Is Overnight Volatility Overlooked?

Abstract

The equity index futures, traded almost 24 hours, reveal the overnight price process. We use both daytime and overnight high-frequency futures price data to estimate the realized volatility (RV) of the S&P 500 and NASDAQ 100 indexes. We find that more than 40% of daily volatility is overnight volatility, which is omitted by many previous works. We demonstrate strong predictive power of overnight RV on daytime RV and vice versa. By incorporating overnight RV, our best-proposed model reduces the forecasting mean squared error of the S&P 500 daytime RV by 27.8% compared to the benchmark model. Based on the inter-correlation between daytime and overnight volatility, we propose a novel Day-Night Realized Stochastic Volatility (DN-SV-RV) model, where the daytime and overnight returns are jointly modeled with their RVs, and their latent volatilities are correlated. The proposed DN-SV-RV model has the best out-of-sample return density forecasting among benchmark models. Under this innovative framework, the model estimation shows that daytime and overnight volatility complement each other, and volatility clustering is persistent during and after regular trading hours. Finally, by jointly estimating daily close-to-close return and realized volatility measures, we show that the daily RV, estimated from the almost 24-hour traded futures, is an accurate measure of daily return volatility by effectively measuring the overnight volatility, compared to daytime RV and squared overnight returns.

Keywords: Overnight volatility, Realized Volatility, Overnight Returns, Daily Realized Volatility, Overnight Jumps, Stochastic Volatility, Bayesian MCMC, Predictive Density

JEL Classification: C11, C15, C22, C52