Measuring mutual fund herding - A structural approach

Abstract

This paper proposes a methodological improvement to empirical studies of herd behavior based on investor transactions. By developing a simple model of trading behavior, we show that the traditionally used herding measure produces biased results. As this bias depends on characteristics of the data, it also affects the robustness of previous findings. We derive a new measure that is unbiased and shows superior statistical properties for data sets commonly used. In an analysis of the German mutual fund market, our measure provides new insights into fund manager herding that would have been undetected under the traditional statistic.