DUE: 1/5/2017 @ 1 pm est.

161THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY VOLUME 40 NUMBER 2 2007 PP. 161–178

Address for correspondence: Ibolya Losoncz, Senior Research Analyst, Research and Analysis Branch, Department of Families, Community Services and Indigenous Affairs, PO Box 7788, Canberra Mail Centre ACT 2610. E-mail: [email protected]

Parental Shaming and Adolescent Delinquency: A Partial Test of Reintegrative Shaming Theory Ibolya Losoncz1 and Graham Tyson Charles Sturt University, Australia

The past decade has seen an increase in the application of Braithwaite’sreintegrative shaming theory as a framework for restorative justice programs. However, to date the theory has received little empirical attention. The current study set out to contribute to the empirical testing of the theory by exploring the appropriateness of the causal model put forward by Braithwaite. One-hundred-and-seventy Year 9 and Year 10 high school students from 2 government high schools in the Australian Capital Territory completed a survey capturing projected delinquency, delinquent peers and family processes. Principal component analysis found an overlap between aspects of shaming with reintegration and stigmatisation. Furthermore, not all facets of reintegration and stigmatisation were found to be discrete concepts. Results from subsequent structural equation modelling were largely supportive of RST, particularly the theory’s emphasis on the harmful effects of stigmatisation and the beneficial effects of reintegration. However, shaming, as defined in the theory, may not affect predatory crime in the way it is predicted by RST.

Reintegrative shaming theory (RST) was first put forward by Braithwaite in the late 1980s. The basic idea of RST is that societies, communities and families where shame is communicated effectively and reintegratively are less likely to experience predatory crime than places where shame is communicated in a stigmatising approach, or not communicated at all. The key concept in the theory is shaming, partitioned into two types: stigmatising and reintegrative. Stigmatising shaming treats the wrongdoer with disrespect, as a bad person and as an outcast. Reintegrative shaming, on the other hand, recognises that the wrongdoer is a good person who has done a bad act; and although the wrongful act will be confronted, after the confrontation special efforts will be made to show the wrongdoer that he/she is still valued and respected. The theory argues that reintegrative shaming reduces crime, while stigmatising shaming increases it (Braithwaite, 1989). While the theory enjoys much attention and acclaim, to date it has received limited

empirical testing. The aim of the current research is to contribute to the experimen- tal testing of the major constructs and propositions of RST by first developing adequate quantitative measures of key concepts in the theory, and then exploring the adequacy of the causal model put forward by Braithwaite.

In his theory, Braithwaite drew heavily on preexisting theories of control, subculture, opportunity, and labelling theory, integrated into a learning theory framework. According to the labelling perspective on delinquency, negative actions or reactions by other people affect the beliefs and feelings individuals develop about themselves (Erikson, 1962; Taylor, Walton, & Young, 1973). However, critics of labelling theory argue that any such statement needs to be refined by identifying alternative outcomes of labelling and specifying the conditions under which each is likely to occur (Grimes & Turk, 1978).

Braithwaite (1989) suggested that his theory refined and clarified labelling theory by partitioning stigmatising or labelling from reintegrative shaming. He argued that while labelling causes offenders to become outcasts, or to view themselves as outcasts, and adopt a deviant identity, RST offers an alternative outcome, as long as it is the act being labelled and not the person. Following the labelling of the act as bad, and holding offenders responsible for their law breaking, offenders are forgiven and accepted back into their family and community.

Braithwaite (1989) identifies several conditions that facilitate reintegration, the most important of which are a communitarian society and a strong family system characterised by a sense of interdependency. That is, individuals who are bonded to conventional society, or their family, are easy to shame back into conforming behaviour. Identifying gender as a predictor of interdependency is another signifi- cant element of the theory, as it accommodates the explanation of the relatively low predatory offence rates among females. Braithwaite (1989) proposes that females experience more attachment to their families, as they are more socialised into being interdependent through the caretaking of others, as well as more likely to be encouraged to fit back into the family. Therefore, they are more susceptible to shaming, particularly reintegrative shaming.

The apparent popularity and growing support for reintegrative shaming-based restorative justice models in Australia, New Zealand and North America calls for the evaluation of RST in the areas of procedural outcomes and research evidence for the model. In Australia, research on procedural outcomes of conferencing based on RST has largely focused on participants’ perceptions of the fairness of the process and their satisfaction with the process and outcome. The most consistent finding, to date, is that conferences are perceived as fair and participants are satisfied with the process and outcomes (Strang, Barnes, Braithwaite, & Sherman, 1999). Concerning the effects of restorative justice conferences on recidivism, a recent meta-analytic review of the perti- nent literature by Bonta, Jesseman, Rugge and Cormier (2006) found that on average restorative justice interventions are associated with reduction in recidivism. While the effects are small, they were found to be significant, particularly for low risk offenders and for programs operating in a noncoercive environment. As for empirical support, the most important criterion by which to judge a theory (Akers, 2000), the number of studies which have used direct quantitative measures of the key concepts in the theory or have investigated the predicted causal effects of reintegration and shaming on criminal behaviour are, to date, rather limited (Ahmed & Braithwaite,

162

IBOLYA LOSONCZ AND GRAHAM TYSON

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

2004; Bennett, 1996; Harris, 2001; Hay, 2001; Makkai & Braithwaite, 1994; Zhang, 1995; Zhang & Zhang, 2004 ).

Researchers who have set out to test RST are confronted by two tasks: the opera- tionalisation of key concepts in the theory and testing the causal structures proposed by the theory. Most of the studies faced serious limitations in their attempt to adequately operationalise the key concepts in RST. This was largely attributed to the use of exist- ing datasets, most of which did not contain variables that sufficiently captured the main concepts in RST. However, there were two important studies that specifically investi- gated Braithwaite’s assertion (1989) of four distinct facets of shaming and reintegration (i.e., low shaming–low reintegration, low shaming–high reintegration, high shaming–low reintegration, and high shaming–high reintegration). The first study, undertaken by Makkai and Braithwaite (1994), investigated the compliance with governmental regulatory standards in nursing homes. Specifically, the study examined changes in compliance level of nursing homes following the visits of government inspectors identified as either supportive of reintegrative shaming or characterised by stigmatising attitudes. Consistent with the prediction of the theory, the four facets of shaming and reintegration formed separate dimensions.

However, a more recent research by Nathan Harris (2001) provided contradic- tory results. In his study Harris (2001) factor analysed data collected from 900 drink-driving cases randomly assigned to a restorative justice conference or to a court. The study used observational and self-report measures to assess the relation- ship between shaming, reintegration and stigmatisation. The prediction that the four facets of shaming and reintegration would form separate dimensions was not supported. Instead, shaming was found to be independent from both reintegration and stigmatisation. Furthermore, the study found reintegration and stigmatisation to be independent concepts rather than opposite poles of a continuum.

Research studies that examined the major causal structures of RST reported conflicting support for the key hypotheses of RST. The study by Makkai and Braithwaite (1994) found strong support for all predictions of RST. That is, high levels of shaming combined with high levels of reintegration by inspectors resulted in a greater increase in compliance by nursing homes than high levels of shaming that were stigmatising, or low levels of shaming that were reintegrative. The key concept of interdependency, measured by the presence or absence of existing meetings between the director of the home and the inspector prior to the regulatory encounter, was also found to have a strong positive effect on compliance.

However, the study by Bennett (1996), which set out to partially assess RST at the individual level using the survey data from the Rochester Youth Development Study (RYDS),2 found only limited support. While Bennett established a positive correlation between family interdependency and family reintegration, reintegrative shaming was not found to be a significant determinant of the frequency of offending. In fact, the only significant determinants of frequency of repeated delinquency were stigmatising shaming and delinquent peers. This finding on the effect of delinquent peers on delin- quency is consistent with the criminology literature, which shows that the number of delinquent friends is one of the strongest predictor of self-reported delinquency (Agnew, 1995; Hirschi, 1969; Matsueda ,1982; Warr & Stafford, 1991).

Similarly, a micro-level exploratory research by Hay (2001) found only limited support for the predictions of RST. While analysis of the survey data, collected from

163

PARENTAL SHAMING AND ADOLESCENT DELINQUENCY

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

adolescents on their parents’ use of reintegration and shaming and their reports of projected predatory delinquency, found a significant negative correlation between reintegration and delinquency, and between shaming and delinquency, the study found no interactive effect of shaming and reintegration on projected delinquency. Further, contrary to the prediction of RST, the reintegration effect disappeared after controlling for interdependency. In fact, multiple regression analysis of the variables suggested that the effect of reintegration is spurious and may only be a byproduct of parent–child interdependency. An explanation proposed by Hay (2001) is that it is reintegrative sanctioning that will lead to high parent–child interdependency rather than vice versa.

The most recent study of RST by Zhang & Zhang (2004), using data from the National Youth Survey (NYS),3 found support for only some of the elements of Braithwaite’s theory. While parental forgiveness and peer shaming had a significant negative effect on the likelihood of predatory offending, the study found no effect of either parental or peer reintegrative shaming (measured as a product term of reinte- gration and shaming) on predatory delinquency when controlling for the effects of other important predictor variables, such as peer delinquency, gender, age, race and socioeconomic status.

The relationship between gender and shaming and reintegration was also addressed in the study by Bennett (1996). However, contrary to the prediction of RST, the study found no support for the hypothesis that females are more likely to experience family shaming and reintegration than males. In fact, males had higher interdependency scores than females. However, the sample comprised only those juveniles who had self-reported predatory offences, and therefore was not represen- tative of the whole juvenile population.

In summary, empirical research has found limited support for Braithwaite’s RST. This lack of strong support for the theory could be partly due to the conceptually weak measures of the major concepts of RST in some of the studies. To some extent this relates back to the lack of clear definitions of key concepts in the theory (Gibbons, 1991) and the challenges of measuring some of the nuanced concepts in the theory. Another weakness of the initial RST identified by Braithwaite himself is the failure to concentrate on the emotion of shame (Braithwaite & Braithwaite, 2001).

To translate Braithwaite’s comprehensive theoretical model into a testable research model is a challenging task. Braithwaite did not present empirical testing of RST in his book (Braithwaite, 1989), but he did include a chapter on ways of testing his theory. In this he identified ethnographic work as the most appropriate methodology to evaluate macro and micro-level variables and their complex inter- action, and the use of survey research for measuring the effects of individual-level variables. However, at this point it is more achievable to test the model partially. Therefore, the current study sets out to test the causal model put forward in RST (Braithwaite, 1989) in the realm of the family, concentrating only on the individual level variables using survey methodology.

Because of the newness of this theory and the limited empirical testing it has received to date, there are still questions as to how its variables should be measured and how the interactions between variables should be tested. In this research, in line with the revised recommendation of Braithwaite (Braithwaite & Braithwaite, 2001), shaming, reintegration and stigmatisation will be treated as three indepen- dent concepts. While some studies on this topic have investigated the interaction of

164

IBOLYA LOSONCZ AND GRAHAM TYSON

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

shaming with reintegration and stigmatisation, this article will test the causal linear model of RST as it was proposed in Braithwaite, 1989, p. 99, and later reviewed by Braithwaite and Braithwaite, 2001, with no interaction between shaming and reintegra- tion or shaming and stigmatisation. According to the theory, parental shaming, reinte- gration, stigmatisation and delinquent peers would all have a main effect on projected predatory delinquency of adolescents. Parental shaming and reintegration would have a negative effect, whereas parental stigmatisation and delinquent friends would have a positive effect. Furthermore, the impact of interdependency would be mediated by shaming, reintegration, stigmatisation and delinquent peers. These relationships are shown in Figure 1.

Data and Method Participants were obtained from two government high schools in the Australian Capital Territory. Positive consent was sought from parents of all Year 9 and 10 students at the schools. The questionnaire, which took approximately 20 minutes to complete, was administered by nominated teachers in class to those students whose parents had given written permission for their child to participate. A total of 169 usable questionnaires was obtained. The exact response rate could not be determined as the information sheets seeking active parental consent were given out by nominated teachers rather than the researchers. However, our estimated return rate is around 40%. The final sample consisted of 96 (56.8%) males and 73 (43.2%) females ranging in age from 14 to 17 years (M =15.28 years, SD = .60). The majority of students (65.0%) were from intact families. The collected demographic data, shown in Table 1, indicated that the partici- pating youths and their families were largely representative of Australian families on the measures of cultural background, marital status and educational attainment.

165

PARENTAL SHAMING AND ADOLESCENT DELINQUENCY

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

FIGURE 1

Model 1: Family model of reintegrative shaming.

Delinquency Scale The current study only measured projected involvement in crime by respondents. The authors would have liked to include a measure of childhood antisocial behaviour, to control for its affect on parental shaming practice and subsequent delinquency involve- ment, and to avoid causal order problems that often arise in cross-sectional delinquency research. However, the inclusion of such measures would have jeopardised parental as well as principal approval to administer the survey to students. To measure projected delinquency, modified items from the Self-Report Delinquency Scale for Australian Adolescents, which was developed by Mak (1993) to indicate participation in delin- quent activity, was used. The current scale contained 22 items relating to delinquent offending and four ‘lie items’ to screen for possible cases of underreporting. To avoid causal order problems that could arise in cross sectional research, items in the present study were preceded with the following sentence: ‘If you found yourself in a situation where you had the chance to do the following things, how likely is it that you would do each one?’ Participants were instructed to give their answer using a 6-point rating scale from 1, Never do, to 6, Definitely do.

Delinquent Peers Scale The 7 items comprising the delinquent peers scale in this study were adopted with minor alterations from Bennett (1996). Respondents were asked how many of their friends had committed any of the seven predatory delinquent acts in the last 12 months. Responses to individual items ranged from 1, None of them, to 4, Most of them.

Family Interdependency Scale The measure of interdependency for this research focused on interdependency within families, measuring the following four qualities identified by Braithwaite (1989); attachment to parents, communication, closeness to parents and respect toward parents. The 8 items4 in the scale were adopted from the Hudson’s Child’s Attitude Toward Mother Scale (Thornberry, Lizotte, Krohn, Farnworth, & Jang, 1991) and from Bennett (1996). Participants were asked to rate their relationship with both parents5 on a 4-point rating scale ranging from 1, Never, to 4, Often.

Parental Shaming, Reintegration and Stigmatisation Scales There is a lack of established and validated scales measuring RST’s main concepts (i.e., shaming, reintegration and stigmatisation) in a family context. Therefore, scales to measure these concepts had to be developed for this study. Items for this purpose were adopted from previous studies (Bennett, 1996; Harris, 2001; Hay, 2001) based on the following criteria: (1) face validity, (2) theoretical salience in the context of RST research, (3) significant correlates and/or predictors in earlier analyses by other researchers, (4) no overlap either theoretically or empirically with other measures.

While Braithwaite’s original theory (1989) evaluated all shaming as either reintegrative or stigmatic, in his later work (Braithwaite & Braithwaite, 2001), based on the results from Harris (2001), Braithwaite recommended treating shaming, reintegration and stigmatisation as three independent concepts. The current research followed this latest recommendation. Selection of items measuring

166

IBOLYA LOSONCZ AND GRAHAM TYSON

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

shaming was driven by the definition from Braithwaite: ‘Shaming means all societal processes of expressing disapproval which have the intention or effect of invoking remorse in the person being shamed and/or condemnation by others who become aware of the shaming’ (1989, p. 100). From this quote we can see that shaming does not need to be verbal, public, humiliating, intentional, or even an extraordi- nary event. In a family context it can be a subtle part of child-rearing, such as a discussion between a child and a parent on how the child’s wrongdoing might have impacted on other peoples’ lives. Seven items (items 13, 14, 15, 16, 18 and 12, 17 reverse coded, in Table 2) were developed for the current study in an attempt to measure parental shaming. They consisted of nonphysical actions communicating to the adolescent that their wrongdoing has not gone ignored or

167

PARENTAL SHAMING AND ADOLESCENT DELINQUENCY

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

TABLE 1

Summary of Participants’ Demographic Data

N %

Age (years)

14 63 37.3 15 87 51.5 16 18 10.6 17 1 0.6

Parents’ Marital Status Married or de facto 104 61.6 Separated 19 11.2 Divorced 29 17.2 Parent/s deceased 8 4.7 Not indicated 9 5.3

Mothers’ country of birth Australia 120 71.0 New Zealand, north-west Europe and north Americas 17 10.1 Other 32 18.9

Fathers’ country of birth Australia 113 66.9 New Zealand, north-west Europe and north Americas 24 14.2 Other 31 18.3 Not indicated 1 0.6

Mothers’ highest educational attainment Left school before 15 22 13.0 High school 59 34.9 Trade or technical course 25 14.8 University 41 24.3 Not indicated 22 13.0

Fathers’ highest educational attainment Left school before 15 12 7.1 High school 40 23.7 Trade or technical course 33 19.5 University 53 31.4 Not indicated 31 18.3

unpunished. The items set out to measure action and inaction, as well as the feeling of shame by the respondent. Most items were adopted with some alter- ations from previous studies.

Parental stigmatisation is the least successfully captured concept in the RST research literature to date. The 6 items (items 1–6 in Table 2) used in the current analy- sis were adopted from previous studies to reflect the following features of stigmatisation identified by Braithwaite (1989): shaming and labeling the person not the act, humilia- tion of the person, disrespect and labeling with an outcast identity.

The 5 items (items 7–11 in Table 2) developed to measure reintegration attempted to capture parental integration and reintegration by reflecting approval of the person and positive behaviour; rituals, words or gestures of forgiveness, trust and care; and sustaining pride in having the person included as a member of the family.

All items, except item 10, were introduced by the statement, ‘When you do things that your family or school teachers would generally consider to be the wrong thing to do, and your mum and/or dad finds out about it, how often do they …’. Item 10 was introduced with the statement, ‘Of course, on many occasions we do nice things as well, things that normally would please our family and teachers. When you do things that your parents like or approve of how often do they …’. For all 18 items, which are shown in Table 2, the response options ranged from 1, Never, to 4, Often.

Findings A principal component analysis with varimax rotation was performed on the 18 items that measured shaming, reintegration and stigmatisation. A Kaiser measure of sampling adequacy was .842, indicating that the group of items measuring shaming, reintegration and stigmatisation was appropriate for factor analysis. Based on the recent assertion of Braithwaite that shaming, reintegra- tion and stigmatisation are likely to be independent concepts (Braithwaite & Braithwaite, 2001) items were forced into three factors. Loadings on the three components are shown in Table 2. Items that either did not reach the cut–off of .45, or did not load the highest on to a factor for which conceptual support was evident, were deemed to be excluded. The result of the analysis was found to be consistent with the literature and has largely supported the proposed structure.6

Sixteen of the 18 items were retained. Component 1, which was labelled parental stigmatisation, had an eigenvalue of 5.57 and accounted for 32.76% of the variance. The items loading on component 2, which had an eigenvalue of 3.24 and accounted for 19.10% of the variance, reflected approval of the adoles- cent and their behaviour and was consequently labelled parental reintegration. The items that loaded on component 3 communicated to the adolescent that their wrongdoing had not been ignored or unpunished and hence were labelled parental shaming. This component had an eigenvalue of 1.21 and accounted for 9.76% of the variance. All the scales were subjected to a reliability analysis. The alpha values listed in Table 3 are indicative of a satisfactory degree of internal consistency for all scales. Also shown in Table 3 are the descriptive statistics for each of the scales. The differ- ences between sexes were only significant on the projected delinquency scale where girls scored significantly lower (M = 1.79, SD = 0.78) than boys (M = 2.12, SD = 0.93),

168

IBOLYA LOSONCZ AND GRAHAM TYSON

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

t(167) = –2.420, p < .05, 2-tailed; and the delinquent peers scale where girls scored significantly lower (M = 1.45, SD = 0.58) than boys (M = 1.65, SD = 0.59), t(167) = –2.253, p < .05, 2-tailed.

Causal Structure of the Model To test the proposed model of RST (Figure 1), Structural Equation Modeling (SEM)7

was performed using AMOS 5. The model consisted of 5 observed exogenous variables (parent–child interdependency, parental shaming, parental reintegration, parental stigmatisation and delinquent peers) and one observed endogenous variable (projected delinquency). The model predicted that the negative relationship between parent–child interdependency and delinquency will be mediated by parental shaming,

169

PARENTAL SHAMING AND ADOLESCENT DELINQUENCY

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

TABLE 2

Rotated Component Matrix of Key Concepts of RST Items

Items Stigmatisation Reintegration Shaming

Stigmatisation 1 Show disappointment in you as a person .84

2 Make you feel ashamed of yourself .77

3 Make negative judgments about what kind of person you are .74 –.37

4 Treat you as though you are a bad person .71 –.32

5 Treat you like a criminal .66 –.48

6 Think of you as a bad person .62 –.34

Reintegration 7 Let you know in some way that they

still trust you .80

8 Let you know in some way that they still care about you –.33 .77

9 Discuss what happened with you .71

10 Mention it to someone else (when you do good things) .63

11 Forgive you after a while –.34 .61

Shaming 12 Do nothing about it (reversed) .65

13 Show disappointment in what you did .56 .58

14 Punish you in some way .41 .49

15 Try to have you ‘make up’ for what you did .33 .42 .47

16 Try to convince you that what you did was wrong or unfair .42 .45

Not included 17 Take your side (reversed) –.62 .49

18 Make you feel guilty or ashamed of what you did .64 .37

reintegration and stigmatisation and delinquent peers. Interdependency will have a direct positive effect on shaming and reintegration, and a direct negative effect on stigmatisation and delinquent peers. In turn, parental shaming, reintegration and stigmatisation will all have a direct effect on adolescent delinquency. Parental shaming and reintegration will have a negative effect, whereas parental stigmatisa- tion will have a positive effect. Furthermore, the positive relationship between stigmatisation and delinquency will be mediated by delinquent peers. Gender of the adolescent is proposed to increase the effectiveness of reintegrative shaming. Specifically, it is expected that girls will have a higher parent–child interdepen- dency, which will increase the prospect of shaming and reintegration and decreases the chances of stigmatisation. This, in turn, will lead to lower delinquency by females. This proposed effect will be tested by comparing the results of the pooled or whole sample model with the results generated by estimating the same model separately on the two subsamples of males and females.

The ratio of cases to observed variables was 24:1 and the ratio of cases to parameters is 9:1; both a good ratio (Byrne, 2001). There were no missing data. The estimation of parameters was based on the maximum likelihood method.

Assumptions The assumptions of univariate normality were evaluated through histograms and summary descriptive statistics. Four of the seven observed variables were significantly skewed; projected delinquency,8 delinquent peers, family interdependency, and parental stigmatisation. Appropriate transformations to normalise these variables were performed, which reduced the skew although the projected delinquency scores were still skewed. As the use of the transformed variables did not alter the results of SEM analy- sis, the untransformed scores were used in the final analysis. It was not feasible to examine all pairwise scatterplots to assess linearity and so randomly selected pairs of scatterplots were examined. All the selected pairs of scatterplots appeared to be linearly related, and no univariate outliers were detected.

Model Estimation The model went through three revisions before reaching its final form. Analysis of the initial Braithwaite model (Model 1 presented in Figure 1) found only marginal

170

IBOLYA LOSONCZ AND GRAHAM TYSON

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

TABLE 3

Summary of Scale Reliability and Scores of Participants

α M SD Min Max

Projected delinquency scale .93 1.98 .88 1.00 5.17

Delinquent peers scale .85 1.56 .57 1.00 3.43

Family interdependency scale .93 3.30 .65 1.06 4.00

Parental shaming scale .67 3.07 .57 1.50 4.00

Parental reintegration scale .84 3.07 .68 1.25 4.00

Parental stigmatisation scale .88 1.89 .72 1.00 4.00

support of the model in terms of χ2 test statistic, goodness of fit index (GFI)9 and comparative fit index (CFI),10 χ2(6, 169) = 75.52, p < .001; GFI = 0.91; CFI = 0.81. It should be noted that the significant χ2 indicates that the model and the observed covariance structure are dissimilar. Further, the value for CFI was below the recom- mended value of (> .95) of a well fitting model (Bentler, 1988). Another concern is the root mean square error of approximation value (RMSEA = .26). RMSEA11

values less than .08 indicate a good fit, values ranging from .08 to .10 indicate a mediocre fit, and those greater than .10 indicate a poor fit (Byrne, 2001). The high RMSEA of Model 1 indicated a poor fit.

Post hoc model modifications were performed in an attempt to develop a better fitting, more parsimonious model. The high modification index values between stigmatisation and shaming (MI = 20.81) and reintegration and shaming (MI = 14.85) indicated cross-loadings between shaming and the concepts of stigma- tisation and reintegration. Furthermore, the marginal and nonsignificant squared multiple correlation (R2 = .01) for shaming indicated that practically no variance associated with parental shaming is accounted for by its predictor — interdepen- dency, and consequently parental shaming was removed from the model. The result- ing model (Model 2 shown in Figure 2), fitted the data better, χ2(3, 169) = 10.12, p > .02; GFI = .98; CFI = .98; RMSEA = .11.

To explore the possibility of a direct relationship between interdependency and delinquency, an alternative model that included a causal path from interdependency to delinquency was tested. Analysis of this alternative model found a negligible and nonsignificant association (.01) between interdependency and delinquency and there- fore this alternative model was rejected in favour of Model 2.

The validity of Model 2 was checked separately on the male and female sub- samples. The model fitted the data better for females χ2(3, 73) = 10.52, p > .02; GFI = .95; CFI = .95; RMSEA = .19, than for males χ2(3, 96) = 14.43, p > .002; GFI = .94; CFI = .93; RMSEA = .20.

Standardised regression weights and squared multiple correlations for the final model (Model 2), are given in Figure 2 (values for the pooled data are bolded, followed by values for the male and female subsamples respectively). For the pooled sample, interdependency was positively predictive (.62) of reintegration, and the expected negative effect of reintegration on delinquency was notable and signifi- cant (–.25). As predicted, interdependency negatively affected both stigmatisation (–.47) and delinquent peers (–.38), both of which had a positive effect on delin- quency. Delinquent peers were positively predictive (.51) of delinquency, and they mediated the relationship between interdependency and delinquency as well as between stigmatising and delinquency. The association between parental stigmatisa- tion and delinquent peers was positive (.10), but nonsignificant. Interdependency between the parent(s) and the child explained 22% of the variation in parental stigmatisation and 38% in parental reintegration. Furthermore, interdependency and parental stigmatisation explained 19% of the variation in delinquent peers. This model explained 52% of the variation in projected delinquency.

For girls the relationship between interdependency and delinquency was more likely to be mediated by delinquent peers than by stigmatisation. The relationships between interdependency and delinquent peers (–.54), and delinquent peers and delinquency (.65) was considerably stronger for girls than for boys (–.33 and .41

171

PARENTAL SHAMING AND ADOLESCENT DELINQUENCY

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

respectively), and interdependency between parent and child explained a higher variation in delinquent peers for girls (.28) than for boys (.17). Furthermore, the negative relationship between reintegration and delinquency was considerably weaker, and nonsignificant, for girls (–.10) than for boys (–.33).

Discussion In the past decades reintegrative shaming has become an important concept as the framework for restorative justice programs and diversion conferencing. However, to date, RST has received little empirical attention. This study set out to contribute to the empirical testing of the theory by examining the appropriateness of the causal model put forward by Braithwaite. While a firm conclusion regarding empirical support for the theory based on this study cannot be made, there are a number of findings that are worth discussing.

Key Concepts of the Theory All key concepts of the theory were adequately captured in this study. Factor analy- sis of the survey data did not find shaming to be independent of reintegration and stigmatisation as proposed by Braithwaite and Braithwaite (2001). In fact, nearly all items measuring shaming also have been found to load onto either reintegration or stigmatisation. This tends to support Braithwaite’s original theory (1989), which evaluated all shaming as either reintegrative or stigmatic.

All the features of stigmatisation identified by Braithwaite (i.e., shaming and labelling the person, not the act; labelling with an outcast identity; and humiliation

172

IBOLYA LOSONCZ AND GRAHAM TYSON

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

FIGURE 2

Model 2: Final model of family model of reintegrative shaming. Note: Standardised regression weight and squared multiple correlations for pooled data, male and females

respectively.*p < .05.

and disrespect of the person) loaded well on this factor. However, our factor analysis found that over half of the items measuring stigmatisation and reintegration had shown a considerable negative loading on the other scale, suggesting that stigmati- sation and reintegration are not completely independent from one another. For example, many of the items set out to measure stigmatisation, by capturing negative judgment and labelling by the parent, has also negatively loaded on the factor or reintegration. Further, the item measuring care, identified by Braithwaite as a feature of reintegration, has also negatively loaded on the factor of stigmatisation, suggesting that although its presence indicated reintegration, its absence indicated stigmatisation. Given that lack of care is associated with labelling, it is not surpris- ing that they loaded on the reintegration factor, as well as on the stigmatisation factor set out to capture labelling of the wrongdoer. In addition, the item measuring forgiveness, a central aspect of reintegration in RST, has also negatively loaded on stigmatisation, indicating that lack of forgiveness may also give the impression of an outcast status (a feature of stigmatisation) to the adolescent. For RST it means that not all facets of reintegration and stigmatisation are discrete concepts.12

Parental shaming, the third key concept in the theory, proved to be the concept least successfully captured in this study. Braithwaite (1989) himself identified the difficulties of measuring shaming through survey research, and warned of the limita- tions of measuring only the respondent’s perception of having been shamed or been forgiven. Accordingly, the current study incorporated the adolescent’s perception of having been shamed and whether they felt ashamed. However, only the items measuring the adolescent’s perception of having been ashamed loaded high enough on the factor of shaming to be retained.

While Braithwaite (1989) predicted that lower delinquency rates among girls are related to higher level of interdependency and reintegration and lower level of stigmatisation among girls, the current study found that while girls reported a signif- icantly lower level of delinquent behaviour and delinquency among their peers, the level of interdependency, reintegration and stigmatisation reported by girls was only marginally different from the level reported by boys.

Testing the Causal Path of the Model SEM analysis found that both stigmatisation and reintegration had a significant causal effect on projected offending. Consistent with the prediction of RST, the effect of parent–child interdependency on delinquency was found to be indirect, mediated by delinquent peers and parental stigmatisation and reintegration. As expected, a strong relationship was revealed between the level of parent–child interdependency and parental reintegration and stigmatisation. This strong effect supports the idea that parents with close relationship with their children are more likely to use reintegration and less likely to use stigmatisation. Children with close relationships with their parents also reported having less delinquent friends. This association was particularly strong for girls. As predicted by RST, interdependency contributed both directly and indirectly to having delinquent peers. Although the indirect contribution through parental stigmatisation was small and nonsignificant, it is still of some interest because it underscores the dual role played by this relationship.

173

PARENTAL SHAMING AND ADOLESCENT DELINQUENCY

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

In addition to the significant positive effects of stigmatisation on delinquency, the current study also found a significant negative effect of reintegration on delinquency. Contrary to the study by Bennett (1996), which did not find reintegrative shaming to be a significant determinant of the frequency of offending, or the research by Hay (2001), which found that the initial reintegration effect on delinquency had disap- peared after controlling for interdependency, the result of this study seems to indicate that reintegration has a negative effect on delinquency. It should be noted, however, that this negative relationship was nonsignificant for girls.

The strongest predictor in this study of projected delinquency was association with delinquent peers, indicating that adolescents who reported more association with delinquent peers are also more likely to be involved in offending. This associa- tion was particularly strong for girls. However, this result needs to be interpreted with caution, as the delinquent peers scale utilised in this study is based on the adolescent’s perceived belief of their peers’ participation in delinquent acts, a measure that can be inaccurate as adolescents often misjudge the degree of similar- ity between themselves and their friends (Erickson, Crosnoe, & Dornbusch, 2000). Even so, other research studies have also found this association to be one of the strongest predictors of delinquency (Agnew, 1995; Matsueda,1982; Warr & Stafford, 1991). Furthermore, when testing for the predictions of RST, Bennett (1996) found that delinquent friends had the greatest effect on persistence in delin- quency, followed by prior offending and stigmatising. It needs to be noted, however, that a lack of measure of a prior delinquency in the current study may mean that the effects of reintegration, stigmatisation and delinquent peers on delinquency may be spurious due to the effects of prior delinquency.

Another noteworthy result is the lack of contribution of parental shaming to the model. Its effect on future delinquency was marginal and statistically nonsignificant. In fact, its exclusion from the model resulted in a better fitting model. Previous research by Bennett (1996) has also found a lack of association between shaming and repeated offending. This lack of contribution of parental shaming to the model may be due to the considerable overlap of shaming with stigmatisation and reintegration. It could also point to the question of construct validity. That is, shaming did not have the predicted effect on delinquency because all the items were measuring attempts of shaming by parents rather than emotional responses of shame by the adolescent. This explanation is supported by the research study of Harris (2001), in which shaming did not independently predict the emotion of shame. Instead, the relationship was depen- dent upon the amount of respect that the participants had for those who disapproved of their actions. This makes intuitive sense. For example, a youth shamed by a person whom she does not respect is unlikely to feel shame, and may even feel pride. On the other hand, if the person’s opinion is respected then a critical judgment is much more likely to result in feelings of shame. This could have an important practical contribu- tion for practitioners in the field of restorative justice. Many young offenders often feel, and may in fact are, marginalised by society, and lack a psychological connection to people in mainstream institutions and in their community or to members of their family. Inclusion of these nonrespected others in conferences would lack positive outcomes, and may, in fact, undermine the intention of the process.

It is becoming increasingly evident that the process of shaming and its emotional consequences are complex. For many, certain aspects shaming, or at least some aspects

174

IBOLYA LOSONCZ AND GRAHAM TYSON

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

of shaming, is inherently stigmatising. This leads to some concerns as to whether shaming should be considered as relevant to restorative justice. Further, the apparent lack of research on shame prevents researchers from finding consistent answers to questions about its role in criminal justice. In summary, empirical validation of the main concepts of RST and its proposed causal effects need to be further tested. This further evaluation and development of RST seems largely dependent on a more complete knowledge of shame and how the feeling of shame is resolved.

Evaluation, Future Research and Implications The data used in this study were a school-based sample. As with any school-based survey, there are certain limitations, such as the exclusion of truant, sick, and tardy students, and the possible under-representation of high-risk youths. Furthermore, the use of active parental consent in this study could lead to concerns of representativeness. However, the main aim of the current study was to explore the relationships between variables, as opposed to measuring the prevalence of those variables in the population, and therefore the effect of active parental consent is likely to be minimal on the results of this study (Dent, Sussman & Stacy, 1997). Further, the relatively small sample size13

of this study has, to a degree, reduced the statistical power of the models and observed relationships. This may have contributed to the observed nonsignificant relationships between some of the variables, particularly in the sex-specific models.

Due to constraints, this study only examined the individual level predictions of RST. Future research should try to include some of the macro-level variables as well. In addition, future research on RST should be aimed at refining measurement of the key concepts in the theory, with particular attention to shaming and the array of feelings induced by shaming. These studies should include a control for the child’s prior involvement in delinquency (an apparent limitation of the current study) and consider the use of a longitudinal design to better assess the complex causal processes, and the inclusion of interactions between the adolescent and sanctioning agents other than parents. Most importantly, future studies should evaluate the recent critical refinement that individuals are only responsive to shaming by those whom they respect.

In conclusion, Braithwaite’s RST is unquestionably popular in the restorative justice arena, even though there appear to be some ambiguities as to what each of the key concepts implies and how to best maximise the positive aspects of their applica- tion. Undoubtedly, the theory is very broad in scope and it may take considerable effort to develop it into an empirically testable theory. However, if RST is to contribute to our understanding of the causes of crime and how best to deal with offenders, then it requires further refinement and increased systematic empirical research. The findings of this study are largely supportive of RST, particularly the theory’s emphasis on the harmful effects of stigmatisation and the beneficial effects of reintegration. However, shaming, as defined in the theory, may not affect predatory crime in the way it is predicted. The result of this study is not intended to be a defini- tive statement on the construct validity of RST. Rather, it is a contribution to the set of empirical research on RST, and it should be examined in the context of these empirical tests.

175

PARENTAL SHAMING AND ADOLESCENT DELINQUENCY

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

Endnotes 1 This article is based on work for Ibolya Losoncz’s Honours thesis. The opinions, comments and/or

analysis expressed in this article are those of the author’s and do not necessarily represent the views of the Minister for Families, Community Services and Indigenous Affairs or the Australian Government Department of Families, Community Services and Indigenous Affairs, and cannot be taken in any way as expressions of Australian Government policy.

2 The RYDS was designed to examine the development of delinquent behaviour and drug use among adolescents. The 4-year study consisted of interviews with approximately 1000 youths and their parents at 6-month intervals.

3 The NYS is an American national probability sample of 1725 adolescents (11–17 years). The first wave was collected in 1977.

4 How often would you say that: You get along with your parents? You feel that you can really trust your parents? You really enjoy your parents? You feel proud of your parents? You have a lot of respect for your parents? Your parents ask you what you think before making a decision about you? Your parents comfort you and help you? How close do you feel to your parents? (Never, Seldom, Sometimes, Often)

5 Adolescents who spent more time living with a step-parent, were instructed to record their relationship with the stepparent. Children who lived away from their mother or father and did not spend much time with him or her were instructed to leave the relevant column empty.

6 The items measuring reversed shaming (17 — Take your side) and whether the respondent felt ashamed at the time (18) did not load sufficiently onto the component of shaming to be included.

7 SEM allows for the exploration of various constructs and their interrelationships as well as their causal contribution to other constructs. Furthermore, it allows for a better index of fit of a causal model to the data and testing alternative models (Byrne, 2001).

8 Previous research in the area of self-reported delinquency (Thornberry et al., 1991) found delin- quency scales to be skewed.

9 GFI compares the hypothesized model with no model at all. Values range from zero to 1.00, with values close to 1.00 being indicative of good fit (Byrne, 2001).

10 CFI assesses the fit of the model relative to other models. The CFI is normed to the 0 to 1 range, and does a good job of estimating model fit even in small samples. Because of the small sample size, values over .90 are desirable (Bentler, 1988).

11 The RMSEA takes into account the error of approximation in the population and asks the question, ‘How well would the model, with unknown but optimally chosen parameter values, fit the population covariance matrix if it were available?’ (Browne & Cudeck, 1993, pp. 137–138).

12 A bivariate correlation analysis found considerable correlations between the three main concepts in theory. The highest correlation was found between parental reintegration and stigmatisation (r = –.45, p <.001), followed by the association between reintegration and shaming (r = –.31, p < .001), and stigmatisation and shaming (r = –.29, p < .001).

13 SEM relies on tests which are sensitive to sample size as well as to the magnitude of differences in covariance matrices. In the literature, it is recommended that sample size should be at least 50 more than 8 times the number of variables in the model or to have at least 15 cases per measured variable or indicator Stevens (1996). Bentler and Chou (1987) recommend at least 5 cases per parameter estimate (including error terms as well as path coefficients). The sample size of the current study complies with most of these recommendations.

References Agnew, R. (1995). Testing the leading crime theories: An alternative strategy focusing on motivational

processes. Journal of Research in Crime and Delinquency, 32, 363–398. Ahmed, E., & Braithwaite, V. (2004). ‘What, me ashamed?’ Shame management and school bullying.

Journal of Research in Crime and Delinquency 41, 269–294.

176

IBOLYA LOSONCZ AND GRAHAM TYSON

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

Akers, R.L. (2000). Criminological theories: Introduction, evaluation, and application. (3rd ed.) Los Angeles, CA: Roxbury.

Bennett, K.J. (1996). A family model of shaming and delinquency: A partial test of Braithwaite’s Reintegrative Shaming Theory. Unpublished Ph.D. dissertation. Sam Houston State University, Huntsville, TX.

Bentler, P.M., & Chou, C.P. (1987). Practical issues in structural modeling. Sociological Methods and Research. 16(1), 78–117.

Bentler, P.M. (1988). Comparative fit indexes in structural models. Psychological Bulletin, 107, 238–246. Bonta, J., Jesseman, R., Rugge, T., & Cormier, R. (2006). Restorative justice and recidivism. Promises

made promises kept? In D. Sullivan & L. Tifft (Eds.), Handbook of restorative justice. London: Routledge.

Braithwaite, J. (1989). Crime, shame and reintegration. Cambridge: Cambridge University Press. Braithwaite, J., & Braithwaite V. (2001). Shame, shame management and regulation. In E. Ahmed, N.

Harris, J. Braithwaite & V. Braithwaite (Eds.), Shame management through reintegration (pp. 3–69). Cambridge: Cambridge University Press.

Browne, M.W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K.A. Bollen & J.S. Long (Eds.), Testing structural equation models (pp. 445–455). Newbury Park, CA: Sage.

Byrne, B. (2001). Structural equation modeling with AMOS: Basic concepts, applications, and programming. Mahwah, NJ: Lawrence Erlbaum Associates.

Dent, C.W., Sussman, S., & Stacy, A.W. (1997). The impact of a written parental consent policy on estimates from a school-based drug use survey. Evaluation Review, 21(6), 698–712.

Erickson, K.G., Crosnoe, R., & Dornbusch, M. (2000). A social process model of adolescent deviance: Combining social control and differential association perspectives. Journal of Youth and Adolescence, 29, 395–420.

Erikson, K.T. (1962). Notes on the sociology of deviance. Social Problems, 9, 16–29. Gibbons, D.C. (1991). Book reviews: Crime, shame, and reintegration. Crime and Delinquency, 37(4),

579–581. Grimes, R.M., & Turk, A.T. (1978). Labeling in context. Conflict, power, and self-definition. In M.D.

Krohn & R. Akers (Eds.), Crime, law and sanctions: Theoretical perspectives (pp. 39–58). Beverly Hills, CA: Sage.

Harris, N. (2001). Shaming and shame: Regulating drink-driving. In E. Ahmed, N. Harris, J. Braithwaite & V. Braithwaite (Eds.), Shame management through reintegration (pp. 73–205). Cambridge: Cambridge University Press.

Hay, C. (2001). An exploratory test of Braithwaite’s reintegrative shaming theory. Journal of Research in Crime and Delinquency, 38(2), 132–153.

Hirschi, T. (1969). Causes of delinquency. Berkeley, CA: University of California Press. Mak, A.S. (1993). A self-report delinquency scale for Australian adolescents. Australian Journal of

Psychology, 45(2), 75–79. Makkai, T., & Braithwaite J. (1994). Reintegrative shaming and compliance with regulatory standards.

Criminology, 32, 361–385. Matsueda, R.L. (1982). Testing control theory and differential association: A causal model approach.

American Sociological Review, 47, 489–504. Stevens, J. (1996). Applied Multivariate Statistics for the Social Sciences (3rd ed.). Mahwah, NJ: Lawrence

Erlbaum Associates. Strang, H., Barnes, G., Braithwaite, J., & Sherman, L. (1999). Experiments in restorative policing: A

progress report on the Canberra shaming experiments (RISE). Canberra, Australia: Law Program, Research School of Social Sciences, Australian National University. Retrieved May 10, 2002, from http://www.aic.gov.au/rjustice

Taylor, I., Walton, P., & Young, J. (1973). The new criminology: For a social theory of deviance. London: Routledge and Kegan Paul.

177

PARENTAL SHAMING AND ADOLESCENT DELINQUENCY

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

Thornberry, T.P., Lizotte, A.J., Krohn, M.D., Farnworth, M., & Jang, S.J. (1991). Testing interactional theory: An examination of reciprocal causal relationships among family, school, and delinquency. Journal of Criminal Law and Criminology, 82(1), 3–35.

Warr, M., & Stafford, M. (1991). The influence of delinquent peers: What they think or what they do? Criminology, 29, 851–866.

Zhang, L., & Zhang, S.X. (2004). Reintegrative shaming and predatory delinquency. Journal of Research in Crime and Delinquency, 41, 433–453.

Zhang, S.X. (1995). Measuring shaming in an ethnic context. British Journal of Criminology, 35, 248–262.

178

IBOLYA LOSONCZ AND GRAHAM TYSON

THE AUSTRALIAN AND NEW ZEALAND JOURNAL OF CRIMINOLOGY

■ ■ ■