The information content of overnight information for volatility forecasting: evidence from China's stock market

Using overnight volatility as the proxy for overnight information, this paper models future Chinese stock market realized range–based volatility (RRV) within a class of heterogeneous autoregressive models augmented by this proxy. We confirm the important role of overnight information in volatility forecasting models with strong evidence from in-sample and out-of-sample analyses. Moreover, such forecasting improvement is considerable at the short-term prediction horizon but weakens as the prediction horizon extends. We conduct numerous robust tests to strengthen our findings, with alternative rolling window lengths, alternative loss criteria, and alternative volatility estimators. We also provide evidence that our forecasting model incorporating overnight volatility performs extremely well in volatility forecasting during times of market turbulence.