Reproduce “Escaping the Subprime Trap in Algorithmic Lending” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Adam Bouyamourn, Alexander Williams Tolbert
Bouyamourn and Tolbert present a theoretical framework demonstrating how mainstream banks can fall into a "Subprime Trap" by denying loans to minority groups due to inflated prior beliefs about repayment variance, perpetuating disparities. They show that finite, temporary subsidies can induce banks to lend, acquire crucial data to correct misperceptions, and ultimately lead to long-term equitable loan access. Empirical analysis of 2024 HMDA data supports this by showing Black applicants with missing income data faced 31.5 percentage points higher loan denial rates than White applicants.
Key insights
- Risk management constraints, such as VaR, can induce inequalities in loan approval decisions even for equally creditworthy applicants, as mainstream banks tolerate less perceived variance than subprime banks.
- A "Subprime Trap" equilibrium exists where mainstream banks perpetually deny loans to minority groups due to inflated prior variance beliefs, preventing any data acquisition and self-correction.
- Learning through lending is critical; banks can converge to correct actual risk perceptions if induced to lend and observe repayment outcomes, thereby breaking cycles of exclusion.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.