Propensity Score Matching to Control for Consumer Bank Churn: A Case of Causal Inference
American Journal of Technology · 19.05.2026 · 5 (6) · 41-64
DOI: 10.58425/ajt.v5i6.543 ↗Kurz gefasst
Fachartikel in der Zeitschrift American Journal of Technology, online erschienen am 19.05.2026. Verfasst von Abhinav Kumar. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Aim
This paper aimed to explore the application of Propensity Score Matching (PSM) to a publicly available dataset of 10,000 European bank customers in order to identify high-propensity churn customers amenable to targeted retention interventions.
Methods
The study applied Propensity Score Matching (PSM), a causal inference technique introduced by Rosenbaum and Rubin (1983), using a two-step feature selection process that combined logistic regression with Random Forest variable importance analysis. The number of banking products held was identified as a proxy for treatment assignment. Nine pre-treatment covariates were balanced across treatment and control groups using five matching methods: nearest neighbour, optimal, full, coarsened exact, and subclass matching. Balance diagnostics were conducted using standardized mean differences, variance ratios, and love plots.
Results
The findings showed that full matching achieved robust covariate balance while preserving all observations. The estimated average treatment effect indicated that treated customers were 12.2% less likely to churn (SE = 0.14, p < 0.001). The study provides preliminary evidence that PSM can complement traditional predictive approaches in designing customer retention strategies in consumer banking.
Conclusion
The study concludes that Propensity Score Matching can serve as a complementary causal framework alongside traditional machine learning approaches for understanding retention outcomes and customer churn in consumer banking, although the findings remain contingent on the stated assumptions and dataset limitations. Recommendation: The study recommends the use of Propensity Score Matching alongside traditional predictive approaches when designing targeted customer retention strategies in consumer banking.
Zitieren
APA
Kumar, A. (2026). Propensity Score Matching to Control for Consumer Bank Churn: A Case of Causal Inference. American Journal of Technology, 5(6), 41-64. https://doi.org/10.58425/ajt.v5i6.543
Vancouver
Kumar A. Propensity Score Matching to Control for Consumer Bank Churn: A Case of Causal Inference. American Journal of Technology. 2026;5(6):41-64. doi:10.58425/ajt.v5i6.543
BibTeX
@article{Kumar2026,
author = {Kumar, Abhinav},
title = {Propensity Score Matching to Control for Consumer Bank Churn: A Case of Causal Inference},
journal = {American Journal of Technology},
year = {2026},
volume = {5},
number = {6},
pages = {41--64},
doi = {10.58425/ajt.v5i6.543}
}
EUVIMED