Merging with Census Rank-Swap Values

I am trying to merge rank-swap values that go back to 1975 available from census to IPUMS CPS data using year, household sequence number (h_seq), and person record (pppos). The data from census has only year, household sequence number (h_seq), and person record (pppos). So there is no other way for me than to use these three variables. I think hseq available on IPUMS is equivalent to h_seq. However I can’t find a variable that is equivalent to pppos. I believe that pernum is not equivalent to pppos as per this post. Is there a variable on IPUMS equivalent to pppos or is the only workaround for me to merge IPUMS to NBER CPS by exploiting the same order of NBER and IPUMS CPS to obtain pppos and then merge to census rank-swap values using year, hseq, and pppos?

I think your proposed solution of merging IPUMS files to NBER files sequentially, and then linking to the rank-swap values is still your best bet. Please be sure to verify your matches on AGE, SEX, and RACE.

1 Like

Great, thanks! Just to make sure that I would do the merge correctly. I would download the raw ASEC CPS data from IPUMS in stata format and leave it in the order it came in. I would also download the raw ASEC CPS data from NBER in dat format, use a dictionary to convert it to stata, and leave it in this order. Then I will merge the ASEC CPS data from IPUMS with ASEC CPS data from NBER sequentially, so first line to first, second to second, etc. And finally I would check if age, sex and race is the same across each match.

Since NBER CPS data comes from the census, I could also use the raw CPS data files from census, right?

Exactly! And yes, you could use the files direct from Census, too.

1 Like

Sorry, one last question. Would the CPS data from CEPR work? Do you have any experience with the extracts from CEPR?

I don’t know if the CEPR files have the same ordering, you’d need to reach out to them or check it yourself.

Just wanted to give an update if people look at this thread in the future. I checked CEPR and the data is not in the same order as IPUMS CPS or NBER CPS or Census CPS. So the only solution is to use CPS from NBER or Census and merge it sequentially to IPUMS CPS if you need variables that are not available in IPUMS CPS. As of right now NBER does not have dictionaries and stata do-files for years before 1987. So you would have to create your own dictionaries and do-files. But it shouldn’t be too hard if you need only few variables.

1 Like

Thanks for the update. You should be able to link CEPR and IPUMS CPS files using linking keys (specifically MONTH, STATEFIP, HRHHID, HRHHID2, SEX, AGE, and PERNUM), at least for some years. Please see this thread and the threads linked therein for more information on this. This is also explained at this blog post.

1 Like