I wonder how to explain individuals, not in identifiable metropolitan areas (met2013 == 0) but with an average 2010 population of 2013 metropolitan areas not equal to zero (metpop10 != 0)?
Is it possible to get an average 2010 population of 2013 metropolitan areas (metpop10) for individuals who were not in metropolitan areas of residence one year ago (migmet131 == 0)? In general, I am interested in getting metropolitan areas’ populations before and after for individuals who have moved between metropolitan areas in the recent year.
METPOP10 does not rely on a household’s metropolitan area of residence in MET2013, but rather uses its own model and criteria for its estimates. Firstly, as the description tab notes, METPOP10 will report a metro area’s population for households residing in a PUMA only in cases where the PUMA is fully within a single metro area. This is more strict than identification in MET2013, which requires that only a majority of PUMA households reside within a metro area. In cases of overlap between a PUMA and a metro area, including whether or not MET2013 is identified, METPOP10 instead reports an approximation of the size of the commuting system PUMA households reside in. Specifically, it uses the populations of all overlaps to arrive at a population-weighted geometric mean of the size of the commuting system. For respondents living in PUMAs with no metro area overlap, METPOP10 uses a similar method with county populations instead.
The METPOP10 description links to a working paper with example figures showing how PUMAs match up with the populations used to compute METPOP10. There’s now also an open version of the published article via PubMed Central.
Regarding your second question, IPUMS has not computed average metro/micro populations for MIGPUMAs. If you’re interested in doing this yourself, we recommend using the working paper as a guide together with the migration PUMA composition files (for relationships to PUMAs) and GeoCorr (for relationships from PUMAs to counties and 2013 CBSAs).