Estimation Of Sub-District Level Income And Nightlights Data: Towards Developing A Methodology

Abstract

Calculation of GDP constitutes an important exercise for every country. However, due to limited state capacity, the requisite data collection infrastructure is poor in a lot of mid and low income countries. Owing to this, a lot of international agencies have begun to advocate the use of big data as a proxy in socio-economic studies. In this broad context, the dissertation studies satellite collected nightlights data and its usefulness for estimating economic activity. The focus of the dissertation is particularly on district and sub-district level for which the official figures come with a few years of lag. The study also looks at the potential of nightlights data in studying economically disruptive events such as demonetisation. The study finds nightlights corresponding with income at all levels. The analysis shows it can be used as a predictive variable to estimate income across regions. The study recommends that the study be scaled up to train models better.