Estimating the Economic Impact of COVID-19 in India Using Night Lights

Abstract

The outbreak of COVID-19 in early 2020 heralded a deep global recession not seen since the Second World War. With a billion people out of work and entire countries in lock-down, the burgeoning economies of countries like India has plunged into a downward spiral. The conventional instruments of estimating the economic impact of a pandemic is limited. And, it is in this backdrop, that we investigate the promise of using an unconventional datasource - night-time images of lights on Earth, taken by satellites, to measure the economic cost. Electricity usage, which is also known to track economic measures is included as an additional regressor. First, a novel processing framework for a new state-of-the-art version of nightlights is developed. Second, using panel regression, the elasticity of nightlights to National GDP and Subnational GSVA is estimated. Machine learning is then used to predict the YoY change in metrics between Apr-Jun,2020. A strong relationship between both Electricity Usage and Nightlights to GDP was observed, with Electricity Usage having a higher predictive power. The model predicted a contraction of 24% in FY2020Q1, almost identical to the official GDP decline of -23.9% later published by the Indian Government. However, the performance of the machine learning model at state level was suboptimal and requires further analysis. Based on the findings, we conclude that nightlights along with electricity usage can be invaluable proxies for estimating the cost of short-term supply-demand shocks, such as COVID-19 and should be explored further.