New Credit-Risk Model Integrates Debit Data to Better Predict Delinquency

Researchers from BI Norwegian Business School and NHH Norwegian School of Economics have developed a new behavioral credit-risk model that integrates credit and debit transactions, significantly outperforming state-of-the-art machine learning methods in predicting credit card delinquency while offering clearer insight into the behavioral drivers behind repayment problems. The study, published in The Journal of Finance and Data Science, demonstrates that combining credit card data with customers' debit transactions substantially improves the ability to predict credit card delinquency, moving beyond traditional models that rely on monthly aggregates like balance and credit limit.

First author Håvard Huse explains that credit data alone provides only a partial picture of a customer's financial situation, while integrating debit transactions reveals payday spending patterns, repayment behavior, and income dynamics—factors strongly influencing payment default risk. The research team developed a hierarchical Bayesian behavioral model that consistently outperforms leading machine-learning algorithms including XGBoost, GBM, neural networks, and stacked ensembles by capturing behavioral dynamics such as how repayment patterns evolve over time and how spending spikes after payday.

The model draws on detailed credit and debit transaction data from a large Norwegian bank and improves prediction accuracy at the individual level while identifying distinct behavioral segments with different "memory lengths"—the extent to which past financial states affect current repayment behavior. Co-author Auke Hunneman notes that customers in financial distress tend to be more influenced by earlier months' behavior, and the new model captures this dynamic far better than standard machine-learning tools. The approach not only performs better than state-of-the-art algorithms but is also more interpretable, allowing banks to understand which behavioral patterns drive risk rather than just receiving accurate predictions.

Using a three-month prediction horizon, early detection of at-risk cardholders could generate substantial cost savings by enabling timely intervention and reducing losses. Co-author Sven A. Haugland emphasizes that for banks, this represents more than an accuracy improvement—it provides a way to proactively help customers avoid serious financial problems. The findings highlight an emerging shift in credit scoring from traditional static models toward richer behavioral analytics based on a full picture of customer transactions, as detailed in the study available at https://doi.org/10.1016/j.jfds.2025.100166.