Abstract
Weather-induced agricultural shocks are geographically selective, as exposure reflects both exogenous atmospheric dynamics and predetermined local geography. We exploit the knowledge of shock determinants to isolate the exogenous component, relying on counterfactual simulations to measure medium-term farmers’ adaptation strategies. Using microdata from Ethiopia and an outbreak of desert locust swarms as source of the shock, we find a shift in agricultural production composition and an increase in household farm labour to smooth consumption. A producer–consumer approach reveals the farmers’ response not to be productivity-neutral in the medium-term, while individual factors’ intensity is used to offset a decrease in partial productivity. Estimating a Cobb–Douglas production function, we find a drop in total factor productivity in treated areas, a decline that exceeds the reduced-form output loss and indicates that available adaptive responses preserve subsistence output while carrying substantial productivity costs.