## ISSUES IN MODEL RISK

**the elephant in the data centre**

**(see below)**

## .................................."Mitigating Model Risk" (from the Xenomorph blog) <the starter for 10>

**some useful pointers (link) [1]:**

**UPDATED (FEBRUARY 2012) TO BECOME **

**A GENERAL MODEL RISK ISSUES REPOSITORY**

although the initial blog post was a simple reference to a very useful blog post in the Xenomorph blog, I want to update & develop this page to include more general references to issues of model risk & in particular the engineering type model risk which can flow from technical distinctions between statistical products, in this instance with particular focus on the R & SAS applications.

**As I have said in a related blog post (here [2]) **

Doing Quantitative Stuff accurately means doing it properly first time around not “falderal” in big meeting rooms with laptops. Investing in it properly, modeling over time; requires ‘proper’ Computing Power. Actually being able to execute and test in one day DOES make a difference to everyone from the beginning. Particularly when one does have to engage colleagues who may not be as mathematically literate as necessary in the first instance. But that is just the way it is today; there is a set with an undefined taxonomy of complex statistical values which constitutes a fundamental ‘social value’ today; i.e. that set of risk values which is socialised as necessary and sufficient to validate the 'safety' of a financial institution, it is a sufficient social risk in my view that such is set is actually not clearly defined either by accounting standards or the Basel or the Solvency rules (since by not clearly defining the standards opens the doors to corruption).

**June 2012**