This study examines the differential effect of each of the ten largest bank failures on shareholders' wealth of non-failed banks over the period from 1973 through 1984. It examines how contagion and information effects of major bank failures have changed over time. FDIC policy for settling failures has important implications for system stability, and has changed over time. This study's purpose is to provide empirical evidence on the effects of FDIC policy. The FDIC's handling of the Penn Square failure signaled a policy shift and offers a unique opportunity to examine changes in market reactions to large bank failures. The literature on the capital market effects of major bank failures provides limited evidence on the impact of bank failures and related FDIC policy. Most fail to discriminate between contagion and information effects, and conduct analysis on one (or a few) bank failure(s) in the mid-1970s using traditional event study methodology. This study considers multivariate regression (MVRM) an appropriate methodology for bank failures which are likely to have simultaneous impact on non-failed banks. MVRM, which accounts for contemporaneous cross-sectional dependence of residuals, has three advantages over standard residual analysis: no "event clustering" problem, multiple hypotheses tests, and computational efficiency. This study uses daily stock-return data for fifty-one non-failed commercial banks. For each bank failure, the non-failed banksare grouped into three portfolios: "information-related," "large," and "small." The impact on each portfolio is tested for an average effect and joint hypotheses on excess return. This study offers evidence on no contagion effects and lack of information effects before Penn Square, strong information effects since Penn Square, contagion effects in post-Penn Square failures, and capital market discipline on large banks since Penn Square. There has been a change in the nature of the impact of bank failures since Penn Square.