The Federal Reserve Bank of Philadelphia has released a working paper, “Credit Risk Modeling in Segmented Portfolios: An Application to Credit Cards,” that analyzes credit card default rates over the last 15 years, segmented by FICO score. This is one of the best longitudinal studies we have seen on this topic. It seeks to explore the question of how much one can expect default rates to increase in a recession, and how sensitive the change is to credit characteristics of the borrower. The paper also contains a review of related literature and studies on the topic, and empirically compares various models’ performance through the Great Recession.
The key finding from this study is in the following sentence from the abstract:
“Prime segments are proportionally more severely impacted by a downturn in economic conditions relative to the subprime or near-prime segments.”
The researchers examined a large set of credit card account performance data from Equifax from 2001 to 2011, and segmented annual default rate and loss rates by FICO score, total credit limits, and utilization rates. As expected, they found that subprime borrowers uniformly have higher annual default rates than near-prime borrowers, who have higher default rates than prime borrowers, etc. As well, those borrowers with very low credit limits have higher default rates than those with high credit limits; and those with high utilization rates have the highest default rates.
But the somewhat surprising finding was that as you examine the pre-recession default rate baseline for each of these cohorts, and compare to the peak default rate in 2009, prime and super-prime borrowers showed a higher increase in default rates (about 3x) compared to near-prime and subprime borrowers. The paper further examines the intersection of credit variables such as utilization and credit score. For example: “Low utilized accounts among subprime borrowers are 46 percent less likely to default (odds ratio of 0.54) relative to high utilized accounts among the same group of borrowers, but this effect is even stronger among the near-prime and prime borrowers, in which the likelihood of default for low utilized accounts is 67 percent and 73 percent lower, respectively.”
We believe this study should help investors in this asset class develop a range of expectations regarding the magnitude of increase in defaults in a recession relative to today, and help them better anchor those expectations to the credit characteristics of their portfolio. It also highlights the need for independent loss modeling for borrowers at different points on the credit spectrum, even in a benign economic environment.