| Volatility and covariance estimation are of great importance to asset pricing, portfolio management, risk management and asset allocation. The availability of high frequency financial data facilitates the empirical studies of the asymptotic properties of various variance-covariance estimators. However, things such as market microstructure noise and asynchronous trading times need to be taken into account when estimating volatility and covariance.;In our research, we have studied different methods to estimate the stock returns' daily integrated variance and covariance using high frequency data. We extend the Bayesian estimation method for integrated volatility to the estimation of daily integrated covariance of two assets. Simulation studies show that the new estimators have both a small relative bias and root mean squared error for both the univariate and bivariate case. We also propose jump robust two time scale variance covariance estimators to estimate the daily integrated variance and covariance of stock returns in the presence of jumps, market microstructure noise as well as asynchronous trading. We show that our estimators outperform other popular jump robust estimators both in relative bias and root mean squared error through simulation studies. We further study the pre-averaging approach for integrated volatility where we derive the analytical form of the optimal sampling frequency of the estimator to minimize the mean squared error both when noise is absent and present. |
