Study On The Consumer Credit Risk Management Of P2P Lending In China

Due to the rapid development of economy and information technology(IT),Internet finance is generated by integrating IT with traditional finance area.Among the industries of internet finance,peer-to-peer(P2P)lending has grown significantly in the recent years.Because of its convenience,P2 P lending has relieved the financial exclusion and financial depression,which have long existed in China.P2 P lending also plays an important role in overcoming the financial difficulties of individuals and small and medium-sized enterprises.However,P2 P lending is now facing a severe consumer credit risk,which hinders the sustainable development of the industry.Consequently,we focus on the consumer credit risk issue in P2 P lending in China.We aim to draw lessons from the advanced consumer credit risk management in traditional financial institutions,and fully consider the characteristics of P2 P lending in China to build an efficient consumer credit risk management system for P2 P lending industry in China.This thesis starts from summarizing the existing literature from the perspective of consumer credit risk identification,measurement,and control.Therefore,some important concepts in this thesis are discussed and analyzed.After comparing the P2 P lending and commercial banks in terms of institutional characteristics and consumer credit risk management,we propose that it is of importance to draw lessons from the process and tools of consumer credit risk management in commercial banks,and fully consider the characteristics of P2 P lending,to achieve the sufficient management of consumer credit risk in P2 P lending.We summary the theories of asymmetric information and modern financial institutions build the theoretical fundamental of consumer credit risk management of P2 P lending in China.Based on the theoretical fundamental and practices,we propose the endogenous mechanisms of consumer credit risk in P2 P lending,including asymmetric information,inherent high risk of P2 P lending,and mandate relationship of participators.Some critical exogenous mechanisms of consumer credit risk,which consist of inefficient credit reporting system,lack of internal credit risk management tools,and exogeneous determinants of borrowers' liabilities and payment intention.A theoretical path analysis framework of mechanisms of consumer credit risk of P2 P lending in China is proposed to investigate into the transmission mechanisms of each component in the framework.Under the principle of “risk-identification,risk-measurement,and risk-control”,as well as following the transmission path of “exogeneous determinants ? borrowers' liabilities and payment intention ? internal credit risk management tools ? capability of investors' risk identification ? consumer credit risk”,we carry out the further empirical studies.Concerning the consumer credit risk identification,based on the data from a mainstream P2 P lending platform in China,a binary quantile regression model is employed to identify key determinants of consumer credit risk of P2 P lending.The empirical results are compared with that of logit regression and probit regression.The results indicate that demographic characteristics,loan characteristics,and borrowers' creditworthiness are significant determinants of consumer credit risk in P2 P lending.The binary quantile regression model can handle the heterogeneity,outliers and heteroscedasticity,which often occur in economic data.The cross-validated prediction accuracy shows that binary quantile regression model is superior to logit or probit regression,which further illustrates the superiority of binary quantile regression model in identifying key factors to consumer credit risk in P2 P lending.Regarding the consumer credit risk measurement,we build a selective heterogeneous ensemble credit scoring model(i.e.,bstacking)to overcome the drawbacks of base models and fusion strategies in current ensemble credit scoring model studies.The proposed bstacking credit scoring model is expected to improve the performances of probability of default prediction.The proposal is validated over several datasets and evaluation metrics and compared with a variety of benchmark models.The comparison results demonstrate the superiority of bstacking approach over several evaluation metrics.Bstacking credit scoring model is shown to achieve the optimal performance when the number of iteration lies between 40 and 60.The ROC curve analysis illustrates that bstacking model performs better in discriminating potential default borrowers when the decision threshold is low.We also discuss whether the proposed bstacking model can be applied to real-world practice from the perspective of cost/benefit analysis and financial regulation.The conclusions can be summarized as follows.The marginal cost of employing accurate yet complex credit scoring models is decreasing.Thus,the managers may attempt these models at proper time,whereas much attention should be paid on the privacy protection and legislation.In terms of consumer credit risk control in P2 P lending,this thesis fully considers the special requirements of decision-makings for lenders in P2 P lending and proposes a decision support system based on cost-sensitive learning.The system aims to guide lenders make profitable portfolio by means of risk aversion and diversified investment,and thus,control the consumer credit risk in P2 P lending in advance.Validated by multiple datasets,the empirical results demonstrate that under the guidance of the proposed system,the expected annualized investment return is significantly higher than that of most mainstream models.The comparisons of ARR curves also support the argument that the proposed model can provide higher expected return than benchmark models at similar rejection rate.The proposed decision support system provides new ideas to consumer credit risk control in P2 P lending domain.Moreover,we discuss controlling the consumer credit risk in P2 P lending in the event and after the event.We consider that much attention should be paid on the internal controls,collection,and punishment on dishonesty.