Open AccessBankruptcy prediction using machine learning and an application to the case of the COVID-19 recession
Author(s) -
Aditya Narvekar,
Debashis Guha
Publication year - 2021
Publication title -
data science in finance and economics
Language(s) - English
Resource type - Journals
ISSN - 2769-2140
DOI - 10.3934/dsfe.2021010
Subject(s) - bankruptcy , bankruptcy prediction , recession , covid-19 , sample (material) , artificial intelligence , economics , machine learning , econometrics , actuarial science , computer science , finance , macroeconomics , medicine , chemistry , disease , pathology , chromatography , infectious disease (medical specialty)
Bankruptcy prediction is an important problem in finance, since successful predictions would allow stakeholders to take early actions to limit their economic losses. In recent years many studies have explored the application of machine learning models to bankruptcy prediction with financial ratios as predictors. This study extends this research by applying machine learning techniques to a quarterly data set covering financial ratios for a large sample of public U.S. firms from 1970–2019. We find that tree-based ensemble methods, especially XGBoost, can achieve a high degree of accuracy in out-of-sample bankruptcy prediction. We next apply our best model, using XGBoost, to the problem of predicting the overall bankruptcy rate in USA in the second half of 2020, after the COVID-19 pandemic had necessitated a lockdown, leading to a deep recession. Our model supports the prediction, made by leading economists, that the rate of bankruptcies will rise substantially in 2020, but it also suggests that this elevated level will not be much higher than 2010.