This weeks Graduate Student Seminar speaker is William Nicholson.
Talk Title: Strategies for High Dimensional Volatility Modeling
Abstract
The Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model proposed by Bollerslev (1986) is widely accepted as the standard approach for forecasting univariate conditional variance in financial applications. However, the multivariate extension, especially in high dimensions, remains an open problem. Moreover, because the true conditional variance is never observed, even ex-post, existing forecast evaluation methods can produce ambiguous conclusions. I will present an overview of several popular multivariate forecasting methods, including Orthogonal GARCH (Alexander, 2000), Dynamic Conditional Correlation (Engle, 2002), and Dynamic Orthogonal Components (Matteson and Tsay, 2011). I will also discuss several popular forecast evaluation methods, specifically the Model Confidence Sets approach of Hansen et. al. (2011).