Testing for Endogeneity of Irregular Sampling Schemes

Abstract

In the context of high frequency financial data, it is often assumed that sampling times are nonendogenous. We derive statistical tests capable of determining whether, and to what extent, this hypothesis is rejected by the data. We propose two kinds of testing procedures: one suitable for moderate sampling frequencies, where the observed price can be assumed to be a noiseless Brownian semi martingale, and a second applicable even for tick-by-tick sampling. Using a vast dataset of financial asset prices, we give empirical evidence that trade arrival times do not show dependence on the efficient component of the observed price process and none of the kind that may jeopardize well-known results on convergence of power variations. Extensive Monte Carlo simulations confirm the good finite-sample performance of the proposed tests and provide micro-founded justification for the empirical results.