Predictors of financial inclusion in Africa: a propensity score matching, intraclass correlation, and accounting decomposition analysis

Abstract Despite the expansion of mobile financial technology, a substantial share of Africa's adult population remains financially excluded. This study investigates the predictors of financial inclusion using micro-level data from the 2025 Global Findex Database ( N = 36,237 individuals across 37 African countries) and applies a three-stage empirical strategy. First, propensity score matching shows that mobile phone ownership shares a robust adjusted association with an 18.6 percentage point higher likelihood of formal bank account ownership among comparable individuals. Decomposing the combined metric isolates this true behavioral association from the 32.0 percentage point mechanical tautology of mobile money (which strictly requires a device), remaining robust to Coarsened Exact Matching and Oster bounds. Second, a multilevel mixed-effects probit model yields a latent intraclass correlation coefficient of 0.212 (with a simulated variance partition coefficient of 11.5 percent), suggesting that a significant proportion of variation in financial inclusion is attributable to unobserved country-level factors. Third, interaction analysis finds significant gender heterogeneity in the association between mobile phone ownership and financial inclusion. Although access to mobile technology increases inclusion for both men and women, the difference in predicted probabilities is smaller among non-owners than among owners. Finally, an accounting decomposition based on predictive margins estimates a 29.1 percentage point difference in predicted probabilities (within the observed support of the data), highlighting the role of mobile infrastructure in expanding financial access. Overall, the findings suggest that achieving broad-based financial inclusion requires policy attention not only to network expansion but also to device affordability and gender-responsive regulation.