Dr. Andrew Jennings, Chief Research Officer, FICO
The recent economic crisis has led some to wonder whether analytics that measure consumer credit risk are fundamentally broken. I can reliablystate that they're not broken - the robust measures, such as the FICO score, still rank-order consumers by credit risk.But many lenders were caught off-guard when credit default rates rose significantly above the rates historically associated with standard credit score ranges. In 205 - 2006, for instance, a default rate of 2% was associated with a FICO score of approximately 650-670. By 2007 - 2008, this default rate was associated with a score of about 710-720 - a 50 to 60 point shift. In other words, accounts scoring 710 in late 2008 were much riskier than they used to be as a group, behaving more like an "old" 650.
The fact is that while even an economic crisis doesn't shake these scores' ability to distinguish "good" payers from "bad" payers, such a disruptive change does mean you'll see many more "bad" payers than you're used to in a given score range. The important question for lenders is: How many more?
Today we have new analytic methods for forecasting how such macro-economic conditions are likely to impact customer behavior and change our results. These analytics can generate an index of how much default rates are likely to increase or decrease under a range of economic forecasts. In other words, the analytics consider the macroeconomic view of market conditions within the micro-analysis of individual consumer risk. Based on past dynamics, the analytics derive the empirical relationship between the default rates observed at different score ranges (e.g., the risk score's odds-to-score relationship) as seen on the lender's portfolio, and historical changes in economic conditions. Using this derived relationship, lenders can then input current and anticipated economic conditions to project the expected odds-to-score outcome under those conditions.
With this kind of economic impact simulation, lenders can select the economic forecast they believe is probable, then use the associated index metrics to adjust their score cutoffs for credit approval. In this way, it's possible to maintain a fairly consistent default rate across changing economic conditions. This kind of analytic fusion - economic forecast and credit risk assessment - is imperative to help us learn from rather than repeat the recent crisis.