A data efficient framework for analyzing structural transformation in low and middle income economies

Structural transformation, the reallocation of labor and output from agriculture to industry and services, is central to economic development but remains difficult to measure in low- and middle-income countries (LMICs) due to incomplete and inconsistent data. This paper proposes a unified framework that integrates Bayesian hierarchical modeling, machine learning-based imputation, and factor analysis to address this challenge. Using World Bank data (2000–2020) from Kenya, Nigeria, and Ghana, we simulate data sparsity and evaluate three imputation techniques. SoftImpute achieves the lowest RMSE for sectoral indicators, while k-Nearest Neighbors excels in reconstructing GDP. Factor analysis distills latent drivers of productivity change, and the Bayesian model incorporates sectoral and temporal heterogeneity under uncertainty. Empirical results reveal distinct national trajectories, service-led growth in Kenya, oil-linked industrial volatility in Nigeria, and balanced expansion in Ghana. Compared to traditional models, the framework offers greater accuracy and interpretability under missingness, providing a scalable tool for structural diagnostics and data-informed policymaking in LMICs.