Simple ordinary least squares estimates indicate that absent fathers boost probabilities of adolescent criminal behavior by 16–38%, but those numbers likely are biased by unobserved heterogeneity. This paper first presents an economic model explaining that unobserved heterogeneity. Then turning to empirics, fixed effects, which attempt to address that bias, suggest that absent fathers reduce certain types of adolescent crime, while lagged-dependent variable models suggest the opposite. Those conflicting conclusions are resolved by an approach that combines those two estimators using an orthogonal reparameterization approach, with model parameters calculated using a Bayesian algorithm. The main finding is that absent fathers do not appear to directly affect adolescent criminal activity. Rather, families with absent fathers possess traits that appear to correlate with increased adolescent criminal behaviors.
