Anchoring credit default swap spreads to firm fundamentals

The literature has shown that structural models tend to generate biased credit spread predictions on average and short-term changes in firm fundamentals do not explain much of short-term changes in credit spread. In this paper, we explore the capability of structural models, and more generally firm fundamentals, in explaining the cross-sectional variation of credit default swap spreads. The paper starts with a new implementation of the Merton (1974) structural model, highlighting its cross-sectional explanatory power, and then proposes a Bayesian shrinkage method to combine the additional predictions from a long list of firm fundamental variables. A comprehensive analysis based on 579 U.S. non-financial public firms over a period of 351 weeks shows that, with the new implementation, the structural model can explain over 66% of the cross-sectional variation on average. Incorporating additional fundamental variables can increase the average cross-sectional explanatory power to 77% while also making the performance more uniform over time.

When the market CDS observation deviates from the fundamental-based prediction, the deviation forecasts future market CDS movements as the market CDS converges to the fundamental-based prediction in the future. By estimating a vector error correction model, we show that the fundamental-based prediction contributes to a larger proportion of the permanent CDS movements than does the market observation. Through an out-of-sample investment analysis, we further show that the predictions are not only statistically significant, but also economically important.

Our fundamental-based predictions can prove useful in at least two economically important applications. First, broker dealers and investors can use the fundamental-based predictions to improve and expand CDS marks both for market making and for marking to market of CDS positions. In the U.S. non-financial public company universe, only hundreds of companies have market CDS quotes. Our methodology can generate WCDS predictions on thousands of these companies and can thus greatly expand the universe with CDS marks. For companies with existing market CDS quotes, investors can also use the fundamental-based prediction to improve the accuracy of the marks by removing the transitory noise in the market quotes. Second, investors can potentially use the deviations between market quotes and the fundamental-based predictions to form CDS investment strategies that generate high returns and low risks while providing liquidity to the market.