We estimate the factors determining specialization of crop choice at the level of individual fields, distinguishing between the role of natural advantage (soil characteristics) and economies of density (scale economies achieved when farmers plant neighboring fields the same). Using rich geographic data from North Dakota, including new data on crop choice collected by satellite, we estimate a model of how a farmer plants adjacent fields under the farmer's control. We find planting decisions on a field are heavily dependent on the soil characteristics of adjacent fields. Through this relationship, we back out the structural parameters of economies of density.

This paper shows that plants located in areas where an industry concentrates are larger, on average, than plants in the same industry outside such areas. In some sectors, such as manufacturing, the differences are substantial. The connection between size and concentration is stronger than what we would expect to find if plants were randomly distributed like darts on a dartboard.

Theory suggests that vertical disintegration should be greater in areas where industries localize. This paper provides some evidence that this implication is true for the U.S. manufacturing sector. Purchased inputs as a percent of the value of output is used as a measure of vertical disintegration. To measure the localization of industry, for each manufacturing plant the amount of employment in neighboring plants in the same industry is determined.