I would like to merge IPUMS CPS records to unharmonized “raw” Basic Monthly Files back to 1976. There are a number of forum questions that relate to merging records starting in (say) 1989, but I couldn’t find any questions about merging all the way back to 1976.
A sequential merge does not seem possible, because the order of CPS records in the unharmonized data (available from NBER) doesn’t seem to match the order of the records in the IPUMS data, as determined by the IPUMS-supplied variables SERIAL and PERNUM.
There are a number of unharmonized variables in the IPUMS data that could be used as merging keys, and I have found that using many of them can eliminate all duplicates and thus allow for a potential one-to-one match with the raw CPS files. This method works well starting in 1978, but for 1976 and 1977, there are still some duplicates even using a long list of variables (see below).
I am wondering whether I am missing some other way of identifying individual records in the CPS data that would be more efficient. For example, is there another way besides using SERIAL and PERNUM that indicates the order of the IPUMS records on the original CPS tape?
Here is the list of variables I am using to identify individual CPS records in 1976 and 1977:
- uh_hhid_b1
- huhhnum (the harmonized version appears to be populated more often)
- uh_lineno_b1
- uh_grdhi_b1
- uh_grdcom_b1
- uh_ind_b1
- uh_occ_b1
- uh_vet_b1
- uh_untno_b1
- uh_docno_b1
- uh_marstat_b1
- wtfinl (using harmonized version – same as unharmonized version)
- statefip (using harmonized version – same as unharmonized version)
- uh_intrv_b1
- uh_day_b1
- uh_cpschk_b1
- uh_hhdseq_b1
Here are the variables I use after 1977, which all seem to work well:
- For 1978-1983: hhid hhnum lineno mis month stfip race sex age
- For 1984-1993: hhid hhnum lineno mis month
- For 1994-1995: hrhhid myhhid2 pulineno hrmis gestfips hrmonth
- For 1996-2004: hrhhid myhhid2 pulineno hrmis
- For 2005: hrhhid hrhhid2 pulineno hrmonth
- For 2006-present: hrhhid hrhhid2 pulineno hrmis