AbstractChina has the largest, and fastest growing, peer-to-peer (P2P) lending market in the world. P2P lending is a novel mode of alternative financing which operates outside of the more traditional institutional finance market by directly connecting consumers and businesses who are excluded from traditional lending channels and private individuals and households with surplus cash. The mechanism by which the two groups are connected to each other is the P2P platform. Individuals with a need (demanders) for funds post their request on the platform and individuals with surplus cash balances (suppliers) make an offer to provide the necessary funds. The unique feature of the
Chinese P2P model is that there are two alternatives: the non-guarantee model, and; the guarantee model. The guarantee model is unique as it plays an active role in the market process in the form of a loan guarantee unlike the non-guarantee model in which the P2P platform simply acts as an intermediary and ‘platform’ for connecting demand and supply. It is this unique aspect of the guarantee model which disciplines the platform into acting as an honest information intermediary but, on the other hand, it also creates potential financial instability given the platform is also the credit risk taker.
Formally analysing the benefits and risks of the guarantee model and comparing it to the non-guarantee model can help policymakers design a more efficient P2P lending market and a bespoke regulatory framework. To do so, regulators need to understand and compare the different platforms’ behaviours as information producers across the two lending models. And this is precisely the research gap this thesis intends to fill. To date, the extant literature on P2P lending mainly focuses on borrower-lender interactions, and the platforms are treated as honest brokers. To fill the gap, this research explores how different P2P lending models affect the P2P lending platform’s screening
and pricing strategies, and overall social welfare.
First, we develop game-theoretic models of the lending processes to derive the platform’s optimal screening and pricing strategies under the two lending models. Under the non-guarantee model, the platform faces a dynamic trade-off between overstating borrower credit quality to increase short-term profits and honestly disclosing borrower credit risk to improve the platform’s long-term reputation. Under the guarantee model, the platform only faces a trade-off between setting a higher guarantee fee (but with lower funding success probability) or a lower guarantee fee (but with a higher probability of funding success). We find that, under the non-guarantee model, the platform chooses the loosest screening standard, i.e., it approves a known “bad” borrower with the highest possible probability. This means reputation concerns are not enough to discipline the platform and consequently the non-guarantee model lowers the screening efficiency as known “bad” borrowers obtain credit. The main reason for this result is a bad loan cannot be unambiguously attributed to the platform’s dishonest information disclosure (deliberately lax screening standards) due to the imperfection of the platform’s screening technology. By contrast, the platform always screens borrowers truthfully under the guarantee model. The optimal pricing strategy of the guarantee model reflects a risk-sharing arrangement between the platform and the lender.
Next, we perform a welfare comparison of the two lending models. The welfare analysis shows that the guarantee model generates greater social welfare than the non-guarantee model. Then, we relax the assumption that the platform is rational/well-calibrated regarding its screening ability and analyse the welfare effect of overconfidence. We find that under the guarantee model, if the platform is overconfident about its own screening precision, it tends underprice borrower risk, which in turn creates welfare losses.
Finally, we develop an empirical procedure to examine whether the platform underprices the credit risk of P2P loans under the guarantee model. By using loan-level data from a Chinese P2P lending platform, we find that the guarantee fees preset by the platform are sufficient to neither cover the ex-post realized loan losses nor the ex-ante predictable loan losses, thus suggesting that the platform underprices the loan risk in both ex-ante and ex-post sense. This implies that the guarantee model could jeopardize the platforms’ soundness and further financial stability. In general, the theoretical and empirical findings together suggest that policymakers should balance financial stability against screening efficiency when developing regulations that define the role and function of P2P lending platforms.
|Date of Award
|Marc Cowling (Supervisor)