金融科技与普惠金融：实证论文综述.pdf

Center for Law and Social Science Fintech and Financial Inclusion:A Review of the Empirical Literature(Southern California Law Review(2023)Carter Faust University of Southern California Marshall School of Business Anthony J.Dukes University of Southern California Marshall School of Business and D.Daniel Sokol University of Southern California Gould School of Law;University of Southern California Marshall School of Business Center for Law and Social Science Research Paper Series No.23-16 May 7,2023 Electronic copy available at:https:/ 2022 FINTECH AND FINANCIAL INCLUSION 137 and shocks,and promote economic mobility.4 Thus,the importance of bringing financial services to the unbanked has captured the attention of many researchers.Online platforms have an important role to play in financial inclusion.Numerous studies have demonstrated that fintech services,such as mobile money,digital payment solutions,and digital lending platforms,have the potential to enable account ownership among the unbanked.5 Further research has shown that countrywide fintech adoption can decrease income inequality by up to 23%.6 Overall,research points to the fact that fintech can have a positive impact on financial inclusion,yet the magnitude of its effects are dependent on relevant infrastructure and policies.7 Recently,governments and global organizations have begun to recognize the need for harnessing the power of fintech to promote financial inclusion.For example,the Group of Twenty(“G20”)High-Level Principles for Digital Financial Inclusion emphasize the importance of utilizing fintech to achieve financial inclusion and reduce global income inequality.8 Additionally,the United Nations(“U.N.”)2030 Agenda for Sustainable Development calls for innovation and development of fintech to spur economic growth among emerging and developing countries.9 This research commentary surveys key research related to fintech and its implications for global financial inclusion.Specifically,it provides an overview of studies regarding digital lending,digital payment,and mobile money platforms and how these services can bridge the financial gap for traditionally unbanked and underserved communities.In terms of the legal role through public and private law,it also identifies common concerns and challenges associated with the adoption of fintech,as well as relevant policies to mitigate these concerns and foster financial inclusion.4.See id.at 114.5.See,e.g.,CAMBRIDGE CTR.FOR ALT.FIN.,WORLD BANK GRP.&WORLD ECON.F.,THE GLOBAL COVID-19 FINTECH MARKET RAPID ASSESSMENT STUDY 8(2020),https:/www3.weforum.org/docs/WEF_The_Global_Covid19_FinTech_Market_Rapid_Assessment_Study_2020.pdf https:/perma.cc/M8HN-GXFP.6.See Ayse Demir,Vanesa Pesqu-Cela,Yener Altunbas&Victor Murinde,Fintech,Financial Inclusion,and Income Inequality:A Quantile Regression Approach,28 EUR.J.FIN.86,95(2020).7.See Purva Khera,Stephanie Ng,Sumiko Ogawa&Ratna Sahay,Measuring Digital Financial Inclusion in Emerging Market and Developing Economies:A New Index 1617(Intl Monetary Fund,Working Paper No.21/90,2021).8.GLOB.PSHIP FOR FIN.INCLUSION(“GPFI”),G20 HIGH-LEVEL PRINCIPLES FOR DIGITAL FINANCIAL INCLUSION(2016),http:/www.gpfi.org/sites/gpfi/files/documents/G20-HLP-Summary_ 0.pdf http:/perma.cc/UY3D-HC28.9.U.N.Inter-Agency Task Force on Fin.for Dev.,United Nations Secretary Generals Roadmap for Financing the 2030 Agenda for Sustainable Development 20192021,at 9(2020),http:/www.un.org/sustainabledevelopment/wp-content/uploads/2019/07/EXEC.SUM_SG-Roadmap-Financing-SDGs-July-2019.pdf http:/perma.cc/GJ66-CCRF.Electronic copy available at:https:/ copy available at:https:/ 138 SOUTHERN CALIFORNIA LAW REVIEW POSTSCRIPT Vol.95:PS135 I.DIGITAL CREDIT A.HOW DOES FINTECH FILL THE CREDIT GAP?1.Expanding Credit to Underserved Borrowers Limited access to credit is one of the largest barriers to financial inclusion.Many studies have shown that digital lending platforms fill the credit gap by expanding credit services to traditionally underserved borrowers.Recent research compared account-level data between digital lending platform LendingClub and U.S.banks to examine whether fintech fills credit gaps in regions underserved by traditional banks.The results showed that LendingClub increased credit availability in areas that could benefit from additional credit supply,including both highly concentrated and underserved bank markets.10 Economists built upon this research by studying marketplace lending at the business and consumer levels across 109 countries from 2015 to 2017.They found that marketplace lending was more prevalent in lower-income economies and filled the credit gap when access to traditional banks and lenders decreased.11 2.Providing Alternative Sources of Data Fintech lending platforms often use alternative data sources to evaluate customer creditworthiness,as compared to traditional lenders that use standard measures such as credit score.Common sources of alternative data include utility bills,bank transactions,online footprints,and personal data such as occupation and education information.These alternative information sources can address information asymmetries and thus benefit borrowers who would typically be classified as subprime by traditional lenders.12 Furthermore,the use of data and machine learning by fintech platforms can increase efficiency and thus decrease cost,as opposed to traditional lenders.Fintech lenders can process mortgage applications 20%faster than traditional lenders without compromising on default prediction accuracy.13 Numerous studies examine the accuracy of loan prediction by fintech 10.Julapa Jagtiani&Catharine Lemieux,Do Fintech Lenders Penetrate Areas that Are Underserved by Traditional Banks?,100 J.ECONS.&BUS.43,53(2018).11.Majid Bazarbash&Kimberly Beaton,Filling the Gap:Digital Credit and Financial Inclusion 1920(Intl Monetary Fund,Working Paper No.20/150,2020).12.Julapa Jagtiani&Catharine Lemieux,Fintech Lending:Financial Inclusion,Risk Pricing,and Alternative Information 3437(Fed.Rsrv.Bank of Phila.,Working Paper No.17-17,2017).13.OECD,DIGITAL DISRUPTION IN BANKING AND ITS IMPACT ON COMPETITION 12(2020)http:/www.oecd.org/competition/digital-disruption-in-banking-and-its-impact-on-competition-2020.pdf http:/perma.cc/DG5U-ZS47.Electronic copy available at:https:/ copy available at:https:/ 2022 FINTECH AND FINANCIAL INCLUSION 139 lenders.Although the correlation between Fair,Isaac and Company(“FICO”)scores(used by traditional banks)and rating grades(used by fintech platform LendingClub)decreased from 80%in 2007 to 35%in 2015,rating grades continued to serve as an accurate predictor for loan default.14 This highlights how alternative data sources and machine learning used to assign rating grades within fintech platforms can serve as accurate predictors of loan delinquency,even for borrowers who lack traditional indicators of creditworthiness like credit scores.Further research demonstrated that applying big data methods to credit screeners significantly strengthened the lenders accuracy of loan default prediction.Predicted default probabilities decreased most among small businesses and lower-tier cities that previously faced information disadvantages during risk assessment by traditional firms.15 Given that lower risk of default increases the likelihood of obtaining loans,their work suggests that the information advantage provided by fintech can play a key role in expanding credit access to underserved borrowers.B.BIAS IN FINTECH LENDING 1.Evidence of Discriminatory Lending In recent years,there has been increased debate as to whether discriminatory lending exists within fintech.While digital lending platforms have been shown to increase credit accessibility,there is a long history of discrimination in the lending industry.Fair-lending laws in many countries are designed to prohibit biased lending,yet there is still significant evidence that some lenders discriminate on the basis of characteristics such as gender and race.16 Researchers compared lending discrimination among fintech and traditional lenders.They merged data on government-sponsored enterprise(“GSE”)and Federal Housing Administration(“FHA”)loans with information on borrowers race and ethnicity.The results showed that fintech lenders charged Black and Latinx borrowers higher rates for FHA and GSE purchase loans,as well as GSE refinance loans,highlighting similar rate disparities for minority borrowers among traditional and fintech lenders.17 A similar study analyzed crowdfunding projects launched on the platform 14.Julapa Jagtiani&Catharine Lemieux,The Roles of Alternative Data and Machine Learning in Fintech Lending:Evidence from The LendingClub Consumer Platform 26(Fed.Rsrv.Bank of Phila.,Working Paper No.18-15,2019).15.Yiping Huang,Longmei Zhang,Zhenhua Li,Han Qiu,Tao Sun&Xue Wang,Fintech Credit Risk Assessment for SMEs:Evidence from China 3335(Intl Monetary Fund,Working Paper No.20/193,2020).16.See,e.g.,Robert Bartlett,Adair Morse,Richard Stanton&Nancy Wallace,Consumer-Lending Discrimination in the Fintech Era,143 J.FIN.ECONS.30,5556(2022).17.See id.Electronic copy available at:https:/ copy available at:https:/ 140 SOUTHERN CALIFORNIA LAW REVIEW POSTSCRIPT Vol.95:PS135 Kickstarter to test for bias against minority founders.The analysis proved that,compared to non-Black founders,Black founders raised on average 86.1%less for their projects,and prospective supporters held an unconscious bias against Black founders.18 Even when sensitive attributes like race and gender are not explicitly used as inputs for machine learning algorithms,they can be correlated with other input features which affect the prediction outcomes.19 Factors such as education level,gender,and income can influence perceptions of borrower trustworthiness and thus bias lending decisions.An analysis of 247,115 loans on Renrendai,one of the largest debt crowdfunding platforms in China,revealed that borrowers regional social capital had a positive relationship with funding success and loan size.20 The findings demonstrate how fintech lenders utilize alternative soft information and personal perceptions to inform their lending decisions and that these factors can be used to bias lending.2.Debiasing with Artificial Intelligence Improvements in machine learning that remove potential for biased decision-making hold promise in increasing credit opportunities for historically disadvantaged borrowers and contribute to financial inclusion.Evidence of lending discrimination has inspired researchers to examine ways in which machine learning algorithms can be designed to correct for bias.A recent study proposed a debiasing algorithm that makes input features independent of sensitive attributes and applied it to a previously biased machine learning algorithm for a peer-to-peer lending platform.The results showed that the differences in the probability of being funded between male and female,as well as non-Black and Black borrowers,were statistically insignificant,indicating the debiasing algorithm effectively removed previous biases.21 Likewise,fintech lending platforms may only be harmful to minority groups if the algorithm includes both a proxy for group membership and explicitly contains prejudice against the group.Research found that removing these proxies and developing debiasing algorithms can mitigate bias if it exists and ultimately benefit minority groups that are historically 18.Peter Younkin&Venkat Kuppuswamy,The Colorblind Crowd?Founder Race and Performance in Crowdfunding,64 MGMT.SCI.3269,327374(2018).19.Runshan Fu,Yan Huang&Param Vir Singh,Crowds,Lending,Machine,and Bias,32 INFO.SYS.RSCH.72,8889(2021).20.Iftekhar Hasan,Qing He&Haitian Lu,Social Capital,Trusting,and Trustworthiness:Evidence from Peer-to-Peer Lending,57 J.FIN.&QUANTITATIVE ANALYSIS 1409,1449(2022).21.See Fu et al.,supra note 19,at 7274.Electronic copy available at:https:/ copy available at:https:/ 2022 FINTECH AND FINANCIAL INCLUSION 141 harmed by bias in traditional lending.22 A similar study revealed that by correcting for soft information bias and using hard information more effectively,machine learning algorithms can be fairer for disadvantaged borrowers than traditional lenders.23 II.DIGITAL PAYMENTS A large body of literature examines the ways in which digital payment services enable financial inclusion.Digital payment technologies support services such as salary disbursements,bill payments,peer-to-peer transfers,credit payments,and consumer-good payments.Quasi-experimental research on the impacts of global mobile money services has shown that digitizing payments has positive effects on the value chain and can benefit disadvantaged groups.24 A.THE ROLE OF GOVERNMENTS IN DIGITIZING PAYMENTS Several studies have examined the effects of digitization of government payments.Digital government payments have particularly large implications for overall fintech adoption,as a government shift toward digital finance could push private sectors to do the same.On a broader scale,transitioning to digital government-to-person(“G2P”)payments can have large effects on incorporating the unbanked into the financial system.Digital G2P payments may facilitate account ownership among 160 million currently unbanked adults who receive government payments exclusively in cash.25 Digitizing G2P payments has been shown to benefit both governments and recipients by decreasing costs and increasing efficiency.For example,the Mexican government decreased its spending on G2P payments by 3.3%annually by shifting to digital payments.26 Similarly,analysis of a social transfer program in Niger found mobile transfers decreased variable cost by 20%.27 Digital payments also benefit recipients by decreasing travel time to collect payments,which translates to saved money in terms of travel 22.See Thomas Philippon,On Fintech and Financial Inclusion 17(Natl Bureau of Econ.Rsch.,Working Paper No.26330,2019).23.See Prasanna Tantri,Fintech for the Poor:Financial Intermediation Without Discrimination,25 REV.FIN.561,590(2021).24.Yan Dong,Moonwon Chung,Chen Zhou&Sriram Venkataraman,Banking on Mobile Money:The Implications of Mobile Money Services on the Value Chain,21 MFG.&SERV.OPERATIONS MGMT.290,30506(2019).25.Leora Klapper&Dorothe Singer,The Opportunities and Challenges of Digitizing Government-to-Person Payments,32 WORLD BANK RSCH.OBSERVER 211,217(2017).26.Id.at 213.27.Jenny C.Aker,Rachid Boumnijel,Amanda McCelland&Niall Tierney,Payment Mechanisms and Antipoverty Programs:Evidence from a Mobile Money Cash Transfer Experiment in Niger,65 ECON.DEV.&CULTURAL CHANGE 1,5(2016).Electronic copy available at:https:/ copy available at:https:/ 142 SOUTHERN CALIFORNIA LAW REVIEW POSTSCRIPT Vol.95:PS135 expenses and lost wages.Researchers further examined these benefits by analyzing a government-sponsored debit-card program in Mexico.In 2009,the government issued debit cards to members of Prospera,a cash-transfer program for und