The Effect of Socioeconomic Status on Delinquency Across Urban and Rural Contexts

Abstract

An extensive line of research has demonstrated that low socioeconomic status (SES) is a risk factor for adolescent delinquent behavior. The possibility that low SES affects adolescent’s risk for engaging in delinquent behavior has garnered a significant amount of empirical and public attention, given its implications for delinquency prevention. However, few studies have examined the association between low SES and delinquent behavior across urban and rural contexts in the United States. Moreover, much is unknown about the strength of the association between low SES and delinquency across urban and rural context after controlling for common genetic liabilities that often cluster within different levels of SES. The present study aimed to address these existing gaps in the literature by conducting a genetically informed analysis of sibling pairs from a nationally representative sample of U.S. youth. The results revealed that shared environmental factors accounted for 17% of the population variation in adolescent delinquent behavior among adolescents growing up in urban contexts, and 3% of this family-wide environmental effect was accounted for by SES. No evidence of a family-wide environmental effect on population variation in delinquent behavior was found among adolescents from rural contexts. Findings from the present study suggest that the association between low SES and delinquency in urban contexts in the United States may be a true environmental effect and highlight the utility of using genetically informed research designs to better understand the extent to which social contexts influence adolescent delinquent behavior.

Keywords

delinquency socioeconomic status behavior genetics urban rural

There is considerable evidence to demonstrate that low socioeconomic status (SES) is associated with negative adolescent outcomes, including poor mental health and increased antisocial and delinquent behaviors (Devenish, Hooley, & Mellor, 2017; Piotrowska, Stride, Croft, & Rowe, 2015; Reiss, 2013). Several factors common to low SES areas can increase the likelihood of individuals’ participation in delinquent behavior. For example, the lack of monetary ability to acquire needed goods and services can increase delinquent involvement to reduce economic strain (Agnew, Matthews, Bucher, Welcher, & Keyes, 2008). Other risk factors include parental stress from raising children in poverty, a lack of parental supervision, living in neighborhoods with increased concentration of poor schools, and increased exposure to delinquent peers and criminal gangs (Costello, Keeler, & Angold, 2001). Furthermore, recent research has found that exposure to environmental toxins such as lead, which has been linked to cognitive deficits and criminal behavior (Boutwell et al., 2016; Wright et al., 2008), are more common in neighborhoods characterized by low SES and social disadvantage. Recent events in Flint, Michigan provide a clear example of this where children in socioeconomically disadvantaged neighborhoods experienced the largest increase in blood lead levels from 2013 to 2015 compared to children in less disadvantaged neighborhood environments (Hanna-Attisha, LaChance, Sadler, & Shnepp, 2016). Ultimately, an aggregate of negative factors associated with low SES may accumulate over time to increase the likelihood of adolescent delinquent behavior and future criminality.

Although a great deal of research has revealed that youth growing up in low SES urban environments are exposed to weak social institutions (e.g., collective efficacy, community organizations, and public schooling) and opportunities for offending (e.g., easy access to criminal gangs, lack of supervision, easy access to firearms), much less is known about whether and to what extent the link between low SES and delinquency remains for adolescent youth growing up in rural environments. Only recently have researchers begun to explore the varying effects of low SES on adolescent delinquency across urban and rural contexts. Contemporary research flowing from this body of work suggests that low SES is indeed associated with delinquent behaviors across both urban and rural environmental contexts (Bouffard & Muftic, 2006; Jiang, Sun, & Marsiglia, 2016). However, no research to date has controlled for genetic liabilities that often cluster within families nested within different levels of SES. Given that variation in antisocial and delinquent behavior has been found to be heritable (Boisvert, Wright, Knopik, & Vaske, 2012; Connolly & Beaver, 2014; Ferguson, 2010; Miles & Carey, 1997; Rhee & Waldman, 2002), it is critical to control for the confounding effects of genetic factors that cluster within families. Using a genetically informed design can allow for a better understanding of how specific social environments, such as low SES, are implicated in the development of adolescent delinquent behavior across geographic context.

The Link Between Low SES and Delinquent Behavior Across Urban and Rural Contexts

Considering the concentration of poverty in inner-city environments, many of the early studies concerning delinquency and low SES focused on urban populations (Bursik, 1988; Sampson, Morenoff, & Earls, 1999). In 1981, Lyerly and Skipper examined differential rates of urban and rural delinquency and found that males from urban areas were more likely to fall into the highly delinquent category. Specifically, of the urban sample, 68% of males were ranked as highly delinquent while only 32% of the rural males were similarly categorized (Lyerly & Skipper, 1981). Other studies have shown similar findings that when lower socioeconomic urban and rural areas are compared, urban youths exhibit higher levels of delinquency. For instance, Farrell, Sullivan, Esposito, Meyer, and Valois (2005) sampled 667 urban students and 301 rural students as part of a violence prevention program in middle schools and found that urban students exhibited higher levels of aggression, drug use, and delinquency compared to rural students. They also found that urban children showed earlier signs of aggressive and delinquent behavior and that these behaviors increased at a faster rate than in their rural counterparts (Farrell, Sullivan, Esposito, Meyer, & Valois, 2005). Hope and Bierman (1998) also examined children’s antisocial behaviors displayed at home and at school in both rural (n = 89) and urban (n = 221) areas. While the authors found no significant differences between urban and rural children for delinquency in the home setting, urban children reported significantly higher levels of externalized behavioral problems at school (Hope & Bierman, 1998).

There are several explanations proposed in the urban–rural literature concerning higher levels of delinquency in urban areas. These include the increased risk of exposure to deviant and delinquent peers, weaker social bonds, and less informal controls inherent of low socioeconomic inner-city areas (Hope & Bierman, 1998; Lyerly & Skipper, 1981). Therefore, researchers attribute the differences in levels of delinquency to characteristic differences between urban and rural communities. Specifically, rural communities, even low SES communities, exhibit characteristics that act as social buffers against delinquency. That is, rural communities have been found to be more conservative than their urban counterparts concerning factors that can influence delinquency, such as religion, morality, family authority patterns, and alcohol consumption (Glenn & Hill, 1977; Lowe & Peek, 1974). Furthermore, members of rural communities experience less complex social networks due to a lower number of residents in their proximity. In a similar vein, populations in rural environments tend to be more homogenous and generally maintain similar beliefs concerning behavior and the importance of said beliefs (Lyerly & Skipper, 1981). The maintenance of similar beliefs allows for informal social controls such as neighbors, families, churches, and schools to be more effective, hence acting as a buffer against delinquent behavior (Biggar, Forsyth, Chen, & Richard, 2016; Lyerly & Skipper, 1981). Contrary to rural areas, low-SES urban environments tend to exhibit lower levels of informal social controls, have higher population heterogeneity and residential mobility, and report fewer social bonds among community members (Agnew et al., 2008; Lyerly & Skipper, 1981). Thus, although low SES is associated with increased delinquency in general, when compared to one another, we see that urban and rural areas experience a different set of characteristics that may play a role in the prevalence of delinquency. However, recent research has found that the same negative social factors that increase delinquency in urban areas have similar effects in rural environments (Bouffard & Muftic, 2006). For example, low SES can reduce family structure stability in rural environments in a similar fashion as urban environments (Bouffard & Muftic, 2006). With weakened familial structures, informal social controls are decreased, leading to increased violent behavior (Bouffard & Muftic, 2006). This is not to say that urban and rural environments exhibit the exact same social pressures, as past research has shown they do not (Biggar et al., 2016; Jiang et al., 2016; Lyerly & Skipper, 1981).

Genetic and Environmental Contributions to Delinquent Behavior Across Urban and Rural Contexts

There is reason to suspect that the magnitude of genetic and environmental effects on delinquent behavior may vary across urban and rural contexts. Based on mainstream gene–environment explanations for child antisocial behavior (Beaver & Connolly, 2013), it is possible that siblings growing up in low-SES urban contexts are much more likely to be influenced by shared environmental experiences (i.e., environmental experiences that make siblings more similar to one another). This may be possible because siblings growing up in poor inner-city urban contexts are more likely to be exposed to common home stressors (i.e., abuse or neglect, family conflict, toxin exposure), school stressors (i.e., bullying victimization, high percentage of high school dropout, gang presence), and neighborhood stressors (i.e., code of the street, high levels of crime, low levels of collective efficacy) in a small concentrated area. The accumulation of stressors and disadvantage in a small area may also eliminate potential opportunities for growth and new nonshared environmental life experiences that create individual differences and increase social capital, such as moving away and attending college. Indeed, previous research has found that shared environmental influences have a greater influence on verbal intelligence for students coming from extremely low-SES backgrounds (Schwartz, 2015). Importantly, verbal intelligence has also been associated with risk of contact with the criminal justice system (Beaver et al., 2013). Taken together, many siblings residing in low SES urban contexts may be more likely, compared to siblings in rural contexts where risk is less concentrated, to be exposed to a host of high-risk environments that overwhelm genetic predispositions for prosocial development. One hypothesis that supports this argument is Raine’s (2002) social push hypothesis which argues that genetic influences are more likely to explain variation in antisocial behavior among individuals in low-risk environments, compared to high-risk environments, where there are several more opportunities for selection into delinquent peer groups and other criminogenic environments.

Recently, several behavior genetic studies have begun using quasi-experimental, genetically informed research designs to control for genetic liability shared between parents and offspring (Caspi, Taylor, Moffitt, Plomin, 2000; Rowe, 1994; Rowe & Rodgers, 1997). If adolescent problem behaviors are passed down from parents, who also have behavioral problems, this may interfere with the parent’s capacity to gain steady employment, provide a healthy home environment, and secure housing in a desirable urban or rural neighborhood. If this is the case, this would create a correlation between low SES and adolescent delinquency that would be net of any casual influence from low SES. Only a few genetically informed studies have tested this hypothesis using U.S. samples and found that after adjusting for familial confounds, the association between low SES and childhood problem behaviors remains, but is largely attenuated (Blau, 1999; D’Onofrio et al., 2009; Hao & Matsueda, 2006). To date, however, no research has used a genetically informed research design to examine whether the influence of low SES on adolescent delinquent behavior is different across urban and rural contexts in the United States. As such, much is unknown about whether low SES is a pervasive environmental risk factor for adolescent delinquency regardless of geographic context and genetic liability, and thus should continue to be a prime target of U.S. prevention programs. By controlling for genetic resemblance between family members, behavioral genetic designs can begin to answer this question and provide strong tests about which environments affect adolescent problem behaviors (Caspi et al., 2000; Thapar & McGuffin, 1996). Results from behavioral genetic research designs can therefore begin to unpack whether low SES is a universal risk factor for adolescent delinquency across urban and rural contexts.

The Current Study

In this study, we seek to expand on the existing literature on low SES and delinquency in two distinct ways. First, we examine the association between low SES and delinquency across urban and rural settings while controlling for several confounders to assess whether low SES is a risk factor for adolescent delinquent behavior in different geographical contexts. This first step will use multivariate linear regression, a standard social science method, designed to assess the association between an observable independent variable and an observable dependent variable while controlling for a host of observable confounds. For this segment of the analysis, we hypothesize that low SES will predict higher levels of delinquency among youth growing up in urban and rural contexts. Second, we utilize a genetically informed research design to further test the hypothesis that variation in SES partially explains individual differences in delinquent behavior above and beyond the influence of unobservable genetic liability. Because there is such a scarcity of research on this topic, it is difficult to put forth any a priori hypothesis regarding the association between SES and delinquency after considering unobservable genetic and shared environmental confounds. However, based on the findings from contemporary behavioral genetic research showing that delinquency is partially heritable and the fact that siblings are clustered within families, which are clustered within different categories of SES, we hypothesize that the effect of SES will be greatly attenuated once familial confounds are considered.

Method

Participants

This study draws on data from the National Longitudinal Survey of Youth 1997 (NLSY97). The NLSY97 is nationally representative sample of youth born between 1980 and 1984 in the United States. Data collection efforts have been funded by the Bureau of Labor Statistics and carried out by the National Opinion Research Center at Chicago University since 1996. Stratified, multistage household probability sampling techniques were used to generate two samples. The first sample was a “cross-sectional” sample (N = 7,335), which was designed to be a nationally representative sample of youth without survey weights, and the second sample was a “supplemental” sample (N = 2,463), which oversampled Black and Hispanic youth. Youth between the ages of 12 and 16 during the initial wave of data collection were asked to participate in the NLSY97, even if they resided in the same household. Because multiple children from the same household were included in the NLSY97, respondents were asked to report any biological or social relationship shared between themselves and other household members. Overall, over 3,500 biological sibling relationships were reported to exist between respondents in the NLSY97. Response categories used to help respondents report the degree of biological relatedness shared between household members were identical twin, full-brother, full-sister, half-brother, half-sister, male cousin, and female cousin. Researchers have recently used this information to assign additive genetic values to identified sibling pairs to conduct genetically informative analyses (Connolly & Beaver, 2016). To do so, researchers have assigned identical (or monozygotic [MZ]) twins an additive genetic coefficient of r = 1.00 since they share 100% of their additive genetic material, nonidentical (or dizygotic [DZ]) twins and full-siblings an additive genetic coefficient of r = .50 since they share, on average, 50% of their additive genetic material, and half-siblings an additive genetic coefficient of r = .25 since they have 25% of their additive genetic material. Because surveyors did not oversample for twins in the NLSY97, full-siblings constituted 90% of the sibling sample. To better even the distribution of sibling pairs across sibling categories, a random sample of full-sibling pairs was selected to be included in the final sample. Only one sibling pair per household with valid data on all variables analyzed in the present study was included in the final sample. As a result, the final analytic sample consisted of 1,532 respondents. Of the 1,532 respondents, 34 were identical twins (17 MZ twin pairs), 40 were nonidentical twins (20 DZ twin pairs), 1,204 were full-siblings (602 full-sibling pairs), and 224 were half-siblings (112 half-sibling pairs).

Measures

Delinquency

A 6-item self-report scale was used to measure delinquency during the second wave of the NLSY97 (1998). Respondents were asked to indicate whether they had committed one of the following delinquent acts since the date of their last interview: (1) carried a handgun (other than a rifle or shotgun); (2) purposely damaged or destroyed property that did not belong to them; (3) stolen something from a store or something that did not belong to them worth less than US$50; (4) stolen something from a store, person or house, or something that did not belong to them worth US$50 or more including stealing a car; (5) committed other property crimes such as fencing, receiving, possessing or selling stolen property, or cheated someone by selling them something that was worthless or worth less than what they said it was; or (6) attacked someone with the idea of seriously hurting them or have had a situation end up in a serious fight or assault of some kind. Response categories were 0 = no and 1 = yes. Responses were summed together to create a variety index of delinquency. The scale showed acceptable internal consistency (Cronbach’s α = .65), thus providing a reliable index of delinquent behavior during adolescence. After examining the association between SES and delinquency using standard social science methodologies, the measure of delinquent behavior was then standardized for behavioral genetic modeling. Table 1 presents values from both raw and transformed scales of delinquent behavior.

Table 1.

Descriptive Statistics.

Variable	Mean/%	SD	Min	Max
Delinquency₁₉₉₇
Raw	1.04	1.34	0	6
Transformed	−1.66	1	−0.75	3.76
Delinquency₁₉₉₈
Raw	0.55	1.04	0	6
Transformed	1.46	1	−0.53	5.23
Family income/SES
Raw	$45712	$45497	0	$425586
Transformed	10.33	1.04	2.63	12.96
Social context
Rural context	37.99%	–	–	–
Urban context	62.01%	–	–	–
Demographic controls
Age	14.26	1.46	12	17
Black/Hispanic/Other	47.45%	0.49	0	1
Male	50.46%	0.50	0	1

Note. N = 1,532. SD = standard deviation; SES = socioeconomic status.

Socioeconomic status (SES)

SES was measured by reports of gross household family income during the first (1997) and second wave (1998) of the NLSY97. In line with previous research studying the link between SES and delinquent behavior, we elected to use a measure of family income as an indicator of SES (Agnew et al., 2008; Hauser, 1994; Yu, 2016). Our reasoning for this was based on an extensive body of research showing that family income predicts a wide variety of different life outcomes also associated with SES, including differences in healthy child development (Linver, Brooks-Gunn, &Kohen, 2002), economic opportunity (Corak, 2013), and overall health and well-being for parents and their offspring (Alaimo, Olson, Frongillo, & Briefel, 2001). Family income or SES was log transformed to approximate normality. Table 1 presents descriptive statistics for the raw and log-transformed measure of SES used in the present study. As can be seen, the average household income for families in the sibling sample was $45,712.

Urban and rural context

Urban and rural context was measured by indictors of whether respondents lived in metropolitan statistical areas that were categorized as “urban” or “rural” regions based on information from the 1994 U.S. Census Bureau’s County and City Data Book (U.S. Census Bureau, 1994). Specifically, respondents living in closely settled, communities with a combination of residential, commercial, and retail areas, and have a population greater than 2,500 (Center for Human Resource Research, 2003) were coded as living in an urban context. Respondents living in areas that did not fit this definition were coded as living in a rural context. Urban and rural contexts were measured by a dichotomous measure where 0 = rural context and 1 = urban context. Table 1 shows that 62.01% of siblings in the analytic sample were identified as living in an urban context, while the remaining 38.99% of siblings were identified as living in a rural context.

Statistical controls

Age (measured in years), race (0 = non-Black, non-Hispanic, 1 = Black, Hispanic, Other), sex (0 = female, 1 = male), and prior delinquency measured at Wave 1 (1997) were controlled for in the analysis. To best control for prior delinquency, the same items that were used to measure delinquency during Wave 2 (1998) as the primary outcome were also used to measure prior delinquency. Measures of age, race, sex, and prior delinquency were used as statistical controls in each multivariate regression model and regressed on delinquency before behavior genetic model-fitting approaches were used to estimate the extent to which additive genetic factors and environmental factors, including SES, explain variance in delinquency. Table 1 presents the descriptive statistics for each control variable. As can be seen, the average age of siblings in the sample was 14 years old; 47.45% of the sample was Black, Hispanic, or another race; and 50% of the sample was male.

Data Analysis

The analysis for the present study was performed in a series of linked steps. First, three separate multivariate regression models were estimated to examine the effect of SES on self-reported delinquency while controlling for social context, prior self-reported delinquency, age, race, and sex. The first regression model examined the effect of SES and urban/rural context on delinquency, while the second model examined the effect of SES on delinquency among siblings residing in a rural context. The third and final model estimated the effect of SES on delinquency with siblings residing in an urban context.

The second step of the analysis focused on examining intraclass correlations between sibling pairs growing up in urban and rural contexts across different percentiles of SES. This stage of the analysis was conducted to assess the level of between-sibling concordance for self-reported delinquency across sibling pairs who shared different levels of additive genetic material. If concordance for delinquency is stronger for MZ twins (who share roughly 100% of their additive genetic material) compared to DZ twins and full-siblings (who share, on average, 50% of their additive genetic material) and stronger for DZ twins and full-siblings compared to half-siblings (who share 25% of their additive genetic material), then this can be interpreted as preliminary evidence that genetic factors explain a proportion of the population variation in delinquency.

After examining the intraclass correlations across different categories of sibling pairs, structural equation modeling was used to estimate a series of ACE models to test different hypotheses about how much of the variation in delinquency is influenced by three sets of latent influences—additive genetic influences (symbolized as A), shared or family-wide environmental influences (symbolized as C), and nonshared or child-specific environmental influences (symbolized as E, which also includes measurement error). Shared or family-wide environmental influences are environmental experiences that operate to make siblings develop comparable temperaments or traits that make them similar to one another (e.g., neighborhood influences, school influences, SES). Nonshared or child-specific environmental influences are environmental experiences unique to each sibling that make him or her develop different temperaments or traits compared to his or her co-sibling (e.g., different levels of toxin exposure, different levels of victimization, different peer groups). In addition to estimating the influence of additive genetic, shared environmental, and nonshared environmental factors on variation in delinquency, we examined the contribution of a shared environmental risk factor, SES, on individual differences in self-reported delinquency. Figure 1 illustrates the path model for this analytic method. To assess whether the effects of SES were context-specific, we examined sibling pairs residing in urban and rural contexts separately. The twin method (which includes ACE model-fitting approaches) has been found to be a robust analytical strategy capable of providing stable and reliable estimates of genetic and environmental influences (Barnes et al., 2014; Wright et al., 2015). Recent research has also showed that violating the equal environments assumption (a critical assumption of the classic twin design which argues that twins and siblings have similar shared environmental experiences) does not significantly inflate heritability estimates (Conley, Rauscher, Dawes, Magnusson, & Siegal, 2013; Felson, 2014). Moreover, a recent study by Kendler, Lönn, Maes, Sundquist, and Sundquist (2015) provides additional support for the robustness of the twin design in finding very similar heritability estimates for criminal behavior when examining full- and half-sibling pairs compared to MZ and DZ twin pairs.

Figure 1.

Path diagram for ACE model with socioeconomic status specified as a family-wide environment. Rectangles represent measured delinquency scores for siblings, and circles represent latent variables. Path estimates a1, c1, e1, a2, c2, and e2 represent the path coefficients from the latent variables on delinquency score variables. Variance in delinquency is partitioned into three latent components: additive genetic effects (A), shared or family-wide environmental effects (C), and nonshared child-specific environmental effects (E). Socioeconomic status was included as a measure of the shared or family-wide environment in the present study. This ACE model assumes that socioeconomic status is therefore not a factor separate from C, but one of the underlying sources that contributes to C, or the shared environment. Thus, the model provides two important variance estimates: (1) the variance in delinquency that is explained by socioeconomic status and (2) the variance in C that can be predicted by socioeconomic status (symbolized as m).

All ACE models were estimated using the structural equation program Mplus Version 7.1 (Muthén & Muthén, 1998–2012). We use three different model selection criteria to identify the best fitting and most parsimonious model. The first is the Santorra-Bentler scaled-difference chi-square (Δχ²) statistic (Satorra & Bentler, 2001). Based on the Santorra-Bentler scaled-difference statistic, a nonsignificant change in χ² indicates that the submodel with fewer parameters fit the data equally well compared to the baseline model and should be preferred based on parsimony. The second model selection statistic is the Akaike’s information criteria (AIC), with negative values indicating a better fitting model and the model minimizes AIC the most commonly selected as the best fitting model (Akaike, 1987). The third model selection statistic is the Bayesian information criteria (BIC), with increasing negative values indicating better model fit (Raftery, 1995).

Results

Table 2 presents the results from three separate multivariate regression models examining the effect of SES on delinquency after controlling for prior delinquency, age, race, and sex. Model 1 examines the full sample and shows that, on average, low SES was associated with higher levels of self-reported delinquency (b = −.04, p < .05) over and above the effect of social context, previous delinquent behavior, age, race, and sex. The results from Model 1 also revealed that social context significantly predicted delinquency with urban context increasing the amount of delinquency reported by siblings (b = .09, p < .05).¹ Model 2 presents findings from a second multivariate model examining the effect of SES on delinquent behavior among siblings growing up in a rural context. As presented, low SES was associated with higher levels of self-reported delinquency (b = −.02, p < .05) after controlling for prior delinquency, age, race, and sex. Model 3 examined the effect of SES on delinquency for siblings growing up in an urban context and revealed that low SES was also associated with higher levels of delinquency with the effect size being slightly larger than that found when examining siblings growing up in a rural context (b = −.06, p < .01).²

Table 2.

Multiple Regression Models Predicting Delinquency Across Urban and Rural Contexts.

	Model 1 (N = 1,532) Full Sample			Model 2 (N = 582) Rural Sample			Model 3 (N = 950) Urban Sample
Variable	b	SE	95% CI	b	SE	95% CI	b	SE	95% CI
SES	−.04*	.02	[−.08, −.01]	−.02*	.01	[−.05, −.01]	−.06**	.02	[−.09, −.02]
Urban context	.09*	.04	[.01, .18]	–	–	–	–	–	–
Delinquency₁₉₉₇	.29**	.02	[.24, .33]	.22**	.02	[.17, .28]	.31**	.01	[.27, .35]
Age	−.04**	.01	[−.06, −.01]	−.02	.02	[−.07, .01]	−.04*	.01	[−.08, −.01]
Black/Hispanic/Other	−.11**	.04	[−.20, −.02]	−.27**	.08	[−.44, −.11]	−.06	.05	[−.17, .04]
Male	.19**	.04	[.11, .27]	.15*	.07	[.01, .29]	.21**	.05	[.11, .32]

Note. SE = standard error of the mean; CI = confidence intervals; SES = socioeconomic status. 95% confidence intervals presented in brackets.

**p < .01. *p < .05.

Table 3 presents the intraclass correlations for delinquency across percentiles of SES in urban and rural contexts. The top half of the table presents the results for siblings growing up in an urban context. The results show that siblings growing up in an urban context in the bottom 25th percentile of SES reported, on average, the highest levels of delinquency (mean = .71) with siblings from the top 25th percentile reporting, on average, the lowest levels of delinquency (mean = .49). Across each percentile of SES, MZ twins reported stronger concordance for delinquency compared to DZ twins and full-siblings, and DZ twins and full-siblings reported stronger concordance for delinquency compared to half-siblings. The bottom half of Table 3 presents the intraclass correlations across percentiles of SES for siblings growing up in a rural context. Comparable to the results from the analysis examining siblings growing up in an urban context, siblings growing up in a rural context and in the bottom 25th percentile of SES reported, on average, the highest levels of delinquency (mean = .54). Siblings from the 75th and top 25th percentile reported, on average, the lowest levels of delinquency (mean = .42). MZ twins from a rural context demonstrated stronger concordance for delinquency compared to DZ twins and full-siblings across all percentiles of SES, while DZ twins and full-siblings demonstrated stronger concordance than half-siblings.

Table 3.

Intraclass Correlations for Delinquency Across SES in Urban and Rural Contexts.

Geographic Context	Bottom 25th Percentile	50th Percentile	75th Percentile	Top 25th Percentile
Urban context
Mean delinquency (n = 956)	0.71	0.53	0.56	0.49
SD	1.10	0.97	0.98	0.94
MZ twins (n = 22)
r	0.80	0.77	0.84	0.86
n pairs	3	3	2	3
DZ twins (n = 20)
r	0.61	0.53	0.60	0.74
n pairs	2	4	3	1
Full-siblings (n = 762)
r	0.57	0.56	0.48	0.50
n pairs	89	107	125	60
Half-siblings (n = 152)
r	0.35	0.24	0.26	0.19
n pairs	16	29	23	8
Rural context
Mean delinquency (n = 576)	0.54	0.37	0.42	0.42
SD	0.98	0.72	0.75	0.88
MZ twins (n = 12)
r	0.86	0.73	0.80	0.77
n pairs	1	2	2	1
DZ twins (n = 20)
r	0.68	0.71	0.62	0.69
n pairs	2	3	1	4
Full-siblings (n = 472)
r	0.47	0.50	0.53	0.52
n pairs	50	89	73	24
Half-siblings (n = 72)
r	0.18	0.26	0.25	0.31
n pairs	11	10	12	3

Note. SES = socioeconomic status; SD = standard deviation; MZ = monozygotic; DZ = dizygotic.

After establishing that SES predicted delinquency across urban and rural context and intraclass correlations indicating that genetic influences most likely explain a portion of the variance in delinquency across different levels of SES, the next step in the analysis was to estimate the contribution of additive genetic, shared environmental, and nonshared environmental influences on delinquency. Table 4 shows standardized ACE parameter estimates from all estimated models. As can be seen, an ACE model (Model 1) including all three latent parameters adequately fit the data for sibling pairs growing up in an urban context (AIC = 10.25, BIC = −26.59). However, model fit statistics indicated that Model 4 (AC_SESE), which included a specific measure of the shared environment to test the hypothesis that SES explains a proportion of the variance in adolescent delinquency independent of genetic influences, was a better fitting model (Δχ² = 6.35, p = .19, Δdf = 6, AIC = −8.91, BIC = −42.64). Therefore, including a measure of SES to the ACE model helped to account for familial aggregation of sibling delinquent behavior. According to standardized parameter estimates, the shared environment or family-wide environment accounted for 17% of the population variance in delinquency and SES accounted for 6% of this family-wide environmental effect, thus explaining 3.54% (17.03/6.04 = 3.54) of the total population variation in delinquency for siblings growing up in an urban context. Additive genetic factors accounted for 36% of the population variation in delinquency, while nonshared environmental factors accounted for 47% of the population variation.

Table 4.

Parameter Estimates From ACE Models With SES Specified as a Family-Wide Environment.

Model	A	C	E	SES	Δχ²	Δdf	AIC	BIC
Urban context
1. ACE	.36*	.17*	.47**	.00	–	–	10.25	−26.59
	[.28, .48]	[.06, .26]	[.39, .62]	[.00, .00]
2. CE	.00	.29**	.71**	.00	47.02**	5	156.43	117.62
	[.00, .00]	[.12, .42]	[.58, .88]	[.00, .00]
3. AE	.48**	.00	.52**	.00	19.03*	5	29.04	−19.32
	[.27, .60]	[.00, .00]	[.40, .73]	[.00, .00]
4. AC_SESE	.36*	.17*	.47**	.06*	6.35	6	−8.91	−42.64
	[.24, .40]	[.12, .23]	[.42, .58]	[.03, .09]
5. A_SESE	.41**	.00	.53**	.00	71.30**	6	132.68	108.35
	[.30, .58]	[.00, .00]	[.42, .70]	[.00, .00]
6. C_SESE	.00	.34**	.66**	.14*	83.61**	6	179.59	127.44
	[.00, .00]	[.24, .48]	[.52, 76]	[.08, .22]
Rural context
1. ACE	.43*	.09	.52**	.00	–	–	26.55	−18.91
	[.32, .50]	[−.01, .11]	[.47, .62]	[.00, .00]
2. CE	.00	.18**	.82**	.00	56.27**	5	93.20	83.75
	[.00, .00]	[.08, .31]	[.69, .92]	[.00, .00]
3. AE	.52**	.00	.48**	.00	10.84	5	−4.57	−32.06
	[.43, .62]	[.00, .00]	[.38, .57]	[.00, .00]
4. AC_SESE	.41*	.05	.38**	.02	21.15**	6	20.84	−18.70
	[.30, .58]	[−.01, .07]	[.21, .50]	[−.01, .03]
5. A_SESE	.48**	.00	.50**	.02	14.51**	6	10.65	−21.70
	[.36, .64]	[.00, .00]	[.34, .63]	[−.01, .04]
6. C_SESE	.00	.21**	.53**	.04	89.90**	6	189.76	140.32
	[.00, .00]	[.13, .37]	[.63, .87]	[−.01, 06]

Note. Standardized parameter estimates are presented. Best fitting models are given in boldface. SES = socioeconomic status; A = additive genetic effects; C = shared environmental or family-wide environmental effects; E = nonshared environmental effects; df = degree of freedom; AIC = Akaike’s information criteria; BIC = Bayesian information criteria.

*p < .05. **p < .01.

The bottom half of Table 4 presents the standardized parameter estimates from a series of ACE models examining the extent to which genetic and environmental influences contribute to individual differences in delinquent behavior among siblings growing up in a rural context. As presented in Table 4, an AE model fit the data best (Δχ² = 10.84, p = .25, Δdf = 5, AIC = −4.57, BIC = −32.06), with genetic influences explaining 52% of the population variation in delinquency and the nonshared environment explaining the remaining 48% of the population variation. Taken together, the family-wide environment, including SES, was found to not significantly explain any of the population variation in delinquency among youth growing up in a rural context.

Sensitivity analyses

Additional sensitivity analyses were conducted to assess the stability of the reported results. First, MZ and DZ twins were dropped from the sample and univariate ACE models were reestimated with only full- and half-siblings. This was done to evaluate whether excluding MZ and DZ twin pairs (only 5% of the sibling sample) produced different parameter estimates of additive genetic and environmental effects on delinquent behavior. Parameter estimates from this analysis did not produce substantively different results from the ACE models analyzing MZ and DZ twins and full- and half-siblings that are reported in the present study (results available upon request). Second, other indicators of SES were used to test whether a different measure of SES produced different results. Self-report measures of mother and father educational attainment were added to reported total net family income to create a variety index measure of SES. Mother and father educational attainment were both significantly correlated with logged family income (mother education: r = .23, p < .01; father education: r = .30, p < .01), and the created index measure of SES demonstrated acceptable internal reliability (Cronbach’s α = .62). Results from the reestimated multivariate regression and biometric models did not, however, produce substantively different results from the results reported in the present study. The only notable difference was that the SES parameter estimate from the best fitting AC_SESE model for youth growing up in urban contexts was slightly smaller (.03), but still statistically significant (p = .019). The results from these analyses are available upon request.

Discussion

Previous research on SES and delinquency has produced an impressive body of evidence suggesting that low SES, or neighborhood disadvantage, operates as a risk factor for adolescent delinquency (D’Onofrio et al., 2009; Goodnight et al., 2012; Hao & Matsueda, 2006; Leventhal & Brooks-Gunn, 2000; Rutter, Giller, & Hagell, 1998). While previous research has helped to show that SES has a small—albeit significant—effect on adolescent delinquency, very little research has examined the varying effects of SES on delinquency across geographic context in the United States. More importantly, virtually no research within criminology has evaluated the influence of SES on variation in delinquency while controlling for genetic liability shared among siblings growing up in the same urban or rural context. With this mind, the present study took advantage of both standard social science and genetically informative research designs to (1) assess the association between SES and delinquency among youth growing up in urban and rural geographical contexts and (2) disentangle genetic from environmental influences on variation in delinquency to offer a rigorous test of the causal hypothesis that low SES increases risk for adolescent delinquency. Overall, three key findings emerged from the analysis that warrant discussion.

First, low SES predicted higher levels of adolescent delinquency in the full sample and in the sample of only urban youth and only rural youth. The effect was slightly stronger when examining youth from urban contexts. SES remained a significant predictor of adolescent delinquency across context while controlling for prior levels of delinquency, age, race, and sex. Evidence of a small—yet significant—association between low SES and self-reported delinquency coincides with a well-developed body of research, suggesting that low-SES settings provide youth with many opportunities for offending (Sampson & Laub, 1994). Youth growing up in low-SES urban or rural environments may be exposed to more delinquent youth, poor school systems, criminal gangs, and opportunities to experiment with drugs and alcohol. Although a great deal of attention has been paid to examining the link between low SES and delinquency in urban contexts, such as inner-city neighborhoods (Anderson, 2000), results from the first stage of the analysis indicate that SES may also influence delinquent behaviors in rural contexts. As such, results from the standard social science models offered preliminary evidence, suggesting that low SES may be a pervasive risk factor for adolescent delinquency in both urban and rural places.

The second major finding from the present study was that SES explained roughly 3% of the population variation in urban youth’s delinquent behaviors after controlling for underlying genetic liability. While this finding cannot be interpreted as causal evidence, since only randomized experimental designs can produce this level of scientific support, the finding does indeed indicate that low SES is one specific environmental risk factor that increases delinquency in urban settings. Although the proportion of population variation explained by SES was small, the influence may be masked by several other more proximal risk factors related to early delinquent behavior, such as poor nutrition (Jackson, 2016), low academic achievement (Jaggers, Robison, Rhodes, Guan, & Church, 2016), subcultural influences (Anderson, 2000), delinquent peers (Connolly, Schwartz, Nedelec, Beaver, & Barnes, 2015; Thomas, 2015), and gang membership (Pyrooz, Turanovic, Decker, & Wu, 2016). Indeed, many of these risks have been found to cluster within disadvantaged urban environments (Leventhal & Brooks-Gunn, 2000). It is important to note that while other genetically informed studies have found similar inverse associations between family income and youth conduct disorder after controlling for familial confounds (D’Onofrio et al., 2009; Hao & Matsueda, 2006), a recent population-based study by Sariaslan, Larsson, D’Onofrio, Långström, and Lichtenstein (2014) found that the association between family income and violent criminality among registered inhabitants of Sweden was entirely accounted for by unobservable familial factors. The disparity in findings could be due to differences in measured outcomes (conduct problems/delinquency vs. violent criminality) or differences in geographic context (Sweden vs. United States). Future genetically informed research is needed to examine the effect of SES on variation in delinquency across context within and between countries. Using such designs will aid in verifying that contexts, like low SES urban environments, affect the development of delinquent behaviors and provide more precise effect size estimates for environmental risk factors.

The third major finding from the present study was that SES did not explain any of the variation in delinquency among youth from rural places. ACE model results revealed that including the shared environmental parameter did not improve model fit. Although speculative, one reason why SES may not have explained any of the variation in adolescent delinquent behaviors among rural youth could be because SES-related risk factors are not as concentrated in rural contexts as they are in urban contexts. It is possible that youth growing up in urban contexts may encounter a host of risk factors right outside their home, while youth growing up in rural contexts may not be exposed to the same type of concentrated risk and thus be more likely to select into certain social environments that provide opportunities to offend. This hypothesis would align with the results from the present study, which showed that additive genetic influences accounted for 52% of the population variation in delinquency among youth from rural contexts and nonshared environmental influences accounted for the remaining 48% of the population variation. Sources of additive genetic influence may include heritable traits such as anger, aggression, low self-control, and psychopathy. Sources of nonshared environmental influence may include differential association with peers, differential exposure to stress or strain, or differential exposure to opportunities to engage in delinquent behavior. Notably, these sources of influence have been purported by prominent criminological theories to operate as risk factors for delinquency regardless of context (Agnew, 1992; Gottfredson & Hirschi, 1990; Sutherland, 1947). Although this may only apply to the sibling sample analyzed in the present study, the results from the present study suggest that SES is not a significant contributor to individual differences in delinquency among rural youth once genetic liabilities are considered. As such, findings from the present study highlight the importance of using genetically informed models to better evaluate the effects of the environment on delinquent behavior across social contexts.

A few limitations of the present study should be noted. First, the sibling sample used in the present study was quite small. This was largely because NLS staff did not oversample for twin pairs when originally recruiting youth to participate in the NLSY97. Future research examining the effects of SES on adolescent delinquency should use larger twin and sibling samples to validate the results from the present study. Second, SES was measured by single item of family net income. Although a long line of research has used information about family income to measure SES, it will be important for future research to use multi-item geodemographic measures of SES that include information on neighborhood conditions, resident work status, home ownership, and public housing. Doing so may provide new information on the association between low SES and context-specific delinquency or confirm findings from the existing body of research. Third, social contexts were classified as urban or rural. Future work should look to expand on findings from previous research on urban and rural risk factors by examining risk factors prevalent in suburban contexts. Focusing on suburban social contexts will help to develop a cleaner picture of risk factors for adolescent delinquent behavior that are unique and common to urban, rural, and suburban settings.

Conclusion

Low SES has long been known as a risk factor for delinquent behavior in adolescence, but little research has examined the effect of SES on delinquency across geographic and social contexts while controlling for genetic liability. The present study, using longitudinal sibling data from a nationally representative sample of youth and genetically informed research designs, suggests that low SES is an environmental risk factor for adolescent delinquency in urban, but not rural contexts. Our results indicate that prevention efforts for delinquency in urban contexts should continue to focus on common environmental risk factors associated with low SES, but target a much wider range of familial risk factors when addressing adolescent delinquency in rural contexts.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Agnew

(1992). Foundation for a general strain theory of crime and delinquency. Criminology, 30, 47–88.

Agnew

Matthews

S. K.

Bucher

Welcher

A. N.

Keyes

(2008). Socioeconomic status, economic problems, and delinquency. Youth & Society, 40, 159–181.

Akaike

(1987). Factor analysis and AIC. Psychometrika, 52, 317–332.

Alaimo

Olson

C. M.

Frongillo

E. A.

Jr. Briefel

R. R.

(2001). Food insufficiency, family income, and health in US preschool and school-aged children. American Journal of Public Health, 91, 781–786.

Anderson

(2000). Code of the street: Decency, violence, and the moral life of the inner city. New York, NY: W.W. Norton.

Barnes

J. C.

Wright

J. P.

Boutwell

B. B.

Schwartz

J. A.

Connolly

E. J.

Nedelec

J. L.

Beaver

K. M.

(2014). Demonstrating the validity of twin research in criminology. Criminology, 52, 588–626.

Beaver

K. M.

Connolly

E. J.

(2013). Genetic and environmental influences on the development of childhood antisocial behavior: Current evidence and directions for future research. In Gibson

C. L.

Krohn

(Eds.), Handbook of life-course criminology (pp. 43–55). New York, NY: Springer.

Beaver

K. M.

Schwartz

J. A.

Nedelec

J. L.

Connolly

E. J.

Boutwell

B. B.

Barnes

J. C.

(2013). Intelligence is associated with criminal justice processing: Arrest through incarceration. Intelligence, 41, 277–288.

Biggar

R. W.

Jr. Forsyth

C. J.

Chen

Richard

T. A.

(2016). Protective factors for deviance: A comparison of rural and urban youth. Deviant Behavior, 37, 1380–1391.

10.

Blau

D. M.

(1999). The effect of income on child development. Review of Economics and Statistics, 81, 261–276.

11.

Boisvert

Wright

J. P.

Knopik

Vaske

(2012). Genetic and environmental overlap between low self-control and delinquency. Journal of Quantitative Criminology, 28, 477–507.

12.

Bouffard

L. A.

Muftic

L. R.

(2006). The “rural mystique”: Social disorganization and violence beyond urban communities. Western Criminology Review, 7, 56–66.

13.

Boutwell

B. B.

Nelson

E. J.

Emo

Vaughn

M. G.

Schootman

Rosenfeld

Lewis

(2016). The intersection of aggregate-level lead exposure and crime. Environmental Research, 148, 79–85.

14.

Bursik

R. J.

(1988). Social disorganization and theories of crime and delinquency: Problems and prospects. Criminology, 26, 519–552.

15.

Caspi

Taylor

Moffitt

T. E.

Plomin

(2000). Neighborhood deprivation affects children’s mental health: Environmental risks identified in a genetic design. Psychological Science, 11, 338–342.

16.

Center for Human Resource Research. (2003). NLSY97 user’s guide: A guide to the rounds 1-5 data: National Longitudinal Survey of Youth 1997. Columbus, OH: The Ohio State University.

17.

Conley

Rauscher

Dawes

Magnusson

P. K.

Siegal

M. L.

(2013). Heritability and the equal environments assumption: Evidence from multiple samples of misclassified twins. Behavior Genetics, 43, 415–426.

18.

Connolly

E. J.

Beaver

K. M.

(2014). Examining the genetic and environmental influences on self-control and delinquency: Results from a genetically informative analysis of sibling pairs. Journal of Interpersonal Violence, 29, 707–735.

19.

Connolly

E. J.

Beaver

K. M.

(2016). Considering the genetic and environmental overlap between bullying victimization, delinquency, and symptoms of depression/anxiety. Journal of Interpersonal Violence, 31, 1230–1256.

20.

Connolly

E. J.

Schwartz

J. A.

Nedelec

J. L.

Beaver

K. M.

Barnes

J. C.

(2015). Different slopes for different folks: Genetic influences on growth in delinquent peer association and delinquency during adolescence. Journal of Youth and Adolescence, 44, 1413–1427.

21.

Corak

(2013). Income inequality, equality of opportunity, and intergenerational mobility. The Journal of Economic Perspectives, 27, 79–102.

22.

Costello

E. J.

Keeler

G. P.

Angold

(2001). Poverty, race/ethnicity, and psychiatric disorder: A study of rural children. American Journal of Public Health, 91, 1494–1498.

23.

Devenish

Hooley

Mellor

(2017). The pathways between socioeconomic status and adolescent outcomes: A systematic review. American Journal of Community Psychology, 59, 219–238. doi:10.1002/ajcp.12115

24.

D’Onofrio

B. M.

Goodnight

J. A.

Van Hulle

C. A.

Rodgers

J. L.

Rathouz

P. J.

Waldman

I. D.

Lahey

B. B.

(2009). A quasi-experimental analysis of the association between family income and offspring conduct problems. Journal of Abnormal Child Psychology, 37, 415–429.

25.

Farrell

A. D.

Sullivan

T. N.

Esposito

L. E.

Meyer

A. L.

Valois

R. F.

(2005). A latent growth curve analysis of the structure of aggression, drug use, and delinquent behaviors and their interrelations over time in urban and rural adolescents. Journal of Research on Adolescence, 15, 179–204.

26.

Felson

(2014). What can we learn from twin studies? A comprehensive evaluation of the equal environments assumption. Social Science Research, 43, 184–199.

27.

Ferguson

C. J.

(2010). Genetic contributions to antisocial personality and behavior: A meta-analytic review from an evolutionary perspective. The Journal of Social Psychology, 150, 160–180.

28.

Glenn

N. D.

Hill

Jr. (1977). Rural–urban differences in attitudes and behavior in the United States. The Annals of the American Academy of Political and Social Science, 429, 36–50.

29.

Goodnight

J. A.

Lahey

B. B.

Van Hulle

C. A.

Rodgers

J. L.

Rathouz

P. J.

Waldman

I. D.

D’Onofrio

B. M.

(2012). A quasi-experimental analysis of the influence of neighborhood disadvantage on child and adolescent conduct problems. Journal of Abnormal Psychology, 121, 95–108.

30.

Gottfredson

M. R.

Hirschi

(1990). A general theory of crime. Stanford, CA: Stanford University Press.

31.

Hanna-Attisha

LaChance

Sadler

R. C.

Champney Schnepp

(2016). Elevated blood lead levels in children associated with the Flint drinking water crisis: A spatial analysis of risk and public health response. American Journal of Public Health, 106, 283–290.

32.

Hao

Matsueda

R. L.

(2006). Family dynamics through childhood: A sibling model of behavior problems. Social Science Research, 35, 500–524.

33.

Hauser

R. M.

(1994). Measuring socioeconomic status in studies of child development. Child Development, 65, 1541–1545.

34.

Hope

T. L.

Bierman

K. L.

(1998). Patterns of home and school behavior problems in rural and urban settings. Journal of School Psychology, 36, 45–58.

35.

Jackson

D. B.

(2016). The link between poor quality nutrition and childhood antisocial behavior: A genetically informative analysis. Journal of Criminal Justice, 44, 13–20.

36.

Jaggers

J. W.

Robison

S. B.

Rhodes

J. L.

Guan

Church

W. T.

(2016). Predicting adult criminality among Louisiana’s urban youth: Poverty, academic risk, and delinquency. Journal of the Society for Social Work and Research, 7, 89–116.

37.

Jiang

Sun

Marsiglia

F. F.

(2016). Rural and urban disparities in adolescent risky behaviors: A family social capital perspective. Journal of Community Psychology, 44, 1027–1039.

38.

Kendler

K. S.

Lönn

S. L.

Maes

H. H.

Sundquist

(2015). The etiologic role of genetic and environmental factors in criminal behavior as determined from full- and half-sibling pairs: an evaluation of the validity of the twin method. Psychological Medicine, 45, 1873–1880.

39.

Leventhal

Brooks-Gunn

(2000). The neighborhoods they live in: The effects of neighborhood residence on child and adolescent outcomes. Psychological Bulletin, 126, 309–337.

40.

Linver

M. R.

Brooks-Gunn

Kohen

D. E.

(2002). Family processes as pathways from income to young children’s development. Developmental Psychology, 38, 719–734.

41.

Lowe

G. D.

Peek

C. W.

(1974). Location and lifestyle: The comparative explanatory ability of urbanism and rurality. Rural Sociology, 39, 392.

42.

Lyerly

R. R.

Skipper

J. K.

(1981). Differential rates of rural–urban delinquency: A social control approach. Criminology, 19, 385–399.

43.

Miles

D. R.

Carey

(1997). Genetic and environmental architecture of human aggression. Journal of Personality and Social Psychology, 72, 207–217.

44.

Muthén

L. K.

Muthén

B. O.

(1998–2012). MPlus user’s guide. Los Angeles, CA: Author.

45.

Piotrowska

P. J.

Stride

C. B.

Croft

S. E.

Rowe

(2015). Socioeconomic status and antisocial behaviour among children and adolescents: A systematic review and meta-analysis. Clinical Psychology Review, 35, 47–55.

46.

Pyrooz

D. C.

Turanovic

J. J.

Decker

S. H.

(2016). Taking stock of the relationship between gang membership and offending: A meta-analysis. Criminal Justice and Behavior, 43, 365–397.

47.

Raftery

A. E.

(1995). Bayesian model selection in social research. Sociological Methodology, 25, 111–163.

48.

Raine

(2002). Biosocial studies of antisocial and violent behavior in children and adults: A review. Journal of Abnormal Child Psychology, 30, 311–326.

49.

Reiss

(2013). Socioeconomic inequalities and mental health problems in children and adolescents: A systematic review. Social Science and Medicine, 90, 24–32.

50.

Rhee

S. H.

Waldman

I. D.

(2002). Genetic and environmental influences on antisocial behavior: A meta-analysis of twin and adoption studies. Psychological Bulletin, 128, 490–529.

51.

Rowe

D. C.

(1994). The limits of family influence: Genes, experience, and behavior. New York, NY: Guilford Press.

52.

Rowe

D. C.

Rodgers

J. L.

(1997). Poverty and behavior: Are environmental measures nature and nurture? Developmental Review, 17, 358–375.

53.

Rutter

Giller

Hagell

(1998). Antisocial behavior by young people: A major new review. Cambridge, UK: Cambridge University Press.

54.

Sampson

R. J.

Laub

J. H.

(1994). Urban poverty and the family context of delinquency: A new look at structure and process in a classic study. Child Development, 65, 523–540.

55.

Sampson

R. J.

Morenoff

J. D.

Earls

(1999). Beyond social capital: Spatial dynamics of collective efficacy for children. American Sociological Review, 64, 633–660.

56.

Sariaslan

Larsson

D’Onofrio

Långström

Lichtenstein

(2014). Childhood family income, adolescent violent criminality and substance misuse: Quasi-experimental total population study. The British Journal of Psychiatry, 205, 286–290.

57.

Satorra

Bentler

P. M.

(2001). A scaled difference chi-square test statistic for moment structure analysis. Psychometrika, 66, 507–514.

58.

Schwartz

J. A.

(2015). Socioeconomic status as a moderator of the genetic and shared environmental influence on verbal IQ: A multilevel behavioral genetic approach. Intelligence, 52, 80–89.

59.

Sutherland

E. H.

(1947). Principles of criminology: A sociological theory of criminal behavior. Philadelphia, PA: Lippincott.

60.

Thapar

McGuffin

(1996). Genetic influences on life events in childhood. Psychological Medicine, 26, 813–820.

61.

Thomas

K. J.

(2015). Delinquent peer influence on offending versatility: Can peers promote specialized delinquency? Criminology, 53, 280–308.

62.

U.S. Census Bureau. (1994). County and city data book. Washington, DC: U.S. Dept. of Commerce, U.S. Census Bureau.

63.

Wright

J. P.

Barnes

J. C.

Boutwell

B. B.

Schwartz

J. A.

Connolly

E. J.

Nedelec

J. L.

Beaver

K. M.

(2015). Mathematical proof is not minutiae and irreducible complexity is not a theory: A final response to Burt and Simons and a call to criminologists. Criminology, 53, 113–120.

64.

Wright

J. P.

Dietrich

K. N.

Ris

M. D.

Hornung

R. W.

Wessel

S. D.

Lanphear

B. P.

,… Rae

M. N

. (2008). Association of prenatal and childhood blood lead concentrations with criminal arrests in early adulthood. PLoS Medicine, 5, 0732–0740.

65.

J. J.

(2016). First delinquent behavior among pre-pubescent children: A development paradigm alongside family SES. International Journal of Offender Therapy and Comparative Criminology, 60, 1609–1622.