CFCM 14/15: The Importance of a Time-Varying Variance and Cross-Country Interactions in Forecast Models

Abstract

This paper examines growth forecasts of models that allow for cross-country interactions and/or a time-varying variance plus feedback from volatility to growth. Allowing for these issues is done by augmenting an autoregressive model with cross-country weighted averages of growth and/or the GARCH-M framework. The models also allow for structural breaks in the mean and variance of growth. The obtained forecasts are then evaluated using statistical criteria, i.e. point and density forecasts, and an economic criterion, i.e. forecasting recessionary events. The results show that the two components are important to obtain improved point and density forecasts, but that forecasting recessionary events remains difficult.

Download the paper in PDF format

Authors

Steven Trypsteen

View all CFCM discussion papers | View all School of Economics featured discussion papers