Title: A stochastic volatility model with two macro-financial components
Authors: Yuze Liu - University of Cologne (Germany) [presenting]
Abstract: The use of macroeconomic variables to improve the accuracy of daily financial volatility estimation has been discussed in the past years. Volatility is typically decomposed into two components: a short-run component from the classical financial volatility estimation, and a long-run component corresponding to macroeconomic determinants. However, there are some important limitations: First, high-frequency volatility is usually estimated by using the GARCH-framework. Second, the low-frequency macroeconomic variables are included via discrete piecewise constant step-functions during a month or a quarter. A new and simple stochastic volatility model with two components is proposed. The short- and long-run component are mapped into two latent variables, where the long-run component is varying simultaneously with the short-run component at the high-frequency. Despite the increased flexibility of the newly proposed model, it retains the classical state-space representation which can be efficiently estimated by using standard Kalman Filter techniques. The finite sample properties are investigated by using extensive Monte Carlo simulations. The proposed model is compared against the classical one component SV and GARCH models, as well as two components GARCH-MIDAS models. As an empirical application, the US stock market volatility is investigated.