Fill and Sign the Exide Life Insurance Policy Bond Download Form

DOES THE EFFECT OF IMPULSIVITY ON DELINQUENCY VARY BY LEVEL OF NEIGHBORHOOD DISADVANTAGE? ALEXANDER T. VAZSONYI Auburn University H. HARRINGTON CLEVELAND Texas Tech University RICHARD P. WIEBE Fitchburg State College The authors examine the importance of the person-context nexus in adolescent deviance. Using the nationally representative National Longitudinal Study of Adolescent Health (Add Health) data set of more than 20,000 male and female adolescents, the authors are interested in testing whether the relationship between impulsivity and a variety of deviance measures varies as a function of neighborhood disadvantage. Results suggest that whereas levels of impulsivity and deviance vary by level of neighborhood disadvantage, relationships between impulsivity and AUTHORS’ NOTE: This research was based on data from the National Longitudinal Study of Adolescent Health (Add Health), a program project designed by J. Richard Udry (principal investigator) and Peter Bearman, and funded by Grant P01-HD31921 from the National Institute of Child Health and Human Development to the Carolina Population Center, University of North Carolina at Chapel Hill, with cooperative funding participation by 17 agencies. Special acknowledgement is due to Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining data files from Add Health should contact Add Health, Carolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 275162524 (www.cpc.unc.edu/addhealth). We are indebted to study participants and project staff, to the three thoughtful reviewers, and to the editor. Correspondence concerning this article should be addressed to Alexander T. Vazsonyi, Department of Human Development and Family Studies, Auburn University, 284 Spidle Hall, Auburn, Alabama 36830; e-mail: vazsonyi@auburn.edu. CRIMINAL JUSTICE AND BEHAVIOR, Vol. 33 No. 4, August 2006 511-541 DOI: 10.1177/0093854806287318 © 2006 American Association for Correctional and Forensic Psychology 511 512 CRIMINAL JUSTICE AND BEHAVIOR deviance do not. This same finding is made for both male and female study participants, though there is some modest evidence of moderation in female youth. Together, these results have important implications for social disorganization theory, the general theory of crime, and for personality research on the etiology of crime and deviance. Keywords: T deviance; general theory of crime; social disorganization; self-control; race o bring a more sophisticated perspective to the debate regarding the relative importance of the person or situation, recent researchers have integrated intrapersonal and contextual variables (Bronfenbrenner, 1979; Magnusson & Stattin, 1998) into the study of crime and deviance (e.g., Sampson, 2001). Employing an integrative perspective, researchers can examine whether personal characteristics retain their predictive power when situational factors are considered. Perhaps more important, they can also examine the relationship between personal characteristics and offending across different contexts. Our study brings an integrative perspective to an examination of the relationship between impulsivity and delinquency across levels of neighborhood disadvantage. Contextual factors, such as neighborhood disadvantage, and personality factors, such as impulsivity, have emerged as predictors of delinquency from discrete theoretical traditions. Social disorganization theory (Bursik, 1988; Shaw & McKay, 1942) suggests that neighborhood characteristics, such as level of poverty, percentage of single parent families, or residential mobility, contribute to greater levels of social disorganization and thus, directly or indirectly, to greater levels of crime and deviance (Bursik & Grasmick, 1993; Sampson, 1991; Sampson & Groves, 1989). The theory postulates that where the developmental context is inherently disorganized, unable to provide a community structure conducive to conformity (Bursik & Grasmick, 1993), impulsive individuals are more likely to engage in conduct that violates community norms, including delinquency. In more organized neighborhoods with high levels of “collective efficacy” (Sampson, Raudenbush, & Earls, 1997), informal social controls ensure that impulsivity and similar traits are suppressed rather than expressed. Thus, neighborhood-level differences in social and organizational characteristics, rather than individual traits such as impulsivity, explain neighborhood variations in crime Vazsonyi et al. / EFFECT OF IMPULSIVITY ON DELINQUENCY 513 rates, implying that in neighborhoods with high levels of collective efficacy, impulsivity should not significantly correlate with offending (see Kubrin & Weitzer, 2003). Because it suggests that impulsivity would be expressed more freely when social controls are absent, social disorganization theory appears to predict a differential relationship between impulsivity and delinquency across contexts. The alternative view, of course, is that impulsivity will predict delinquency similarly across neighborhood contexts. This prediction stems from the personality-based approach to explaining crime. This approach, which has used different terminology to generally describe the inability to delay gratification, to modulate affect (negative emotionality), or to have poor ego control (Block & Block, 1980), holds that traits such as (a) impulsivity (e.g., Eysenck, 1977; Wilson & Herrnstein, 1985), (b) low self-control or impulse control (Pulkkinen, 1982, 1986), and (c) weak constraints (Caspi et al., 1994) or sensation seeking (Zuckermann, 1994; Zuckermann, Bone, Neary, Mangelsdorff, & Brustman, 1972) predispose individuals for crime and delinquency. The personality-based approach generally does not emphasize the influence of contextual constraints or social stressors on offending, implying that the relationship between offending and relevant traits should remain invariant across different environmental contexts. Studies in this tradition have found that impulsivity (both behavioral and cognitive) and low self-control are consistent predictors of delinquency, especially serious delinquency (e.g., Moffitt, Caspi, Harrington, & Milne, 2002; White et al., 1994). Longitudinal investigations from Scandinavian countries (Pulkkinen, Virtanen, Klinteberg, & Magnusson, 2000), England (Farrington, 1990, 1995), New Zealand (Moffitt et al., 2002), and the United States (White et al., 1994) that have followed cohorts from early childhood into adolescence and adulthood have provided substantial evidence that childhood impulsivity is an important predictor of later delinquency as well as other antisocial and deviant behaviors. The prediction that context will not affect the relationship between delinquency and impulsivity is also consistent with a prominent theoretical tradition in criminology, the general theory of crime, proposed by Gottfredson and Hirschi (1990). This theory postulates that low self-control is “the property of individuals that explains variation in the likelihood of engaging in such [criminal and crime analogous] 514 CRIMINAL JUSTICE AND BEHAVIOR acts” (Hirschi & Gottfredson, 1994, p. 2). The self-control trait described by Gottfredson and Hirschi (1990) is multidimensional and hence broader than the impulsivity construct favored by personality researchers. In fact, Gottfredson and Hirschi argue that the evidence from personality-crime research is unimpressive, in part because of the conceptual and methodological overlap between dependent and independent variables. In addition, the general theory is also broader (and more cognitive) than personality-based crime theories because it emphasizes individual choice and decision making. However, the general theory does not appear to contradict traditional personality research regarding the identity of the basic intrapersonal correlates of crime and delinquency. Gottfredson and Hirschi (1990) conclude the following: The search for personality characteristics common to offenders has thus produced nothing contrary to the use of low self-control as the primary individual characteristic causing criminal behavior. People who develop strong self-control are unlikely to commit criminal acts throughout their lives, regardless of their other personality characteristics. In this sense, self-control is the only enduring personality characteristic predictive of criminal (and related) behavior. People who do not develop strong self-control are more likely to commit criminal acts, whatever the other dimensions of their personality. (p. 111) Perhaps the main difference between the general theory and traditional personality–crime research lies not in the identity of relevant traits—impulsivity is an important component of low self-control—but in their etiology. Gottfredson and Hirschi (1990) suggest that individuals “differ in the extent to which they are restrained from criminal acts” (p. 88) as the result of ineffective “training, tutelage, or socialization” (p. 95) rather than as the result of genetic bases (Rowe, 1994), although they also acknowledge individual differences pending socialization pressures. However, this difference is unimportant to the current question. Regardless of the difference between etiological explanations for the development of relevant constructs, neither personality-based theories nor the general theory suggest that relevant traits predict crime differently across neighborhood contexts. Vazsonyi et al. / EFFECT OF IMPULSIVITY ON DELINQUENCY 515 Gottfredson and Hirschi (1990) propose that the calculation or choice to engage in norm violating conduct is not determined by the probabilistic construct of low self-control. Rather, individual dimensions of low self-control affect the calculation of potential consequences of one’s behavior. In turn, this “hedonic calculus” is not context dependent (Vazsonyi, 2003). Individuals that score low on measures of low self-control reside in both affluent neighborhoods and in poor neighborhoods, they work as low-wage manual laborers and as white-collar executives, and of import for the current study, they are members of both minority and majority populations (Gottfredson & Hirschi, 1990). In addition to the observation that individuals differ in self-control within any single context, Gottfredson and Hirschi (1990) expressly argue that the connection between low self-control and offending should remain invariant across contexts or developmental ecologies. (They also argue for no differences by sex, ethnicity, or nationality.) Clearly, contexts do vary in their levels of opportunity for crime, and for any given level of opportunity, the lower an individual’s selfcontrol, the higher the risk of offending. Thus, self-control theory and the personality-based approach to crime both hypothesize that the relationship between impulsivity and delinquency will remain constant across diverse developmental contexts. This hypothesis may be contrasted with the hypothesis from social disorganization theory that only in disorganized neighborhoods will impulsivity correlate significantly with delinquency. In the following section, we review prior empirical work that has specifically focused on the question of whether there exists variability in the person-context interaction important for the explanation of crime and deviance. PREVIOUS EMPIRICAL INVESTIGATIONS A review of the literature reveals that few previous studies have directly tested the person-context nexus in crime and delinquency with a particular focus on impulsivity or low self-control across different neighborhoods. This is also true in general of efforts that have attempted to study the impact of contextual effects on individual behavior. Kubrin and Weitzer (2003) provide a discussion of the small number of studies that have done so. In the current review, 516 CRIMINAL JUSTICE AND BEHAVIOR we focused on studies that had direct relevance for the current investigation. Based on both cross-sectional and longitudinal data from the Pittsburgh Youth Study, Lynam et al. (2000) found that although impulsivity predicted offending for boys within poor neighborhoods, it did not predict offending among boys within better neighborhoods. Furthermore, they concluded that official census–based data had weaker effects on the impulsivity-offending relationship, whereas selfreported socioeconomic status (SES) indicators had stronger (possibly inflated) ones. The authors interpreted the difference in the relationship between impulsivity and delinquency across neighborhood contexts as consistent with the perspective that lower levels of informal social controls and higher levels of social disorganization allowed impulsivity to be expressed in offending behaviors, counter to the expectations of both the personality-based approach and self-control theory. In a similar effort, framed as a risk-protective factor study, which was also based on the Pittsburgh Youth Survey, Wikström and Loeber (2000) examined (among other things) whether neighborhood context moderated the importance of hyperactivity-impulsivity-attention problems in delinquency. The measure of risk and protective factors included five additional indicators: one individual characteristic (lack of guilt) and four social situation measures (poor supervision, low school motivation, peer delinquency, and positive attitudes toward antisocial behavior). Findings were mixed. Among high-risk boys, neighborhood context was unimportant for offending. On the other hand, for boys classified as balanced between risk and protective factors, the authors found evidence of a consistent contextual effect for youth from poor neighborhoods. Finally, youth who had high protective factor scores seemed to be most influenced by neighborhood context. It is interesting that the authors found that neighborhood context seemed unimportant in rates of serious offending; in addition, context was unimportant for all early onset youth, although it showed a moderate effect for late onset youth for the high protective factor and the balanced groups. Taken together, the two studies based on the Pittsburgh data suggest that variability in neighborhood context conditions the impulsivity–delinquency relationship and that disadvantaged neighborhoods potentiate offending among youth not otherwise at risk. Vazsonyi et al. / EFFECT OF IMPULSIVITY ON DELINQUENCY 517 THE CURRENT INVESTIGATION In the present study, we tested the two competing hypotheses discussed above: an enhanced effect for impulsivity in poor neighborhoods (the “potentiation” or social disorganization hypothesis) and no interaction between impulsivity and neighborhood (the personalitybased and self-control theory derived “invariance” hypothesis). We attempted to test the same questions posed by Lynam et al. (2000) using data from the Add Health Project.1 An advantage of using data from this nationally representative project is that it uses more extensive measures of neighborhood factors and contains a wider sample of neighborhood contexts than the data used by the Lynam study. Rather than being confined to a single city, the Add Health is a national study that systematically sampled schools that were representative for region, urbanicity, school type, ethnicity, and school size (Udry, 1998). Of course, using these data is not without disadvantages. Specifically, the measure of impulsivity available within the Add Health is less extensive than that used in the Lynam study and thus may limit comparability of findings. Thus, our study should not be construed as an attempt to replicate Lynam et al.’s work. Like Lynam et al. (2000), we examined whether the impulsivitydelinquency relationship varied across levels of neighborhood disadvantage. In addition, we also took advantage of the large sample size of the Add Health data set to examine mean levels of deviance and impulsivity across 10 groups of participants classified by neighborhood disadvantage/quality. Also like Lynam et al., we included a variety of deviance measures: general delinquency, nonviolent delinquency, and aggression. Unlike Lynam et al., we extended our analytic framework based on the complex sampling used in the Add Health data to examine potential multilevel effects (school clusters). In addition, we also examined the same set of questions on a sample of female adolescents. RESEARCH QUESTIONS AND ANALYTIC STRATEGIES Our main research question was whether the impulsivitydelinquency relationship varied as a function of neighborhood context. We derived decile groupings of neighborhoods based on the distribution of neighborhood disadvantage scores of participants. 518 CRIMINAL JUSTICE AND BEHAVIOR This allowed us to compare mean levels of deviance and impulsivity across these 10 neighborhood levels. Finally, we tested the same predictive models in male and female adolescents. To address our research questions, we employed the following analytic strategy. First, we computed descriptive statistics for the main study variables. Second, we compared mean levels of impulsivity and deviance measures across the neighborhood deciles by sex, followed by correlational analyses of the main variables in the study. Third, the primary hypothesis was examined with multilevel models run using SAS PROC MIXED. Multilevel modeling provides a regression approach to analyses that differs from ordinary least squares (OLS) methods in that the calculation of standard error takes the clustered nature of the data into account. In the case of this study, it handles data of individuals that are nested within school clusters. Because two levels of data were included, these models apportion the variance not associated with specified fixed effects into school cluster (intercept) and individual (residual) levels. In addition, a random effect for impulsivity was entered to examine whether the effect of this variable varied significantly across individuals. METHOD SAMPLE The data for the current study were based on the In-Home data collection of the National Longitudinal Study of Adolescent Health (Add Health). The Add Health project consists of several data sets, including the In-School data set, the Wave 1 and Wave 2 In-Home data sets, a school administrator data set, a parent data set, and a contextual data set. The sampling plan for the project as a whole was organized around 80 high schools selected by a probability sampling mechanism to be representative of schools in the United States with respect to region, urbanicity, school type, ethnicity, and school size. Because not all of these high schools included seventh, eighth, and ninth graders, 52 feeder schools were added. All students in 7th through 12th grades—more than 120,000 students were enrolled in these schools—were eligible for In-School data collection (Udry, Vazsonyi et al. / EFFECT OF IMPULSIVITY ON DELINQUENCY 519 1998). Collected during the 1994-1995 school year, the In-School data set, with a sample size of more than 90,000, included questions about adolescents’ demographics, parental characteristics, household structure, and risk behaviors. Starting the summer after the In-School data collection, detailed In-Home interviews that focused on healthrelated behaviors were given to 20,745 adolescents. These data constitute the Wave 1 In-Home data set and are the source of most of the items used in this study’s analysis. A year later, 14,738 of the Wave 1 respondents—Wave 1 high school seniors were excluded—were interviewed for the Wave 2 In-Home data collection. It is important to emphasize that the selection of the students for the Wave 1 In-Home data collection was not limited to the 90,000 students who participated in the In-School data collection. Instead, the sampling frame for the In-Home data collections included all students who were eligible for the In-School data collection. The full Wave 1 In-Home (N = 20,745) sample itself comprises a core sample (n = 12,105), which is nearly self-weighting; a “saturated” sample from 16 schools; ethnic oversamples of Cubans, Chinese, and Puerto Ricans, as well as highly educated Blacks; and a genetic pairs sample. This data set includes sampling weights to insure the generalizability of the full In-Home sample. As part of the Wave 1 InHome data collection, 17,700 parent interviews were conducted. These interviews collected data on parents themselves, household characteristics, and parents’ perceptions of their children. The analyses reported in this article were completed separately for males and females. Of the 20,745 eligible Wave 1 In-Home respondents, 20,367 had valid values for biological sex. Of the 10,097 (49.5%) males, 83 had missing data for one or more outcome measures. For those with complete outcome data, 28 others (.3%) were missing data for our impulsivity measure, and 156 were missing census tract information used to construct the neighborhood disadvantage measure. Patterns were similar among the 10,270 females (50.5%): 69 had missing data for one or more outcome measures, 36 others were missing impulsivity data, and 153 (1.5%) were missing census information. The final data set on which most analyses were completed consisted of interviews with 9,830 male and 10,012 female adolescents. The average age during the summer of 1995, at the time data 520 CRIMINAL JUSTICE AND BEHAVIOR collection began, was 16.4 (SD = 1.7) for males and 16.2 (SD = 1.7) for females. This data set was ethnically diverse. Approximately half the sample was White (51% for both sexes), followed by Black (22% for males and 24% for females), Hispanic (17% and 16%), Asian (7% and 6%), and American-Indian/Other (3% for both). The mean household income of sample participants was $46,070. Its median was $38,000. MEASURES Background variables. Similar to Lynam et al. (2000), we classified family structure as traditional or nontraditional (intact or nonintact). Approximately half of the sample (51%) resided in traditional families, defined as families with two biological parents. To control for ethnicity and race, we created dummy-coded variables for Black, Hispanic, Asian, and American-Indian/Other that used Whites as the comparison group. Neighborhood disadvantage. Three U.S. Census Bureau indicators of neighborhood disadvantage were drawn from the contextual data set of the Add Health and used to form a neighborhood disadvantage composite: (a) the proportion of single-parent households with children, (b) the proportion of households with less than $15,000 in annual income, and (c) the unemployment rate in block groups of respondents.2 In the average U.S. census block group of the respondents, the proportion of single-parent households with children was .24 (SD = .17), the proportion of households with less than $15,000 in annual income was .19 (SD = .16), and the unemployment rate was 7.6 (SD = 5.6). A factor analysis revealed that the three items loaded strongly on a single factor for both males and females. Factor loadings were .86, .85, and .89 for Indicators 1, 2, and 3, respectively. These factor loadings were used to create a neighborhood disadvantage factor score, with high scores indicating high neighborhood disadvantage. This approach is similar to that used by Sampson, Morenoff, and Earls (1999) to measure what they referred to as “concentrated disadvantage” (see also Sampson et al., 1997). General delinquency. General delinquency was assessed by averaging 15 self-report items from the Add Health Delinquency Scale Vazsonyi et al. / EFFECT OF IMPULSIVITY ON DELINQUENCY 521 that asked about a broad range of antisocial behaviors within the past 12 months, from minor acts, such as shoplifting and lying to parents about whereabouts, to more serious offenses, such as being in a serious fight and selling drugs. Responses ranged from 0 to 3, representing never, one or two times, three or four times, and f ive or more times. Participants entered their responses to these items while listening to an audio cassette recording of the questions. High scores on this measure indicated high levels of delinquent behaviors. Cronbach’s alpha for this scale was .85 for males and .81 for females. The appendix includes all delinquency items. Nonviolent delinquency. The 10 items from the general delinquency scale that assessed nonviolent delinquent behaviors were used to construct a nonviolent delinquency scale. Behaviors covered included painting graffiti or signs on someone else’s property or in a public place, running away from home, stealing something less than $50, stealing something more than $50, and deliberately damaging property. Cronbach’s alpha for this 10-item scale was .82 for males and .78 for females. Aggression. Aggression was assessed by an average score of 7 items drawn from fighting and violence and joint occurrences sections of the Wave 1 In-Home interview as well as from the delinquency scale. The three items from the delinquency section were as follows: (a) “How often did you get into a serious physical fight?” (b) “How often did you hurt someone badly enough to need bandages or care from a doctor or nurse?” and (c) “How often did you take part in a fight where a group of your friends was against another group?” Items from the joint occurrences section were as follows: (a) “Have you ever carried a weapon at school?” and (b) “Have you ever used a weapon in a fight?” Items from the fighting and violence section were as follows: (a) “You got into a physical fight” and (b) “You pulled a knife or gun on someone.” All of these items assessed the frequency of these behaviors during the past 12 months. The fighting and violence and joint occurrences items were assessed in the same manner as the delinquency scale, using audio cassettes and laptop computers. Unlike the responses for the items drawn from the delinquency section, which ranged from 0 to 3, the items drawn from the joint 522 CRIMINAL JUSTICE AND BEHAVIOR occurrences sections provided only a 0 to 1 response pattern, indicating whether the event did or did not happen. The two items from the fighting and violence section provided a response pattern that ranged from 0 to 2, representing never, one time, and two or more times. Despite the response patterns differences, most responses across the five items with either 0 to 2 or 0 to 3 response patterns were either 0 or 1. However, to avoid the possibility that the resulting scale would be affected disproportionately by items with greater response ranges (e.g., the delinquency items), prior to creating the scale, each response for the five “larger range” items that were greater than 1 were recoded as a 1 to be consistent with the 0 to 1 format used by the joint occurrence items. The highest correlation for scale items (.60) was between “How often did you get into a serious physical fight?” and “You got into a physical fight.” The lowest correlation (.21) between scale items was between “How often did you get into a serious physical fight?” and “Have you ever carried a weapon at school?” The Cronbach’s alpha of this scale was .78 for males and .74 for females and was not reduced by deleting any item. High scores indicated higher levels of aggressive behavior. Impulsivity. Impulsivity was measured with the mean of four items from the personality and family section of the Add Health In-Home Interview.3 Items asked respondents to indicate on a 5-point scale from strongly agree to strongly disagree whether they agreed with the following four statements: (a) “When you have a problem to solve, one of the first things you do is get as many facts about the problem as possible”; (b) “When you are attempting to find a solution to a problem, you usually try to think of as many different ways to approach the problem as possible”; (c) “When making decisions, you generally use a systematic method for judging and comparing alternatives”; and (d) “After carrying out a solution to a problem, you usually try to analyze what went right and what went wrong.” The Cronbach’s alpha of this 4-item scale was .73 for males and .75 for females. RESULTS In a first step, descriptive statistics were computed for each of the individual variables of interest: the three outcome measures and Vazsonyi et al. / EFFECT OF IMPULSIVITY ON DELINQUENCY 523 TABLE 1: Descriptive Statistics of Deviance and Impulsivity Measures Males n M SD Females Original Adjusted Skew Skew n M SD Original Adjusted Skew Skew General 10,022 0.34 .40 delinquency 2.14 .43 10,203 0.24 .29 2.41 .47 Aggression 10,022 0.24 .26 0.99 .15 10,204 0.12 .20 1.82 .93 Nonviolent 10,019 0.35 .44 delinquency 2.08 .47 10,203 0.27 .34 2.24 .43 Impulsivity 0.47 – 10,193 2.19 .62 0.40 10,041 2.18 .63 impulsivity. Means and standard deviations for each of these four variables, as well as the skew for both unadjusted and squareroot corrected versions of the three outcome variables, are shown in Table 1. As seen in this table, square-root corrections reduced skew substantially for the three outcome measures. The unadjusted mean response for each of the outcomes was around .3, indicating that the average study participant had refrained from most of the delinquent acts during the past 12 months. ANALYSES OF VARIANCE Neighborhood effects. Next, analysis of variance (ANOVA) was used to search for the main effect of neighborhood disadvantage on delinquency. Although the primary analyses rely on a regression approach that conceptualizes neighborhood disadvantage as a continuous variable, ANOVA was used across 10 neighborhood groupings to provide the greatest opportunity to consider differences in mean levels of the outcomes and impulsivity measures across levels of neighborhood disadvantage. To balance this openness to differences across levels of disadvantage, Tukey comparisons, which reduce the chance of Type 1 error, were used to examine differences between these decile groupings. These decile groupings were formed by classifying each respondent into 1 of 10 neighborhood disadvantage groups based on the composite neighborhood disadvantage measure. Although it is unlikely these decile groups conform specifically to different neighborhood types, their use here allows for detailed 524 CRIMINAL JUSTICE AND BEHAVIOR consideration of the distribution of both the outcome measures and impulsivity along the spectrum of neighborhood disadvantage and takes advantage of the considerable sample size of the Add Health. Table 2 provides the mean levels of these three outcome measures and impulsivity by these 10 neighborhood groupings. Findings were mixed. The distributions of the measures across these deciles of neighborhood disadvantage were similar by sex. The evidence suggested that although general delinquency did not significantly vary by neighborhood disadvantage for males, F(1, 9828) = 2.43, p > .05, or females, F(1, 10012) = 0.01, p > .05, aggression and nonviolent delinquency were associated with context. It is interesting that these effects were in different directions, depending on the criterion variable. For aggression, levels were greater in more disadvantaged neighborhoods for both males, F(1, 9828) = 60.24, p < .05, and females, F(1, 10012) = 200.92, p < .05. For males, ANOVA with Tukey comparisons revealed that the 10th decile of neighborhood disadvantage (the most disadvantaged) was significantly more aggressive (.28) than all but the 8th and 9th deciles, F(9, 9820) = 8.51, p < .05. The 9th decile (.27) was significantly more aggressive than all deciles with less neighborhood disadvantage, save the 3rd (.24). The fewest aggressive behaviors were reported in the least disadvantaged neighborhood decile (.20). Male aggression in this decile was significantly less than that reported in deciles 3 and 8, as well as from 9 and 10. Although a number of contrasts reached statistical significance, it is important to note that using a standardized effect size metric, namely Cohen’s d, the differences were modest in magnitude (largest d = .31)—the 1st decile compared to the 10th decile. In fact, on average, effects sizes were d = .1 or smaller. ANOVA results for aggression among females were more dramatic, F(9, 9820) = 23.28, p < .05. Aggression in the 10th decile of neighborhood disadvantage (.18) was significantly greater than aggression in all other deciles. Aggression in the 9th decile (.15) was significantly greater than all deciles from the 7th and lower, and aggression in the 8th decile (.14) was greater than all deciles 6th and lower. The least aggression (.08) was again reported in the neighborhood decile with the lowest disadvantage. In addition to being significantly less than deciles 8 through 10, aggression in this decile was significantly different from aggression in deciles 4 and 7. 525 0.39 Nonviolent Impulsivity 0.23 0.08 0.29 2.22 General delinquency Aggression Nonviolent delinquency Impulsivity Females 2.28 0.20 Aggression delinquency 0.34 General delinquency Males M .60 .34 .16 .27 .61 .46 .24 .39 SD 1 (0 to 10) 2.25 0.39 0.10 0.26 2.25 0.41 0.22 0.37 M .64 .38 .19 .32 .65 .48 .26 .41 SD 2 (10 to 20) 2.22 0.30 0.11 0.25 2.18 0.36 0.24 0.34 M .59 .39 .19 .34 .63 .45 .26 .40 SD 3 (20 to 30) 2.22 0.28 0.11 0.24 2.17 0.36 0.23 0.33 M .60 .34 .19 .28 .62 .43 .27 .39 SD 4 (30 to 40) 2.19 0.28 0.10 0.23 2.20 0.37 0.22 0.34 M .62 .34 .18 .28 .62 .45 .25 .39 SD 5 (40 to 50) 2.22 0.29 0.10 0.24 2.20 0.33 0.23 0.31 M .63 .35 .18 .30 .64 .43 .26 .39 SD 6 (50 to 60) 2.21 0.27 0.12 0.24 2.18 0.34 0.23 0.33 M .61 .34 .19 .29 .64 .45 .26 .40 SD 7 (60 to 70) 2.17 0.25 0.14 0.23 2.14 0.33 0.25 0.33 M .62 .32 .21 .28 .65 .44 .27 .40 SD 8 (70 to 80) 2.13 0.25 0.15 0.23 2.13 0.32 0.27 0.33 M .64 .32 .22 .29 .62 .42 .28 .39 SD 9 (80 to 90) 2.12 0.25 0.18 0.25 2.11 0.32 0.28 0.33 M .63 .31 .22 .29 .62 .43 .28 .41 SD 10 (90 to 100) TABLE 2: Means and Standard Deviations of Deviance and Impulsivity Measures by Neighborhood Disadvantage Decile Groups 526 CRIMINAL JUSTICE AND BEHAVIOR Again, using Cohen’s d, effect sizes were small to moderate, ranging from d = .17 to d = .53. Unlike aggression, nonviolent delinquency decreased with neighborhood disadvantage for both males, F(1, 9826) = 27.71, p < .05, and females, F(1, 10012) = 22.12, p < .05. Among males, nonviolent delinquency was highest in the 2nd decile (.41). ANOVAS revealed significant differences between the nonviolent delinquency in this decile and the nonviolent delinquency in deciles 6 (.33), 7 (.34), 8 (.33), 9 (.33), and 10 (.32). Nonviolent delinquency in decile 1 (.39) was significantly greater than deciles 9 and 10 (d = .16). Nonviolent delinquency was also highest in decile 2 (.39) for females, which differed significantly from deciles 8 (.25), 9 (.25), and 10 (.25; d = .43). Impulsivity. The distribution of impulsivity varied significantly by neighborhood disadvantage for both males, F(1, 9876) = 47.12, p < .05, and females, F(1, 10043) = 32.73, p < .05. Male impulsivity was highest in low-disadvantage neighborhoods and declined with increasing neighborhood disadvantage. ANOVA results, F(9, 9868) = 6.97, p < .05, revealed that impulsivity in the 1st decile (2.28) was significantly higher than that of deciles 3 (2.18), 4 (2.17), 7 (2.18), 8 (2.14), 9 (2.13), and 10 (2.11). Decile 2 (2.25) was significantly less than deciles 8, 9, and 10 (ds = .22 or less). As with the males, female impulsivity was generally lower in neighborhood deciles with greater disadvantage, F(9, 10035) = 4.41, p < .05. It was highest in the 2nd decile (2.25), which significantly differed from both the 9th (2.14) and 10th (2.12) deciles. The 10th decile was also significantly lower than deciles 1, 3, 4, 6, and 7 (ds = .16 or less). Before concluding the consideration of mean differences by neighborhood deciles, two issues should be emphasized. First, the approach used was chosen to maximize the opportunity to examine patterns of mean level differences across levels of neighborhood disadvantage. Second, although Tukey tests were used to minimize Type 1 error, the Add Health’s large sample size provides substantial opportunities to identify small, albeit significant, differences. CORRELATIONS Table 3 provides the zero-order correlations among the three deviance measures, impulsivity, and neighborhood disadvantage for Vazsonyi et al. / EFFECT OF IMPULSIVITY ON DELINQUENCY 527 TABLE 3: Correlations Between Deviance Measures, Impulsivity, and Neighborhood Disadvantage 1 1. General delinquency 2. Aggression 3. Nonviolent delinquency 4. Impulsivity 5. Neighborhood disadvantage .65* .95* .16* .00 2 3 4 5 .68* .95* .50* .14* .09* .15* −.02 .08* −.05* −.07* .45* .07* .14* .10* −.05* −.06* Note. Correlations for males (n = 9,826) are above the diagonal and for females (n = 10,012) are below the diagonal. *p < .0001. male and female adolescents. Associations among the outcomes ranged from .45 (between aggression and nonviolent delinquency for females) and .95 (between general and nonviolent delinquency for both sexes). This latter correlation reflects the influence of nonviolent delinquency on variance in general delinquency. As would be expected from the ANOVA results, associations between the outcomes and impulsivity and neighborhood disadvantage were small but, with the exception of the association between general delinquency and neighborhood disadvantage, statistically significant. All three outcomes were positively associated with impulsivity. MULTILEVEL REGRESSION MODELS Models with fixed effects for impulsivity, neighborhood disadvantage, and the interaction between impulsivity and neighborhood disadvantage were estimated to examine the hypothesis for each of the three delinquency outcomes. In addition to these three fixed effects, each model included a random effect for the school cluster variable intercept as well as for impulsivity. We did not specify a random effect for the covariance between impulsivity and the school cluster intercept, as this random effect would have confounded the examination of the specific impulsivity by neighborhood disadvantage interaction that operationalized our hypothesis. To account for the skew of the original outcome variables, these models used square-root corrected scores for the outcomes variables (see Table 1). Table 4 provides the male results on the left and female results on the right. For males, the analyses sample was limited to n = 9,830 528 .108 .111 Impulsivity Residual Total .001 .002 .000 4.86 69.58 0.00 9,682 9,682 –0.84 0.89 15.47 .0001 .360 .050 .253 –.015 Intercept Impulsivity Nhood Dis IMP * ND .005 .050 .037 .005 142 9,682 9,682 9,682 –0.29 6.57 9.57 66.47 .0001 Intercept Nhood Dis IMP * ND ns Impulsivity .0001 .0001 Total Residual Intercept Impulsivity IMP * ND ns ns Nhood Dis ns .0001 Impulsivity .0001 Aggression (male observations = 9,830; female observations = 10,014) .003 .000 Intercept Covariance estimates .051 .038 9,682 .043 .353 .035 .203 .087 .082 .001 .004 –.092 .092 .034 .006 –.043 Est. Fixed Effects IMP * ND p< Nhood Dis t .377 df .080 Z .082 SE Impulsivity Est. Delinquency (male observations = 9,830; female observations = 10,014) Intercept .461 .006 142 74.77 .0001 Intercept Fixed Effects Males .043 .031 .005 .006 .001 .000 .001 .046 .033 .006 .007 SE 69.81 1.85 5.65 Z Females 9,868 9,868 9,868 142 9,868 9,868 9,868 142 df 1.00 11.27 7.06 32.66 –1.97 2.80 14.21 57.76 t ns .0001 .0001 .0001 .0001 .05 .0001 .05 .01 .0001 .0001 p< TABLE 4: Mixed Model Results Predicting Deviance Measures From Impulsivity, Neighborhood Disadvantage, and Their Interaction 529 .103 .105 Impulsivity Residual Total .000 .001 .000 4.43 69.61 0.00 Total Residual ns .0001 Intercept Impulsivity .0001 –.055 –.044 Impulsivity Nhood Dis IMP * ND .134 Residual Total .008 .002 .000 .001 .055 .043 .006 69.57 0.00 5.43 142 9,682 9,682 9,682 58.46 –0.80 –1.29 16.31 Total Residual Impulsivity ns .0001 Intercept IMP * ND ns .0001 Nhood Dis ns Intercept Impulsivity .0001 .0001 .004 .105 .099 .001 .005 –.144 –.021 .090 .397 .078 .074 .000 .001 .001 .001 .001 .051 .037 .006 .007 .001 .000 5.69 69.83 1.96 5.17 69.82 0.87 142 9,868 9,868 9,688 56.32 –2.81 –1.03 14.40 .0001 .05 .0001 .01 ns .0001 .0001 .0001 ns .0001 Note. Est. = unstandardized estimate; Nhood Dis = neighborhood disadvantage; IMP*ND = impulsivity and neighborhood disadvantage. .000 .129 Impulsivity .005 Intercept Covariance estimates .448 .095 Intercept Nonviolent delinquency (male observations = 9,828; female observations = 10,014) .002 .000 Intercept Covariance Estimates 530 CRIMINAL JUSTICE AND BEHAVIOR individuals. For general delinquency, the fixed effect for impulsivity was significant. However, the fixed effects for neighborhood disadvantage and the impulsivity by neighborhood disadvantage interaction were both nonsignificant. Results for the random effect indicated significant variance among school clusters. However, 97% of the variance in general delinquency remained in the residual, indicating that very little was associated with school cluster differences. For aggression, the fixed effects for both impulsivity and neighborhood disadvantage were significant, with both increasing the likelihood of aggressive behaviors. Similar to the results for general delinquency, however, the impulsivity by neighborhood disadvantage interaction was not statistically significant. Covariance estimates revealed that as with general delinquency, nearly all (98%) of the variance in aggression was located within, rather than between, school clusters. Results for nonviolent delinquency were also similar. Impulsivity had a significant and positive fixed effect on nonviolent delinquency. The fixed effects for neighborhood disadvantage and the interaction term, however, were both nonsignificant. Most of the variance remaining in the conditioned error terms was apportioned to the residual (96%) rather than to the intercept or impulsivity terms. On the right in Table 4 are the analysis results for females. For the first outcome, delinquency, the impulsivity fixed effect was statistically significant (p < .0001). Unlike the results based on male youth, the fixed effects for both neighborhood disadvantage (p < .01) and the impulsivity by neighborhood disadvantage interaction (p < .05) were statistically significant. Each of the covariance estimates was statistically significant. Yet similar to findings based on the male sample, 94% of the variance in general delinquency existed within rather than between clusters. Female results for aggression were similar to those of males; both impulsivity and neighborhood disadvantage were associated with greater aggression. However, the effect of impulsivity was not moderated by neighborhood disadvantage. The majority of the remaining variance (95%) was accounted for by the residual. Nonviolent delinquency was also significantly associated with the fixed effect for impulsivity but not neighborhood disadvantage. The impulsivity by neighborhood disadvantage interaction was statistically significant Vazsonyi et al. / EFFECT OF IMPULSIVITY ON DELINQUENCY 531 (p < .01). Again, covariance analysis results indicated that random effects for the school cluster intercept (p < .0001) and impulsivity (p < .05) were statistically significant. However, as found for previous models, variance of female nonviolent delinquency was nearly entirely (94%) at the individual (residual) level. In summary, results differed slightly between males and females. Among males, impulsivity was significantly associated with increases in each of the three delinquent outcomes, whereas neighborhood disadvantage was associated only with an increase in aggression. Most important, the fixed effect for the impulsivity by neighborhood disadvantage interaction was not statistically significant for any of the three outcome measures. Among females, impulsivity was significantly associated with greater amounts of each of the three delinquency outcomes, and neighborhood disadvantage was associated with variance in general delinquency and aggression but not nonviolent delinquency. Unlike the results among males, the impulsivity of female participants significantly interacted with neighborhood disadvantage for both general delinquency and nonviolent delinquency. To provide the greatest opportunity for the trait by context hypotheses, the models described above were run without covariates (race/ethnicity and family structure status). However, to insure that the significant results obtained were not a product of factors that might covary with impulsivity and neighborhood disadvantage influences, models were repeated with race/ethnicity (dummy coded variables for Black, Hispanic, Asian, and American Indian/Other) and traditional (i.e., intact) family status covariates. The entry of these covariates had a minimal effect on the above presented results. Specifically, each of the six significant fixed effects (p < .0001) for impulsivity as well as the significant fixed effects for neighborhood disadvantage on aggression among both males (p < .0001) and females (p < .0001) maintained their significance. For males, the significance of the fixed effect for neighborhood disadvantage on aggression was slightly reduced from less than .0001 to p = .0060. For females, the fixed effect for neighborhood disadvantage on general delinquency, which was significant at the .01 level in the presented models, was no longer significant. The two significant interactions, both among females, retained their significance levels. These interactions are examined fully below. 532 CRIMINAL JUSTICE AND BEHAVIOR Among the covariates, being part of an intact family was significantly (p > .0001) related to less of each of the three delinquency outcomes for both males and females. Relative to the Whites of the same sex, general delinquency was higher among Hispanic (p < .05) and American Indian/Other females (p < .01), as well as Black (p < .05) and Hispanic (p < .0001) males. Aggression was greater among Blacks, Hispanics, and American Indian/Other females and males (all at the p < .0001 level). Finally, nonviolent delinquency was higher among Black (p < .05) males and Hispanic females (p < .0001). FOLLOW-UP ANALYSES BY DECILES Out of a possible six interactions considered by our mixed model analyses, results revealed two significant fixed effects indicating the moderation of impulsivity by neighborhood disadvantage. Given the substantial sample size of the analyses samples, it is important to consider the size of these interaction effects. To do this, we computed standardized associations between impulsivity and the three delinquency outcomes across decile groups of neighborhood disadvantage. This provided an opportunity to examine the magnitude of any differences in the association between impulsivity and the delinquency outcomes. A second reason for considering the distribution of the associations across neighborhood disadvantage deciles is that prior research has revealed that the effects of disadvantaged neighborhood context may not accrue linearly. Instead, such effects may occur above cutoffs of disadvantage (Crane, 1991), something our linear models would not necessarily have documented. Accordingly, we took advantage of the substantial sample size offered by the Add Health to compute correlations between impulsivity and the three outcome measures across deciles of neighborhoods defined by their level of neighborhood quality. Table 5 includes correlations between impulsivity and the three behavioral outcomes for both sexes by decile group. On inspection, the patterns of correlations across deciles of neighborhood disadvantage appear to contradict the notion that the association between impulsivity and delinquency is conditioned by neighborhood disadvantage. For example, among males, associations for general delinquency, although peaking at .22 in the 6th decile, are .14 and .13 in 533 .13 .07 .14 .20 .07 .20 .14 .07 .16 .13 .03 .14 .19 .14 .19 .12 .10 .12 .17 .09 .18 .14 .11 .15 .21 .05 .23 .13 .10 .13 Note. These associations are not conditioned by the entry of control variables. Males General delinquency Aggression Nonviolent delinquency Females General delinquency Aggression Nonviolent delinquency 1 2 3 4 5 (0 to 10) (10 to 20) (20 to 30) (30 to 40) (40 to 50) .10 .04 .11 .22 .16 .23 .11 .03 .11 .13 .08 .13 6 7 (50 to 60) (60 to 70) .16 .10 .17 .12 .08 .13 .09 .05 .08 .14 .08 .14 8 9 (70 to 80) (80 to 90) TABLE 5: Correlations Between Deviance Measures and Impulsivity by Neighborhood Disadvantage Decile Groups .21 .16 .18 .14 .10 .14 10 (90 to 100) 534 CRIMINAL JUSTICE AND BEHAVIOR the two least disadvantaged neighborhoods and .14 in the two most disadvantaged. Patterns are similar for aggression and nonviolent delinquency. Associations for aggression begin at .07 for the first two deciles and increase only to .08 and .10 in the last two. Corresponding correlations for nonviolent delinquency are .16, .14, .14, and .14. Similar to general delinquency, with which it correlates at .95, nonviolent delinquency’s association with impulsivity peaks in the 6th decile (.23). Results were less clear among females. For general delinquency, associations did not appear to systematically change from lowdisadvantage to high-disadvantage neighborhoods; standardized associations for the first two deciles were .13 and .20, and .09 and .21 for the last two. In contrast, correlations for aggression suggest that neighborhood disadvantage may increase the impact of impulsivity, rising from .03 and .07 in the lowest deciles to .05 and .16. This pattern did not hold across all deciles, however, as the 3rd decile’s association was .14. Consistent with the significant fixed effect (p < .01) for the impulsivity by neighborhood disorganization interaction in Table 4, correlations for nonviolent delinquency suggest a modest decrease in association across levels of neighborhood disadvantage. However, this effect is small and may be driven by the 9th decile’s .08 mean. This is seen by comparing the average of the first three deciles and the last three, which are .176 and .143, respectively. This suggests that despite the statistical significance of neighborhood disadvantage’s moderation of the influence of impulsivity on nonviolent delinquency in this large sample, the magnitude of its effect is quite modest. It is also worth noting that the patterns of associations across these deciles do not provide any evidence of a curvilinear effect. DISCUSSION The purpose of the current investigation was to test competing predictions by the general theory of crime and personality research and by the social disorganization tradition on the relationship between impulsivity, one manifestation of low self-control, and different measures of deviance across different levels of neighborhood disadvantage. As such, the study tested whether contexts or developmental Vazsonyi et al. / EFFECT OF IMPULSIVITY ON DELINQUENCY 535 ecologies are strong predictors of deviance. Based on a nationally representative sample of more than 20,000 male and female adolescents, the answer appears to be that they are not. With small or absent interactions among male youth between the impulsivity–delinquency relationship and neighborhood type and only small interactions among females, our study also failed to reproduce parallel findings (male only) to the Lynam et al. (2000) study, whose measures of impulsivity were admittedly different. Instead, our findings are more consistent with the invariance hypothesis derived from personality research and the general theory of crime. It is a strength of our study that our sample size was large enough to detect small interaction effects. However, subsequent examination across deciles of neighborhood disadvantage supported the hypothesis that the impulsivity–delinquency association is largely independent of contextual influences. Our results provide little support for social disorganization theory, especially concerning the importance of impulsivity in predicting delinquency.4 Whereas general levels of delinquency and aggression increased with neighborhood disadvantage, nonviolent delinquency decreased. The social processes proposed by social disorganization theory—high levels of social disorganization and low levels of social efficacy—should increase both aggression and nonviolent delinquency. Although we did not predict level results, nor were we primarily interested in level differences across neighborhoods in the current study, the different patterns of aggression and nonviolent delinquency across neighborhoods deserve comment. Perhaps the relative paucity of adults concerned with resolving conflicts of interest before they escalate into criminal and aggressive acts in disadvantaged neighborhoods contributes to the higher amount of adolescent aggression in these neighborhoods (see Sampson, 1991). In contrast, the life strategies of adults residing in middle-class neighborhoods may make these communities especially vigilant for and intolerant of physical aggression. Similar to normative adolescent substance use, a “kids will be kids” allowance of nonviolent delinquency might exist in middle-class neighborhoods. Although the evidence suggests that impulsivity was related similarly to behavioral outcomes across neighborhood contexts, other factors, such as family influences (e.g., parental monitoring), may be 536 CRIMINAL JUSTICE AND BEHAVIOR more important in some contexts than others. For example, using a behavioral genetic approach to examine contextual effects on adolescent aggression with the same data set used in this study, Cleveland (2003) found that shared environmental influences accounted for significantly greater variance within disadvantaged neighborhoods than in adequate neighborhoods. The importance of positive socialization pressures and effects by parents or caregivers cannot be emphasized enough; perhaps this is best exemplified by the fascinating work that tested for the existence of a genetic marker for violence, aggression, and crime, the MAOA gene. This work has provided evidence that the gene is polymorphic (Caspi et al., 2002). The researchers failed to find evidence of a “gene main effect” in predicting antisocial behavior. However, they did find a main effect of child maltreatment and of the Child Maltreatment X Gene interaction, where individuals with high MAOA activity were affected to a lesser degree by maltreatment than individuals with low activity. Finally, the importance of consistent socialization pressures also clearly manifested itself in the current study through the finding that traditional, two-biological-parent families were less likely to have male and female youth involved in all forms of delinquency in comparison to adolescents from other family forms. This finding clearly provided evidence that proximal socialization efforts and effects may be at the very least more efficient in one family form versus another. Of course, how and whether systematic differences existed in the quality and effects of this process is beyond the information we currently tested in the study, although clearly of interest and of importance for future work. Our study also provided little support for an alternate interpretation of social disorganization theory, namely that offending in disorganized neighborhoods can be explained by social disorganization itself without the need to measure the characteristics of the individual. In this model, individuals not particularly biased toward offending might offend anyway because of environmental factors, including but not limited to increased opportunities for crime, peer pressure, and a lack of external social controls. This suggests that if impulsivity predicts delinquency at all, it should do so only in relatively organized neighborhoods. Thus, a trait by context interaction should be Vazsonyi et al. / EFFECT OF IMPULSIVITY ON DELINQUENCY 537 observed but one wherein the association between impulsivity and delinquency would be greater in advantaged or adequate neighborhoods than in disadvantaged neighborhoods. Taken together, these findings provide further support for the empirical work connecting impulsivity and impulse control with crime and delinquency (Farrington, 1995; Moffitt et al., 2002; Pulkkinen, 1982, 1986; Pulkkinen et al., 2000; White et al., 1994) as well as for predictions by the general theory (Gottfredson & Hirschi, 1990), although neighborhood disadvantage itself did have some explanatory power in our models (for a further discussion of both traditions, see Wiebe, 2003, 2004). To date, empirical work on the general theory has not explicitly tested variability in SES as a moderator of the self-control–deviance relationship (Pratt & Cullen, 2000), although some work has focused on the context-invariance thesis through empirical tests of the theory that have been completed in different countries (Forde & Kennedy, 1997; LaGrange & Silverman, 1999) or that have been completed as cross-national comparative tests (Vazsonyi, Pickering, Junger, & Hessing, 2001; Vazsonyi, Wittekind, Belliston, & Van Loh, 2004) or cross-cultural/ethnic ones (Vazsonyi & Crosswhite, 2004). In general, these studies have provided further support for the theory. At the same time, impulsivity constitutes only a small part of the much broader self-control construct, and thus, conclusions or generalizations about the empirical support for the theory are tentative. Yet the present study, with its extensive test of the impulsivity–deviance relationship, provides evidence in support of the predictions made by self-control theory. Findings about the modest amount of variance explained by impulsivity are consistent with some previous empirical studies that have assessed the importance of individual selfcontrol dimensions (e.g., Longshore, Turner, & Stein, 1996). However, other examinations of the impulsivity–delinquency relationship have found more substantial associations based on more comprehensive assessments of both behavioral (19% of variance explained) and cognitive (4% of variance explained) impulsivity (Moffitt et al., 2002; see also Caspi et al., 1994). In conclusion, the current study provides strong evidence based on a nationally representative sample of more than 20,000 male and female adolescents supporting the invariance of the 538 CRIMINAL JUSTICE AND BEHAVIOR impulsivity–deviance relationship across varied levels of neighborhood ecologies, a finding that is particularly consistent with theoretical predictions by the general theory of crime as well as previous empirical findings on the etiology of deviance from the personality literature. Though a more comprehensive assessment of self-control would have allowed broader generalizations of the findings, particularly for the general theory of crime, we believe that impulsivity, one dimension of the self-control construct, provides good confirmation. This evidence also casts some doubt on the importance of specific developmental ecologies or contexts in the understanding of crime and deviance, although only additional empirical work can ultimately provide the answer. APPENDIX DELINQUENCY ITEMS In the past 12 months, how often did you . . . 1. paint graffiti or signs on someone else’s property or in a public place? 2. deliberately damage property that didn’t belong to you? 3. lie to your parents or guardians about where you had been or who you were with? 4. take something from a store without paying for it? 5. get into a serious physical fight? 6. hurt someone badly enough to need bandages or care from a doctor or nurse? 7. run away from home? 8. drive a car without its owner’s permission? 9. steal something worth more than $50? 10. go into a house or building to steal something? 11. use or threaten to use a weapon to get something from someone? 12. sell marijuana or other drugs? 13. steal something worth less than $50? 14. take part in a fight where a group of your friends was against another group? 15. act loud, rowdy, or unruly in a public place? NOTES 1. Because we were interested in testing the same questions posed by Lynam et al. (2000), we chose an analytic framework that would maximize the opportunities to find significant effects. Specifically, we chose to use OLS regressions without a repeated measures format to provide the greatest chance of revealing significant interactions. Vazsonyi et al. / EFFECT OF IMPULSIVITY ON DELINQUENCY 539 2. Sampson (1997) and Sampson, Morenoff, and Earls (1999) have used a five indicator scale for neighborhood disadvantage. We used three indicators instead of five because additional indicators would have been redundant or inconsistent with the neighborhood disadvantage theory guiding the composite’s construction. In addition to the three indicators used here, Sampson et al. (Sampson et al., 1999; Sampson et al., 1997) used proportion receiving public assistance and proportion Black. The percentage on public assistance was excluded here because it did not change the classification of neighborhoods. Proportion Black was excluded for two reasons. First, its loading (.60) was much lower than the loadings for the other indicators, which ranged from .86 to .94. Second, research by Simons, Johnson, Beamon, Conger, and Whitebeck (1996) suggests that the effects of disadvantage similarly accrue in rural areas. The exclusion of Black also accords with the original theory of Shaw and McKay (1969), which focuses on structural, rather than cultural, barriers. For these two reasons, using this indicator did not seem appropriate. 3. The items we selected from the family and personality section of the Add Health study tap into what we believe to be indicators of self-control most generally speaking. However, we did not think that we should or could call this self-control, as the measure did not assess all theoretically specified components of self-control. We believe, though, that the scale does tap into impulsivity, even though it does not directly compare to the one used by Lynam et al. (2000). Based on an examination of the face validity of the items, we found this scale an appropriate approximation of impulsivity because the items assess a lack of deliberate thinking/planning, an inability to delay gratification, an unwillingness to weigh different consequences of a decision or a behavior, and a “here and now” orientation. 4. It is important for us to note that the limited and inconsistent empirical support that exists for social disorganization theory has been questioned because of potential biases in official statistics and because of potential biases against poor and/or minority youth and adults documented in the criminal justice system. Thus, inasmuch as our data do not provide empirical support for social disorganization theory or predictions, this is consistent with some previous work based on self-report methodology. REFERENCES Block, J. H., & Block, J. (1980). The role of ego-control and ego-resiliency in the organization of behavior. In W. A. Collins (Ed.), Minnesota symposia on child psychology (Vol. 13, pp. 39-101). Hillsdale, NJ: Lawrence Erlbaum. Bronfenbrenner, U. (1979). The ecology of human development: Experiments by nature and design. Cambridge, MA: Harvard University Press. Bursik, R. J. (1988). Social disorganization and theories of crime and delinquency. Criminology, 26, 519-551. Bursik, R. J., & Grasmick, H. G. (1993). Neighborhoods and crime: The dimensions of effective community control. New York: Lexington. Caspi, A., McClay, J., Moffitt, T. E., Mill, J., Martin, J., Craig, I. W., et al. (2002). Role of genotype in the cycle of violence in maltreated children. Science, 297, 851-854. Caspi, A., Moffitt, T. E., Silva, P. A., Stouthamer-Loeber, M., Krueger, R. F., & Schmutte, P. S. (1994). Are some people crime-prone? Replications of the personality-crime relationship across genders, races, and methods. Criminology, 32, 163-195. Cleveland, H. H. (2003). Disadvantaged neighborhoods and adolescent aggression: Behavioral genetic evidence of contextual effects. Journal of Research on Adolescence, 13, 211-238. 540 CRIMINAL JUSTICE AND BEHAVIOR Crane, J. (1991). The epidemic theory of ghettos and neighborhood effects on dropping out and teenage childbearing. American Journal of Sociology, 96, 1226-1259. Eysenck, H. J. (1977). Crime and personality. London: Routledge. Farrington, D. P. (1990). Implications of criminal career research for the prevention of offending. Journal of Adolescence, 13, 9