Abstract
Theoretical and empirical research has linked variation in parental and peer socialization patterns as well as criminogenic traits, particularly psychopathy and low self-control, to criminal involvement. Findings from this body of scholarship, however, have generally been produced without adequately controlling for genetic confounding. The current study addresses this gap in the literature by analyzing data from the National Longitudinal Study of Adolescent to Adult Health using a genetically informative research design. This study employs monozygotic difference scores analyses in order to examine the effects of psychopathic personality traits (PPTs), low self-control, and nonshared environmental factors on involvement with criminal behavior while controlling for genetic factors. The results indicated that of the four outcomes examined, PPTs were only associated with involvement in violent behavior. In addition, the results revealed that delinquent peers was the only nonshared environmental factor associated with any of the outcome measures.
Mainstream criminological research has tended to focus on environmental explanations for criminal behavior while downplaying the role of individual-level criminogenic factors (Gottfredson & Hirschi, 1990; Hirschi, 1969; Sampson, Raudenbush, & Earls, 1997). Traditional criminological explanations, for example, implicate families, peers, and neighborhoods in the development of criminal behavior and delinquency. There has been a slight shift recently with more and more research examining the potential role of individual-level factors in explanations for involvement with criminal behavior (Walsh & Wright, 2015). Findings from this area of research have revealed that psychopathy and low self-control are two trait-based characteristics consistently associated with criminal behavior (DeLisi, 2009; Pratt & Cullen, 2000).
Despite findings from studies that link environmental and individual-level factors with criminal behavior, these studies generally do not control for genetic confounding. As a direct result, these findings could be biased in some capacity, as previous research has revealed that failing to control for genetic confounding may lead to inflated parameter estimates for environmental variables (Barnes, Boutwell, Beaver, Gibson, & Wright, 2014; Harris, 1998; Wright J. P. & Beaver, 2005). In order to address this issue, the current study employs a research design that is capable of estimating the effects of nonshared environmental factors and individual-level factors while controlling for genetic influences.
Behavioral Genetics Research
Methodological and statistical techniques employed in behavioral genetics can be used to estimate the effects of genetics and environmental factors on phenotypic variance (Plomin, DeFries, McClearn, & Rutter, 1997). The environmental component of the variance of phenotypes is further broken down into shared and nonshared environmental factors. 1 Shared environmental factors refer to elements in the environment that make siblings similar to each other, whereas nonshared environmental factors refer to elements in the environment that make siblings different from each other. The majority of behavioral genetics studies are conducted using twin research designs where cross correlations on a trait are compared between monozygotic (MZ) twins and dizygotic (DZ) twins. MZ twins share 100% of their DNA and DZ twins share 50% of their dissenting DNA on average. As a result, it is possible to use comparisons between these two types of twins to determine the percentage of variance in phenotypes that is due to genetic and environmental factors. If the assumptions of twin-based research are met—and mathematical proofs and simulations indicate that they are (Barnes, Wright, et al., 2014)—then the greater similarity of MZ twins compared to DZ twins would be due to genetic influences.
A large body of behavioral genetics research has estimated the influence of genetic and environmental factors on antisocial behavior and criminal involvement. According to a recent meta-analysis that examined twin studies over the last 50 years, including tests of more than 17,800 traits and using more than 14,500,000 twin pairs, 49% of the variance in all human phenotypes, including personality and behavioral phenotypes, is due to genetic factors (Polderman et al., 2015). Of particular importance, genetic factors were found to account for 49% of the variance in conduct disorder, 44% of the variance in temperament and personality functions, and 62% of the variance in personality disorders, all of which have been linked with criminal involvement (Caspi et al., 1994; Fridell, Hesse, Jæger, & Kühlhorn, 2008; Mordre, Groholt, Kjelsberg, Sandstad, & Myhre, 2011; Murray & Farrington, 2010). Other meta-analyses that focus exclusively on antisocial behavior have revealed similar findings indicating that approximately 50% of the variance in antisocial behavior is due to genetic factors. The remaining 50% of the variance in antisocial behavior is explained by environmental factors with approximately 40–50% of the variance explained by nonshared environmental factors and approximately 0–10% of the variance explained by shared environmental factors (Ferguson, 2010; Mason & Frick, 1994; Miles & Carey, 1997; Rhee & Waldman, 2002).
Psychopathic Personality Traits (PPTs), Low Self-Control, and Criminal Involvement
A body of research has revealed a connection between trait-based characteristics and criminal involvement (Caspi et al., 1994; Krueger et al., 1994; Pratt & Cullen, 2000; Vazsonyi, Pickering, Junger, & Hessing, 2001). For instance, variation in psychopathy and low self-control has been found to be consistently associated with criminal and antisocial behavior (DeLisi, 2009; Pratt & Cullen, 2000). While these associations appear consistent in the literature, the majority of studies examining the associations between trait-based characteristics and crime have not controlled for genetic factors. As a result, findings from these studies may be confounded by genetic influences and may have produced upwardly biased estimates for environmental factors.
According to the existing research, psychopathy and PPTs are associated with violent, antisocial, and criminal behavior (DeLisi, 2009; Hare, 1993; Vaughn & DeLisi, 2008). Psychopathy is a personality disorder that is characterized by a combination of behavioral, affective, and lifestyle attributes. Psychopaths are generally characterized as irresponsible, impulsive, short tempered, lacking empathy, and lacking guilt (Hare, 1996). Previous research on psychopathy and criminal involvement indicates that psychopaths are responsible for a disproportionate amount of serious and violent crime (Blackburn & Coid, 1998; Hare, 1993; Vaughn, Howard, & DeLisi, 2008). For instance, estimates indicate that psychopaths may be responsible for more than 50% of serious crimes committed (Hare, 1993) and that psychopaths make up between 15% and 25% of prison populations (Hare, 1996).
Research examining the underpinnings of psychopathy suggests that variation in PPTs is primarily explained by genetic and nonshared environmental factors (Beaver, Vaughn, & DeLisi, 2013; Blonigen, Hicks, Krueger, Patrick, & Iacono, 2005; Taylor, Loney, Bobadilla, Iacono, & McGue, 2003; Waldman & Rhee, 2006). For example, a study by Viding, Blair, Moffitt, and Plomin (2005) revealed that genetic factors explain 67% of the variance in extreme callous–unemotional traits and 81% of the variance in antisocial behavior in children with psychopathic tendencies. Shared environmental factors, in contrast, account for only 6% of the variance in callous–unemotional traits and 0% of the variance in antisocial behavior in children with psychopathic tendencies. Additionally, a meta-analysis examining studies on psychopathy revealed that 49% of the variance in psychopathy is attributable to genetic factors and 51% of the variance is attributable to nonshared environmental factors (Waldman & Rhee, 2006).
Similarly, low self-control has also been consistently linked with involvement in antisocial behavior (Pratt & Cullen, 2000). People with low levels of self-control are described as being impulsive, irresponsible, self-centered, insensitive to others, and prone to risky behavior (Gottfredson & Hirschi, 1990). Empirical research on self-control indicates that low self-control is associated with involvement in delinquency, criminal behavior, antisocial behavior, and violence (Chapple, 2005; Evans, Cullen, Burton, Dunaway, & Benson, 1997; Vazsonyi et al., 2001; Wright B. R. E., Caspi, Moffitt, & Silva, 1999).
Research on the etiology of self-control indicates that genetics contribute to variation in levels of self-control and that levels of self-control may be primarily the product of genetic and nonshared environmental influences (Beaver, DeLisi, Vaughn, Wright, & Boutwell, 2008; Beaver, Wright, DeLisi, & Vaughn, 2008). These findings run directly counter to Gottfredson and Hirschi’s (1990) assertion that variation in levels of self-control is mainly the result of parental socializing and is not produced by genetic influences. To illustrate, according to a study by Beaver et al. (2009), 40–56% of the variance in self-control is explained by genetics, 44–60% of the variance is explained by nonshared environmental factors, and 0% of the variance is explained by shared environmental factors. Similarly, a study by Wright J. P., Beaver, DeLisi, and Vaughn (2008) found that genetic factors account for 25–40% of the variance in self-control and nonshared environmental factors account for the remaining 60–74% of the variance. In combination, these findings indicate that self-control is largely the product of genetic and nonshared environmental factors.
Research findings revealing that genetic factors explain approximately 50% of the variance in psychopathy and self-control indicate that it is necessary to control for genetics when examining factors predicted to explain criminal behavior in order to avoid genetic confounding. Consequently, studies that do not control for genetics are likely confounded and may report inflated estimates of the influence of environmental measures on involvement with criminal behavior and the criminal justice system (Harris, 1998; Wright J. P. & Beaver, 2005). A study by Nedelec and Beaver (2014) underscores this possibility. In this study, they compared estimates of the influence of self-control on social outcomes using standard social science methodologies (SSSMs) and a genetically informed research design. They employed DeFries–Fulker (DF) modeling in order to determine the proportion of the variance in the social outcomes that is attributable to genetic and environmental factors. Their findings revealed that the estimates of the influence of self-control on social outcomes using SSSMs were confounded. Specifically, the findings from the genetically informed design (DF), when compared to the SSSMs, revealed that the relationship between self-control and many of the outcomes either no longer reached significance or switched directions. That is, after controlling for genetic factors and shared environmental factors, the relationship between self-control and many social outcomes either decreased or was in the reverse direction than according to the findings from the SSSMs. These findings highlight the need to control for genetic factors while analyzing the development of criminal behavior.
The Current Study
Findings from a broad array of studies have revealed an association between variation in trait-based characteristics and criminal behavior. The robustness of these findings, however, is difficult to evaluate because they generally fail to account for the possibility of genetic confounding. The current study addresses this gap in the literature by employing an MZ difference scores analysis in order to determine if twin differences in trait-based characteristics and nonshared environmental factors during adolescence is associated with involvement in criminal behavior in adulthood while controlling for genetic factors.
Method
Data
The current study employs data drawn from the National Longitudinal Study of Adolescent to Adult Health (Add Health; Udry, 2003). The Add Health is a nationally representative and longitudinal sample of adolescents who were enrolled in middle or high school in 1994–1995. The sample consists of four waves of data. The first wave of data was collected from more than 90,000 adolescents and covered topics related to daily activities, relationships with parents, friends, and participation in delinquent behavior. The second wave of the survey was administered in 1996 and covered many of the same topics as Wave 1. The third wave of the survey was conducted in 2001–2002 when most of the subjects were between the ages of 18 and 26. The fourth wave was completed in 2007–2008 when the respondents were between the ages of 24 and 32 and was completed by more than 15,000 of the original participants from Wave 1 (Harris et al., 2003). Waves 3 and 4 of the survey covered topics similar to the first wave, but the surveys were adjusted to ask questions that were more age appropriate. For example, Waves 3 and 4 asked questions pertaining to educational histories, marital status, labor force participation, and involvement with the criminal justice system.
The Add Health survey oversampled siblings and twins allowing for genetic analysis of the data. During the first wave of the survey, respondents were asked to indicate if they were part of a twin pair or if they lived with a sibling, stepsibling, or cousin. If the respondent indicated that they were part of a twin pair, the co-twin was added to the sample. In addition, if respondents indicated that they lived with a sibling, and the sibling was between the ages of 11 and 20, then their sibling was also included in the sample. The full sibling and twin sample in Add Health contains more than 3,000 sibling pairs and includes 289 MZ twin pairs. After removing cases with missing data using listwise deletion, the final analytic sample for this study spanned from 172 to 173 MZ twin pairs.
Measures
Outcome measures
Violent criminal behavior
A Violent Criminal Behavior Scale was created by using six questions asked at Wave 4. Specifically, participants were asked how often in the last 12 months they deliberately damaged property, were in a serious physical fight, hurt someone bad enough to need care from a doctor or a nurse, were involved in a group fight, shot or stabbed someone, and pulled a knife or a gun on someone. These items were coded so that higher values indicate higher levels of involvement with violent criminal behavior. Responses to these items were summed together to create a Violent Criminal Behavior Scale (α = .603). This scale is similar to previous scales that have been used to measure involvement in violent delinquency and violent crime (Guo, Roettger, & Shih, 2007). Descriptive statistics for this measure and all of the other measures included in the analyses are presented in Table 1 (See Appendices A and B for a correlation matrix and descriptive statistics of the original variables).
Descriptive Statistics for MZ Difference Scores Included in the Analyses.
Note. MZ = monozygotic; CJ = Criminal Justice.
*p < .05.
Nonviolent criminal behavior
A Nonviolent Criminal Behavior Scale was created using 5 items asked at Wave 4. For example, participants were asked how often in the last 12 months they stole something worth more than US$50, stole something worth less than US$50, sold drugs, bought or sold stolen property, and wrote bad checks. These items were coded so that higher values indicate higher frequencies of involvement with nonviolent criminal behavior. Answers to these items were summed together in a Nonviolent Criminal Behavior Scale (α = .532). This scale is similar to previous scales that have been used to measure nonviolent criminal behavior and nonviolent delinquency (Guo et al., 2007).
Arrest
Arrest was measured using a single item asked during Wave 4. Respondents were asked to indicate if they had ever been arrested. Responses were coded so that 0 = no and 1 = yes. 2
Criminal justice involvement
A Criminal Justice Involvement Scale was created using four questions regarding involvement with the criminal justice system. At Wave 4, respondents were asked if they had ever been arrested, if they have ever been convicted of charges more serious than a traffic violation, if they have ever been on probation, and if they have ever been incarcerated. Items were coded so that 0 = no and 1 = yes. Answers to these questions were summed together to create a Criminal Justice Involvement Scale (α = .520).
Trait-based measures
Psychopathic personality traits (PPTs)
A continuous measure of PPTs was created using items measured at Wave 4. The items used in this scale were originally intended to measure personality traits according to the five-factor model (FFM). While there is a currently a debate in the psychopathy literature discussing the appropriate measures to use to tap PPTs, there is a significant body of research suggesting that continuous PPTs measures constructed from the FFM are a valid and reliable way to measure psychopathy (Derefinko & Lynam, 2007; Lynam & Miller, 2015; Lynam et al., 2005; Miller & Lynam, 2015). 3 In total, this scale includes 23 items that measure the interpersonal, affective, and behavioral attributes of psychopathy. The items in this scale were coded so that higher values indicate higher levels of PPTs. All of the items were standardized and then summed together to make a scale of PPTs (α = .814). This scale of PPTs is identical to PPTs Scales used previously with Add Health data (Beaver, Barnes, May, & Schwartz, 2011).
Low self-control
A wealth of previous research links low levels of self-control with increased involvement in delinquency and criminal behavior (Pratt & Cullen, 2000). In line with previous research, we constructed a scale for low self-control from five questions asked during Wave 1 (Perrone, Sullivan, Pratt, & Margaryan, 2004). For example, the respondents were asked how well they could concentrate and focus in school and how well they got along with their teachers. Answers to these items were coded so that higher values indicate lower levels self-control. These items were standardized and then summed together in a Low Self-Control Scale (α = .664). This Low Self-Control Scale is similar to previous Low Self-Control Scales that have been used with Add Health data (Beaver, 2008).
Environmental measures
Maternal involvement
A Maternal Involvement Scale was included at Wave 1 in order to control for the influence of maternal involvement during adolescence on later involvement in criminal behavior. In line with previous research (Crosnoe & Elder, 2004), we constructed a Maternal Involvement Scale using 10 items from Wave 1 regarding activities the respondents had participated in with their mothers over the previous 4 weeks. Respondents, for instance, were asked if they had gone to a movie, went shopping, or played a sport with their mother in the last 4 weeks. Responses were coded so that 0 = no and 1 = yes. Responses to these items were summed together in a Maternal Involvement Scale (α = .553).
Maternal attachment
We have included a Maternal Attachment Scale at Wave 1 in order to control for maternal attachment in adolescence on the prediction of involvement with criminal behavior in adulthood. The Maternal Attachment Scale was constructed by summing together answers from two questions concerning emotional connectedness between respondents and their mothers. Specifically, respondents were asked how much they think their mothers care about them and how close they feel to their mother. Responses to these questions were summed together in a Maternal Attachment Scale (α = .640). This scale is similar to Maternal Attachment Scales that have been in used in previous research with Add Health data (Beaver, 2008).
Maternal rejection
We have created a Maternal Rejection Scale at Wave 1 in order to control for maternal rejection during adolescence in the prediction of later criminal behavior. A Maternal Rejection Scale was created by summing together answers to five questions regarding respondent’s relationships with their mothers. Respondents, for example, were asked how much they talked with their mother, how warm their mother is, and the overall quality of their relationship with their mother. Items were coded so that higher values indicate a greater level of maternal rejection. Responses to these items were summed together in a Maternal Rejection Scale (α = .836).
Parental permissiveness
We have included a scale for parental permissiveness in order to control for parental permissiveness in the prediction of criminal behavior. The Parental Permissiveness Scale was constructed using answers to seven questions asked at Wave 1 concerning how permissive respondent’s parents were. For example, respondents were asked if their parents let them make their own decisions concerning what they ate, about their friends, and about curfews. Answers to these items were coded so that 0 = no and 1 = yes. Answers to these items were combined together in a Parental Permissiveness Scale (α = .631). This scale is coded so that higher values indicate higher levels of parental permissiveness. This scale is identical to Parental Permissiveness Scales that have been used previously with Add Health data (Beaver et al., 2013).
Delinquent peers
We constructed a Delinquent Peers Scale in order to control for the influence of delinquent peers in adolescence on the prediction of involvement with criminal behavior during adulthood. Delinquent peers may be an important environmental factor influencing later criminal behavior as associating with delinquent peers has been shown to be consistently associated with involvement in delinquent and criminal behavior (Haynie, 2001; Warr, 2002). The Delinquent Peers Scale was constructed using 3 items concerning peer involvement in substance use at Wave 1. Respondents were asked how many of their three best friends drink alcohol at least once a month, smoke marijuana more than once a month, and smoke at least one cigarette a day. These items were coded so that higher values indicate higher levels of delinquent peer behavior. Responses to these items were summed together in a Delinquent Peers Scale (α = .756). This scale of delinquent peers is identical to scales of delinquent peers that have been used with Add Health data previously (Beaver, 2008).
Analytic Strategy
This study implements an MZ difference scores analysis in order to determine the effects of trait-based characteristics and nonshared environmental factors on involvement with criminal behavior. The MZ difference scores analysis is used to determine how differences in PPTs and self-control between MZ twins influence differences in involvement with criminal behavior. MZ twins share 100% of their DNA and therefore phenotypic differences between MZ twins can only be explained by environmental factors that differ between them. MZ difference scores can be used to determine which nonshared environmental factors are responsible for variation in behavioral phenotypes by subtracting one MZ twin’s score on an outcome from their co-twins score. As a result, MZ difference scores are a valuable analytic strategy for empirical research, as they are able to assess the influence of environmental factors while controlling for genetic factors. Due to the simplicity of the analysis and the ability to control for genetics and shared environmental factors, the MZ difference scores approach is considered the “gold standard” for establishing causality in observational data for nonshared environmental factors (Asbury, Dunn, Pike, & Plomin, 2003; Beaver, 2008; Caspi et al., 2004; Pike, Reiss, Hetherington, & Plomin, 1996; Rovine, 1994; Vitaro, Brendgen, & Arseneault, 2009). MZ difference scores have been used previously to determine the influence of nonshared environmental factors on behavioral phenotypes such as depression (Kendler & Gardner, 2001), criminal involvement (Beaver, 2008), and delinquent behavior (Caspi et al., 2004; Pike et al., 1996).
The analysis of this study proceeded in three steps. First, one twin from each pair was randomly selected as Twin 1 and the other was selected as Twin 2. Second, for this analysis, MZ difference scores were created by subtracting Twin 2’s scores from Twin 1’s scores for each variable. MZ difference scores were calculated for each of the criminal involvement indicators and Criminal Involvement Scales along with all of the predictor variables. 4 Third, after the MZ difference scores were created, ordinary least squares regression analysis was used to test for associations between the criminal outcome measures and the predictor variables. 5
Results
The analysis began by examining the effects of the nonshared environmental variables and the trait-based measures on involvement with criminal behavior. As can be seen in Table 2, PPTs are positively associated with involvement in violent criminal behavior. This finding indicates that the twin with a higher score on the PPTs Scale, relative to their co-twin, reported more involvement with violent criminal behavior. Further examination of Table 2 reveals that none of the nonshared environmental measures are significantly associated with involvement in violent criminal behavior or nonviolent criminal behavior. In addition, low self-control does not appear to explain a significant amount of variation in involvement with violent or nonviolent criminal behavior.
The Effects of Trait-Based Differences and Nonshared Environmental Factors on Monozygotic Differences in Criminal Behavior.
Note. SE = standard error.
*p < .05.
Next, we examined the effects of the nonshared environmental variables and the trait-based measures on involvement with the criminal justice system. Examination of Table 3 reveals that the only nonshared environmental factor to explain a significant amount of variation in arrest is delinquent peers. This finding indicates that the twin with a higher score on the Delinquent Peers Scale, relative to their co-twin, was more likely to report an arrest. None of the other nonshared environmental variables are significantly associated with arrest or the Criminal Justice System Involvement Scale. Additionally, neither of the trait-based measures are significantly associated with arrest or involvement with the criminal justice system.
The Effects of Trait-Based Differences and Nonshared Environmental Factors on Monozygotic Differences in Criminal Justice Outcomes.
Note. SE = standard error.
*p < .05.
Discussion
Findings from studies consistently reveal that trait-based characteristics and environmental factors are associated with criminal behavior, but the majority of these studies do not take into the account the role of genetics. The current study addressed this gap in the literature by using MZ difference scores analysis to examine the influence of PPTs, low self-control, and nonshared environmental factors on involvement with criminal behavior and the criminal justice system. Results of the MZ difference scores analyses revealed two main findings.
First, PPTs were found to be associated only with involvement in violent criminal behavior even after controlling for genetic factors. The significant association between PPTs and violent criminal behavior is consistent with a body of literature documenting that psychopaths are disproportionately involved in serious and violent crime (DeLisi, 2009; Hare, 1993). Keep in mind, however, that PPTs were not related to any of the other outcomes. These findings are in contrast to much of the literature on PPTs (DeLisi, 2009; Hare, 1993) and leave open the possibility that the relationship between PPTs and nonviolent crime, along with involvement in the criminal justice system, may be confounded by genetic factors.
Also of importance was that levels of self-control were not found to be significantly associated with either criminal behavior or involvement in the criminal justice system. These results stand in stark contrast to a large body of literature linking variation in self-control to involvement in criminal behavior (Pratt & Cullen, 2000). Importantly, these findings are consistent with studies suggesting that the effects of low self-control are exaggerated when genetic factors are not taken into account (Nedelec & Beaver, 2014). Taken together, these findings indicate that studies examining the connection between self-control and criminal behavior need to employ genetically sensitive designs in order to avoid genetic confounding.
The second key finding to emerge from our study was that only one of the five nonshared environmental factors was associated with criminal justice outcomes: delinquent peers. Specifically, the Delinquent Peers Scale was found to be positively associated with arrest, indicating that the twin who reported having more delinquent peers was more likely to report being arrested at some point. This finding is consistent with criminological research indicating that delinquent peers are a risk factor for criminal involvement (Haynie, 2001; Warr, 2002). Despite this finding, it should also be noted that neither delinquent peers nor any of the other nonshared environmental factors were significantly associated with the Criminal Justice System Involvement Scale.
More telling from these results may be that none of the nonshared environmental factors were found to be significantly associated with violent or nonviolent criminal behavior. In addition, only one of the nonshared environmental factors, delinquent peers, was found to be significantly related to any of the criminal justice outcomes. These results indicate that many of the environmental variables predicted to be associated with criminal behavior and involvement with the criminal justice system are not significant when using a conservative research design (MZ difference scores) that controls for genetic factors.
The findings of this study need to be interpreted in light of several limitations. First, the items used to measure PPTs were originally designed to measure broad personality traits and not PPTs. While using items that were not originally designed to measure PPTs may be viewed as a limitation, previous research has employed similar personality trait–based scales to measure psychopathy (Lynam & Miller, 2015; Miller & Lynam, 2015). Despite empirical studies finding support for the use of PPT Scales based off broad personality traits, it is still possible that a different pattern of results may have emerged if we employed a different measurement strategy for psychopathy (e.g., Hare’s Psychopathy Checklist–Revised; Hare, 1991). Second, the reliability (as measured via Cronbach’s α) for some of the scales included in the analyses was relatively low. These relatively low reliability estimates may have contributed to the null findings and thus future research needs to use different measures to determine whether the findings would be replicated. Third, the findings of our study may be limited by reliability issues in conjunction with the use of MZ difference scores. However, this is an issue that has been examined with MZ difference scores and there is evidence indicating that MZ difference scores can be reliable (Asbury et al., 2003; Ragosa, Brandt, & Zimowski, 1982; Ragosa & Willet, 1983). Moreover, previous studies have demonstrated that difference scores are a reliable analytic strategy even when there is significant measurement error in the variables being examined (Asbury et al., 2003; Ragosa et al., 1982). Fourth, the analysis of this study was restricted to 172–173 MZ twin pairs leaving open the possibility that the analysis was limited by the sample size. The small sample size used in the study may help to explain why our analysis did not detect significant findings between PPTs and many of the outcome variables. Perhaps if we had used a larger sample, a different pattern of results may have emerged that may conform more closely to the existing literature on PPTs and low self-control. Future studies will have to be conducted in order to determine if the nonsignificant findings in this study are due to having low statistical power. However, it should be noted that studies employing MZ difference scores have previously used this sample (Beaver, 2008; Crosnoe & Elder, 2002) and several studies have used even smaller samples (Deater-Deckard et al., 2001; Pike et al., 1996). Fifth, the MZ difference scores analytic strategy analyzes only MZ twins and leaves open the possibility that findings concerning MZ twin pairs may not be generalizable to singletons. Previous research examining the twin sample in the Add Health data, however, has revealed that there are no systematic differences between the twins and the singletons on behavioral measures (Barnes & Boutwell, 2013). Future research in this area should examine these relationships using different analytic strategies and different samples in order to determine if these results hold with larger sample sizes and different populations.
Footnotes
Appendix A
Zero-Order Correlation Matrix of Variables Included in the Analyses.
| Variables | X1 | X2 | X3 | X4 | X5 | X6 | X7 | X8 | X9 | X10 | X11 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Psychopathic personality traits | X1 | 1 | ||||||||||
| Low self-control | X2 | 0.252* | 1 | |||||||||
| Violent crime | X3 | 0.187* | 0.097* | 1 | ||||||||
| Nonviolent crime | X4 | 0.146* | 0.113* | 0.092* | 1 | |||||||
| Arrest | X5 | 0.096* | 0.165* | 0.062 | 0.174* | 1 | ||||||
| Criminal Justice involvement | X6 | 0.082 | 0.185* | 0.072 | 0.225* | 0.854* | 1 | |||||
| Maternal attachment | X7 | −0.091 | −0.092* | 0.009 | 0.044 | 0.062 | 0.026 | 1 | ||||
| Maternal involvement | X8 | −0.085 | −0.031 | −0.025 | −0.059 | −0.089 | −0.079 | 0.111* | 1 | |||
| Maternal rejection | X9 | 0.097* | 0.271* | −0.047 | 0.033 | 0.018 | 0.047 | −0.375* | −0.169* | 1 | ||
| Parental permissiveness | X10 | −0.158* | −0.040 | −0.103* | −0.022 | −0.078 | −0.071 | −0.006 | −0.047 | 0.014 | 1 | |
| Delinquent peers | X11 | 0.054 | 0.274* | 0.020 | 0.109* | 0.283* | 0.245* | −0.109* | −0.012 | 0.119* | 0.102* | 1 |
Note. N = 578.
*p < .05, two-tailed.
Appendix B
Descriptive Statistics for the Original Scores of All of the Variables Used in the Analyses.
| Variables | Mean | Standard Deviation | Range | N |
|---|---|---|---|---|
| Psychopathic personality traits | −.590 | 10.036 | −28.05 to 34.89 | 477 |
| Arrest | .250 | 0.433 | 0 to 1 | 477 |
| Conviction | .098 | 0.297 | 0 to 1 | 482 |
| Probation | .098 | 0.297 | 0 to 1 | 482 |
| Incarceration | .137 | 0.344 | 0 to 1 | 482 |
| Criminal Justice involvement | −.017 | 3.281 | −1.63 to 10.32 | 477 |
| Violent criminal behavior | −.271 | 3.299 | −1.48 to 28.65 | 480 |
| Nonviolent criminal behavior | −.549 | 1.998 | −0.91 to 25.88 | 480 |
| Maternal attachment | .158 | 1.623 | −12.13 to 0.90 | 520 |
| Maternal involvement | .006 | 4.432 | −8.51 to 12.84 | 521 |
| Maternal rejection | −.151 | 3.602 | −4.49 to 17.81 | 519 |
| Parental permissiveness | −.393 | 4.185 | −13.46 to 4.35 | 553 |
| Low self-control | −.245 | 3.340 | −5.60 to 13.06 | 560 |
| Delinquent peers | .008 | 2.506 | −2.34 to 6.03 | 554 |
Note. Descriptive statistics based off 578 monozygotic twins.
Acknowledgments
This research uses data from Add Health, a program project designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and is funded by grant PO1-HD31921 from the National Institute of Child Health and Human Development, with cooperative funding from 17 other agencies. Special acknowledgement is due to Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining data file from Add Health should contact Add Health, Carolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 27516-2524 (
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
