Improving Performance for Pooling Averaging Strategies in Forecasting Sovereign CDS Spreads

This paper examines the estimation and forecasting of heterogeneous panel data models by various estimation techniques, of which a particular pooling averaging approach is of central interest. Furthermore, the research addresses the potential time-varying nature of financial products. This is done by incorporating optimal window size selection in the estimation of regression parameters. These methods are applied to forecast sovereign CDS spreads. By considering two different angles of approach for window determination, it is found that these methods can help improving forecasting performance significantly.