题目：Regressions at High Frequency

主讲人：Professor Joon Y. Park, Indiana University

Abstract

We establish the asymptotic theories for the general classes of high frequency regressions, including both the common trend models and the conditional mean models. Our asymptotic theories are based on two dimensional asymptotic, assuming that the sampling interval shrinks down to zero as well as the sample span increases up to infinity. Consequently, they reveal many novel and interesting statistical aspects of high frequency regressions, which cannot be analyzed by the usual one dimensional asymptotic relying only on the sample size. For the common trend models, we investigate the regressions involving nonstationary covariates generated by continuous time processes with common stochastic trends. On the other hand, for the conditional mean models, we study two different types of regressions identified respectively by the martingale condition and the orthogonality condition. A wide range of models for asset pricing are of one of these two types. In the development of our asymptotic, we pay a particular attention to the persistency of stochastic volatility that is observed commonly in many financial time series. A variety of empirical illustrations will also be given during the presentation.