Delinquent Peer Group Formation: Evidence of a Gene × Environment Correlation
Abstract
Emergence of biosocial explanations for adolescent development. Variants of specific genes may influence youths to seek out or associate with antisocial peers.
Study uses genotypic data (N = 1,816) from Add Health to test whether the dopamine transporter gene DAT1 (particularly the 10R allele) is associated with delinquent peer affiliation.
Key finding: The 10R allele of DAT1 is associated with higher self-reported delinquent peer affiliation for male adolescents from high-risk environments, even after controlling for delinquent involvement, self-control, and drug/alcohol use. β-range for this effect: eta ext{ in } [0.13, 0.14].
Implication: Supports a biosocial, gene × environment correlation (rGE) framework for adolescent peer selection and antisocial development.
Key Concepts and Definitions
Homophily (Birds of a feather): tendency to associate with similar others (talents, beliefs, behaviors).
Evidence: associations by race/ethnicity, age/education, etc. (Giordano 2003; McPherson et al. 2001; others).
Delinquent peers: peers engaged in delinquency, drug use, and related antisocial behaviors; strong predictor of antisocial conduct.
Warr (2002) and Warr (1998, 2002) emphasize persistent link between delinquent peers and misconduct.
Gene × Environment Correlation (rGE): when an individual’s genotype is partially responsible for shaping the environment they experience.
Types of rGE:
Passive rGE: parents pass on genes and provide environment correlated with those genes.
Evocative rGE: an individual’s genetically influenced traits evoke specific responses from the environment.
Active rGE: individuals select environments compatible with their genotypes.
References: Caspi & Moffitt (1995); Scarr & McCartney (1983); Rutter (2006); Walsh (2002).
DAT1 (Dopamine transporter gene): genetic polymorphism with 40-base pair VNTR in the 3'-UTR, typically 9R or 10R alleles.
The 10R allele is considered the risk allele, associated with ADHD, externalizing problems, and related maladaptations (Comings et al. 2001; Rowe et al. 1998; Gill et al. 1997).
Genotype coding in Add Health: number of 10R alleles, values 0, 1, or 2 (0 = no 10R alleles, 1 = one, 2 = two).
Add Health (NSDUH Add Health): nationally representative longitudinal study of adolescents in grades 7–12 with genetic data available at Wave III.
Final analytic sample for this study: N = 1{,}816 (after removing one twin from each MZ pair and after data cleaning).
Family risk composites: combining maternal attachment, maternal involvement, and maternal disengagement into a single family risk factor; dichotomized into low (0) vs high (1) risk groups.
Delinquent peers measure: three-item scale at Wave 1 assessing friends’ smoking, drinking, and marijuana use; responses summed and standardized (α = .76).
Control variables: age, paternal criminal history, delinquency, low self-control, and drug/alcohol use.
Theoretical Background: Why rGE Matters for Delinquent Peers
Traditional debate: social causation (peers cause delinquent behavior) vs self-selection (individual differences lead to selection of delinquent peers).
Walsh (2002) highlights that people seek environments compatible with their genetic dispositions; this motivates testing for rGE in delinquent peer networks.
rGE provides a mechanism by which genotype may influence exposure to antisocial environments, potentially explaining part of the association between genes and delinquent outcomes.
Prior genetic research on peer context (Iervolino et al. 2002; Cleveland et al. 2005; Kendler et al. 2007) suggests genetic contributions to exposure to antisocial peers, though findings are mixed across samples and designs.
Present Study: DAT1 and Delinquent Peer Selection
Research question: Is there a gene × environment correlation between the DAT1 gene and assortative formation into delinquent peer networks?
Data source: Add Health Wave III genetic subsample (N = 1,816 after data cleaning).
Hypothesis: DAT1 variation, particularly the 10R allele, would be associated with delinquent peer affiliation, especially among males in high-risk family environments.
Rationale: Active rGE predicts that youths with certain genotypes may actively select into environments that maximize gene expression (e.g., antisocial peer groups).
Context: Previous research shows rGEs in other domains (ADHD, externalizing problems, substance use) and suggests gene × environment interplay in social contexts.
Methods
Sample and Genotyping
Add Health Wave III: includes genetic data; final analytic sample: N = 1{,}816.
Genotyped DAT1 locus focused on 40-bp VNTR with alleles 9R and 10R.
Allele distribution in this sample:
0 copies of 10R: 5.3 ext{%},
1 copy: 35.1 ext{%},
2 copies: 59.6 ext{%}.
Hardy–Weinberg equilibrium: χ²(1, N = 1,816) = 0.0239, p > .05.
DAT1 variable recoded as ext{DAT1}_{10R} o ext{0,1,2}, representing the number of 10R alleles.
Coding: 0 = no 10R alleles, 1 = one 10R allele, 2 = two 10R alleles.
Measures
Delinquent peers (Wave 1): three items indicating whether close friends smoke ≥1 cigarette/day, drink alcohol once a month, or smoke marijuana ≥ once a month.
Scale: sum of three items; transformed to a standardized scale (higher values = more involvement with antisocial friends).
Reliability: ext{α} = .76.
DAT1: as described above (0, 1, 2 copies of 10R).
Family risk (Wave 1): three scales combined via factor analysis, then dichotomized at the mean into low-risk (0) vs high-risk (1) family environments.
Maternal attachment: two items; higher scores indicate less attachment after recoding (α = .64).
Maternal involvement: 10 activities with mother; coded 1 if engaged in activity, 0 otherwise (α = .55).
Maternal disengagement: five items indicating perceived maternal coldness/withdrawal (α = .84).
Result: a single factor captures global family risk; dichotomized as described.
Control variables:
Age (years at Wave 1).
Criminal father: whether the respondent’s biological father was ever incarcerated (dichotomous).
Delinquency: Wave 1 delinquency scale (15 items; 0,1,2,3 coding; α = .78).
Low self-control: five-item scale (α = .63).
Drug and alcohol use: composite of alcohol use and marijuana use in the past 12 months (Wave 1).
Analytic Strategy
Primary approach: Ordinary Least Squares (OLS) regression to predict delinquent peers from DAT1 while controlling for covariates.
Stratified analyses by gender and by race (non-Hispanic White and Black) to assess potential differential DAT1 effects.
Subgroup analyses by family risk level (low vs high) to test for moderation by environment.
Additional spuriousness checks: include delinquency, low self-control, and drug/alcohol use to test whether the DAT1 effect persists when controlling for related antisocial traits and tendencies.
Note on sample size: high-risk male subsample used for some robustness checks to preserve degrees of freedom (n ≈ 341 in Table 3 analyses).
Results: Direct Effects of DAT1 on Delinquent Peers
Overall model characteristics (Table 1, full sample and subgroups):
Full sample (N = 1,816):
DAT1: b = 0.05, ext{ } eta = 0.03, ext{ } SE = 0.04, ext{ } R^2 = 0.09, not significant overall.
Male adolescents (n = 849):
DAT1: b = 0.17, ext{ } eta = 0.10^{*}, ext{ } SE = 0.06, ext{ } R^2 = 0.10.
Age: b = 0.18, ext{ } eta = 0.29^{*}, ext{ } SE = 0.01.
Criminal father: b = 0.18, ext{ } eta = 0.07^{*}, ext{ } SE = 0.07.
Note: asterisk indicates p < .05.
Female adolescents (n = 967):
DAT1: b = -0.06, ext{ } eta = -0.03, ext{ } SE = 0.04, ext{ } R^2 = 0.09.
White adolescents (n = 1,464):
DAT1: b = 0.05, ext{ } eta = 0.03, ext{ } SE = 0.04, ext{ } R^2 = 0.09.
Black adolescents (n = 352):
DAT1: b = 0.12, ext{ } eta = 0.07, ext{ } SE = 0.09, ext{ } R^2 = 0.07.$
Key takeaway from Table 1: DAT1 is predictive of delinquent peers only for the subsample of male adolescents; for females and for most race groups, the direct DAT1 effect is non-significant in the full models.
Gender-specific interpretation: The DAT1 effect on delinquent peers is present for males (β ≈ .10; p < .05) but not for females.
Mediation/Moderation: DAT1 by Family Risk (Table 2)
Researchers tested whether the DAT1 effect is conditioned by family risk level (low vs high).
Findings for male adolescents (White and Black; combined) show:
Low-risk families (n for White and Black combined): age is the main predictor; DAT1 effect not significant.
High-risk families (n for White and Black combined): DAT1 is significantly related to delinquent peers across the models, indicating a robust gene × environment interaction for DA transporters in adverse family contexts.
Overall interpretation: For male adolescents, the DAT1 × family-risk interaction suggests that the genetic propensity (10R allele) to affiliate with delinquent peers is amplified in high-risk environments and attenuated or null in low-risk environments.
Robustness Checks: Spuriousness Tests (Table 3)
High-risk male adolescents (N ≈ 341) tested to ensure DAT1’s association with delinquent peers is not due to confounding by delinquency, self-control, or drug/alcohol use.
Model specifications incorporated additional controls (Delinquency, Low Self-Control, Drug/Alcohol use) in various combinations (Models 1–4):
DAT1 coefficients remain significant and positive across all four models when predicting delinquent peers in high-risk males.
Coefficients for DAT1 ranged approximately b ext{ around } 0.23 ext{ to } 0.25, with standardized effects eta ext{ around } 0.13 ext{ to } 0.14, and SEs around 0.08 ext{ to } 0.09.
Model comparison indicates incremental explanation of variance, with total R^2 values spanning roughly 0.23 to 0.32 across models (Model 4 R^2 ext{ around } 0.32).
Implication: The DAT1 effect on delinquent peers in high-risk males persists even after controlling for direct delinquency and other risk factors, suggesting a genuine rGE signal rather than a spurious correlation due to correlated behaviors.
Tables: Core Numerical Findings (Summary)
Table 1: Direct Effects of DAT1 on Delinquent Peers (N = 1,816)
Full sample: $b = 0.05$, $eta = 0.03$, $SE = 0.04$, $R^2 = 0.09$;
Male: $b = 0.17$, $eta = 0.10^{*}$, $SE = 0.06$, $R^2 = 0.10$;
Female: $b = -0.06$, $eta = -0.03$, $SE = 0.04$, $R^2 = 0.09$;
White: $b = 0.05$, $eta = 0.03$, $SE = 0.04$, $R^2 = 0.09$;
Black: $b = 0.12$, $eta = 0.07$, $SE = 0.09$, $R^2 = 0.07$.
Table 2: DAT1 Effects by Family Risk (Male Adolescents, White/Black; N = 814 for low-risk/ high-risk groups combined per subgroup)
Low-risk: DAT1 effect not statistically significant; age is the main predictor; other covariates non-dominant.
High-risk: DAT1 significantly related to delinquent peers; effect persists across models; pattern suggests environment moderates genetic effect.
Table 3: Spuriousness Tests (High-Risk Male Adolescents; N ≈ 341)
Model 1: $DAT1
ightarrow ext{delinquent peers}$; $b = 0.23$, $eta = 0.13$, $SE = 0.09$;Model 2: adding Age and Criminal father; $b = 0.25$, $eta = 0.14$, $SE = 0.09$;
Model 3: adding Delinquency; $b = 0.22$, $eta = 0.13$, $SE = 0.13$;
Model 4: adding Delinquency + Low self-control + Drug/Alcohol; $b = 0.21$, $eta = 0.13$, $SE = 0.08$; $R^2$ values: $0.23$, $0.10$, $0.26$, $0.32$ respectively.
Overall pattern: In high-risk males, DAT1’s association with delinquent peers is robust to multiple controls and remains statistically significant in all tested specifications.
Discussion and Interpretation
Main finding: An rGE exists between the DAT1 gene and delinquent peer affiliation, but it is not uniform across all groups.
Specifically, male adolescents with more DAT1 risk alleles (10R) are more likely to affiliate with delinquent peers, but primarily within high-family-risk environments.
In low-risk family contexts, the DAT1 effect on delinquent peer affiliation is not statistically significant.
Female adolescents show no significant DAT1 association with delinquent peers in any environ- mental context studied here.
The results support an active rGE mechanism: genetic propensities may lead some male youths to seek out delinquent peer groups, especially when the family environment provides fewer protective resources.
Theoretical implications:
Confirms the idea that genotype and environment are interwoven in complex ways, where genes influence the environments to which individuals are exposed, and environments may activate or mute genetic propensities.
Extends prior findings on MAOA and environmental risk to DAT1 and antisocial peer networks.
Explanations for gender differences:
Parental monitoring may be more effective for sons in low-risk families, potentially attenuating gene-driven selection into delinquent peers for males in low-risk environments.
In high-risk families, the 10R DAT1 allele may interact with contextual stressors to promote selection into antisocial peer groups.
Limitations and cautions:
The DAT1 effect did not generalize to females; results are strongest for male adolescents in high-risk contexts.
Delinquent peers measure at Wave 1 captured only drug-using behaviors of friends, not more serious violence or broader antisocial activities.
Genotyped subsample (1,816) is substantial but not fully representative; generalizability is limited.
Observational design cannot prove causality; however, the robust associations after controlling for multiple covariates strengthen the rGE interpretation.
Relation to prior work:
Aligns with Iervolino et al. (2002) and Kendler et al. (2007) suggesting genetic contributions to exposure to antisocial peers, while highlighting environmental moderation (family risk) as crucial.
Connects to a broader biosocial literature that emphasizes gene–environment interplay in antisocial and externalizing outcomes (Caspi et al. 2002; Moffitt 2005; Walsh 2002).
Implications, Applications, and Future Directions
Biosocial approach: the study supports integrating genetic data into criminological and developmental research to better understand how youths select or are exposed to antisocial environments.
Policy implications: interventions that bolster family environments (attachment, involvement, and reducing disengagement) may dampen genetic propensities to seek delinquent peers, particularly for at-risk males.
Future research directions:
Replication in independent samples with more diverse measures of peer delinquency (including violence) and objective behavioral outcomes.
Exploration of additional genes and polygenic risk scores to capture a broader genetic architecture of peer selection tendencies.
Longitudinal analyses to examine how rGE evolves across adolescence and into early adulthood.
Ethical and Practical Considerations
Use of genetic data in criminology requires careful handling to avoid determinism and stigma; emphasis should be on understanding mechanisms to inform prevention and intervention.
Findings should not be used to label individuals as “genetically predisposed criminals” but to recognize complex gene–environment interplay and identify protective factors.
Data privacy and consent for genetic data in minors are critical; analyses should ensure robust protection of participant information.
Connections to Foundational Principles
Differential association and social learning theory underpin the interpretation that peers shape delinquent behavior, while gene–environment interplay provides a mechanism for why some youths opt into or are exposed to these networks.
Self-control theory and general strain theory are consistent with the idea that family risk and environmental stressors interact with genetic predispositions to influence antisocial outcomes.
The findings reinforce Scarr & McCartney’s (1983) genotype–environment effects framework by demonstrating a measurable rGE in a real-world, nationally representative sample.
Key Formulas and Notation (LaTeX)
Allele counts for DAT1:
ext{DAT1}_{10R} o ext{number of 10R alleles} \ ext{Possible values: } 0,1,2
Distribution in Add Health sample: P(0)=0.053,\, P(1)=0.351,\, P(2)=0.596
Delinquent peers measure:
DP = x1 + x2 + x3, ext{ where } xi ext{ indicates peer behavior (0/1)}
Family risk construct:
Let FR^ ext{econ} o egin{cases} 0, & ext{low-risk} \ 1, & ext{high-risk} \ ext{(composite score from maternal attachment, involvement, disengagement)} \ ext{split at mean} \ ext{dichotomized} \ ext{(FR ∈ {0,1})} \ ext{Higher FR indicates higher family risk} \
Regression models (example):
Base model: DP = eta0 + eta1 ext{DAT1}{10R} + eta2 ext{Age} + eta_3 ext{CriminalFather} + oldsymbol{
u}
Subgroup model for spuriousness check:
DP = eta0 + eta1 ext{DAT1}{10R} + eta2 ext{Age} + eta3 ext{CriminalFather} + eta4 ext{Delinquency} + eta5 ext{LowSelfControl} + eta6 ext{DrugAlcoholUse} + ext{error} $$
References (Selected from Article)
Iervolino, C. et al. (2002). Genetic and environmental influences in adolescent peer socialization. Child Development.
Cleveland, H. H., Wiebe, R. P., & Rowe, D. C. (2005). Sources of exposure to smoking and drinking friends among adolescents: A behavioral-genetic evaluation. The Journal of Genetic Psychology.
Kendler, K. S., et al. (2007). Creating a social world: A developmental twin study of peer-group deviance. Archives of General Psychiatry.
Caspi, A., & Moffitt, T. E. (1995, 2002). Gene–environment interplay and antisocial behavior literature cited in discussion.
Scarr, S., & McCartney, K. (1983). How people make their own environments: A theory of genotype–environment effects. Child Development.
Walsh, A. (2002). Essay review. Companions in crime: A biosocial perspective. Human Nature Review.
Udry, J. R. (1998, 2003). Add Health design and data files.
Warr, M. (1996, 1998, 2002). Works on delinquent peers and social structure in crime.
Moffitt, T. E. (2005). The new look of behavioral genetics in developmental psychopathology.
Caspi, A., et al. (2002). Role of genotype in the cycle of violence in maltreated children. Science.
Summary Takeaways
DAT1 10R allele is linked to higher delinquent peer affiliation, but primarily for male adolescents in high-risk family environments.
The effect persists after accounting for delinquency, self-control, and drug/alcohol use, supporting a genuine gene–environment correlation rather than a spurious association.
The results highlight the importance of examining interactions between genes and environment, and they demonstrate that genetic influences on social experiences can be conditional on context (family risk) and gender.
The study advances a biosocial framework in criminology and suggests avenues for prevention that bolster family environments to potentially mitigate genetic propensities toward delinquent peer networks.