Mutual Fund Survivorship

Abstract

Survivorship induces a variety of biases in mutual fund research. I show analytically that biases in performance estimates depend on sample length and whether funds disappear after one or many poor returns. Using a sample free of survivor bias, I document higher risk and predominantly multiple-year underperformance in nonsurviving funds. This causes the bias in mean return estimates to increase in the time-length of the sample. In my data set, the bias is 0.43 percent per year in five-year samples and approximately one percent for samples longer than fifteen years. I also find downward bias in persistence tests and both upward and downward bias in the relations between performance and fund attributes depending on the type of selection bias. The results cast doubt on the conclusions of many published mutual fund studies.

JEL Classification: G23, C40