A Multiscale Estimator for Pricing Errors in High-Frequency Financial Markets

For a Lévy process corrupted with microstructure noise, I derive the sampling distributions for the information-related and information-unrelated pricing error parameters and for the variance of latent true price returns (a noise-robust and consistent estimator of realized variance). The test statistics converge in distribution to the standard normal distribution, while statistics for joint tests, tests for intraday seasonality, and tests for time varying parameters converge in distribution to the χ^2 distribution. Simulation evidence verifies that test statistics display good properties. As an empirical example, the proposed tests are taken to a sample of exchange rates, commodities, and index futures