The longitudinal trajectory of delinquency – A genetically informed study


Ksinan, Albert J. & Vazsonyi, Alexander T. (2018). The longitudinal trajectory of delinquency – A genetically informed study. Biennial Meeting of the Society for Research on Adolescence. Minneapolis, MN.


Recent studies have shown support for considerable heritability of delinquency (Barnes, Wright, et al., 2014), indicating approximately 50% of variance in antisocial or criminal behaviors might be explained by genetic influences (Barnes, Boutwell, et al., 2014). Nevertheless, there continues to be a paucity of developmental studies focused on delinquency. The current study sought to assess relative genetic and environmental contributions to the level of as well as the development of delinquent behavior from adolescence to adulthood. Data from four waves of the Add Health Project were used. Due to well-known sex differences in rates of depressive symptoms, we decided to reduce the potential complexity by focusing only on same-sex pairs of twins, resulting in the final twin sample of 285 monozygotic (MZ) twins and 246 dizygotic (DZ) twins. Delinquency was measured by a 6-item self-report measure asking about the frequency of delinquent behaviors in the past 12 months (e.g., fighting, selling drugs, stealing, or damaging property). First, ACE models were tested separately for each wave to estimate the relative contributions of additive genetic (A), shared environmental (C), and nonshared environmental (E) variance. Next, a biometric quadratic latent growth model (LGM) using the four waves of data was tested (McArdle & Plassman, 2009; see Figure 1). This model decomposes variance of the intercept, slope, and quadratic terms into the A, C, E parts, enabling to asses genetic and environmental influence on levels as well as development of delinquency. ACE model results for each wave showed minimal variance explained by heritability at Waves 1 and 2, with the shared environment explaining 30%, and 33% respectively. However, at Waves 3 and 4, the heritability coefficient increased (38% and 44% respectively), while no effect of the shared environment was found. The biometric LGM confirmed that there was a trivial and non-significant contribution of heritability (h2) on the intercept (0.1%), whereas the shared environment (c2) explained half of the variance in intercept, c2= .53, BcCI [.38, .68], followed by the nonshared environment, e2 = .47, BcCI [.31, .62]. However, for the variance in slope, h2 explained 23.9% of variance, BcCI [.02, .63], while c2 remained significant, c2= .30, BcCI [.09, .58], as well as e2= .46, BcCI [.20, .67]. Finally, the variance of quadratic had a larger heritable influence, h2= .41, BcCI [.10, .80], along with c2= .23, BcCI [.02, .51] and e2= .40 BcCI [.10, .66] effects. The present results underscore dynamic changes in variance estimates of heritability and environmental effects of delinquency from adolescence to adulthood. Heritability estimates at the first two timepoints were practically non-existent, while they became more prominent at Waves 3 and 4. This was also confirmed in the biometric model, showing that initial levels of delinquency were not affected by heritability. However, the slope and the quadratic terms showed a strong genetic component, suggesting that delinquency might be perhaps normative and primarily influenced by the environment (e.g., peer affiliation) during adolescence, whereas developmental changes in delinquency into adulthood showed a significant genetic contribution.


Reference Type

Conference proceeding

Book Title

Biennial Meeting of the Society for Research on Adolescence


Ksinan, Albert J.
Vazsonyi, Alexander T.

Year Published


City of Publication

Minneapolis, MN

Reference ID