We investigate whether FinTech alleviates human biases in lending decisions under asymmetric information. Using proprietary data from a large auto equity loan company in China, we find that nonlocal borrowers obtain a smaller loan-to-value (LTV) ratio than their local counterparts, even after controlling for collateral value and other bor- rower characteristics. Using two quasi-experiments where the lender adopts different financial technologies, we find that replacing human decision-making with FinTech al- gorithms significantly reduces both the LTV ratio and default rate differences between local and nonlocal borrowers, mitigating lending biases against nonlocal borrowers. However, the introduction of FinTech credit scores to assist human decision-making through information provision has no impact. Our results thus demonstrate the po- tential of algorithms in correcting human biases and promoting financial inclusion.
Keywords: FinTech, algorithm, human biases, lending, quasi-experiments
JEL classification: G21, G41, G51
Download the full paper：../../docs/20230320095046504176.pdf