Zombie firms and credit risk: a micro–macro-analysis based on supervisory data

This paper investigates the implications of zombie lending for credit risk and financial stability, addressing three objectives: (1) identifying and characterizing zombie firms among non-financial enterprises; (2) examining which of these firms maintain credit exposures to commercial banks; and (3) evaluating the impact of zombie loans on micro-level credit risk and macro-level financial fragility. The empirical analysis employs a unique supervisory panel dataset covering 2007–2021, integrating supervisory, financial, behavioral, and registry data. A two-stage methodological framework is applied. At the micro-level, the relationship between zombie status and firm-level probability of default (PD) is estimated using logistic regression, machine learning (CatBoost), and treatment effect models. At the macro-level, a satellite model assesses the predictive power of sectoral zombie firm shares and their debt levels for sector-wide default rates. The findings reveal that zombie loans are associated with substantially higher credit risk. Zombie firms exhibit PDs 61–64 percentage points higher than non-zombie peers. Sectoral zombie exposure—measured by firm share and debt—provides additional short-term predictive value for default rates, beyond autoregressive and macroeconomic components. Non-performing loan (NPL) ratios for zombie exposures persistently exceed those of the general corporate loan portfolio. Notably, in contrast to logistic regression results, zombie status emerged as the most influential feature in the CatBoost model, underscoring its relevance for default classification. These results contribute to the literature on financial fragility by quantifying the risk of zombie exposures. The findings also have policy relevance, supporting the calibration of higher capital requirements for zombie loans under the Internal Ratings-Based (IRB) framework.