Scholarlink

<p>Digital payments are central to financial inclusion strategies, yet adoption remains highly uneven across individuals and countries. Using Global Findex 2025 microdata, this paper combines an interpretable econometric benchmark with explainable machine learning to examine who adopts digital payments and through which observable mechanisms. A baseline logistic regression highlights familiar socio-demographic gradients, but shows limited discriminative power. An enriched predictive specification incorporating digital income channels, payment use cases, remittances, and a connectivity index substantially improves performance. SHAP-based interpretation reveals that participation in digitally mediated income and payment ecosystems, together with connectivity, dominates prediction, while traditional socio-demographic variables play a more indirect role. These attributions are explicitly interpreted as predictive contributions rather than causal effects. Methodologically, the paper demonstrates how econometrics and explainable AI can be used jointly: the logit model provides a globally interpretable benchmark, while machine learning and SHAP identify where predictive signal concentrates in high-dimensional digital-behaviour data. Substantively, the findings point toward ecosystem-oriented policy approaches that expand digital income channels, payment acceptance networks, and connectivity infrastructure. Although descriptive and subject to sample-selection constraints, the framework offers a scalable approach for integrating survey microdata and explainable AI in the study of digital financial inclusion.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Explaining digital payments adoption with econometrics and explainable machine learning: cross-country evidence from a global household survey

  • 期刊论文

  • 发表日期：

    2026年3月3日

查看全文
  • Saida Hajjaji

摘要

Digital payments are central to financial inclusion strategies, yet adoption remains highly uneven across individuals and countries. Using Global Findex 2025 microdata, this paper combines an interpretable econometric benchmark with explainable machine learning to examine who adopts digital payments and through which observable mechanisms. A baseline logistic regression highlights familiar socio-demographic gradients, but shows limited discriminative power. An enriched predictive specification incorporating digital income channels, payment use cases, remittances, and a connectivity index substantially improves performance. SHAP-based interpretation reveals that participation in digitally mediated income and payment ecosystems, together with connectivity, dominates prediction, while traditional socio-demographic variables play a more indirect role. These attributions are explicitly interpreted as predictive contributions rather than causal effects. Methodologically, the paper demonstrates how econometrics and explainable AI can be used jointly: the logit model provides a globally interpretable benchmark, while machine learning and SHAP identify where predictive signal concentrates in high-dimensional digital-behaviour data. Substantively, the findings point toward ecosystem-oriented policy approaches that expand digital income channels, payment acceptance networks, and connectivity infrastructure. Although descriptive and subject to sample-selection constraints, the framework offers a scalable approach for integrating survey microdata and explainable AI in the study of digital financial inclusion.