Zipcodes are also too large to provide much fine grain, even if they were homogeneous and the population within each zip were evenly distributed. It is simply a mistake to expect very many households in a zip code to conform to the overall demographic profile for that zip. By contrast, the profiles of households will relate much better to that of the Census block group to which they belong -- and even then, much deviation is possible, especially in neighborhoods and cities experiencing demographic shifts (redevelopment, general plan amendments, gentrification, etc.) This is why so much has been written and talked about recently regarding household-level data -- which has its own inherent problems, another subject entirely.

Do you have technical questions relating to zipcodes? Try Jonathan Sperling at the U.S. Census Bureau.

Much thanks to Deane Merrill at Lawrence Berkeley for the initial catalog of these resources.