Replication Data and Code for “Digital Connectivity, Productive Capacity, and Trade Openness in Africa: A Machine Learning Approach”

This dataset supports the study titled “Digital Connectivity, Productive Capacity, and Trade Openness in Africa: A Machine Learning Approach.” The study examines how digital connectivity, productive capacity, infrastructure, and macroeconomic conditions predict trade openness across African economies. The repository contains the harmonised dataset and Python code used for the empirical analysis. The data cover 54 African countries from 1985 to 2024, producing a balanced country-year panel of 2,160 observations. The main outcome variable is trade openness, measured as trade as a percentage of GDP. The predictors include internet use, mobile subscriptions, access to electricity, GDP per capita, GDP growth, foreign direct investment inflows, industry value added, services value added, inflation, and population density. The accompanying Python notebook reproduces the machine learning workflow used in the study. The analysis estimates Elastic Net, Random Forest, Extra Trees, Gradient Boosting, and HistGradientBoosting models. Model performance is assessed using a random train-test split, repeated-split validation, and a temporal holdout test. The repository also supports the permutation importance analysis, principal component analysis, and K-means clustering used to identify country-level digital-trade readiness regimes. The files are shared to support transparency, replication, and further research on digital connectivity, productive capacity, and trade openness in Africa. The data were drawn from publicly available World Development Indicators sources and processed for the purpose of this study.