Essentially, Any county-level estimate of population density is trash. What is needed is a continuous measure of density, rather than an areal unit of density. The pixel maps of employment density generated at <http://onthemap.ces.census.gov/> represent such a method. It would be difficult but feasible to generalize such a method to GIS. It would require assuming population uniformity at the block level, but doing so would certainly be better than doing so at the County level.
If you aren't familiar with them, I recommend Jarrett Walker's (Human Transit) blog; specifically, the blog posts on 'The Perils of Average Density' and 'Portrait of a City as a Squiggly Line'.