Measuring the Local Impact of Renewable Energy Plants Using Satellite Data: Quasi-Experimental Evidence from Africa

Despite accounting for less than 2% of global installed renewable capacity, Africa is ac- celerating utility-scale solar and wind deployment under acute development and climate pressures, yet rigorous micro-level evidence on the causal territorial consequences of this expansion remains scarce. This thesis provides causally identified, spatially explicit, and technology-comparative evidence on the local economic and environmental impacts of renewable energy infrastructure across the African continent. The empirical strategy combines precise utility-scale solar and wind plant geolocation from the Renewable Power Plants for Africa (RePP Africa) database (Peters et al., 2023) with a suite of satellite-derived outcome variables—VIIRS Nighttime Lights (NTL) as a proxy for economic activity, MODIS NDVI for vegetation health, and MODIS MCD12Q1 categorical land cover shares for structural land-use transitions—observed annually from 2012 to 2024. A spatial buffer design around each power plant (500 m to 20 km radii) gen- erates treatment zones based on the plant’s commissioning year and geographically prox- imate control areas (25–150 km annuli) for a panel of 259,480 observations across 3,992 spatial units (372 solar plants, 83 wind plants, and 3,537 control areas). Causal identifica- tion exploits staggered variation in plant commissioning dates through the doubly robust, treatment-effects estimator of Callaway and Sant’Anna (2021), which addresses the “for- bidden comparisons” bias inherent in standard two-way fixed effects designs with stag- gered adoption (Goodman-Bacon, 2021), and conditions on rich geographic, topographic, climatic, and accessibility covariates to satisfy a conditional parallel trends assumption. Solar photovoltaic plants exhibit a multi-scalar impact signature. Vegetation is sig- nificantly depleted within the immediate plant footprint (500 m buffer), consistent with industrial land clearance. At the 20 km buffer, nighttime luminosity increases significantly, indicating infrastructure-associated economic spillovers whose precise mechanisms—grid extension, commercial agglomeration, or ancillary construction—cannot be isolated from the satellite record alone. Concurrently, barren land share expands significantly at the same spatial scale, suggesting peripheral land degradation that co-occurs with, but may be physically distinct from, the economic activity captured by luminosity. Wind power plants exhibit a contrasting profile: nighttime luminosity is statistically unchanged across all spatial scales, while barren land share declines significantly and consistently at every buffer from 500 m to 20 km, accompanied by localised vegetation expansion at 2 km. The most striking finding is a qualitative reversal of environmental impact between the two technologies: solar infrastructure is associated with barren land expansion at the 20 km scale, while wind infrastructure is associated with barren land reduction across all scales examined. This finding challenges the assumption that “renewable energy” consti- tutes an environmentally homogeneous category, and has direct implications for spatial planning frameworks, technology prioritisation under land constraints, and the design of benefit-sharing mechanisms for host communities navigating the energy transition.