Estimating the probabilities of default under the assumption of unobserved heterogeneity

Bond default and rank transition are often modeled as a Markov chain with an absorbing state. However, recent studies have shown that the theory does not match the empirical data. We suggest that this mismatch possibly arises from unobserved heterogeneity and we examine via a numerical example whether increased heterogeneity reduces, as expected, the accuracy of the estimated defaults. The extent to which this reduction is economically significant is also considered. We then suggest a methodology for identifying the heterogeneity parameters and for testing their explanatory power.