ISCAP Proceedings: Abstract Presentation
Mixed Signals: An Empirical Study of the Alignment (and Misalignment) of Risk Signals with Actions and Outcomes in P2P Lending Markets
Hoda Atef Yekta
James Madison University
Robert Day
University of Connecticut
Justin Tersoglio
James Madison University
Abstract
Peer-to-peer (P2P) lending markets are internet-enabled platforms that create a venue for matching lenders and borrowers with lower overhead, hence the potential for more significant profit margins than traditional financial institutions. These markets generally have three types of participants or players. Borrowers each request a loan amount and return period (e.g., 36-month or 60-month loans) and provide details of their financial history and personal information, such as their occupation and the purpose of the loan. The market owner or platform serves as a mediator between borrowers and investors, receiving the borrowers’ requests and details and making potentially profitable loan requests available to lenders for selection. Though auction-based platforms (with pricing determined through a competitive process) have existed, we focus on markets in which the platform prescribes an interest rate based on their analysis of credit information and borrower analysis. In this case, the platform serves as an expert on risk assessment to prescribe the interest rate rather than distribute this task among investors (as in the auction case). In many cases, the platform will also provide additional signals (beyond defining the interest-rate terms of the loan), such as a risk score, a grade label, and an estimated loss rate, to facilitate differentiation of the loan applications for investors. The third set of players are investors who receive all this information from the platform and decide how much to invest in each loan.
All players may behave strategically in this market. For example, although borrowers cannot change their financial history, they may be able to select the “purpose of the loan” strategically, given that the non-institutionalized and unsecured (i.e., with no collateral) market platform usually cannot verify this information. On the other hand, the platform wants to select a format of information transmission (the loan terms, types of signals, and level of detail) that most efficiently clears loans and results in the most significant profit over time. (Platforms are paid a small percentage of every payment from borrowers to the investors, aligning their incentives partially but also receiving a one-time fee.) Investors evaluate the ex-post outcome of the loans precisely and accurately predict each potential funding decision's risk and benefit.
Having collected a large dataset of publicly available loan information for over four years of loan origination requests (with all follow-up data through the completion of 36-month loan terms) from an anonymous lending platform, this study seeks to shed light on the interplay between the borrowers, lenders, and the market owner in one of the most popular peer-to-peer (P2P) lending markets. We show how signals from one class of participants affect the behavior of others using data analytics. In particular, we first explore the borrowers’ disclosed personal information and analyze the platform's response to these signals and the investors'. Then, we analyze the response of investors to the platform signals and examine how closely investors follow the signals provided by the platform. Finally, we study the efficiency of both the borrower’s self-reported information as well as the P2P lending platform signals in predicting the success of each loan. Our results suggest that certain platform signals might be misleading, as well as significant trends in which the self-reported loan purpose might also be a signal that investors are misreading.