Duan, Yaxin, Fund Flow Betas and the Cross-Section of Stock Returns (August 8, 2014). Available for download at SSRN: http://ssrn.com/abstract=2478033

“I document a robust new fact about the cross-section of stock returns: stocks whose returns co-vary more with flows into the entire mutual fund sector earn significantly higher average returns. I consider three potential explanations for this finding: that it reflects cross-sectional differences in liquidity across stocks; that it reflects the hedging concerns of risk-averse fund managers; and that it stems from the fact that flows into the entire fund sector are a measure of good and bad economic times, so that stocks that co-vary more with flows are riskier. I show that the evidence is most consistent with the last interpretation. In short, my results provide support for a basic principle in asset pricing: that stocks that do well in good times and poorly in bad times should earn higher average returns.

I test three hypotheses for this empirical finding. The first hypothesis is based on the general concept of illiquidity and states that the spread reflects cross-sectional differences in liquidity across stocks. Amihud (2002) measures illiquidity as the average price impact of order flows. Fund flows into the mutual fund sector are likely to increase order flows, hence generating price pressure on financial assets. Stocks that co-vary more with aggregate fund flows may therefore be illiquid stocks. I provide three pieces of evidence that go against this hypothesis. First, the correlation between illiquidity (measured as in Amihud (2002), which is a measure of the price impact of order flows) and fund flow beta is very small, at 0.04. Second, the return spread between the high and low portfolios generated by sorting stocks on fund flow beta alone is much larger, both in magnitude and in significance, than the spread generated by sorting stocks on illiquidity alone. Third, I perform a double sort analysis by first sorting stocks on illiquidity, and then on fund flow beta. The results show that fund flow beta is still related to cross-sectional returns, even for stocks within the same illiquidity group.

The second hypothesis is that my main empirical finding reflects the hedging concerns of risk-averse fund managers. It is common practice for mutual fund industry to compensate a fund manager in proportion to their total amount of assets under management (AUM). A fund manager therefore has an incentive to maximize his fund’s AUM. Given that the growth in a fund’s AUM is determined both by the return on the fund’s portfolio as well as by the flows into the fund, a risk-averse fund manager would require a premium for holding stocks that co-move positively with fund flows because these stocks would increase the volatility in the fund’s AUM. Under this hypothesis however, fund managers only care about the co-variance of their fund flows with the assets that they hold. In particular, a manager whose fund does not invest in stocks would not care whether a stock co-varies with his fund flows. Hence, flows into stock funds alone should more accurately capture the volatility risk that fund managers are concerned about. Therefore, the empirical results should become stronger if I consider only flows into stock funds rather than flows into the entire fund sector. To the contrary, I find that the spread between high flow beta and low flow beta stocks drops substantially when I replace aggregate fund flows by flows into stock funds alone. Furthermore, I decompose aggregate fund flows into two orthogonal parts. I show that the main empirical result still holds if I estimate fund flow betas using the variation in aggregate flows that is orthogonal to stock fund flows. This is inconsistent with the second hypothesis as fund managers should not care about the variation orthogonal to their fund flows.

Finally, the third hypothesis is that fund flows into the mutual fund sector measure good and bad economic states, so that co-variation with these flows earns a risk premium. In particular, that a high inflow reflects a good economic state, while a low inflow or an outflow reflects a bad economic state. Basic finance theory predicts that stocks that co-move positively with a measure of the state of the economy should have higher average returns as compensation for their higher systematic risk. I provide empirical evidence that is consistent with this hypothesis. First, I construct a measure of general economic conditions by extracting the principal component from a group of macro variables commonly associated with the state of the economy. I show that fund flows are positively related to this measure. Second, I also show that this measure is positively related to the return difference between the high fund flow beta and low fund flow beta portfolios: in short, consistent with the hypothesis, high(/low) fund flow beta stocks do perform well(/poorly) in good times and poorly(/well) in bad times. The third hypothesis interprets the spread between high and low fund flow beta portfolios as a risk premium. As such, the spread may help explain stock returns. I perform GRS (Gibbons, Ross and Shanken (1989)) tests on common sets of portfolios and find that adding the spread to existing risk factor models as an additional factor improves the models’ fit in some settings.”