Abstract
An initial reaction among many observers of urban America to recently released data showing a net migration out of its metropolitan areas is that this phenomenon must reflect nothing more than an accelerated expansion of these areas beyond their conventionally defined borders. This paper tests this hypothesis by tracing out the behavior of the Hoover index for five levels of areal disaggregation in the US. In Statistical Geography, Duncan and his collaborators found that for the period 1900–1950 there was dispersal at the finest and coarsest grains of areal disaggregation (reflecting city—suburb and East—West dispersal, respectively) and concentration at intermediate grains (reflecting rural-urban migration). We find that, by 1970, dispersal was occurring at all levels of areal disaggregation. That is, using the county as our basic unit of analysis, and building up increasingly more aggregated regions based on these units, we are unable to find an increase in concentration at any level of areal aggregation. We conclude that dispersal is more than a statistical artifact of the way in which metropolitan areas are defined.