Travelers Insurance - Vice President of Analytic Strategy
From 2003 to 2008, Fannie Mae and Freddie Mac wrongly estimated the default risk on $5 trillion of residential mortgages. This error is now acknowledged as being instrumental in creating the current recession.
We will begin with a brief introduction to the size and importance of financial services and then mention some well know commercial failures, failures that have a common cause. Behind each of the disasters is the failure of the underlying statistical models to properly estimate the risk.
The Financial services industry is in an exciting transition as it learns how not just to do numerology, but to effectively and efficiently build and use statistical models. An essential element of this transition is the increasing sophistication, and increasing adoption, of formal statistical corporate governance. Corporate governance is the set of rules, enforced on management by the board, that directly addresses topics for which management is not economically incented. Typical corporate governance examples include finance, HR, and audit. When statistical governance is done well, it enables diverse analytic professionals to rapidly build high performing, trustworthy models. If Fannie and Freddie had had a mature statistical governance process, there would have been enough pressure on the irrational exuberance in housing prices to have induced appropriate modulation.
The core of the presentation will be a brief introduction to best practices in statistical model governance: ownership, review, documentation, and process. The reason for each component will be highlighted with examples of failures. The value of the statistical governance process will be highlighted with examples of somewhat unsung, but transformationally important, successes. For example, the broad availability of unsecured credit and the ensuing increase in street safety resulting from people no longer carrying material amounts of cash is a consequence of statistical governance enabling safe and effective models.
Through the last century, the profession of statistics has been instrumental in the development of many revolutionarily important industries including agronomy, pharmaceuticals, semiconductors, and genomics. In this long tradition, the profession is now becoming instrumental in the creation of a reliable and effective financial services industry.