Divergence of credit valuation in Germany: Continuous theory and discrete practice

Lending is associated with credit risk. Modelling the loss stochastically, the cost of credit risk is the expected loss. In credit business the probability that the debtor will default in payments within one year, often is the only reliable quantitative parameter. Modelling the time to default as continuous variable corresponds to an exponential distribution. We calculate the expected loss of a trade with several cash flows, even if the distribution is not exponential. Continuous rating migration data show that the exponential distribution is not adequate in general. The distribution can be calibrated using rating migrations without a parametric model. A practitioner, however, will model time as a discrete variable. We show that the expected loss in the discrete model is a linear approximation of the expected loss in the continuous model and discuss the consequences. Finally, as costs for the expected loss cannot be charged up-front, the credit spread over risk-free interest is derived.