Climate-Related Credit Risk in Banking: A Critical Review of Empirical Evidence and Methodological Approaches

This paper provides a critical narrative review of the literature on climate-related credit-risk and its implications for the banking sector. Banks’ exposure to physical and transition climate risks can significantly affect credit quality, lending conditions, and overall financial stability, yet the methods used to study these effects vary widely, and each faces limitations. This paper systematically assesses the strengths and weaknesses of the main empirical approaches—including regression models, stress testing, and scenario analysis—in the presence of rare default events, unobserved heterogeneity, and quasi-complete separation and highlights unresolved tensions in the evidence base. Building on this assessment, it discusses the Bayesian multilevel logistic model (BMLM) as a conceptual methodological framework for firm-level credit-risk analysis and provides a structured comparison of seven modelling families—including deep learning architectures and survival analysis methods—across eight evaluation dimensions. This paper’s principal contribution is a structured mapping of research objectives, data characteristics, and modelling choices, rather than the advocacy of a single preferred approach.