JULLIAN JACKSON AND TAFIRENYIKA SUNDE TELECOMMUNICATIONS

The study uses a harmonised country-year panel covering 45 sub-Saharan African countries from 2004 to 2023. The balanced levels panel contains 900 observations, while the dynamic estimation sample contains 855 observations after the first annual observation for each country is used to construct the lagged dependent variable. The dependent variable, formal financial-system access (FI), is the unweighted mean of automated teller machine density and commercial bank-branch density per 100,000 adults. The principal explanatory variable, FINTECH, is the percentage of adults who own a mobile-money account, obtained from the World Bank Global Findex database. Regulatory quality (REGQUAL) comes from the World Bank Worldwide Governance Indicators. Infrastructure (INFRA) is constructed as the mean of second-generation mobile-network coverage and access to electricity. Real GDP per capita, inflation and internet penetration are included as control variables. The financial-literacy measures are constructed from the 2014 S&P Global Financial Literacy Survey and annual adult-literacy data. FINLIT2, the principal literacy variable, represents the share of adults answering at least three of four financial-knowledge concepts correctly and is annualised using changes in adult literacy. FINLIT1 provides an alternative component-knowledge measure. The dataset explicitly records whether observations are directly observed, internally interpolated, boundary-completed or model-estimated. Within the dynamic sample, FINTECH comprises 88 directly observed observations (10.3%), 132 internally interpolated observations (15.4%), 426 boundary-completed observations (49.8%) and 209 telecom-only panel estimates (24.4%). These provenance classifications are retained for audit and sensitivity analysis. The principal sources are the IMF Financial Access Survey, World Bank Global Findex, World Development Indicators, Worldwide Governance Indicators, International Telecommunication Union DataHub and S&P Global financial-literacy records. Provider-side mobile-money indicators, including active accounts, registered agents and transaction value, are retained only as separate robustness measures because they differ conceptually and in measurement units from person-level Findex mobile-money ownership.