l , LIBRARY
0.3;" 33 Michigan State
0 V University

This is to certify that the
thesis entitled

A Meta-Analytic Review of Social Bonding Theory and
Juvenile Delinquency, with Moderated Meta-Analysis by
Gender

presented by

Rebecca Michele Hayes

has been accepted towards-fulfillment
of the requirements for the

MS. degree in Criminal Ju$tice

Major'ProfeésoflsSignature
O _ 0 7/// 3/7 5'

Date

 

 

 

MSU is an Affinnative Action/Equal Opportunity Institution

 

PLACE IN RETURN Box to remove this checkout from your record.
TO AVOID FINES retum on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

DEC 1 12008
E

l-l l'
(2.1:
ht
m

 

 

 

 

5.,

 

 

 

 

 

 

 

 

 

 

 

2/05 czlci'ficmateomjrm-ws

 

 

A META-ANALYTIC REVIEW OF SOCIAL BONDING THEORY AND JUVENILE
DELINQUENCY, WITH MODERATED META-ANALYSIS, BY GENDER *’
By

Rebecca Michele Hayes

A THESIS
Submitted to
Michigan State University
In partial fulfillment of the requirements
For the degree of
MASTER OF SCIENCE
School of Criminal Justice

2005

ABSTRACT

A meta-analysis was conducted to examine the true score (population) correlation
for Hirschi’s social bonding theory and juvenile delinquency. Findings revealed
considerable inconsistencies in indicators that measure the four social bonding
variables—attachment, involvement, commitment and belief. The true score correlations
for the overall model and also for the models using gender as a moderator ranged from
-.08 (attachment analyzed for males only) to -.33 (beliet). The explained variance in the
distribution of correlations for the meta-analysis ranged from 2.41 (beliet) to 26.26 (for
the overall model for females only, which means there was considerable (as much as
97.59 per cent) unexplained variance, pointing to the likelihood of one or more

moderators of the relationship between social bonding and juvenile delinquency.

Table of Contents

Introduction ............................................................................................ 1
Social Bonding Theory .............................................................................. 2
Issues in Social Bonding Research ................................................................. 5
A Broad Review of Social Bonding Theory ...................................................... 7
Gender Differences and Social Bonding Theory ................................................. 9
Thesis Hypotheses .................................................................................. 11
Methods .............................................................................................. 13
Results ........................... , ..................................................................... 16
Discussion ........................................................................................... 18
Conclusion .......................................................................................... 22
Table ................................................................................................. 23
References .......................................................................................... 24

iii

INTRODUCTION

This study presents a meta-analysis of the relationship between social bonding
theory and juvenile delinquency. The goal of the study is to determine which of the four
social bonding variables-- commitment, attachment, involvement, and belief» best
predict juvenile delinquency. In addition, the study evaluates the hypothesis that gender
is a moderator of the relationship (correlation) between social bonding and juvenile
delinquency. Thus, the study includes a moderated meta-analysis of gender and social
bonding to see how much difference, if any, there is in the amount of explained variance
for males versus females.

Hundreds of studies have been conducted on the relationship between social
bonding theory and delinquency, but research findings are inconsistent. Few studies find
the same or similar correlations. There are several possible reasons for these
inconsistencies in the literature including the use of different sample sizes, which can
contribute to sampling error (Hunter and Schmidt, 1990). In addition, the reliability of
scales used to measure social bonding and delinquency differ across studies. The use of
different measures contributes to measurement error (Hunter and Schmidt, 1990).
However, meta-analysis can correct for these errors to reveal the “true score correlation,”
which is the population correlation. Thus far, no such meta-analysis has been conducted
on social bonding theory and delinquency. This study seeks to fill this gap in the
scientific literature.

The meta-analysis in this study will clarify: (1) the magnitude of the true
correlations (rhos) between each of the four social bonding variables and delinquency,

and (2) the differences, if any, in the magnitude of these correlations for males versus

females. The meta-analysis method used will be that of Hunter and Schmidt (1990). In
light of the extensive literature and studies available on social bonding theory, the
possibility exists that a clearer picture of the relationship between social bonding and
delinquency will emerge as a result of this meta-analysis.
SOCIAL BONDING THEORY

According to Hirschi (1969), the theory of social bonding, also called “social
control,” asserts that delinquent acts occur when one's bond to society is either weak or
broken. An individual's bond to society is achieved through the socialization process and
the level of conformity to social norms (W iatrowski, Griswold and Roberts, 1981).
Different elements that contribute to this bond between individual and society are defined
by four variables: attachment, commitment, involvement and belief (Hirschi, 1969;
Wiatrowski, Griswold and Roberts, 1981; Agnew, 1985, 1991). Each of these variables
will be discussed in greater detail. According to Hirschi’s theory, these variables
discussed further below, are related to each other and weaknesses in one or all of them
lead a juvenile to commit delinquent behavior (Hirschi, 1969).

Social bonding theory presents a different explanation of criminal behavior than
its predecessors that placed the causal importance on motivational processes (i.e.
differential association, labeling, strain; Lasley, 1988). Instead of explaining the
propensity to commit crimes like most theories, social bonding theory seeks to explain
why individuals do n_ot commit crimes (Hirschi, 1969). The theory offers an explanation
based on an individual’s bond to society. It assumes that all individuals would commit
crime but that because of these “social bonds” individuals are disinclined from doing so

(Hirschi, 1969). According to this theory, all individuals have unfulfilled needs and

desires. Because of low social control, people use unconventional means to attain their
needs and desires (Agnew, 1993). In essence, social bonding theory is claiming to be a
general theory that would include everyone and their propensities to attain needs and
desires based on their personal level of social control.

Four elements purportedly contribute to an individual’s bond to society.
Weakness in one or more of these variables provides can lead to criminal behavior
(Hirschi, 1969). The four elements are: attachment, commitment, involvement, and
belief. Many of the studies find differences in the magnitudes of the correlations between
delinquency and each of these four social bonding variables. The social bonding variables
and their indicators are described by Hirschi (1969) and Agnew (1993) as follows:

I Attachment: the affection and respect that the
adolescent holds toward significant others, such as
parents, teachers and friends

0 Indicators include: opinions of affection
they receive from parents, teachers and
friends; expectations of and closeness of
parents, teachers and friends; and amount of

caring towards these people.

' Commitment: the adolescent’s actual or anticipated
investment in conventional activities.
0 Indicators include: Interest in school, sports

or other extracurricular events; school

grades; college and career aspirations; life

goals, etc.

I Involvement: the amount of time spent engaged in
these conventional activities.
0 Indicators include time spent in school,

sports, or other extracurricular events.

- Belief: the adolescent’s commitment to the central
value system of the society or social norms, which
also can be gauged by their religious beliefs.

0 Indicators include: opinions of morality,
honesty, and what is considered wrong by

social standards.

According to Hirschi, the social bond elements are interconnected with each other and, if
one aspect is weak, the others will be weakened as well (1969). For example if a child’s
attachment to his or her family is weak, commitment to school is also assumed to be
weak.

There is much empirical support for social bonding theory. However, as
previously stated, studies’ findings conflict with one another (Agnew, 2003). There is
also a debate as to whether commitment and involvement should be considered the same
variable or if involvement is already tested within the other three variables of

commitment, attachment, and belief (Gardner and Shoemaker, 1989; Junger-Tas, 1992;

Krohn, Lanza-Kaduce and Akers, 1984; Junger and Marshall, 1997; Agnew, 1991). It is
hard to distinguish between commitment and involvement indicators when studies use
them interchangeably, this meta-analysis will statistically untangle this issue.
Furthermore, belief indicators are inconsistent among researchers with some using norm
values and others using religious beliefs and not both (Rosenbaum ,1987; Huebner and
Betts, 2002). For this thesis however, Hirschi’s original propositions are adhered to as
closely as possible, by making sure that the studies remain consistent with social bonding
theory and its corresponding variables.
ISSUES IN SOCIAL BONDING RESEARCH

Hirschi’s Social Bonding theory has been at the forefront of many analyses of
crime, more specifically delinquency, since its introduction in 1969. In 1999, Ellis and
Walsh deemed it the most endorsed theory in criminology. According to Junger and
Marshall(l997), the theory has been tested, challenged, and revised numerous times since
its publication in 1969 (p. 81). Despite all the literature available concerning this theory,
the focus is directed almost entirely on delinquency; there are few studies on adth
criminality based on social bonding (Kempf, 1993).

Agnew's (1991) study on the social bonding variables concluded that attachment,
commitment and belief showed strong significance for the prediction of delinquent
behavior. The fourth variable, involvement, was only weakly related to delinquency
(Agnew, 1991). But ten years prior, Wiatrowski et a1. (1981) conducted a study and
found very different results. According to Wiatrowski's study, attachment was the
strongest predictor of delinquency followed by involvement (Wiatrowski et a1., 1981).

The results of this study showed that the belief variable was not significant when the

other variables were present, and the commitment variable was a weak predictor of
delinquency (Wiatrowski et al., 1981). In yet another study by Wiatrowski and Andersen
(1987), contrary to his previous findings, belief was found as the second most significant
variable preceded by attachment.

Greenberg (1999) conducted a reanalysis of Hirschi’s 1969 study and found
weaker support than Hirschi claimed to have found. The significance of this reanalysis to
the social bonding literature was that it was one of the few studies that claimed to show
minimal relationship between attachment and delinquency (Greenberg, 1999). However,
this reanalysis did not examine all facets of attachment. The focus was on attachment
with only the father (Greenberg, 1999).

Another issue with the variable attachment is that of peer attachment. Many
studies shows a positive relationship with delinquency but, Hirschi’s propositions deem it
to be a negative relationship with delinquency (Agnew, 1985; Barton and Figueira-
McDonough, 1985; Linden, 1978; Marcos, Baht and Johnson, 1986; Agnew, 1991;
Freidman and Rosenbaum, 1988).

In addition to the above, many studies do not always include all of the social
bonding variables, yet they claim to support social bonding theory, (Bishop, 1982;
Rankin and Kern, 1994; Huebner and Betts, 2002; Thompson, Mitchell and Dodder,
1984; Agnew, 1991; Marcos, Bahr and Johnson, 1986; Krohn and Massey, 1980;
Linden,1978; Hepburn, 1976). An example is the study by Rankin and Wells (1990)
which examined only attachment and found it significant in explaining delinquency.
However, Conger (1976) studied only attachment and found weak relationships between

attachment and delinquency.

Further, Thaxton and Agnew (2004) recently studied only attachment, without the
other social bonding variables. This study claimed to have found significant support for
the theory of social bonding. However, these authors only found support for one
component of the theory not the entire theory. All of the studies that were reviewed for
this thesis included the variable ‘attachment’. In the literature, attachment has been found
to explain the most variance compared to the other three variables. However attachment
is the only variable that is present in many studies, such as those in Hepburn (1976),
Freidman and Rosenbaum (1988), Rankin and Kem(l994), Poole and Regoli (1979),
Jensen and Browfield (1983) and Thaxton and Agnew (2004).

Another problem that plagues the literature on social bonding theory is that social
bonding is a vague term. Many studies claim to be social bonding studies but they do not
clearly contain social bonding variables (Heimer, 1997; Jensen and Eve, 1976; Patterson
and Dishion, 1985). Other studies include social bonding variables but these studies
integrate other theories so the results are unclear (Polakowski, 1994; Conger, 1976;
Matsueda, 1982). According to Le Banc and Caplan (1993), until social bonding theory is
properly formalized, inconsistencies will persist in social bonding studies. A formalized
theory would have a clear, concise definition of terms and measurements associated with
that theory. Furthermore, if social bonding is not properly formalized it can not
accurately be integrated with other theories (Le Banc and Caplan, 1993).

A BROAD REVIEW OF SOCIAL BONDING THEORY
Kempf(1993) conducted a broad qualitative review of the many studies on social
bonding. Kempfs review was conducted because of a lack of a systematic critique of

social bonding theory. It examined research on social bonding theory from 1970 to 1991.

The study was comprised of seventy-one studies with the majority of those studies based
on white adolescent males (Kempf, 1993). However, Kempf does not claim her
evaluation is exhaustive due to the three criteria she applied in selecting the studies for
the review.

Kempfs criteria were as follows: "first, there must be an acknowledged test of
control theory; second, Hirschi (1969) must be cited; and third, the study must be
published" (Kempf, 1993, p. 148). Kempt‘s selection criteria imply that in order to
validate the authenticity of a study of social bonding theory the study must mention
Hirschi’s original propositions. Published studies are presumed to be credible, probably
due to the journal review process (Kempf, 1993).

According to Kempf (1993), the limitations of past studies were: a lack of
variation in design elements; little attention to the construct validity of the four elements
within the social bond; the use of only cross-sectional data and when longitudinal data
were used they were analyzed as if cross-sectional. Furthermore, different and often
contrary results were found between studies, and the studies failed to show causation in
order to give the theory scientific merit (Kempf, 1993). The limitations do not invalidate
the usefulness of social bonding theory but rather imply that in order to improve much
work remains to be done (Kempf, 1993). Although Kempf recommended ways to
address these problems, these same issues mentioned above persist today.

Kempfs study, however, was a narrative review and no comparative analyses
were computed. Furthermore, differences in social bonding for male and female
offenders were not explored (Rosenbaum, 1987; Krohn and Massey, 1980). Finally,

Kempf‘s study concluded that more research is needed to further develop social bonding

theory (1993)
GENDER DIFFERENCES AND SOCIAL BONDING THEORY

Based on the increasing numbers of female offenders since the early 1980’s, there
is a growing interest in trying to explain female criminality (Daly and Chesney-Lind,
1988; Daly, 1989). Unfortunately, there is a shortage of criminological theories that
address female criminality. According to Daly and Chesney-Lind (1988), this is a gap in
the literature that needs to be filled. Most feminist theorists tend to ignore the
mainstream theories of crime, because those theories were written by, for, and about men
(Costello and Mederer, 2003).

Costello and Mederer (2003) suggest that such a disregard for current theories and
how they pertain to females might be a mistake by the feminist theorists, who might
consider current theories in order to expand or improve upon them. Therefore, gender
differences will be examined in this thesis to determine if the social bonding variables
have equivalent exploratory power with regard to both sexes and delinquency.

Gender differences manifest themselves according to the manner in which males
and females are socialized into their respective gender roles. According to Costello and
Mederer (2003), males and females have the same innate tendency to pursue self-
interests, which suggests, something is causing the differences in the male and female
crime rates. Therefore, it might be ascertained that the self-interest of females is carried
out in different ways than that of males based on differences in socialization. So the
question may be, if women do commit fewer crimes than men, what are the motivations
for the ones who commit those crimes (Rumgay, 2004). Zeitz (1981) did a study on male

and female white collar offenders. The major difference was women were sorry for their

actions substantially more ofien than their male counterparts. This again brings up the
interesting question of whether or not women’s and men’s motivations are similar or very
different (Costello and Mederer, 2003). In regard to social bonding and gender,
Rosenbaum (1987) examined gender, social bonding, and delinquency and found better
explanatory power for social bonding theory with females versus males. Specifically,
attachment was the best predictor for both males and females but attachment showed
greater predictive validity for the females (Rosenbaum, 1987).

According to Rosenbaum (1987), this finding is not surprising because females
are required to conform to a greater extent than males so it would seem understandable
that a theory that measures one's social bonds would be more explanatory for females
than for males (Rosenbaum, 1987). For example, females in general seem to have a
stronger attachment to their families and maintain this attachment longer than do males
(Rosenbaum, 1987).

Another example of differences in gender socialization is that at a young age,
what is socially acceptable for females is not socially acceptable for males and vice a
versa. For example, males are taught to not play with female toy but it is often acceptable
for females to play with male toys. Also, as males get older they often attain more
freedom than do females.

Costello and Mederer (2003) assessed gender differences in crime and
delinquency by using control theory. Costello and Mederer (2003) believe that the focus
of studies should be on the similarities between the types of crimes committed by females
and males. These authors also believe that one of the reasons that females commit less

crime is due to their socialization (Costello and Mederer, 2003). Yet, other scholars

10

believe that women who commit crimes are taking on male behaviors (Akers and Sellers,
2004). Some scholars claim that it is important to look at each individual as well as the
group, in order to fully comprehend the motivations, beliefs, and propensities for
offending (Daly and Chesney-Lind, 1988).

Costello and Mederer (2003) argue that, in order to lessen male crime the existing
method of male socialization must change; however, they also claim that cannot be
accomplished without an understanding of the reasons why women do and do not commit
crimes. That is why it is important to look at the occurrences of female criminality and
their propensity to commit these acts (Daly and Chesney-Lind, 1988). Although various
theories have been proposed, there is as yet no dominant theory for gender and
criminality.

Based on all of the above literature on gender and social bonding, it seems
important to further examine the relationships between gender and delinquency.
Therefore in this thesis, gender will be examined as a moderator of social bonding and
delinquency.

THESIS HYPOTHESES

Based on the above literature review of social bonding theory, five hypotheses
are proposed and will be tested meta-analytically. First, several studies indicate that
attachment is the higher predictor of delinquency, relative to the other three social
bonding variables (Costello and Vowell, 1999; Rankin and Kern, 1994; Rankin and
Wells, 1990; Agnew, 1993; Kempf, 1993; Rosenbaum, 1987; Wiatrowskieta1., 1981).
Therefore,

Hypothesis 1: The magnitude of the true score (meta-analyzed) validity for

11

predicting juvenile delinquency from attachment will be larger than the

predictive validity for the other three social bonding variables

(commitment, involvement, and belief).

Second, many other studies show that commitment also strongly, but to a lesser
extent than attachment, predicts juvenile delinquency (Agnew, 1993; Agnew, 1991;
Kempf, 1993; Junger-Tas, 1992; Wiatrowski et al., 1981; Krohn and Massey, 1980),
therefore,

Hypothesis 2: Afler attachment, the magnitude of the true score

validity for the prediction of delinquency from commitment will be larger

than the predictive validity for the involvement or belief variables.

Third and fourth, considerable research shows mixed results on the relative
extent to which involvement and belief predict delinquency, and some studies find no
significant effects at all. However, some studies indicate that the belief variable may be
more strongly related to delinquency than does involvement (Costello and Vowell, 1999;
Junger and Marshall, 1997;Wiatrowski etal., 1981; Kempf, 1993;Agnew, 1993; Kempf,
1 993; Hirschi, 1969). Therefore,

Hypothesis 3: After attachment and commitment, the magnitude
of the true score validity for belief and the prediction of delinquency will
be greater than for involvement.

Hypothesis 4: The magnitude of the true score validity for
involvement and the prediction of delinquency will be smaller than the
other three social bonding variables

Last, based on the gender differences and social bonding theory studies by

12

Rosenbaum (1987) and Krohn and Massey, (1980),
Hypothesis 5: Attachment, commitment, belief and involvement

are stronger predictors of delinquency for females versus males.

METHOD
The Sample

The following procedures were used to select the appropriate studies for
inclusion in this meta-analysis. First, a search of the literature was conducted to locate
studies that report correlations between each of the four social bonding variables--
attachment, commitment, belief and involvement-~and juvenile delinquency, for both
males and females. To locate these studies, a comprehensive search of the following
computer databases was completed: The Criminal Justice Abstracts, ProQuest, First
Search, J STOR, Social Sciences Abstracts, Sociology Abstracts and the National
Criminal Justice Research Service. The literature search was performed using the
following keywords: attachment, commitment, belief, involvement, delinquency, gender,
males, females, and social bonding.

Second, a search of the literature was also conducted of the following journals:
Criminology, Justice Quarterly, Journal of Research in Crime and Delinquency,
Quantitative Criminology, American Sociological Review and Sociological Quarterly.

Finally, the reference lists of all the retrieved studies and books concerned with
social bonding theory were searched in an effort to find additional studies that may have
not been in any of the above computer databases or journals. This "snowballing" method

turned up additional studies including theses, dissertations and conference presentations.

13

The final search generated 74 studies for possible inclusion in the meta-analysis.
However, three rules were established to identify data that were appropriate for this meta-
analysis:

(1) The study had to have delinquency or a type of delinquency as the
dependent variable;

(2) One or all the social bonding variables had to be present (attachment,

commitment, involvement, belief) and;

(3) The indicators for the independent variables must be consistent with
Hirschi’s original study and clear in which of the four variables they are
attempting to measure.

In addition, exclusionary criteria included

(1) studies that reported regression statistics with no other statistics

available to calculate correlations, and

(2) studies that did not have adequate data to compute the statistics needed

for a meta-analysis.

As a result of these selection criteria, the original 74 studies were reduced to 26
studies. These 26 were further reduced due to the lack of adequate statistics from which
to compute correlations.

In an effort to increase the number of studies in the analysis, several authors
(Agnew, 1985, 1991 and 1993; Freidman and Rosenbaum 1988; Costello 1999 and
Rankin and Wells 1990), were contacted and asked for the data necessary to compute
correlations. Only Robert Agnew complied with the request. The final count was 20

studies that provided the necessary data for the meta-analysis. Most of the data came

14

from journals and two from dissertations. The articles ranged in date from 1976-2002.

Data Coding

The four independent variables (attachment, commitment, involvement and
belief) were coded based on correlations available for each of them. Gender correlations
were also reported when available. The reliability coefficients were also coded when they
were reported from measures of delinquency, attachment, commitment, involvement and
belief.

Three of the 20 articles misrepresented Hirschi's original variables. That is, in
those studies the indicators were misaligned with the variables there were supposed to
represent. For example, Gardner and Shoemaker (1989) examined three social bonding
variables: belief, attachment, and commitment. These authors, however, measured
commitment using indicators that, according to Hirschi's theory, intend to measure
involvement. In this case (and in the other two), because this thesis is testing Hirschi's
theory, the indicators were coded as measures of the variable Hirschi intended each to
represent: in this example, involvement. The final database included a table of all

correlations for each study and the variables; attachment, commitment, involvement, and

belief.

Statistical Analysis
Using a database composed of the correlations from 20 studies , the Hunter and
Schmidt (1987, 2004) meta-analysis program was used to correct those correlations for

sampling error due to differences in the studies' sample sizes. The meta-analysis method

15

also corrects for measurement error due to differences in reliabilities of scales used in
those studies to measure each of the four predictors, that is, the four social bonding
variables(attachment, commitment, involvement, and belief) and juvenile delinquency.

When corrected for sampling and measurement errors, it is possible to estimate
the "true score" validity--a predictor coefficient for which the statistical variance is free
of sampling and measurement errors. In this way, even correlations reported in studies as
non-significant may, in fact, be significant afier all, to the extent the scales used in the
study lack perfect reliability and the sample sizes are small.

The purpose of the moderator analysis is if the correlations in the meta-analysis
explain a small amount of variance then it can be deduced that something moderates
those relationships. A 80% credibility interval was used to observe where the true score
correlation lay within that distribution (Hunter and Schmidt, 1990). If the credibility
interval is large or includes zero, the correlations between social bonding and

delinquency is being moderated by some other variable or variables (Hunter and Schmidt,

1990)

RESULTS
The results are reported in Table 1. The table includes first the overall social
bonding model followed by a moderator analysis by gender. Second, the subset analysis
shows the meta-analysis for the four variables separately (attachment, commitment,
involvement and belief) and also by gender, for attachment and belief. Gender was not
used as a moderator for commitment or involvement as there were inadequate statistics

for these two variables.

16

The meta-analysis of the overall social bonding model was conducted using a
sample size of 29,476. The true score correlation was -.23 and SD =.l7. The 80%
credibility interval is wide and is very close to zero (-.46 to -.01), which suggests the
presence of moderators. Another indication of moderators was the negligible 4.79%
variance explained. Since there is such a strong indication of moderators, gender was
then tested to determine if the variance explained would be higher for either males or
females.

The true score correlation for males was -.15 and for females -.22 indicating a
stronger relationship between social bonding and delinquency for females versus males.
The interesting part was that the explained variance for males was 20.74% and for
females was 26.26% a substantial jump in magnitude from the 4.79% variance explained
of the overall model. However, the explained variance by gender is still low so a subset
analysis was conducted to explore the relationships between each of the social bonding
variables and delinquency.

Afier correcting for sampling error and measurement errors, the true score
correlations were; -.24(attachment), -.25(commitment), -.20(involvement), and
-.33(belief). The strongest correlation was for belief and delinquency but the highest
explained variance was for attachment. All the explained variance percentages, however,
remain relatively low, as in the overall social bonding model.

The last two meta-analysis models were gender and attachment and gender and
belief. For attachment, the correlation for females was -.19 and for males it was -.08. For
belief the correlation for females was -.22 and for males it was -.23. The explained

variance for the correlation between attachment and delinquency was slightly higher for

males (11.8%) than for females (7.4%) the differences are not large. After all the analyses
were completed, considerable unexplained variance remained. The lack of explained
variance means that some other factor or factors are involved in the prediction of juvenile
delinquency from social bonding theory.

DISCUSSION

This discussion will follow the order of variables as listed in Table 1. First,
however, I wish to address the issue of longitudinal versus cross-sectional research on
social bonding theory and delinquency. The social bond literature is sparse when it comes
to longitudinal studies: the review generated only three studies, two of those studies
contained correlations and the third contained no statistics from which to compute
correlations. It was, therefore, not possible to conduct a moderated meta-analysis to
compare the two different types of studies (longitudinal versus cross-sectional), with only
two correlations for the longitudinal studies.

Insofar as the overall socialpbonding model did not reveal strong correlations with
delinquency. There was considerable unexplained variance for the correlation between
social bonding and delinquency, which called for moderated meta-analysis to determine
whether other (moderating) factors might explain some of that variance.

The relationship between gender and social bonding was the most supported
hypothesis in this study, but there remained unanswered questions. Although the
explained variance did increase when using gender as a moderator, the increase was
modest and did not meet the 75% criterion that would lend support to the social bonding
model.

The above findings for the model overall and with gender as a moderator

l8

indicates the need for further research on social bonding. For example, gender would be a
good topic to examine further with regard to the attachment variable. The correlation for
attachment was stronger for females than for males, a finding that is consistent with the
previous literature on social bonding theory and other studies of gender and delinquency
(Rosenbaum, 1987).

That is, considerable literature on gender and delinquency focuses on
socialization through parents, peers and teachers, which also are the indicators for
attachment in social bonding theory. Attachment, therefore, might possibly be the best
predictor for female delinquency and social bonding, aside from the other three social
bonding variables (belief, commitment, and involvement).

Studies that would focus specifically on females and attachment are
recommended because many problems exist with this current literature. For example,
studies on females and social bonding almost always include males, and all four variables
of social bonding-«attachment, commitment, involvement and belief, but these studies are
few. Also, the majority of the studies are male samples. Females need to be included as
samples in more research on social bonding.

In the present thesis, only four studies were available that included correlations
for both males and females. However, using four studies capitalizes on chance; that is,
although the aggregated sample §i_z_e is large the number of correlations in the meta-
analysis is small. These results therefore must be considered speculative until further
female-social bonding research is available.

For attachment the true score correlation with social bonding was about the same

 

as for the overall model, and the moderator subset analysis revealed more explained

19

variance than did the variance for other three variables (commitment, involvement and
belief). However, the variance explained by attachment was only slightly larger in
magnitude (relative to the overall model), and this value failed to meet the 75% criterion
for ruling out moderators. Thus, gender was examined as a moderator.

As in the overall social bonding model, the explained variance increased when
controlling for gender. There remained nonetheless a large amount of unexplained
variance which maybe due to at least two factors: (1) other moderators, such as socio-
economic status, ethnicity, urban/rural residence; would modify the size of the correlation
and the amount of variance explained, and/or the variable called attachment lacks
construct validity. Construct validity means that a variable measures
all of the characteristics of a concept (e. g., attachment) and nothing else; for example,
attachment would be construct valid if there were evidence to show that the items that
measure attachment are all inclusive and exclude others that are unrelated to attachment.

The meta-analysis for commitment and involvement revealed results similar (in
correlation magnitude) to the overall model and attachment. However, again, for both
variables, the considerable unexplained variance in the distribution of correlations
indicates a moderator effect operating. Unfortunately, the studies that included the
commitment and involvement variables did not contain data for males and females. Thus,
no gender moderator analysis could be conducted.

The highest true score correlation for social bonding and delinquency was
between b_el_ie_f and juvenile delinquency, but not all of the indicators for belief were
included in all of the studies. For example, some studies included religion as an indicator

of belief and others included morals as an indicator, but not all studies included both

20

measures. If both measures, or indicators, are included in future research, the correlation
between social bonding (as measured by belief) and delinquency may be larger in
magnitude than in this study. The results showed the possibility of moderators, which
were conducted, by gender.

Once again, the true score correlations for gender (and a moderator of belief) are
similar in size as for the overall model, attachment, commitment, involvement, and
belief—only the variance for the distribution of correlations is slightly larger in
magnitude. What this means is that yet other factors are moderating the correlation; that
is, if one knew of the other factors, other moderator analysis could be performed that
would explain the large unexplained variance. As well, the issue of construct validity
may be involved, and this will be addressed further below.

In general, the problems that plague the social bonding studies are in their lack of
consistency: the studies claim to examine social bonding theory but the studies use
different indicators for different social bonding variables. Until this thesis, the research
has not addressed this issue of inconsistencies across studies.

Also, the validity of measures (indicators) used to measure the constructs—the
four social bonding variables, is in question. There are no studies that test the construct
validity using statistical procedures, such as structural equation modeling, that would
reveal whether or not the variables represent what they purport to—attachment,
commitment, involvement, or belief.

Another potential problem is the social bonding may include other variables that
were not proposed in Hirschi’s theory and that are therefore not explored in current

research. So perhaps, the theory is subject to modification, should other criminal justice

21

theories point to such variables. In other words, although social bonding itself may be a
well-grounded theory, this theory may be incomplete—which is why further research is
needed.

Yet another possible explanation for the relatively small true score correlations
and unexplained variance may be that other moderators are involved. For example, few
studies on social bonding include measures of socioeconomic status, race, and region.
Demographic variables could explain differences in how people bond and are therefore
become subject to acts of delinquency. These and other variables could be examined
further in primary (versus secondary meta-analysis) studies. First, however, the construct
validity of the social bonding variables must be established because this is where the
inconsistencies and conflicting findings of studies may reside.

CONCLUSION

The results of this meta-analysis showed that the correlations and
explained variance between delinquency and social bonding theory is minimal. That is to
say that the relationship between the two is weak, even when gender was used as a
moderator. Corrections for sampling error and measurement error increased the size of
the correlation for the overall model and also for the subset models, however, the
correlations still were marginal. Evidence from these results indicates that the correlation
between social bonding and delinquency is moderated by variables other than, but
including, gender. Future research is needed to disambiguate the social bonding theory. A
final recommendation is that all future studies would report descriptive statistics to
provide data for meta-analyses that can reveal the true score correlation for social

bonding and delinquency.

22

Table l. Meta-Analysis of Correlations Between Social Bonding & Juvenile Delinquency

 

 

 

80% Credibility % Var.
Model K N 5 SD, p SDp Interval Explained
Overall 20 29,476 -.17 .13 -.23 .18 -.46 to -.01 4.79
Males 4 2,059 -.11 .09 -.15 .11 -.29 to -.01 20.74
Females 4 1,675 -.17 .08 -.22 .1 l -.36 to -.09 26.26
Subset Analysis
Attachment 20 29,476 -.18 .13 -.24 .17 -.45 to -.028 4.87
Males 4 2,059 -.06 .12 -.08 .16 -.28 to .11 .1 1.8
Females 4 1,675 -.15 .17 -.19 .22 l -.48 to .09 5.44
Commitment 17 27,218 -.19 .18 -.25 .23 -.54 to .04 2.38
Involvement 11 16,405 -.14 .17 -.20 .24 -.51 to .11 2.72
Belief 17 26,957 -.25 .21 -.33 .27 -.68 to .02 2.41
Males 4 2,059 -.20 .18 -.23 .20 -.49 to .03 5.07
Females 4 1,675 -.20 .19 -.22 .21 -.49 to .04 6.04

 

Note: K = number of correlations; N = number of subjects; r = sample size weighted

mean observed correlation; SDr = sample size weighted observed standard deviation; p =

true score correlation; SDp = standard deviation of the true correlation; 80% Credibility

Interval = p = +/- 1.28*SD; % Variance Explained = percent of variance in the observed

correlations attributable to sampling error and measurement error.

23

References

Agnew, R. (1993). Why Do They Do It? An Examination of the Intervening Mechanisms
Between "Social Control" Variables and Delinquency. Journal of Research in
Crime and Delinquency, 30(3), 245-266.

*Agnew, R. (1991). The Longitudinal Test of Social Control Theory and Delinquency.
Journal of Research in Crime & Delinquency, 28(2), 126-156.

*Agnew, R. (1985). Social Control Theory & Delinquency: A Longitudinal Test.
Criminology, 23(1), 47-61.

Agnew, R. (2003). The Interactive Effects of Social Control Variables on Delinquency.
In C. L. Britt, Gottfredson, M.R. (Ed.), Control Theories of Crime and
Delinquency (V 01. 12, pp. 53-76). New Bnmswick: Transaction Publishers.

Akers, R. L. Sellers, C. (2004). Criminological Theories Introduction, Evaluation, and
Application (Fourth ed.). Los Angeles: Roxbury Publishing Company.

*Aultman, MG. (1979) Delinquency Causation: A Typological Comparison of Path
Models. The Journal of Criminal Law & Criminology, 70(2), 152-163.

*Barton, W.H., Figueira-McDonough J. (1985). Attachments, Gender and Delinquency.
Deviant Behavior, 6, 119-144.

*Bemburg, J .G., Thorlindsson, T. (1999). Adolescent Violence, Social Control, and the
Subculture of Delinquency Factors Related to Violent Behavior and NonViolent
Delinquency. Youth & Society, 30(4), 445-460.

*Bishop, D.M. (1982). Deterrence and Social Control: A Longitudinal Study of the
Effects of Sanctioning and Social Bonding on the Prevention of Delinquency.
Unpublished doctoral dissertation, University of New York, Albany.

Chesney-Lind, M. (1997). The Female Ofi’endenGirls Women and Crime. Thousand
Oaks, CA:Sage Publications.

Conger, RD. (1976). Social Control and Social Learning Models of Delinquent
Behavior; A Synthesis. Criminology, 14(1), 17-40.

Costello, B. J ., Mederer, H. J. (2003). A Control Theory of Gender Difference in Crime
and Delinquency. In C. L. Britt, Gottfredson, M.R. (Ed.), Control Theories of

Crime and Delinquency (Vol. 12, pp. 77-107). New Brunswick: Transaction
Publisher.

Costello, B.J., Vowell, PR. (1999). Testing Control Theory and Differential Association:

24

A Reanalysis of the Richmond Youth Project Data. Criminology, 37(4), 815-842.

Daly, K. (1989). Gender and Varieties of White- Collar Crime. Criminology, 27(4), 769-
794.

Daly, K., Chesney-Lind, M. (1988). Feminism And Criminology. Justice Quarterly, 5(4),
497-538.

Ellis, L. Walsh, A. (1999). Criminologists Opinions About Causes and Theories of Crime
and Delinquency, The Criminologist, 24 (4).

Freidman, J. Rosenbaum, DP. (1988). Social Control Theory: The Salience of
Components by Age, Gender, and Type of Crime. Journal of Quantitative
Criminology, 4 (4), 363-381.

*Gardner, R.L. (1984). Social Bonding and Delinquency: A Multivariate Analysis.
Unpublished doctoral dissertation, Virginia Polytechnic Institute and State
University, Blackburg.

Gardner, L., Shoemaker, DJ. (1989). Social Bonding and Delinquency: A Comparative
Analysis. The Sociological Quarterly, 30(3), 481-500.

Greenberg, DP. (1999). The Weak Strength of Social Control Theory. Crime &
Delinquency, 45(1), 66—81.

Heimer, K. (1997). Socioeconomic Status, Subcultural Defintions and Violent
Delinquency. Social Forces, 75, 799-833.

*Hepbum, J .R. (1976). Testing Alternative Models of Delinquency Causation. The
Journal of Criminal Law & Criminology, 67(4), 450-460.

Hirschi, T. (1969). Causes of Delinquency. Berkeley: University of California Press.
*Huebner, A.J., Betts, SC. (2002). Exploring the Utility of Social Control Theory for
Youth Development Issues of Attachment, Involvement, and Gender. Youth &

Society, 34(2), 123-145.

Hunter, J .E., Schmidt, FL. (1990). Methods of Meta-Analysis. Newbury Park, CA: Sage
Publications.

Jensen, G.F. & Brownfield, D. (1983). Parents and Drugs. Specifying the Consequences
of Attachment. Criminology, 21, 543-554.

Jensen, G.F. & Eve, R. (1976). Sex Differences in Delinquency: An Examination of
Popular Sociological Definitions. Criminology, 13, 427-448.

25

*Junger, M., Marshall I.H. (1997). The Interethnic Generalizability of Social Control
Theory: An Empirical Test. Journal of Research in Crime and Delinquency,
34(1), 79-112.

J unger-Tas., J. (1992). An Empirical Test of Social Control Theory. Journal of
Quantitative Criminology, 8(1), 9-28.

Kempf, K. L., (1993). The Empirical Status of Hirschi's Control Theory. In Adler, Laufer
(Ed), New Directions in Criminological Theory. Advances in Criminological
Theory(Vol. 4, pp.143-185). New Brunswick: Transactions Publisher.

*Kempf-Leonard, K., Decker, S.H.(1994). The Theory of Social Control: Does it Apply
to the Very Young? Journal of Criminal Justice, 22(2), 89-105.

*Krohn, M.D., Lanza-Kaduce, L., Akers, R. L. (1984). Community Context and Theories
of Deviant Behavior: An Examination of Social Learning and Social Bonding
Theories. The Sociological Quarterly, 25, 353-371.

*Krohn, M.D., Massey, IL. (1980). Social Control and Delinquent Behavior: An
Examination of the Elements of the Social Bond. Sociological Quarterly, 529-
544.

Lasley, J. R. (1988). Toward a Control Theory of White-Collar Offending. Journal of
Quantitative Criminology, 4(4), 347-362.

*Linden, R. (1978). Myths of Middle Class Delinquency: A Test of the Generalizability
of Social Control Theory. Youth & Society, 9(4), 407-432.

LeBanc, M. and Caplan, A. (1993). Theoretical Formalization, a Necessity: The Example
of Social Bonding Theory. In Adler, Laufer (Ed.), New Directions in
Criminological Theory. Advances in Criminological Theory(V 01. 4, pp. 23 7-336).
New Brunswick: Transactions Publisher.

*Marcos, A.C., Bahr, S.J., Johnson, RE. (1986). Test of a Bonding/Association Theory
of Adolescent Drug Use. Social Forces, 65(1), 135-161.

Matsueda, R.L. (1982). Testing Control Theory and Differential Causation: A Causal
Modeling Approach. American Sociological Review, 47, 489-504.

Patterson, G.R. & Dishion, T.J. (1985). Contribution of Families and Peers to
Delinquency. Criminology, 23, 63-80.

Poole, E. D. & Regoli, RM. (1979). Parental Support, Delinquent Friends and

Delinquency: A Test of Interaction Effects. Journal of Criminal Law and
Criminology, 70(2), 1 88- 193.

26

*Rankin, J.H., Kern, R. (1994). Parental Attachments and Delinquency. Criminology, 32,
495-515.

Rankin, J .H., Wells, LE. (1990). The Effect of Parental Attachments and Direct Controls
on Delinquency. Journal of Research in Crime and Delinquency, 27, 140-165.

*Rosenbaum, J .L., (1987) Social Control, Gender, and Delinquency: An Analysis of
Drug, Property and Violent Offenders. Justice Quarterly, 4(1), 117-132.

Rumgay, J. (2004). Scripts for Safer Survival: Pathway Out of Female Crime. The
Howard Journal of Criminal Justice, 43(4), 405-419.

Thaxton, S., Agnew, R. (2004). The Nonlinear Effects of Parental and Teacher
Attachment on Delinquency: Disentangling Strain from Social Control
Explanations. Justice Quarterly, 21(4), 763-791.

*Thompson, W.E., Mitchell, J ., Dodder, RA. (1984). An Empirical Test of Hirschi's
Social Control Theory of Delinquency. Deviant Behavior, 5, 11-22.

Tubre, T.C., Collins, J .M. (2000). Jackson and Schuler (1985) Revisited: A Meta-
Analysis of the Relationships between Role Ambiguity, Role Conflict, and Job
Performance. Journal of Management, 26(1), 155-169.

Wiatrowski, M.D. , Anderson, KL. (1987). The Dimensionality of the Social Bond.
Journal of Quantitative Criminology, 3(1), 65-81.
*Wiatrowski.,M.D., Griswold, D.B., & Roberts, MK. (1981). Social Control Theory and

Delinquency. American Sociological Review ,46, 525-541.

Zietz, D. (1981). Women Who Embezzle or Defraud: A Study of Convicted Felons. New
York: Praeger Publishers

27

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

llllfllllllllllllllllllllllllill/171111