Country Default Probabilities : Assessing and Backtesting

We present statistical backtesting procedures applicable in situations where few and heterogeneous probabilities have to be evaluated, as is typically the case when forecasting country default risk. These tests are applied to a sample of default probabilities assessed for 19 emerging market and transition countries by estimating a Merton-type credit risk model using bond market data. For the estimation, we use a maximum likelihood technique based on time series of market data. Thereby, we avoid the drawbacks of alternative approaches commonly used in the literature