Abstract:We introduce and evaluate mixed-frequency multivariate GARCH models for forecasting low-frequency (weekly or monthly) multivariate volatility based on high-frequency intra-day returns (at five-minute intervals) and on the overnight returns. The low-frequency conditional volatility matrix is modelled as a weighted sum of an intra-day and an overnight component, driven by the intra-day and the overnight returns, respectively. The components are specified as multivariate GARCH (1,1) models of the BEKK type, adapted to the mixed-frequency data setting. For the intra-day component, the squared high-frequency returns enter the GARCH model through a parametrically specified mixed-data sampling (MIDAS) weight function or through the sum of the intra-day realized volatilities. For the overnight component, the squared overnight returns enter the model with equal weights. Alternatively, the low-frequency conditional volatility matrix may be modelled as a single-component BEKK-GARCH model where the overnight returns and the high-frequency returns enter through the weekly realized volatility (defined as the unweighted sum of squares of overnight and high-frequency returns), or where the overnight returns are simply ignored. All model variants may further be extended by allowing for a non-parametrically estimated slowly-varying long-run volatility matrix. The proposed models are evaluated using five-minute and overnight return data on four DJIA stocks (AXP, GE, HD, and IBM) from January 1988 to November 2014. The focus is on forecasting weekly volatilities (defined as the low frequency). The mixed-frequency GARCH models are found to systematically dominate the low-frequency GARCH model in terms of in-sample fit and out-of-sample forecasting accuracy. They also exhibit much lower low-frequency volatility persistence than the low-frequency GARCH model. Among the mixed-frequency models, the low-frequency persistence estimates decrease as the data frequency increases from daily to five-minute frequency, and as overnight returns are included. That is, ignoring the available high-frequency information leads to spuriously high volatility persistence. Among the other findings are that the single-component model variants perform worse than the two-component variants; that the overnight volatility component exhibits more persistence than the intra-day component; and that MIDAS weighting performs better than not weighting at all (i.e., than realized volatility).

A coupled component DCS-EGARCH model for intraday and overnight volatility

A Coupled Component GARCH Model for Intraday and Overnight Volatility

Prediction of Volatility Based on Realized-Garch-kernel-type Models: Evidence from China and the U.S.

Incorporating Overnight and Intraday Returns into Multivariate GARCH Volatility Models

Coupled GARCH(1,1) model

Asymmetry of return distribution, abnormality of intradaily risk based on quantile regression model with high-frequency data from Shanghai stock market

Prediction of Realized Volatility Based on Realized-GARCH-Kernel Model: The Comparison of CHINA and US

Forecasting Stock Volatility Using Time-Distance Weighting Fundamental’s Shocks

Modeling the Impact of Overnight Surprises on Intra-daily Stock Returns

Forecasting Intraday Volatility and Var Using Multiplicative Component Garch Model

Mixed-Frequency Multivariate GARCH

Overnight GARCH-Itô Volatility Models

A time-varying GARCH mixed-effects model for isolating high- and low- frequency volatility and co-volatility

Sparse Multivariate GARCH

Forecasting risk measures using intraday and overnight information

Improving Quantile Forecasts via Realized Double Hysteretic GARCH Model in Stock Markets

The fine structure of volatility feedback II: Overnight and intra-day effects

Volatility Spillovers: A Sparse Multivariate GARCH Approach with an Application to Commodity Markets

An Intraday GARCH Model for Discrete Price Changes and Irregularly Spaced Observations

Efficient Semiparametric Garch Modeling of Financial Volatility

Extended Multivariate EGARCH Model: A Model for Zero‐Return and Negative Spillovers