Are FICO Scores “Artificially Inflated?”

A recent Bloomberg article asserted that “consumer credit scores have been artificially inflated over the past decade,” as credit scores have steadily increased over the past decade of economic expansion.  The conclusion cited is that “debtors are riskier than their scores indicate because the metrics don’t account for the robust economy, skewing perception of borrowers’ ability to pay bills on time”.

So are FICO^® Scores “artificially inflated?"  The simple answer is no.

FICO Scores Are Not Fixed Estimates of Credit Risk

The FICO^® Score is designed to rank-order the likelihood that a borrower will repay their loan(s), with higher scoring borrowers representing lower risk, and lower scoring borrowers representing higher risk.  The aim of the FICO^® Score is to ensure that a pool of borrowers scored as a 660 at a given point in time represent lower risk of default than a pool of borrowers scored as a 620 at that same point in time.  Figure 1 provides a conceptual example of this: for every one borrower at a score of 660 that goes on to  not pay their loan as agreed, there are 10 borrowers at that score level that will.  This 10:1 odds of repayment at a 660 is more than triple the odds of repayment of 3:1 at a 620.

Figure 1. FICO® Score is designed to rank order credit risk (purely illustrative data).

Note the use in the paragraph above of the phrase “at a given point in time”.  As covered here, the relationship between repayment odds and the FICO^® Score can and does shift over time, largely driven by shifts in factors outside the credit file that also play an important role in consumers’ ability to pay their credit obligations.  As noted in the section “Risk in Bankcard Originations on the Rise” here, this relationship is currently trending downwards for certain products such as bankcards, such that repayment risk at a given FICO^® Score is higher than it was several years prior.   FICO has for many years closely tracked trends in the odds-to-score relationship, and advocated lenders to do the same on their portfolios in case the findings merited an adjustment in underwriting criteria such as the FICO^® Score cutoff used. The Federal Reserve, OCC and FDIC have all recognized the importance of such careful monitoring, and issued regulatory guidance (SR 11-7 and OCC 11-12) that requires lenders to validate their models annually.

Shift in Odds to Score ≠ Score Inflation

Evidence of a changing relationship between the FICO^® Score and repayment odds is not evidence of score inflation.  It is likely indicative of a shift in the aforementioned factors outside the credit file that also play an important role in consumers’ go-forward risk of repayment.  So what are some of these factors?

• Macroeconomic conditions such as unemployment rate and Housing Price Index (HPI) changes.
• Lender underwriting practices, such as underwriting borrowers with higher debt-to-income ratios, lower down payments, or lengthier repayment periods---none of these practices can be identified in the traditional credit bureau file.

FICO has long advocated that regardless of whether the current state of lending is healthy or stressed, it is essential to a robust understanding of borrower repayment risk to consider all of the “four Cs” of credit: past Credit behaviors, Capacity, Collateral, and Conditions (including economic factors).  Over-reliance on any of of these dimensions can lead to a less robust understanding of the true credit risk a borrower represents.  The Bloomberg article cited rising defaults in the subprime auto loan market, while also including a quote from an investor expressing concern that “deep subprime auto lenders may be deeply reliant on credit scores”.  There has been considerable coverage of the trend in auto lending towards loans with longer repayment terms. A number of economists and analysts have noted similar factors in credit underwriting practice in mortgage lending as well, evidenced by higher DTI and LTV ratios, a development that warrants close monitoring.

From an investor perspective, it is also important to ensure that a fresh view of the borrower’s credit standing is also leveraged.  By also utilizing a recently calculated FICO^® Score for each loan/borrower being examined, the investor will avoid basing its risk and valuation assessments on a FICO^® Score pulled back when the loan was originally underwritten, based on credit file information that may now be outdated.

An Effective Rank Ordering Tool Through The Cycle

We regularly track FICO^® Score dynamics and consistently observe that it is highly effective at rank ordering consumer repayment risk (what it is designed to do) through different economic cycles and underwriting environments.  If anything, recent validations indicate that the FICO^® Score is actually more effective at differentiating the repayment likelihood of low and high scorers than it was 10 years ago.  Figure 2 below shows the Kolmogorov-Smirnov (K-S)* statistic---a widely used measure of the rank ordering effectiveness of a predictive model---calculated for the FICO^® Score 8 model over the past decade.  As shown, the K-S of the FICO^® Score has been very stable over the past 6+ years, and is slightly higher today than it was 10 years ago.

Figure 2. FICO® Score 8 Kolmogorov-Smirnov (KS) Statistic on Existing Accounts (All Products) Over Time; ‘defaulters’ defined as 90+ dpd accounts over 24 month outcome period subsequent to Scoring Date.

FICO^® Scores have improved for many consumers over the past 10 years.  What has driven the improvement in national average FICO^® Score over that period is improvement in the underlying consumer credit data.  We have observed fewer credit blemishes and higher available credit lines (which can serve as safety nets in times of temporary financial strain), both of which reflect improving consumer financial health as would be expected during a period of economic expansion.  While the relationship between the FICO^® Score and repayment odds can and does continue to shift, prudent lenders and investors who leverage the full breadth of information available to them as they assess borrower credit risk are well-positioned to navigate the road ahead.

*This statistic is the maximum difference between the cumulative distributions of non-defaulters and defaulters. A zero value indicates that the credit score fails to differentiate between defaulters and non-defaulters; a value equal to 100 indicates that the credit score perfectly differentiates defaulters from non-defaulters