Professor Francesco D’Acunto will present the paper “How Costly Are Cultural Biases? Evidence from FinTech”.
We propose a revealed-preferences approach to assess the costs of cultural biases faced by discriminators by comparing peer-to-peer loans the same lenders make alone and after observing suggestions by an automated robo-advisor in India. Discrimination across religions and against low-caste borrowers is substantial and stronger in locations with higher religious and caste animus. When unassisted, lenders face 8% higher default rates and earn 7.3pp lower returns on loans to their favored groups. Biased beliefs about borrowers’ quality can explain our results better than taste bias because (i) lenders barely override robo-advised matches to borrowers against whom they discriminated when unassisted, (ii) incentives do not affect discrimination, and (iii) low-caste lenders discriminate against low-caste borrowers more than higher-caste lenders.