A Midas Multinomial Logit Model with Applications to Bond Credit Rating

In the multinomial classification analysis, mixed frequency data issue is becoming more and more common. In order to solve the multi-classification problem in the mixed frequency data environment with a large frequency ratio, we combine the restricted mixed data sampling (MIDAS) method with the multinomial logit (MLogit) model to construct a MIDAS-MLogit model. The MIDAS-MLogit model imposes functional constraints on parameters of high frequency variables and makes full use of high frequency information to complete multi-classification tasks in the mixed frequency data environment. Through extensive Monte Carlo simulations, we prove that the novel constructed MIDAS-MLogit model has better classification performance than the conventional MLogit model and the U-MIDAS-MLogit model in solving multi-classification problems with a large frequency ratio. In addition, we use the MIDAS-MLogit model to predict credit rating of corporate bonds issued by Chinese listed companies from 2011 to 2020, and find that it outperforms several competitive models in terms of prediction accuracy. The empirical results also show that the covariates from the bond level, corporate level and financial environment level have different effects on the credit rating of corporate bonds