Essays on Mutual Funds

Abstract

My dissertation comprises of three essays on mutual funds. In the first essay, I test whether fund investors rationally incorporate portfolio manager ownership disclosure in their portfolio allocation decisions. Using a natural experiment, regulations that mandate portfolio manager ownership disclosure, I find that investor flows respond to higher percentage ownership. The relationship between investor flows and percentage ownership is persistent well after the regulatory change in 2005, suggesting that the investor responses are permanent rather than transient and are robust to several controls as well as unobserved heterogeneity reflected in fund family and manager fixed-effects. The investor responses to ownership are rational, as investors investing in higher percentage ownership funds are rewarded back in terms of higher risk-adjusted performance. Finally, I shed light on the channels through which higher ownership translates into better investor rewards. I show that agency alleviation is one of the channels through which ownership translates into better investor rewards. These findings are consistent with a "rational investor" viewpoint in which, at least, some investors incorporate managerial ownership in their portfolio allocation decisions. In the second essay, I analyze the "herding" (trading together) behavior of managers, conditional on their ownership stakes in the funds they manage. I find that the funds with low and high managerial ownership have economically distinct patterns in their herding behavior. Each herd has its own distinct trading style and different qualitative and quantitative effect on stock prices. Low ownership funds herd more and engage in positive-feedback trading that is followed by stock price reversals. High ownership fund herding is followed by more stable price adjustments. Low ownership herding effects appear to dominate in the full sample where herding causes price reversal. These findings suggest that there is heterogeneity in the herding behavior of mutual funds, which appears to be related with ownership. It is costly for the high ownership managers to ignore their substantive information due to reputational concerns, or to engage in uninformed trading, and thus herding by such managers results in more informative prices. On the other hand, lower ownership fund herding appears to be driven by agency that generates temporary price movements that are reversed. In the third essay, I and my co-authors, Gerard Hoberg and N. R. Prabhala, propose new techniques for identifying benchmark peers for mutual funds. We identify the location of funds in the space of stock style characteristics. All funds within a pre-specified normed distance are a fund's peers. Our benchmark peers are customized to each fund, intransitive, and employ techniques that are readily scalable across dimensions and loss functions. We show that peers derived in this fashion are significantly different cross-sectionally from conventional peers and exhibit considerable dynamic churn. The customized peers we construct outperform traditional peers in out of sample prediction tests, have lower tracking error, and our peer-excess alphas predict the future Characteristic-Selectivity and Carhart alphas of funds. We find that measures of competition derived from our peers predict performance persistency of funds for up to four quarters.