ASECWT is based on the inverse probability of selection into the sample and adjustments for the following factors: failure to obtain an interview; sampling within large sample units; the known distribution of the entire population according to age, sex, and race; over-sampling Hispanic persons; to give husbands and wives the same weight; and an additional step to provide consistency with labor force estimates from the basic survey.
However I find that for some observations, ASECWT is not equal to ASECWT_SP. Granted, those are very few cases corresponding to 0.01% of my data. Is there more documentation on how those weights are constructed?
The attached characteristics features are based on the IPUMS-derived family interrelationship variables.These relationships may not always match couples defined by the Census Bureau/BLS. I looked at a few ASEC samples (one per decade) and found that the bulk of the mismatched ASECWT ASECWT_SP cases (70% in my quick check) are householders linked to unmarried partners (which IPUMS classifies as partnered via SPLOC and therefore the attach characteristics features would capture this as well). I looked at the SPRULE values as well, which report how the spousal link was determined (and our confidence in that linkage). Based on the SPRULE values for the remaining cases, it doesn’t seem unreasonable to me that these couples are only partnered by IPUMS and not in the original CPS data. You may be interested in using SPRULE to determine if you are interested in keeping couples as defined by SPLOC.
For more information on the construction of weights, I would encourage you to look at CPS Technical Paper 77.
Thank you for the clarification. Just one more follow up. Are those links that are made by IPUMS reflected in the marst variable? That is, for people that are matched by IPUMS will always show up as married according to marst?
Also, I realize that I didn’t label my original question as “CPS” so it shows up as “General.” I wanted to correct that, but can’t find how to do it, or even if it is possible, is it?
Not necessarily. For example, “unmarried partners” should not have a MARST value that indicates they are married, but the IPUMS family interrelationship variables will still pair this person to the householder (and vice versa) via SPLOC and SPRULE. MARST is in the original CPS data, while SPLOC and SPRULE are created by IPUMS.
Also, I was able to update the topic of your forum post to “CPS” instead of “General.”