This paper evaluates the relationship between urban productivity and density using data on metropolitan areas. This is an alternative measure of the urban economy to the one employed by Ciccone and Hall (1996). They used data on output and education by state and employment and education by county, which excludes agricultural and mining sectors. Instead, our U.S. metropolitan area data are defined contemporaneously for the five available census years from 1950 to 1990. These data allow us to conduct both cross-sectional and panel analyses. Furthermore, since we use a model where income is a linear function of density, these data allow us to evaluate the urban system in its own right. Our results replicate the findings of Ciccone and Hall (1996). We find that a doubling of population density leads to about a 6% increase in productivity. Our results establish an important role for Jacobs externalities, measured by metropolitan area population.