This project investigates the time-varying variance and covariance matrix of stock and bond returns in China using the GARCH-MIDAS and DCC-MIDAS approach. GARCH is a univariate time series model that captures time-varying volatility and conditional heteroskedasticity. DCC is a multivariate GARCH model that estimates time-varying conditional correlations between multiple variables. MIDAS is a regression-based approach that allows for the inclusion of high-frequency and low-frequency data in the same model. The combined GARCH-MIDAS, DCC-MIDAS model are used to study the variance and correlation of stock and bond returns respectively. The finding shows that the variance of bond return is mainly affected by the lagged realized volatility itself and the variance of stock return is affected by both the lagged realized volatility and macroeconomic variables. Correlation between stock and bond returns in China is generally very weak and the sign of the correlation is different from the those obtained from the U.S financial market.