Higher urban densities are almost a Holy Grail among many urban planners and urban economists. But as I may have said before, because of data limitations we usually take these measures over areas that are too large and varied for the averages to be useful.
The PUMS areas (PUMAs) are smaller and closer in size to what we would call a neighborhood. There are just over 2,000 of them in the U.S. that are within metro areas; these 2000+ had an average population just over 139,000 in 2005.
Trouble is that as of now, there are only data for these areas for the years 2005-2008. In that interval, many areas became denser and many became less dense. The relationship between initial density and increased densification was very weak (very complex) with a correlation of less than -0.10.
Some would say that the benefits of density are clear and the dense should/would become even denser. This is apparently not what's going on.
We will know more once we get results from the 2010 census, but only if we stop discussing city-wide or metro area-wide averages.