Abstract

The peer-to-peer lending industry has experienced recent turmoil, posing risks to fintech companies and banks. Based on a random sample of 33,669 borrowers who had downloaded peer-to-peer lending platforms prior to submitting loan applications to a well-known fintech company, Du Xiaoman Financial (formerly Baidu Finance), this article evaluates the predictive power of borrowers’ internet behaviours on credit default risk. After controlling for borrowers’ basic characteristics that are widely used in academic research and enterprise practices, the coefficients of key factors selected from 3,100 variables are economically and statistically significant. The average Kolmogorov-Smirnov value of the prediction model calculated using the hold-out method is approximately 37.09%. The results remain robust in several additional analyses. This study indicates the importance of non-credit information, particularly borrowers’ internet behaviours, in supplementing borrowers’ credit records for both fintech companies and banks.

摘要

点对点贷款行业最近经历了动荡，对金融科技公司和银行构成了风险。基于对33,669位借款人的随机样本，这些借款人在向著名的金融科技公司Du Xiaoman Financial（原百度财务）提交贷款申请之前已经下载了对等借贷平台，本文评估了借款人互联网的预测能力信用违约风险行为。在控制了在学术研究和企业实践中广泛使用的借款人的基本特征之后，从3,100个变量中选择的关键因素的系数在经济和统计上都是有意义的。使用保留方法计算的预测模型的平均Kolmogorov-Smirnov值约为37.09％。在其他一些分析中，结果仍然很可靠。