How Local Are Labor Markets?: Evidence from a Spatial Job Search Model

Abstract

We use data at the Census ward level to investigate the extent to which labor markets are "local". We present some non-structural estimates in which the probability of filling a vacancy is influenced by unemployment and vacancies in the surrounding areas, but we argue that these estimates cannot adequately estimate the true cost of distance. We then present a simple model of job-search across space that allows us to estimate a matching process with a very large number of segments. We find that the cost of distance is relatively high. That is, the utility of being offered a job decays at exponential rate around 0.25 with distance (in km) to the job. Also, workers are discouraged from appling to jobs where they expect a large number of applications. Finally, we find evidence of constant returns to scale in matching markets. The estimated model seems to replicate fairly accurately actual commuting patterns across Census wards, although it tends to underpredict the proportion of individuals who live and work in the same ward. Finally, we find that, despite the fact that labor markets are relatively "local", local stimulus to labor demand is fairly ineffective in raising the local unemployment outflow, because labor markets overlap, and the associated ripple effects in applications largely dilute the effect of local stimulus across space.