I have two questions related to the CPS ASEC weight (survey year 2000-2005). I am trying to structure the data at the family level by (i) focusing on either primary family, related subfamily or unrelated subfamily; and (ii) considering the head of the family, the head’s spouse (if there is) and the head’s children as one family. In this case, is it correct that I need to use ‘asecfwt’ family weight in my analyses?
Second, after the data is set up, I need to collapse the data to the state and the demographic group level (such as number of children in family) and run a regression. I was wondering where I should apply the survey weight in this case. Should I be applying the weight 1. when I collapse the data; or 2. when I run a regression (after collapsing the data); or 3. is it unnecessary to use weights in this aggregated data?
Any advice would be highly appreciated.
Hello! I’d highly appreciate it if anyone can help me understand the questions I asked above. Thank you.
You are correct that the family weight is appropriate in analyses of families rather than households. I would, however, encourage you to be specific about how you are defining families and ensure that the ASECFWT variable is assigning values to families in the ways that you expect.
Within primary families and subfamilies (whether related or not), everyone has the same family weight.
Secondary individuals also have their own family weight, unless they are an unmarried partner of the householder, in which case they get the householder’s family weight. However their family ID is NOT the same as the householder.
The unharmonized variable UH_FAMNUM_A2 can be used to identify families as it relates to ASECFWT (and the related variables FTYPE, FAMKIND, and FAMREL).
It isn’t obvious to me what your ultimate goal is in the regression–if you are interested in getting an accurate count of families at a given geographic level, you should use the weights when generating your count. It is ultimately beyond the role of IPUMS user support to provide analytical advice about specific models, but if you have a question about which weight to use with a clear unit of analysis and set of focal variables, we can provide more detailed guidance. I hope this helps.