A CLOSER LOOK AT THE EPPS EFFECT

Abstract

Epps [17] reported empirical evidence that stock correlations decrease when sampling frequency increases. This phenomenon, named Epps effect, has been observed in several markets. In this paper, the dynamics underlying the Epps effect are investigated. Using Monte Carlo simulations and the analysis of high frequency foreign exchange rate and stock price data, it is shown that the Epps effect can largely be explained by two factors: the non-synchronicity of price observations and the existing lead-lag relationship between asset prices. In order to compute co-volatilities, an original method based upon the Fourier analysis is adopted. This method performs well in estimating correlations precisely, as illustrated by simulated experiments. Being naturally embedded in the frequency domain, this estimator is well suited to the study of the Epps effect.