FACULTÉ & RECHERCHE

We examine whether the managers of equity mutual funds exhibit reinforcement learning, investing more heavily in firms in which they previously experienced higher returns. The results reliably support this hypothesis. Experienced returns are related to managers' re-balancing decisions in response to flows. Experienced returns do not play a role for index-tracking funds. When new managers come in a fund, their experiences with stocks in their old funds, influence the investments in these stocks by their new funds. Funds managed by managers who rely more on reinforcement learning, earn lower returns. Experienced returns, when aggregated across managers for each stock, predict persistent lower stock returns. Overall, our evidence suggests that reinforcement learning plays a role in the stock-specific return trades of portfolio managers, with important implications for fund performance and asset prices.