Abstract
A sizable majority of individuals report involvement in at least some delinquency during their lives. The small percentage of individuals who abstain from delinquency represents an interesting, yet underdeveloped area of research. The purpose of this article is to provide a general model for abstention from learning and integrated theories and propose a standard framework for operationalizing and measuring delinquency abstention. Data from Waves 1 to –9 of the National Youth Survey Family Study were used. Results are supportive of the integrated approach using suggested abstention measurement procedures. Predictors from learning theory had a more robust impact compared to social control variables.
Introduction
A sizable majority of individuals report involvement in at least some delinquent or criminal behavior during their lives (e.g., Brezina & Piquero, 2007; Elliott, Huizinga, & Menard, 1989; Piquero, Brezina, & Turner, 2005). This implies that the proportion of people who refrain from all criminal and delinquent behaviors throughout the life course is small, deviant from the statistical, if not the cultural, norm. Most research suggests that abstainers represent only 6–15% of the population (e.g., Boutwell & Beaver, 2008; Brezina & Piquero, 2007; Krueger et al., 1994; Leifman, Kühlhorn, Allebeck, Andréasson, & Romelsjö, 1995; Moffitt, Caspi, Dickson, Silva, & Stanton, 1996; Piquero et al., 2005), with a smaller number of studies reporting higher percentages, from 25% to nearly 40% (e.g., Dunford & Elliott, 1984; Kivivuori, 2000; Pulkkinen, Lyyra, & Kokko, 2009), depending at least in part on the offenses included and the definition of abstention used in the study.
Despite some increase in attention to life-course abstention from crime and delinquency in recent years, the literature remains limited with respect to both explicit theoretical explanations and tests of existing explanations of abstention. The purposes of this article are (a) to propose standard definition of delinquency abstention on conceptual and operational levels, (b) to suggest a set of hypotheses to explain abstention based on an integrated theoretical model, and (c) to test the model using prospective self-report data from a national sample.
Review of the Delinquency Abstention Literature
Most theories of criminal and delinquent behavior do not explicitly attempt to explain complete abstention throughout the life course. Piquero et al. (2005) note that the absence of risk factors for delinquency serves as an implicit explanation for abstention, but such an approach has not been tested to compare abstainers with all other offenders. In fact, the only theory that includes an explicit explanation for complete abstention from offending is Moffitt’s (1993) developmental taxonomy. This taxonomy focuses on two groups, adolescence-limited and life-course-persistent offenders, the first characterized by rates of offending that are low before and after adolescence but peak during the adolescent years, and the second characterized by rates of offending that are relatively high throughout the life course. The taxonomy also includes a third group of individuals who abstain from all delinquency, suggesting important differences between abstainers and the normative adolescence-limited offenders (Moffitt, 1993). Moffitt (1993) hypothesized that abstainers could be distinguished from the less serious adolescence-limited delinquents based on three general factors. First, the lack of a “maturity gap” due to late-onset of puberty during adolescence may prevent onset of antisocial behavior. Second, abstainers may lack structural opportunities for delinquent behavior (e.g., living in a rural area). Finally, Moffitt also suggests that certain personality characteristics such as being overly tense, having problems with interpersonal skills, and lacking the ability to display emotions may lead to being excluded from normative social groups.
Existing research examining differences between abstainers and offenders is nearly as limited as the theoretical literature. Piquero et al. (2005) cite between 15 and 20 studies with some form of abstention research. However, these studies vary considerably in conceptualization and operationalization of abstention, sample composition, time frame (e.g., longitudinal vs. cross sectional), overall methodological quality, and the range of behaviors included in the study. For instance, some studies used abstention measures based solely on substance use items (e.g., Hogan, Mankin, Conway, & Fox, 1970; Leifman et al., 1995; Shedler & Block, 1990; Vaillant, 1993) or criminal conviction data (e.g., Farrington & West, 1993).
Abstainers were compared with different types of offender groups using data from the Dunedin Multidisciplinary Health and Development Study in three studies as part of larger research Moffitt’s (1993) perspective. (Krueger et al., 1994; Moffitt, Caspi, Harrington, & Milne, 2002; Moffitt et al., 1996). Each used a combination of self-reported and informant-reported (parents and teachers) delinquency variables from multiple time points during childhood and adulthood. Self-reported delinquency items included a number of different behaviors ranging from minor infractions (e.g., traffic violations) to serious violent offenses (e.g., aggravated assault), with slight variations in specific items used in each article. Krueger et al. (1994) examined bivariate personality differences between respondents reporting no delinquency up to age 18 and “normative” delinquents. Abstainers tended to be more traditional (socially conservative, higher moral values), controlled (cautious, careful, and rational), less aggressive, and less interested in leadership.
Moffitt, Caspi, Dickson, Silva, and Stanton (1996) included a group of abstainers in their comparisons of offending pathways among males in the Dunedin study, using a less strict method for identifying abstention. Respondents were classified as abstainers if they had one or zero acts of antisocial behavior based on childhood informant reports (taken every 2 years from age 5 to 11) and later self-reports (at 15 and 18). They identified abstainers (compared with life-course-persistent and adolescence-limited groups) as timid, awkward, late to heterosexual relationships, exceptionally good students, and controlled. A follow-up study (Moffitt et al., 2002) presented a somewhat different profile of male abstainers at age 26. Essentially, abstainers “bloomed” into successful adults, showing little-to-no evidence of mental disorder and having better outcomes regarding educational attainment, relationship status, and job status compared to those in other offending pathways.
To date, only a few published research efforts focused attention directly on abstention from delinquency in American samples (Barnes, Beaver, & Piquero, 2011; Boutwell & Beaver, 2008; Brezina & Piquero, 2007; Chen & Adams, 2010; Piquero et al., 2005). Piquero et al. (2005) used a national sample of 17-year-olds in the United States to compare individuals who reported never engaging in a variety of delinquent behaviors ranging from felonies to trivial status offenses with other offenders. They found that abstainers had fewer delinquent peers, fewer dating experiences, stronger teacher attachment, fewer symptoms of depression, and less personal autonomy. Brezina and Piquero (2007) used a large citywide sample of American adolescents to examine the impact of moral beliefs and peers on abstention. Two-wave logistic regression models indicated that abstainers had stronger moral beliefs and fewer delinquent peers than offenders.
Three studies (Barnes et al., 2011; Boutwell & Beaver, 2008; Chen & Adams, 2010) used data from the National Longitudinal Study of Adolescent Health (Add Health). Barnes et al. (2011) found that abstainers spent less time with peers, had fewer friends who used drugs and tobacco, and generally were physically matured later than delinquents (using Waves 1–2). Chen and Adams (2010) also found that abstainers were generally less popular than delinquents but were not socially isolated from peer groups altogether (Using Waves 1–3). Rather, they were part of prosocial peer groups that avoided delinquent involvement. Finally, Boutwell and Beaver (2008) used a biosocial model to compare abstainers from “nonabstainers” by examining the effects of two dopamine receptor genes linked to antisocial behavior. Using data from the National Longitudinal Study of Adolescent Health (Add Health), they found that nonabstainers had lower self-control and more involvement with drug-using peers. Significant effects for the genetic factors were limited to male-only regression models for one of the two dopamine receptor genes.
Research Problems
The Concept of the “Delinquency Abstainer”
The working definition of delinquency abstention proposed here is that abstainers refrain from all forms of delinquent and criminal behavior after age 11, the youngest age after which juveniles in all states are subject to original juvenile court jurisdiction (Snyder & Sickmund, 2006). Delinquent and criminal behaviors refer here to acts that are illegal and would likely result in official intervention (including but not limited to apprehension) among most or all jurisdictions within the target population. This definition serves as the basis for measurement decisions in the present analysis.
This is a relatively strict definition of abstention, more so than definitions used in some prior research. Some prior studies have limited abstention to a specific age range, for example, adolescence and ignored offenses committed prior or subsequent to that age span (Boutwell & Beaver, 2008; Moffitt, 1993; Moffitt et al., 1996). Others have been limited in the range of offenses covered, for example, as noted earlier, focusing only on substance use. Still others have allowed individuals who had committed few, but more than zero, offenses to be classified as abstainers or nondelinquents (e.g., Dunford & Elliott, 1984; Moffitt et al., 1996, 2002). Such limitations on the definition of abstention beg the question of whether abstention from offending really occurs. A stricter definition reduces the number of cases that are classified as true abstainers, and may include cases that are not true abstainers (because even a relatively comprehensive list of offenses may miss unique offense patterns of some individuals), but it comes closer to real abstention, as most would commonly understand the term.
Lack of Abstention Theory
We examine abstention based on a model using concepts from social learning theory (Akers, 1985) and Elliott’s learning-centered integrated theory (Elliott, Ageton, & Canter, 1979; Elliott et al., 1989; Elliott, Huizinga, & Ageton, 1985), which is also consistent with elements of Agnew’s (2005) general theory of crime. Based on the results of prior tests of the combination of learning and control elements in these theories for a wide range of measures of delinquency and crime using these data (e.g., Elliott et al., 1985, 1989; Menard & Elliott, 1990, 1994; Roitberg & Menard, 1995), it is expected that exposure to friends who are involved in delinquency will have the strongest effect on abstention (decreasing the likelihood of abstention). Increased involvement with peers in general is also expected to decrease the likelihood of abstention, corresponding to the assumption that abstention is not a normative outcome (at least not statistically normative). Social control variables such as moral beliefs against criminal behavior and family involvement are expected to have weak or nonsignificant direct effects (higher social control predicts abstention) on abstention when the learning measures are included. Although the results should show that increased time with peers reduces abstention likelihood and higher social control scores increases the likelihood, abstention for all nine waves requires stability. The same outcome for all nine waves (i.e., zero offenses) would be less likely if other factors change over the life course.
Data
For this study, data from Waves 1 to 9 of the National Youth Survey Family Study (NYSFS: Formerly the National Youth Survey) were used (Dunford & Elliott, 1984; Elliott et al., 1979, 1985, 1989). The NYSFS is a prospective longitudinal study of youths between 11 and 17 years of age during the first wave of data collection (1976), representative of adolescents in the United States. Data were collected annually for the first five waves (1976–1980), with Waves 6–9 (1983–1992) measured at 3-year intervals. The original sample consisted of 1,725 participants, with 78% remaining at Wave 9. Menard, Mihalic, and Huizinga (2001) reported no significant demographic differences between those who dropped out of the study and those who remained by Wave 9, and others have noted that any nonrandom differential attrition was too small to impact substantive findings (Bosick, 2009; Brame & Paternoster, 2003; Brame, Bushway, & Paternoster, 1999; Elliott et al., 1989; Menard, Mihalic, & Huizinga, 2001).
Some respondents may have started and ended a short delinquency career before participation in the study, which would have inaccurately placed them in the abstainer category. Because this scenario is more likely to affect older adolescents, only those who were between 11 and 13 years old during Wave 1 (28–30 by Wave 9) were included in this study to minimize incorrect classification. In addition, analyses were based on respondents with complete offending data for Waves 1–9. This reduces the sample size from 778 (among those between 11 and 13 years of age at Wave 1) to 513. Analyses were also conducted allowing for different levels of missing data. Models allowing for one and two waves of missing offending data were compared with results based on no missing data. The different models produced generally similar results.
Measures
Dependent Variable: Delinquency Abstention
At each measurement period, respondents were asked to recall the number of times they engaged in a wide variety of delinquent behaviors during the previous year. Delinquency abstention is a dichotomous variable differentiating individuals reporting no involvement in any delinquent or criminal behavior throughout all nine waves (abstainers) from all others (offenders). The behaviors included for measurement are motor vehicle theft; theft of something worth more than $50; theft of something worth between $5 and $50; theft of something worth less than $5; strong-arm robbery; breaking and entering a building; purchasing stolen goods; selling marijuana; selling hard drugs; using marijuana, using hard drugs (cocaine, hallucinogens, heroin, inhalants, and barbiturates), hitting or threatening to hit others; public disorder; taking a vehicle for a ride without permission (joyriding); aggravated assault; gang fighting (except for Wave 9); destruction of property; and sexual assault. This list includes a variety of serious and relatively nonserious offenses are violations of criminal law and could reasonably result in arrest if noticed by authorities.
Items were not used if they met one of the following criteria. First, trivial behaviors (including status offenses) that may not be arrestable or even illegal at older ages were excluded. Some evidence suggests that trivial acts are distinct from more serious behaviors and should be treated as predictors of delinquency rather than as instances of delinquency (e.g., Elliott & Huizinga, 1989; Huizinga & Elliott, 1986; Junger-Tas & Marshall, 1999). Respondents reporting one or more delinquent or criminal behaviors in any of the nine waves were classified as offenders, leaving as abstainers those who reported zero delinquent acts for every item in every wave. Abstainers comprise 6.82% (n = 35) of the total sample, which is in line with previous research. 1
Independent Variables
A model of delinquency abstention encompassing 16 years of the life course necessarily differs from previous research using the integrated theory (e.g., Elliott et al., 1985, 1989; Elliott & Menard, 1996; Menard & Elliott, 1990, 1994; Menard & Huizinga, 1994; Roitberg & Menard, 1995), necessitating alternative measurement and analytical approaches. The dependent variable is binary and based on multiple items across nine waves of data. Because abstention status is permanently lost once an offense is reported and respondents become first-time offenders at every wave, the time frame linking potential causes of becoming an offender likely vary across the sample, and important factors influencing abstention may have varying levels of consequence across time.
Because this research is based on the idea that lifetime abstainers are substantively different from all offenders, the dependent variable is static, in addition to being restricted to two categories. This prevents the use of analytical techniques that incorporate change over time (e.g., growth curve modeling), and restricts options for how independent variables are entered into the model as well. All of the independent variables (other than demographic items) are best implemented as composite, continuous measures, and each corresponding scale is measurable at multiple waves (Waves 1–5 for school-related constructs; Waves 1–8 for all others). Despite the need for a single time-aggregated variable to represent each theoretical construct given the dependent variable, the scales should incorporate measures from all previous waves. To do this, several methods were explored to create a single measure of each predictor using information from multiple waves.
Once scales were constructed for a given measure at each wave (discussed below), various summary measures were explored including using the highest scale score across waves, lowest scores (scores created without accounting for possible problems with minimum possible values and floor effects, as well as transformed scores in an attempt to alleviate skewness), and the mean across each wave. In the present research, the mean scores are used. Means are calculated using every score, are less susceptible to the presence and consequences of outliers and wide ranges, and have less variability across time periods on a given measure within individuals. All scales from individual waves were calculated by taking the average of indicators multiplied by the number of items used in the scale. This allows for the inclusion of cases with very little missing data without altering the scale.
It is expected that individuals with few or no friends have limited opportunities to model nonnormative behaviors, including delinquency (which may be normative in adolescence). A measure of peer involvement is based on the amount of time respondents reported spending with friends. In Waves 1–8, they reported the number of evenings spent with friends on school nights (0–5); the number of afternoons they spent with friends after school (0–5); and relative amount of time spent with friends on weekends (very little, not too much, some, quite a bit, or a great deal [1–5]). The overall mean score was taken from scales in Waves 1–8. From the perspective of Hirschi’s (1969) social bonding theory, peer involvement should reduce illegal behavior; but from a social learning theory perspective (and based on prior research; see Roitberg & Menard, 1995), we expect peer involvement to lead to increased illegal behavior.
Also from social learning theory, a measure of exposure to delinquent peers is based on items asking respondents about the number of friends (none, very few, some, most, or all) that participate in the following behaviors: Damaged property, used marijuana, stole something worth less than $5, hit someone, broke into a vehicle, sold hard drugs, stole something worth more than $50, and suggested that the respondent do something against the law. The final exposure to delinquent peers measure was created by taking the mean for exposure over Waves 1–8.
It has also been suggested that individuals who never offend are more likely to participate in more prosocial behaviors such as schoolwork and spending time with family members (Piquero et al., 2005; Thornberry, 2005). A measure of school involvement was created based on Waves 1–5. School involvement is based on 3 items asking respondents how many evenings they study on school nights (0–5), how many afternoons they study after school (0–5), and how much time they study on weekends (very little, not too much, some, quite a bit, or a great deal [1–5]). We took the mean over Waves 1–5 to produce the final school involvement variable.
A measure of school isolation was included, which is based on 5 items (Waves 1–5) tapping into feelings of social isolation felt by respondents at school. Respondents were asked to indicate their level of agreement (strongly agree, agree, neither agree nor disagree, disagree, strongly disagree) with the following statements pertaining to them: Teachers don’t call on me in class, even when I raise my hand; I often feel like nobody at school cares about me; I don’t feel as if I really belong at school; even though there are a lot of kids around, I often feel lonely at school; and teachers don’t ask me to work on special classroom projects. The final school isolation measure was created by taking the mean of the scales over Waves 1–5.
A measure of family involvement consists of items asking respondents how many evenings they spend with their family during the school week (0–5), how many afternoons they spend with their family during school week (0–5), and the amount of time spent with their family on weekends (very little, not too much, some, quite a bit, or a great deal [1–5]). The mean of scales from Waves 1 to 8 was used for the final measure.
A measure of family isolation was included, which is based on 5 items (Waves 1–5) tapping into feelings of social isolation felt by respondents pertaining to familial relations. Respondents were asked to indicate their level of agreement (strongly agree, agree, neither agree nor disagree, disagree, strongly disagree) with the following statements pertaining to them: I feel like an outsider with my family; my family is willing to listen if I have a problem (reverse coded); sometimes I feel lonely when I am with my family; I feel close to my family (reverse coded); and my family doesn’t take much interest in my problems. This measure was created using the mean of scales from Waves 1 to 5.
The final independent variable is moral beliefs. It would be expected that individuals who are strongly opposed to criminal behaviors are more likely to be abstainers. This needs to be qualified by noting that past results on the impact of belief on illegal behavior suggest that the impact of belief may be indirect, operating via its influence on exposure to delinquent peers (e.g., Elliott et al., 1989; Menard & Elliott, 1994), so failure to find a direct influence of belief on abstention would not necessarily indicate that belief is unimportant as an influence on abstention. We created a “belief” measure based on items (Waves 1–8) asking respondents how wrong they think it is (not wrong at all, a little bit wrong, wrong, or very wrong) to engage various deviant behaviors. With the exception of “suggested that you do something against the law,” the belief items are the same as the items used for exposure to delinquent peers. For the belief measure, we took the mean over Waves 1–8.
Finally, gender (female = 1 and male = 0) and ethnicity were included as demographic variables. Ethnicity was dichotomized into a White/non-White (majority/minority) measure due to low frequencies for minority racial groups other than African Americans.
Analytical Approach
Bivariate and multivariate analyses were conducted to examine differences between abstainers and offenders. Abstainers were coded 1 and offenders were coded 0 so the models are predicting abstention over the life course (with the reference category being offending at least once over the life course). Independent samples t tests were used for bivariate comparisons between offenders and abstainers. Binary logistic regression was used to predict abstention in multivariate models. Relative strength of independent variables was assessed by calculating fully standardized logistic regression coefficients (Menard, 2002, 2004). 2
Results
Bivariate Comparisons
As previously mentioned, 6.82% (n = 35) of respondents reported no involvement in delinquent behavior from Waves 1 to 9. Abstainers were evenly distributed across the three age cohorts (7.60% at age 11, 6.90% at age 12, and 5.95% at age 13) and between non-White and White respondents (7.07% for non-White and 6.76% for White). Females (10.83%, n = 30), however, were significantly more likely to be abstainers than males (2.12%, n = 5).
Mean differences between abstainers and offenders are shown in Table 1 for continuous independent variables. With the exception of family and school isolation, mean differences were statistically significant (p < .05) and in the expected direction. Abstainers had higher levels of family involvement, school involvement, and moral beliefs compared to offenders. Offenders had significantly higher levels of exposure to delinquent peers and peer involvement. The correlation matrix shown (Table 2 ) displays bivariate correlations for all measures used. The relationships between independent variables are in the expected directions among significant associations.
Mean Comparisons of Abstainers and Offenders
Note. *p < .01.
**p < .001.
Correlation Matrix (N = 513)
Note. *p < .05.
Multivariate Analysis
The binary logistic regression model comparing abstainers and offenders is shown in Table 3 . Statistically significant direct effects on abstention in the expected theoretical direction were found for exposure to delinquent peers, peer involvement, family involvement, belief, and gender. Based on the fully standardized coefficients, exposure to delinquent peers was by far the strongest predictor (−.28), with all other coefficients below .10 (absolute values of coefficients). The model does well in predicting abstention based on McFadden’s R2, which indicates that the model explains nearly half of the variation in the log odds of abstention.
Binary Logistic Regression Predicting Delinquency Abstention (N = 513)
Note. *p < .05.
**p < .01.
***p < .001.
The significant coefficient for belief was negative, which contrasts with expectations. Bivariate differences show abstainers with significantly higher moral belief means, indicating a change in direction caused by characteristics of the regression model. Further investigation suggests that the result has little meaning to the overall model. First, the coefficient just reached statistical significance (p = .049) and had the weakest effect among significant predictors. Second, separate models were estimated (8) each with one of the eight independent variables removed to assess changes in the belief coefficient. The two least influential measures (family isolation and ethnicity) resulted in similar weak negative coefficients, while removal of the other variables resulted in nonsignificant effects for belief. Removal of exposure to delinquent peers alone changed the sign for a nonsignificant belief coefficient. Only after removing exposure to delinquent peers and peer involvement or gender did a significant, positive belief coefficient appear. Note, too, that there is a tendency for both delinquency and strength of beliefs that it is wrong to violate the law to decline over the life course, potentially resulting in an artifactual rather than explanatory positive correlation between the two in longer-term longitudinal research (e.g., Menard, 1992).
Discussion and Conclusion
Overall, the binary logistic regression model was consistent with expectations based on the integrated theory (Elliott et al., 1985, 1989), particularly the learning component of that perspective. Regarding the abstention hypothesis, there is evidence that being excluded from many peer-normative activities can isolate individuals and reduce opportunities for involvement in any delinquent behavior. Support for the learning component of the theory is seen in the strong relationship between peer measures and abstention, as well as the nonsignificance of direct effects of social control variables. Indeed, when exposure to delinquent peers and peer involvement are not included, belief represents the strongest overall predictor. Otherwise, learning variables are the dominant factors. This is consistent with past research (e.g., Elliott et al., 1989; Menard & Elliott, 1994) and suggests that the influence of moral beliefs may operate indirectly through exposure to delinquent peers. 3
Overall, the results are consistent with the prediction that peer influences trump the effects of social control factors (at least as direct effects) in explaining long-term abstention from delinquency and crime. Exclusion from social groups (i.e., low-peer involvement), distinguishes abstainers from offenders, as suggested in previous research by Moffitt (1993) and Chen and Adams (2010), but as suggested by Chen and Adams, the groups from which the abstainers are being excluded are not necessarily as normative (in a behavioral rather than a statistical sense) as Moffitt (1993) seems to suggest. Correspondingly, exposure to delinquent peers is consistently the strongest predictor abstention, which parallels results from prior research. Also parallel to previous studies (see particularly Elliott et al., 1989), the effect of belief that it is wrong to violate the law appears to be primarily indirect, operating via exposure to delinquent friends. Family involvement is the only significant predictor in the expected direction among social bonding measures as family isolation, school involvement, and school isolation failed to have any direct influence.
These findings raise important questions. First, to what extent can these results be generalized across different types of offenses? For example, would the same model of abstention apply equally well if we limited the scope of offenses considered to (a) violent offenses, (b) property offenses, or (c) illicit substance use or sales? Second, how do these results compare to different samples and time periods (or equivalently to different cohorts)? Finally, if the key to abstention is stable, favorable levels of exposure to delinquent friends and social bonding (here, in particular, belief that it is wrong to violate the law, whose effect appears to be indirect, but which does change in tandem with exposure to delinquent friends and delinquent behavior), what in turn would account for the high degree of stability of those variables for abstainers (as opposed to the sharp changes that occur for nonabstainers)?
In particular, are there influences on stability in key variables that are amenable to policy intervention? Although there were early suggestions that abstention might be indicative of emotional or psychological problems (Moffitt, 1993; Moffitt et al., 1996), subsequent research (Chen and Adams, 2010; Moffitt et al., 2002) suggest that any small short-term disadvantages to abstention would be substantially outweighed by its long-term advantages. Taking the attitude that some participation in illegal behavior is in some sense beneficial (having some emotional and psychological advantage) would be, based on these findings, misguided, and identifying ways to encourage complete abstention from illegal behavior would be well advised. The present results suggest that while abstainers do indeed appear to be a unique group, the explanation for their abstention is unique not in what influences matter (the present results suggest that they are the same as the influences that distinguish between low- and high-frequency offenders), but in how stable those things that matter are over the life course for abstainers. If such stability is possible, then finding ways to promote that stability may offer a key to more effective interventions to encourage complete abstention (or as close to complete abstention as possible) from illegal behavior.
There are number of ways to approach future abstention research, given the relative lack of past theoretical and empirical attention. For instance, it would be desirable to have data that could more thoroughly test Moffitt’s perspective on abstention by including neuropsychological deficits and differences in the maturity gap (information that is not available in the data used for the present study). 4 In addition, it would be desirable if the data had the characteristics of the present NYSFS study to address the limitations of prior abstention research by using a nationally representative sample over a sufficiently long span of the life course, with a sufficiently broad range of offenses used to define abstention, to avoid falsely identifying as abstainers many individuals who are at some point involved in illegal behavior. Prospective research should also expand to assess differences between abstainers and different types of offenders, providing a more accurate depiction of abstainer–offender differences (e.g., differences between abstainers and normative delinquents as opposed to differences between abstainers and chronic offenders). This presents additional avenues of study to create offending measures that account for development and change in offending seriousness and frequency over time along with the inclusion of an unchanging abstention group.
Footnotes
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Work on this project was supported in part by grants from the Antisocial and Violent Behavior Branch, National Institute of Mental Health (MH27552), the National Institute for Juvenile Justice and Delinquency Prevention, U.S. Department of Justice (78-JN-AX-003), the National Institute on Alcohol Abuse and Alcoholism and the Office of Behavioral and Social Science Research (AA11949), and the National Institute on Drug Abuse (DA015983).