Essays on Mutual Fund Performance

In this dissertation we use a network approach based on cross-fund correlation to calculate metrics that can help to describe the existence of common features not captured by common factors. Previous studies have found that measurements of performance adjusted by risk factors are explained by some fund characteristics, like size, age, expenses and turnover. This dissertation specifically studies performance in the mutual fund industry utilizing a network approach to address the following three questions. Do the characteristics of the mutual fund network help to explain the performance of mutual funds? Is the position of a mutual fund in the network related to persistence of abnormal performance of the fund? Do changes in the structure of the network of mutual funds signal the presence of herd behavior? In the first essay of this dissertation we argue that performance is also explained by other features of the funds related to informational linkages. We argue that managers with different types of information will invest differently. In the same line of argument, managers with access to only free public information and poor private information should behave similarly, and their investments should perform close to the market, while managers of the type that holds quality private information should outperform the market. In that sense, funds with similar type of information will be correlated. We propose that informational linkages can be approximated by measurements of centrality in a network of mutual funds. Centrality of a fund in the network, as explained later, can be interpreted as a measure of how closely correlated the returns of a fund are to other funds in the market. In the second essay we build on the foundations of the first essay to include measures of the mutual fund network in order to establish a link between persistence and correlation across funds. If there are informational linkages, those connections could help to test for persistence. Then, we expect that the returns of those funds will be correlated to each other; if the returns of those funds were correlated by luck, it is expected that in the next period, those linkages will disappear. However, when those linkages remain from one period to another, we can consider that the returns of the funds are not correlated by chance and we expect to find persistence. Finally, we consider how correlation across funds can help to explain herd behavior in the mutual fund industry. The third essay argues that an aggregate centrality measure in the network of mutual funds could signal the existence of herd behavior in the market