For mutual fund analysts, new insights into evaluating a fund’s performance don’t come along every day. Lately, though, financial experts at Wharton and other schools have collaborated on research that challenges some long-held notions about which funds are best for investors.

Earlier this year, for instance, Wharton Finance Professor Andrew Metrick and graduate students Klaas Baks of Brown University and Jessica Wachter of Harvard posed the thesis that actively managed funds, in which expert managers carefully choose a portfolio of investments, can in some cases outperform passively managed funds whose portfolios mimic the returns and risks of elite companies in a given category or index. Their paper was titled, The 6% Factor: Which Fund Managers Will Outperform Index Funds?

Now Wharton Finance Professor Robert F. Stambaugh and the National Bureau of Economic Research, along with Lubos Pastor of the University of Chicago’s Graduate School of Business, have presented a paper, “Evaluating and Investing in Mutual Funds,” showing that, with a little statistical tweaking, standard benchmarks used to evaluate and predict fund performance can be improved on.

Evaluating mutual fund performance, the authors note, requires a combination of data and judgment. The judgment comes in when picking what standard, or benchmark, to use as the basis for calculating how well a fund is doing. It also may involve one’s faith, or lack of it, in the ability of a fund manager to pick profitable investments. The Stambaugh-Pastor model provides a framework to incorporate both potential managerial skill and the adequacy of chosen benchmarks into the investment decision.

Stambaugh and Pastor accomplish this by introducing the notion of using non-benchmark data on a fund’s profitability to estimate its success. First, they defined a benchmark that measures a fund’s performance in terms of its “alpha” (a performance index established by calculating how well the fund did in a given time period compared with an appropriate measuring point, such as the S&P 500). Then they added to that a variety of additional calculations reflecting returns on passively managed assets in a fund’s portfolio. The addition of these non-benchmarks helped them estimate a fund’s performance better than was possible by using the benchmark alone.

Examples of non-benchmark measurements used in the study included a calculation that correlated a fund’s long and short positions with its market capitalization as well as with its book-to-market-price ratio; a measurement of a fund’s momentum, which averages out its recent history of financial returns; and the differences among a fund’s returns earned on small, medium and large company stocks as well as stocks with low, medium and high book-to-market ratios.

They then applied these evaluations to a broad selection of sample fund portfolios. For instance, they looked at a sample universe of 413 small-company growth funds, using information on non-benchmark assets as well as the capital asset pricing model (CAPM), a common benchmark that relates risk and expected return for a given investment. Typically, adding the non-benchmark information to the CAPM made the Stambaugh-Pastor estimates between 7.2% and 8.3% per year more accurate than the CAPM alone, depending on assumptions made about the fund managers’ skills.

Another analysis of portfolios of no-load funds under various assumptions about manager skill and book-to-market price came to similar impressive conclusions. The study constructed portfolios with maximum Sharpe ratios from a universe of 505 no-load equity funds that had at least 36 months of return history under its most recent manager, existed at the end of 1998 and had published data on their expense ratios and turnover rates. (A Sharpe ratio, named after its inventor, William Sharpe, is found by subtracting the risk-free yield rate, such as on 91-day Treasury bills, from a portfolio’s total return, then dividing this number by a measure of the fund’s volatility over a certain period). The Stambaugh-Pastor analysis, measured against the no-load funds’ CAPM, was more accurate for predicting actual returns by between 1.73% and 3.93%.

Although it’s unlikely that an individual investor would use the advanced statistical analysis employed in this study, the findings will be of interest to analysts who recommend the best mutual funds for a given investment strategy. The conclusion for analysts to take to heart is that simply comparing a fund’s rate of return with, say, the S&P 500 index does not provide a complete picture of how well the fund is performing.