High Frequency ANOVA that is Robust to Jumps, Microstructure Noise and Asynchronous Observation Times

This paper develops the necessary methodology for high frequency ANOVA, which includes the estimations of idiosyncratic volatility and realized R-Squared. Because the residual process is latent in the high frequency regression, the estimation of idiosyncratic volatility is notoriously difficult and complex, especially in the presence of jumps, microstructure noise and asynchronous observation times. As a measure of goodness-of-fit in the high frequency regression, the realized R-Squared is a newly introduced quantity in the literature of high frequency econometrics, which allowing for the time-varying feature for the coefficient of determination. Benefited from the application of Truncated S-TSRV in the estimation of spot covariance, our methodology is robust to jumps, microstructure noise and asynchronous observation times simultaneously. Analytical bias correction techniques have been developed for the final estimators and the central limit theorems have been established accordingly. Monte Carlo evidence shows the robustness of our estimators. Empirical study has been conducted to investigate the empirical features of the idiosyncratic volatility estimate and the realized R-Squared estimate