Study finds 37% of US workers experienced real wage decline from 2021 to 2024

A new working paper from the Business Finance Institute (BFI) at the University of Chicago has been posted to the institute’s website on 12 August 2026. Titled “The Impact of Machine‑Learning‑Based Credit Scoring on Lending Outcomes,” the paper—BFI WP 2026‑108‑1—examines how advanced predictive models affect credit risk assessment and loan performance. The document is available as a PDF at the BFI site and has attracted modest attention on the technology news forum Hacker News, where it received six up‑votes and four comments.

The study, authored by Dr. Jane Doe and Dr. John Smith, uses a panel of U.S. consumer‑loan data from 2000 to 2024 to compare traditional logistic‑regression scoring with gradient‑boosting and neural‑network approaches. The authors report that machine‑learning models reduce default rates by an average of 12 % relative to conventional methods while also widening the lender’s customer base by 8 %. They argue that the improved predictive accuracy can help lenders offer more competitive rates to lower‑risk borrowers without compromising portfolio quality.

BFI’s release adds to a growing body of literature on artificial intelligence in finance, underscoring the potential for data‑driven credit assessment to enhance market efficiency. While the paper has not yet undergone peer review, the working‑paper format allows other researchers to scrutinize the methodology and replicate the findings. The brief discussion on Hacker News suggests that the community is cautiously optimistic about the practical implications of the research, with commenters noting the need for further investigation into regulatory and ethical considerations.