ZIP Codes shouldn't be used for collecting statistics.
It is a sad fact that the primary statistical collection unit used by businesses in the United States today is the ZIP Code. Unfortunately, they're collecting a lot of mostly meaningless data that can't be compared over time.
ZIP Codes are used by the U.S Postal Service to balance mail delivery workload. They are useless for collecting statistics because ZIP codes change every day.
In my little corner of the world, representing about 1% of the population of the United states, in one year alone, the U.S. Postal Service split six ZIP codes into twelve.
Not only that, there were thousands of minor changes -- that is, one segment of street had its ZIP code reassigned to a neighboring ZIP code so that one letter carrier has the same workload as his/her neighbor.
I'm sure there are a lot of salesmen wondering why sales numbers for their territory suddenly dropped in half. I'm sure their bosses wonder that too.
There is no way you can collect meaningful statistics using collection units that writhe around like amoebae!
In an attempt to remedy this, the US Census Bureau has devised a system of "Zip Code Tabulation Areas", abbreviated "ZCTA", that approximate the Zip codes of the United States as of the 2000 Census. You can now get summary Census data by ZIP code. Unfortunately for this idea, people will continue to report that they live in the Postal Service's ZIP codes. I'm just waiting for the 2010 Census, when people will report the population of a split ZIP Code has dropped in half.