Is the flow-performance relationship really convex? - The impact of data treatment and model specification

This paper challenges the convexity of the flow-performance relationship, according to which investors strongly chase top-performing funds, while fund flows exhibit little to no sensitivity to past performance within the segment of poorly performing funds. Our results suggest that the flow-performance relationship is not convex, but rather linear. In contrast to prior studies, we use reported (i.e., exact) instead of approximated fund flow data, we trim (instead of winsorize) outliers, and we account for persistence in fund flows. We find that each factor contributes to serious biases. For example, investor reactions to poor performance only appear insignificant when outliers are winsorized instead of trimmed. And it is even more evident that fund investors flee poorly performing funds when the model incorporates lagged flows to account for fund flow persistence. Furthermore, our results provide evidence that the degree to which investors chase top-performing funds appears to be slightly upward biased if approximated fund flows are used. Our findings have important implications for the potential moral hazard of fund managers