Estimating municipal economic activity: An alternative data-based approach

Abstract

Remote sensing techniques permit characterizing urban landscapes through urban spectral indices that distinguish built from unbuilt areas. These indices are considered a good proxy for economic activity as they reflect the development of the country, state, or municipality. However, urban spectral indices easily confuse built-up areas with bare land, compromising economic estimates. This confusion is associated with the resolution of the satellite image, the spectral index, and the threshold value that classifies pixels as built-up or unbuilt areas. This research introduces the Constrained Double-threshold method that allows for the reduction of this confusion. The method allows constraining the confusion between built and unbuilt areas to a certain percentage (e.g., 10%), reducing this error by 2.9 times and the overall error by 7%. The paper results show that the capacity of the spectral index to estimate economic activity depends more on the confusion between built-up and non-built-up areas than on the overall classification error. The two best-performing spectral indexes showed high explanatory power (R = 0.88; R = 0.86) and prediction (R2 = 90; R2 = 0.62) of the economic activity of the municipalities. The estimated growth in economic activity between 2018 and 2019 obtained from satellite images of 7.2% is consistent with official data for civil construction of 6.9%.