Résumé en anglais

This paper develops a dynamic approach for assessing hedge fund risk exposures. First, we focus on an approximate factor model framework to deal with the factor selection issue. Instead of keeping the number of factors unchanged, we apply Bai and Ng (2002) and Bai and Ng (2006) to select the appropriate factors at each date. Second, we take into account the instability of asset risk profile by using rolling period analysis in order to estimate hedge fund risk exposures. Individual fund returns instead of index returns are employed in the empirical application to better understand the covariation structure of the data: the common behavior of hedge fund returns is filtered not only from the past historical data (time- series dimension), but also from the cross-section of returns. Finally, we apply our approach to equity hedge funds and replicate the returns of the aggregated index.