Working Paper: NBER ID: w31154
Authors: Mark J. Johnson; Itzhak Bendavid; Jason Lee; Vincent Yao
Abstract: FinTech lending—known for using big data and advanced technologies—promised to break away from the traditional credit scoring and pricing models. Using a comprehensive dataset of FinTech personal loans, our study shows that loan rates continue to rely heavily on conventional credit scores, including 45% higher rates for nonprime borrowers. Other known default predictors are often neglected. Within each segment (prime/nonprime) loan rates are not very responsive to default risk, resulting in realized loan-level returns decreasing with risk. The pricing distortions result in substantial transfers from nonprime to prime borrowers and from low- to high-risk borrowers within segment.
Keywords: No keywords provided
JEL Codes: G21; G23; G50
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| reliance on FICO scores (G21) | collider in causal graph (C22) |
| neglected predictors of default (G33) | confounders for nonprime borrowers' risk (G21) |
| pricing discrepancies between prime and nonprime loans (G21) | borrower outcomes (G51) |
| FICO scores (G51) | fintech loan pricing (G13) |
| higher FICO scores (G51) | lower interest rates (E43) |
| nonprime borrowers (G21) | 45% premium in loan pricing (G19) |
| higher credit quality borrowers (G51) | subsidize lower credit quality borrowers (H81) |
| average overpayment by nonprime borrowers (G51) | systematic mispricing (G19) |