I develop an analytically tractable dynamic asset pricing model to study expected returns of financially distressed firms in the presence of learning about firm fundamentals and endogenous information acquisition by active investors. The model reveals that learning and information acquisition critically affect lowfrequency risk exposures close to default and can, counter to standard models, rationalize low and even negative expected return premia for firms with high default risk. Similar to Schumpeter’s (1934) argument that recessions have a positive, cleansing effect on the economy, the model reveals that equity holders naturally benefit from an increased speed of learning about truly insolvent firms in downturns, which positively affects the value of their default option in these times. Equity holders’ option value is similarly enhanced by the ability to partially freeride on active investors’ information acquisition. Learning thus dynamically affects distressed firms’ exposures to business-cycle frequency risks and can rationalize striking, momentum-type dynamics in risk premia.