15:40 1/4/2019 RFS-OP-REVF180102.tex Page: 1939 19391982 Marketplace Lending: A New Banking Paradigm? Boris Valle Harvard Business School Yao Zeng University of Washington Marketplace lending relies on screening and information production by investors, a major deviation from the traditional banking paradigm. Theoretically, the participation of sophisticated investors improves screening outcomes and also creates adverse selection among investors. In maximizing loan volume, the platform trades off these two forces. As the platform develops, it optimally increases platform prescreening intensity but decreases information provision to investors. Using novel investor-level data, we find that sophisticated investors systematically outperform, and this outperformance shrinks when the platform reduces information provision to investors. Our findings shed light on the optimal distribution of information production in this new lending model. (JEL G21, G23, D82) ReceivedAugust29,2018;editorialdecisionJuly1,2018byEditorItayGoldstein.Authors have furnished an Internet Appendix, which is available on the Oxford University Press Web site next to the link to the final published paper online. Lending marketplaces, also commonly referred to as peer-to-peer lending platforms, such as Lending Club, Prosper, and Funding Circle, have been rapidly gaining market share in consumer and small business lending over We are grateful to Itay Goldstein, Wei Jiang, Andrew Karolyi (the Editors), Shimon Kogan, Gregor Matvos, Adair Morse (discussants), Chiara Farronato, Simon Gervais, David Scharfstein, Andrei Shleifer, Jeremy Stein, and Xiaoyan Zhang and three anonymous referees for helpful comments. We also thank seminar and conference participants at the RFS FinTech Workshops at Columbia University and Cornell Tech, NY Fed/NYU Stern Conference on Financial Intermediation, NYU Stern FinTech Conference, Texas Finance Festival, Bentley University, Chinese University of Hong Kong Shenzhen, London School of Economics, Peking University, Tsinghua PBC School of Finance, the University of Oregon, and the University of Washington. The authors have receivedin-kindsupportfromLendingRobotforthisprojectintheformofaproprietarydatasetofLendingRobot transactions. These data were provided without any requirements. We thank Yating Han, Anna Kruglova, Claire Lin, Kelly Lu, Lew Thorson, and, in particular, Botir Kobilov for excellent research assistance. All remaining errors are our own. Supplementary data can be found on The Review of Financial Studies Web site. Send correspondence to Boris Valle, Harvard Business School, Baker Library 245, Boston, MA 02163; telephone: 617-496-4604. E-mail: bvalleehbs.edu. The Author(s) 2019. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For permissions, please e-mail: . doi:10.1093/rfs/hhy100 Downloaded from by Renmin University user on 05 December 2019 15:40 1/4/2019 RFS-OP-REVF180102.tex Page: 1940 19391982 The Review of Financial Studies / v 32 n 5 2019 the last decade. 1 This rapid development has important implications on the consumer lending market, and more broadly on banking. 2 Designed as a two-sided platform, a lending marketplace bears no balance sheet exposure to the issued loans but brings innovations to traditional banking on both the borrower and investor side. The innovation on the borrower side relies on low-cost information technology to collect standardized information from dispersed individual borrowers on a large scale. With such information, lending platforms prescreen loans in the sense that they use scalable algorithm to gauge the riskiness of the underlying loan applications, list some of them on the platform, and allocate the listed applications into risk buckets. Innovation on the investor sidethat is, the focus of this paperis equally important and complementary. Instead of pooling loans and issuing safe, information-insensitiveclaims,platformsprovidedetailedinformationoneach prescreened loan application to investors, relying on investors to further screen individual borrowers and to directly invest in individual loans. These investors include informationally sophisticated investors, both among retail and institutional investors, as well as passive and unsophisticated investors. 3 These diverse investors fully bear loan risks and perform large-scale borrower screening, and the resultant joint information production between the platform and investors challenges the traditional role of banks as being the exclusive informationproduceronbehalfofinvestors(GortonandPennacchi1990;Dang et al. 2017). The joint information production between the platform and investors of varying sophistication poses several research questions, which we address in thispaper.First,aremoresophisticatedinvestorsonlendingplatformsscreening borrowers more intensively and thereby systematically outperforming less sophisticated investors? If so, how does sophisticated investor outperformance relate to the platforms prescreening and information provision? Finally, given the heterogeneity of investors, what is the optimal platform design in terms of loanprescreeningandinformationprovisiontoinvestorstomaximizevolumes? Answering these questions is essential for understanding how platform and investor information production interact with each other. This understanding could further speak to the promises and pitfalls of marketplace lending. Ourstudyisalsomotivatedbyapuzzlingeventduringthedevelopmentofthe marketplace lending industry. On November 7, 2014, Lending Club, the largest lendingplatform,removedhalfofthe100variablesonborrowercharacteristics 1 Loans issued by these platforms represented one-third of unsecured consumer loans volume in the United States in 2016. Their revenues are predicted to grow at a 20% yearly rate over the next 5 years (see IBIS World 2017). 2 Currently,morethanhalfofretailborrowersonlendingplatformsusemarketplaceloanstorefinanceexistingloans or pay off credit balances. Although the refinancing market represents an attractive entry point for platforms to achieve scale rapidly, it does not represent their whole addressable market, which includes any form of consumer or small business loan. 3 The platforms are typically segmented between retail and institutional investors and randomly allocate loans between the two pool of investors. Platforms in general offer a passive investing feature as well. 1940 Downloaded from by Renmin University user on 05 December 2019 15:40 1/4/2019 RFS-OP-REVF180102.tex Page: 1941 19391982 Marketplace Lending: A New Banking Paradigm? that it previously provided to investors. This change was unanticipated and surprisedmanymarketparticipants,becauseitwastheonlyinvestor-unfriendly moveinLendingClubshistory. 4 Giventhatinformationtransparencyiscrucial for investor screening, what is the economic rationale behind this reduction of the information set provided to investors? In addressing the previously mentioned questions, we develop a model for marketplace lending and test its predictions using a novel data set that includes borrower and investor data. To start, we theoretically argue that informationally sophisticated investors actively use information provided by the platform to screen listed loans (beyond platform prescreening) and identify good loans to invest. In contrast, unsophisticated investors do not screen; they invest in a listed loan passively if the platform prescreens intensely enough so that they can break even on average, or they do not invest at all. Hence, sophisticated investors outperform unsophisticated ones. Because sophisticated investors can identify good loans and finance them, their participation helps boost the volume of loans financed on the platform when unsophisticated investors are reluctant to invest. However, the heterogeneity in investor sophistication creates an endogenous adverse selection problem among investors, which can hurt volume. Because sophisticated investors can identify and finance good loans before unsophisticated investors invest, sophisticated investor participation lowers the average quality of loans eventually facing unsophisticated investors. Being aware of this adverse selection problem, unsophisticated investors require a higher interest rate, or equivalently a lower loan price to break even, resulting in a higher prevailing interest rate on the platform. This higher interest rate (i.e., lower loan price) reduces the amount of loan applications on the platform, hurting volume. If adverse selection becomes too severe, unsophisticated investors may not break even and may exit the market as a whole, leading to even lower volume. Hence, to maximize volume, the platform optimally trades off these positive and negative effects of sophisticated investor participation. When platform prescreening cost is initially high, the platform optimally chooses a low prescreening intensity but distributes more information to investors. This environment encourages sophisticated investor participation, boosting volume even if unsophisticated investors do not participate. When platform prescreening cost becomes low as the platform develops, it optimally reverses thepoliciesbychoosingahighprescreeningintensitysuchthatunsophisticated investors are willing to invest, but at the same time distributes less information to mitigate the adverse selection caused by sophisticated investors. Testing the model predictions crucially relies on data of investors of heterogenous sophistication. Although borrower-level data are made public 4 We provide more relevant institutional details, including about this specific event in Section 1.2. 1941 Downloaded from by Renmin University user on 05 December 2019 15:40 1/4/2019 RFS-OP-REVF180102.tex Page: 1942 19391982 The Review of Financial Studies / v 32 n 5 2019 by the platforms, data on investor characteristics and their loan portfolios is not publicly available. 5 Fortunately, we obtain a rich data set provided by LendingRobot,analgorithmicthirdparty,whichincludesportfoliocomposition for a large set of retail investors on two largest lending platforms, Lending Club and Prosper. Therefore, we can study sophisticated investor screening and outperformance within the same investor segment and across platforms. Importantly, our sample includes a significant source of heterogeneity in terms of sophistication: some investors invest by themselves, whereas others rely on thevariousscreeningandorder-placingtechnologiesofferedbyLendingRobot. Our empirical analysis progresses in several steps. First, we show that more sophisticated investors rely on different loan characteristics to screen the loans they finance, which points to their information advantage. Being selected by sophisticated investors predicts a significantly lower probability of default for a given loan, meaning that sophisticated investors systematically outperform unsophisticated investors over time and across all risk buckets. We find that loans selected by sophisticated investors have a default rate on average 3% lowerthantheaverageloan,orloanspickedbyunsophisticatedinvestors,which correspondstoareductionofmorethan20%oftheaveragedefaultrisk.Indeed, sophisticated investors produce information. Using the 2014 Lending Club event described above, we then implement a difference-in-differences methodology to establish causal evidence of the impact of a large reduction in platform information provision on sophisticated investors performance. We find that sophisticated investor outperformance drops by more than half at the time of the reduction. We rationalize this unanticipated event, corresponding to the platform “evening the playing field,” by referring to the theoretical argument that platforms actively manage adverse selection. Under our rationalization, this event suggests that lending platforms valueunsophisticatedinvestorsandthereforeactasifthey“protect”them,even in the absence of specific regulation. Finally, we find that platforms prescreening intensity also has been improving in the sense that platform risk buckets are increasingly precise at predicting default. In addition to attracting unsophisticated investors directly, this increased precision is also likely to mitigate adverse selection by reducing theheterogeneityofloanswithinagivenriskbucket.Consistentwithplatforms managing adverse selection, we also find robust time-series evidence that sophisticated investor outperformance has become lower in recent years. Although our empirical tests mainly rely on the heterogeneity within the retail investor segment, our findings have external validity for the institutional 5 Traditionally,thebreakdownbetweenretailandinstitutionalinvestorsrepresentsanaturalsourceofheterogeneity in terms of sophistication, and platform public data allow to identify which loans are sold to retail (fractional loans)orinstitutionalinvestors(wholeloans).However,thisdistinctionisnotinformativeinmarketplacelending, as the allocation between the retail and institutional investor segments is randomized by platforms, and each segment itself also has a large heterogeneity of investor sophistication. Therefore, studying the impact of investor sophistication needs to be conducted within these segments. 1942 Downloaded from by Renmin University user on 05 December 2019 15:40 1/4/2019 RFS-OP-REVF180102.tex Page: 1943 19391982 Marketplace Lending: A New Banking Paradigm? investor segment of marketplace lending, as the heterogeneity in sophistication is comparable across investor segments. Many institutional investors, such as pension funds, only apply a little screening (for instance, only relying on a grade threshold) as retails investors do; while other institutional investors, such as hedge funds, develop highly sophisticated investment strategies that are comparable to what LendingRobot offers to investors. We focus on the robust features in the development of marketplace lending so far while leave a number of interesting questions for future research. These include, for example, the overall welfare implications of marketplace lending, and whether marketplace lending poses any financial stability concerns due to the increasing participation of institutional investors. Ourpapercontributestotheburgeoningliteratureonmarketplacelendingby directly examining the sharing of information production between platforms and investors, one key factor that makes marketplace lending special as a new banking model. So far, the literature of marketplace lending has mainly focused on how borrowers soft information improves lending outcomes (e.g., Duarte, Siegel, and Young 2012; Iyer et al. 2015; see Morse 2015 for a review). To the best of our knowledge, we are the first to study how investors characteristics affect loan screening outcome and how the participation of sophisticated investors interacts with the optimal platform desi