Does FinTech reduce human biases? Evidence from advisory vs. automated FinTechs in lending

Abstract

We investigate whether FinTech can mitigate human biases in lending decisions using proprietary loan-level data from a Chinese auto equity lender. The lender first integrated big data credit scoring as an advisory tool to enhance its traditional lending model, subsequently transitioning to algorithmic decision-making with optional human override. Our findings reveal that cognitive biases decrease significantly when loan officers use algorithmic lending decisions, substantially reducing disparities in loan-to-value ratios between local and nonlocal borrowers without exacerbating default differentials. Notably, the discretionary adjustments made by loan officers remain modest. In contrast, advisory credit scores alone exhibit no discernible bias-reducing effects. Our study is among the first to demonstrate that automation and choice architecture – specifically, nudging via algorithmic defaults – is more effective than mere information provision in combating discrimination and promoting financial inclusion.