Bootstrapping autocorrelated financial time series

Abstract

Bootstrap replications of financial time series are used to test the efficacy of portfolio selection algorithms, arbitrage strategies, and other risk management techniques. This paper shows several methods for bootstrapping financial time series with the objective of preserving autocorrelation structure and allowing time-varying variance in the replicated series. The replicated series are tested in terms of their resulting distributions by two methods: comparison of the distributions' moments and the autocorrelation coefficients.