This paper analyzes the extent of risk-sharing among stockholders and nonstockholders. To evaluate the empirical importance of market incompleteness, it is essential to determine whether idiosyncratic shocks are important for the wealthy who have access to better insurance opportunities, but also face different risks, than the average household. We study a model where each period households decide whether to participate in the stock market by paying a fixed cost. Due to this endogenous entry decision, the testable implications of perfect risk-sharing take the form of a sample selection model, which we estimate using a semiparametric GMM estimator proposed by Kyriazidou (2001). Using data from PSID, we strongly reject perfect risk-sharing among stockholders, but perhaps surprisingly, do not find evidence against it among nonstockholders. This result appears to be robust to several extensions. This finding suggests further focus on risk factors that primarily affect the wealthy, such as entrepreneurial income risk.