Fintech firms that use machine learning and “non-traditional” data sources may be able to generate a more accurate picture of the credit risks posed by their customers, a working paper published by the Bank for International Settlements finds.

Leonardo Gambacorta, Yiping Huang, Han Qiu and Jingyi Wang study transaction data from a “leading Chinese fintech company”, which asked to remain anonymous. This allows the authors to separate traditional data (such as credit card information) from more