Enhancing the Attractiveness of Online Lending for Lenders, Borrowers, and the Platform
COM2 Level 4
Executive Classroom, COM2-04-02
Abstract:
Online lending has grown rapidly over the past decade, yet its market share remains small relative to other asset classes. Perceived loss risk, stringent regulation, and limited awareness of benefits have constrained adoption. Because online loans are a new asset class, investors may not know how their performance compares with traditional assets, and it is unclear whether they merit a place in asset allocation. We introduce general characteristics-based portfolio policies (GCPP), a framework that models a loan's portfolio weight as a flexible function of its characteristics, avoiding direct estimation of loan return distributions. Using more than one million LendingClub loans from 2013 to 2020, GCPP portfolios deliver average annualized internal rates of return (IRR) of 8.86% to 13.08%, significantly outperforming an equal-weight portfolio. Relative to six benchmark indices spanning stocks, bonds, and real estate, online-loan portfolios earn competitive or higher returns, and their IRRs exhibit low correlations with the benchmarks, indicating diversification benefits. Together, these results show how GCPP can help platforms serve borrowers and lenders while growing the market.
Our analysis also considers how interest rates for unsecured personal loans should be set. The literature and practice favor risk-based pricing, yet they often overlook the requirement that prices deliver a consistent risk-return trade-off. This omission leads to mispricing in which risky borrowers pay disproportionately high premia. Applying GCPP, we find that platform rates are not always appropriate. The resulting price distortions create a sizable funding-probability gap between subprime borrowers (660 < FICO < 700), and we document evidence consistent with bias against African Americans in rate setting. Building on these findings, we develop a generative-adversarial pricing framework that treats the platform as the rate-setting "generator" and rational lenders as the "discriminator" whose portfolio choices reveal mispricing. We formalize a fairness condition in which rates are fair if and only if the equal-weight portfolio is optimal, and we train the system to minimize deviations from that condition while targeting the platform's average rate and satisfying operational and legal constraints (for example, ECOA). Using LendingClub (2013–2016) and Prosper (2013–2021) in an expanding-window design, the trained system preserves historical average rates, reduces portfolio-weight deviations out of sample, and substantially narrows disparities by race and income source without using protected attributes, while maintaining risk-based pricing and correcting excess spreads across FICO tiers.
Bio:
Ram D. Gopal is the Information Systems Society’s Distinguished Fellow and a Professor of Information Systems Management and Analytics at the Warwick Business School. He also serves as the Academic Director of the Gillmore Centre for Financial Technology at the Warwick Business School. He previously served as the Pro-Dean for Research at the Warwick Business School (2020-2023) and as Head of the Department of Operations and Information Management in the School of Business, University of Connecticut (2008-2018). He has a diverse and a rich portfolio of research that spans analytics, health informatics, financial technologies, information security, privacy and valuation, intellectual property rights, online market design and business impacts of technology. He has served on the editorial boards of top journals including Information Systems Research and has served as the President of the Workshop on Information Technologies and Systems organization from 2016 to 2018. At the Warwick Business School, he teaches 'Digital Transformation' on the Full-time MBA and Executive MBA (London), as well as 'Digital Finance, Blockchain & Cryptocurrencies' on the MSc in Management of Information Systems and Digital Innovation, and 'Generative AI and AI Applications' on the MSc in Business Analytics.