A better delineation of U.S. metropolitan areas

Abstract

Metropolitan areas are a fundamental unit of economic analysis. Broadly defined, they are unions of built-up locations near each other among which people travel between places of residence, employment, and consumption. Despite the importance of metropolitan areas, metropolitan Core-Based Statistical Areas and other official U.S. delineations considerably stray from this broad definition. We develop a simple algorithm to better match it, using commuting flows among U.S. census tracts in 2000. Three judgmental parameters govern the threshold strength of commuting ties between locations to include them in the same metropolitan area, the maximum separating distance between locations, and the threshold density of outlying settlement. A parameterization that balances encompassing commuting flows and excluding sparsely settled land delineates 361 Kernel-Based Metropolitan Areas (KBMAs), in aggregate capturing almost all the population and employment of metropolitan CBSAs in a small fraction of their land area. We benchmark KBMAs against two alternative parameterizations, one that prioritizes encompassing commuting flows and one that prioritizes excluding less built-up and less near locations.