It has been argued that rare economic disasters can explain most asset pricing puzzles. If this is the case, perceived risk associated with a disaster in stock markets should be revealed in household portfolios. That is, the framework that solves these pricing puzzles should also generate quantities that are consistent with the observed ones. This paper estimates the perceived risk of disasters (both probability and expected size) that is consistent with observed portfolios and consumption growth between 1983 and 2004 in the United States. I find that the portfolio choices of households that have less than a college degree can be partially explained by expectations of stock market disasters only if one allows for a large probability of labor income loss at the same time. Such disaster expectations, however, are not revealed in the portfolios of educated and wealthier households: simple per-period participation costs of the stock market coupled with preference heterogeneity explain their participation and investment patterns.