Using satellite imagery to map rural marketplaces and monitor their activity at high frequency
In many rural areas of low- and middle-income countries, weekly gatherings of buyers and sellers are the most tangible manifestation of the market economy. Knowing these markets' whereabouts and activity over time could provide insights in otherwise data-scarce environments, helping researchers and policymakers to better understand poor rural economies. But these markets are by nature informal and scattered widely across often-remote regions. As a result, data on this fundamental institution are sparse and inconsistent. We develop, test, and apply a method to fill this gap, leveraging market activity's unique temporal and visual signature in satellite imagery. Using secondary data from Kenya, Malawi, and Mozambique, we first confirm that we detect markets with high sensitivity and specificity. We then derive a map of 1,776 markets in Ethiopia and track their activity at up-to-weekly frequency between 2017 and 2024. Measured market activity exhibits seasonal patterns following local agricultural calendars and responds to weather and conflict shocks. Our approach is applicable wherever satellites can regularly acquire images of rural periodic markets and requires no ground data. Once markets are mapped, our approach can be fully automated to produce an up-to-weekly measure of economic conditions in areas where such data is otherwise generally not available.