Economically calibrating risk predictions

While there are many commentators trying the define what the “new normal” will look like in a post-crisis economy, there’s still a lot of new learning to be done and not much in the way of economic consistency. So it seems a good time to talk about companies that have stopped waiting for the dust to settle and started moving ahead.

One such FICO client is a leading European banking group that was experiencing a problem common to many affected by the global financial crisis: It couldn't recalibrate its models fast enough to keep up with the increasing risk—despite monthly updates.

Decisions were being made based on relationships between risk score and default rate that were no longer reliable for predicting customer behavior. Delinquencies were soaring. And the situation was creating a regulatory issue, since the company was having difficulty achieving stability in the overall capital requirements for its retail group.

This dilemma makes clear the inadequacies of traditional methods of recalibrating models, which rely on backward-looking, point-in-time techniques. Instead, the company is now implementing methods that anticipate where it’s headed. The key is economically calibrated models that predict shifts in default rates at specific score bands under various future economic scenarios.

If the company is forecasting deteriorating economic conditions, it now has a scientific method of determining how much to raise score cutoffs in its strategies to keep default rates steady. If the forecast is for improving conditions, the company can lower its cutoffs in a timely way to avoid depressing revenue by being overly conservative.

The new approach, which was tested on credit cards, personal loans and mortgages across the group, was first implemented in one pilot country. In that successful debut, calibrations took into account not only statistically valid macroeconomic time-series characteristics such as household savings and debt, unemployment rate/growth in employment, average monthly oil price and changes in home prices, but also currency relationships affecting mortgage payment behavior.

The project has now expanded to 15 markets across Europe, with models being calibrated to the different economic prospects of each. By adjusting its policies flexibly to economic developments in these specific markets, the company expects to be able to "grow in good and bad economic periods."

FICO is also helping the banking group improve availability of capital to fuel growth by applying economic impact analysis at the portfolio level. In stress tests, the lender's portfolio-level projections of the impacts of future economic scenarios are now firmly rooted in account-level economically calibrated risk scores and decision strategies. The company is able to more accurately project its loss reserve requirements, facilitating regulatory compliance (including with the Basel III counter-cyclical capital buffer requirement), without tying up capital unnecessarily.