Mutual fund behavior under the microscope

This dissertation examines mutual fund data through the “microscope” of modern econometrics to uncover new insights into mutual fund behavior. It demonstrates that, even from the limited and infrequent data reported by mutual funds, advanced econometric techniques can extract new and revealing evidence on how mutual funds operate, how skilled managers make investment decisions, and how investor behavior interacts with sustainability preferences.

The first study introduces a novel methodology to impute daily mutual fund trades using quarterly holdings, monthly asset data, and daily returns. This approach enables the reconstruction of 10.5 million daily mutual fund trades across thousands of U.S. mutual funds. The resulting dataset opens the door to a new generation of empirical research on mutual fund behavior at the daily level.

The second study investigates whether the investment decisions of top-performing mutual fund managers can be replicated using publicly available information. By training machine learning models on the high-conviction holdings of these managers, the study shows that their stock selection process reflects systematic patterns in observable stock characteristics that can, to some extent, be learned and replicated.

The third study analyzes whether investors benefit from allocating capital to sustainable mutual funds. It documents a striking paradox: while sustainable funds underperform on a buy-and-hold basis, their investors earn higher dollar-weighted returns than investors in conventional funds. The findings suggest that this performance difference is largely explained by the tendency of investors with better timing behavior to self-select into sustainable funds based on fund attributes unrelated to sustainability.