Predicting default of listed companies in mainland China via U-MIDAS Logit model with group lasso penalty

Abstract

We introduce the group LASSO penalty into the U-MIDAS logistic regression context to develop a U-MIDAS-Logit-GL model. The U-MIDAS-Logit-GL model enables us to identify important variables at group level in high dimensional mixed frequency data analysis. We then apply it to a real-world application on studying the default of listed companies in mainland China. The U-MIDAS-Logit-GL model is able to effectively identify important determinants from high-frequency financial factors and low-frequency corporate governance profiles simultaneously. It also successfully predicts the default and outperforms the other competitive models for both in-sample and out-of-sample tests.