Abstract
We examined how parenting is directly and indirectly associated with adolescent delinquency. We derived four possible mechanisms from major criminological theories and examined their relative contribution to explaining the relationship between parenting and delinquency: self-control theory (that is, self-control), differential association theory (that is, delinquent attitudes and peer delinquency), and routine activity theory (that is, time spent in criminogenic settings). In addition, we examined how changes in different aspects of parenting during adolescence were directly and indirectly related to changes in delinquency. Results of multilevel structural equation modeling on two waves of panel data on 603 adolescents indicated that parenting was indirectly related to delinquency through self-control, delinquent attitudes, peer delinquency, and time spent in criminogenic settings. However, only when examined together these variables, derived from major criminological theories, almost fully mediate the effects of parenting. Furthermore, changes in parenting during adolescence were indirectly related to changes in delinquency through changes in delinquent attitudes and in peer delinquency.
Keywords
Introduction
A large body of research has demonstrated that parenting is related to adolescent delinquency (for a meta analytical overview, see Hoeve et al., 2009). In the literature, however, the attention has been given to the mechanisms that might explain this relationship. Parents might be indirectly protective for involvement in delinquent behavior. In the present study we examined the extent to which the most important mechanisms, stemming from three major criminological theoretical approaches (self-control theory, differential association theory, and routine activity theory), uniquely contribute to explaining the associations between parenting and adolescent offending.
According to self-control theory (Gottfredson and Hirschi, 1990), self-control is the key cause of crime involvement. People with lower levels of self-control are more impulsive, tend to engage in risk-taking activities, and prefer easy or immediate gratifications of desires compared with people with higher levels of self-control (Gottfredson and Hirschi, 1990; Grasmick et al., 1993). The theory states that low levels of self-control imply the inability to refrain from delinquent behavior when temptations or provocations to engage in criminal behavior are present.
Whereas Gottfredson and Hirschi (1990) stressed the importance of self-control in explaining engagement in delinquent behavior, differential association theory (Sutherland, 1947) proposed that delinquent attitudes and peer delinquency are the most important factors related to delinquent behavior. Delinquent attitudes refer to an individual’s views about whether delinquent acts are acceptable or unacceptable. The more an individual holds attitudes that approve of delinquent behavior, the more likely he or she is to engage in delinquent behavior.
Although differential association is not limited to interaction with peers, peer associations are a major part of the differential association process, especially during adolescence. Differential association refers to an individual’s exposure to attitudes that are more or less favorable towards delinquent behavior and implies that these attitudes are learned in interaction with others (Sutherland, 1947). Delinquent peers are expected to be an important source of adopting delinquent attitudes, and therefore related to an individual’s delinquent behavior.
An alternative explanation for engagement in delinquency is offered by routine activity theory (Cohen and Felson, 1979). According to the routine activity perspective, opportunities that arise in routine everyday life are central in explaining criminal behavior. Certain settings provide opportunities for delinquent behavior, and the degree of involvement in delinquent behavior depends on the amount of time a person spends in these criminogenic settings. The social and environmental characteristics of settings provide temptations, opportunities, and controls that make delinquent behavior more or less attractive (Felson and Boba, 2010; Wikström et al., 2012). Osgood et al. (1996) applied the routine activity perspective to explaining individual offending by time spent unsupervised with peers in unstructured activities. The presence of peers is believed to make criminal behavior more rewarding, and the absence of adult supervision indicates low social control over the potential offender. Furthermore, unstructured socializing is suggested to leave time available for delinquent behavior because it provides few constraints on how time is spent (Osgood et al., 1996).
In addition to these social characteristics of settings, environmental characteristics are also expected to provide opportunities for crime. The presence of signs of disorder (for example, litter, graffiti, decaying houses) may communicate lack of control over an area, which might reduce the perceived risk of being caught when committing a crime (Felson and Boba, 2010; Sampson and Raudenbush, 2004). Keizer et al. (2008) offer another explanation why settings with higher levels of disorder are criminogenic. In settings where norms and rules are violated, the concern for appropriate behavior weakens, which results in more violations of norms and rules.
Whereas the theoretical perspectives focus on different factors as the most important direct cause of delinquency, self-control theory and differential association theory acknowledge parenting as important indirect cause. The role of parenting is, however, not directly addressed by the routine activity perspective. According to Gottfredson and Hirschi (1990), the development of self-control is a result of parental socialization during early childhood. More specifically, parental monitoring, discipline, and support are necessary to foster self-control in young children. It has been demonstrated that the relationship between parenting and delinquency is partially mediated by the level of self-control (Burt et al., 2006; Hay, 2001; Perrone et al., 2004; Unnever et al., 2006; Vazsonyi and Belliston, 2007).
Similarly, the differential association perspective also contends that parental monitoring and providing consequences for misbehavior are important indirect causes of delinquency. Individuals who receive less parental monitoring and control are more likely to acquire delinquent attitudes and to engage with delinquent peers, and are therefore more likely to engage in delinquent behavior. Previous research indeed indicated that adolescents who received more effective parenting are less likely to have delinquent attitudes and delinquent peers (Knoester et al., 2006; Pardini et al., 2005; Ragan et al., 2014; Warr, 2005). By supervising and monitoring the child, parents may intentionally or unintentionally encourage the child to associate less with delinquent peers. Adolescents with a high-quality relationship with their parents may be more likely to seek out friends who their parents will like to avoid parental disapproval or disappointment (Knoester et al., 2006; Warr, 2005). In addition, previous work suggests that parenting has a possible spillover effect, meaning that parents affect not only the behavior of their children but possibly also the behavior of their children’s peers (Shakya et al., 2012).
Although the role of parents is not directly elaborated in the routine activity perspective, parents are expected to restrict their children from spending time in criminogenic settings in order to keep them out of trouble (Felson and Boba, 2010). Osgood and Anderson (2004) have shown that parental monitoring is negatively related to unstructured socializing with peers. A recent study has shown that adolescents who have a relationship of high quality with their parents and who perceive more parental monitoring and limit-setting spend less time in criminogenic settings (Janssen et al., 2014).
The first aim of the present study is to determine the relative contribution of the proposed mediators (self-control, delinquent attitudes, delinquent peers, and time spent in criminogenic settings) to explaining the relationship between parenting and adolescent delinquency. Previous research has shown that the proposed mediators of the relationship between parenting and adolescent delinquent behavior are interrelated. For example, adolescents with delinquent peers are more likely to have delinquent attitudes and to seek out situations to gain opportunities to engage in delinquency (Bernburg and Thorlindsson, 2001; Pardini et al., 2005). Because of this interrelatedness, it is important to examine these mediators simultaneously in order to determine their relative contribution to explaining the relationship between parenting and adolescent delinquency. We examine how various dimensions of parenting (that is, parental monitoring, parental limit-setting, and the quality of the parent–adolescent relationship) are related to self-control, delinquent attitudes, peer delinquency, and time spent in criminogenic settings, and how these factors are in turn related to involvement in delinquency. We use the most important mechanisms as derived from self-control theory, differential association theory, and routine activity theory in order to test whether these mechanisms are interrelated and uniquely contribute to explaining how parenting is related to adolescent delinquency.
The second aim of the present study is to examine, in addition to differences between individuals, the extent to which changes over time in self-control, delinquent attitudes, peer delinquency, and time spent in criminogenic settings can account for the relationship between changes in parenting and changes in delinquency over time. This is of specific importance when considering the period of adolescence, which is characterized by many changes at the individual and contextual levels (Mulvey, 2014; Smetana et al., 2006; Steinberg and Silk, 2002). Naturally, adolescents strive for more freedom and independence from their parents. Peers become more important and adolescents spend more time away from home and their parents (Smetana et al., 2006). The parent–child relationship changes during adolescence as parents have to learn to allow more freedom and independence and at the same time continue to be supportive (Galambos et al., 2003; Simons et al., 2007). Furthermore, it has been demonstrated that self-control might be not as stable as Gottfredson and Hirschi (1990) predicted (Burt et al., 2014) and found that changes in levels of self-control could be explained by changes in social factors, including parenting (Burt et al., 2006). During adolescence, individuals develop increasingly tolerant views about delinquency (Pardini et al., 2005). Association with delinquent peers is also likely to change over time. In general, exposure to delinquent peers peaks in adolescence and decreases when entering adulthood (Warr, 1993). Osgood et al. (1996) have shown that within-individual changes in spending time unsupervised and unstructured with peers were related to changes in delinquency.
Present study
In the present study we examined (1) the extent to which four components derived from different criminological theories mediate the associations between parenting and adolescent delinquency (between-person) and, (2) the extent to which changes in these factors over time mediate the associations between changes in parenting and changes in delinquency over time (within-person). We expect to find that all concepts derived from competing theoretical frameworks contribute to explaining the relationship between parenting and adolescent delinquency.
In the present study we attempt to contribute to the theoretical understanding of the relationship between parenting and delinquency by examining the most important mechanism from three major criminological theories simultaneously. Although a fairly large number of studies have already examined various mediating mechanisms, including self-control, delinquent attitudes, and peer delinquency, time spent in criminogenic settings has not been examined previously as a mediator between parenting and delinquency. The present study focuses on the relative contribution of the proposed mechanisms to explaining the relationship between parenting and delinquency, because we know from previous studies that the mediators are interrelated. By examining the mechanisms simultaneously, we can determine whether they all explain a portion of the association between parenting and delinquency.
In addition, we advance our theoretical understanding of the relationship between parenting and delinquency by examining the extent to which the mechanisms derived from self-control theory, differential association theory, and routine activity theory are able to explain the effects of changes over time in parenting on changes in delinquency. Instead of using a static approach, the present study applies a dynamic approach in which change is possible. Because the parent–child relationship is likely to change during adolescence (Galambos et al., 2003; Simons et al., 2007), it is important to examine the extent to which these mechanisms can explain how these changes in parenting are related to changes in delinquency.
In addition to these theoretical contributions, the present study makes methodological contributions as well. First, related to the above-mentioned theoretical contribution, we examined both between-person differences and within-person differences. Therefore we are able to examine how differences between adolescents are related to differences in delinquency, but also how changes in parenting over time are related to changes in delinquency over time. Whereas a majority of previous longitudinal studies usually examined how risk factors at T1 are related to delinquency at T2, examining how within-person changes in risk factors over time are related to changes in delinquency over time could produce a more refined picture (Mulvey, 2014).
Second, we examined three parenting dimensions simultaneously to determine their relative contribution to explaining adolescent delinquency (Simons et al., 2007). In general, three parenting dimensions can be derived from the literature as important for explaining adolescent delinquency: parental monitoring, which entails efforts by parents to track their children’s behavior, parental limit-setting, which involves setting rules and providing consequences for misbehavior, and the quality of the parent–adolescent relationship (Smetana et al., 2006; Wright and Cullen, 2001).
Third, most previous studies used questionnaires about how many hours per week in general adolescents spent with peers away from home (Osgood and Anderson, 2004; Osgood et al., 1996; Osgood et al., 2005; Siennick and Osgood, 2012). In the current study, time spent in criminogenic settings is measured with greater precision and more detail using space–time budget data that recorded hour by hour where respondents were, what they were doing, and who they were with during a period of four days. Because the geographical small-scale location of each hour was known, we were able to enrich these space–time budget data with information about the level of physical disorder, which was conducted using systematic social observation (Sampson and Raudenbush, 1999). We used small geographical units of analysis of 200 by 200 meters (0.04 square kilometers) because they more closely approximate behavior settings and are more likely to be homogeneous in terms of environmental characteristics (Oberwittler and Wikström, 2009).
Method
Sample
The Study of Peers, Activities and Neighborhoods (SPAN) is a longitudinal study conducted by the Netherlands Institute for the Study of Crime and Law Enforcement (NSCR). The SPAN data consist of two waves of data collection among adolescents (11–17 years of age at T1) from The Hague and neighboring suburbs in the Netherlands; 40 secondary schools were approached and 10 agreed to participate in the study. The first wave of data collection took place in 2008/9 and the second wave in 2010/11. In total, 615 adolescents (52 percent boys) participated fully in both waves of the study.
Three data sources from the SPAN project are used. A self-report questionnaire is used to measure delinquency, parenting, self-control, delinquent attitudes, and delinquent peers. The questionnaire was individually conducted in groups of four adolescents during a school hour of about 45 minutes, supervised by a research assistant. In addition, space–time budget interviews combined with systematic social observation are used to measure time spent in criminogenic settings. The space–time budget interview is a structured personal interview, which was conducted individually and face to face with the respondents, and is used to measure the social characteristics of time spent in criminogenic settings. The instrument was originally developed by Wikström and Butterworth (2006) in the Peterborough Youth Study and refined in its successor, the Peterborough Adolescent Delinquency Study (PADS+). During the interview, the activities of the adolescent during each hour of four recent days were recorded (always including the previous Friday and Saturday), including the nature of the main activity (for example, sports, learning), the function of the place (for example, soccer field, school), persons present in the setting (for example, teacher, parents), and the geographical location (see also Bernasco et al., 2013).
Additionally, to assess the level of physical disorder of the settings where the adolescents spend time, systematic social observation was carried out during the first half of 2012. A grid of 200 by 200 meters that overlaid the maps of The Hague, on which the respondents indicated their locations, was used to select the street segments for the systematic social observation. As trained observers walked the street, they completed an observation checklist, which was based on the instrument used by Raudenbush and Sampson (1999), capturing physical disorder by 13 items (for example, ‘How much trash or broken glass is on the street or sidewalks?’). All items had three response categories (none, one, and more); alpha was .62. 1 A setting was indicated as highly disordered when it belonged to the top 25 percent of locations with the highest scores on physical disorder (see also Janssen et al., 2014). 2
Time spent in criminogenic settings could not be determined for all respondents because not all hours covered by the space–time budget are spent in the geographical study area. 3 This resulted in a final sample of 603 respondents. The majority of the respondents are of Dutch origin (53 percent); the sample also included a relatively high proportion of respondents of non-Western origin (38 percent) and 9 percent of the respondents are of other Western origin.
Dependent variable
Self-reported delinquency was a summary construct based on the scale developed by Wikström and Butterworth (2006) consisting of 20 items asking how often the respondent committed various types of crimes during the past year. The offense types ranged from minor (for example, vandalism) to serious (for example, robbery). For each item, the following response categories were used: 0 times; 1 time; 2 times; 3–5 times; 6–10 times; more than 10 times. These responses were coded 0 through 5, respectively, and then summed. The scale ranges from 0 (that is, zero delinquent acts) to 100 (that is, all 20 acts more than 10 times). Cronbach’s alpha was .88 at T1 and .83 at T2. The percentage of respondents who reported having committed zero acts of delinquency is 36.5 percent (n = 197) at T1 and 29.8 percent (n = 173) at T2.
To complement the frequency-based scale we also constructed a variety scale as recommended by Sweeten (2012). The variety scale is highly correlated with the frequency-based measure of delinquency (ρ = .97 at T1 and T2). The variety scale indicated the number of different types of delinquent behavior an individual reported having committed. The variety scale ranges from 0 to 20, indicating how many of the 20 different delinquent acts the respondent reported having committed at least once during the past year. Cronbach’s alpha was .85 at T1 and .80 at T2. Supplemental analyses using this variety scale produced similar results to those obtained using the frequency scale. The results of the analysis with the variety scale are provided in Appendix B.
Independent variables
Parental monitoring was measured by the use of a summary construct based on the scale developed by Kerr and Stattin (2000; Stattin and Kerr, 2000) consisting of five items asking whether the adolescent has to inform his/her parents about his/her whereabouts (for example, ‘If I go out, my parents want me to tell them where I go, with whom and what I’m going to do’) using a five-point scale from 0 (totally disagree) to 4 (totally agree). Cronbach’s alpha was .77 at T1 and .82 at T2.
Parental limit-setting is a summary construct based on the scale developed by Wikström and Butterworth (2006) consisting of four items that reflect the extent to which parents intervene in rule-breaking behavior (for example, ‘If you had been beating up or threatening somebody at school, your parents would tell you off or punish you’) using a five-point scale from 0 (totally disagree) to 4 (totally agree). Cronbach’s alpha was .62 at T1 and .58 at T2.
The quality of the parent–adolescent relationship was based on the scale developed by Wikström and Butterworth (2006) and measured by seven items (for example, ‘Do you talk to your parents when you have a problem or feel sad about something?’) using a four-point scale from 0 (never) to 3 (every day). Cronbach’s alpha was .68 at T1 and .70 at T2.
Mediators
Self-control was measured by the use of a summary construct based on the scale developed by Grasmick et al. (1993) and consisted of 10 items asking about the respondent’s general behavior (for example, ‘I sometimes find it exciting to do things that may be dangerous’) using a five-point scale from 0 (totally agree) to 4 (totally disagree). Cronbach’s alpha was .75 at T1 and .72 at T2.
Delinquent attitudes were operationalized as the adolescent’s beliefs about the acceptability of several delinquent acts. The construct was based on a scale that was developed by Loeber et al. (1998) and consisted of 16 items asking the respondent about how wrong it would be for someone his/her age to engage in the behavior (for example, ‘Ride a bike through red light’) using a four-point scale from 0 (not wrong at all) to 3 (very wrong). The scale was reversed so that a higher score indicated more delinquent attitudes. Cronbach’s alpha was .91 at T1 and .88 at T2.
Peer delinquency was a summary construct based on the scale developed by Wikström and Butterworth (2006) consisting of six items asking about the amount of delinquent behavior of the adolescent’s peers (for example, ‘How often do your friends steal something from others or from shops?’) using a four-point scale ranging from 0 ((almost) never) to 3 (very often (each week)). Cronbach’s alpha was .82 at T1 and .72 at T2.
Time spent in criminogenic settings was measured as the total number of hours (on the four days covered by the space–time budget interviews) spent in unstructured socializing with peers, without adult supervision in settings with high disorder. For each respondent, we summed the number of hours that met all the following conditions: (1) they were spent with at least one peer, (2) they were spent in the absence of authority figures, (3) they included socializing or ‘hanging around’ as the main activity, and (4) they were spent outside a household setting in an area with high physical disorder.
Control variables
Following the definition of Statistics Netherlands (2014), ethnicity was measured by two dummy variables, with Dutch as the reference category, indicating whether the respondent is of non-Western or Western origin. In the case of mixed ethnicity, the origin of the mother was conclusive in determining the ethnicity of the respondent (Statistics Netherlands, 2014). Gender is measured with a dummy variable, with girl as the reference category, and age at T1 is measured in years. We also included two measures of household structure: a dummy variable that indicated whether the respondent lived in a single-headed household, and a variable indicating family size.
Analytical approach
We applied multilevel structural equation modeling in Mplus (Version 7, Muthén and Muthén, 1998–2012). The multilevel structure consisted of two levels: time at Level 1, which is nested in persons at Level 2. The Level 1 model addressed within-person change in delinquency, whereas the Level 2 model explains time-stable differences between individuals. For each independent variable, two variables were constructed: a between-person variable and a within-person variable. The between-person variables were computed by averaging the scores on the independent variables across both waves for each respondent. The within-person variables specify the deviation from the score at T1 (Hoffman and Stawski, 2009; Snijders and Bosker, 1999). As standard practice in longitudinal multilevel analysis, a dummy variable for wave was included in the models, which indicates average change between the two occasions.
Because the dependent variable (total delinquency frequency) was a right-skewed count variable, negative binomial models were estimated using maximum likelihood estimation with robust standard errors (Hox et al., 2010).
As a first step, we estimated a model predicting delinquency by including the three parenting variables and control variables (gender, age, and ethnicity) as predictors (Model 1). This model tested the direct relationship between (change in) parenting and (change in) delinquency.
In order to determine the relative contribution of the mediators, the analyses to examine the indirect effects were twofold. 4 First, we estimated a single mediation model separately for all four mediators (see Figure 1). Model 2 included self-control as a mediator, Model 3 delinquent attitudes, Model 4 peer delinquency, and Model 5 time spent in criminogenic settings. Second, we estimated a multiple mediation model that included all mediators simultaneously to examine their relative contribution (see Figure 2). In this model, a specific indirect effect represented the ability of the mediator to mediate the effect of parenting on delinquency while controlling for all other mediators (see Preacher and Hayes, 2008).

Single mediation models.

Multiple mediation model.
Results
Descriptive statistics
Means, standard deviations, stability coefficients, and Spearman’s rank correlations between all variables are reported in Appendix A. All correlations between the core theoretical variables were statistically significant and in the expected directions across both waves. Stability coefficients indicated moderate relative stability in all variables, ranging from .254 to .567. Comparing the mean scores of both waves showed that, overall, parental monitoring (t = 3.866, p < .000) and delinquency (t = 2.475, p = .013) decreased between the two waves, whereas delinquent attitudes (t = −7.771, p < .000) and time spent in criminogenic settings (t = −3.664, p < .000) increased over time. There were no significant changes over time in parental limit-setting, the quality of the parent–adolescent relationship, the level self-control, and peer delinquency.
Parenting and delinquency: Direct effects
The results of the multilevel path model including parenting and control variables as predictors of delinquency are presented in Model 1 of Table 1. The between-person results indicate that adolescents who perceive more parental monitoring, more parental limit-setting and a higher-quality relationship with their parents reported being less delinquent. The exponentiated value of the coefficient – the incidence rate ratio (IRR) – for the between-person effect of parental monitoring is 0.92. This indicates that an increase of one unit in parental monitoring is related to an 8 percent decrease in the count of delinquency. The IRR for parental limit-setting is 0.87, which indicates that a one-unit increase in parental limit-setting is related to a 13 percent decrease in the count of delinquency. The IRR for the quality of the parent–adolescent relationship is 0.90.
Single mediation multilevel path models predicting delinquency (N = 603).
Notes: Unstandardized coefficients. All coefficients of the direct paths to delinquency represent changes in the expected log count. Robust standard errors in parentheses.
p ⩽ .05; ** p ⩽. 01; *** p ⩽ .001.
The within-person results show that decreases in parental limit-setting (IRR = 0.92) and in the quality of the parent–adolescent relationship (IRR = 0.94) were related to increases in delinquent behavior over time. Decreases in parental monitoring were not related to change in delinquent behavior.
Single mediation models
Model 2 to Model 5 in Table 1 show the results of the single mediation models, in which only one mediator was included in the multilevel model. The between-person direct effects of all three parenting dimensions remained significantly related to delinquency in all single mediation models, controlled for the respective mediator. The within-person direct effects indicate that decreases in parental limit-setting and in the quality of the parent–adolescent relationship remained significantly related to increases in delinquent behavior, even when controlled for changes in the respective mediator. These findings show that none of the mediators is able to fully explain the associations between parenting and delinquency.
The between-person indirect effects of parenting on delinquency were statistically significant in all models, meaning that the associations between parenting and delinquency are partially mediated by self-control, delinquent attitudes, peer delinquency, and time spent in criminogenic settings. The within-person indirect effects showed that the effects of change in parenting were partially mediated by change in self-control (except for change in parental monitoring), change in delinquent attitudes, and change in peer delinquency. Change in the amount of time spent in criminogenic settings did not mediate the associations between change in parenting and change in delinquency. Gender was significantly related to delinquency in all models, indicating that boys reported higher levels of delinquency than girls. Age and ethnicity were related to delinquency in only some of the models. The measures for household structure were not related to delinquency. In all models the dummy variable for wave was negatively related to delinquency, meaning that, on average, adolescents reported less delinquency at wave 2 compared with wave 1.
Multiple mediation model
In Table 2, results of the multiple mediation model, which included all mediators simultaneously, are reported. The between-person direct effects of parental limit-setting and the quality of the parent–adolescent relationship were no longer significantly related to delinquency. Together, self-control, delinquent attitudes, peer delinquency, and time spent in criminogenic settings fully mediated the relationship between these parenting dimensions and delinquency. Parental monitoring, however, remained directly related to delinquency (IRR = 0.97). The within-person direct effects of parental monitoring and parental limit-setting were also no longer significantly related to changes in delinquency over time, indicating that changes in self-control, delinquent attitudes, peer delinquency, and time spent in criminogenic settings fully mediated these relationships.
Multiple mediation model predicting delinquency (N = 603).
The between-person indirect effects for self-control, delinquent attitudes, and peer delinquency remain statistically significant when including all mediators simultaneously. Adolescents who perceived more parental monitoring, more parental limit-setting, and a relationship with their parents of higher quality were less involved in delinquent behavior. This could partially be explained by their higher levels of self-control, lower levels of delinquent attitudes, and less delinquent peers. Controlling for all other indirect effects, time spent in criminogenic settings no longer mediated the relationship between parenting and delinquency.
The within-person indirect effects showed that the association between change in parental monitoring and change in delinquency was mediated by change in delinquent attitudes and change in peer delinquency. The effects of change in parental limit-setting and change in the quality of the parent–adolescent relationship were mediated by change in delinquent attitudes.
Discussion and conclusion
The main purpose of the present study was to examine how various aspects of parenting are related to adolescent offending. We used the most important mechanisms from three major criminological theoretical approaches to explain the relationship between parenting and delinquency. We used self-control theory (Gottfredson and Hirschi, 1990), differential association theory (Sutherland, 1947), and routine activity theory (Cohen and Felson, 1979). We examined the extent to which associations between parenting and adolescent offending are mediated by the level of self-control, delinquent attitudes, peer delinquency, and time spent in criminogenic settings. Because these mediators are interrelated, it is important to determine their relative contribution to explaining the relationship between parenting and delinquency.
The findings of the present study have shown that the theoretical approaches contain some empirically supported propositions, but the explanatory power of any single mechanism tends to be limited. Pratt and Cullen (2000) have also observed in their meta-analysis that variables derived from differential association theory and self-control theory, when included simultaneously, both contributed to explaining delinquency. In addition to this, the present study indicated that these variables also contribute to explaining the relationship between parenting and delinquency. Whereas in the single mediation models the parenting dimensions remained directly related to delinquent behavior, the results of the multiple mediation model, including all mediators simultaneously, suggest that the impact of these parenting dimensions was almost completely mediated. Although self-control theory and differential association theory are usually seen as competing theoretical frameworks (Pratt and Cullen, 2000), the findings of the present study suggest that they both partially explain the relationships between parenting and adolescent delinquency.
According to self-control theory, more effective parenting results in higher levels of self-control, which in turn is related to lower levels of delinquency (Gottfredson and Hirschi, 1990). The results of the present study support these propositions, as self-control continued to mediate the associations between parenting and delinquency when controlling for competing mechanisms derived from differential association theory and routine activity theory. The findings of the present study are also consistent with the mechanisms proposed by differential association theory (Sutherland, 1947). Adolescents who perceived more effective parenting were less likely to have delinquent attitudes and delinquent peers, and these adolescents were less likely to engage in delinquency. Controlling for the mechanisms proposed by self-control theory and differential association theory, time spent in criminogenic settings, derived from routine activity theory (Cohen and Felson, 1979), did not mediate the association between parenting and delinquency.
Because adolescence is characterized by changes, in addition to differences between persons, we also examined the extent to which changes in parenting over time were related to changes in delinquency and the extent to which these relationships could be explained by changes in self-control, delinquent attitudes, peer delinquency, and time spent in criminogenic settings. The findings of the present study suggest that changes in delinquent attitudes mediated the effect of changes in parental monitoring, in parental limit-setting, and in the quality of the parent–adolescent relationship, whereas changes in peer delinquency over time mediated only the effect of changes in parental monitoring. Changes in self-control and changes in time spent in criminogenic settings did not mediate the effect of changes in the parenting dimensions.
Like any study, the present study has some limitations that need to be addressed. First, all of the measures, except for time spent in criminogenic settings, concern the adolescent’s perception. Although adolescent perceptions of parenting are found to be better predictors of adolescent behavior than actual measurements of parenting (Abar et al., 2014), and adolescents are valid informants regarding their level of self-control (Duckworth and Kern, 2011) and delinquency (Thornberry and Krohn, 2000), there are certainly disadvantages of using the perceptions of the adolescents. For example, it is possible that parents do not change their behavior, rather adolescents perceive less parental control as they begin to engage in delinquency and realize that they can get away with more than they originally thought. In addition, perceptions of peer delinquency might also be problematic. Recent studies have indicated that individuals tend to project their own attitudes and behavior onto their peers and that the effect of peer delinquency therefore is overestimated (Haynie and Osgood, 2005; Young and Weerman, 2013; Young et al., 2014). To overcome this same-source bias, as suggested by Haynie and Osgood (2005), different sources of information, including peers, parents, teachers, and direct observations, should be used to measure parenting and peer delinquency.
Second, the longitudinal data used in the present study consisted of two waves of panel data, with two years in between covering only part of the adolescent period. Future research could include more waves and smaller time intervals between the waves in order to explore development throughout adolescence in more detail.
Third, the alpha level of parental limit-setting is low, particularly at T2 (.58). Although low reliabilities reduce statistical power (see Bacon, 2004), the correlations as well as the path coefficients of parental limit-setting in its relationship with the other variables are in line with the expected direction and previous research (Luthar and Sexton, 2007; Lahey et al., 2008). However, for future research the scale can be improved and extended.
Fourth, we were not able to control for genetic factors in our models. Parenting and children’s behavior could be associated because they are both influenced by genes or other biological factors (Wright and Beaver, 2005). The associations found in the present study may be confounded with genetic influences, and might therefore be overestimated.
Notwithstanding these limitations, a major strength of the present study is the unique dataset from SPAN. The data included the most important concepts from self-control theory, differential association theory, and routine activity theory. The space–time budget data enriched with systematic social observation data made it possible to examine with great detail and precision where and in what conditions adolescents spent their time. Furthermore, the data included multiple parenting dimensions, which made it possible to examine constructs of parental control as well as the quality of the parent–adolescent relationship. Last, the use of longitudinal data made it possible to examine changes over time, which reflects the developmental nature of adolescence.
The findings offered useful insights for understanding the processes that may give rise to offending during adolescence. Adolescents, by increasing in age, become more oriented towards peers and spend more time outside the home (Keijsers et al., 2012; Sullivan, 2014). Although during this developmental period adolescents gain greater freedom and independence from their parents, parents remain important for the continuing socialization of adolescents (Halgunseth et al., 2013; Pardini et al., 2005). The findings of the present study indicate that the aspects of parental control and parental support are both important, and that parenting might indirectly be protective for involvement in delinquent behavior, by affecting self-control, delinquent attitudes, and peer delinquency.
Footnotes
Appendix A
Descriptive statistics and Spearman’s correlations at T1 (n = 580) and T2 (n = 539).
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | M | SD | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1. Parental monitoring | .440 | .438 | .323 | .232 | −.465 | −.366 | −.223 | −.388 | −.221 | 17.35 | 4.05 |
| 2. Parental limit-setting | .433 | .458 | .223 | .219 | −.353 | −.332 | −.192 | −.345 | −.169 | 16.86 | 2.66 |
| 3. Quality of relationship | .179 | .193 | .557 | .286 | −.372 | −.350 | −.111 | −.329 | −.164 | 23.05 | 3.30 |
| 4. Self-control | .183 | .231 | .310 | .548 | −.281 | −.393 | −.209 | −.472 | −.016n.s. | 29.74 | 6.35 |
| 5. Delinquent attitudes | −.358 | −.355 | −.274 | −.303 | .567 | .502 | .215 | .450 | .338 | 46.94 | 9.06 |
| 6. Delinquent peers | −.359 | −.303 | −.275 | −.324 | .514 | .459 | .315 | .517 | .431 | 8.64 | 3.17 |
| 7. Time spent in criminogenic settings | −.232 | −.251 | −.186 | −.142 | .234 | .308 | .254 | .239 | .215 | 0.73 | 1.93 |
| 8. Delinquency | −.315 | −.289 | −.250 | −.463 | .460 | .499 | .215 | .524 | .058n.s. | 5.50 | 9.00 |
| 9. Age T1 | −.156 | −.062n.s. | −.077n.s. | .079n.s. | .029n.s. | .263 | .193 | −.022n.s. | – | 13.90 | 1.65 |
| M | 16.36 | 16.61 | 22.76 | 30.34 | 42.94 | 8.91 | 1.22 | 4.31 | – | – | – |
| SD | 4.56 | 2.25 | 3.42 | 5.76 | 8.17 | 2.85 | 2.43 | 7.15 | – | – | – |
Notes: The values above the diagonal represent correlations and descriptive statistics at T1, the values below the diagonal represent correlations and descriptive statistics at T2, the values on the diagonal represent stability coefficients. All correlations are significant at p < .01, except where indicated with n.s. (not significant). Abbreviations: M = mean, SD = standard deviation
Appendix B
Multiple mediation model predicting variety of delinquency (N = 603).
| Between | Within | |
|---|---|---|
| Direct effects | ||
| Parental monitoring | −.031** (.012) | −.014 (.010) |
| Parental limit-setting | −.014 (.019) | −.017 (.015) |
| Quality of relationship | .000 (.012) | −.011 (.012) |
| Self-control | −.052*** (.008) | −.015 (.008) |
| Delinquent attitudes | .026*** (.007) | .026*** (.006) |
| Delinquent peers | .128*** (.019) | .032** (.012) |
| Criminogenic settings | .055** (.020) | −.003 (.014) |
| Indirect effects | ||
| Parental monitoring → Self-control → Delinquency | −.010* (.004) | .000 (.001) |
| Parental limit-setting → Self-control → Delinquency | −.016** (.006) | −.003 (.002) |
| Quality of relationship → Self-control → Delinquency | −.027*** (.006) | −.005 (.003) |
| Parental monitoring → Delinquent attitudes → Delinquency | −.018*** (.005) | −.007** (.002) |
| Parental limit-setting → Delinquent attitudes → Delinquency | −.013* (.005) | −.019*** (.006) |
| Quality of relationship → Delinquent attitudes → Delinquency | −.014** (.004) | −.012** (.004) |
| Parental monitoring → Delinquent peers → Delinquency | −.023*** (.005) | −.003* (.001) |
| Parental limit-setting → Delinquent peers → Delinquency | −.038*** (.008) | −.005 (.003) |
| Quality of relationship → Delinquent peers → Delinquency | −.025*** (.006) | −.005* (.002) |
| Parental monitoring → Criminogenic settings → Delinquency | −.005* (.002) | .000 (.001) |
| Parental limit-setting → Criminogenic settings → Delinquency | −.006* (.003) | .000 (.000) |
| Quality of relationship → Criminogenic settings → Delinquency | −.003 (.002) | .001 (.003) |
| Control variables | ||
| Wave (Ref. = T1) | – | −.292*** (.053) |
| Gender (Ref. = Girl) | .287*** (.075) | – |
| Age T1 | −.052* (.024) | – |
| Ethnicity (Ref. = Dutch) Non-western | .218** (.079) | – |
| Western | −.054 (.125) | – |
| Single-headed household | .131 (.087) | – |
| Family size | .018 (.027) | – |
Acknowledgements
We thank Per-Olof Wikström for sharing the questionnaire and the space–time budget interview developed in the Peterborough Adolescent and Young Adult Development Study (PADS+); Beth Hardie, Caroline Moul, and Neema Trivedi for helping train our interview staff; and Kirsten Grandia, Evelien Hoeben, and Lieneke Spel for coordinating the data collection and managing the SPAN fieldwork.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Netherlands Organization for Scientific Research (NWO) [grant number 431-09-021].
