Selection Bias in Mutual Fund Fire Sales

Elizabeth A. Berger

Liquidity trading following mutual fund outflows creates a potentially powerful empirical setting in which stock price variation is unrelated to changes in firm fundamentals. Instrumental variables (IV) drawn from this setting impose an additional assumption that managers sell firms in proportion to portfolio weights. I show that this assumption causes selection bias in these IVs. It misallocates large price impacts to poorly performing, illiquid firms with lower growth — firms that managers systematically avoid selling. Simulations show that selection bias doubles the magnitude of regression coefficients and precludes potential fixes. Numerous recent studies exploiting these IVs should be re-evaluated.