Estimation of District-Level GDP using Night Lights : The Case of Bangladesh

This paper uses a harmonized dataset of annual nightlight to predict district-level GDP for the 64 districts of Bangladesh over the period 1992-2020. The premise of the paper originates from the work of Henderson et al. (2012), who assert that a simple constant elasticity relationship encapsulates the relationship between nightlights and GDP - with the former being dependent on the latter. We initially follow Henderson et al. (2012) to estimate sectoral nightlight elasticities for Bangladesh as a whole. However, not all nightlights contribute to GDP, hence we only use nightlight emitted by enterprises. Furthermore, we also incorporate the elasticity of sectoral nightlight to get a measure of nightlight that incorporates the productive nature of value addition to GDP. Our approach to estimating district-level GDP yields a prediction error of 22.24%, which is smaller than previous estimates. Furthermore, we also estimate a composite measure of GDP for Bangladesh and the 64 districts that is devoid of measurement error. We use this composite measure of GDP to estimate the true structural sectoral nightlight elasticities. Lastly, we also address the long-standing debate of differing economic progress of the eastern and western regions of Bangladesh using nightlights and predicted district GDP