Dynamic Hedge Fund Exposures:
One Size Estimation Interval Doesn’t Fit All

Abstract

Investors frequently use rolling regressions to estimate dynamic factor exposures. The estimation interval is an important parameter, but is often set unsystematically. The author studies interval selection through a model in which a fund’s factor exposure switches between high and low states, and investors track or clone this exposure with a rolling regression. The mean square error between the fund and clone depends on the fund’s exposure turnover and idiosyncratic risk, as well as the estimation interval. Lower idiosyncratic risk and higher turnover favor shorter intervals, but when turnover is sufficiently high, an expanding window is superior to fixed intervals. Consistent with the article’s model, expanding interval clones of some popular hedge fund indexes have smaller errors than those based on rolling intervals of any length.