The Effect of Snow on Urban Density Patterns in the United States

Abstract

It is well known that the metropolitan areas of the American heartland generally exhibit a more spatially concentrated population distribution than do the cities of the Sunbelt. Social scientists have generally attributed this difference to social and historical factors related to city age (i.e., to variation in the timing of cities' growth, and to related differences in their economic and social characteristics). This article seeks to establish the direct explanatory importance of climate itself in determining the population concentration of cities. The data are from Guterbock and Edmonston's Metropolitan Deconcentration Project, and pertain to a large sample of monocentric 1970 Urbanized Areas (UAs) of the United States. Population concentration is measured by the Density Distribution Index (DDI), a size-independent measure based on the parameters of an estimated negative exponential density pattern for each UA. It is shown that regional differences in urban concentration cannot be adequately explained away by the city age variable. Average annual snowfall is strongly correlated with DDI, and this loglinear relationship remains strong in multivariate regressions that control for city age, growth timing, geographic shape, population size, median family income, and racial composition. The intervening effect of automobile ownership is elucidated in a path model that shows both a direct and an indirect effect of snowfall on DDI. Theoretical links between climate and urban form are discussed, and some reasons are proposed for the conspicuous omission of climate from previous research on urban concentration.