| The Vintage curve is currently the industry standard for estimating the bad debt rate of credit assets in the P2 P industry.The Vintage curve is similar to the Hazard Function in the survival analysis.The abscissa of the curve is the month of book of the credit assets(the duration of the credit assets).And the ordinate of the curve is the bad debt rate for the corresponding aging(the credit asset’s bad debt rate at the corresponding aging).The Vintage curve can dynamically describe the migration of bad rate over the life of the credit asset,and it is currently the most direct data to reflect the risk management level of the P2 P platform.It can accurately expose the bad debt rate of credit assets,help the platform to timely link the pre-lending and post-loan departments to adjust risk strategies,lock high-risk groups,formulate more reasonable capital asset plans,and buffer the liquidity of the platform.At present,the Vintage curve in the P2 P industry usually uses the Kaplan-Meier method in the survival analysis,and the estimated value of the Vintage curve is the bad debt rate corresponding to the month of book.The method has few restrictions and is used widely,but it cannot describe the impact of bad debt rate and credit characteristics of credit assets,which makes the estimation of Vintage curve not specific enough.Credit assets characteristics include credit risk rating,presence or absence of property,job nature,income,etc.This paper will introduce the Cox model to solve the above problems.It can analyze the impact of multiple characteristic variables on the bad debt rate of credit assets at different month of book.In this paper,through the deep mining of the Y platform data,the sample data of the loan time between January 1,2017 and January 31,2018 is obtained.The sample data must be the completed order,then describe the characteristic variables and construct model.In this paper,we obtained 10 characteristic variables and more than 4,500 sample data.The sample data was tested and the Cox proportional hazards model was used to measure the Vintage curve of credit assets.The results show that the Cox proportional hazard model can quantify the impact of multiple characteristic variables on the credit asset Vintage curve.The four characteristic variables tested by the significance test and the premise hypothesis were included in the Cox proportional hazards model.The SPSS statistical software was used to analyze the proportion of each variable to the bad debt ratio of credit assets,and we can get the Vintage curve estimation function based on Cox proportional hazard model. |
