We evaluate the usefulness of satellite-based data on nighttime lights for the prediction of annual GDP growth across a global sample of countries. Going beyond traditional measures of luminosity, such as the sum of lights within a country's borders, we propose several innovative distribution- and location-based indicators attempting to extract new predictive information from the night lights data. Whereas our findings are generally favorable to the use of the night lights data to improve the accuracy of simple autoregressive model-based forecasts, we also find a substantial degree of heterogeneity across countries on the estimated relationships between light emissions and economic activity: individually estimated models tend to outperform pooled specifications, even though the latter provide more efficient estimates for out-of-sample forecasting. The estimation uncertainty affecting the country-specific estimates tends to be more pronounced for low and lower middle income countries. We conduct bootstrapped inference in order to evaluate the statistical significance of our results.
night lights remote sensing business cycles leading indicators panel models