Standard Errors of Risk and Performance Estimators for Serially
Correlated Returns.

A new method for computing the standard errors of returns-based risk and performance measure estimators for serially correlated returns is described. The method estimates the spectral density at the origin of a time series of influence function transformed returns by fitting a polynomial to the periodogram using a generalized linear model for exponential distributions with elastic net regularization. We show that the method works well for hedge fund returns, and Monte Carlo studies for first-order autoregressions show that the mean-squared-error performance of our new method is better than Newey-West type methods for a number of common risk and performance estimators.