Estimating Impact in Partial vs. General Equilibrium: A Cautionary Tale from a Natural Experiment in Uganda

This paper provides an example where sensible conclusions made in partial equilibrium are offset by general equilibrium effects. We study the impact of an intervention that distributed information on urban market prices of food crops through rural radio stations in Uganda. Using a differences-in-differences approach and a partial equilibrium assumption of unaffected urban market prices, the conclusion is the intervention lead to a substantial increase in average crop revenue for farmers with access to the radio broadcasts, due to higher farm-gate prices and a higher share of output sold to traders. This result is consistent with a simple model of the agricultural market, where a small-scale policy intervention effects the willingness to sell by reducing information frictions between farmers and rural-urban traders. However, as the radio broadcasts were received by millions of farmers, the intervention had an aggregate effect on urban market prices, thereby falsifying the partial equilibrium assumption and conclusion. Instead, and consistent with the model when the policy intervention is large-scale, market prices fell in response to the positive supply response by farmers with access to the broadcasts, while crop revenues for farmers without access decreased as they responded to the lower price level by decreasing market participation. When taking the general equilibrium effect on prices and farmers without access to the broadcasts into account, the conclusion is the intervention had no impact on average crop revenue, but large distributional consequences.