Further Biases in Using Dollar-Weighted Returns to Infer Investment Timing Effects

Abstract

The IRR (dollar-weighted return) reflects the periodic addition or withdrawal of funds by investors, and the difference between IRR and geometric mean is widely used to indicate the impact that the timing of these flows has had on investor returns. This is a biased measure, since it is also affected by investment flows which “chase” previous strong returns. A method has previously been derived for separating this bias from genuine timing effects. This paper demonstrates that using in-sample mean returns for this decomposition causes an additional bias which again misleadingly suggests bad investor timing. This paper quantifies this bias, allowing unbiased investor timing effects to be estimated. A proper understanding of these biases is of significant practical importance, since investors are often presented with biased timing indicators based on IRRs.