The (Bad?) Performance of Mutual Fund Investors

Abstract

This paper examines monthly data on aggregate U.S. equity mutual funds, and provides a new look at the performance of mutual fund investors. In particular we examine the relationship between investors' aggregate net flows into and out of equity funds and the returns of the funds in subsequent periods. As an indication, we compare the average monthly return on the funds with the ratio of the average dollar return to average monthly value of assets in the years 1984-2003. Although in these years the monthly average return was relatively high, 1.07%, the ratio of average dollar return to average monthly value of assets was relatively low, only 0.56%. We develop a formal statistical test, based on bootstrapping of accumulated monthly returns in a 20 year period, and significantly reject the null hypothesis that returns are not related to lagged flows. This negative relationship causes mutual fund investors, as a group, to realize a lower long term accumulated return than the long term accumulated return on a buy and hold position in these funds. The bad performance of mutual funds' investors can be explained either by behavioral explanations, such as investor sentiment, or by efficient market explanations that are based on time varying risk premiums. We present a simple model where flows to equity funds are positively correlated with changes in the aggregate demand curve for stocks. We assume that the flows into and out of equity funds are not related to information about future cash flows (dividends), but they are caused by changes in other factors affecting the demand for stocks. Hence, a positive (negative) net flow in a given month implies a positive (negative) price change in the same month, but a lower (higher) expected future returns. We show that in each month the change in the expected future returns may be relatively small (relative to the return variance), but the accumulated effect of these changes may be significant. This result may explain why previous studies, using monthly data of flows and returns in either simple regression models or VAR, could not have significantly detect the monthly change in the expected future returns even in a 15-year sample.