The Basel II IRB approach revisited: do we use the correct model?

In this paper I question whether the risk weights in the advanced (IRB) approach of the Basel 2 regulation are appropriate, on a strictly theoretical ground. The major concern is that the model behind the regulation considers defaults only at the end of the time horizon for which capital is to be held - whereas defaults in the whole time interval should be taken into consideration. This latter approach is represented by a model that is different from the Basel model. It follows, as I show, that the Basel model should be viewed just as a technical tool to turn the expected value of the unconditional loss distribution into a given percentile of the same distribution - making use of conditional (on the systemic factor) default probabilities - and should not be interpreted as describing even 'virtual' firms and asset values. More importantly, I also show that a logical step in the theoretical foundation of the model is missing which raises the question whether the risk weights calculated with the model are indeed appropriate. Due to difficulties in the calculation in the alternative approach of the percentiles of the loss distribution no clean-cut answer is given in this paper.