I develop a model of agriculture on heterogeneous land to study the relation between trade, productivity, and welfare in Peru, where farmers face high internal and external trade costs. I quantify the model with new data on crop prices, yields, and land allocations. I then measure the effects of changes to trade opportunities. A policy of paving roads raises aggregate productivity (4.9%) and the median farmer’s welfare (2.7%), but increased competition from remote suppliers harms 20% of farmers. An increase in international grain prices spreads unevenly across regions, benefiting farmers but hurting urban consumers close to ports.

We introduce new measurement tools to understand the sources of earnings differences across space. Based on the natural language employers use in job ads, we develop granular measures of job tasks and of worker specialization. We find that jobs in larger commuting zones involve greater interpersonal interactions and have higher computer software requirements. Between 10% and 50% of task and technology variation between large and small commuting zones exists within occupations. Furthermore, workers in larger markets are more specialized. Tasks, technologies, and worker specialization account for a substantial portion of the market size premium even within occupations.