|
On the Robustness of Performance Measures in Fund Persistence Authors: Yin-Ching Jan, Su-Ling Chiu Source: Journal of Performance Measurement, vol. 11, n°3 Date: Spring 2007 |
Persistence in mutual fund performance is of great interest to investors; indeed, it can determine their choice of investment. Numerous studies have used different methods of performance measurement to investigate the possible existence of persistence in fund performance. In the present article, Jan and Chiu examine whether the results depend on the measure used.
They use a data sample including 228 equity funds from the Taiwan Economic Journal Mutual Fund Database. The data cover a period beginning in January 1993 and ending in December 2004. The database includes defunct funds. Jan and Chiu evaluate annual fund performance using several methods including raw adjusted returns (fund return minus market return), Sharpe ratio, Jensen’s alpha based on the one-factor model, Fama-French’s alpha based on the three-factor model, Carhart’s alpha based on the four-factor model, the Portfolio Change Measure developed by Grinblatt and Titman (1993), which evaluates the covariance between the change in portfolio holdings and the realized returns, and also the Modified Portfolio Change Measure, which divides the Portfolio Change Measure by its standard deviation in order to adjust it for risk.
To ascertain whether or not the funds exhibit performance persistence, they look at whether the funds keep the same rank year after year. For each performance measure, and for each two-year interval in the period, they computed the Spearman rank correlation coefficient and test the null hypothesis that the coefficient equals zero. Their results show that the degree of persistence exhibited by the funds differs depending on the performance measure used. Most measures do not enable identification of persistence in performance. Only the Portfolio Change Measure makes it possible to identify persistence for a majority of the two-year intervals considered. To test the robustness of these results, the authors perform the same computation using semi-annual performance and two half-year intervals. Here again they find that Portfolio Change Measure is the best measure to identify persistence in performance, confirming the robustness of the results.
In addition, they investigate the autocorrelation of the performance measures, as a high degree of autocorrelation is a sign of persistence. They find that the average autocorrelation is positive for all the categories of funds with the Portfolio Change Measure, while the other measures show negative autocorrelation in some cases. Based on all their results, the authors conclude that the Portfolio Change Measure provides a consistent measurement to predict future performance. They do no extend this conclusion to markets other than the Taiwan mutual fund markets they investigated, leaving this work for further investigation.
Reference
Grinblatt, Mark, and Sheridan Titman, “Performance Measurement without Benchmarks: An Examination of Mutual Fund Returns”, Journal of Business, vol. 66, 1993, p. 47-68.




