[OSM-talk] Post code areas
david.fawcett at gmail.com
Fri Apr 2 15:48:19 BST 2010
Another thing to note.
In the US, Zip Codes do change. In fact, due to closing post offices,
the data is more volatile than it has been in a long time.
On Fri, Apr 2, 2010 at 9:13 AM, Paul Houle <paul at ontology2.com> wrote:
> Anthony wrote:
>> The actual areas are basically only useful for reverse geocoding
>> (click a spot on the map and get the postal code). But whether or not
>> that's even possible is highly dependent on whether or not the post
>> office provides such information. For some post offices, such
>> information is not meaningful. What is the postal code for the middle
>> of a highway? Maybe there is one defined, which represents what the
>> postal code would be if there were a post box there. But maybe there
>> isn't. It depends on the post office.
> In the US, ZIP codes are really sets of points (addresses that get
> delivered to), not geographies.
> There are commercially available ZIP code boundaries that are made
> by a process of assigning plausible shapes that fit the points known to
> be in the ZIP codes. I suppose these are done with alpha shapes or
> maximum-margin or some similar algorithm.
> There's a lot of demand for ZIP code-indexed data in the US because
> everybody is familiar with them, and because they're roughly on the
> scale of marketing activities: you're likely to drive to stores that
> are in your ZIP code or nearby ZIP code. On the other hand, the more I
> learn about ZIP codes the more I learn how bad they are: for instance,
> ZIP codes are insufficient to determine a person's congressional
> district, commonly cross counties, and I'm aware of at least one ZIP
> code that spans two US states. ZIPs are also too big to do effective
> geodemographics: for instance, the ZIP code 14850 (most of "Ithaca,
> NY") contains some very rich neighborhoods, some very white
> neighborhoods, and also some neighborhoods that are poor and minority.
> There are good commercial databases, however, that give geodemographic
> profiles at the census block or individual household level: enough that
> you could find out that the most of the 'rich' people in 14850 are
> college professors who don't spend ostentatiously so they wouldn't
> support a Nordstrom's or a Jaguar dealership.
> Personally I like the TIGER county shapes for spatial control in the
> US. These are accurate and tile nicely and, I find I that the union of
> several counties is generally a good proxy for the kind of 'semantic
> regions' that I work with... For instance, even if I'm targeting a
> city, the noosphere density falls off so much in the suburbs that the
> county is an effective boundary: and if something out in the 'burbs has
> that much 'interestingness' it's probably semantically associated with
> the city anyway.
> I'm currently establishing a spatial control system for the world
> and it's probably going to be based on second-level administrative
> divisions, though I've got good third-levels for a lot of interesting
> talk mailing list
> talk at openstreetmap.org
More information about the talk