Effects of Family Composition on Investment in Human Capital : Estimation Methods

Abstract

The effect of sibling sex composition on parental investment in human capital is typically estimated by fixing sibsize in regression models. But the estimated?parameters are ill-defined; sibsize cannot be fixed because it changes with sibling gender. We well-define the causal effects by counterfactual notions whereby potential sibsize could be fixed although observed sibsize couldn't. We propose a new method to separate direct from indirect effects using stronger exogeneity conditions. The ?effect of sibling sex composition, if any, is driven almost entirely by the indirect ??effect via changed sibsize, contrary to what the conventional estimates would have suggested.