Spatial Business Cycles

This paper uses micro-spatial data and develops a simple theory that explicitly includes a spatial dimension to better understand the start and the diffusion of business cycles. We first document how unemployment has diffused across US counties during recent recessions and find a recurring spatial pattern. Unemployment does not increase in all counties at the same time; at the start of the recession it increases in a few specific areas and then spreads from there to the rest of the country, following an epidemic pattern, i.e. increasing first in areas that are closer to areas of high unemployment. Our theory develops a stylized model of a country, composed by many locations, where each location is connected to its neighbors by explicit trade links. In particular we assume that each location produces a differentiated good which is demanded more heavily by locations which are nearby. Labor markets are segmented at the location level and in each location there can be unemployment due to sticky wages. We then model an aggregate shock as an event that affects, at the same time, the productivity or the demand (modeled as government demand for a good produced by a given location) of many different (random) locations. We have two main findings. The first is that our setup is able to account for the spatial pattern of business cycles we document in the data. The second is that the intensity of the local trade links is an important determinant of the amplification and propagation of the initial aggregate shock. If trade links are strong, small shocks tend to be muted and not persistent, while large shocks can be very disruptive. If trade links are weak, the opposite is true, small shocks