accurate zip code boundaries? no such thing
Q: [1] Does anyone have what they deem to be accurate zip code boundaries? [2] If so, where did you get them or how did you build them? I have not yet found a reliable source of zip code boundary data… including the Post Office.
A: [1] There are no good zip code boundaries, because zip code boundaries are interpolated—much like an isoline—from addresses. The actual address location is given a zip code first, then the boundary is interpolated. Therefore the boundary can vary due to different source data used [parcels, centerlines, census blocks, etc.]*.
[2] The best way to make zip code boundaries is to use your local data. Use the zip codes from parcels, centerlines, etc.… use the local mailing address zip codes and work from there.
[ note why would one want to use zip codes as boundaries anyway? cartomio, if you or mb are listening, feel free to comment, because we know you have thoughts on this issue. ]
sourced from a forwarded thread on a list
* Conversations with postmasters in the Madison area revealed some interesting zip code facts:
- they can overlap [two zips for same location]
- they can be stacked [more than one zip in the building]
- they might not exist at all [for very rural areas]
- they can change or split [due to new development]
- the USPS does not define zip code boundaries [they are a cluster of addresses]
- ZIP = Zone Improvement Plan [for trivia fans]
The post masters also said that zip codes are assigned based on efficiency of delivery—but that's not always the case. Example: recently a subdivision was added in Madison, WI near post office station 53716, but that station was at capacity, so a station in McFarland [53558] agreed to take on the additional mail, even though delivery would be far less efficient.
December 15th, 2006 at 4:47 pm
Why would one want zip code boundaries? So you can do displays based on zip codes. I’ve got a client right now who wants that.
You can sort of fake it and use the Census ‘ZCTA’ - zip code tabulation areas. These are not true zip codes, but mostly they are fine. And arguably, for doing analysis they are better than zip codes, since they are roughly the same as the zip code areas, but they are consistant, so your data won’t jump about simply because a new zip code is created by splitting addresses from an existing zip.
I am fiddling with ZCTA’s right now. I’ve got the data loaded into Postgis, but there is some screwy/broken topology in the data that is keeping me from generating a kml file to export.