Some subprime borrowers get lower-priced credit at LendingClub

AI system

System: LendingClub rating-grade model (grades A to G)
Model: Not assessed
Release: Not assessed
Application developer: LendingClub
Role: Not assessed
Configuration: Not assessed
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: Account-level LendingClub data compared with Y-14M bank data for similar loans, 2007 to 2015 vintages. Lender grading that increasingly used information not captured by FICO.
Human oversight: Not assessed
Relation to AI: The paper attributes the regrading of subprime borrowers to alternative data used by the lender. It does not establish that machine learning produced the grades. LendingClub developed and deployed its own model. The Federal Reserve authors are independent.
AI contribution: Not assessed

Outcome identifiers: Not assessed
Benefit: about 8 percent of A-rated and 28 percent of B-rated borrowers had FICO scores below 680
Category: Increased Inclusion
Status: Not assessed
Beneficiary groups: People. Society.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes
Costs and harms: Not assessed

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: about 8 percent of A-rated and 28 percent of B-rated borrowers had FICO scores below 680
Comparison: classification of borrowers by traditional credit criteria
Results: Reference: classified as subprime by traditional criteria. Result: slotted into “better” loan grades, which allowed them to get lower-priced credit.
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Conditions for replication: Alternative data that predicts repayment beyond FICO. Grades that keep predicting loan performance (the paper reports good prediction over two years after origination).
Replication attempts: 1 follow-up entries. Outcome unresolved: 1.
Later findings: No later finding on LendingClub's grades was found; the 2022 'Invisible Primes' paper concerns a different platform.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Assessed benefits: People. Society.
Depth: Not assessed
Reasons: People: Effects on borrowers are reported as vintage-level shares and spreads, not as outcomes for identifiable individuals. Society: Financial-inclusion effects are discussed at the aggregate level only.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: Independently reported

Kind: Not assessed
Related entries: Not assessed
Field: Consumer credit. Marketplace consumer lending.
Purpose: Assign rating grades and price consumer loans on the LendingClub marketplace.
Dates: Date basis unassessed: 2019 (precision unassessed). Updated: 21 October 2026.
Stage: Not assessed
Version: SHA-256: 3b29bea167a2…