The Analytics of Risk Model Validation

Wayne Holland [*] and ManMohan S. Sodhi [*]
Under Basel II, banks are encouraged to develop their own advanced measurement approaches (AMA), allowing different approaches to flourish. However, this flexibility may pose a problem for the regulators in approving a particular bank s proposed capital requirement as indicated by the bank s own AMA model. We propose a simple method for regulators to cross-check any bank s proposed capital reserve using a lower bound function on the capital requirement. This bound is a function of the weight of the bank s estimate of its expected losses relative to the pooled loss experience of other banks. The regulator can use the bound to indicate potential problems with an individual bank s AMA model at the line of business level. We illustrate our approach with a bound that is easy to compute. Banks can also use this approach to cross-check their own models.
[*] Cass Business School, City University London, UK
We seek to provide a simple way for regulatory bodies to cross-check the operational risk loss model and the proposed capital requirement at the line of business level for a particular bank by using the pooled loss experience of other banks. For regulators, as well as for any bank developing its operational risk models using advanced measurement approaches (AMA), there are at least two challenges: (1) the variety of sophisticated statistical and other probability-based ideas for modelling and (2) the paucity of loss data available for constructing loss distributions. Our contribution...