The value of announced and predictable change- AKA Why I hate ZIP codes
Today someone was looking for housing data by zipcode. I hate ZIP codes. Well I don't hate them they are decent geographic units that most people have some sense about (unlike census geography like block groups which most people don't have any sense of)
I strongly suggested that they reconsider and do the analysis by block group. This is for a couple of reasons, but they are all related. In general serious geographic data is stored by census geography, using ZIP codes means conversion. In general people who deal with serious demographic data from a spatial perspective are suspicious of any analysis done on ZIP codes. But these are both really kind of symptoms of the bigger problem.
Zip Codes Change.
Census geography does too (they try to keep block groups at around 600-3,000 people) but it changes every ten years on a predictable schedule with geographic revisions carefully noted and historical data mostly back tabulated to reflect the new boundaries. This predictable change is VERY important when you are trying to look at timespan data.
Zip Codes change when the post office decides they need to for postal, political, or other reasons. It isn't a regular timetable and while it is announced to the households who are undergoing the change finding a statewide, much less a nationwide diff file is really hard.
It is totally common for people who don't think about the geographic nature of demographic data to compare zip codes when at least one code involved has changed over the period being analyzed. It is so common that I tend to assume that IS the case for any ZIP code analysis.
Changes often happen as suburbs become urban, but they have also happened in rural areas where there have been pushes to readdress for reverse 911, sometime urban zips just grow too much. Change you don't understand or can't document is just a feature I have come to expect with ZIP code based analysis.
So even if you are not a data geek or a GIS person there is something you can learn here too- there is a real value in data analysis to have a predictable change cycle for any units or subdivisions you create so that they can be accounted for in your forward projections.
This is often a thing people don't think of, especially when you are dealing with "hidden geography" like business divisions by management structure or delivery routes, but it is worth it to think about how you change and how you might make that more predictable and easier to account for in analysis
(The picture is a graph of each zip code attache to the next numerically higher one after the state prefix and the first locality number so XX101 connects to XX102 and so forth but XX199 does not connect to XX200 it gives you a feel for how they cluster and related to geography from https://eagereyes.org/zipscribble-maps/united-states