Improving Commuting Zones Using the Louvain Community Detection Algorithm

Abstract

Using the Louvain community detection algorithm, I produce two new commuting zone delineations ``TS Louvain" and ``Sum Louvain" to define U.S. local labor markets. I show that TS Louvain and Sum Louvain better define local labor markets and improve statistical precision. I evaluate the TS Louvain, Sum Louvain, and ERS commuting zone delineations on several metrics: Sum Louvain has a higher proportion of workers who work and live in the same commuting zone; the modularity metric is greater for TS Louvain and Sum Louvain than ERS. Additionally, I conduct two case studies in which TS Louvain produces smaller standard errors compared to ERS, with the exception of one regression where TS Louvain produces the same size standard error as ERS. Researchers can access these commuting zone definitions at bit.ly/LouvainCZ.

Keywords: Local labor markets, commuting, measurement error

JEL Classification: C13, J21, C63