Digital Transformation and Crop Production in Sub-Saharan Africa: Evidence from Dynamic Panel Data Analysis
Peter Makieu
*
Department of Agribusiness Management, School of Agriculture and Food Sciences, Njala University, Njala Campus, Sierra Leone.
Saffa Mohamed Massaquoi
Department of Agribusiness Management, School of Agriculture and Food Sciences, Njala University, Njala Campus, Sierra Leone.
Samba Kamara
Department of Agribusiness Management, School of Agriculture and Food Sciences, Njala University, Njala Campus, Sierra Leone.
Andrew Success Howe
School of Environmental Science and Engineering, Suzhou University of Science and Technology, Jiangsu Province, China.
Marie Kargbo
School of Environmental Science and Engineering, Suzhou University of Science and Technology, Jiangsu Province, China.
Daniel Karlay Hinneh
Department of Material Science and Engineering, University of Liberia, Monrovia, Republic of Liberia.
Aruna James Kabia
School of Architecture Rural and Ubran Planning, Suzhou University of Science and Technology, Jiangsu Province, China.
*Author to whom correspondence should be addressed.
Abstract
Digital transformation can reduce information frictions and expand access to markets, extension services, and financial tools, but its relationship with agricultural outcomes in Sub-Saharan Africa (SSA) remains sensitive to measurement and econometric specification. This study examines the association between a connectivity-based Digital Transformation Index (DTI) and crop production in 40 SSA countries over 2000-2022 (N = 769 country-year observations). The DTI combines mobile-cellular subscriptions, internet use, and fixed-broadband subscriptions, while the outcome is the natural logarithm of the FAO Crop Production Index (2014-2016 = 100), an aggregate production index rather than an input-normalised productivity measure. Entity fixed effects (FE) provide the main static estimate, while pooled OLS, random effects, two-way FE, System GMM, Difference GMM, component-level models, income-group estimates, and a quadratic specification are used for sensitivity analysis. Entity FE yields a positive DTI coefficient (beta = 0.948, p < 0.001), whereas the estimate is smaller and only marginally significant with year effects (beta = 0.523, p = 0.095). In the dynamic models, System GMM produces a positive but imprecise DTI estimate (beta = 0.031, p = 0.409), while Difference GMM yields a negative and imprecise estimate (beta = -0.059, p = 0.581). Separate income-group FE models show a larger coefficient in low-income countries (beta = 1.841) than in lower-middle- and upper-middle-income groups (approximately 0.48), although no formal cross-group coefficient-equality test is reported. The quadratic FE specification is concave and implies a sample-specific turning point near DTI = 0.709, located in the upper tail of the observed distribution. Overall, the results support a positive long-horizon within-country association in the static FE model, but the weaker time-controlled and dynamic estimates warrant cautious, non-causal interpretation.
Keywords: Digital transformation, ICT for development, crop production, digital divide, Sub-Saharan Africa, fixed effects, dynamic panel data