Does Stockholding Provide Perfect Risk Sharing?

This paper analyzes the extent of risk-sharing among stockholders. To provide a benchmark, we ask if stockholders are able to share risk more effectively than non-stockholders, where the latter serves as a control group. We study a dynamic structural model where each period households compare the benefits from stockholding with a per-period trading cost and decide whether to be in the stock market. Due to the endogenous entrym decision, the testable implications of perfect risk-sharing take the form of a sample selection model. To eliminate the selection bias, we implement a semiparametric estimation method (kernel-weighted GMM) recently proposed by Kyriazidou (2001). Using data from the Panel Study of Income Dynamics on U.S. households we find a surprising result: We strongly reject perfect risk-sharing for stockholders, but find no evidence against it among non-stockholders. Moreover, we also strongly reject risk-sharing for the whole population consistent with existing literature. We offer two explanations. First, if financial markets are incomplete, holding stocks may expose owners to extra risks more than it insures existing risks. Second, stockholders also own a large part of the entrepreneurial wealth which is largely uninsurable. Thus, stockholders not only have more assets for risk sharing but also have more risks to insure. These results suggest that these two groups face different economic environments and solve quite different optimization problems. Finally, we find significant heterogeneity in the risk aversion and leisure elasticities between the two groups. We conclude that the distinction between stockholders and non-stockholders is crucial in exploring many questions in macroeconomics and finance.