Written by Saida Hajjaj
This paper studies the determinants of digital payment adoption by individuals using the Global Findex Database 2025, published by the World Bank and based on a nationally representative household survey conducted in 141 countries in 2024. We combine standard econometric approaches with explainable machine learning (ML) techniques to provide both transparent global benchmarks and detailed policy-relevant insights into digital payment behavior.
First, we estimate a parsimonious logistic regression model on a broad multicountry sample of 5,189 adults from 97 countries, focusing on core sociodemographic and structural characteristics commonly used in the financial inclusion literature. Although this baseline specification supports established patterns related to education, income, and access to digital tools, its predictive performance remains limited, as expected given the limited set of covariates and the objective of cross-country comparability.
We then leverage a richer subsample of 5,183 employers in 74 countries for which detailed information on digital income receipt, payment use cases, and digital connectivity is available. Estimate regularized logistic regression, random forest, and gradient boosting models based on this enriched sample. Prediction performance improved significantly, with the area under the ROC curve increasing from approximately 0.61 for the baseline logit to up to 0.94 for the best performing ML model. Using SHapley Additive exPlanations (SHAP), we show that the adoption of digital payments is primarily driven by participation in digital income and payments ecosystems (e.g., digital receipts of wages, remittances, pensions, agricultural income, domestic remittances, payments to merchants and utilities), along with a synthetic index that captures digital connectivity.
Overall, our results demonstrate strong complementarity between traditional econometrics and explainable machine learning. While econometric models provide interpretable global benchmarks, the ML-SHAP framework reveals heterogeneous mechanisms that are not captured by standard specifications. These findings have direct implications for the sequence of digital financial reforms, the design of digital payment strategies, and the transition to retail central bank digital currencies, especially in emerging and developing countries.
