We estimate quarterly return series from March 1984 through December 1989 for 10 classes of thrift assets using the statistical cost‐accounting methodology of Hester and Zoellner (1966). We then use these return series to estimate mean‐variance efficient frontiers for all thrifts, for thrifts that were well capitalized two years earlier and for thrifts that were insolvent two years earlier. Our results show that neither the asset restrictions existing before nor those in effect after passage of the Financial Institutions Reform, Recovery and Enforcement Act of 1989 would have prevented thrifts from reaching most of the portfolios along the efficient frontier. The actual portfolio chosen by well‐capitalized thrifts is close to the estimated efficient frontier, while the actual portfolio chosen by insolvent thrifts is located far from the frontier in the high‐risk end of investment space. These findings, coupled with the high proportion of nontraditional assets in the actual portfolio chosen by insolvent thrifts, support the hypothesis that moral hazard induced thrifts to take on investments that were excessively risky from the deposit insurer's point