Nicolas P.B. Bollen

Owen Graduate School of Management
401 21st Avenue South
Nashville, TN 37203
office: 615.343.5029
fax: 615.343.7177

Research Story 2: Daily mutual fund performance

Back when I was a freshly minted Assistant Professor at the University of Utah, I was working on a small consulting project for the Australian Stock Exchange. They wanted a study that demonstrated the benefits of option trading for portfolio managers. I wrote a nice report, entitled Measuring the Benefits of Option Strategies for Portfolio Management, that argues why standard utility functions may reward the changes that option positions could make to the shape of a portfolio's return distribution. In the study, I also created a simulation exercise that shows the change in the distribution of a mutual fund's final NAV per $1,000 of initial investment, both with and without the protection that put options provide:

You can see that the put-protected portfolio has a large number of small losses, but that losses are capped at that level. This insurance is not free - the cost of the put protection slightly lowers the probability of gains at all levels.

At about that time, we were interviewing PhD students for possible additions to our faculty. One of the recruits was Jeff Busse, who studied at NYU. His dissertation used a database of daily mutual fund returns that he had put together. I started thinking about academic studies along the lines of the consulting report that I wrote, and I realized that Jeff's data would be necessary in order to turn from the simulation that I had done to analysis of actual fund performance. The reason is that estimating changes in the shape of a return distribution is highly data-intensive, so that the greater frequency data was important. I phoned Jeff (he took a job at Emory University in Atlanta) and we started working together.

The ability of mutual fund managers to trade options is highly limited by the SEC's Investment Company Act. However, fund managers can replicate some of the features of option payoffs in the returns of their funds through dynamic trading. Our first project was a study of market timing in mutual funds. Market timing is a form of dynamic trading which can generate similar patterns to positions in options. For example, if a manager forecasts bad market returns, he or she can reduce exposure to the market, thereby replicating a position in put options, which would generate positive payoffs when the market does poorly, thereby offsetting losses in the fund. We argued that existing studies of market timing are likely to be hampered by their use of monthly data. We then examined the daily data to show that a greater number of funds appeared to feature market timing when the higher frequency observations are used. The results were published in a 2001 Journal of Finance paper On the Timing Ability of Mutual Fund Managers.

Our second project continued our analysis of performance. Another issue that the daily data could help us with is whether performance of mutual funds persists. There are two reasons. First, with daily data, subsets can be more easily formed to determine whether performance in one time period predicts performance in a later period. Second, with daily data, shorter subsets can be analyzed, which is important if persistence is a short-term phenomenon. We were able to measure risk-adjusted performance over three month intervals, something impossible without daily data. The following picture is from the resulting 2004 Review of Financial Studies paper, Short-term Persistence in Mutual Fund Performance:

Here we show for the top 10% of funds and bottom 10% of funds as ranked by prior quarter performance, the average daily performance over the following quarter. You can see that the winners keep winning and the losers keep losing. We also find that over the following year, the persistence eventually dissipates. This is consistent with short-term informational advantages. It is also consistent with a well-cited article by Berk and Green, who argue that persistence can disappear if capital flows from performance-sensitive investors eliminate scale-dependent performance.

The third of my papers with Jeff was prompted by a discussion I had with my students in one of my Equities Markets classes. We were discussing the move from trading stocks in eighths to sixteenths and ultimately decimals. I argued, based on my earlier work with Bob Whaley on the subject, that mutual funds are likely to suffer from decimalization if liquidity is hurt by the smaller tick size. One of my students said "Why don't you make that your next research project?" and I said "Great idea!" That summer, Jeff and I worked at a furious pace, because I knew decimalization was a hot-topic. The resulting paper, Tick Size and Institutional Trading Costs: Evidence from Mutual Funds was published in the Journal of Financial and Quantitative Analysis and attracted a great deal of attention in the popular press:

The paper was also cited in a letter from the Security Traders Association to Chairman Donaldson of the SEC.

Listed here are some representative articles from the popular press, some which are supportive of our finding that actively managed mutual fund trading costs increased, on average, following decimalization.

"Decimalization hits U.S. Investors" Lauren Foster, Financial Times 5/12/03

"Decimalization under attack on Wall St." Floyd Norris, New York Times 5/16/03

"On pricing, is glass .5 full or .5 empty?" Chuck Jaffe, Boston Globe 5/18/03

"Revisiting decimals" Erin Arvedlund, Barron's 5/19/03