Research On Early Warning Of Bond Default Risk Of Real Estate Enterprises In My Countr

In recent years,the default of bonds in the real estate sector has become increasingly prominent,and real estate enterprises have always had the characteristics of "capital intensive and high leverage",with supply and demand relying on financing,and the overall financial leverage is high,and the pressure of capital turnover is high.Especially with the introduction of a series of combined tightening policies by the state,such as "housing is not speculation" continues to emphasize the housing residential attributes,the "New Regulations on Housing Enterprise Financing" sets three red lines to standardize the scale of financing,and the "Real Estate Loan Concentration Management System" tightens the residential housing loan policy,the pace of real estate financialization slows down,investment attributes are limited,housing demand gradually declines,and the speed of repayment of housing enterprises slows down,so housing enterprises with high debt and high turnover begin to have large-scale and large-scale bond default events.With the help of real estate enterprises’ actual default events,this paper uses machine learning methods to study the early warning of bond default risks in the real estate industry,that is,to answer three research questions:(1)What are the common reasons for frequent bond defaults of many real estate enterprises?(2)Can machine learning effectively warn of bond default risks in the real estate industry? If so,what method has a better warning effect?(3)For identifying the default risk of real estate enterprise bonds,what dimension of early warning indicators is better?This paper uses the actual default events of real estate enterprises since 2018 to realistically portray the default of real estate enterprises and analyze the deep-seated common causes of the outbreak of default risk,and constructs an early warning index system for bond default risk of real estate enterprises from five dimensions:macroeconomy,bond characteristics,issuer characteristics,financial characteristics and business characteristics.Then,eight commonly used machine learning methods,including Logistic Regression,Lasso Regression,Ridge Regression,Elastic Net,Support Vector Machine SVM,Random Forest RF,Extreme Gradient Boosting Tree XGBoost and KNN algorithm,were used to construct a bond default risk early warning model,and the early warning effect of machine learning methods on bond defaults in the real estate industry and the early warning ability of five-dimensional early warning indicators were empirically tested.The results show that:(1)The direct cause of the current round of bond default is that real estate enterprises have fallen into a "liquidity crisis",and the deep-seated reason is that real estate enterprises have a "linkage effect" between their poor business operation,financial difficulties,and the long-term performance of urgently needed funds such as low valuation and high interest on bonds and the external macroeconomic downturn and tightening of housing market regulation.(2)Through the analysis of model evaluation indicators,the accuracy,recall rate,F1 and AUC values of the eight models were more than 64%,indicating that the machine learning method can effectively warn the default risk of real estate enterprise bonds.Through the comparison and evaluation index,it is found that the ability F1 of the extreme gradient boosting tree XGBoost model to divide positive cases reaches 0.8303,the AUC value of the model evaluation effect is 0.8648,the total number of defaulted bonds is 2528,the model successfully predicts 2011,the recall rate is 0.89;the normal bond sample is misjudged as 305 defaulted bonds,and the accuracy rate is 88.52%,indicating that the model has a more prominent early warning performance in the default problem of housing enterprise bonds.It is more suitable for the early warning of default risk of housing enterprise bonds.(3)Based on the superior early warning method XGBoost,the contribution degree of characteristic indicators of the early warning model of real estate enterprise bond default risk is analyzed,and all dimensions have an early warning effect on the default risk of real estate enterprise bonds.It is found that the early warning contribution of bond characteristics such as coupon rate,issuance period,and issuance market is 33.26%,which is the most important type of early warning indicator that triggers the default risk of housing enterprises.Secondly,early warning indicators representing financial characteristics,such as operating capacity,cash quality,profitability and solvency,will also significantly affect the possibility of default of housing enterprise bonds,with an early warning contribution of 25.14%.Finally,the contribution of early warning was20.39% for business characteristics,16.9% for macroeconomic characteristics and4.31% for issuer characteristics.Based on this,the following policy suggestions are put forward: First,improve the full-caliber monitoring and evaluation platform and risk early warning system for bond defaults of real estate enterprises.Second,real estate enterprises must change their business philosophy,operate steadily,and improve the management level of real estate enterprises.The third is to establish and improve the bond market information disclosure system and improve the macro-prudential management policy system of real estate finance.