Parsimonious Estimation of Credit Spreads

Abstract

The traditional method of credit spread estimation is based on subtracting independently estimated risk-free and risky term structures of interest rates which in many cases yields unrealistically shaped and often irregular credit spread curves. A parsimonious joint estimation of the risk-free term structure and the credit spread as proposed by Houweling et al. (2001) might serve as a valuable alternative to overcome this drawback but it is hard to decide whether a seemingly irregular shape of the credit spread curve is economically caused by the data or is only an artefact of the functional form of the estimation model. Results of an empirical examination of EMU government bond data show that traditional estimation models with different functional forms yield differing irregularities in the credit spread curves whereas joint estimation procedures result in well-behaving and coinciding curves. Moreover, the explanatory power of the more parsimonious joint estimation procedures is virtually equal to the traditional methods. This is strong evidence for the superiority of a joint estimation procedure of credit spread curves. Finally, we conclude that a simple linear joint cubic splines specification performs surprisingly well compared to a numerically more affording non-linear model.