Networks and Performance of Index-Benchmarked Mutual Funds

Abstract

This paper establishes empirical evidence that network among mutual funds that share a similar benchmark plays an important role in changes in fund net asset value (NAV) and total net assets (TNA). Using a matrix that indicates the level of overlapped security holdings of funds, we obtain eigenvector centrality, which measures the `influence' of each fund. A fund that has high eigenvector centrality is one that has high fund holdings overlap with other peers, who themselves have high overlap with others. A panel VAR analysis shows significant positive relation of changes in lagged eigenvector centrality with changes in fund NAV and TNA. Changes in eigenvector centrality `Granger causes' changes in fund's NAV and TNA and vice versa. A shock in centrality leads to an increase in fund NAV and TNA. These results make eigenvector centrality a feasible indicator to measure the benchmark-induced herding of individual funds. High centrality funds that have the highest herding tendencies among their peers have the largest average total assets under management. In general, high centrality funds have the lowest tracking error, fund beta, expense and turnover ratios and management fees compared to their peers. To achieve these criteria, high centrality funds have larger exposure to the `low-beta' stock market segment compared to their peers, as well as small but nimble positions in the `high-beta' stock market segment.