Nighttime lights and wealth in very small areas: Namibian complete census versus DHS data

Nighttime lights observed from satellites are a widely accepted proxy measure for economic development. This is mainly based on cross-country evidence that finds strong correlations between lights and Gross Domestic Product. Yet, the evidence on the correlations at local levels is scarce, and it often relies on randomly sampled survey data. We contribute by enhancing the understanding of the relationship between light and development at local levels. First, we use complete (non-publicly available) census data from Namibia to evaluate the findings based on the randomly sampled Demographic and Health Surveys data. We find that the census data provides a stronger association between light and wealth at local levels. Second, we criticize the practice of aggregating light from buffers around survey cluster locations. Instead, we recommend aggregating data in grid cells, and studying the relationship in different grid sizes. In our study correlations based on grid cells remain significant from a 0.5 degree grid to the smallest 0.0083 degree grid (similar to 1 km(2)) allowed by the nighttime light data. Third, we supplement the commonly used relative wealth index by using individual asset variables as proxies for the total stock of wealth. The stock variables reveal a significant association between changes in light and wealth which cannot be found using our relative wealth index. Altogether, our results show that nighttime lights provide an even stronger signal of economic development at local levels than the current survey-based results in the literature suggest.

Automated summary

Nighttime lights observed from satellites are widely accepted as a proxy measure for economic development, showing strong correlations with Gross Domestic Product. This study uses complete census data from Namibia to reveal a stronger association between lights and wealth at local levels, and suggests aggregating data in grid cells for studying the relationship in different grid sizes. The study also finds that individual asset variables can provide significant insights into changes in light and wealth, offering a stronger signal of economic development at local levels compared to existing survey-based results.