Responding to: “I am trying to figure out the county composition of pwmetro, can I use the compositin file for “metarea”(https://usa.ipums.org/usa/volii/count…)?”
In short: yes and no.
The composition tables you linked to accurately describe the metro areas on which both METAREA and PWMETRO are based. I.e., we use the same metro area definitions as the basis for both METAREA and PWMETRO.
However, neither METAREA nor PWMETRO can exactly identify the populations (or worker populations) of the specific set of counties comprising every metro area. Census PUMS files don’t identify counties or metro areas; they identify only PUMAs (and place-of-work PUMAs), which sometimes straddle multiple counties. As noted in the PWMETRO variable description: “PWMETRO identifies the metropolitan area in which the respondent worked, if the respondent’s workplace was in an identifiable metropolitan area, given confidentiality restrictions,” and METAREA uses a similar protocol. This means that any persons who lived in a PUMA, or worked in a PWPUMA, that straddled a metro area boundary are assigned “Not identifiable” codes in METAREA or PWMETRO, respectively.
This page identifies the portions of each metro area’s population that are coded as “Not identifiable” in the METAREA variable. Because place-of-work PUMAs are occasionally larger than PUMAs, and may straddle metro area boundaries even where PUMAs do not, PWMETRO will have higher rates of unidentified populations for some metro areas. Unfortunately, however, we do not at this time have any summary information available on the unidentified populations for PWMETRO. In order to determine what portion of a metro area is not identified by PWMETRO, you’d have to use the metro area county composition info in conjunction with PUMA and PWPUMA composition info. If you have any familiarity with GIS or mapping software, you may also use boundary files from IPUMS USA and IPUMS NHGIS to determine how PWPUMAs correspond to metro areas.
The complexity of this problem is why we took a different approach for the MET2013 variable, and will be taking a different approach in a new version of PWMETRO (“PWMET13”) for 2000 and ACS samples that we plan to release later this fall. For these new variables, we identify metro area codes only when PUMAs align reasonably well (within a 15% population tolerance) with the metro area, and we provide more complete composition information, as previewed in my answer to your other User Forum question.