In this part of the book we deal with the problem of creating a ‘good’ joint distribution of risk factors that takes into account (‘knows about’) the exceptional events which have informed the construction of our Bayesian net.

Building a ‘good’ distribution entails two different things.

First, we must make sure that we combine (‘splice together’) the normal-times and excited distributions in an effective manner. As we shall see, there is no unique, ‘right’ way to do so, as any choice will depend on how we think that normal and exeptional regimes interact. (If this statement is somewhat cryptic now, it will become clear in Section 20.5.)

Second, a good spliced joint distribution will have to respect some fundamental, if somewhat stylized, relationships between risk and return. In Chapter 21 we shall therefore employ in a normative manner a simple asset-pricing model such as the CAPM in order to guide us in the construction of the spliced distribution. In doing so, we do not subscribe to the view that a model such as the CAPM is ‘right’ (whatever that means), but that it is ‘reasonable’ (whatever that means).