Journal of Youth and Adolescence

, Volume 41, Issue 2, pp 121–145

It’s a Two-Way Street: The Bidirectional Relationship Between Parenting and Delinquency

Authors

    • Department of SociologyBaylor University
Empirical Research

DOI: 10.1007/s10964-011-9656-4

Cite this article as:
Gault-Sherman, M. J Youth Adolescence (2012) 41: 121. doi:10.1007/s10964-011-9656-4

Abstract

As the primary socializing institution of youth, the family has long been recognized as important for predicting delinquency. Social control theory focuses on the effects of parental behavior on adolescent delinquency but fails to take into account the effect of adolescent delinquency on parental behaviors. This study addresses this problem by estimating eighteen models examining bidirectional relationships between general, property, and violent delinquency and parental attachment, monitoring, and involvement. The magnitude of both parent and child effects were examined using data from 12,505 youth ages twelve to seventeen who participated in the Add Health study. These youth were an average age of 14 and were predominantly white (65%); just over half (50.42%) were female. Cross-lag regressions showed that while parental attachment has an effect on delinquency, an adolescent’s delinquency also impacts parental attachment, regardless of the type of delinquency. Findings also revealed no significant parental effects of monitoring or involvement on any type of delinquency, and the only child effects revealed for monitoring or involvement were for involvement, which decreases when overall or property delinquency increases. The findings regarding parental attachment provide strong evidence for the existence of a reciprocal relationship between parenting and delinquency, consistent with the transactional and interactional models of reciprocal parent–child relationships.

Keywords

Social bondsDelinquencyParentingAttachment

Introduction

As the primary socializing institution of youth, the family has long been recognized as important for predicting delinquency. Criminological theories, especially social control theories, consider parenting as a, if not the, key predictor of delinquency. Most criminological theories that incorporate parents either assume a unidirectional influence of parent to child or do not thoroughly discuss the possibility of a bidirectional influence. The shortfall of criminological theories that emphasize parenting (e.g., control theory) is that they seldom take into account that children may affect the kind of parenting they receive.

There are several existing developmental theories that attempt to account for this interactive relationship, including transactional theory (Sameroff 1975), coercion theory (Patterson 1982), and interactional theory (Thornberry 1987). The common thread of each of these theories is that the behaviors of parents and children influence each other, causing changes in both over time; the empirical studies of these developmental theories will be discussed later. Within the field of criminology, a small number of researchers recently have found some support for these theories (e.g., Jang and Smith 1997; Liska and Reed 1985; Stewart et al. 2002).

Given the interactive nature of the parent–child relationship, one might expect parents and children to influence each other’s behavior. Liska and Reed (1985) were the first criminologists to test empirically the idea of bidirectionality in the association between parenting and delinquency, arguing that there is every reason for researchers to assume not only does attachment affect delinquency but also that delinquency itself affects attachment. Their view is echoed by other criminologists who have examined this reciprocal relationship. These include studies by Jang and Smith (1997), who found evidence of reciprocal relationships and concluded that previous research findings on the relationship between parenting and delinquency may require reinterpretation from an interactional perspective, and Agnew (1991), who stated that the “failure to consider such [reciprocal] effects means that one’s model is misspecified and that parameter estimates may be biased” (p. 131). Despite the fact that bidirectional effects have been incorporated into developmental theories, and are seen in the work of behavioral geneticists (e.g., Moffitt 2005), such effects have yet to be incorporated into any major criminological theory and only recently have been incorporated into empirical research on delinquency. Indeed, most criminological studies that consider bidirectionality do so by controlling for child effects, such as self-control (Burt et al. 2006; Hay 2001; Simons et al. 2007), rather than directly assessing their magnitude.

The key problem this study addresses is a problem with one of the apparent assumptions of social control theory, which calls into question the cross-sectional results of past studies. Social control theories assume a unidirectional relationship where bonds such as attachment affect later delinquency. Criminological research that actively measures the magnitude of a bidirectional relationship between parent and youth is quite rare, occurring fewer than ten times over the past twenty-four years, though numerous studies control for child effects. With a few exceptions (Agnew 1991; Liska and Reed 1985), studies have used small, select samples or focused on specific types of parents, such as nonresident fathers (e.g. Hawkins et al. 2007). Because of this, the field still has relatively little compelling, generalizable evidence of a reciprocal relationship between parenting and delinquency.

To fill this gap, this study used longitudinal, nationally representative data on youth ages twelve to seventeen to examine the reciprocal relationship between parenting and delinquency, accounting for family structure, age, and delinquent peers, among other factors. The results of this study will aid criminologists in further understanding the dynamic relationship between parenting and delinquency, with the hope that as the empirical evidence grows the field will continue to incorporate reciprocal effects into research on delinquency.

Social Control and Social Bonds

Social Bonds Framework

The social bonds framework begins with the assumption that to benefit themselves people are naturally inclined to engage in self-serving behaviors, behaviors that may harm or take advantage of others. The research question asked within the social bond and social control framework is: why do some people not engage in crime? Hirschi (1969:34) believed that “we would if we dared.” What stops us, according to the social bonds framework, are the bonds we have to conventional others and beliefs.

The social bonds framework, often referred to as control theory, has a lengthy history. Early theorists, such as Reiss (1951) and Nye (1958), saw social control as involving both inner and outer forces (Akers 2000). Nye (1958) further expanded Reiss’ original classifications by breaking social control into three categories: direct control, indirect control, and internal control. Portions of other early control theories, such as Walter Reckless’ containment theory (Reckless 1961, 1967) and Sykes and Matza’s (1957) techniques of neutralization and drift are associated with control theories, as well. These early theories set the groundwork for what would become one of the most tested criminological frameworks: social bonds theory. In 1969, Travis Hirschi published a book outlining the causes of delinquency and testing his theory, which was social control theory. This theory is the basis for the social bonds framework used in the current study, and includes four key elements: attachment, commitment, involvement, and belief.

Hirschi’s Social Control Theory

At the beginning of Hirschi’s book he notes that, according to control theories or social bond theories, “a person is free to commit delinquent acts because his ties to the conventional order have somehow been broken” (1969:3). This is the essence of the social bonds framework. Hirschi argued that we ought to take deviance for granted and attempt, unlike other criminological theories, to explain conformity. Hirschi described four elements of the social bond, each of which is related in some way to delinquent behavior. Hirschi argued that the more bonded one is to society through one of these elements, the more likely one is to be bonded to society through the other elements.

Attachment

Most people have some concern for the opinions of others. They want other people to think they are smart or kind or moral, among other things. Because we are concerned with what others think of us, we often tailor our behavior in ways that would gain approval from those about whom we care the most. For example, a wife may be attached to her husband, caring about what he wants her to wear or cook, who he wants her friends to be, and how he expects her to treat him. An adolescent may be attached to his teacher, showing it by trying hard on his homework, showing up to school on time, sitting quietly during lessons, and playing nicely with others on the playground. If the wife is not attached to her husband, or the adolescent is not attached to his teacher, then the wife and adolescent are free to behave, however, they wish, since they have no concern for what the husband and teacher might think about their behavior. A person is attached to someone if they are concerned about the wishes and expectations of that other person.

Hirschi comments on the fact that this element of the bond (attachment) is not a new idea; it has been included in other social control theories. It is linked to Nye’s internal and indirect controls, as well as Reiss’s personal controls. Hirschi expands on these prior renditions of the theory by adding three additional elements: commitment, involvement, and belief.

Commitment

The element of commitment is best understood as an investment. People make investments in life: investing time, energy, money, and reputation in various projects, such as getting a college degree, obtaining a good job, or getting married. Since the majority of the projects to which we are committed are conventional, engaging in delinquent or criminal behavior puts those investments at risk. If, for example, I have spent years earning an education, getting a steady job, getting married, and raising children, I risk losing all of those things were I to engage in crime. All of those investments—school, work, and relationships—are jeopardized by engaging in crime. The element of commitment prohibits people from engaging in delinquency because of what that line of behavior endangers. Commitment thus refers to future goals. If a high school boy plays on the football team and hopes to get a scholarship to play in college soon, he may be less likely to go out with his friends and get caught using illegal substances because this could cost him the scholarship opportunity. Commitment to conventional activities acts as a control preventing delinquent behavior because of the risks associated with losing the investments one has made.

Involvement

The element of involvement is about opportunity. It is the idea that someone can be too busy to get into trouble, assuming one’s major involvements are in conventional activities. Consider today’s middle class youth—school, sports, dance, music lessons, boy/girl scouts, volunteer activities, family time (Lareau 2003). An adolescent with a busy schedule of conventional activities has little time for delinquency. Having conventional activities in which one is involved reduces the opportunity to be in a situation where delinquent opportunities are available and where there is time to engage in them. Time and energy are limited, and most people fill them with conventional things. Someone who does not have conventional ways of occupying his/her time may be more inclined to fill that time with deviant activities. The issue of involvement may be particularly relevant to working class adolescents because they have relatively few responsibilities, plenty of time, and often lots of things they want (Lareau 2003). Adolescents who lack involvement in conventional activities have more time to engage in delinquent activities.

Belief

In Hirschi’s discussion of the element of belief, he notes that in most societies, there is a common value system. We codify this value system into law, and most people make a concerted effort to keep their behavior within the boundaries of the law. They do this because they believe in the validity of the law; they have been socialized to think the law is right and worthy of being followed. It is when people do not hold such beliefs that they are then free to engage in delinquent behavior. It is the beliefs themselves that control the person. As beliefs in the legitimacy of law weaken, the risk of engaging in delinquency is increased.

Of course, social control theory is not without its critics. It has been critiqued for many reasons, some of the most common of which are faulty assumptions about human nature (i.e. all humans are inherently selfish or bad and need to be controlled), failure to consider structural causes of crime or issues of power and inequality, and failure to include any discussion of the issues of race, class, or gender. Though these issues are not addressed here, it is important to note that social control theory is not accepted by all and is not always supported by the empirical evidence. This study focuses on only one of the flaws of the social bonds theory, specifically as explicated by Hirschi (1969), the failure to discuss child effects and their magnitude in any meaningful manner.

Sampson and Laub’s Age-Graded Social Control Theory

Sampson and Laub (1993) developed a theory of age-graded social control using the Glueck data. Their theoretical framework has three themes: informal family and school social controls mediate the relationship between structural context and delinquency in childhood and adolescence, there is continuity in antisocial behavior from childhood through adulthood, and regardless of individual differences in propensity to commit crime, informal social capital in adulthood explains desistence in crime over the life course. Most relevant to the current study, however, is their discussion of reciprocal relationships between parent and child in relation to delinquency. They test for child effects on delinquency by controlling for early childhood characteristics, such as temperament, and find that while child effect do exist, they do not fully explain delinquency. So, although they acknowledge the importance of considering the effect of children on their own delinquency, especially difficult children who may disrupt effective parenting, their findings reveal that the majority of the effect is still parental.

Empirical Tests of the Social Bonds Framework

In his explication of the theory, Hirschi (1969) also tested the main portions of the theory and found support for the majority of his assertions. He found that, except for involvement, the weaker the bonds, the higher the probability of delinquency. Others have since engaged in numerous tests of the theory, making it one of the most tested theories in criminology. While the results are mixed, this is not surprising given the variance in samples, methodologies, and measures used. As described below, for the most part, the theory receives weak to moderate support.

In addition to Hirschi’s own work, there have been many tests supporting the theorized effect of parental attachment on adolescent delinquency. The support has been found using a diverse set of samples, such as samples from rural areas (Jensen 1972), at-risk samples (Jang 1999), historical samples (Laub and Sampson 1988), and large, nationally representative samples (Rankin and Kern 1994). This wide range of studies has shown the positive impact parental attachment can have on reducing adolescent delinquency. In all of these studies, higher parental attachment was associated with lower rates of delinquency.

There has also been some support for the theorized effect of parental monitoring on delinquency, specifically that higher parental monitoring of adolescents will reduce delinquent behavior. Again, Hirschi found support for this in his own data, but others have found evidence in support of it, as well. For example, using a large data source from a metropolitan area, Unnever et al. (2006) found that low monitoring produced delinquency, albeit indirectly through low self-control and aggression. Others found support in historical data (Laub and Sampson 1988) and data from metro-area youth who were smokers (Ary et al. 1999). In all of these cases, lower parental monitoring or supervision was associated with higher risk of delinquent behavior.

Parental support and involvement also have been found to reduce the risk of delinquency, though Hirschi did not find support for this relationship in his data. Several large studies found that parental support or involvement was linked to lower rates of delinquency (Hundleby and Mercer 1987; Parker and Benson 2004; Wright and Cullen 2001). The same was found in a smaller sample of males by Simons et al. (1998). The finding appears to hold across age groups (Wright and Cullen 2001), as well. Like parental attachment and parental monitoring, parental involvement has been shown to have a significant negative effect on delinquent behavior.

In conclusion, delinquency prediction studies have consistently shown that parental discipline, child-rearing practices, and other family variables, all of which are important in social bonding theory, are among the best predictors of subsequent delinquency (McCord and McCord 1959; Loeber and Stouthamer-Loeber 1986). On the whole, social bonding theory has received verification from empirical research; the magnitude of the relationships between social bonding and deviant behavior has ranged from low to moderate (Akers 2000).

Family Social Bonds and Delinquency

Hirschi (1969) spent a great deal of time discussing attachment to parents and considers it a central variable in social control theory. Children who are attached to their parents may be less likely to get into trouble because they spend more time with their parents; in addition to time spent together, strong parental attachment can lead the parents to have what Hirschi calls psychological presence (1969:88). In other words, though the parent may not be standing next to the child when the opportunity for delinquency arises, the child knows, through their strong attachment to the parent, what the parent would say or think if the parent were actually there.

This is especially true for children who have been consistently monitored, as the child then knows that while their parent is not there, the parent does know where they are and with whom, making an even stronger psychological connection. Monitoring is an important part of the parent–child bond in that it fosters the psychological presence of the parent, besides enabling misbehavior to be detected directly. In sum, Hirschi makes it clear that if a child is strongly attached to her parents, then she is likely to be bound by their expectations. Within the family, the social bond of attachment, including a monitoring aspect, is most important in preventing delinquency.

Strength of the Family/Parenting and Delinquency Relationship

There has been a long-standing interest in the relationship between the family and delinquency in the social sciences. The literature examining this relationship is extensive and varied, making concise conclusions difficult. Parenting and the family are complex phenomena to measure; subsequently, different studies focus on different aspects of each. In an attempt to evaluate the strength of the effect of parenting (and the family) on delinquency, some researchers have conducted meta-analyses. These meta-analyses provide a fairly concise review of the literature by sifting through numerous empirical studies and focusing on the strongest, most comparable results from each. The first meta-analysis on parenting and delinquency was conducted by Loeber and Dishion (1983) focusing on early predictors of male delinquency. The authors found that composite measures of parental management techniques (such as lack of supervision and discipline) were most predictive of delinquency. In 1994, Simourd and Andrews conducted a meta-analysis of sixty studies on juvenile delinquency and found that poor parent-adolescent relations, such as problems in attachment or supervision, were among the strongest predictors of delinquency.

A more recent meta-analysis by Cottle et al. (2001) examined the risk factors associated with juvenile recidivism. Examining twenty-three studies and eight groups of predictors, the authors found that offense history was most predictive of recidivism, but that the domain of family was also consistently associated with recidivism. Finally, as part of her dissertation, Hoeve (2008) conducted a meta-analysis of 144 studies, mostly from the United States, published between 1950 and 2005. The analysis found a mean significant relationship between the quality of family relationships and delinquent behavior. This significant mean effect was found for nineteen of twenty different aspects of parenting and the parent-adolescent relationship, clearly demonstrating that parenting has a significant impact on delinquency, but also that no single parenting characteristic is responsible for the association. Both of these meta-analyses were conducted fairly recently, and, like earlier meta-analyses, found that at least some family characteristics, including parental behaviors, had significant impacts on delinquent behavior.

To summarize, four separate meta-analyses conducted over a span of 25 years found that the family/parenting and delinquency relationship was quite strong. In fact, the studies found that various parenting characteristics were more predictive of delinquency than parental criminality, poor academic achievement, or socioeconomic status (Loeber and Dishion 1983), lower social class, personal distress, or family structure (Simourd and Andrews 1994), and school attendance/achievement, a history of substance abuse, or severe clinical pathology, (Cottle et al. 2001). Due to the evidence of the strength of this relationship, the current study is an important step in further increasing our understanding of the complete process that occurs between parents and adolescents that leads to delinquency.

Prior Developmental Research on the Reciprocal Relationship

Within the field of child development, a number of studies have analyzed the reciprocal influence of parent and child behaviors. The results are somewhat mixed. In many cases, a reciprocal relationship between parenting and child behavior was found. For example, using a lagged multiple regression, Felson and Zielinski (1989) found that parents’ supportive behaviors affected their child’s self-esteem and that the child’s self-esteem affected perceptions of parental support. Brody and Ge (2001) studied 120 white children from two-parent families and found that harsh parenting lowered the child’s ability to self-regulate, and the child’s lowered ability to self-regulate increased harsh parenting. In a clinical sample of boys, Burke et al. (2008) found a reciprocal relationship between oppositional defiant disorder and timid discipline strategies, as well as a reciprocal relationship between conduct disorder and timid discipline strategies.

The majority of studies from the child development literature consists of small, nonrepresentative samples, sometimes containing only one gender, limited family or racial diversity, or high risk samples. Only a few have used nationally representative samples, or at least more diverse samples (Dishion et al. 2004; Laird et al. 2003; Henrich et al. 2006; Ream and Savin-Williams 2005). More specifically, Dishion et al. (2004) found that not only do parents need to monitor their adolescents, but effective parental leadership depends in part on some level of acceptance of that supervision by the adolescents themselves. These nationally representative studies provide further support for the existence of a reciprocal relationship between parental monitoring, involvement, and closeness, and delinquency.

Not all studies find evidence of reciprocal relationships between parents and adolescents, and, interestingly, some of those that find a unidirectional relationship find evidence of child effects, while others find evidence of parental effects. Evidence of child-only effects were found by Stice and Barrera (1995), using a community sample, and by Huh et al. (2006), using a sample of adolescent girls. In these studies, parental behaviors did not have any significant effects on delinquency, but delinquency did impact the behavior of parents.

A few studies found only a parental effect. For example, in a sample of adolescents in two-parent families, Coley et al. (2008) found that in their within-person results, positive family and parenting variables predicted lower rates of substance use in youth, but substance use by youth did not affect either family or parenting variables. Lifford et al. (2008) studied the associations between ADHD and the parent–child relationship. They found that the child’s ADHD symptoms affected the mother–child relationship, but that the father-child relationship affected the child’s ADHD symptoms. While these studies do not support a reciprocal relationship between parenting and delinquency, they do provide evidence of a parental effect, evidence in support of the idea that child effects alone cannot explain the relationship.

Prior Criminological Research on the Reciprocal Relationship

The examination of reciprocal relationships between parents and children has a 40-year history in the developmental psychology literature. Over the past 20 years, a small body of research has examined the reciprocal relationship between parenting and delinquency by directly assessing the magnitude of the child effects in addition to the parental effects. In these studies, delinquency is viewed not simply as an outcome of parenting practices but as part of a transactional process whereby parenting affects delinquency but is also affected by it. A search of the literature revealed a number of studies that focus specifically on the reciprocal relationship between adolescent delinquency and parenting, directly measuring child effects.

Studies examining the reciprocal relationship between parents and youth are easiest to examine by type of social bond studied. Several studies looked for a reciprocal relationship between parental attachment and adolescent delinquency of various types. Some found evidence of both parental and child effects, indicating a reciprocal relationship (Buist et al. 2004; Liska and Reed 1985; Thornberry et al. 1991); others found only a small parental effect of attachment on delinquency (Agnew 1991) or only a child effect of delinquency on parental attachment (Jang and Smith 1997). While most of the studies examining parental attachment found reciprocal effects, the support is certainly not unanimous.

Other studies have examined the relationship between parental monitoring or supervision and delinquency, looking again for evidence of a reciprocal relationship. Each of the studies found that examined this relationship found evidence of both parental and child effects (Jang and Smith 1997; Kerr et al. 2010; Laird et al. 2003; Stewart et al. 2002). Of particular interest in this line of studies is the recent study by Kerr et al. (2010) in which they reiterate their argument that parental monitoring is really parental knowledge, supplied primarily through youth disclosure. Because of such criticisms of this measure, the current study will use a measure that asks youth to report what kind and how rules are established in the home, rather than relying on a measure of parental monitoring that is really a measure of child disclosure. Kerr et al. use longitudinal data from seventh and eighth graders, and their parents, from a small city in Sweden. Delinquency scales were created from both parent (23 items) and youth (18 items) reports. They find evidence of a bidirectional relationship between parental monitoring, whatever its source, and adolescent delinquency, separated by 2 years. It is interesting to note that none of the studies found report only child effects or only parental effects between parental monitoring and delinquency, but consistently find a reciprocal relationship.

Finally, a couple of studies examined the relationship between parental involvement and delinquency and found no evidence of reciprocal effects (Coley and Medeiros 2007; Hawkins et al. 2007). This is not surprising since Hirschi himself found no support for the theory that parental involvement reduces adolescent delinquency. At the very best, we might expect to find evidence of child effects whereby delinquency reduces the amount of parental involvement a youth experiences. This is exactly what Hawkins and colleagues found; however, Coley and Mederios found evidence of only parental effects. In each study, however, the focus was on non-resident father involvement. The current study examines the relationship between resident parents’ involvement and youth delinquency, so the findings may be quite different.

Together, the results of these studies were mixed. However, the weight of the evidence seems to be for the existence of a reciprocal relationship between delinquency and some form of parenting. There were a few studies that found a unidirectional relationship, typically between involvement and delinquency, but most—especially those involving attachment—found reciprocal relationships. The evidence indicates there is a strong possibility that not only do parents affect adolescent’s behavior, as suggested by the social bonds framework, but that the behavior of adolescents also affects parenting.

In sum, while a fair amount of evidence suggests that a reciprocal relationship exists between parenting and various child outcomes, including delinquency, the findings in the literature are not conclusive; there is, however, enough evidence of a reciprocal relationship to warrant further investigation. Additionally, while the study of reciprocal relationships between parents and children is not new or particularly uncommon in the developmental psychology literature, it has only recently begun to be studied by criminologists. Moreover, the few criminological studies that examine this relationship suffer from various shortcomings, predominantly sampling issues, discussed in the next section.

Limitations of Prior Research

While prior research on the reciprocal relationship between parenting and delinquency is not new, each study suffers from important limitations that will be improved upon in this study. There is no apparent link between the type of sample and the findings of prior research, and the findings are mixed enough to warrant further study. Each study had strengths, but each study also suffered from at least one of the following limitations: small sample sizes; studying only males; samples composed entirely or primarily of whites; samples of high school students only; high risk samples; respondents from two-parent families only or predominantly; nonresident fathers only; failure to include lagged effects (controlling for prior delinquency prior parenting behaviors); failure to control for delinquent peer effects; examination of only one type of parenting or one type of delinquency. The majority of these are sampling issues. In fact, the only studies that used nationally representative samples were those that included only nonresident fathers or failed to include lagged effects in the models. The current study addresses these issues.

The Current Study

This study examined the reciprocal relationship between three parental behaviors and delinquency in a nationally representative sample of adolescents. Reciprocal relationships between the key variables were tested using lagged regression equations. Eighteen separate models examined whether there is a significant parental effect on adolescent delinquency and whether there is a significant adolescent delinquency effect on parental behaviors. The examination of a reciprocal relationship between parenting and delinquency is of vital importance. The assumption in social control theory, the criminological theory that focuses most on parental attachment and supervision, is that parents affect children. If, however, there are both parent and child effects, this has great implications for theory.

This study examines bidirectional effects using two points in time. Like most prior research, delinquency is measured using youth reports of multiple items of delinquency ranging in severity from minor property crimes to serious violent crimes. The failure of previous criminological studies to test for such reciprocal relationships means that these studies may have overestimated the effect of parenting on delinquency, as the effect of delinquency on parenting was not generally considered. Based on prior research in both the criminological literature and the child development literature, six hypotheses will be tested. The first three hypotheses relate to parental effects. First, it is hypothesized that there will be a significant, negative effect of parental attachment on each type of delinquency, the rationale being that the stronger the bond between parent and child, the less likely that child is to engage in delinquent behavior (H1). This has been supported in the literature. Second, there will be a significant, negative effect of parental monitoring on each type of delinquency. As parents increasingly monitor their children, the opportunity and/or willingness of the child to engage in delinquent behavior should diminish (H2). Third, there will be a significant, negative effect of parental involvement on each type of delinquency (H3), as found by Unnever et al. (2006). Parents who are involved are, in most cases, involved in conventional activities with their child. This decreases the opportunity for engaging and delinquency and increases parental monitoring as time is spent together. The final three hypotheses relate to child effects. First, there will be a significant, negative effect of each type of delinquency on parental attachment (H4). The rationale here is that delinquent behavior on the part of the child will negatively impact the parent–child relationship, resulting in a decrease in the level of parental attachment felt. Second, there will be a significant, positive effect of each type of delinquency on parental monitoring (H5). The rationale behind this hypothesis is that as parents observe their child engaging in delinquent behavior, they will begin to increase their level of monitoring of that child in order to limit the opportunities for delinquent behavior. Third, there will be a significant, negative effect of each type of delinquency on parental involvement (H6). As noted in the literature review, Ary et al.’s (1999) found that increased family conflict, which is a likely consequence of delinquent behavior on the part of the child, leads to lower levels of parental involvement. Children who are engaging in delinquent behavior probably do not wish to spend time with their parents, and parents and children who are in conflict over the child’s behavior also are unlikely to want to spend time together. Each of these hypotheses will be tested using regression models.

Methods

Data

The data used in this research were collected for the National Longitudinal Study of Adolescent Health (Add Health). Add Health is a nationally representative, school-based longitudinal study of adolescents in grades seven through twelve in the United States in 1994–1995. The data were collected from adolescents, fellow students, school administrators, parents, siblings, friends, and romantic partners (Harris et al. 2008). There are currently four waves of data. The final sample of adolescents that participated in both Waves I and II of the study was 13,570, which is the base sample for this study. After a loss of data for respondents who were not between the ages of twelve and seventeen, the final sample size was 12,505 with 50.42% female.

Wave 1 of Add Health was collected in several stages. First, a stratified, random sample of all high schools in the United States was identified. Schools were eligible if they included an eleventh grade and if they had a minimum enrollment of thirty students. A school that sent graduates to the high school and that included a seventh grade was also included from each community. A sample of 80 high schools and 52 middle schools was selected with unequal probability of selection, resulting in an in-school sample of over 90,000 respondents. High schools were then stratified into 80 clusters based on region, urbanicity, school size, school type, percent white, percent black, grade span, and curriculum (Harris et al. 2008). The second stage of wave I was an in-home sample of 20,745 adolescents drawn from each school where an in-school survey was completed. Special oversamples also were selected. Parents were interviewed in wave I only.

The wave II in-home sample is almost the same as the wave I in-home sample. Key differences are that most of the twelfth-grade respondents were removed from wave II, there are a few respondents in wave II who were not interviewed in wave I, and parents were not interviewed in wave II (Harris et al. 2008). Wave II was collected approximately one year after wave 1; the interviews were generally similar.

Delinquency and Parenting Measures

Delinquency

Three delinquency scales were created for both waves of data using self-report questions. Parents likely respond differently to a child that steals versus a child that harms others, and given that the parental behavior regarding different types of behavior may vary, the child’s response may also vary. To address this, each reciprocal relationship is examined using separate property and violent delinquency scales, as well as one overall combined delinquency scale.

Overall Delinquency

The overall delinquency scales were created as a count of delinquent acts reported by the respondents. Status offenses (e.g. running away or lying to one’s parents) were not included in any of the delinquency scales in an effort to reduce the positive skew of the scale towards the less serious items. The overall delinquency scales consisted of acts listed in both the property and violent delinquency descriptions below this section. Each item was dichotomized, and the items were then summed to create scales ranging from zero (no items) to eleven (Wave I; all items) and zero to ten (Wave II). The alpha for this scale in both waves is 0.82.

Property Delinquency

The property delinquency scales are a count of only those delinquent acts that are considered property crimes: graffiti, vandalism, shoplifting, auto theft, stealing more/less than $50, and burglary. As with the overall delinquency scales, each item was dichotomized and summed to create scales ranging from zero (no to all) to seven (yes to all). The alpha for this scale in both waves is 0.78.

Violent Delinquency

The violent delinquency measure is somewhat different from the previous two delinquency scales in that the final measure is a dichotomous variable rather than a count. This is due to the rarity of each act, given that these are much more serious delinquent behaviors. A count was created including the following violent crimes: fighting, group fighting, robbery, pulling a knife or gun on someone, and shooting or stabbing someone. Each of these items are dichotomized (0 = no, 1 = yes) and summed. If a respondent scored a one or greater on the created scales, they were coded has “1” for the dichotomous violence measure. If a respondent scored a zero on the scales, they were coded as “0” for the measure. Various methods can and have been used in prior research to address the skewed distribution of this type of variable; dichotomization was chosen to provide the most simple interpretation of the findings. The alpha for the scale from which this measure was calculated is 0.65 for Wave I and 0.69 for Wave II (five items).

Parenting

The parenting measures are divided into three separate scales with no overlap between them: parental attachment, parental monitoring, and parental involvement. Prior research has tended to focus on these areas of parenting, and they closely mirror the social bonds described by Hirschi (1969). Though these measures could each be considered as aspects of attachment, the types of behaviors involved may be expected to relate to delinquency in different ways. This might be especially true for the effect of delinquency on parenting.

Parental Attachment

This measure was calculated using five questions asked of the adolescent regarding the mother and five questions asked of the adolescent regarding the father. The same questions were asked for both waves. Respondents were asked how close they feel to their mom (1 = not at all to 5 = very much) and how much they think she cares about them (1 = not at all to 5 = very much). The respondents were then asked to what extent they agreed or disagreed with the following: most of the time, your mom is warm and loving toward you (1 = strongly agree to 5 = strongly disagree); when you do something wrong that is important, your mom talks about it with you and helps you understand why it is wrong (1 = strongly agree to 5 = strongly disagree); you are satisfied with the way you and your mom communicate with each other (1 = strongly agree to 5 = strongly disagree). The scores were recoded as necessary so that a high score indicated higher parental attachment. The average of these five items was then taken to compute maternal attachment (alphas for both waves were 0.86).

Respondents were also asked how close they feel to their dad (1 = not at all to 5 = very much) and how much they think he cares about them (1 = not at all to 5 = very much). They were then asked to what extent they agreed or disagreed with the following: most of the time, your dad is warm and loving toward you (1 = strongly agree to 5 = strongly disagree); you are satisfied with the way your dad and you communicate with each other (1 = strongly agree to 5 = strongly disagree); overall, you are satisfied with your relationship with your dad (1 = strongly agree to 5 = strongly disagree). These were recoded and averaged as those items involving the mom (alphas for both waves were 0.86). The average of the mother-item mean and the father-item mean was then taken to account for overall average parental attachment between parents and child for both waves. If the child does not have a father present, the section answering the questions for the father was skipped; in this case, the parental attachment measure is simply the mean of reported attachment to the mother.

Parental Monitoring

Parental monitoring is measured using seven dichotomous items (0 = no, 1 = yes) that asked respondents about the kinds of decisions their parents allow them to make. This is a similar measure to that used recently by Keisner et al. (2010) in response to criticisms that suggesting that measures of parental monitoring are often measuring how well informed parents are rather than how much they control youth behavior. In the current study, youth were asked if their parents let them make their own decisions about the following: the time they must be home on weekend nights; the people they hang around with; what they wear; how much television they watch; which television programs they watch; what time they go to bed on week nights; and what they eat. These items are used here as a proxy for the establishment of rules in the household. If parents are establishing rules in the home, then it is assumed that they are monitoring the behavior of their children. Research that uses measures of regarding parents’ knowledge of their child’s whereabouts, activities, friends, etc. has been criticized for measuring adolescent disclosure more than parental monitoring (Stattin and Kerr 2000). This measure avoids that problem by asking youth about how rules are established in the household. These seven items were summed to create a monitoring scale for both waves (alphas ranged from 0.61 to 0.65 across the waves).

Parental Involvement

The involvement measure was created using nine items meant to determine what the respondents have done with their mother and/or father in the last month, for both waves. Respondents were shown a card with a list of nine activities and were asked if they had done any of them in the past four weeks (all answer choices were yes/no): gone shopping; played a sport; gone to a religious service, or church-related event; talked about someone they are dating, or a party they went to; gone to a movie, play, museum, concert, or sports event; had a talk about personal problems they (the adolescents) were having; talked about their school work or grades; worked on a project for school; talked about other things they are doing in school. This is similar to how Goncy and van Dulmen (2010) recently measured one aspect of parental involvement, which they called “shared activity participation.” Shared activity participation is a common way of measuring parental involvement that has been linked to reductions in depression (Yuan and Hamilton 2006), school failure (Menning 2006), and alcohol use (Jordan and Lewis 2005). In the current study, the respondents were asked to respond to each of the previously mentioned items for both their mother and their father; the maximum score for each item was used (i.e. if they did something with their mother but not their father, they still scored a “1”). The maximum score for each item was then summed to calculate the parental involvement score (alphas ranged from 0.56 to 0.57 across both waves). Respondents who shared many activities with their parents scored higher on this variable, with a possible range of zero (did none of these things with either parent) to nine (did each of these things with at least one parent).

Controls

Delinquency has been found to be associated with a number of variables that necessitate their inclusion as control variables into any analysis of delinquent behavior and, in this case, variables are included as controls that relate to both parenting and delinquency. For this study, age, gender, race, parent’s education, family structure, grades, and delinquent peers are included as controls. All control variables were measured using the data from wave I. A correlation matrix of all variables included in the analyses can be found in Tables 1 and 2.
Table 1

Correlation matrix of all variables, part 1

 

Overall Del. T1

Overall Del. T2

Property Del. T1

Property Del. T2

Violent Del. T1

Violent Del. T2

Female

Grades

Two-Parent Family

Age

Parent’s education

Overall Del. T1

1

          

Overall Del. T2

0.55

1

         

Property Del. T1

0.94

0.52

1

        

Property Del. T2

0.49

0.92

0.50

1

       

Violent Del. T1

0.57

0.31

0.35

0.22

1

      

Violent Del. T2

0.35

0.54

0.26

0.32

0.39

1

     

Female

−0.18

−0.15

−0.13

−0.10

−0.19

−0.16

1

    

Grades

−0.24

−0.18

−0.19

−0.12

−0.24

−0.21

0.14

1

   

Two-parent family

−0.07

−0.04

−0.05

−0.02

−0.08

−0.06

−0.02

0.13

1

  

Age

0.02

−0.03

0.02

−0.05

−0.05

−0.04

−0.04

−0.12

−0.03

1

 

Parent’s education

−0.05

−0.03

−0.02

0.001

−0.12

−0.09

−0.03

0.26

0.16

−0.05

1

Attachment T1

−0.20

−0.13

−0.20

−0.12

−0.09

−0.07

−0.10

0.12

0.003

−0.18

0.04

Attachment T2

−0.17

−0.17

−0.17

−0.15

−0.09

−0.09

−0.09

0.10

0.002

−0.11

0.03

Monitoring T1

0.03

0.01

0.04

0.01

−0.03

−0.02

0.01

0.03

−0.03

0.27

0.05

Monitoring T2

0.04

0.03

0.05

0.02

−0.02

−0.004

0.01

0.02

−0.04

0.28

0.06

Involvement T1

−0.09

−0.06

−0.08

−0.04

−0.07

−0.06

0.10

0.19

0.11

−0.06

0.19

Involvement T2

−0.11

−0.05

−0.09

−0.03

−0.09

−0.06

0.15

0.21

0.12

−0.11

0.18

White

−0.06

−0.03

−0.03

0.0004

−0.11

−0.07

−0.004

0.14

0.21

−0.06

0.13

Hispanic

0.08

0.06

0.07

0.05

0.06

0.05

−0.01

−0.13

−0.02

0.07

−0.28

Black

−0.01

−0.02

−0.05

−0.06

0.10

0.05

0.03

−0.10

−0.23

−0.02

0.05

Other race

0.02

0.01

0.03

0.01

−0.02

−0.01

−0.03

0.07

−0.02

0.05

0.07

Delinquent peers

0.38

0.26

0.33

0.19

0.22

0.20

−0.02

−0.27

−0.06

0.25

−0.08

Table 2

Correlation matrix of all variables, part 2

 

Attachment T1

Attachment T2

Monitoring T1

Monitoring T2

Involvement T1

Involvement T2

White

Hispanic

Black

Other race

Del. peers

Overall Del. T1

           

Overall Del. T2

           

Property Del. T1

           

Property Del. T2

           

Violent Del. T1

           

Violent Del. T2

           

Female

           

Grades

           

Two-parent family

           

Age

           

Parent’s education

           

Attachment T1

1

          

Attachment T2

0.61

1

         

Monitoring T1

−0.20

−0.06

1

        

Monitoring T2

−0.13

−0.05

0.45

1

       

Involvement T1

0.28

0.22

−0.03

−0.03

1

      

Involvement T2

0.22

0.31

−0.04

−0.04

0.45

1

     

White

0.01

−0.001

0.07

0.08

0.08

0.08

1

    

Hispanic

−0.03

−0.03

−0.05

−0.05

−0.04

−0.04

−0.46

1

   

Black

0.04

0.06

−0.03

−0.03

−0.02

−0.01

−0.55

−0.23

1

  

Other race

−0.03

−0.05

−0.01

−0.03

−0.04

−0.06

−0.34

−0.14

−0.17

1

 

Delinquent peers

−0.20

−0.17

0.14

0.14

−0.10

−0.12

0.08

0.01

−0.09

−0.03

1

Table 3

Descriptive statistics before and after multiple imputation

Variables

Before imputation

After imputation

Number imputed

Mean or percentage

Standard deviation

Mean or percentage

Standard deviation

Female

50.97%

50.42%

0

Grades

2.83

0.77

2.81

0.32

293

Family structure

66.39%

64.93%

0

Age

14.81

1.45

14.82

0.60

0

Parent’s education

5.90

2.29

5.84

0.95

566

Race

     

 White

65.69%

65.69%

0

 Black

15.25%

15.25%

0

 Hispanic

11.85%

11.85%

0

 Other

6.77%

6.77%

0

Delinquent peers

2.31

2.60

2.34

1.07

416

Overall delinquency, T1

0.93

1.44

1.36

0.75

2732

Overall delinquency, T2

0.66

1.23

1.06

0.68

2386

Property delinquency, T1

0.91

1.45

0.94

0.59

144

Property delinquency, T2

0.72

1.32

0.75

0.54

102

Violent delinquency, T1

36.37%

36.67%

809

Violent delinquency, T2

24.75%

23.94%

898

Parental attachment, T1

4.36

0.59

4.35

0.24

104

Parental attachment, T2

4.25

0.61

4.25

0.25

306

Parental monitoring, T1

4.94

1.55

4.16

0.57

162

Parental monitoring, T2

5.33

1.53

4.49

0.56

330

Parental involvement, T1

4.11

1.98

4.09

0.81

110

Parental involvement, T2

4.12

2.00

4.10

0.82

309

Race, Gender, and Parent’s Education

Prior research has consistently shown variations in delinquency by gender (see Steffensmeier and Allan 1996, for a review). Research on the effects of SES and race has been more equivocal and tends to depend on whether the delinquency is official or self-reported, with self-reported delinquency having a weaker association to race and social class (see Elliott and Ageton 1980; Hindelang et al. 1979, for discussions of the discrepancy between self-reported and official delinquency and the correlates of delinquency). Some studies have found racial differences in serious self-reported delinquency, where blacks engage in more serious and violent delinquency than do whites (Salts et al. 1995) and in high frequency self-reported delinquency, where blacks report higher frequencies of delinquency than do whites (Elliott and Ageton 1980). Elliott and Ageton (1980) also found social class differences in high frequency self-reported delinquency, with respondents from the lower class reporting higher frequency of delinquency. Additionally, parenting practices have been associated in the literature with SES (see Bradley and Corwyn 2002, for a brief discussion), as have differences in parenting by gender of child (e.g. Barnes and Farrell 1992; Starrels 1994; see Rossi 1984, for a discussion) and race (e.g. Barnes and Farrell 1992; Hill and Sprague 1999).

With these data, the race and gender measures are fairly straightforward, single survey questions. Parent’s education is measured using the maximum of the mother and father’s highest level of education attained. Respondents were asked how far in school both their mom and their dad went, with answers ranging from “never went to school” to “professional training beyond a four-year college or university.

Family Structure

Family structure has also been associated with delinquency. Family structure has been found to have a strong relationship with violent crime, with youth living in single-parent homes being most at risk for such behavior (Dornbusch et al. 1985; Paschall et al. 1996; Sampson 1987). Here family structure was measured using a single variable from the parent survey, asking the parent (usually the mother) what her current marital status is. Those respondent’s whose parent answered that they were currently married were coded as “1” for living in a two-parent household; all others were coded as “0” for living in a single-parent household or other type of family arrangement. The rationale for creating the family structure variable as a dichotomous two-parent family versus others is really about opportunity. Adolescents who live in a home with two parents have two sets of eyes helping monitor them, two parents with whom they could engage in activities (involvement), and two parents with whom they could develop an attachment. Given that the family structure of two-versus-one parent could influence the parenting measures and the delinquency measures, family structure is measured as a dichotomy.

Age

The association between delinquency and age is well document and one of the most accepted relationships in criminology (see Steffensmeier et al. 1989 for a discussion). Delinquency peaks in adolescence, and aging out of crime is the norm, even for serious offenders, as they reach their twenties and thirties (Sampson and Laub 2003). Respondents were asked to report their birthday, which was then used in conjunction with the interview date to calculate their age at wave I. Age has an impact on the parent–child relationship, as well. Adolescence is a time of transition as children who once relied solely on their parents are beginning to explore autonomy, perhaps weakening their ties to their parents and strengthening their ties to their peers. It is imperative to control for age here because the age range of twelve to seventeen covers the beginnings of that potential change. Respondents who are twelve may still have closer ties to parents than peers, while respondents who are seventeen may have closer ties to peers than parents.

Grades

School failure has been associated with an increased risk for delinquent behavior. Wang et al. (2005) found that delinquent youth attained lower grade point averages than nondelinquent youth, and Rhodes and Reiss (1969) found that delinquency was one of several adaptations to school failure. However, Felson and Staff (2006) argue that this relationship may be spurious, as both delinquency and school failure may be caused by low self-control. Though not examined by Felson and Staff, it also seems plausible that both delinquency and school failure could be influenced by poor parenting practices. To assess school failure or success, respondents reported their grade for four core courses (English/language arts, mathematics, history/social studies, and science). Those four grades were then used to calculate an average school grade on a scale of one to four, with four being the highest score (equivalent to an “A”).

Delinquent Peers

While it is often found that peer delinquency is a stronger predictor of delinquency than are parenting measures (e.g. Aseltine 1995; Matsueda 1982), there have also been some findings showing that parenting can influence delinquency through delinquent peers (e.g. Warr 2005), though this will not be tested here. The measure of peer delinquency used in this study is a count of the number of delinquent peers based on three questions in Add Health. Youth were asked how many of their three best friends smoke at least one cigarette a day, drink alcohol at least once per month, and/or use marijuana at least once per month. The scale ranges from zero (none of three best friends are delinquent) to three (all three best friends are delinquent). It has an alpha of 0.76.

Data Analysis

The Add Health data are clustered data collected with an unequal probability of selection. Because of this, the data were analyzed in Stata, which is capable of handling the clustered design. This allowed for the obtaining of unbiased estimates of population parameters and standard errors (Chantala 2006).

The study included some missing data. While the sample I studied used only respondents who participated in both waves, respondents sometimes failed to answer certain questions. The variables with the most missing data were the delinquency variables, which is not surprising; youth may be hesitant, even on a confidential survey, to reveal participation in illegal activities. Stata will automatically use listwise deletion to deal with missing data, which significantly reduces the sample size, resulting in a loss of power and potential loss in generalizability. In this study listwise deletion would have resulted in the loss of approximately 3,000 cases. To avoid these losses, multiple imputation in Stata was used.

Imputation is the “filling in” of missing data with plausible values (Schafer 1999). In multiple imputations, missing data are replaced in a number of simulated versions of the data. Each of these imputed datasets is then analyzed as usual, and the results of each analysis are combined to produce estimates that take into account missing-data uncertainty. Multiple imputation takes advantage of all of the available data and avoids the bias that could result from using only one imputed data set. In the current study, five imputations were calculated resulting in five imputed data sets. Following standard procedures, every independent and dependent variable in the models is included for the imputation procedure. After imputation, the final sample was 12,505 respondents,1 50.42% of which were female. These are respondents who participated in both waves, had a valid weight (making the sample nationally representative), and were between the ages of twelve and seventeen. Imputation was not used to account for those respondents who only participated in one wave as imputing an entire wave of data for a given respondent is more of a stretch on reality than imputing a single variable.

Because the outcome varies in each model, the type of regression analysis used for each model varies, as well. For the overall and property delinquency outcomes, negative binomial regressions are used. Normal OLS regression assumptions are violated due to the skewed distribution of delinquency scales. Because the violent delinquency measure is a simple dichotomy, logistic regressions are used for models in which this is the outcome. When any of the three parenting scales are the outcome in a model, a standard linear regression is used because these variables tend to be normally distributed.

In order to test for bidirectionality, eighteen main models will be estimated, accounting for three delinquency outcomes, each with one of the three parenting measures as the key predictor, and three parenting outcomes, each with one of the three delinquency measures as the key predictor. In each model, the time 1 measure for the outcome will be controlled in a cross-lagged regression to control for prior delinquency or parenting behaviors. This will allow for the estimation of the effect of each key predictor on change in the relevant outcome, thus factoring in individual differences related to each of the dependent variables of interest.

It is possible to examine this relationship using structural equation modeling. Liska and Reed (1985) conducted their study of the reciprocal relationship using both cross-lagged regressions and simultaneous structural equations models. They found that the process was essentially the same using either method, though the estimates were somewhat smaller in the lag model. The question determining which method is best depends on whether the effects are simultaneous or instantaneous. The rationale used in the current study mirrors that of Felson and Zielinski (1989). Parenting and child development are on-going processes that occur over time, not instantaneously. The assumption made here is that the parental effects and youth effects are not simultaneous or instantaneous. Even if this assumption were found to be invalid, it is likely that both lagged and simultaneous/instantaneous effects are still in the same direction. Though this assumption is not tested here, I argue that parent and youth behavior is relatively stable over time, and that reactions to either parent or youth behavior reflect the long-term impact of those behaviors rather than instant impact, as similarly argued by Felson and Zielinski (1989).

Descriptive and Bivariate Results

Descriptive Findings

Delinquency Variables

Table 4 shows the demographic characteristics of the sample. The average level of overall delinquency at Time 1 was 1.36 acts of delinquency, out of thirteen possible acts. At Time 2, the average level was slightly lower at 1.06 acts of delinquency. The average level of property delinquency at Time 1 was 0.94, less than one act of property delinquency; at Time 2 the average was 0.75 acts of property delinquency. At Time 1, nearly 37% of respondents reported engaging in at least one act of violent delinquency (which includes being in a fight or a group fight); at Time 2, nearly 24% of respondents reported this.
Table 4

Descriptive statistics

Variable

Mean or percentage

Standard deviation

Range

Female

50.42%

0–1

Grades

2.806

0.319

1–4

Family structure

64.93%

0–1

Age

14.823

0.596

12–17

Parent’s education

5.844

0.951

0–9

Race

   

 White

65.69%

0–1

 Black

15.25%

0–1

 Hispanic

11.85%

0–1

 Other

6.77%

0–1

Delinquent peers

2.336

1.074

0–9

Overall delinquency, T1

1.360

0.748

0–11

Overall delinquency, T2

1.057

0.684

0–10

Property delinquency, T1

0.943

0.593

0–7

Property delinquency, T2

0.753

0.543

0–7

Violent Delinquency, T1

36.67%

0–1

Violent delinquency, T2

23.94%

0–1

Parental attachment, T1

4.353

0.243

1–5

Parental attachment, T2

4.246

0.250

1–5

Parental monitoring, T1

4.160

0.566

0–6

Parental monitoring, T2

4.493

0.561

0–6

Parental involvement, T1

4.092

0.812

0–9

Parental involvement, T2

4.103

0.822

0–9

Parenting Variables

As seen in Table 4, the average level of parental attachment was 4.35 (out of five, high scores indicating higher levels of attachment) at Time 1; this was reduced only slightly over time to 4.25 at Time 2. The average level of parental monitoring reported by respondents at Time 1 was 4.16, which means that on average, respondents reported having approximately four of the six house rules listed on the survey. At Time 2, the average level of parental monitoring had gone up slightly to 4.49. The most common house rules were having a weekend night curfew and having a bed time on week nights. Finally, respondents reported an average of 4.09 for parental involvement at Time 1, indicating that they participated with their parents in approximately four of nine activities listed on the survey. By Time 2, this average had barely increased to 4.10 activities. Despite slight fluctuations, these averages indicate that levels of parental attachment, monitoring, and involvement were fairly stable over time. This is not to imply that there are no within-individual changes over time but that the average levels for all respondents do not change over time. This is not surprising given that these are long-term relationships, and the measures were assessed one to one and a half years apart.

Bivariate Relationships

Delinquency

Fourteen of eighteen relationships are significant at the bivariate level. Consistent with the social bonds framework, as parental attachment increases, risk of each of the three types of delinquent behavior decreases (b = −0.33 for overall delinquency, b = −0.34 for property delinquency, and b = −0.34 for violent delinquency). As parental involvement increases, risk of delinquency decreases (b = −0.039 for overall delinquency, b = −0.031 for property delinquency, and b = −0.067 for violent delinquency). Since Hirschi included parental monitoring in his discussion of social control theory and attachment, parental monitoring is often viewed as part of parental attachment. Parental monitoring evinced no significant associations with any of the three types of delinquency, indicating that existence of basic house rules does not seem to influence risk of delinquency for adolescents.

Parenting

At the bivariate level, overall delinquency has a significant association with only parental attachment. Overall delinquency has a negative and significant association with parental attachment (b = −0.064). With parental involvement, the relationship is negative (b = −0.126), though not significant, and with monitoring, the relationship is positive (b = 0.031), though not significant. There are significant negative associations between property delinquency and parental attachment (b = −0.080) and parental involvement (b = −0.128), and a significant positive association between property delinquency and parental monitoring (b = 0.047). The picture is slightly different with violent delinquency. Violent delinquency has a significant, negative association with parental attachment (b = −0.112) and parental involvement (b = −0.433). There is, however, no significant association between violent delinquency and parental monitoring.

Multivariate Models

The Effect of Parenting on Delinquency

Table 5 presents the results of the regression analyses predicting each type of delinquency using each type of parental behavior. There are nine models in Table 5; these nine models are the parental effects models, the first half of the potential bidirectional relationship between parenting and delinquency (testing H1, H2, and H3). Each column of coefficients represents a separate delinquency outcome: overall delinquency in column 1, property delinquency in column 2, and violent delinquency in column 3. The coefficients presented reflect changes in each type of delinquency over time, controlling for demographic variables and prior delinquency.
Table 5

Time 2 delinquency outcomes regressed on time 1 parenting predictors

Variables

Attachment predicting delinquency

Monitoring predicting delinquency

Involvement predicting delinquency

Overall

Property

Violent

Overall

Property

Violent

Overall

Property

Violent

Model 1

Model 2

Model 3

Model 4

Model 5

Model 6

Model 7

Model 8

Model 9

Attachment

0.13 (0.03)***

0.13 (0.03)***

0.18 (0.06)**

Monitoring

 

0.003 (0.01)

0.003 (0.02)

0.01 (0.02)

Involvement

 

0.01 (0.01)

0.01 (0.01)

0.01 (0.01)

Delinquency

0.28 (0.01)***

0.39 (0.02)***

1.46 (0.07)***

0.28 (0.01)***

0.40 (0.02)***

1.48 (0.07)***

0.28 (0.01)***

0.40 (0.02)***

1.47 (0.07)***

Female

−0.30 (0.04)***

−0.20 (0.04)***

−0.55 (0.06)***

−0.28 (0.04)***

−0.17 (0.04)***

−0.52 (0.06)***

−0.28 (0.04)***

−0.17 (0.04)***

−0.52 (0.06)***

Grades

−0.10 (0.03)***

−0.08 (0.03)**

−0.29 (0.05)***

−0.11 (0.03)***

−0.09 (0.03)**

−0.30 (0.05)***

−0.11 (0.03)***

−0.10 (0.03)**

−0.30 (0.05)***

Two-parent family

−0.04 (0.04)

−0.03 (0.05)

−0.07 (0.07)

−0.04 (0.04)

−0.03 (0.05)

−0.07 (0.07)

−0.04 (0.04)

−0.03 (0.05)

−0.08 (0.07)

Age

−0.11 (0.01)***

−0.13 (0.02)***

−0.17 (0.02)***

−0.11 (0.01)***

−0.13 (0.02)***

−0.16 (0.02)***

−0.11 (0.01)***

−0.12 (0.01)***

−0.17 (0.02)***

Parent’s education

0.0002 (0.01)

0.01 (0.01)

−0.04 (0.01)**

0.0004 (0.01)

0.01 (0.01)

−0.04 (0.01)

−0.000 (0.01)

0.01 (0.01)

−0.04 (0.01)**

Race

 Hispanic

0.14 (0.06)*

0.14 (0.07)*

0.26 (0.10)**

0.14 (0.06)*

0.14 (0.07)*

0.27 (0.10)**

0.14 (0.06)*

0.14 (0.07)*

0.27 (0.10)**

 Black

−0.05 (0.06)

−0.18 (0.07)*

0.14 (0.08)

−0.06 (0.06)

−0.18 (0.07)**

0.12 (0.08)

−0.06 (0.06)

−0.18 (0.07)*

0.12 (0.08)

 Other

0.03 (0.08)

−0.05 (0.09)

0.31 (0.11)**

0.03 (0.08)

−0.04 (0.09)

0.32 (0.11)**

0.03 (0.08)

−0.04 (0.09)

0.32 (0.11)**

Delinquent peers

0.06 (0.01)***

0.04 (0.01)***

0.12 (0.01)***

0.06 (0.01)***

0.04 (0.01)***

0.13 (0.02)***

0.06 (0.01)***

0.05 (0.01)***

0.13 (0.01)***

Constant

2.02 (0.27)***

1.80 (0.30)***

2.74 (0.52)***

1.33 (0.22)***

1.13 (0.24)***

1.79 (0.40)***

1.32 (0.22)***

1.11 (0.24)***

1.75 (0.40)***

Standard errors are in parentheses

p < 0.05, ** p < 0.01, *** p < 0.001

The first set of models (1–3) is for the parental attachment predictor. As shown in the table under model 1, controlling for demographic variables as well as overall delinquency at Time 1, the coefficient for the effect of parental attachment at Time 1 on overall delinquency at Time 2 is significant (b = −0.129, s.e. = 0.028, p = 0.001), providing partial support for H1. Parental attachment has a significant negative effect on overall delinquency; as the level of attachment increases, the prevalence of overall delinquency decreases.

As shown in model 2, the coefficient for the effect of parental attachment at Time 1 on property delinquency at Time 2 was also significant and negative (b = −0.125, s.e. = 0.033, p = 0.001), providing additional support for H1; as the level of parental attachment increases, property delinquency also decreases. Finally, the same was found for violent delinquency, as shown in model 3; parental attachment at Time 1 had a significant and negative effect on violent delinquency at Time 2 (b = −0.196, s.e. = 0.060, p = 0.002), providing final support for H1. Respondents who reported higher parental attachment were less likely to engage in violent delinquency. In sum, parental attachment had a significant negative effect on each type of delinquency. These findings provide full support of hypothesis 1.

The second section of Table 5 shows the results for the three models involving parental monitoring as a predictor of each type of delinquency. Controlling for overall delinquency at Time 1, results under model 4 show that parental monitoring at Time 1 had a positive but nonsignificant effect on overall delinquency at Time 2, failing to support H2. The same was found for parental monitoring at Time 1 on property and violent delinquency at Time 2 (models 5 and 6), failing to provide any support for H2. Unlike parental attachment, parental monitoring at Time 1 does not have an effect on any of the three types of delinquency at Time 2 examined; there are no parental effects in these models. The existence of rules in the household does not appear to have a significant influence on changes in self-report delinquency levels.

The final section of Table 5 shows the results for the three models where parental involvement is a predictor of each type of delinquency outcome. Similar to parental monitoring, models 7–9 show that parental involvement at Time 1 has no significant effect on any of the three types of delinquency at Time 2; the coefficients are all positive but nonsignificant, failing to provide any support for H3. This indicates that the level of parental involvement has little or no impact on the delinquency levels of the child. In sum, parental attachment has a significant effect on each type of delinquency, however, parental monitoring and parental involvement had no significant effects on any of the delinquency measures.

All prior models have included each parenting measure in separate regressions. Though this shows the impact of each individual parenting behavior on each delinquency outcome, it may also be useful to test the effect of all three parenting measures in one model, as these three parenting behaviors are not likely to be practiced in isolation. The results of this test are not shown but indicated that parental attachment swamps the other parenting measures in predicting overall (b = −0.145 for attachment, b = 0.002 for monitoring, b = 0.017 for involvement), property (b = −0.142 for attachment, b = 0.002 for monitoring, b = 0.019 for involvement), and violent delinquency (b = −0.206 for attachment, b = −0.007 for monitoring), with the exception of a small significant effect of parental involvement on violent delinquency (b = 0.028). This indicates that given all three possible parenting behaviors, parental attachment likely has the greatest impact on delinquency, above and beyond any effect of parental monitoring or parental involvement.

The Effect of Delinquency on Parenting

Table 6 presents the results of the regression analyses predicting each type of parenting behavior using each type of delinquency. There are nine models in Table 6 (10–18), providing examination of the second half of hypothesized bidirectional effects (H4, H5, and H6). Each column of coefficients represents a separate parenting outcome: parental attachment in column 1, parental monitoring in column 2, and parental involvement in column 3. The coefficients reflect changes in each type of parenting over time, controlling for demographic variables and prior parenting behaviors.
Table 6

Time 2 parenting outcomes regressed on time 1 delinquency predictors

Variables

Overall Del. predicting parenting

Property Del. predicting parenting

Violent Del. predicting parenting

Attachment

Monitoring

Involvement

Attachment

Monitoring

Involvement

Attachment

Monitoring

Involvement

Model 10

Model 11

Model 12

Model 13

Model 14

Model 15

Model 16

Model 17

Model 18

Overall delinquency

0.02 (0.004)***

0.02 (0.01)

0.02 (0.01)*

Property delinquency

 

0.03 (0.01)***

0.02 (0.01)

0.03 (0.01)*

Violent delinquency

 

0.04 (0.01)**

0.03 (0.04)

0.05 (0.05)

Parenting

0.60 (0.02)***

0.40 (0.01)***

0.39 (0.02)***

0.60 (0.02)***

0.40 (0.01)***

0.39 (0.02)***

0.61 (0.02)***

0.40 (0.01)***

0.39 (0.02)***

Female

−0.06 (0.01)***

0.03 (0.04)

0.39 (0.05)***

−0.06 (0.01)***

0.03 (0.04)

0.39 (0.05)***

−0.06 (0.01)***

0.03 (0.04)

0.39 (0.05)***

Grades

0.03 (0.01)**

0.08 (0.03)**

0.25 (0.03)***

0.03 (0.01)**

0.08 (0.03)**

0.25 (0.03)***

0.03 (0.01)**

0.08 (0.03)*

0.25 (0.03)***

Two-parent family

−0.01 (0.02)

−0.18 (0.04)***

0.19 (0.05)***

−0.01 (0.02)

−0.18 (0.04)***

0.19 (0.05)***

−0.004 (0.02)

−0.18 (0.04)***

0.19 (0.05)***

Age

−0.001 (0.004)

0.17 (0.01)***

−0.11 (0.02)***

−0.001 (0.004)

0.17 (0.01)***

−0.11 (0.02)***

−0.000004 (0.004)

0.17 (0.01)***

−0.11 (0.02)***

Parent’s education

−0.0001 (0.003)

0.02 (0.01)*

0.08 (0.01)***

−0.0004 (0.003)

0.02 (0.01)*

0.08 (0.01)***

−0.002 (0.003)

0.02 (0.01)*

0.08 (0.01)***

Race

 Hispanic

0.02 (0.02)

−0.28 (0.08)***

0.06 (0.07)

0.02 (0.02)

−0.28 (0.08)***

0.06 (0.07)

0.01 (0.02)

−0.28 (0.08)***

0.06 (0.07)

 Black

0.06 (0.02)***

−0.16 (0.05)**

−0.02 (0.08)

0.05 (0.02)***

−0.15 (0.05)**

−0.03 (0.08)

0.06 (0.02)***

−0.16 (0.05)**

−0.02 (0.08)

 Other

−0.03 (0.03)

−0.18 (0.07)*

−0.15 (0.10)

−0.03 (0.03)

−0.18 (0.07)*

−0.15 (0.10)

−0.04 (0.03)

−0.18 (0.07)*

−0.16 (0.10)

Delinquent peers

−0.01 (0.003)

0.02 (0.01)*

−0.02 (0.01)

−0.01 (0.003)*

0.02 (0.01)**

−0.02 (0.01)

−0.01 (0.003)**

0.02 (0.01)**

−0.02 (0.01)*

Constant

1.66 (0.11)***

0.58 (0.21)**

2.74 (0.32)***

1.65 (0.11)***

0.59 (0.21)**

2.72 (0.33)***

1.60 (0.12)***

0.60 (0.21)**

2.73 (0.33)***

Standard errors are in parentheses

p < 0.05, ** p < 0.01, *** p < 0.001

The first section of Table 6, model 10, shows the results of regressing overall delinquency at Time 1 (plus controls) on parental attachment at Time 2. Overall delinquency had a significant negative effect on parental attachment (b = −0.019, s.e. = 0.004, p = 0.001), indicating that those respondents who reported lower levels of delinquency at Time 1 reported their attachment to their parents as higher at Time 2, indicating a child effect on parental attachment and providing partial support for H4. Similar to the lack of effects found for parental monitoring on delinquency, overall delinquency at Time 1 has no significant impact on parental monitoring at Time 2 (model 11), failing to provide support for H5. Finally, as with parental attachment, overall delinquency had a significant negative effect on parental involvement (b = −0.023, s.e. = 0.011, p = 0.041), indicating another child effect (model 12). Respondents who reported lower levels of delinquency at Time 1 were more likely to report higher levels of parental involvement at Time 2, providing partial support for H6. It appears that parents spend more time involved in the lives of their children when those children are not engaging in delinquency.

The results for the models including property delinquency as a predictor of each type of parental behavior are in the second section of Table 4 (13–15). Controlling for parenting at Time 1, property delinquency at Time 1 had a significant negative effect on parental attachment at Time 2 (b = −0.027, s.e. = 0.005, p = 0.001), indicating evidence of a child effect, providing partial support for H4. Youth who reported higher levels of property delinquency at Time 1 also reported lower levels of parental attachment at Time 2. The effect of property delinquency at Time 1 on parental monitoring at Time 2 was positive but not significant, failing to provide support for H5; the property delinquency of the child had little or no impact on change in the monitoring behavior of the parents. Property delinquency at Time 1 did, however, have a significant effect on parental involvement at Time 2 (b = −0.025, s.e. = 0.014, p = 0.043), providing partial support for H6.

Finally, the results for the violent delinquency predictors are in the third section of Table 6, models 16–18. As shown for model 16, violent delinquency at Time 1 had a significant effect only on parental attachment, providing final support for H4; respondents who reported engaging in violent delinquency at Time 1 were less likely to report high levels of parental attachment at Time 2 (b = −0.038, s.e. = 0.013, p = 0.004). Violent delinquency at Time 1 did not have a significant impact on change in parental monitoring or change in parental involvement, failing to provide final support for H5 or H6. Hypothesis 5 received no support from any of the models; hypothesis 6 received some support.

In conclusion, the only models that showed evidence of bidirectional relationships (both parent and child effects) were the parental attachment models. Parental attachment was associated with a significant negative change in each of the three types of delinquency; and each of the three types of delinquency were associated with a significant negative change in parental attachment. The other significant effects found were unidirectional in nature. Involvement at Time 1 had no significant association with change in any of the three types of delinquency at Time 2 (no parental effects), but overall delinquency and property delinquency did have a significant association with changes in parental involvement (two child effects). Figure 1 shows the results of all eighteen models in a diagram for easier review.
https://static-content.springer.com/image/art%3A10.1007%2Fs10964-011-9656-4/MediaObjects/10964_2011_9656_Fig1_HTML.gif
Fig. 1

Summary of model results

Discussion

The family, and parenting in particular, is consistently recognized as an important predictor of adolescent delinquency. One theory that focuses on this relationship is social control theory, or the social bonds framework, whereby the effect of parenting behavior on youth is examined. However, this theory fails to take into account the effect of adolescent delinquency on parenting itself. When child effects are not taken into account, the effects of parental behaviors may be overestimated. Criminological research that actively measures the magnitude of a bidirectional relationship between parent and youth is quite rare. The field still has relatively little compelling evidence of reciprocal relationships between parental behaviors and delinquency due to limitations of prior research, especially in regards to limited samples (e.g. Hawkins et al. 2007; Jang and Smith 1997). A few studies, however, have shown evidence of reciprocal relationships between parental attachment and delinquency (e.g. Liska and Reed 1985) and of parental monitoring and delinquency (Kerr et al. 2010), though not parental involvement and delinquency (e.g. Coley and Medeiros 2007). The current study examined the bidirectional effects of parenting and delinquency using a nationally representative sample of 12,505 adolescents. Using lagged effects in the models, the analysis examines both parental and child effects in an effort to establish whether or not there is a bidirectional relationship between parental behaviors and delinquency.

One main research question guided this research. Is there a reciprocal relationship between parenting behaviors and delinquency? In other words, are there both parental effects on delinquency and child effects (through delinquency) on parenting? To answer this question, this study employed cross-lag regressions using data from Add Health. Results showed that there are bidirectional effects between parental attachment and each of the three types of delinquency. The level of parental attachment negatively influenced the prevalence of delinquent behavior in youth; conversely, the level of delinquency in which the youth engaged negatively influenced the level of parental attachment. There were no bidirectional effects for parental monitoring or parental involvement. Rather, no significant parental effects between parental monitoring and delinquency or between parental involvement and delinquency were found, and no significant child effects between parental monitoring and delinquency were found. Overall delinquency and property delinquency both had negative effects on parental involvement, providing evidence of child effects. Parental involvement does not prevent delinquency; rather, an increase in delinquency is associated with less parental involvement.

The significant effect of parental attachment on delinquency is supportive of the social bonds framework. Attachment to one’s parent(s) can serve as a protective force against delinquent behavior, above and beyond variables already known to be associated with delinquency, such as race, gender, SES, and especially delinquent peer associations. This is consistent with prior research on the link between attachment and delinquency (e.g. Hirschi 1969; Laub and Sampson 1988). Consistency with prior research lends credence to the findings in this study, but this study goes further than prior studies of parental attachment and delinquency by examining child effects as well.

The findings regarding parental attachment also provide strong evidence for the existence of a reciprocal relationship between parenting and delinquency, consistent with both the transactional and interactional models of reciprocal parent–child relationships but inconsistent with social bonds theory. Contrary to prior criminological research that has focused on or assumed a unidirectional relationship of parents on children, the data here show that the influence goes both ways in the relationship between parental attachment and delinquency. Parents, as posited by the social bonds framework, can and do affect the risk of delinquency in their children through the level of parental attachment. However, in addition to the unidirectional proposition of the social bonds framework, the data here show that the behavior of the child also influences the parent in that children who engage in delinquency report lower levels of parental attachment. If the current data had additional waves instead of only two with similar questions, we might find that this is a cycle of weakening bonds and increasing delinquency, at least for the period of adolescence, but that is for future research to determine.

The lack of a significant effect of parental monitoring or parental involvement, when framed as a type of parental attachment, on any type of delinquency is in contradiction to the social bonds framework, though it may be due at least partially to the way these variables were measured in the current study. In the past, a measure of parental monitoring would likely include questions about the parents’ knowledge of the adolescent’s whereabouts when not at home, who they are with, and what they are doing. This has been criticized, however, for relying on and essentially measuring adolescent disclosure, the argument being that parents’ knowledge of their teenager’s activities is mostly dependent upon the teenager telling the parent what they are doing and with whom they are doing it. This form of measuring parental monitoring is not used here because these questions are not asked in Add Health; however, even if they were, a measure created using them would be subject to these same criticisms. Instead, the measure used in the current study is comprised of questions regarding the existence of household rules. Some of the items included tap into parental knowledge concerning the adolescent’s activities, such as requiring a curfew or monitoring what they see on television. Though it avoids the issue of child disclosure, it is certainly not a perfect measure of parental monitoring as it does not effectively measure parental knowledge, an important aspect of parental monitoring, however, flawed previous ways of measuring it may have been. Perhaps future studies should aim to use questions that ask adolescent respondents how often their parents ask about their activities, location, and associations, as well as how often the respondent honestly answers.

Parental involvement as measured here is a good measure of the level of involvement between parent and adolescent; however, it is not the kind of involvement Hirschi (1969) included in his formulation of social control theory. In his theory, involvement is not limited to parental involvement; it is meant as a measure of involvement in a range of conventional activities resulting in loss of opportunity to engage in delinquency. The measure of involvement in the current study does give an indication of loss of opportunity to engage in delinquency, but it refers specifically to activities that parents and adolescents share and is therefore also another way of measuring attachment. Presumably parents and adolescents who have a high level of attachment spend more time together than those who are less attached, while those who do not feel a strong bond with each other would be less likely to spend large amounts of time together.

The existence of child effects on two of three types of parental behaviors suggests the need for criminological theories that allow for reciprocal relationships between parents and children. As recent evidence in the criminological literature has suggested, children influence the type of parenting they receive which can, in turn, further influence the behavior of the child, at least in the case of parental attachment. The social bonds framework argues, in part, that delinquency is caused by weakened social bonds, such as low parental attachment. The current study showed that not only does low parental attachment increase the risk of delinquency, but that delinquency reduces parental attachment. Given the social bonds framework, then, this subsequent lowered parental attachment should lead to further delinquency.

The current study built on and expanded upon important previous studies that examined the bidirectional relationship between parenting and delinquency, and the finding of a bidirectional relationship between parental attachment and delinquency is consistent with most of what can be found in the criminological literature on reciprocal relationships. Both significant parental effects and child effects were found by Liska and Reed (1985), Thornberry, Lizotte, Krohn, Farnworth, and Jang (1991), Buist, Dekovic, Meeus, and Van Aken (2004), though not by Agnew (1991), who found no evidence of a child effect and only weak evidence of an indirect effect of parental attachment on delinquency through commitment to school.

In their study of high-risk youth, Jang and Smith (1997) found evidence of reciprocal effects between parental monitoring and delinquent behavior. The difference in their results and the results of the current study may be due to differences in the sample. For youth who are already at an increased risk of delinquent behavior, closer parental monitoring may indeed reduce that risk, whereas youth who are not at increased risk for delinquency may not be as influenced by the monitoring activities of their parents. It may be that parental monitoring only influences the delinquency of high-risk youth, as found by Jang and Smith (1997), and not a more general adolescent population, as was studied here.

Contrary to this idea, Stewarts et al. (2002) also found reciprocal effects between parental supervision and delinquency using a sample of Iowa youth who were predominantly white and predominantly from two-parent families, characteristics that would not typically be deemed as placing the youth at “high risk.” As will be discussed in the limitations section, however, these differences may be due to the way the measure was created. Very different measures of supervision were used in the current study and in the study by Stewarts and colleagues.

The parental involvement effects are surprising in that involvement does not protect against delinquency, but it is not surprising that there are adolescent effects on involvement. A delinquent adolescent can cause strain, frustration, anger, or hurt on the part of the parent, in which case spending large amounts of time with the adolescent may not be a priority of the parent. It is disheartening and frustrating to constantly spend time with your adolescent who is causing or getting into trouble, especially when the time spent with them doesn’t seem to be doing any good. It is also possible that delinquent youth are less drawn to spending time with their parents. These findings do not appear consistent with the literature, but differences in sample and outcome may be the reason.

Ream and Savin-Williams (2005) found evidence of reciprocal effects between parental involvement and adolescent behavior, but their study focused on the sexual activity of the youth rather than delinquency per se. Coley and Medeiros (2007) found the opposite, that parental involvement predicted delinquency but not that delinquency predicted parental involvement. Theirs was, however, a restricted sample in that the only parental involvement measured involved nonresident father involvement of low-income, minority youth. Using a more representative sample than Coley and Medeiros (2007)—the Add Health data—Hawkins et al. (2007) also examined the relationship between nonresident father involvement and delinquency and found effects similar to those found in the current study, that non-resident father involvement did not predict delinquency but that delinquency did predict nonresident father involvement.

These differences in the relationship between involvement and delinquency or other adolescent outcomes may be due to differences in the outcome measured, or they may be due to the ages of respondents. For example, the Coley and Medeiros (2007) study examined slightly younger children than the Hawkins et al. (2007) study or the current study, both of which found child effects only. It is possible that younger children are more influenced by their parents wanting to spend time together, whereas older youth are less positively influenced by parental involvement because they are becoming more autonomous. Perhaps being forced to go to the movies with one’s parents is not helpful to teenagers whereas being allowed to go to the movies with one’s parents is helpful as a younger child. These data, however, do not allow us to examine the “why” behind time spent together, so we cannot know whether the respondent wanted the parent to be involved.

The current study has several advantages over previous work in this area. First, nationally representative data was used, allowing for greater generalization of the results. Prior studies had sample limitations such as examining only high-risk youth, only male youth, or predominantly white youth. Second, the data used here were longitudinal in nature. This allowed for the analysis of the data using cross-lag regressions. Cross-lag regressions permit the examination of change in the outcome over time by controlling for the outcome at a prior time period. In the case of the current study, I was able to examine the changes in parenting by controlling for prior parenting in each model; the same was true for examining changes in delinquency. Third, multiple measures of parenting and multiple measures of delinquency were included, allowing for a more thorough examination of the process between parents and children in relation to delinquency using all the same data. Finally, association with delinquent peers was controlled which, given its strong relationship with both delinquency and parenting, is an important and necessary control that many prior studies of the reciprocal relationship between parenting and delinquency have lacked.

Implications for the Social Bonds Framework

The current study demonstrates the importance of considering reciprocal relationships between parenting behaviors and delinquency, using a nationally representative, longitudinal data set of youth. The social bonds framework has enjoyed a long history of empirical support and theoretical discussions, but the evidence is mounting that the framework is incomplete. This study supports the conclusion that the assumption of unidirectional effects between parents and children within the social bonds framework is an inaccurate picture of how this process works. Efforts to improve the social bonds framework must address the issue of reciprocal influences in the development of delinquent behavior. Parents and children are indeed involved in a dynamic, interactive relationship comprised of both parent and child effects. Criminological theories, especially those that fit within the social bonds framework, need to take this into account when attempting to explain what causes (or controls) delinquency in youth as evidence continues to show that the youths themselves have an impact on this process; as Laub and Sampson (2003) argued, human agency exists and is an important consideration in the explanation of delinquency. Adolescents have a personal choice regarding their behavior, and this choice may be influenced by social bonds such as attachment to parents. Future studies of the relationship between parents and delinquent behavior should investigate bidirectional effects in order to account for human agency.

Directions for Future Research

Future studies of the process leading up to delinquent behavior should examine both parent and child effects. The current study suggests that both parents and children influence the processes that culminate in delinquency. In light of the findings and limitations of the current study, future studies are needed that examine more than two waves and that includes a broader range of ages of youth. This would allow us to understand how the reciprocal relationship between parenting and delinquency changes as children age, so that we may better understand relative importance of parent and child effects as children develop. While Add Health has a third and fourth wave, the respondents are no longer youth living at home, changing the meaning of parental behaviors, and respondents are not asked the same questions in these waves as in the first two waves used in the current study. Inclusion of young adults may also be of interest so that in the future we can test how and when this process shifts away from parental influence and toward other forms of informal social control.

In addition, future studies should include more exhaustive measures of parental behaviors and include multiple sources of data collection. Scales of parental behaviors would benefit from having not just children responding to the survey items but parents as well; in fact, future studies should also examine mother and father effects separately to see if the process is specific to one parent or if it describes “parents’ in general. Because the available data for both waves include only youth responses regarding parental attachment, what the scale truly measures is perceived parental attachment. The youth are answering the questions regarding parent–child closeness, parent–child communication, etc. from their point of view. Parents may have a different point of view concerning the state of the relationship, though the small test run in the current study does not indicate that this is the case. A model of parental attachment predicting delinquency was run using a parental attachment scale created using parent responses, and the results were the same: parental attachment is significantly negatively associated with delinquency.

An ideal measure of parental monitoring would likely include questions about the parents’ knowledge of the child’s whereabouts when not at home, who they are with, and what they are doing. The measure used in the current study is comprised of questions regarding the existence of household rules, the only monitoring behaviors available in the survey. Some of the items included tap into parental knowledge concerning the child’s activities, such as requiring a curfew or monitoring what they see on television, but it is certainly not a perfect measure of parental monitoring. Future research should include these more ideal questions, though finding them in a nationally representative dataset is challenging.

Finally, future research should test whether these bidirectional influences are truly lagged or whether they might be instantaneous. For the current study, the assumption was made that the effects of parents on children and children on parents would be lagged, as it seems that the parent–child relationship is something that develops over time, rather than instantaneously. However, it is possible that there are also instantaneous effects, or that one of the directions of effects—parent or child—may be lagged while the other is instantaneous. Perhaps for parents, their child’s pattern of behavior is what influences their parenting decisions, in which case a lag of approximately one year might be reasonable, however, it may be that youth react to current (instantaneous) parental behaviors, rather than reacting to the pattern of their parent’s behavior. For youth, it may be more instantaneous; if a parent increases the harshness of the household rules in reaction to the behavior of the child, whether it is to a pattern of delinquent behavior or current delinquent behavior, the child may very well react to that change instantly, acting out in anger, frustration, or strain. Future research should test for these different possibilities.

This study shows that parents and children are in a dynamic, interactive relationship and that failure to consider the possibility of both parent and child effects limits our understanding of this important developmental process. Though criminologists began to examine the possibility of such reciprocal relationships more than twenty years ago, few studies have actually tested this relationship. The current study built on the seminal article by Liska and Reed (1985), the first truly criminological examination of this reciprocal process. This study was followed by several others, such as the 1997 study by Jang and Smith upon which the current study also built. However, recent studies of the reciprocal relationship between parenting and delinquency has focused on select samples, such as only those children or youth with nonresident fathers (Coley and Medeiros 2007; Hawkins et al. 2007). In fact, the only criminological examination of this relationship using a nationally representative sample was by Agnew (1991), who found no reciprocal relationship between parental attachment and delinquency.

The current study was undertaken to address limitations of prior criminological research on the reciprocal relationship between parenting and delinquency and to examine the consequences of a fundamental problem within the social bonds framework. The findings provided evidence of this bidirectional relationship when analyzed using cross-lagged regressions. Future research that includes simultaneous analyses and different measures of parental monitoring may provide further insight into the parent–child relationship as related to the serious topic of adolescent delinquency.

Footnotes
1

A comparison of means and standard deviations of all variables before and after imputation are shown in Table 3. This table also shows the number of values imputed for each variable.

 

Acknowledgments

The author wishes to thank Eric Silver, Darrell Steffensmeier, Jeremy Staff, Daphne Hernandez, and Jerry Park for their contributions to this work.

Copyright information

© Springer Science+Business Media, LLC 2011