I am hoping to determine how many total millennial homeowners (ages 23-38 in 2019) there are in the U.S. by state (using as recent of data as possible). Could you possibly help me determine which variables would be the most appropriate and which weights to use?
Would I need to use the OWNERSHP and RELATE variables as described in this blog post?
Thanks for any help you might be able to provide,
Matthew
The post you linked by Joe Grover is still valid and covers most of the issues you should consider. I would suggest looking at the same variables in the ACS data available through IPUMS USA, which has a much larger sample size. Since there’s not clear data on joint homeownership (at least in the CPS or ACS), you may want to report on the number of homes owned by millennials, instead of the number of millennials who own homes. In that case you would keep only those people with PERNUM=1 and use HHWT and conduct the analysis at the household level. If you want to try to identify the number of joint owners as well (e.g. by counting spouses of the householder), then you would use PERWT.
Is that PERWT in conjunction with RELATE=1 and OWNERSHIP =1?
The weight PERWT should be used with most person-level analyses of IPUMS USA data. When working with ACS data, the particular variables you are using do not generally determine which weight you should use—rather, the level of your analysis is the key deciding factor. There are cases where users may want to analyze households rather than householders, in which case they would use one observation per household (you can do this by retaining only people with PERNUM=1, or you can retain only householders by retaining only people with RELATE=1) and use the household-level sampling weight HHWT. For example, a household-level analysis would estimate the share of households that reside in rented dwellings. A person-level analysis would estimate the share of people who live in rented dwellings. You can read more about weights on our FAQ page.