Hello! I am trying to create consistent county (or some other sub-state) identifiers using decennial Censuses between 1940 and 1960. I understand all counties are identified in 1940 because of the 72 years rule, but that’s not the case for the later years (only bigger counties will be identified).
However, in the 1% 1950 sample, only 17% of the respondents have a valid county assigned, which is much lower than the 59% in 1960. Why is there such a discrepancy?
Alternatively, I could create SEA identifiers, but the variable only exists for 1940 and 1950. Is there a crosswalk/way to assign SEAs to respondents in the 1960 5% census? Or is there any other sub-state geographical identifier that can be used over the period 1940-1960 in order to maximize geographical coverage? Thanks!
It sounds like your primary focus is on increasing the coverage of your sub-state identifier in 1950. IPUMS USA now includes the 1950 full count file; I would recommend using this instead of the 1% sample.
I can think of two contributing factors to the difference in the share of respondents living in an identified county in 1960 and 1950. First, the 1960 1% file does not include county-level identifiers; these are only available in the 5% file; the sample density certainly contributes to some of the difference. Additionally, counties in the 1960 5% file are inferred from mini-PUMAs (see variable PUMAMINI), which have a population threshold of 50,000 and were custom defined by IPUMS in the creation of the 1960 5% sample to correspond to a set of counties and census tracts. In contrast, the 1950 1% sample infers counties from the SEA variable, which enforces a population threshold of 100,000. Between the lower population threshold and the intentional design to nest within counties, I am not surprised that the county identification rate is higher in the 1960 5% file.