We are conducting county-level analysis for Westchester, Rockland, and Putnam counties in New York State. However, I can only find Rockland County in the COUNTYFIP variable (with STATEFIP = 36) in both the 2010 and 2014 ACS 1-year and 5-year IPUMS microdata extracts.
What could explain why Rockland is included, while a larger county like Westchester is not?
My understanding is that COUNTYFIP comes from the original Census ACS data. If Westchester and Putnam do not appear in IPUMS, can we confidently conclude that they are also not identifiable in the original ACS Census data?
Additionally, are there any workarounds or recommended approaches to obtain reliable population estimates for these counties using IPUMS or related data sources?
The COUNTYFIP variable actually does not come from the original Census ACS; instead, IPUMS add the COUNTYFIP variables to microdata records based on the spatial relationship between PUMAs (the smallest geographic unit identified in the original Census ACS microdata) and counties.
Part of Westchester county belongs to PUMA 03101. PUMA 03101 also includes part of Putnam County. IPUMS only identifies counties if PUMAs nest perfectly within counties. Since one small part of Westchester county belongs to a PUMA that crosses its boundary, we do not identify Westchester.
The US Census Bureau has a census tract to PUMA crosswalks, and you could use them to determine the fraction of Westchester county’s population that falls into PUMA 3101. If that percentage is small enough for you, you could then try to create your own Westchester county from the other PUMAs that cover it. When IPUMS identifies metropolitan areas in the microdata, we allow for a small level of mismatch between the PUMAs and the metro areas. You could implement a similar strategy for counties. I recommend reading the METFIPS description to learn more about our implementation.
Dave Van Riper
IPUMS research scientist
I wanted to add, in case you weren’t aware, that a key source of “reliable population estimates for these counties” is ACS summary tables. The Census Bureau publishes both microdata and summary tables from the ACS program. IPUMS USA redistributes the microdata. IPUMS NHGIS (and the Census Bureau’s site, data.census.gov) provide the summary tables. The ACS 5-year summary tables include data for all standard levels of census geography down to census tracts and block groups, including all counties in the U.S.
Another option for more up-to-date county-level population estimates is the Census Bureau’s Population Estimates Program, which provides annual county-level population counts broken down by a few key demographics. They just recently released their 2025 estimates. (We don’t include data from the Population Estimates Program in IPUMS NHGIS, so you’d have to get them through the Census Bureau’s site.)
These options, unlike the microdata, are all limited to the particular cross-tabulations that the Bureau has published for counties, but at least they do identify all counties.