Modeling of Credit Risk: An Empirical Analysis of the European CDS Market

Thomas Bo Kragh & Jacob Vilby

Student thesis: Master thesis


The European bond market has been expanding over the past 15 years. This makes the credit market more accessible for investors as they earlier had to achieve their credit exposure through credit institutions. Therefore, credit risk is an important topic. The definition of credit risk is the likelihood of a borrower not repaying his obligation and the lender thus losses the face value of the debt and the interest related to the loan.
A credit default swap’s most important driver is credit risk and thus the spreads of a CDS is largely given by the credit risk of a given company. Therefore, CDS spread can be used as a proxy for the credit risk of an individual company at a given time.
This report examines and develops a model to estimate CDS spreads. The starting point of the model is an extended Merton model. The extended Merton model shows limited precision in its estimates and an analysis shows that one reason for this is a limited number of input variables. Therefore, further explanatory variables are identified and a panel data regression is performed on the difference between the observed CDS spreads and the extended Merton model estimated spreads. In the panel data regression 15 variables are tested for a significant and consistent contribution to the model throughout different periods. Seven variables are found significant and consistent, of which three are already included in the extended Merton model and are found to have significant additional information. Further, four new variables are found significant and consistent. These four variables include two proxies for the market environment, one explaining the profitability of the firm and one describing the illiquidity of the CDS. Though the last mentioned does not describe credit risk, it is included to avoid omitted variable bias and further give a more precise estimation.
The panel data regression estimates are combined with the extended Merton model’s estimations and a combined estimation of spreads from the two models are found. The spreads from our model are compared to the observed CDS spreads. Our model shows a significant improvement of the extended Merton model increasing the precision from 0.47 to 0.73. Thus, the study concludes that the extended Merton model is significantly enhanced by the panel data regression as a further extension

EducationsMSc in Finance and Accounting, (Graduate Programme) Final Thesis
Publication date2017
Number of pages136