1. Platform risk: A new type of risk in online lending
As an initiative in the FinTech revolution in recent years, online lending (i.e., marketplace lending), the practice of connecting borrowers with (individual or institutional) lenders via an online platform, without a traditional financial intermediary (e.g., bank), has changed the brick-and-mortar landscape of lending. Similar to traditional lending, online lending is also accomplished by risks. In online lending, it is difficult for lenders to obtain comprehensive information about borrowers, resulting in a more severe hazard of information asymmetry, compared to traditional bank loans. Credit risk, i.e., the risk of default on a loan by the borrower, is accordingly compelling to industry practice and academic research. At the same time, different from traditional lending systems, this emerging market is enduring some unique risk, platform risk, i.e., the risk that a platform may default (cease operation). Compared to borrower credit risk, platform risk may result in even bigger losses for lenders. While lenders can mitigate credit risk by diversifying their portfolios, if a platform defaults, they subsequently may not receive repayments, even the principals, through the platform. In the Chinese online lending market, a large number of platforms have been launched since 2013, thanks to the laissez faire policy previously taken by the government to foster a more inclusive financial market supporting entrepreneurship and innovation, as “it has been notoriously difficult for start-ups and MSEs to obtain finance from China’s traditional banking system”. However, the rapid market expansion is also accompanied by a very high platform default rate. According to WDTY (www.p2peye.com), more than 3,000 platforms have defaulted by the end of 2020, causing grievous economic losses. For example, Ezubao, once a leading online lending platform, wrung out over 74 billion RMB from more than 900,000 investors, and had about 38 billion RMB unpaid loans.
2. We explore types of information and methods that are effective in predicting platform risk
As for information, to investigate whether and how much each type of available information contributes to predicting platform risk, we identify several factors by categorizing the available features based on the aspects they reflect. As for methods, to deal with the partial observation and long-term survivor problems, we propose the use of survival analysis, especially the mixture survival model, for predicting platform risk. As part of the modeling process, we consider two types of default platforms, i.e., problematic platforms (default due to absconding with funds, withdrawal failure, or being involved in economic investigation) and failed platforms (default due to termination of business), using competing risk analysis. We compare survival analysis methods (i.e., mixture survival model, Cox proportional hazards (PH) model, and random survival forests) with classification methods (i.e., logistic regression and random forests). We examine the predictive utilities of four factors, i.e., platform characteristic, risk management, commercial competition, and online word of mouth (OWOM), respectively and jointly. To distinguish between platform risk under mild and intense policy interventions, we perform a cross-stage analysis, covering two historical stages, i.e., the past stage (providing lessons learned) and the current stage (providing practical implications) separated by the recent dramatic policy intervention.
3. Our main findings
The key findings are summarized from the aspects of methods and information as follows:
In regard to platform risk prediction methods:
● Survival analysis methods are superior to classification methods in platform risk evaluation, especially in predicting default probability over time.
● Compared to traditional survival analysis methods, MSM has additional advantages in identifying the effects of features on whether and when a platform will default separately.
In regard to information for platform risk prediction:
● Problematic and failed platforms are highly heterogeneous, as reflected in the predictive utilities of the identified factors and the corresponding features, and thus should be treated differently in platform risk evaluation.
● Commercial competition plays a major role in platform risk evaluation. While the severity has partly been mitigated by the recent dramatic policy intervention, its impact (especially the impact of overstated reference return rates) is still noteworthy.
● Risk management plays an important role in platform risk evaluation, especially at the current stage, and certain devices for increasing information transparency, especially depositing funds at a custodian bank, could alleviate platform risk.
● Platform risk may vary across platform development periods (start-up and stable) and market stages (past and current) in terms of key factors and top-ranking features.
Figure 1. Cases of High-risk and Low-risk Platforms
4. Lessons Learned and Implications
We may learn from this history that when a novel financial model emerges, market competition must be controlled properly or even strictly, and meanwhile, risk supervision (e.g., operation inspection and continued surveillance) must be reinforced to prevent fraud risks derived from vicious competition. Our practical implications, for policymakers/regulators and lenders, carry over from the past stage to the current stage. Facing the enormous losses arising from platform risk, policymakers and regulators in the Chinese online lending market undoubtedly strive to figure out the key incentives and regulate the market into a benign condition, if possible. In this regard, based on our findings, we would suggest that they consider trying to further contain the competition in the current market. In addition, we would suggest that policymakers and regulators consider increasing the information transparency of platforms. For lenders, they can take the same way to build effective models for predicting platform risk, with our identified factors and recommended survival analysis, and accordingly select where to invest. In addition, lenders should rationally trade off platform risk against benefit and cannot just pursue benefit and ignore risk.
Article Information
Zhao Wang, Cuiqing Jiang, & Huimin Zhao. (2022). Know Where to Invest: Platform Risk Evaluation in Online Lending. Information Systems Research, 33(3), 765-783.
