Forecasting Credit Default Swap spreads

This paper studies the predictability of 5-year Credit Default Swaps by constructing multiple forecasting models that incorporate a wide range of explanatory variables. A model that combines forecasts shows to be able to outperform benchmark models in terms of multiple forecast evaluation measures. Two ‘simple’ trading strategies based on buy/sell signals of this combination of forecasts model turn out to be highly profitable. This model combines the forecasts of ten models that incorporate a wide range of explanatory variables. Six of these models are based on three new variable reduction techniques. A new proposed factor based on the cross-sectional idiosyncratic risk in Credit Default Swap spreads shows to be an addition to the most widely used explanatory variable in forecasting credit related financial instruments. The results show the usefulness of combining forecasts of models based on new variable reduction techniques that incorporate a wide range of explanatory variables in forecasting Credit Default Swap spreads.