Mutual Fund Trading Costs

The profitability of an investment strategy depends on both the information value and the implementation costs of the strategy. While the crucial importance of trading costs is well-recognized in practice, relatively little has been known about the variation in trading costs across portfolio managers and its impacts on portfolio performance. One of the difficulties in understanding these relations is the lack of data that allow joint analysis of portfolio characteristics/performance and trading costs. In this study, we overcome this data limitation by linking together three commonly-used databases: (1) The ANcerno database of institutional trades, which contains detailed trade records of a large number of institutional investors; (2) Thomson mutual fund holdings database, which contains quarterly holdings of US stocks by mutual funds; (3) CRSP survivor-bias-free mutual fund database, which reports monthly return, total net asset value, and other fund level information such as investment objective, expense ratio and turnover. Our matched sample includes more than 500 US equity funds from 1999 to 2010. This large panel data set allows us to measure the trading costs directly at the trade level, and link them to the fund characteristics and performance. Using this data set, we investigate two closely-related questions:(1) How do trading costs differ across funds? (2) How do trading costs affect fund performance? To the best of our knowledge, we are the first to examine the cross-section of equity trading costs among US mutual funds using trade-by-trade data. Previous studies on institutional trading costs are either at the institution level, with limited information about the institutions, or rely on trades inferred from quarterly holdings data and ignore heterogeneity in trading skills across funds. Our trading costs estimated directly from the actual trade prices draw a more accurate picture of the ``invisible costs" of mutual fund trading. Our results are highly relevant for mutual fund managers, investors, and researchers. Specifically, the cross-sectional relation we obtain between trading costs and various fund characteristics provides a simple tool to estimate mutual fund trading costs and evaluate their trading performance. A researcher who has no access to the actual trade data can estimate the trading costs of a mutual fund by applying our regression results. Similarly, a fund management company can detect whether a fund's trading costs are ``abnormally" high or low using our regression results as a benchmark. Furthermore, our results on the relation between trading costs and fund performance can be a valuable guide toward mutual fund investment. Our results also provide important insights into the nature of active portfolio management. For example, the relation between fund/family size and trading costs tells us whether economy or diseconomy of scale dominates in institutional trading. The relation between fund performance and trading costs allows us to quantify the contribution of trade execution to fund performance and to assess whether the information value of mutual fund strategies tend to be offset by their implementation costs.