Genesis of adolescent risk-taking: pathways through family, school, and peers.
Abstract: This paper presents an empirical examination of Sampson and Laub's social control theory. It tests the effects of family structure, family attachment, school attachment and peer attachment on a generalized form of risk-taking behaviour which includes delinquency and drug use. The data come from a single stratified sample of 1,075 high school students in Ontario. The findings suggest that the effect of family attachment on risk-taking is moderated by both school and peer involvement. When family attachment is low, school attachment inhibits risk-taking and strong peer attachment reinforces it.
Resume: Cette etude presente un examen empirique de la theorie du controle social de Sampson et Laub. Cette theorie verifie les effets de la structure de la famille, l'attachement a la famile, a l'ecole et aux pairs sur une forme generalisee de comportements a risques ci-inclus la delinquance et l'usage de drogue. L'information vient d'un echantillon de 1075 ecole secondaires de l'Ontario. Les resultats suggerent que l'effet de l'attachement a la famille sur les comportements a risques est modere par l'implication de l'ecole et des pairs. Quand l'attachement a la famille est faible, l'attachement a l'ecole inhibe les comportements a risques alors que l'attachement aux pairs les renforce.
Recent research has established the centrality of the family as an institutional crucible in the formation of delinquent conduct and attitudes (Loeber and Stouthamer-Loeber, 1986; McCord, 1979; Riley and Shaw, 1985; West and Farrington, 1977). The influence of family structure appears to derive from the informal social control it provides over children and adolescents. In their General Theory of Crime, Gottfredson and Hirschi (1990) stress the role of parental supervision in establishing strong internal self control. They argue further that other socializing institutions are largely inconsequential when the effects of the family are identified. Thus, the school system and peer groups have little additive or independent effect on one's likelihood of criminal behaviour beyond the family. This view is in contrast to the conclusions drawn by Sampson and Laub (1993) in their re-analysis of the historic Glueck and Glueck delinquency data. While acknowledging the central importance of the family as a socializing mechanism, Sampson and Laub argue that "social capital" (i.e. social embeddedness and personal interdependencies) can mediate the influence of families, mitigating the negative influences of criminogenic families where it exists and exposing the off-spring from non-criminogenic backgrounds to delinquent opportunities in its absence. In this analysis we examine some of the central tenets of these theories in order to better specify the determinants of adolescent risk-taking behaviours.
The concept of `criminality' is the positive disposition which makes delinquent choices attractive. Hirschi's (1969) earlier "social bond" theory had outlined the developmental correlates of conformity -- attachment, commitment, involvement and belief -- without which the risk of crime increased. Gottfredson and Hirschi's (1990: 85ff) concept of `criminality' attempts to spell out that underlying disposition affirmatively. The relatively stable, long term disposition based on impulsiveness or low self control is an individual property characterized by the need for immediate gratification of desires through the utilization of simple means ("money without work, sex without courtship, revenge without court delays"). This disposition is coupled with preferences for exciting, thrilling or risky activities, with little interest in long term interpersonal or economic investments, little need for skill or sophisticated criminal planning, and an insensitivity to the pain or discomfort of others (Gottfredson and Hirschi, 1990: 89-90).
In their self control theory, Gottfredson and Hirschi have provided a theoretical baseline for contemporary studies of delinquency by clarifying both the characteristics of crime as well as the characteristics of its perpetrators. "The vast majority of criminal acts are trivial and mundane affairs that result in little loss and less gain." These are events "whose temporal and spatial distributions are highly predictable" -- typically representing unprotected targets close at hand with little or no surveillance or resistance. They are also events that "require little preparation, virtually no specialized knowledge and often do not produce the result intended by the offender" (1990: 16). On this basis, self control theory cautions against misconstruing certain elements of the offender's repertoire as causes of other elements of misconduct, i.e. narcotics as a cause of robbery or assault since all are common derivatives of the underlying "criminality." The type of crime committed is largely opportunistic, and is liable to evidence little preparation or specialization. Self control theory also extends to non-criminal acts of low self-control. "[Such persons] will tend to smoke, drink, use drugs, gamble, have children out of wedlock, and engage in illicit sex...[P]eople who lack self-control will tend to be impulsive, insensitive, physical (as opposed to mental), risk-taking, short-sighted, and non-verbal, and they will tend therefore to engage in criminal and analogous acts" (1990: 90). Such traits appear where parental supervision has been lacking, and/or where discipline has been harsh and inconsistent. Braithwaite (1989) suggests that it can also occur where the children are stigmatized and rejected, as opposed to shamed and reintegrated. Other evidence suggests it occurs as a result of childhood maltreatment (Smith and Thornberry, 1995; Zingraff, Leitor, Myers and Johnsen, 1993). Whatever the precise mechanism, these traits are evident to teachers as early as grammar school and, according to Sampson and Laub (1993) and Nagin and Farrington (1992), they tend to persist throughout the life cycle.
Self control theory is critical of the role assigned to factors outside the family -- the school, delinquent friends, and criminal justice interventions. Contrary to suggestions of school critics (Cohen, 1955; Rosenthal and Jacobson, 1968), the theory contends that the school is not a primary source of socialization which fosters criminal careers (Wineburg, 1987). The school makes demands of control, discipline and accountability which are difficult for the low selfcontrol student to meet, and for this reason, poor performance and early school leaving are outcomes of low self control, not causes of it, and hence, not causes of delinquency. Adolescents from fractious families may be at greater risk of early school leaving just as they are more likely to engage in risk-taking behaviours because their families have ill-equipped them to respond appropriately to classroom discipline or because the parents have already lost any control they had over the children. Thus, for Gottfredson and Hirschi, dropping out of school is another outcome of ineffective familial socialization and further evidence of impulsiveness.
Delinquent peers have traditionally been thought of as a determinant of criminal involvement, especially in the context of the contemporary images of the power of gangs over ghetto adolescents (Katz, 1988: 144-163). Gottfredson and Hirschi argue that the drift of adolescents towards gangs similarly reflects a loss of parental control. Adolescents seek out people like themselves who are thrill seekers, renegades, and who reject conformity. Unlike Sutherland's differential association theory (1939) which suggested that gangs were a sort of counter cultural fraternity where the positive labelling of crime outweighed the negative, and where the gang became a repository for the acquisition of specialized knowledge, contacts and attitudes, self control theory argues that gangs and delinquent associates are only attractive to people who are already impulsive, hedonistic, and lack normal restraints. From this perspective delinquent peer associations should not be a source of juvenile risk-taking behaviour; rather, peers become important as a result of ineffective familial bonding and will not generally cause juvenile risk-taking behaviour.
In contrast to this static/general model which predicts "heterotypic continuity" over the life of the offender (stable expressions of non-specialized misconduct), many criminologists from the life course perspective view crime as "state dependent" and subject to age sensitive dynamics. In their focus on "social capital" as the networks of associations through which informal social control is exercised, Sampson and Laub (1993) return to elements of Hirschi's social bond theory -- attachments to parents and teachers and involvement in prosocial activities. The social capital model is used to explain how emotional attachments to spouses and employers in later life can contribute to criminal desistance among youth with serious adolescent delinquency records -- and is associated with arrest among men without prior adolescent records where such attachments do not materialize. Sampson and Laub found that social capital attachments appear to mediate the tendency towards low self control even though the latter appears phenomenally persistent. In this context, social capital is the existence of personal bonds of emotional interdependencies and role reciprocities which can provide social and psychological resources that individuals can draw on as they move through life's major points of transition (leaving school, establishing a career, developing personal ties, establishing a family, etc.).
In a re-analysis of the Glueck and Glueck study (1950, 1968), Sampson and Laub (1993: 111), found that, in addition to family background (poverty, crowding, residential instability) and family process variables (supervision, acceptance), school attachment was a powerful predictor of delinquency. They found that such family background factors influenced the social control processes in the family, and, in turn, these social control factors influenced the likelihood of normal school completion, adolescent delinquency, and entry into legitimate employment. Finally, they also found that delinquent peer attachment contributed significantly to delinquency in their sample of Boston boys independent of both family and school attachment. If the Sampson and Laub re-analysis of this rich data set from the Glueck and Glueck study accurately captures the independent contribution of delinquent peers and schools while both recognizing the centrality of family processes which persumably contributes to self control and the structural determinants of such processes, then this synthesis represents a major conceptual advance within the control perspective.
In this study, we propose to test key elements of the social control theory on adolescent risk-taking behaviour based on a large stratified sample of subjects designed primarily to examine drug use among Canadian students. We prefer the concept of "risk-taking behaviour" as opposed to delinquency because the idea of adolescent risk-taking behaviour is more consistent with the non-specialization theorem of control theory. This theorem predicts that delinquent acts as well as alcohol, cannabis, tobacco and hard drug use represent opportunistic occasions for illicit pleasure or enjoyment, frequently by illegal means, that involve a risk, danger or peril. Other researchers have explored related "imprudent behaviours" (Arneklev et al., 1993). The first element of the study involves the examination of an inclusive model of risk-taking to determine whether the non-specialization theorem enjoys empirical support.
After assessing whether there is support for the non-specialization theorem, we proceed to examine the causal processes spelled out by Sampson and Laub (1993). Figure 1 presents the conceptual model that describes the relationship between family structure, economic background, family processes, social capital, and risk-taking behaviour. The broken arrows in Figure 1 illustrate where Sampson and Laub expand on self control theory to accommodate both the influence of structural factors on sources of social capital outside the family, specifically school and peer attachment, as well as the potential consequences of these sources of capital on risk-taking behaviour.
Following Sampson and Laub we examine structural factors -- income and family form -- as predictors of attachments to parents, school, and peers. When one examines the contribution of background factors such as poverty, residential mobility etc. in models which also fit parental or family dysfunctions (erratic discipline, poor supervision, parental rejection of child, weak attachment to parents), Sampson and Laub found that the background factors become non-significant as predictors of delinquency whether measured officially or unofficially. However, they found that the background factors predict variations in parental effectiveness, school attachment and delinquent peer association. This means that structural stressors influence delinquency, but only indirectly by undermining parental effectiveness in the families (1993: 81).
In this analysis, because direct measures of self control were unavailable, family relationships were given primary theoretical relevance. Indeed, since one of the most important measures of low self control is delinquency, this variable should probably not appear on the right hand side of the equation. This is consistent with the findings of Sampson and Laub and a host of other criminologists such as Hirschi's earlier control theory and contemporary ethnographic work. As Fleischer remarks in his introduction to Beggars and Thieves, an ethnography of urban street criminals in Seattle, "the nature of natal family life is the central issue that affects criminal's life trajectories" (1995: 4).
We also test to see whether the conjoint effects of family, school, and peer attachment on risk operate independently or if they influence each other. In other words, we test whether the level of involvement with family moderates the effect of school and peer attachment on risk-taking to assess whether school and peers can compensate for, or exacerbate poor family processes.
The data for this study were obtained from the 1993 Ontario Student Drug Use Survey by the Addiction Research Foundation (Adlaf, Smart and Walsh, 1993). This survey is part of a bi-annual, long term, systematic cohort study of drug use by Ontario students that began in 1977. The target population is students across Ontario in the junior high and high school systems. Thus, it excludes students who are not enrolled in the regular school system such as those on Native reserves, Canadian Forces bases, correctional and health institutions, those enrolled in private schools, and those who have dropped out of school (Adlaf et al., 1993). Since some of these populations are at unusually high risk for delinquency, their exclusion is likely to make the conclusions drawn here more conservative.
The design is based on a stratified single-stage cluster probability sample of school classrooms in grades 7, 9, 11, and 13 across four regions of Ontario -- east, west, central and north. In 1993 the sampling design included 25 school boards involving 233 classes in 165 different schools. The participation rate was 77 percent (n = 3,571) (Adlaf et al., 1993). Participation in the survey was voluntary and required approval from school principals, teachers, and the parents of students. The questionnaires were self-administered and results were kept completely anonymous. In this analysis we use the unweighted data. Since we control for both stratification factors (age and region) in the analysis, any bias should be minimal (Groves, 1989).
The questionnaire gathered information on a variety of topics such as demographic and social characteristics, attitudes towards and prevalence of drug and alcohol use, delinquency, and school performance. To reduce the time necessary to complete the survey, two separate instruments, Schedules "A" and "B" were created in which 70 percent of the items were identical and 30 percent of items were unique to each questionnaire (Adlaf et al., 1993). Many of the items essential to this analysis were contained in the unique questions, thus reducing our initial sample by half. Since the two versions of the questionnaire were randomly assigned to respondents, we did not anticipate any systematic bias across the two halves of the sample. The working sample after including cases that answered only one version of the interview schedule was n = 1,725. Missing data from certain questions reduced this sample to 1,075 using listwise deletion for an effective sample size of 62.3 percent of the original students who answered version "B" of the questionnaire. This sample size allows us to detect mean differences of as little as 0.20 of a standard deviation at a significance level of p < .01 with a statistical power greater than 0.95 (Cohen, 1988).
One of the common criticisms about drug use surveys is the tendency for people to report socially acceptable responses to illicit substance usage. Since the target population is students in junior high and high school and the instrument was a self-report questionnaire, one might expect that the respondents might over-report usage, thus biasing the results. To correct for the potential of over-reporting, the Student Drug Use Survey included one item about usage of a fictitious drug called adrenochromes (also known as "wagon wheels"). Of the sample who completed the "B" interview schedule, no student responded positively to this question, adding credibility to the results of this analysis. Adlaf et al. (1995) have found that in successive surveys from 1977 to 1995, over-reporting of drug use based on a fictitious drug in this survey is about 0.5 per cent.
Socio-Demographic Controls: The socio-demographic control variables in this analysis consisted of students' sex and age. Sex was dummy coded so that males = 0 and females = 1. The proportion of males was 0.46 and the number of missing cases was 5. Age was reported in years. The median age was 15 years old and the age ranged from 11 years old to 20 years old. Region was categorized into 4 regions using 3 dummy coded variables. The central region (Metro Toronto) was used as the reference category.
Family Structure: Family structure was broken down into two types of single -- two natural parents versus all others. The other category included single-parent families, families with one natural-parent and one step-parent, and families with neither natural parent. Ten respondents failed to answer this question and were deleted from the analysis. A dummy variable was constructed coding two natural-parent families = 0 and all other families = 1. The proportion of two-parent families in the analysis was 0.75.
Family Financial Situation: This variable asked students to describe their family's financial situation on a 5-point scale ranging from well-below average to well-above average. The nonresponse rate for this question was 2.6 percent (45 cases).
Family Attachment: Respondents answered six questions about their relationship with their parents. Questions asked about how much time during weekends was spent with parents, whether it was important to get along with parents, how well they did get along with them, whether they talked to their mother and father about problems, and if a parent knew where they were when they were away from home. Two questions, whether the student talked to (1) the mother and (2) the father if they were having problems were combined and the highest score from either item was used to construct a single item to compensate for cases of single-parent families. The resulting five items were then standardized and combined to create a scale that measured the child-parent relationship with an internal reliability of 0.72.
School Attachment: Students were asked a series of questions about how well they liked school, how well they were doing at school, whether they skipped classes, and how likely it was that they would complete their high school education. Six items were asked in the specific interview schedule and were used in this analysis. The items were standardized and then summed to create a general measure of involvement with school. The standardized alpha for this scale was 0.69.
Peer Attachment: Students were asked a number of questions about how many nights per week they spent doing certain activities. Three of these activities -- going to a friend's house, going out with a friend, and "just hanging around"--were combined to construct a general measure of peer attachment. It is important to note here that these questions refer to peer association generally and not to specific association with delinquent peers per se. The reliability for this scale was 0.69. This variable was responsible for the greatest number of lost cases for this analysis as 24.7 percent (427) of the respondents did not answer one or more of the items.
Risk-Taking Behaviour: The dependent measure was constructed with five dimensions which include delinquent behaviour (including self-reported theft and assault), as well as the consumption of alcohol, tobacco, cannabis, and hard drugs. To test whether these items were in fact dimensions of an underlying non-specialized type of risk behaviour, a principal components factor analysis was performed. The factor analysis revealed that the separate dimensions loaded highly on one factor, supporting the non-specialization theorem that these behaviours were all dimensions of a more global measure of risk-taking behaviour. The multi-dimensional measure of risk-taking behaviour was created by standardizing and adding together the five behaviours. The reliability for this global measure was 0.72. A total of 213 (12.4%) respondents did not answer one or more of the items that constructed this scale and were deleted from the analysis.
Three dimensions of the dependent variable that measured tobacco, alcohol and cannabis use were constructed based on the frequency of use in the last 12 months. The categories "did not know," "do not use," and "did not use in the last 12 months" were recoded to equal no use. For alcohol, the categories "for special events only" and "less than once per month" were recoded to equal 1. Categories indicating greater consumption were coded to reflect increasing quantity or frequency of use.
The dependent measure also included a dimension on "hard" drug use including usage of glue, solvents, barbiturates (with or without a prescription), heroin, speed, stimulants (with or without a prescription), tranquilizers (with or without a prescription), LSD ("acid"), PCP ("angel dust," "horse tranquilizer," etc.), other hallucinogens such as mescaline and psilocybin ("mesc," "magic mushrooms," etc.), cocaine, crack, and ecstasy. The prevalence for each individual drug tended to be small. As a result, it was necessary to combine these different drugs to create one variable based on the usage versus non-usage of the various hard drugs in the last twelve months. Individual items were recoded to create dichotomous measures where usage = 1. The individual items were then summed to create a total measure of the number of different "hard" drugs used by the student in the previous 12 months.
The final dimension was a measure of delinquent behaviour. It was based on a summative measure of the 12-month prevalence of various delinquent behaviours such as self-reported petty theft, grand theft, trafficking in cannabis or hard drugs, assault, vandalism, breaking and entering, carrying a weapon, and joy riding. These items were also dichotomized to indicate whether the student participated in the behaviour and then all items were summed to create a total count of the different types of delinquent behaviours.
The correlations, means and standard deviations of the variables of interest are presented in Table 1. This table shows the zero order relationships between
[Part 1 of 2] Table 1. Zero-Order Correlations, Means, and Standard Deviations for all
Variables 1 2 3 1. Age - 2. Sex (1) -0.176 - 3. Family Financial Situation -.0453 -.0504 - 4. Family Structure (2) .1032(3) -.0324 -.1935(3) 5. Family Attachment -.1909(3) .0650 .1196(3) 6. School Attachment .0210 .1695(3) .0928(3) 7. Peer Attachment -.1297(3) -.0738(3) .0255 8. Risk-Taking Behaviour .2534(3) -.1407(3) -.0275 Mean 15.49 0.54 3.49 S.D. 2.32 0.50 0.83 [Part 2 of 2] Table 1. Zero-Order Correlations, Means, and Standard Deviations for all
Variables Variables 4 5 6 7 8 1. Age - 2. Sex (1) 3. Family Financial
4. Family Structure (2) - 5. Family Attachment -.1608(3) - 6. School Attachment -.1063(3) .3212(3) - 7. Peer Attachment .0782(3) -.1262(3) -.2342(3) - 8. Risk-Taking Behaviour .1546(3) -.3990(3) -.2851(3) .2065(3) - Mean 0.25 0.41 3.94 -0.07 - 0.22 S.D. 0.43 3.42 3.60 2.42 3 .27 N = 1075 (1) Sex is coded as a dummy variable, males = 0 and females = 1. (2) Family Structure is coded as a dummy variable, two natureal paren ts = 0, others = 1. (3) p < .01.
variables for those students who answered the "B" interview schedule. T-tests (not shown) were computed to compare students across the two interview schedules on variables where this was possible. The only significant difference across the interview schedules was the age of respondents (schedule "A" = 15.26 yr., schedule "B" = 15.52 yr.; p = .001). Child-parent relations as well as peer and school attachment were unique to the "B" interview schedule and could not be tested.
Since there was such a large number of lost cases from the scale that measured peer relations, we also performed a missing case analysis comparing these missing cases to those who responded to the individual items. The respondents differed significantly from missing cases on age (15.34 yr. versus 16.08 respectively; t = -5.70, p < .001) and sex ratio (47% of valid cases versus 66% of missing cases were male; chisq = 10.67, p < .01).
The analysis proceeded in three stages. First, we examined the effects of the predictors on risk behaviour in a hierarchical regression analysis. A hierarchical regression technique allowed us to assess how subsequent variables mediate the relationship between earlier predictors and risk-taking behaviour. Second, we present a causal path diagram to illustrate the pathways between family structure and family, school, and peer attachment on risk behaviour. This model presents a causal framework to better understand the processes of family, peers and school on risk-taking suggested by our theory. Finally, we investigate the potential moderating effects of school and peers on the relationship between family and risk-taking behaviour.
Table 2 presents the results of the hierarchical regression analysis. Model 1 demonstrates that both sex and age were powerful predictors of risk-taking behaviour. That is, males and older adolescents were more likely to participate in these forms of behaviour. This model accounted for just over 8 percent of the explained variance. Since we used the unweighted data for this analysis, any conclusion about the effects of age on risk-taking should be made with caution.
In Model 2 we added family structure and family financial situation. Age and sex remained significant predictors after controlling for these variables. In addition, a child who came from a family that had both natural parents was much less likely to participate in these delinquent activities than children who lived in other types of families. Finally, the financial situation of the family did not have any independent significant effect on risk-behaviour after controlling for family structure and the age and sex of the adolescent. The inclusion of these structural variables only increased the variance explained by 1.5 percent beyond the effect of age, sex, and region.
In Model 3 we included the measure for family attachment. This variable proved to be the most important predictor of risk-taking behaviour. The better the relationship between the child and his/her parents, the less likely the child was to be involved in risky behaviours. The inclusion of this measure increased the explained variance by 11.4 percent (9.8% to 21.2%). In addition, it also reduced significantly the effect of family structure on risk-behaviour. Finally, in Model 4, we included school and peer attachment. Both measures were significant predictors of risk behaviour and their effects were independent and additive beyond the contribution of family relations. The greater one's commitment to school, the less likely one was to be involved in risky behaviour. Conversely, the greater the association with peers, the more likely the adolescent was to be involved in these behaviours. The inclusion of these two variables increased the explained variance by an additional 5.8 percent. Furthermore, the addition of these two variables in conjunction with family relationships mediated completely the effect of family structure on risk-taking behaviour. The age and sex of the child remained significant even after controlling for family, school and peers.
The hierarchical regression analysis assessed the relative strength of each predictor on the likelihood of adolescent participation in risk-taking behaviour. However, it did not give any clues as to the causal pathways in which these processes may occur. To examine the causal pathways between variables, we employed a path analytic technique to test the proposed path model presented in Figure 1.
Figure 2 presents the significant regression effects of family, school and peers on the degree of risk-taking behaviour among adolescents. It demonstrates
Table 2 Adolescent Risk-Taking Behaviour Regressed on Family Structur e, Family Financial Situation, and Family, School, and Peer Attachments (N = 1075) (3) Model 1 Model 2 Model 3 Model 4 [beta] [beta] [beta] [beta] Age .250(2) .238(2) .178(2) .216(2) Sex (4) -.136(2) -.132(2) -.111(2) -.079(1) Family Structure (5) .126(2) .083(1) .064 Family Financial Situation .001 .033 .034 Family Attachment -.348(2) -.280(2) School Attachment -.147(2) Peer Attachment .153(2) R2 .083 .098 .212 .260 (1) p < .01; (2) p < .001. (3) All coefficients are standardized. All equations controlling for
(4) Sex coded as male = 0, female = 1. (5) Family structure coded as two natural parents = 0, other = 1
that family structure had a direct effect on both the family financial situation and on the relationship between the student and his/her parent(s). Families where there were two natural parents versus other family types were in a better financial situation and were found to have stronger child-parent attachments. Family financial situation also had a direct effect on family attachment. Thus, the better the economic well-being of the family, the stronger the family relationships between child and parent(s). As predicted, the structural variables had no direct effect on risk-taking. However, unlike Sampson and Laub, we observe that family structure and financial situation did not have any direct effect on school attachment or peer attachment. The only effect of these variables was mediated through the child-parent relationship. This suggests that even with structural and financial deficiencies at home, children were not destined by economics nor non-traditional families to become risk-takers. Instead, we found that family attachment was the pivotal factor in determining the likelihood of engaging in risk-taking behaviour. This confirms the central importance placed on family processes by Gottfredson and Hirschi.
Finally, both school attachment and peer association had a significant independent effect on risk-taking behaviour apart from the family relationship. Those students with a positive attitude towards school were less likely to engage in risk-taking behaviour. Conversely, those with a stronger commitment to peers were more likely to engage in these types of behaviours. It is interesting to note, however, that the inclusion of these two sources of social capital did not reduce substantially the strong effect of family attachment on risk-taking.
Since school and peer attachment did not mediate the effect of family attachment on risk-taking, we examined the possibility that these variables may instead moderate this relationship. We tested the interaction effects of both peer and school attachment on the relationship between family relations and risk-taking behaviour. This allowed us to explore whether peer or school attachment conditions the effect of family relationships on risk-behaviour. The data suggest that they do. We found that both school attachment (b = .029; t = 4.487, p < .001) and peer attachment (b = -.033; t = -3.534, p <.001) significantly moderated the effect of child-parent relations on risk-taking behaviour. (1) This suggests that while family attachment is a vital predictor of risk-taking behaviour, its effect is conditional upon the adolescent's attachment to both school and peers. This effect remains independent of sex and age since all three-way interactions tested were nonsignificant. Thus, the buffering effects of school and peers on child-parent relationships were not conditional upon the sex or age of the student.
Figure 3 illustrates the effect of school attachment on the relationship between the family relationship and risk-behaviour. Those adolescents who had a positive relationship with their parents were much less likely to partake in risk-taking regardless of their attachment to school. Students who did not have a positive home environment but had a positive attachment towards school had only a slight increase in their likelihood of engaging in risk-taking. However, those students with a weak family attachment in addition to a weak attachment to school were the most likely to engage in risk behaviours. While it is apparent that one's family relationship is crucial, a positive school environment appears to compensate, albeit only partially, for the hazards of a weak child-parent relationship.
Figure 4 presents the interaction effect between peer attachment and child-parent relations on risk-taking behaviour. Those with positive home environments were much less likely to partake in risk-taking behaviours and were much less affected by their peers. Those with negative home environments were more likely to partake in risk-behaviours. However, those who lacked a strong peer attachment were not near as likely to partake in these activities as those with a strong peer association. Thus, while a negative relationship with parents is a crucial predictor for risk-taking behaviour, it also appears that one's social connection with peers is important.
Control theory gives primary importance to low self control as the disposition which supports impulsive choices, including risk-taking behaviours. Parent-child relationships are decisive in socializing children with strong self-control. However, our research also presents compelling evidence that the institution of the family does not operate in a vacuum. Figures 3 and 4 readily illustrate the contribution of school and peers. For both sources of social capital, the effect was most dramatic when family attachment was weak. Thus, when the prophylactic effect of high family attachment is absent, school and peers, which play a central part in the daily lives of adolescents, are far more consequential. (2)
In our analysis, we have inferred a causal ordering which treats family processes as prior and subsequent to peer and school influences. While this is consistent with control theory, the data used are cross-sectional and cannot examine the sequential processes directly. However, in our approach we follow the initiative of Gottfredson and Hirschi (1990) and Hirschi and Gottfredson (1995) who argue that the problem is conceptual, not empirical and that longitudinal research has succeeded only in confirming "the results of cross-sectional designs and as having shown that the methodological criticisms of cross-sectional family research were not justified" (1990: 230). Loeber and LeBlanc (1990) agree, providing additional support for this assertion. They argue that longitudinal design have only served to confirm the findings of previous cross-sectional studies on the correlates of juvenile delinquency. The primary requirement is a causal specification based on conceptual issues and this analysis is firmly based on the strong model presented by Sampson and Laub (1993), using the Glueck and Glueck longitudinal data set. The data prompted Sampson and Laub to suggest that family, school, and peer attachment were key causal determinants of delinquent behaviour and this is the specification that guides this analysis.
The theoretical significance of our findings revolve around the stability theorem. If the effect of schooling and peers were only a proxy measure of impulsiveness we would not predict any interaction effect for school and peer attachment. The fact that high school and peer attachment are more important when a low family attachment exists suggests a more dynamic role for non-familial institutions in the prevention of delinquency. We would have to conclude that the impact of schooling and peers appears to be underestimated by the heterotypic continuity model.
In this work, we have focused on the sociogenic correlates of delinquency. A common criticism of self-control theory is that it places the linchpin of the explanation on individual level differences in self control despite the systematic association of crime with social class and race. By focusing on the family instead of individual traits we keep in sight the structural components which mediate individual dispositions. This is consistent with a control perspective which recognizes that both criminality (i.e., impulsiveness) and social structure (i.e., criminal opportunities) influence criminal acts. Families, schools and associates may not only be instrumental in teaching discipline and inculcating self control. They also structure opportunities independent of the person's general dispositions.
If we understand self control theory correctly, impulsivity is not a fixed psychiatric disorder like schizophrenia which frequently makes those who suffer from it socially dysfunctional. It is a disposition, and as such, it can be tempered by effective social structures and institutions (Gibbs, 1989). Sampson and Laub made the case for adults with long criminal histories. Our data suggest that attachments to schools and peers can similarly moderate underlying dispositions for adolescents, although, like Sampson and Laub, we have only indirect evidence of such dispositions. In the case of schools, even where attachment to families points to divisiveness or distance between parents and children -- situation which ought to contribute to risk -- this is significantly off-set where the child is positively attached to schools.
And in the case of strong bonds within families which ought to predict low levels of risk, this can be elevated by a high investment in peers -- "hanging out" and spending a lot of evenings with friends out of the house. Self control theory discounts the contribution of peers to delinquency since self reported measures of delinquent acts by peers are effectively measures of the respondent's own acts of delinquencies. We avoid this criticism because in this study the measures of peer attachment employed was an investment of time with friends per se and not association with delinquent peers. While self control discounts the contribution of peers to crime, Gottfredson and Hirschi (1990: 158) do acknowledge that "some criminal acts are facilitated by group membership or a group context". Crimes may be committed by young people who normally restrain themselves alone, or alternatively whose risk taking is part of a group's collective recreational activities (Sampson and Laub, 1995). Acknowledging this does not require a dismissal of the primary role of the family in establishing self control. It merely stresses the relative importance of schools and peers, and these ties probably structure opportunities that either inhibit or exacerbate risk taking behaviour without changing the child's disposition.
These implications may be of some relevance in a policy framework. What Barlow calls "the stability postulate" (1991: 235) or the "stability problem" (Gottfredson and Hirschi, 1990: 107) paints a bleak picture of long term impulsivity once low self control has been established in young children. This deterministic approach does not acknowledge the potential of change due to sources of other types of social capital. In order to account for this possibility, we need to advance a "developmental criminology" (Loeber and LeBlanc, 1990) or "life course perspective" (Sampson and Laub, 1995) which captures the transitions in risky conduct over the life cycle. What our data suggest is that the family, while centrally important, is not the only institution which can make a difference in minimizing risky behaviour. We share the conclusion of Sampson and Laub (1993: 246) that "family and school processes of informal social control provide the key causal explanation of delinquency in childhood and adolescence." In future research we need to pay further attention to the interaction between families and other institutions, particularly at different stages in the life cycle if social policies are to evolve to intervene in risk-taking conduct effectively in a way that is consistent with human development. In addition, we need to determine, first, why tendencies which are acquired so early in life are so resistant to subsequent alteration. Second, we need to determine the precise age-sensitive windows of opportunity for effective intervention. And third, we need to identify the psychosocial processes (discipline, integrative shaming, identification, etc.) which switch on such global dispositions as self control when the person -- young or old -- is open to such changes.
Adlaf, Edward M., Reginald G. Smart and Gordon W. Walsh
1993 Ontario Student Drug Use Survey, 1977-1993. Toronto: Addiction Research Foundation of Ontario.
Adlaf, Edward M., Frank J. Ivis, Reginald G. Smart and Gordon W. Walsh
1995 Ontario Student Drug Use Survey, 1977-1995. Toronto: Addiction Research Foundation of Ontario.
Arneklev, Bruce J., Harold Grasmick, Charles Tittle and Robert J. Bursik
1993 "Low self control and imprudent behaviour." Journal of Quantitative Criminology 9(3): 225-247.
Barlow, Hugh D.
1991 "Explaining crimes and analogous acts, or the unrestrained will grab at pleasure whenever they can." The Journal of Criminal Law and Criminology 82: 229-242.
1989 Crime, Shame and Reintegration. Cambridge: Cambridge University Press.
1955 Delinquent Boys: The Culture of the Gang. New York: Free Press.
1988 Statistical Power Analysis for the Behavioural Sciences, 2nd Edition. New York: Academic Press.
1995 Beggars and Thieves: Lives of Urban Street Criminals. Madison: University of Wisconsin Press.
1989 Control: Sociology's Central Concept. Urbana: University of Illinois Press.
Glueck, Sheldon and Eleanor Glueck
1950 Unraveling Juvenile Delinquency. New York: The Commonwealth Fund.
1968 Delinquents and Nondelinquents in Perspective. Cambridge: Harvard University Press.
Gottfredson, Michael and Travis Hirschi
1990 A General Theory of Crime. Stanford: Stanford University Press.
Groves, R. M.
1989 Survey Errors and Survey Costs. New York: Wiley.
1969 Causes of Delinquency. Berkeley: University of California Press.
Hirschi, Travis and Michael Gottfredson
1995 "Control theory and the life-course perspective." Studies on Crime and Crime Prevention 4: 131-142.
Jaccard, James, Robert Turrisi and Choi W. Wan
1990 "Interaction effects in multiple regression." Sage University Paper Series on Quantitative Applications in the Social Sciences, 07-072. Newbury Park, CA.: Sage.
1988 Seductions of Crime. New York: Basic Books.
Loeber, Rolf and Magda Stouthamer-Loeber
1986 "Family factors as correlates and predictors of juvenile conduct problems and delinquency." In M. Tonry and N. Morris, Crime and Justice: An Annual Review of Research, Vol. 7, 29-149. Chicago: University of Chicago Press.
Loeber, Rolf and Magda LeBlanc
1990 "Toward a developmental criminology." In M. Tonry, and N. Morris, eds., Crime and Justice: An Annual Review of Research, Vol. 12, 375-437. Chicago: University of Chicago Press.
Nagin, Daniel and David P. Farrington
1992 "The onset and persistence of offending," Criminology 30 (4) 501-23.
1979 "Some child rearing antecedents of criminal behaviour in adult men." Journal of Personality and Social Psychology 37: 1477-1486.
Riley, David and Michael Shaw
1985 "Parental supervision and juvenile delinquency." Home Office Research Study No. 83. London: HMSO.
Rosenthal, Robert and Elenore Jacobson
1968 Pygmalion in the Classroom. New York: Holt.
Sampson, Robert J. and John H. Laub
1993 Crime in the Making: Pathways and Turning Points Through Life. Cambridge: Harvard University Press.
1995 "Understanding variability in lives through time: Contributions of life-course criminology." Studies on Crime and Crime Prevention 4: 143-158.
Smith, Carolyn and Terence P. Thornberry
1995 "The relationship between childhood maltreatment and involvement in delinquency." Criminology 33 (4) 451-77.
1939 Principles of Criminology. Philadelphia: Lippincott.
West, David and David Farrington
1977 The Delinquent Way of Life. London: Heinemann.
Wineburg, Samuel S.
1987 "The self-fulfilment of the self-fulfilling prophecy." Educational Researcher December: 28-37.
Zingraff, Matthew, Jeffery Leiter, Kristen A. Myers and Matthew C. Johnsen
1993 "Child maltreatment and youthful problem behavior," Criminology 31 (2) 173-202.
(1) . Reported effects are unstandardized regression coefficients. Moderating effects were computed by adding each interaction term separately to the full regression model, Model 4 in Table 2.
(2) . The test of interaction effects between family, school and peers provides additional evidence for how social capital works conjointly with the family to influence risk behaviour. Sampson and Laub (1993) explain that school and peer attachment both exert strong independent effects on risk-taking. Furthermore, while they did test higher order interactions, they discount their importance and provide no indication as to whether these tests were significant because "the interaction terms were highly correlated with constituent terms and did not improve the overall explanatory power of the model" (Sampson and Laub 1993: 273). A high correlation between an interaction variable and its constituent items as well as a negligible change in the variance explained can be expected since an interaction variable is a simple multiplicative combination of variables already included in the model. However, this alone does not discount the substantive importance of an interaction variable if it is significant (cf. Jaccard, Turrisi, and Wan 1990). The inclusion of an interaction term examines whether the effect of one primary variable is independent of or conditional upon the level of another primary variable. In this analysis we found strong evidence to support the conditional nature of both school and peer attachment on risk-taking dependent upon the level of family attachment. Thus, the effect of these elements of social capital are not independent and must be interpreted with respect to the level of family attachment.
|Printer friendly Cite/link Email Feedback|
|Author:||Wade, Terrance J.; Brannigan, Augustine|
|Publication:||Canadian Journal of Sociology|
|Date:||Dec 1, 1997|
|Next Article:||Too close for comfort? Parental assessments of "boomerang kid" living arrangements.|