency than
boys, and abandon it more quickly, it may still be the case that the minority
that do become involved suffer more in later life. Generally less inured
to the idea of female deviance, society is particularly concerned with the
control of female sexuality. Adolescent girls who do break to appropriate
conduct norms may therefore anticipate problematic repercussions across a
variety of adult domains, including work and employment.
Historical evidence certainly suggests that the
police and courts have shown a greater interest in regulating the moral behavior
of girls than boys. It is for such status offences as incorrigibility,
truancy and moral deprivation that girls, rather more frequently than
boys, have been referred to the juvenile court. Similarly, girls have
often received harsher punishment, including institutional sentences, for
violating moral codes than the criminal code (Reiss, 1960). And while girls may
benefit more readily than boys from police discretion when criminal acts are
involved, they have also been more frequently for infringements of status
offences which often connote gender-specific undesirable sexual activity
(Monahan, 1970). Overall, therefore, it seems that boys are punished more
for illegal acts, whereas girls are more vulnerable to judicial concerns about
moral propriety.
Longitudinal Research on Delinquency
Sociological research on education clearly shows
that delinquency has negative and reciprocal associations with school grades and
retention (Stinchcombe, 1964; Polk and Schafer, 1972; Myers, Baker, Milne, and
Ginsburg, 1987; Mensch and Kandel, 1988; Davies, 1995). In turn, volumes of
status attainment research attest to the overwhelming significance of
educational success as a determinant of occupational success (Sewell and Hauser,
1980). Given this research, one might expect delinquents to eventually do
less well in school compared to their more conformist peers, and to subsequently
fare worse in the labor market. Schooling, in this view, would merely reinforce
the corrosive effects initiated by delinquency.
But what about the effects of delinquency over the
long haul? Criminology clearly shows that while most adult criminals have
delinquent pasts, few teenage delinquents become adult criminals (Farrington,
1992; Rutter and Smith, 1995; Sampson and Laub, 1993). If few deviants
grow into hard core criminals, youth deviancy may exert little negative impact
on eventual status attainment. The fact that most adolescents in the general
population grow out of delinquency suggests that it is not always strongly
damaging in the long-term, and may have only a negligible effect upon
occupational attainment.
Existing research on this issue, however, is
incomplete, and does not offer a clear answer, since its findings tend to depend
on the nature of the measures and samples used. For instance, when using
measures of non-criminal or less serious forms of delinquency, such as deviant
leisure or “party subcultures,” some research supports Matza’s image of
inconsequential effects of teen deviance, or shows that the effects of this
deviance are contingent upon other factors. Bynner and Ashford (1992),
when examining teenage lifestyles in a longitudinal study of young people in
Britain, found little evidence to suggest that subcultural activities predicted
pathways into the labor market, and concluded that for the most part, adolescent
leisure has an ephemeral quality. Hagan (1991) found that Toronto-area middle
class males actually benefitted from engaging in a leisure-oriented,
non-delinquent ‘party’ subculture.
Others focus on various contingencies. Monk-Turner (1989) argues that
experiences of school success may break the connection between youth crime and
subsequent occupational under-achievement, claiming that if delinquents do well
in school, occupational future life-chances go un-threatened. Jessor et
al’s (1993) research on adolescents in the American mid-west found a
continuity between adolescence and adulthood - the more problematic an
individual’s behavior during the former, the more problematic an
individual’s behavior throughout the latter - but also found that youthful
deviance had no significant effect upon occupational attainment. They hold that
the consequences of deviant behavior depend in part on one’s social class
background, since middle-class youngsters in well-heeled communities have access
to resources - economic, social, cultural - that provide them with a safety-net
when they ‘get into trouble’, and which protect them from suffering
long-term consequences for their youthful indiscretions. Hagan (1991)
found that the consequences of participation in various youth subcultures
differed sharply by their class background.
Other studies, however, find broader and independent
consequences of less savory forms of youthful deviance. Analyzing data
originally collected by Sheldon and Eleanor Glueck (1950), Laub and Sampson
(1994) found that childhood delinquents, compared to non-deviants, held weak
educational, economic and professional aspirations, were more likely to be on
welfare in both early and later adulthood, and were more likely to have an adult
history of unstable employment. Hagan (1993), using a British sample of
poor and working class Londoners, found higher rates of unemployment among
former delinquents. He proposed that having delinquent friends and
engaging in continuing delinquency leads to adult unemployment through the
formation of a distinct social network that isolates individuals from legitimate
employment opportunities.
Yet, the Gleuck’s samples used by Sampson and Laub,
as well as Hagan’s London sample, are either of a clinical variety that
consist of serious deviants, or are collected from relatively poor
neighborhoods. Both types are overpopulated by those official delinquents
who are most likely to be marred in later life by their adolescent
indiscretions.2 Further, these samples are limited to small single
towns or cities (ie. The Boston area for Sampson and Laub, a Toronto suburb for
Hagan [1991], or a 1-mile radius of working class London (Hagan, 1993).
Perhaps more importantly, many of these studies lack key control variables. In
contrast, many new data samples are large and nationally representative, and
offer the types of control variables needed to truly test for the independent
effects of delinquency. For instance, none of the data sets reviewed above
could objectively control for cognitive skill (as measured on a standardized
test). This is key, since many delinquents’ later school and labor
market attainments may be harmed less by their deviant actions per se than by
their lack of skills. This point has been made most forcefully in the work of
Farkas (1996).
Finally, the above-cited studies point to the
importance of different measures of delinquency. Many offer either “mild” or
non-criminal forms of delinquency, or only a single measure of “harder”
delinquency. Since different forms of delinquency may have different impacts on
life chances (a point discussed further below), studies that employ multiple
measures of delinquency could help to better specify the links between deviance
and resulting life chances.
Research Questions
We wish explore the theme of “whatever happened to
yesterdays rebels” by examining whether and how teen delinquency, in terms of
its total effects or its partial effects net of relevant variables, is
consequential for a variety of adult educational and employment outcomes. The
literature reviewed above, taken as a whole, suggests that the long term impact
of adolescent deviance on various attainments still remains unclear. No one as
yet has established whether teenage delinquency, measured in both
“softer” and “harder” forms, has a clear causal link to adult
attainments among a national, representative sample. The primary issue is what
is the typical outcome, across the life course, caused by different forms of
delinquency, independent of standard attainment variables, among a
representative population. In this paper we are less concerned with the
contingencies of these effects - ie, whether all groups of people are hurt in
the same way or to the same degree - but whether or not most delinquents are
hurt in some way. We explore bivariate models to inspect the total effects
of delinquency on attainments, then explore multivariate models to inspect
independent effects, and employ a series of interaction terms with SES and crime
in order to explore some of these possible contingencies.
Our investigation can be placed also within the
broader context of an ongoing debate within the criminological community about
the relative import of personality characteristics and life events as influences
upon adult attainments. On the one hand, there are those such as
Goffriedson and Hirschi who claim that the trait of self-control - acquired
early in life and relatively invariant in its effects thereafter - is a strong
determinant of both educational and occupational success and involvement in
crime and delinquency. On the other hand, there is Sampson and Laub’s
life course argument which proposes that life events have an impact upon adult
destinies that is not reducible to the motivational characteristics of
individuals exposed to those life-turning events. The present paper is not a
test of these different perspectives: we have no measure of Gottfredson and
Hirschi’s key explanatory variable, self control; and, more importantly,
we do not share the axiomatic assumption that there are long term consequences
of adolescent deviance. Indeed, our paper, in the first instance, is about
establishing whether or not in a large representative sample (as opposed to
smaller, less representative samples), there is a negative dividend for a
misspent youth. Nevertheless, it is possible that the results of our
inquiry will speak to the issues raised by these criminological protagonists.
To address our issues, we have made series of key
decisions concerning our sample and measures. First, we have elected to
test our hypotheses using a nationally representative sample of 14-17 year olds
who were enroled in school in 1979. Our logic is to begin with a sample of
students still in school, and then from that starting position trace the
subsequent impact of their delinquency. This decision, of course, excludes any
youth in that age range who had dropped out of school before that year. We
have done so intentionally in order to focus on mainstream or “ordinary
youth” (Brown, 1987), whether they be deviant and conformist. Relatively few
American youth leave school at earlier ages, and those who do so are likely to
be harder-core delinquents. This is important, since most of the
contemporary influential theorizing about school resistance is based, lest we
forget, on the activities of youth who were still in school at the time of
each study. Willis’ “lads” and Macleod’s “Hallway Hangers” and
“Brothers” were all full-time students, albeit reluctant ones, at their
neighborhood public schools. Thus we are more interested in tracking the
status outcomes of a broad swath of adolescent delinquents - about whom we know
comparatively little - as opposed to documenting the under-performance of
hard-core dropouts and street youth who are most likely to exhibit deviant and
disadvantaged continuities over time (for such studies, see Hagan and McCarthy,
1992; McCarthy and Hagan, 1992; Tanner et al, 1995). By excluding the
relatively anomalous, we can test the impact of delinquency upon
relatively “typical” or ordinary high schoolers - the type of population
examined by Status Attainment researchers. This also makes our statistical tests
tougher and more conservative by eliminating the few most likely candidates for
having long term effects of delinquency. (We ran the same models for the full
sample; the results were entirely consistent with our broad conclusions based on
the student sample.)
Second, model specification is crucial, since the
effects of delinquency may be spurious if other key explanatory variables are
not taken into account. To guard against this, our models include most
predictors of schooling and job attainment used in standard school attainment
and job attainment models. Unlike most studies in this area, we are able to
control for background, school-related abilities, and cultural resources, and
thus test with rigor for the existence of independent effects of delinquency.
This allows us to see whether the adult legacy of youth deviance is a direct
one, or is mediated through schooling. We also provide wide range of
outcomes with multiple measures of educational and occupational outcomes in
order to better capture the multi-dimensional nature of life chances.
Third, we have elected to examine a variety of
separate, independent measures of delinquency, rather than a single, unitary
scale of delinquency. (For statistical purposes we did examine models that
included a single omnibus measure of delinquency, as is discussed in the
Findings section.) While some criminologists hold that the effects of
delinquency, in terms of sanction, stigma, and subsequent defiance (see Sherman,
1993) are more or less constant, it is unclear whether different forms of
delinquency have different causes and consequences (c.f. Osgood, Johnston,
O’Malley and Bachman, 1988). Hagan (1991), for instance, reported that
different types of subcultural preferences had different consequences on
subsequent life achievements. We thus believe it important to test whether or
not all behaviors indulged by adolescents are equally harmful. Everyday
experience, and the prescriptions of the criminal code, certainly suggest
otherwise. Shoplifting from a corner store does not carry the same long-term
consequences as wounding someone with a weapon. Rather than assume that all
miscreant action produces the same effect, our concern is for seeing which
particular forms of delinquent behavior are consequent for various life chances.
Further, as described in the next section, our factor analyses were able to
clearly separate different types of delinquency. We therefore examine five
different types of delinquency, representing a range of severity, in order to
test for their differential impact. Finally, to explore whether the connections
between teen deviance and adult outcomes differ for males and females, we
provide separate models for both men and women.
Method
We use the representative sample (n=6,111) from the
National Longitudinal Survey of Youth (NLSY79). This data set consists of an
annual follow-up of youth aged 14-22 when initially interviewed in 1979. The
NLSY. enjoys an excellent retention rate. By 1993, over 90% of original
representative sample remained. To investigate the enduring impact of teen
delinquency, we selected a subset of individuals: those still enrolled in
school and between the ages of 14-17 in 1979, and examined the educational and
employment outcomes for these individuals in 1990-1992, when they were 25-30
years of age. We exclude youth who had already dropped out of school for the
reasons mentioned in the previous section, and because their inclusion would
make unclear the causal connection between delinquency and some outcomes, since
these youth’s educational attainment would be largely determined from the
outset.
This subset consists of 1,452 males and 1,397 females.
Sample attrition between 1979 and 1992 and list-wise deletion of missing data
reduces this sample to 1,145 males and 1,112 females for most models, yielding
missing case rates of 21% and 20% respectively. These rates are very good for
longitudinal data, and indeed they are better than virtually all other studies
in this area of study. Most of our missing cases stem from missing data on a few
select variables, especially father’s occupation (used to calculate SES
background), though some are due to sample attrition in later years.
Samples are somewhat smaller for our models of occupational status, since many
individuals, particularly females, were not in the labor force in 1992 (n=1,029
for males; 917 for females). Since it is plausible that long-term
delinquents may be over-represented among our missing cases, our tests are again
conservative by being possibly biased against finding those effects that would
support our thesis.
Measures
Our models include variables over a twelve-year span to
afford a glimpse into the long-term impact of teen delinquency. Our main
explanatory variables are five forms of teen delinquency. We constructed
delinquency scales from the NLSY’s retrospective questions pertaining to crime
and deviance. In 1980, interviewees were asked to report their deviant
activities over the previous year (the NLSY unfortunately lacks similar measures
of crime and deviancy in later years). Using factor analysis and alpha
reliability tests (tables available upon request), we created two single item
variables and three scales by summing standardized scores of related survey
items.3 Males and females had almost identical factor loadings, so we
used the same scales for both males and females. Table 1 presents a summary of
all variables.
The first variable is “skipping school”,
represented by a single question about the number of times over the previous
year the respondent skipped school. Both common sense and previous research
suggests to us that while skipping school ultimately depreciates achievements in
the labor market, its effects will be most directly evident in educational
attainments. While this item loaded with a series of questions
regarding drug use, we decided to separate them in order to preserve their
distinct theoretical meanings. The second variable is “drugs”, which
includes the number of times over the previous year the respondent drank
alcoholic beverages, smoked marijuana or hashish, sold marijuana or hashish, and
used other drugs. Standardized item alpha reliability for males =.77; for
females, .74.
The third variable is “property crime”, derived
from the number of times in the previous year the respondent intentionally
damaged property, shoplifted, stole items worth less than $50 or more than $50,
knew someone who held or sold stolen property, and had broken into a building.
Standardized item alpha reliabilities were .84 for males and .69 for females.
The fourth variable is “violence”, consisting of the number of times over
the previous year the respondent used force to obtain things, seriously
threatened someone, attacked someone with the intention of injuring them, or had
fought at school or work. Standardized item reliabilities were .55 for males and
.73 for females. We contend, (uncontroversially, we hope) that crimes
against persons, crimes against property, and drug offenses represent different
types of illegality. Criminal codes throughout much of the
English-speaking world view interpersonal acts of aggression and violence as
more morally wrong, and therefore more serious, than crimes against property.
This hierarchy is reflected in sentencing dispositions. Our assumption is
that engaging in forms of deviance that are generally agreed to be wrong and
harmful, and which, upon apprehension, are severely responded to, will produce
more negative long term consequences than participation in less serious
wrong-doings. Drugs offenses are of course a slightly different story.
There is considerably less consensus about the wrongfulness and harm caused by
drug use - exemplified by periodic efforts to de-criminalize marijuana. At
the same time, much societal concern - in its extreme form, moral panic -
centers on adolescent drug use. While the severity of responses to the
drug problem may be out of proportion to its projected gravity, this should not
obscure the fact that illicit substance use is a risky pleasure, carrying many
potentially negative ramifications if discovered. In sum, we are using a
variety of indicators of delinquency because of the varying levels societal
response to different forms of delinquency; such varying responses, we argue,
may well lead to differing effects of adult attainment.
We also added a fifth measure of delinquency.
“Contact with criminal justice system” is a dummy variable that scores a
value of one if the respondent had ever had any one of the following experiences
by 1980: if they had ever been stopped by police, or booked, or charged, or
convicted. While previous research has occasionally looked at the effects of
these sorts of contacts, the follow-up periods have usually been very
short (ie. within the teenage years; see Sampson and Laub, 1997). In contrast,
we explore the effects of this variables across the adolescent-adult divide, and
assume that the police and courts are most inclined to handle the most
persistent and serious offenders. Adolescent encounters with the criminal
justice system thus become, in the first instance, a proxy measure of
respondent’s own delinquent behavior. In addition, it is also possible that
such encounters are, or will become, generative of adolescent dissent.
Being stopped by police - no matter howl reasonable from the police view - may
provoke a spirit of defiance and disrespect for authority that may make further
offending more likely (Sherman, 1993). Farrington (1977) has demonstrated that
subsequent to first conviction, English delinquents develop more aggressive
attitudes towards the police. Similarly, those stoppages by the police that
result in arrest and conviction, and thus receive official and public
recognition as being criminal, provide a fertile breeding ground for labeling,
the formation of deviant identities, and most concretely, reluctant hiring by
employers.
Our expectations with respect to the effects of
delinquency vary. At the bivariate level, we expect all forms of delinquency to
have a negative impact upon school attainment, for both males and females.
However, at the multivariate level, the more severe forms of delinquency, such
as violence and property crime, may not be significant once key measures of
background and cognitive skill are controlled. Those individuals most likely to
engage in those crimes may come from the types of backgrounds (ie. low SES),
or have the types of student profiles (ie. low standardized tests scores) that
result in poor school attainments. Further, while we expect delinquency to
be important in models of educational attainment, those effects could be weaker
in models of occupational attainment for the simple reason that the later occur
later in time, and are more distant from the occurrence of teenage delinquency.
Table 1: Variable Definitions and Descriptive Statistis
| Variable |
Description |
Male
mean (sd)
N=1,144 |
Female mean (sd)
N=1,112 |
| SES |
Summed z scores of father’s Duncan SEI score,
father’s and mother’s educational attainment when respondent was 14 |
.738
(2.034) |
.612
(2.041) |
| # of Siblings, 1979 |
|
3.041
(2.120) |
3.197
(2.141) |
| Two parent family |
=1 if lived with natural father and mother at age |
.853
(.354) . |
.836
(.370) |
| Age |
Age Years of age in 1979 |
15.550
(1.055) |
15.629
(1.063) |
| African American |
If African American =1 |
.109
(.312) |
.114
(.318) |
| Hispanic |
If Hispanic =1 |
.060
(.238) |
.077
(.267) |
| Cultural Resources |
Additive scale of whether anyone in
respondent’s household at age 14 received magazines newspapers or
owned a library card |
2.268
(.880) |
2.276
(.900) |
| Cog. Skill |
Percentile score on the AFQT |
46.993
(28.900) |
45.753
(26.826) |
| Educational Expectations |
Highest grade respondent expects to complete,
1979, measured in years |
14.108
(2.215) |
14.124
(2.070) |
| Highest grade completed, ‘92 |
Measured in years |
13.348
(2.377) |
13.409
(2.244) |
| Skipped School |
Number of days skipped in 1979 |
1.335
(1.507) |
1.246
(1.390) |
| Drugs |
Summed z scores from 1979, self reported number of
occurrences in the past year: drank alcohol, smoked marijuana or
hashish, used other drugs, sold marijuana or hashish (standardized item
alpha =.770 for males, .740 for females) |
-.054
(3.703) |
.082
(3.752) |
| Property Crime |
Summed z scores from 1980; self report of number of
times in the past year the respondent had intentionally damaged
property, stole belongings worth less than $50, stole belongings worth
more than $50, shoplifted, broken into a building, knew someone who
held/sold stolen goods (standardized item alpha = .837 for males, 692
for females) |
-.117
(4.259) |
.093
(3.867) |
| Violence |
Summed z scores from 1980, self report of number of
times the respondent in the past year fought at school or work, used
force to obtain things, seriously threatened someone, and attacked
someone with the intention to injure (standardized item alpha = .546 for
males, .732 for females) |
-.035
(2.898) |
-.019
(2.655) |
| Contact with criminal justice system |
If ever one of the following: stopped by police,
or booked, or charged, or convicted by 1980 =1 |
.298
(.458) |
.103
(.305) |
| Occupational status in 1992 |
Duncan SEI score |
39.507
(23.676) |
48.829
(20.609) |
| Unemployed in 1990 |
If unemployed for at least 1 week = 1 |
.167
(.373) |
.156
(.363) |
| College graduate in 1992 |
If highest grade completed is equal to 16 or more =1 |
.252
(.434) |
.248
(.432) |
| High School diploma by 1992 |
If received a high school diploma or GED by 1992
=1 |
.889
(.314) |
.919
(.273) |
Our control variables consist of social background and
school measures that are pertinent to educational and occupational attainment.
Background variables include SES, race, family structure, number of siblings,
and cultural capital. SES is a composite of standardized scores for
father's Duncan SEI, mother's education, and father's education (measured
by years of schooling). Two dummy variables represent racial background: African
American and Hispanic; the reference is non-black / non-Hispanic. Family
structure was measured as a dummy variable where living in a two-parent
household at age 14 was coded 1, all other family arrangements coded 0.
Family size is a raw score of the respondent’s number of siblings. We also add
a measure of cultural resources, used by some researchers as an indicator of
family-transmitted cultural capital (see Teachman 1987). This is a composite
ranging from 0 to 3, derived from whether at age 14 the respondent's family
subscribed to magazines, subscribed to newspapers, or owned a library card. This
measure has been shown to predict various educational outcomes (Davies and
Guppy, 1997).
A second block of independent variables consists of
two academic measures which are pertinent to educational and occupational
attainment and which are standard predictors of school outcomes in status
attainment models. “Cognitive skill” is measured by the respondent’s
percentile score on the AFQT (which measures skill and knowledge in reasoning
and language). Educational expectations is the number of years of
schooling the respondent expected to eventually attain.
Our dependent variables consist of education and
employment outcomes. Three educational outcomes represent differing levels of
attainment by 1992: highest grade completed, whether the respondent ever
received a high school diploma or GED, and whether they completed a college
degree. Given that opportunities to return to school have greatly expanded
in the past 20 years, these measures from 1992 offer a longer-term perspective
than cross-sectional measures of dropping out or university attendance. For
employment outcomes, we have two important indicators of life chances and
conditions: respondent’s Duncan SEI in 1992, and whether they were unemployed
in 1990.
Analysis
Our statistical analysis proceeds in four stages. We
commence with a series of bivariate regressions to measure gross effects of
delinquency on educational and occupational attainment. We then introduce
controls in three successive multivariate models that resemble status attainment
models but that contain also delinquency variables. The first multivariate model
examines the partial effects of each delinquency variable separately,
controlling for background variables. The second model adds educational
variables to those equations, while the final model incorporates all delinquency
variables into the same equations in order to estimate the unique effects of the
various forms of delinquency. For models of employment we add the variable
“highest grade completed” to control for eventual educational attainment.
Our statistical analysis consists of ordinary least square regressions for
continuous dependent variables, and logistic regression for the
dichotomous dependent variables.
Findings
Is delinquency related to educational and employment
outcomes? Does delinquency have significant effects on attainment net of
standard status attainment variables? Tables 2 and 3 presents regression
coefficients for delinquency from three models for each dependent variable.
Model one in each instance contains bivariate regression coefficients; model two
consists of background variables and one delinquency variable, and model three
consists of background variables, education variables, and one delinquency
variable (in the latter two models only the partial coefficients for delinquency
are presented for reasons of space). Each coefficient in models two and
three represents a separate regression (ie. one that includes skipped school,
then one that includes drugs, then property crime, etc).
The upper panels of Tables 2 and 3 show that among
both males and females, all forms of delinquency have a significantly negative
impact on the three educational outcomes. Whether at the zero-order
level (model 1), or controlling for background (model 2), or controlling for
both background and education (model 3), each form of delinquency (with only a
single exception) has a significant negative impact on each measure of
educational attainment. Since the magnitudes of the delinquency coefficients are
reduced very little upon the inclusion of background variables (and in some
instance they grow), it does not appear that delinquents merely hail from the
types of backgrounds that tend to produce unsuccessful students. Rather,
the act of delinquency itself has a negative influence, over and above the
demographic characteristics of the delinquent. Further, those negative
influences persist even controlling for key student characteristics such as
cognitive skill and educational expectations. Thus, delinquency does not merely
reduce students’ educational skills and aspirations, nor are youth with lower
expectations and fewer skills merely more likely to be delinquent. Rather,
delinquency exerts a penalty over and above one’s academic characteristics.
These findings thus confirm the contention that teenage delinquents, male and
female, have lessened eventual outcomes in school.
There is, however, a more varied pattern of effects
across the two employment outcomes, as shown in the lower panels of Tables 2 and
3. Among males, all forms delinquency have expected negative impacts on
occupational status in bivariate models and models that control for background.
For instance, whether or not one has contact with the criminal justice system
reduces one’s occupational status score by over seven points; each day
reported of skipping school reduces those scores by over two points.
However, when educational variables (including highest grade completed) are
added in model 3, all delinquency coefficients are sharply reduced.
Indeed, only drug use and property crime have significant direct effects on
occupational status controlling for both background and education. These
findings suggest that while some forms of delinquency have direct impacts on
men’s eventual job attainment, much of that impact is mediated through
education.
Turning to the right-hand side of the lower panel of Table
2, all delinquency variables have significantly negative impacts on unemployment
among males. Men’s odds of being unemployed during 1990 were increased
by involvement in all forms of delinquency in 1980, even controlling for
background and education. Delinquency thus appears to create some insecurity in
males’ subsequent employment. Indeed, Table 2 suggests that for males, high
school delinquency has effects that are felt twelve years later in both
educational and occupational realms.4
Among females, however, the lower panel of Table 3
presents a somewhat different story. Female occupational status is barely
affected by delinquency.5 The most noteworthy effect is found in
model two, wherein contact with the criminal justice system hurts female
occupational status scores by over five points. However, this effect drops
to non-significance once educational variables are introduced in model three,
suggesting that for females, the negative influence on delinquency on
occupational status occurs largely through its hindrance on their educational
attainment. This gender difference in effects can be explained in terms of
women’s continuing need to balance employment and familial obligations.
Women, no matter how well qualified, are still less inclined than men to
maximize their occupational aspirations - settling instead for less prestigious
employment if it helps them make that reconciliation. If this is the case,
delinquency would cause relatively lesser harm in the labor market.
Table 2: Estimates from Regressions of Dependent Variables on Delinquency,
Males
| |
Highest Grade Completed
by 1992 |
High School Diploma by
1992 |
College Degree by
1992 |
| Predictor |
Model 1a |
Model 2b |
Model 3c |
Model 1a |
Model 2b |
Model 3c |
Model 1a |
Model 2b |
Model 3c |
| Skipped school |
-.345***
(.042)
n=1,328 |
-3.24***
(.038)
n=1,217 |
-.208***
(.032)
n=1,217 |
-.318***
(.049)
n=1,329 |
-.319***
(.057)
n=1,218 |
-.325***
(.067)
n=1,218 |
-.323***
(.053)
n=1,328 |
-.358***
(.067)
n=1,217 |
-.269***
(.076)
n=1,217 |
| Drugs |
-.088***
(.021)
n=1,260 |
-.137***
(.019)
n=1,162 |
-.089***
(.016)
n=1,162 |
-.047*
(.026)
n=1,261 |
-.099***
(.029)
n=1,163 |
-.117***
(.035)
n=1,163 |
-.092***
(0.25)
n=1,260 |
-.174***
(.032)
n=1,162 |
-.138***
(.036)
n=1,162 |
| Property crime |
-.050***
(.015)
n=1,258 |
-.067***
(.014)
n=1,161 |
-.041***
(.011)
n=1,161 |
-.031*
(.018)
n=1,259 |
-.041*
(.020)
n=1,162 |
-.040*
(.023)
n=1,162 |
-.063***
(.019)
n=1,258 |
-.115***
(.026)
n=1,161 |
-.090***
(.028)
n=1,161 |
| Violence |
-.104***
(.022)
n=1,280 |
-.099***
(.020)
n=1,182 |
-.044**
(.016)
n=1,182 |
-.059**
(.025)
n=1,281 |
-.070**
(.030)
n=1,183 |
-.036
(.034)
n=1,183 |
-.137***
(.030)
n=1,280 |
-.144***
(.035)
n=1,182 |
-.092**
(-.038)
n=1,182 |
| Contact criminal justice |
-.892***
(.140)
n=1,302 |
-.844***
(.123)
n-1,203 |
-.368***
(.104)
n=1,203 |
-.824***
(.176)
n=1,303 |
-.874***
(.202)
n=1,204 |
-.602**
(.226)
n=1,204 |
-.859***
(.160)
n=1,302 |
-.998***
(.191)
n=1,203 |
-.647**
(.222)
n=1,203 |
Table 2 Estimates from Regressions of Dependent Variables on Delinquency,
Males (cont’d)
| |
Occupational Status 1992 |
Unemployment in 1990 |
| Predictor |
Model 1a |
Model 2b |
Model 3c |
Model 1a |
Model 2b |
Model 3c |
| Skipped school |
-2.139***
(.449)
n=1,188 |
-1.818***
(.448)
n=1,090 |
-.465
(.395)
n=1,090 |
.169***
(.047)
n=1,248 |
.152**
(.040)
n=1,118 |
.135**
(.053)
n=1,118 |
| Drugs |
-.889***
(.221)
n=1,132 |
-.1.140***
(.219)
n=1,044 |
-.380***
(.193)
n=1,044 |
-.058***
(.023)
n=1,189 |
.091***
(.025)
n=1,069 |
.086***
(.025)
n=1,069 |
| Property crime |
-.674***
(.164)
n=1,131 |
-.844***
(.164)
n=1,044 |
-.470***
(.142)
n=1,044 |
.026
(.016)
n=1,187 |
.042**
(.018)
n=1,069 |
.037*
(.018)
n=1,069 |
| Violence |
-.897***
(.241)
n=1,148 |
-.871***
(.240)
n=1,060 |
-.187
(.270)
n=1,060 |
.050*
(.024)
n=1,209 |
.-057*
(.026)
n=1,088 |
.042*
(.027)
n=1,088 |
| Contact criminal justice |
-7.211***
(1.479)
n=1,169 |
-7.295***
(1.448)
n=1,080 |
1.848
(1.271)
n=1,080 |
.485***
(.156)
n=1,229 |
.563***
(.169)
n=1,1106 |
.454**
(.174)
n=1,106 |
Notes:
For highest grade completed and occupational status, cells
contain unstandardized OLS coefficients; all other cells contain logistic
regression coefficients. Standard errors are in parentheses.
* denotes 1-tailed p <.05, ** p< .01,
*** p< .001
a Zero-order coefficients
b Partial coefficients controlling for SES, number of siblings,
family type, age, African American, Hispanic, cultural resources
c Partial coefficients controlling for all variables in b and
cognitive skill, educational expectations.
Table 3: Estimates from Regressions of Dependent Variables on Delinquency,
Females
| |
Highest Grade Completed by
1992 |
High School Diploma
by 1992 |
College Degree by 1992 |
| Predictor |
Model 1a |
Model 2b |
Model 3c |
Model 1a |
Model 2b |
Model 3c |
Model 1a |
Model 2b |
Model 3c |
| Skipped school |
-.249***
(.044)
n=1,289 |
2.33***
(.041)
n=1,162 |
-.134***
(.034)
n=1,162 |
-.287***
(.063)
n=1,290 |
-.315***
(.072)
n=1,163 |
-.277***
(.080)
n=1,163 |
-.284***
(.056)
n=1,289 |
-.318***
(.068)
n=1,162 |
-.262***
(.079)
n=1,162 |
| Drugs |
-.053**
(.020)
n=1,232 |
-.101***
(.018)
n=1,128 |
-.064***
(.015)
n=1,128 |
-.016
(.032)
n=1,233 |
-.080*
(.036)
n=1,129 |
-.086*
(.039)
n=1,129 |
-.080***
(0.24)
n=1,232 |
-.153***
(.031)
n=1,128 |
-.121***
(.034)
n=1,128 |
| Property crime |
-.041**
(.016)
n=1,228 |
-.051***
(.015)
n=1,123 |
-.034**
(.012)
n=1,123 |
-.056**
(.019)
n=1,229 |
-.068***
(.021)
n=1,124 |
-.068**
(.025)
n=1,124 |
-.049*
(.022)
n=1,228 |
-.093***
(.031)
n=1,123 |
-.097**
(.035)
n=1,123 |
| Violence |
-.135***
(.024)
n=1,241 |
-.088***
(.022)
n=1,135 |
-.037*
(.018)
n=1,135 |
-.117***
(.028)
n=1,242 |
-.109***
(.031)
n=1,136 |
-.074*
(.035)
n=1,136 |
-.182***
(.043)
n=1,241 |
-.117**
(.045)
n=1,182 |
-.087**
(-.045)
n=1,135 |
| Contact criminal justice |
-.505***
(.207)
n=1,265 |
-.726***
(.186)
n-1,155 |
-.389**
(.154)
n=1,155 |
-.662**
(.277)
n=1,266 |
-1.112***
(.320)
n-1,156 |
-.732*
(.360)
n=1,156 |
-.474*
(.244)
n=1,265 |
-.805**
(.286)
n=1,155 |
-.652*
(.337)
n=1,155 |
Table 3: Estimates from Regressions of Dependent Variables on Delinquency,
Females(cont’d)
| |
Occupational Status 1992 |
Unemployment in 1990 |
| Predictor |
Model 1a |
Model 2b |
Model 3c |
Model 1a |
Model 2b |
Model 3c |
| Skippsd School |
-.597
(.469)
n=1,060 |
-659
(.478)
n=961 |
-.195
(.453)
n=961 |
.005
(.056)
n=1,228 |
-.032
(.062)
n=1,091 |
-.049
(.063)
n=1,091 |
| Drugs |
-.206
(.206)
n=1,015 |
-.557**
(.207)
n=933 |
-.228
(.197)
n=933 |
.018
(.025)
n=1,177 |
.026
(.026)
n=1,058 |
.022
(.027)
n=1,058 |
| Property crime |
-.222
(.178)
n=1,010 |
-.303*
(.173)
n=928 |
-.169
(.163)
n=928 |
.021
(.018)
n=1,1 76 |
.010
(.021)
n=1,079 |
.010
(.021)
n=1,079 |
| Violence |
-.475*
(.267)
n=1,021 |
-.272
(.265)
n=938 |
-.077
(.249)
n=938 |
.039
(.027)\
n=1,187 |
.033
(.028)
n=1,065 |
.027
(.029)
n=1,065 |
| Contact criminal justice |
-2,709***
(2.152)
n=1,041 |
-5.250**
(2.155)
n=955 |
-2.815
(2.026)
n=955 |
-.277
(.280)
n=1,210 |
-.190
(.293)
n=1,085 |
.257
(.297)
n=1,085 |
Notes:
For highest grade completed and occupational status, cells
contain unstandardized OLS coefficients; all other cells contain logistic
regression coefficients. Standard errors are in parentheses.
* denotes 1-tailed p <.05, ** p< .01,
*** p< .001
a Zero-order coefficients
b Partial coefficients controlling for SES, number of siblings,
family type, age, African American, Hispanic, cultural resources
c Partial coefficients controlling for all variables in b and
cognitive skill, educational expectations.
Furthermore, there is no discernible impact of delinquency
on female unemployment. No delinquency variable is significant in any of
the three models for females. Why is this the case, especially in light of the
links between male delinquency and unemployment? Unemployment may not represent
the same negative status for some women as it does for most men. That is,
among males, deviance represents a negative attribute that penalizes them in the
job market, with unemployment representing one such penalty. However,
given that women are more likely than men to work part-time, being unemployed
temporarily for a woman is not necessarily a penalty if she is seeking full-time
employment as an alternative to part-time work. Thus, since short bouts of
unemployment for women is not necessarily a negative status, it is not
surprising that delinquency is not a predictor of such a status.
Finally, what are the unique effects of the various
forms of delinquency on outcomes? Tables 4 and 5 present the full
multivariate models that include all background, education and delinquency
variables. For these models we present all coefficients in order to
highlight the role of delinquency in the broader context of status attainment.
Starting with the educational outcomes, Table 4
shows that most background measures among males are significant in the expected
direction in all models. SES is a typically strong positive effect on
educational attainment, while number of siblings often has a significant
negative impact. Cultural resources has important positive effects for
males, while cognitive skill and educational expectations are strong predictors
of educational outcomes, adding a significant boost to r square in the model of
highest grade completed (where over 55% of the variance is explained).6
Among females, the same social background and academic variables influence
educational outcomes, though among females there are more race effects.
For our purposes, the important effects are those of
the delinquency. Perhaps the biggest story in Tables 4 and 5 is that there are
fewer significant effects of delinquency. Among both males and females,
only two of five delinquency variables have unique effects on educational
outcomes, and between both genders, only a single delinquency variable is a
significant unique predictor of a single employment outcome. What is
happening in these full models is that the overall impact of delinquency is
shared among its multiple measures. The five delinquency variables are
inter-correlated (ranging from .22 to .52) enough to contribute in similar ways
to the variance explained, but are not correlated so much as to create
collinearity problems (none of the SPSS diagnostics detected any such
difficulties). To investigate this further, we introduced another model
specification (tables not shown; available upon request) in which a unitary
index of delinquency was constructed from the five existing delinquency
variables (each of which was standardized and then summed).
Table 4: Regression Estimates for Educational Outcomes by 1992, Full
Models
| |
Highest Grade Completed |
High School Diploma |
College Degree |
| Predictor |
Male |
Female |
Male |
Female |
Male |
Female |
| SES |
.178***
(.030) |
.217***
(.030) |
.206***
(.085) |
.061**
(.099) |
.297***
(.058) |
.362***
(.061) |
| Number of Siblings |
-.046*
(.024) |
.041*
(.023) |
.016
(.052) |
-.075
(.057) |
-.079
(.054) |
-.047
(.053) |
| Two-Parent Family |
-.110
(.139) |
.115
(.133) |
.328
(.300) |
1.109***
(.307) |
.171
(.306) |
.020
(.283) |
| Age |
.007
(.047) |
-.082*
(.046) |
.080
(.115) |
.233*
(.129) |
-.062
(.095) |
-.030
(.093) |
| African American |
.245
(.174) |
.442**
(.171) |
.592*
(.348) |
.156
(.378) |
.020
(.455) |
.093
(.422) |
| Hispanic |
-.111
(.213) |
.346*
(.195) |
-.237
(.427) |
-.148
(.427) |
.011
(.505) |
-.135
(.481) |
| Cultural Resources |
.187**
(.062) |
-.018
(.062) |
.169
(.131) |
-.039
(.154) |
.368**
(.152) |
-.277
(.138) |
| Skipped School |
-.146**
(.037) |
-.070*
(.039) |
-.247***
(.080) |
-.254**
(.093) |
-.145*
(.086) |
-.128
(0.89) |
| Drugs |
-.048**
(.020) |
-.041*
(.019) |
-.054
(.046) |
-.015
(.052) |
-.064
(.044) |
-.077*
(.041) |
| Property Crime |
-.003
(.015) |
-.007
(.015) |
-.002
(.033) |
-.038
(.031) |
-.032
(.034) |
-.038
(.040) |
| Violence |
-.010
(.020) |
-.005
(.020) |
.050
(.049) |
-.022
(.041) |
-.017
(.042) |
-.021
(.051) |
| Contact with criminal justice |
-.171
(.113) |
-.178
(.166) |
-.329
(.262) |
-.393
(.413) |
-.358
(.244) |
.265
(.369) |
| Cognitive skill |
.029***
(.002) |
.032***
(.002) |
.068***
(.009) |
.056***
(.010) |
.039***
(.005) |
.048***
(.005) |
| Educational Expectations |
.323***
(.028) |
.315***
(.028) |
.454***
(.096) |
.464***
(.102) |
.389***
(.057) |
.343***
(.057) |
| Constant |
7.203 |
6.139 |
-6.825 |
-.8251 |
-9.297 |
-7.974 |
| R² |
.556 |
.519 |
|
|
|
|
| Pseudo R² |
|
|
.406 |
.349 |
.442 |
.419 |
Notes:
N=1,144 for males; 1,1112 for females. Cells for Highest Grade
Completed contain unstandardized OLS regression estimates; all other cells
contain logistic regression coefficients; Standard errors are in parentheses
* denotes 1-tailed p < .05, ** < .01,
*** < .001
Results using this omnibus measure of delinquency were
essentially identical to those in third models of table 2: delinquency was
strongly significant for both male and female educational outcomes and for male
occupational outcomes, but not for female occupational outcomes. Given
this set of shared effects, Table 5 simply alerts us to “unique” effects,
that is, the effects of some delinquency measures independent of related
delinquency variables.
These unique effects are as follows. Among males,
skipping school has a consistently negative impact on school outcomes, net of
all background, educational, and other delinquency variables. Each
reported day of skipping reduced attainment by 0.15 years net of all other
variables, and reduced the odds ratio of getting a college degree to .865 and of
getting a high school diploma to .760. These findings are
particularly interesting, given that skipping is arguably the least serious of
our measures of delinquency. This suggests that this form of delinquency is a
“bad habit” that has autonomously negative effects on educational
attainment. Drug use also has important, independent effects on highest
grade completed, but not on the other outcomes. Among females, skipping school
and drug use each have significant net impacts on two of the three outcomes.
The other delinquency variables (contact with criminal justice system, property
crime and violence), which were all significant table 2, are reduced to
non-significance in table 4, for both males and females. It appears that the
mildest form of delinquency, skipping school, has the most consistent effects on
education because it is the most ubiquitous and common form of deviance across
various social categories of youth. It suggests that such youth fare worse in
school, over and above their background and academic endowments, and their
engagement in other forms of delinquency.
We turn next to employment outcomes. Table 5 shows
that among males, the expected effects of prior educational attainment on both
occupational status and unemployment. The r square of .385 suggests that these
variables are solid predictors of occupational status, but the smaller pseudo r
square [1 - (-2 log likelihood for current model / -2 log likelihood for base
model)] for unemployment suggests the latter is more difficult to predict. There
is, however, only one direct delinquency effect among males: property crime has
a significant independent effect on occupational status. Given the number
of significant effects in model 3 of Table 2, this again dramatizes the shared
effects of delinquency which are masked by when all delinquency variables are
included simultaneously. For females, Table 5 shows similar effects of academic
variables on occupational status, but fewer such effects on unemployment. Among
delinquency variables, there are no unique effects in either model, which is not
surprising given the paucity of significant effects for female occupational
status and unemployment in the third models of Table 3.
Table 5: Regression Estimates for Employment Variables, Full Models
| |
Males |
Females |
| Predictor |
Occupational Status
in 1992
(N=1,029) |
Unemployed
in 1990
(N= 1,053) |
Occupation Status
in 1992
(N-917) |
Unemployed
in 1990
(N=1,045) |
| SES |
-.078
(.365) |
-.008
(.057) |
.-452
(.397) |
.026
(.058) |
| Number of Siblings |
-.283
(.307) |
.045
(.040) |
.130
(.318) |
-.026
(.043) |
| Two-Parent Family |
-2.491
(1.736) |
.218
(.234) |
2.339
(1.763) |
.286
(.233) |
| Age |
-.803
(.577) |
.158*
(.084) |
.397
(.587) |
-.040
(.083) |
| African American |
-.283
(2.244) |
.105
(.295) |
-.939
(2.304) |
-.351
(.306) |
| Hispanic |
2.016
(2.679) |
.296
(.353) |
2.854
(2.561) |
-.179
(.357) |
| Cultural Resources |
-.090
(.791) |
-.095
(.107) |
-.173
(.818) |
-.117
(.110) |
| Skipped School |
-.529
(.459) |
.082
(.062) |
-.592
(.511) |
-.107
(.073) |
| Drugs |
-.145
(.245) |
.049
(.033) |
-.209
(.245) |
.046
(.033) |
| Property Crime |
-.534**
(.188) |
-.009
(.025) |
-.140
(.196) |
.000
(.025) |
| Violence |
.223
(.253) |
.023
(.034) |
.209
(.273) |
.030
(.033) |
Contact with
criminal justice |
-.894
(1.411) |
.279
(.195) |
-2.779
(2.192) |
-.301
(.315) |
| Cognitive skill |
175***
(.030) |
-.011*
(.005) |
.143***
(.033) |
-.012**
(.005) |
| Educational Expectations |
1.012**
(.369) |
-.153**
(.054) |
.298
(.382) |
.043
(.053) |
| Highest grade completed |
3.734***
(.368) |
-.111*
(.056) |
2.434***
(.388) |
-.026
(.056) |
| Constant |
-19.015 |
2.547 |
-5.001 |
-1.26 |
| R2 |
.385 |
|
.204 |
|
| Pseudo R² |
|
.065 |
|
.021 |
Note:
Cells for Highest Grade Completed contain unstandardized OLS
regression estimates; all others contain logistic regression coefficients;
Standard errors are in parentheses
* denotes 1-tailed p < .05, ** < .01,
*** < .001
In sum, our investigation yields two main findings.
First, among both males and females, all forms of delinquency have consistently
negative and significant effects on eventual educational attainment, even
controlling for key background and educational variables. Full models
suggest that skipping school and drug use have unique impacts on school
attainment. Second, there are key gender differences with regard to
employment variables. Among males, there are some direct influences of
delinquency on unemployment and occupational status controlling for background
and education, some indirect influences that are mediated by education
variables, and one significant “unique” effect in the full model.
Among females, delinquency has some indirect influence on occupational status
through its impact on education, but has no direct or unique impact controlling
for education. Further, delinquency does not appear to have any influence
on female unemployment.
Conclusions
We know now what happened to yesterday’s rebels:
according to our evidence, delinquency in adolescence yields a negative dividend
and reduces educational and occupational attainments in young adulthood,
especially among males. While some social scientists might predict this outcome,
and most parents fear it, much of the prior supportive evidence has relied on
data from distinctively disenfranchised populations. Our investigation confirms,
for the first time, that delinquency reduces the attainments of ‘ordinary
kids’ as well as their more disadvantaged counterparts.
That said, the bigger picture is both more
complicated and more qualified because it requires separate considerations of
male and female experiences and of different outcomes. For males, almost all
forms of adolescent misconduct—from serious violent activity to skipping
school — have some consequence for their futures. Some of this status
non-attainment stems from the deleterious effects that deviance has upon
educational performance, though some is directly attributable to that deviant
activity. The educational attainments of young women are also influenced by the
indiscretions of their youth, but not their occupational outcomes. The
latter finding may partly reflect the lesser amount variation in female
occupational outcomes to be explained (see Table 1 for the smaller standard
deviation of occupational status scores among females versus males), though it
likely reflects also the differing roles of employment in male and female lives.
These findings are important for three main reasons.
Firstly, they testify to the importance of avoiding trouble when young: early
deviance, both directly and indirectly, has lingering effects that negatively
influence life chances. Secondly, they illuminate the darker side of
subterranean values as practiced by American adolescents; and, in so doing so,
cast doubt on the notion of benign or innocuous wrong-doings. That tradition,
associated with Matza, argues that involvement in acceptable, if not
respectable, teenage culture inoculates adolescents from involvement in more
serious delinquency at the same time that it prepares them for conventional
adult lifestyles. Our findings do not support this contention, and unlike
Hagan, we find no indication that any sub-group of adolescents benefits from an
earlier initiation into the risky pleasures of minor delinquency. Perhaps
this is because we really are talking about delinquency - not a “party
subculture” that only skirmishes with illegalities (as opposed to having a
“good time”) when drugs are alcohol are factored into the equation.
Third, and perhaps most important, our findings
extend efforts to integrate insights from deviancy and stratification research.
Many sociologists in the area of status attainment rightly emphasize the
distinct impacts on social position of different sources of capital - human,
social, and cultural - as well as the more familiar economic determinants.
Researchers in this tradition show that the presence or absence of these forms
of capital have independent effects upon of the status passage into adult life.
While stratification researchers speak to the import of capital in advancing
social attainment, theorists of subcultural delinquency and resistance, from
Albert Cohen to Paul Willis, in their different ways, emphasize the reverse:
that immersion in oppositional working class cultural practices ensure the
reproduction of class inequality. However, our findings show that status
attainment is stalled, in addition, by participation in delinquent behavior -
independent of the amounts and quality of human and cultural capital that
delinquents possess.
We do not want to overstate our case here; the
distance - temporal and experiential - between delinquency in adolescence and
getting a job in young adulthood limits the scope for direct effects of the
former on the latter. Just as few students of social stratification
(outside strict Marxist orthodoxies) would expect parental social origins to
have an unmediated effect upon occupational destinations, few contemporary
criminologists would expect that delinquency always or even often to have an
unfettered influence upon occupational entry. What is surprising therefore, is
not the modesty of the direct effects of delinquency on occupation, but the fact
that delinquency has any non-reducible impact upon occupational outcomes at all.
And it is this finding which has the most resonance
for the debate referred to earlier on - between those who trace attainments in
adult life back to a fundamental personality characteristic developed in
childhood and those who stress the significance of particular events and
situations to which individuals are exposed. Initial appearances to the
contrary, these are not necessarily mutually exclusive propositions. In
our study, the human resource of “cognitive skill” has the same obdurant
generalizing qualities that Gottfredson and Hirschi attribute to the trait of
self control: it is the strongest and most consistent predictor of educational
and occupational attainments. At the same time, it is also the case that
delinquent events can impede and reverse the advantages that cognitive skill
bestows upon individuals: thus youthful deviance has a relatively autonomous
impact upon life chances.
Just why deviance has this impact is of course
another question entirely. What is it exactly about delinquency that impedes
progress in the labor market? Is it some relatively stable trait or motivational
configuration that is a common source of variation in both legitimate economic
activity and illegitimate criminal activity throughout the life course? Or
do life-transforming events retard occupational progress regardless of the
characteristics of those exposed to them? Delinquency may foster poor work
skills, limit social networks, or bring on the effects of labeling and
stigmatizing.
Sorting out the relative import of these
explanations is obviously the task of future research. We suggest that
responses to deviance may have an effect on life course trajectories - a key
proposition of labeling theory (Sampson and Laub, 1997). In this argument,
contact(s) with the criminal justice system constitutes potential turning points
in individual development. Research indicates that adults who are
officially identified and processed as criminals will find it harder than other
job-seekers to secure employment (Sampson and Laub, 1997; see also Freeman,
1992). It is reasonable to suppose that given a choice, employers will choose to
hire those without police records. Similarly, escalating involvements with
police, courts and custodial institutions will, in due course, stigmatize and
marginalize young people, thereby shrinking their pro-social employment
networks, and cut them off from information about conventional jobs (Hagan,
1993). Finally, even if troublesome behavior does not result in official
recognition by police and courts, it is likely to generate private conflict with
parents and care-givers at home, and more public confrontations with teachers at
school. While such disputes might be smoothed over by skillful counseling,
it can also be assumed that either, or both of, punitive parenting and
authoritarian schooling will engender a negative dividend over time.
Needless to say, confirmation of the explanatory power of labeling theory
(or some version of it) awaits the demonstration that the stigmatizing and
segregating life events provided by the criminal justice system are not simply
the spurious artefact of prior and present traits or criminal orientations.
Furthermore, if labeling is indeed responsible for turning yesterday’s rebels
into today’s occupational low fliers, we do not know whether it affect all
rebels equally, or only those without protection from various forms of capital.
These questions await further research.
Notes
1. Given the spectacular world-wide success of Learning to Labour -
winner of the most citations as an exemplary study in a 1981 survey of British
sociologists, and recently credited in Contemporary Sociology and Footnotes
as one of the most influential books of the past twenty-five years - it would be
intriguing to know what became of Joey, Spanksy and the rest of the 'lads.' Did
they end up in, and stay with, working class jobs when they left school? Or did
they escape that fate by returning to the educational system? Quite obviously, a
proper evaluation of Willis' thesis twenty years on depends upon finding
out what eventually happened to the lads. A rumor circulating in British
sociology is that Joey, the ring-leader of the lads, did indeed re-enter the
chsool system, and is now the proud recipient of a B.A. in English.
2. The contingent and qualified nature of the link between juvenile
delinquency and adult crime is well illustrated by the Philadelphia Cohort
Study, in which less than 10% of the non-delinquents had no adult arrest record
by age 26, compared with about one-third of the delinquent population (Tracy and
Kempf-Leonard, 1997:206). While those findings clearly show that
delinquency predicts adult criminality, they also reveal that for the majority
of delinquent youth, teen deviancy is not an all-determining harbinger of more
serious trouble to come. Moreover, those researchers who identify continuity
between juvenile crime and adult crime, or who view adolescent rebellion as a
crucial cultural agent in reproducing existing social and economic inequalities,
do so by focusing attention upon sites of extreme disadvantage - impoverished
inner city neighborhoods - or by targeting members of a reputed underclass (see
for example Freedman, 1992). While paying attention to such communities is
obviously extremely important, there is also a danger in allowing them to
represent the experiences of all adolescents growing up in the United States.
Ironically, Sampson and Laub’s influential study has been criticized for
excluding high risk black youth (Tracy and Kempf-Leonard, 1997:62).
3. Though the delinquency variables are inter-correlated (see the
correlation matrix in the Appendix), there do not appear to be any problems of
collinearity. The largest correlation between any of the delinquency
variables is only .51; the lowest tolerance for any variable, among either males
and females, is .51, while the largest variance inflation value was 1.9.
None of these indicators suggest any danger of collinearity. This issue is
further discussed in the Findings section.
4. We also investigated the possibility of interaction effects by creating a
series of product terms between SES and delinquency. However, none of
these interaction terms added a significant increment to r square. While this
non-finding may appear on the surface to contrast with Hagan (1991), who found
interactions between class background and the effects of subcultural
participation on early occupational status, it should be remembered that
Hagan’s “party subculture”, which had positive effects for middle class
males, consisted of leisure activities rather than serious forms of delinquency.
5. Coefficients actually increase between models one and two for four of the
five delinquency variables in the female occupational status regressions.
This is because SES is correlated with occupational status, but not delinquency
(see Appendix). Thus the inclusion of SES in models thus clarifies the
delinquency-occupational status link.
6. Age has effects in some models, but it is likely an artefact of our
sample. Since our sample includes only youth still in school, the older
students (aged 17 in 1979) would not include among their ranks any high
school dropouts. Since few 14 or 15 year olds drop out, the younger cohorts
would be less selective, and more representative. This interpretation is
supported by the finding that the age coefficients are no longer significant
once academic variables are added to the models.
References
Aschaffenburg, Karen, and Ineke Maas. 1997. “Cultural and Educational
Careers: The Dynamics of Social Reproduction.” American Sociological Review
62(4): 573-587.
Alexander, Karl, L, Doris, R. Entwisle, and Carrie S. Horsey. 1997. “From
First Grade Forward: Early Foundations of High School Dropout.” Sociology
of Education 70(2):87-107.
Bourdieu, Pierre. 1984. Distinction: A Social Critique of the Judgement of
Taste. Cambridge: Harvard University Press.
Brown, Phillip. 1987. Schooling Ordinary Kids: Inequality,
Unemployment, and the New Vocationalism. London: Tavistock.
Bynner, John and Sheena Ashford, 1992. “Teenage Careers and Leisure Lives:
An Analysis of Lifestyles.” Society and Leisure 15(2):499-520.
Chesney-Lind, Meda and Randall Sheldon. 1992. Girls,
Delinquency and Juvenile Justice. Belmont, California: Brooks/Cole.
Cohen, Albert K. 1955. Delinquent Boys. The Free Press.
Crutchfield, Robert D. and Pitchford, Susan R. 1997. “Work and Crime:
The Effects of Labor Stratification.” Social Forces 76(1):93-118.
Davies, Scott. 1995. “Reproduction and Resistance in Canadian High Schools.
An Empirical Examination of the Willis Thesis.” British Journal of
Sociology 46(4):662-87.
Davies, Scott. and Neil Guppy. 1997. “Fields of Study, College
Selectivity, and Student Inequalities in Higher Education.” Social Forces.
75(4):1417-38.
DiMaggio, Paul 1982. “Cultural Capital and School Success: The Impact of
Status Culture Participation on the Grades of U.S. High School Students.” American
Sociological Review 47(2): 189-201.
Farkas, George. 1996. Human Capital or Cultural Capital? Boulder:
Westview Press.
Farkas, George, P. Grobe, D. Sheehan, and Y. Shuan. 1990. “Cultural
Resources and School Success: Gender, Ethnicity and Poverty Groups Within an
Urban School District.” American Sociological Review 55:127-42.
Farrington, David. 1977. “The Effects of Public Labeling.” British
Journal of Criminology 17:112-125.
Farrington, David. 1992. “Criminal Career research in the United
Kingdom.” British Journal of Criminology, Vol. 32, No 4 (autumn) pp.
531-536.
Freeman, Richard B. 1992. “Crime and the Employment of Disadvantaged
Youths” p201-237 in George E. Peterson and Wayne Vroman (Eds) Urban Labor
Markets and Job Opportunity. Washington, D.C.: The Urban Institute Press.
Frith, Simon. 1985. “The Sociology of Youth.” in Michael Haralabos (ed) Sociology:
New Directions. Ormskirk: Causeway Press.
Glueck, Sheldon and Eleanor Glueck. 1950. Unraveling Juvenile Delinquency.
Cambridge: Harvard University Press.
Hagan, John. 1991. “Destiny and Drift: Subcultural preferences, status
attainments and the risks of youth.” American Sociological Review.
56:567-87.
Hagan, John. 1993. “The Social Embeddedness of Crime and
Unemployment.” Criminology. 31(4):465-490.
Hagan, John and Bill McCarthy. 1992. “Street Life and Delinquency.”
British Journal of Sociology 43(4):533-61.
Jessor, Richard., John Donovan and Frances. Costa. 1993 Beyond
Adolescence: Problem Behaviour and Young Adult Development. New York:
Cambridge University Press.
Keane, Carl, A.R. Gillis, and John Hagan. 1989. “Deterrence and
Amplification of Juvenile Delinquency by Police Contact: The Importance of
Gender and Risk-Orientation.” British Journal of Criminology
29:336-352.
Laub, John. and Robert Sampson. 1994 “Unemployment, Marital Discord, and
Deviant Behavior: The Long-term Correlates of Childhood Misbehavior”, in
Travis Hirschi and Michael Gottfredson (eds.) .1994. The
Generality of Deviance. New Brunswick, New Jersey: Transaction Publishers.
Matza, David. 1961.”Juvenile Delinquency and Subterranean Values” (with
G. Sykes) American Sociological Review 26: 712-29.
MacLeod, Jay. 1995. Ain’t No Makin’ It: Leveled Aspirations in a
Low-Income Neighborhood (2nd Ed.) Boulder, Colo: Westview Press.
McCarthy, Bill and John Hagan. 1992. “Mean Streets: The Theoretical
Significance of Situational Delinquency Among Homeless Youths.” American
Journal of Sociology 98(3):597-627.
Mensch, Barbara S and Dianne Kandel, 1988. “Dropping out of High School and
Drug Involvement.” Sociology of Education. 61(2)95-113.
Monahan, T. 1970. “Police Dispositions of Juvenile Offenders.” Phyhon
31:91-107.
Monk-Turner, Elizabeth. 1989. “Effects of High School Delinquency on
Educational Attainment and Adult Occupational Status,” Sociological
Perspectives. 33 (1) 413-418
Murdock, Graham. 1982. “Mass Communication and Social Violence.” in Peter
Marsh and Ann Campbell (eds) Aggression and Violence. Oxford: Blackwell.
Myers, David E, Ann M. Milne, Keith Baker, Ann M. Milne, and Alan
Ginsburg. 1987. “Student Discipline and High School Performance.” Sociology
of Education 60(1):18-33.
Polk, Kenneth and Walter Schafer (eds) 1972. School and Delinquency
Englewood Cliffs, J.J.: Prentice Hall. 71 -79.
Reiss, Albert. 1960. “Sex Offences: The Marginal Status of the
Adolescent.” Law and Contemporary Problems. 25:309-33.
Robins, Lee. 1966. Deviant Children Grow Up. Baltimore: Williams and
Wilkins.
Sampson, Robert. and John Laub. 1997 . “A Life-Course Theory of Cumulative
Disadvantage and the Stability of Delinquency.” p.133-161 in Terence P.
Thornberry (ed) Developmental Theories of Crime and Delinquency.
Transaction Publishers: New Brunswick, USA.
Sampson, Robert. and John Laub. 1993 Crime in the Making: pathways
and turning points through life. Cambridge, Ma/London: Harvard University
Press.
Sampson, Robert. and John Laub. 1992 . “Crime and Deviance in the
Life Course”. Annual Review of Sociology. 18:1 63-84.
Schneider, Barbara and James S. Coleman (eds) 1993. Parents, Their
Children, and Schools. Boulder: Westview Press.
Sewell, William and Robert Hauser. 1980. “The Wisconsin Longitudinal Study
of Social and Psychological Factors in Aspirations and Achievements.” Research
in Sociology of Education and Socialization 1:59-99.
Sherman, Lawrence. 1993. “Defiance, Deterrence, and Irrelevance: A Theory
of Criminal Sanction,” Journal of Research in Crime and Delinquency. 30
(4) 445-473.
Smith, David. 1995. “Youth Crime and Conduct Disorders: Trends, Patterns
and Causal Explanations,” pp. 389-489. in Rutter, Michael and David
Smith (eds.), Psychological Disorders in Young People: Time Trends and Their
Causes. Chichester: John Wiley and Sons.
Stinchcombe, Arthur. 1964. Rebellion in a High School. Chicago:
Quadrangle Books.
Teachman, Jay D. 1987. “Family Background, Educational
Resources, and Educational Attainment.” American Sociological Review ,
52, 548-557.
Tanner, Julian. 1996. Teenage Troubles: Youth and Deviance in Canada.
Toronto: Nelson.
Tanner, Julian, Harvey Krahn and Timothy F. Hartnagel. 1995. Fractured
Transitions From School to Work: Revisiting the Dropout Problem.
Toronto: Oxford University Press.
Willis, Paul. 1977 Learning to Labour. Farnborough: Saxon House, Teakfield.
Appendix
Correlations Between all Variables Used in the Analysis (N-2,069 -
2,849)
| Variable- |
(1) |
(2) |
(3) |
(4) |
(5) |
(6) |
(7) |
(8) |
(9) |
(10) |
| (1) Female |
|
|
|
|
|
|
|
|
|
|
| (2) SES |
-.029 |
|
|
|
|
|
|
|
|
|
| (3) Siblings |
-.035 |
-.268** |
|
|
|
|
|
|
|
|
| (4) Family |
-.011 |
.102** |
-.082** |
|
|
|
|
|
|
|
| (5) Age |
.030 |
.046* |
.009 |
.040* |
|
|
|
|
|
|
| (6) Black |
.020 |
-.165** |
.193** |
-.2546** |
.023 |
|
|
|
|
|
| (7) Hispanic |
.016 |
-.269** |
.121** |
-.073** |
-.020 |
-.102** |
|
|
|
|
| (8) Culture |
.010 |
.429** |
-.208** |
.163** |
.053** |
-.207** |
-.219** |
|
|
|
| (9) Skipped |
-.041* |
-.055** |
.018 |
-.092** |
.076** |
-.068** |
.003 |
-.001 |
|
|
| (10) Drugs |
.001 |
.086** |
-.020 |
-.021 |
.122** |
-.127** |
-.063** |
.109** |
.472** |
|
| (11) Property |
.002 |
.033 |
-.004 |
-.041* |
-.042* |
-.029 |
-.019 |
.061** |
.344** |
.511** |
| (12) Violence -.001 |
-.057** |
-.057** |
.049* |
-.046* |
-.071** |
.042* |
.003 |
-.019 |
.271** |
.320** |
| (13) Justice |
-.258** |
.006 |
.028 |
-.046* |
-.010 |
-.032 |
-.015 |
.021 |
.242* |
.311** |
| (14) Cog. Skill |
-.015 |
.503** |
-.250** |
.154** |
140** |
-.341** |
–.153 |
.400** |
-.088** |
.031 |
| (15) Ed. Expect. |
-.015 |
.499** |
-.156** |
.046* |
.035 |
-..046* |
-.010 |
.277** |
-.146** |
.087* |
| (16) High Grade |
.022 |
.497* |
-.213** |
.089** |
.083** |
-.122** |
-.101** |
.324** |
-.190** |
-.096* |
| (17) Diploma |
.057** |
.230** |
-.136** |
.116** |
.090** |
-.063** |
-.098** |
.223** |
-.160** |
-.033 |
| (18) Degree |
.057 |
.446** |
-.156** |
.082** |
.041* |
-.136** |
-.077** |
.247** |
-.158** |
-.101** |
| (19) Occ. Status . |
194** |
.291** |
-.122** |
.061** |
.055** |
-.126** |
-.023 |
.200** |
-.102** |
-.076** |
| (20) Unempld. |
-.013 |
-.059** |
.047* |
-.057* |
-.041* |
.050* |
.018 |
-.076** |
.056** |
.048* |
Appendix cont’d
| Variable |
(11) |
(12) |
(13) |
(14) |
(15) |
(16) |
(17) |
(18) |
(19) |
| (1) Female |
|
|
|
|
|
|
|
|
|
| (2) SES |
|
|
|
|
|
|
|
|
|
| (3) Siblings |
|
|
|
|
|
|
|
|
|
| (4) Family |
|
|
|
|
|
|
|
|
|
| (5) Age |
|
|
|
|
|
|
|
|
|
| (6) Black |
|
|
|
|
|
|
|
|
|
| (7) Hispanic |
|
|
|
|
|
|
|
|
|
| (8) Culture |
|
|
|
|
|
|
|
|
|
| (9) Skipped |
|
|
|
|
|
|
|
|
|
| (10) Drugs |
|
|
|
|
|
|
|
|
|
| (11) Property |
|
|
|
|
|
|
|
|
|
| (12) Violence |
.487** |
|
|
|
|
|
|
|
|
| (13) Justice |
.280** |
.218** |
|
|
|
|
|
|
|
| (14) Cog. Skill |
-.005 |
-.134** |
-.080** |
|
|
|
|
|
|
| (15) Ed. Expect. |
-.064** |
-.106** |
-. 129** |
.509** |
|
|
|
|
|
| (16) High Grade |
-.082** |
-.142** |
-.133** |
.623** |
.592** |
|
|
|
|
| (17) Diploma |
-.066** |
-.095** |
-.119** |
.352** |
.300** |
.476** |
|
|
|
| (18) Degree |
-.080** |
-.125** |
-.107** |
.543** |
.484** |
.828** |
.186** |
|
|
| (19) Occ. Status |
-.084** |
-.087** |
-.150** |
.446** |
.379** |
.515** |
.235** |
.462** |
|
| (20) Unempld. |
.040 |
.053** |
.044* |
-.132** |
-.096** |
-.087** |
-.120** |
-.062** |
-.153** |
* Denotes correlation is significant at the 0.05 level (2-tailed).
** Denotes correlation is significant at the 0.01 level (2-tailed).
