Business Failure Prediction: A Comparison of Discriminant Analysis, Logit Regression, and PLS Regression

Prévision de la défaillance des entreprises: comparaison de l'analyse discriminante, la régression logit et PLS Business Failure Prediction: A Comparison of Discriminant Analysis, Logit Regression, and PLS Regression

Rahma Mzouri and Abdelkrim Kandrouch
Additional contact information
Rahma Mzouri: Faculté des Sciences Juridiques, Economiques et Sociales - UM5 - Université Mohammed V de Rabat [Agdal]
Abdelkrim Kandrouch: Faculté des Sciences Juridiques, Economiques et Sociales - UM5 - Université Mohammed V de Rabat [Agdal]

Post-Print from HAL

Abstract: Corporate failure prediction represents a major challenge for lenders, investors, and managers in a context characterized by increasing bankruptcy rates and growing economic uncertainty. Although discriminant analysis and logistic regression models have been extensively employed in the bankruptcy prediction literature, comparative studies incorporating the Partial Least Squares (PLS) method remain relatively limited, particularly in contexts characterized by high multicollinearity among financial variables.This study aims to compare the predictive performance of Linear Discriminant Analysis (LDA), Logistic Regression (Logit), and the PLS method in forecasting corporate failure.The study is based on a balanced sample of 200 Moroccan firms, including 100 failed companies and 100 non-failed companies. Thirty-three financial ratios covering financial structure, liquidity, solvency, profitability, activity, and growth were analyzed over three forecasting horizons prior to failure (T-1, T-2, and T-3). Discriminating variables were selected using Wilks' Lambda and Fisher's statistic before being incorporated into the different prediction models.The results suggest that the Logit model provides the best short-term predictive performance, achieving a classification accuracy of 93.4% at T-1, compared with 91.2% for discriminant analysis and 90.8% for the PLS method. At longer forecasting horizons, the PLS approach appears to be the most robust, with a classification accuracy of 83.2% at T-3, outperforming both discriminant analysis (78.4%) and Logistic Regression (81.3%). Ratios related to working capital, working capital requirements, solvency, and profitability emerge as the most relevant indicators for the early detection of financial distress.These findings highlight the relevance of Logit and PLS approaches for the development of early warning systems and credit risk scoring models used by financial institutions and decision-makers.

Keywords: Logistic Regression; Défaillance des entreprises Prévision de faillite Analyse discriminante Régression Logit Partial Least Squares Ratios financiers Risque de crédit. JEL classification : G32 C38 C51 M41 Recherche empirique Corporate Failure Bankruptcy Prediction Discriminant Analysis Logistic Regression Partial Least Squares (PLS) Financial Ratios Credit Risk. JEL Classification : G32; C38; C51; Credit Risk. JEL Classification : G32; Financial Ratios; Partial Least Squares (PLS); M41 Paper type : Empirical research; Discriminant Analysis; Bankruptcy Prediction; M41 Recherche empirique Corporate Failure; Risque de crédit. JEL classification : G32; Ratios financiers; Partial Least Squares; Régression Logit; Analyse discriminante; Prévision de faillite; Défaillance des entreprises (search for similar items in EconPapers)
Date: 2026-06-09
New Economics Papers: this item is included in nep-for
Note: View the original document on HAL open archive server: https://hal.science/hal-05652823v1
References: Add references at CitEc
Citations:

Published in International Journal of Accounting, Finance, Auditing, Management and Economics, 2026, 7 (6), ⟨10.5281/zenodo.20500100⟩

Downloads: (external link)
https://hal.science/hal-05652823v1/document (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:hal:journl:hal-05652823

DOI: 10.5281/zenodo.20500100

Access Statistics for this paper

More papers in Post-Print from HAL
Bibliographic data for series maintained by CCSD ().