Modeling Corporate Bond Spreads : An Application-Oriented Forecasting Exercise

The question of predictability of credit spreads is of more than just academic interest. The ability to generate unerring spread forecasts is of considerable practical relevance for both treasurers of companies, who want to finance themselves through bonds, and institutional investors, who must manage their fixed income portfolios to an optimum. Against this background, the present paper discusses the quantitative modeling of aggregate credit spreads of Euro-denominated corporate bonds. The analysis comprises three main elements: identification and selection of important determinants of credit spreads, choice and implementation of appropriate econometric methods according to statistical time series properties as well as model evaluation on the basis of simulated real-time predictions. The results of this study show that monthly changes of credit spreads can be predicted quite well, and that the selected quantitative methods generate forecasts that are superior to the benchmark "random walk" model