Scoring Solutions
In my last post, I discussed the specialized logic in the FICO® Score that accounts for loan rate shopping by consumers. As a quick refresh, this includes the score treating all student loan, auto and mortgage inquiries within a 45-day period as a single inquiry—what’s called an inquiry “deduplication (dedupe) window.” Naturally, I’m often asked: why is this dedupe window 45 days? The answer: because the data tells us so.
Here’s a bit of background. During an earlier FICO® Score redevelopment, we revisited the inquiry logic to determine if the dedupe window, which at the time was 14 days, remained ideal for risk prediction. During this time, consumers were becoming more financially savvy, and spending more time in their search for the best interest rates for mortgage and auto loans. We realized that if the 14 day window was too short, too many inquiries would be counted, which would excessively penalize consumers and yield a slightly less predictive score characteristic.
To evaluate whether to lengthen the dedupe window, we analyzed 14, 21, 30, 45, 60 and 90 day windows, and measured the resulting impact to predictiveness. Specifically, we looked at information value, a statistic that measures how well a given characteristic separates goods from bads.
In general, our results showed that by expanding the length of the dedupe window, the characteristic was better at predicting risk. Figure 1 below is based on the performance of all credit accounts on the consumer’s file. Figure 2 is based on the performance of new accounts only. We repeated the same analysis for new accounts (those opened within 6 months following the scoring date), since inquiries can be more relevant in an originations context. Risk patterns are fairly consistent in both contexts.
For the total population, increasing the dedupe window leads to a slightly stronger characteristic—but there is an upper limit to these gains. There is marginal improvement in using either 60 or 90 day windows.
For the clean population (roughly 70% of consumers), the ideal window is 45 days, for both the all accounts and new accounts analyses. Interestingly, we see that the ideal dedupe window for the rest of the population—those with at least one derogatory event on their credit history—may be greater than 45 days. This is intuitive since consumers with blemished credit histories may require more time to find and secure credit.
Ultimately, the 45 day window was chosen. It was more predictive than the prior 14 day window, and the longer 60 and 90 day windows were not substantially more predictive. In fact, for the clean population, the longer window had little predictive merit. Additionally, it would be impractical and confusing to consumers if different windows were used to assess inquiries. Based on this research, we rolled out changes to the FICO® Score at all three credit bureaus.
This illustrates the type of research that we undertake regularly to ensure that the logic underlying the FICO® Score remains sound. And of course, we look beyond just the treatment of inquiries. Our credit scoring models are periodically redeveloped to account for any notable changes in risk patterns. This ensures that the FICO Score is adapting when consumer credit behavior evolves or new financial products are introduced. Our goal continues to be producing the most consistently predictive score possible.
