A Semiparametric Intraday Garch Model

We propose a multiplicative component model for intraday volatility. The model consists of a seasonality factor, as well as a semiparametric and parametric component. The former captures the well-documented intraday seasonality of volatility, while the latter two account for the impact of the state of the limit order book, utilizing an additive structure, and fluctuations around this state by means of a unit GARCH specification. The model is estimated by a simple and easy-to-implement approach, consisting of across-day-averaging, smooth-backfitting and QML steps. We derive the asymptotic properties of the three component estimators. Further, our empirical application based on high-frequency data for NASDAQ equities investigates non-linearities in the relationship between the limit order book and subsequent return volatility and underlines the usefulness of including order book variables for out-of-sample forecasting performance.