Early warning research on corporate bond default risk based on proportional odds model

Bond defaults pose a serious threat to the stability of financial markets, and accurately predicting the occurrence of these events is crucial for risk management. In view of the limitations of traditional statistical models in dealing with temporal dynamics, this paper takes 318 bonds as the research sample and uses proportional odds model to construct a bond default early warning system. Due to the large number of predictor variables involved and in order to enhance the predictive accuracy of the model, two penalty functions, MCP and SCAD, are introduced to utilize the penalized log-likelihood of the proportional odds model for variable screening. Considering that the MM algorithm has the advantages of separating parameters and numerical computation stability, this paper chooses the MM algorithm to regularize the estimation of model parameters, and determines the main factors affecting bond default. The results show that the proportional odds model can effectively capture the time-dynamic characteristics of bond default and has high accuracy in predicting the occurrence of default events.