Financial distress classification in Vietnam: a market-sensitive Z-score framework using quadratic discriminant analysis

Purpose This study develops and validates a multidimensional financial distress classification framework for firms operating in an emerging market environment. While the Altman Z-score model remains one of the most widely used bankruptcy prediction tools, its application often assumes homogeneous covariance structures across risk groups, an assumption that may not hold in practice. This study, therefore, examines whether Quadratic Discriminant Analysis (QDA) provides a more appropriate classification approach than conventional Linear Discriminant Analysis (LDA) for financial distress prediction. Design/methodology/approach The study employs five standardized financial indicators derived from the Altman framework, capturing liquidity, cumulative profitability, operating profitability, market-based solvency and asset utilization. Firms are classified into three financial-risk categories: Extreme Risk, Moderate Risk and Stable. The relative performance of LDA and QDA is evaluated using discriminant analysis, while model robustness is assessed through bootstrap resampling, cross-validation procedures and multiclass classification metrics. Findings The results reveal significant covariance heterogeneity across financial-risk groups, indicating that the assumptions underlying LDA are violated. Consistent with this finding, QDA outperforms LDA and cross-validated LDA, achieving an overall classification accuracy of 92.0%. Canonical discriminant analysis further demonstrates strong separation among risk categories, with the first discriminant function explaining 92.6% of total discriminatory variance and exhibiting a canonical correlation of 0.796. Among the explanatory variables, asset utilization and market-based solvency emerge as the most influential discriminatory factors. The model also maintains strong performance under class imbalance, supporting the robustness of the proposed framework. Originality/value This study extends the financial distress literature by demonstrating the importance of accounting for covariance heterogeneity in multidimensional risk classification. Unlike traditional applications that rely primarily on LDA, the findings provide both theoretical and empirical support for the adoption of QDA in financial distress prediction. The proposed framework offers an interpretable and robust early-warning tool for investors, creditors, regulators and corporate managers seeking to monitor financial risk in emerging markets.