Misperceptions and Product Choice: Evidence from a Randomized Trial in Zambia

Do misperceptions deter small retailers from stocking profitable products? We apply unsupervised machine learning to the inventories of 2000 shops in Lusaka, Zambia, to identify clusters of similar retailers and flag a “target” product—one stocked by some but not all shops in a cluster. Retailers that do not stock the product systematically underestimate its profitability. We test whether misperceptions drive this gap using a randomized trial that temporarily subsidized stocking the product. Treated retailers earn similar profits to peers already selling the target product. They continue stocking it after incentives end, especially when given a longer period to experiment.