Abstract (EN)

This paper analyses the purchase and redemption behaviour of mutual fund investors and its implications on fund liquidity risk. We collect a novel set of proprietary data which contains a large number of French investors holding funds with various degrees of asset liquidity. We build a Self-Exciting Poisson model capturing fund flows' clustering effects and over-dispersion. The model improves the forecast accuracy of future flows and provides a reliable risk indicator (Flow Value at Risk.) Accordingly, we introduce the notion of liability risk where investor's behaviour increases mutual fund liquidity risk. We further decompose fund flows into investor categories. We find that investors exhibit high heterogeneous behaviour, and a lead-lag relation exists between them. Finally, we control flow dynamics for various economic conditions. We show that although flows evolve with economic conditions, investor's behaviour stays the main significant determinant of flows' randomness. Our findings encourage fund manager to adopt an ALM approach.