Working with accurate ZIP code data in your geographic publication (website or report) or demographic analysis can be problematic. The most accurate dataset – perhaps the only one that could be called reliably accurate – is one that you purchase from one of the United States Postal Service’s (USPS) authorized resellers. If you want to skip the introduction on what ZIP codes really represent, jump to “ZIP-code related datasets”.
Understanding what ZIP codes are
In other words the post office’s ZIP code data, which they use to deliver mail and not to locate people like your publication or analysis, is not free. It is also, unbeknownst to many, a dataset that lists mail carrier routes. It’s not a boundary or polygon, although many of the authorized resellers transform it into a boundary so buyers can geocode the location of their customers (retail companies might use this for customer tracking and profiling, and petition-creating websites for determining your elected officials).
The Census Bureau has its own issues using ZIP code data. For one, the ZIP code data changes as routes change and as delivery points change. Census boundaries needs to stay somewhat constant to be able to compare geographies over time, and Census tracts stay the same for a period of 10 years (between the decennial surveys).
Understanding that ZIP codes are well known (everybody has one and everybody knows theirs) and that it would be useful to present data on that level, the Bureau created “ZIP Code Tabulation Areas” (ZCTA) for the 2000 Census. They’re a collection of Census tracts that resemble a ZIP code’s area (they also often share the same 5-digit identifiers). The ZCTA and an area representing a ZIP code have a lot of overlap and can share much of the same space. ZCTA data is freely downloadable from the Census Bureau’s TIGER shapefiles website.
There’s a good discussion about what ZIP codes are and aren’t on the GIS StackExchange.
Chicago example of the problem
Here’s a real world example of the kinds of problems that ZIP code data availability and comprehension: Those working on the Chicago Health Atlas have run into this problem where they were using two different datasets: ZCTA from the Census Bureau and ZIP codes as prepared by the City of Chicago and published on their open data portal. Their solution, which is really a stopgap measure and needs further review not just by those involved in the app but by a diverse group of data experts, was to add a disclaimer that they use ZCTAs instead of the USPS’s ZIP code data.
ZIP-code related datasets
Fast forward to why I’m telling you all of this: The U.S. Department of Housing and Urban Development (HUD) has two ZIP-code based datasets that may prove useful to mappers and researchers.
1. ZIP code crosswalk files
This is a collection of eight datasets that link a level of Census geography to ZIP codes (and the reverse). The most useful to me is ZIP to Census tract. This dataset tells you in which ZIP code a Census tract lies (including if it spans multiple ZIP codes). HUD is using data from the USPS to create this.
The dataset is documented well on their website and updated quarterly, going back to 2010. The most recent file comes as a 12 MB Excel spreadsheet.
2. Vacant addresses
The USPS employs thousands of mail carriers to delivery things to the millions of households across the country, and they keep track of when the mail carrier cannot delivery something because no one lives in the apartment or house anymore. The address vacancy data tells you the following characteristics at the Census tract level:
- total number of addresses the USPS knows about
- number of addresses on urban routes to which the mail carrier hasn’t been able to delivery for 90 days and longer
- “no-stat” addresses: undeliverable rural addresses, places under construction, urban addresses unlikely to be active
You must register to download the vacant addresses data and be a governmental entity or non-profit organization*, per the agreement** HUD has with USPS. Learn more and download the vacancy data which they update quarterly.
Tina Fassett Smith is a researcher at DePaul University’s Institute of Housing Studies and reviewed part of this blog post. She stresses to readers to ignore the “no-stat” addresses in the USPS’s vacancy dataset. She said that research by her and her colleagues at the IHS concluded this section of the data is unreliable. Tina also said that the methodology mail carriers use to identify vacant addresses and places under change (construction or demolition) isn’t made public and that mail carriers have an incentive to collect the data instead of being compensated normally. Tina further explained the issues with no-stat.
We have seen instances of a relationship between the number of P.O. boxes (i.e., the presence of a post office) and the number of no-stats in an area. This is one reason we took it off of the IHS Data Portal. We have not found it to be a useful data set for better understanding neighborhoods or housing markets.
The Institute of Housing Studies provides vacancy data on their portal for those who don’t want to bother with the HUD sign-up process to obtain it.
* It appears that HUD doesn’t verify your eligibility.
** This agreement also states that one can only use the vacancy data for the “stated purpose”: “measuring and forecasting neighborhood changes, assessing neighborhood needs, and measuring/assessing the various HUD programs in which Users are involved”.