This is the accepted version of the paper.This version of the publication may differ from the final published version.
Permanent repository link
AbstractThis paper proposes a new bivariate modeling approach for setting daily equity-trading risk limits using high-frequency data. We construct one-day-ahead Value-at-Risk (VaR) forecasts by taking into account the different dynamics of the overnight and daytime return processes and their covariance. The covariance is motivated by market microstructure effects such as price staleness and news spillover. Among the competitors we include a simpler bivariate model where the overnight return is redefined by moving the open price further into the trading day, and a univariate model based on the close-to-close return and an overnight-adjusted realized volatility. We illustrate the different approaches using data on the S&P 500 and Russell 2000 indices. The evidence in favour of modeling the covariance is more convincing for the latter index due to the lower trading volumes and, relatedly, the less efficient price discovery at market open for small-cap stocks.