Big Data Analysis with No Digital Footprints Available: Evidence from Cyber-Telecom Fraud

Abstract

Cyber-telecom fraud is an increasingly severe problem globally. We focus on a special type of cyber-telecom financial fraud, in which criminals induce innocent people to borrow online. Since no digital footprints are available for the fraudsters behind the borrowing cases, identifying the fraud is difficult. Using a proprietary dataset of online consumer financing from a large Fintech company in China, we estimate the extent to which an intervention based on big data and machine learning can identify this type of fraud and prevent customers’ financial losses. We find that female borrowers are more likely to become victims of fraud generally, that young and inexperienced users are more likely to become victim of fraud schemes targeting a lack of financial literacy, and that experienced and inexperienced users are equally likely to become victims of fraud schemes targeting overconfidence. Overall, the intervention effectively identifies fraud targeting either financial literacy or behavioral biases. However, it is more difficult to persuade victims of fraud targeting behavioral biases to change their behaviors.

Keywords: Fintech, big data, machine learning, cyber-telecom fraud, Internet finance