ACS Data Users Group

HI, My county shares several zip codes with other counties. How can I best determine the population for the portion of the zip code that is only in my county? Thanks!

Using Geocorr at the University of Missouri I got what I think you're looking for by using ZCTA as the "source" and county as the "target" example below, link is

mcdc.missouri.edu/.../geocorr2022.html

Dear Marti,

I'm not sure how Geocorr works. Since I use R I download the mapping data and check for the tracts (or block groups in this case) that have an intersection with the chosen ZCTA geography. This method uses only the "raw" mapping data so it is pretty complicated. Once I have the tracts (or block groups) that intersect the larger geography I add up the populations for the individual tracts (or blockgroups). Likely not too helpful in this case.

Dave

Geocorr works by reducing both source and target geographies into collections of blocks -- the "atoms", if you will. We know the population of every block. So we can take all the blocks that are in ZCTA 21771 (say) and add up the populations of all those blocks that are also in county #24013, divide by the total ZCTA population to get the allocation factor, then repeat for any other counties that are partially covered by that ZCTA. That's why the allocation factors should always add up to 1, although there are sometimes rounding errors.

Yes entirely possible to do it yourself with block definition files and block populations. Did it in Florida after 2000 census to find the 1990 footprint of some areas, for comparison, that had changed radically in 10 years.
it makes more sense now than it will in 5 years as change continues

Dear Glenn,

I'm curious does Geocorr give you the fips codes for the "common" block groups ? What about statistics other than the total population ?

Dave

Geocorr doesn't provide a list of the blocks in common between the source and target geographies, if that's what you mean, although it could be tweaked to do so.

As for other statistics -- Geocorr wasn't designed to be a data tool. You can get total population, land area, and centroids, but that's all.

Since this is based on blocks you could get all the things counted by block in decennial census (subject to differential privacy) — detailed race/Hispanic, two age groups. Not from GeoCORR necessarily but the PL94 block data once you get the block definitions for your county/ZCTA

as I say the passage of time since 2020 can distort things — in Florida there were giant blocks extending across the unpopulated Everglades to pick up a few scattered people, development occurred at the edges and created many populous tracts from one piece of that “block”

Just as a comment, if you use my "brute force" method you can't just check for a "non-null" intersection between polygons. Some "adjacent" polygons may have a non-zero intersecting area because of an intersection along the border of two adjacent polygons. I think that i allow 400 sq meters or so of overlap without including the polygon within the the larger polygon.

Thank you for taking the time to reply! Since I'm not familiar with R, this is a little beyond my skillset, but I appreciate the concept as you've described it.

Hi Tim, Thank you so much for your reply - tremendously helpful! This is a great resource, and I sincerely appreciate the opportunity to expand my knowledge of ACS data. Can't thank you enough!

As a standard disclaimer, this data uses Zip Code Tabulation Areas (ZCTAs) which are technically different than ZIP codes. In some cases this difference doesn't matter, sometimes it's significant, often ZCTAs are the best you can do regardless. ZIP codes don't technically exist as geographies (they're discretely assigned to each address based on post offices and delivery routes), but since we all think of them as geographies, the Census created ZCTAs to approximate the boundaries of addresses in a given ZIP code. Usually, ZCTAs are very close to the actual on-the-ground ZIP codes, but it's uneven. ZCTAs are only updated every 10 years, so right now, they're better than they'll be in 2029. Areas with fast growth (or decline) are most likely to have changes to their ZIP codes, making ZCTA boundaries less accurate. Depending on the precision you need for what you're doing, this could be a non-issue or a showstopper.

One major difference between ZCTAs and ZIP codes is that houses on either side of a street will usually be in the same ZIP code, but they could be in different ZCTAs, because block boundaries run down the middle of streets.

Dear Bernie,

Just as a conceptual issue, addresses (via the master address file) are also used for block groups. In theory you could set up a "correlation" between the census master address file and the post office file. https://geocoding.geo.census.gov/geocoder/geographies/onelineaddress?form Glenn, I sure you know more about this than I do. Does geocorr use the actual addresses or just the Tiger polygons with the longitude and latitude ?

Neither. I use the 2020 ZCTA to Census Tabulation Block Relationship File (https://www2.census.gov/geo/docs/maps-data/data/rel2020/zcta520/tab20_zcta520_tabblock20_natl.txt -- but don't click the link, it's a very large file). The Geocorr database (aka MABLE) is just a list of blocks with every block assigned to one and only one ZCTA (and the other larger geography types).