Is it (at least theoretically) possible to obtain new marriages from 1979 onwards using the CPS? Perhaps comparing marital status between waves of the rotation groups? Just thought about asking before embarking on that task.
Yes, you should be able to do this by making use of the longitudinal aspect of CPS. If you haven’t done this before I suggest checking out the IPUMS page on linking the CPS. Unique individuals are identified by CPSIDP. You should note that CPS samples housing units, and does not follow individuals who move out of their house. So if people get married and move in the same month, they will not be captured in these data.
That’s what I feared. I suppose the same happens for ACS.
What do you think about using differences from say the census in 1980 and the CPS-ASEC in 1981 to assess new marriages by year, and then successive differences in CPS-ASEC until the census in 1990? I’d have to pay close attention to weighting, but as I understand, the Census Bureau calibrates the weighing of the CPS to reflect the actual counts in the census.
The method you propose seems reasonable to me for calculating net changes in number of married people. However, if you want the number of marriages you can’t just count the change in number of married people since there are also divorces and deaths happening. I’d also be cautious about disaggregating too much with the CPS, since you’ll get very small sample sizes. You could use the ACS in the same way starting in 2000, to increase your sample size.
The ACS also has a question (MARRINYR) about whether a person was married in the past year. But it’s only available starting in 2008.
Depending on your needs, you may also want to check out data from the CDC on marriages.