Satellite data for the social sciences: measuring rural electrification with night-time lights

Remote-sensing data has the potential to revolutionize social science. One of the most prominent examples of this is the Nighttime Lights data set, which provides digital measures of night-time luminosity from 1992 to 2013. This study evaluates the Nighttime Lights data against detailed rural electrification data from the 2011 Census of India. The results suggest that many night-time luminosity measures derived from satellite data are surprisingly accurate for measuring rural electrification, even at the village level and using simple statistical tools. We also demonstrate that this accuracy can be substantially improved by using of better GIS maps, basic geoprocessing tools, and particular aggregations of night-time luminosity. Night-time luminosity performs worse in measuring financial inclusion or proxies of poverty, however, and detects rural electrification less accurately when the supply of power is intermittent. These results offer guidelines for when and how remote-sensing data can be used when administrative data is absent or unreliable.