RIEB Discussion Paper Series No.2024-34

Title

Is It Possible to Detect the Insolvency of a Company?

Abstract

As corporate sector stability is critical for economic stability and development, machine learning has become a popular tool for constructing an early warning system (EWS) to detect a company's financial vulnerabilities more accurately. Although most of the EWS literature focuses on constructing bankruptcy prediction models, bankruptcy is not the only indicator of a company's financial fragility. This study uses random forest modelling to systematically investigate the possibility of detecting 1) the financial signs of a company falling into a financially fragile condition of insolvency, and 2) whether insolvent companies fall into bankruptcy. We also analyse how the financial conditions of insolvent companies differ from those of active and bankrupt companies. Our empirical study shows that highly accurate insolvency models can be built to detect status changes from active to insolvent and from insolvent to bankrupt. Our analysis also shows that the financial criteria for the status change from active to insolvent and are quite different from those of a change from insolvent to bankrupt. The criteria of the former are due to structural and operational ratios, whereas those for the latter are due to further financial distress in operational and profitability ratios.

Keywords

Random forest; Data science; Company insolvency and bankruptcy; Financial distress; Financial vulnerability; Economic activity

JEL Classification

G0, C0

Inquiries

Katsuyuki TANAKA
Center for Computational Social Science (CCSS)
Research Institute for Economics and Business Administration(RIEB)
Kobe University
Rokkodai-cho, Nada-ku, Kobe
657-8501 Japan
Phone: +81-78-803-7036
FAX: +81-78-803-7059

Takuo HIGASHIDE
au Asset Management Corporation

Takuji KINKYO
Graduate School of Economics, Kobe University

Shigeyuki HAMORI
Graduate School of Economics, Kobe University
Faculty of Political Science and Economics, Yamato University