Identifying Basic monthly respondents

How can I identify or separate basic monthly respondents from supplement respondents in IPUMS-CPS March samples? Are there other months in which the IPUMS-CPS monthly samples include both Basic Monthly and supplement respondents?

Thank you.

At this time, the only data we have available for March datasets represents the supplement respondents, this means that variables from the basic questionnaire are not available for March (though there are variables in the supplement dataset that are the same as variables in the basic dataset). Because we only offer the supplement data there is no recommended approach for identifying basic supplement respondents.

You could download the Supplement and basic March datasets from NBER and perform a sequential merge between the IPUMS and NBER March Supplement datasets so that the IPUMS-CPS data would now also have the HRHHID1 and HRHHID2 variables. Using these ID variables, attempt to merge persons from the basic file also using individual identifiers such as age, sex, race, and other information available on both files to identify basic March respondents. It is a known phenomenon that Household IDS and the three person-level variables listed above do not match 100% of basic March respondents. How you deal with these unmatched cases is a decision best made in the context of your specific research topic. Because the March supplement includes oversamples (as discussed on the sample design page) not all supplement respondents will be present in the basic March file.

All other CPS supplements represent subsets of basic monthly respondents, so all respondents are both basic and supplement respondents in the context of IPUMS-CPS. If a basic monthly respondent did not fit into the supplement universe they were coded as “Not in Universe” for the relevant variables. I hope this helps.


I can add that I tried the procedure suggested by Joe_Grover for the March samples between 2005 and 2010 and it seems to work.

Moreover, you don’t even need to double check the demographic characteristics of the matches. If you sequentially merge the IPUMS with the NBER March ASEC file* while retaining the NBER’s ‘h_idnum1’ ‘h_idnum2’ and ‘a_lineno’ identifiers and then match these 1:1 with ‘hrhhid’ ‘hrhhid2’ and ‘pulineno’ in the March basic file from the NBER, you are able to match all observations in the NBER basic March file, while approx 73,000 to 74,000 observations from the March supplement remain unmatched (which I suspect would be the supplement over-sample part). The matched individuals would be those who are both part of the supplement and the basic file.

*note: before doing this, you should eliminate non-respondents from the NBER files, and perhaps sort both files by household IDs and person line nrs