Thursday, September 03, 2009

measuring economic growth by lights

Vox | Given the low quality of GDP measures for countries and the almost total absence of GDP measures for sub-national units such as cities, we propose a readily available proxy: satellite data on lights at night. The best use of lights data is to examine growth in GDP rather than GDP levels, so that cross-country differences in how lights spatially and culturally reflect consumption are differenced out.

We start by examining cross-country GDP growth rates, focusing on the period 1992-2003, and develop a statistical framework for optimally combining the growth in lights measure for each country with estimates of GDP growth from the World Development Indicators. We first establish that changes in lights are well related to particular positive or negative economic growth episodes for particular regions and times and, more generally, that growth in lights is a good predictor of growth in GDP measures. As an illustration (Elvidge et al, 2005), Figure 1 contrasts the big increase in lights from 1992 to 2002 in the Eastern European countries of Poland, Hungary, and Romania with the distinct dimming of lights to the east in the former Soviet Republics of Moldova and the Ukraine, which endured a harsh transition process.

Next, we develop a framework to optimally combine measured GDP growth with growth in lights to obtain a best estimate of true GDP growth. The objective is to minimise the variance of true GDP growth from its best estimate. The weights placed on the World Bank GDP growth measure and the lights growth measure depend in part on the ratio of signal to total variance in the World Bank measure.

Applying our method to the countries given a data quality grade D in the Penn World Tables, we get estimates of true GDP growth that are starkly different from conventional measures.