One-Factor-GARCH Models for German Stocks - Estimation and Forecasting

Abstract

This paper presents theoretical models and their empirical results for the return and variance dynamics of German stocks. A factor structure is used in order to allow for a parsimonious modeling of the first two moments of returns. Dynamic factor models with GARCH dynamics (GARCH(1,1)-M, IGARCH(1,1)-M, Nonlinear Asymmetric GARCH(1,1)-M and Glosten-Jagannathan-Runkle GARCH(1,1)-M) and three different distributions for the disturbances (Normal, Student's t and Generalized Error Distribution) are considered. Out-of-sample forecasts for the stock returns based upon these models are computed. These forecasts are compared with forecasts based on individual GARCH(1,1)-M models, static factor models, naive, random walk and exponential smoothing forecasts.