A0264
Title: High-dimensional censored MIDAS logistic regression for corporate survival forecasting
Authors: Wei Miao - KU Leuven (China) [presenting]
Jad Beyhum - KU Leuven (Belgium)
Jonas Striaukas - Copenhagen Business School (Denmark)
Ingrid Van Keilegom - KU Leuven (Belgium)
Abstract: Forecasting corporate distress presents three key statistical challenges: (i) right censoring, (ii) high-dimensional predictors, and (iii) mixed-frequency data. To address these complexities, a novel high-dimensional censored MIDAS (mixed data sampling) logistic regression model is introduced. The method accounts for censoring through inverse probability weighting and ensures accurate estimation with numerous mixed-frequency predictors by incorporating a sparse-group penalty. Finite-sample bounds for the estimation error are established, considering censoring, the MIDAS approximation error, and heavy tails. Monte Carlo simulations demonstrate the superior performance of the approach. An extensive empirical application further showcases its effectiveness in predicting the financial distress of Chinese-listed firms. The proposed methodology is implemented in the R package $\texttt{Survivalml}$.
