数字足迹与违约预测:基于消费信贷的证据  

Digital footprints and default prediction:Evidence from consumer credit

作　　者：胡俊 李强[1] 曾勇[1] 田轩 刘嵩 HU Jun;LI Qiang;ZENG Yong;TIAN Xuan;LIU Song(School of Management and Economics,University of Electronic Science and Technology of China,Chengdu 611731,China;PBC School of Finance,Tsinghua University,Beijing 100083,China;Sichuan XW Bank Corp.,Ltd.,Chengdu 610094,China)

机构地区：[1]电子科技大学经济与管理学院,成都611731 [2]清华大学五道口金融学院,北京100083 [3]四川新网银行股份有限公司,成都610094

出　　处：《管理科学学报》2025年第2期115-139,共25页Journal of Management Sciences in China

基　　金：国家自然科学基金资助项目(71972027,71790591);国家杰出青年科学基金资助项目(71825002);四川省软科学项目(2022JDR0290)。

摘　　要：本研究利用国内某金融科技借贷机构借款人层面的独特数据,在将借款人信息分为自我报告、央行征信、电商信用评分和网购行为共四类信息的基础上,重点考察电商信用评分和网购行为两类数字足迹的违约预测作用及其提高信贷可得性的内在机理.以自我报告和央行征信两类基础信息的预测能力为基准,研究结果表明:两类数字足迹信息能够将违约预测的准确性提高约50%,其中电商信用评分的预测能力明显高于其他三类信息,网购行为足迹不仅能达到与自我报告信息相当的预测效果,而且相对于电商信用评分还具有额外的信息含量.进一步,针对数字足迹提高信贷可得性机理的检验结果表明,利用两类数字足迹进行违约预测,不仅可为传统征信严重缺失的信用白户建立数字信用,还能对信用质量被传统征信错误低估的借款人进行更为准确的风险评估.Using a unique data set obtained from a Chinese fintech lending company,this paper categorizes the information used for loan default prediction into four types:Self-reported information,central bank credit information,e-commerce credit score,and online shopping behaviors.The roles played by the two types of digital footprints,e-commerce credit score and online shopping behaviors,in predicting loan default are investigated and the underlying mechanisms through which digital footprints improve credit availability are explored.Our results show that the digital footprints can help lenders to improve the accuracy in predicting a borrower's default likelihood by about 50%.The prediction power of the e-commerce credit score is significantly the highest among the four types of information.Nevertheless,online shopping behaviors can not only provide additionally useful information beyond the e-commerce credit score,but also perform as well as the self-reported information disclosed by borrowers.Furthermore,the preliminary evidence on the mechanisms through which digital footprints improve credit availability shows that e-commerce credit score and online shopping behaviors can be used to establish credit scores for"unbanked customers"and accurately evaluate the credit worthiness of borrowers whose credit quality is always wrongly underestimated.

关 键 词：金融科技 数字足迹 违约预测 信贷可得性 

分 类 号：F830.589[经济管理—金融学]