Does Stockholding Provide Perfect Risk Sharing?

This paper addresses two questions: One, does stockholding provide perfect risk insurance? Two, do stockholders have significantly different preferences than non-stockholders? We investigate these questions in the context of a dynamic structural model in which households maximize their lifetime utility under price and wage uncertainty, and face complete markets only if they hold stocks. Preferences are parameterized allowing for heterogeneity in stockholders' and non-stockholders' risk aversion and leisure elasticity parameters. Due to self-selection by stockholders the model cannot be estimated consistently using standard methods. So, we implement the (bias-correcting) kernel-weighted GMM procedure proposed by Kyriazidou (1999), which does not require strong distributional assumptions about the error terms. We find that, contrary to the common belief, it is more difficult to reject perfect risk sharing for non-stockholders than for stockholders. The parameter estimates also reveal that stockholders are close to being risk neutral and that non-stockholders are significantly more risk averse. Finally, we find evidence indicating that each group experiences a different aggregate shock level, possibly due to the additional uncertainty transmitted to stockholders by financial securities that non-stockholders do not hold. From a substantive viewpoint, these results suggest that the distinction between stockholders and non-stockholders is crucial in exploring many questions regarding asset markets. From a methodological viewpoint, the bias correction procedure has a dramatic effect on estimates and test statistics in our model.