This paper proposes four metrics to measure sprawl in metropolitan regions as marginal changes in land use over time. The metrics (change in urban housing unit density, marginal land consumption per new urban household, housing unit density in newly urbanized areas and percent of new housing units located in previously developed areas) are computed for all 329 metropolitan areas in the continental USA for 1980 and 2000. Regression analysis is used to explain variations in sprawl metrics across metropolitan areas, incorporating variables representing market, geographic and policy factors. Changes in development patterns reflect interactions of market and geographic structures. States with a substantial state role in planning accommodate a higher percentage of new housing units in previously developed areas and with lower marginal land consumption, suggesting that policy can mitigate sprawl development.