Title: Testing the absence of lead-lag effects in high-frequency data
Authors: Yuta Koike - University of Tokyo (Japan) [presenting]
Abstract: The focus is on the problem of testing whether there exists a (possibly) time-lagged correlation between two Brownian motions based on their high-frequency observation data where the observation times are possibly non-synchronous. The test statistic considered here is the maximum of the absolute value of the empirical cross-covariance function as a contrast function to estimate the time-lag parameter of the lead-lag relationship considered here. The approximation of the null distribution of the test statistic is analytically difficult, so we develop a bootstrap procedure to solve this issue. The validity of the proposed bootstrap procedure is ensured by a version of the Gaussian approximation theory.