The economies of developing countries are often characterized by market failures, but the causes of these inefficiencies remain incompletely understood. Here we use an experimental approach to study market interactions in a developing country from first principles. In particular, we ask whether basic predictions of neoclassical price theory hold in a simple market with participants from an informal settlement in Nairobi, Kenya. In developed countries, neoclassical price theory has been shown to accurately predict convergence and equilibrium in such markets. We use a classic double auction design, in which sellers set a price and buyers make a purchasing decision. All sellers have the same reservation price, and all buyers have the same, higher reservation price, creating a surplus. Since sellers have unlimited supply and buyers freely choose from which seller to buy, the predicted equilibrium transaction price is the sellers’ marginal cost. We find that both offer and transaction prices converge rapidly to the theoretically predicted equilibrium. We find evidence for learning-by-doing, in that sellers learn to optimally set prices in the first few rounds of the game. In addition, we find evidence for learning-by-observing: when buyers switch into the role of sellers, they set prices optimally from the very first round. Optimal behavior, and thus profits, are strongly correlated with cognitive skills, especially mathematical ability. Together, these results suggest that neoclassical price theory accurately predicts basic market interactions in developing countries.