Multiple respondents, same cpsid, same lineno, different cpsidp


I will ask a weirdly specific question, since I suspect the issue appears elsewhere and the logic will generalize. In 1982 November CPS, there is a household identified by cpsid 19811100821000 which has 3 2-year-old female white individuals, each with the same lineno but a different cpsidp. Thus, the cpsidp seems to classify these as different individuals, but since they have the same lineno within the same household, I would think they are the same individual and that the original tapes erroneously included duplicates. Can you offer some guidance? It seems like an error, but the documentation for cpsidp (the Drew, Flood, Warren, 2014 JESM paper) only discusses methods from 1989 onwards, so it’s hard to be sure.

Thank you,


This is an error, but it seems to be in the source variables provided by the BLS. Basically because these records are technically (if not meaningfully) unique, they remain in the IPUMS reformatted data, and are assigned a CPSIDP value as if they were unique people, even though the LINENO is duplicated. In samples prior to 1989, duplicated line numbers for records that are not exactly identical occurs regularly - though this usually affects a very small number of records. Occasionally, there will be meaningful differences in other variables between two such records, and so you might wish to investigate any potential differences before deciding which, if either, of the records to keep.

The authors of the paper you cited above are currently working on a pre-1989 CPSID paper in which they plan to document and discuss any oddities (such as this) and their implications for CPSIDP. In the meantime it may be worthwhile to reach out to the BLS and discuss these oddities. They may have more specific guidance for how to deal with these issues.