CPS person level weights consistency with other sources


I’ve been asked to compare/validate some population estimates obtained previously from the CPS (via final and supplement weights) from dataferret with some of those with overlapping time periods I have obtained through IPUMS. I have found that the estimates are similar although slightly different. Can you provide explanations to account for the discrepancy?


The weight variables available through IPUMS-CPS are directly copied from the original census files so I would not expect any discrepancies based on weights. The differences you are seeing may be the result of recoding variables to enhance comparability across time in IPUMS-CPS. I would recommend comparing unweighted frequencies to establish if the variables provide the same counts between the IPUMS-CPS data and the dataferret data. If the counts are the same, but the weighted estimates are still different, please let us know. If you could provide a list of variables and/or a sample of your code, that would be very helpful in determining the source of these differences.