Thu 01 May 2025 - In capital modelling, one often makes use of sets of pre-simulated outputs from external models, such as ESGs & cat models. Each output has a joint distribution which must be preserved within the calculation kernel. This constrains the additional correlations which can be introduced between the sets, and the variables generated in the calculation kernel.
Market practice is to specify a single correlation between the pre-simulated output & the calculation kernel. One then faces the difficult task of reducing the dependence relationship between ESG & calculation kernel to an assumption between one material pair of variables—e.g. Equity market index & Financial Lines loss ratio. Such an approach can also result in unintended correlation between remaining, unspecified variables.
This talk introduces a method using a vine copula, whereby one may specify, within constraints, additional correlation assumptions between pre-simulated output and the calculation kernel.
For the case where all vine edges are Gaussian, we solve for the additional correlation which can be introduced using this method.
Our target audience is experienced capital modelling actuaries. The talk assumes familiarity with concepts of copulas, external models, and rank correlation.
Key knowledge takeaway is a method to introduce additional correlations over and above those achievable using the apparent industry standard practice.
Tim Harrison, Liberty Global Group