How Normalization Shapes Fairness–Performance Trade-Offs in Threshold-Based Credit Decision Systems
Journal of Risk and Financial Management · 07.11.2026 · 19 (11) · 900
DOI: 10.3390/jrfm19110900 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Risk and Financial Management, online erschienen am 07.11.2026. Verfasst von Essodjolo Kpatcha. Die Arbeit nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Machine Learning (ML) models are increasingly used in credit risk assessment, creating a growing need for decision mechanisms that balance predictive performance and fairness. Fairness-aware threshold optimization has emerged as an attractive post-processing approach because it enables explicit control of fairness–performance trade-offs without modifying the underlying predictive model. These approaches typically aggregate heterogeneous performance and fairness metrics through scalarized objective functions, implicitly relying on normalization to ensure comparability across metrics. Despite its widespread use, the influence of normalization on fairness-aware threshold optimization has received comparatively limited attention, and its impact on optimization outcomes remains insufficiently understood. This paper investigates the impact of normalization on fairness-aware threshold optimization in credit risk modeling. Using a common optimization framework, three representative normalization strategies, namely empirical min–max, theoretical-bound, and distance-to-ideal normalization, are evaluated across Logistic Regression, Random Forest, and XGBoost classifiers on both synthetic and real-world credit datasets. The analysis reveals that normalization substantially influences optimization outcomes. Different normalization schemes modify the representation and aggregation of fairness and performance objectives, alter optimal decision thresholds, and change the resulting fairness–performance trade-offs. As a consequence, identical policy preferences may lead to materially different operational decisions depending solely on the normalization strategy employed, with normalization-induced approval-rate differences reaching approximately 20 percentage points in some cases. The influence of normalization is observed across models and datasets, indicating that normalization is not merely a technical scaling procedure but a modeling choice capable of affecting threshold recommendations and approval decisions. These findings highlight the importance of explicitly assessing normalization choices in ML-based risk modeling. More broadly, they demonstrate that normalization should be regarded as a governance-relevant modeling decision because it can influence fairness-aware credit decision systems, stakeholder preference interpretation, and operational outcomes.
Zitieren
APA
Kpatcha, E. (2026). How Normalization Shapes Fairness–Performance Trade-Offs in Threshold-Based Credit Decision Systems. Journal of Risk and Financial Management, 19(11), 900. https://doi.org/10.3390/jrfm19110900
Vancouver
Kpatcha E. How Normalization Shapes Fairness–Performance Trade-Offs in Threshold-Based Credit Decision Systems. Journal of Risk and Financial Management. 2026;19(11):900. doi:10.3390/jrfm19110900
BibTeX
@article{Kpatcha2026,
author = {Kpatcha, Essodjolo},
title = {How Normalization Shapes Fairness–Performance Trade-Offs in Threshold-Based Credit Decision Systems},
journal = {Journal of Risk and Financial Management},
year = {2026},
volume = {19},
number = {11},
pages = {900},
doi = {10.3390/jrfm19110900}
}
EUVIMED