A1198
Title: Instrumental realized volatility
Authors: Hiroyuki Kawakatsu - Dublin City University (Ireland) [presenting]
Abstract: The use of multiple realized variance measures is considered in observation driven GARCH type and parameter-driven stochastic volatility models. The main idea is to use several noisy realized measurements as instruments of each other to extract signals to help identify the underlying latent variance series. In addition to the commonly used tick, data-based realized measures, information content in the more widely accessible but noisy open-high-low-closed data-based realized measures are considered. The performance of the proposed models is compared using (pseudo) out-of-sample tail risk forecast accuracy.