Differential Susceptibility of Peer Influences: Gene-Environment Interactions and Gene-Environment Correlations

This paper examines potential gene-environment interactions in responses to peer influences on tobacco use. Specifications found in the literature that link own use to school-level tobacco use suggest widespread interactive effects, where individuals with the short/short 5-HTT genetic variant have the largest responsiveness to peer smoking. However, I show that individuals are sorted into schools in ways that suggest important gene-environment correlations may confound these findings. Using an across-cohort, within school strategy to separate school level effects (including school selection bias) and grade-level peer effects, I find evidence of reversals of the baseline specifications, so that the results suggest that individuals with the long/long 5-HTT variant are most susceptible to peer influence, increasing the likelihood of smoking by 3 percentage points per 10% increase in peer smoking. These results are consistent with a broader concern that many gene-environment models may fail to fully account for gene-environment correlation.