Identifying matching individuals within a household in the ASEC and Monthly samples over time.

CPS
1

In producing the fields IPUMS has developed to enable consistent household matching in the CPS, has there been any effort to see if individuals in matching households also match? If so, is there anything in the IPUMS-CPS data that indicates individuals that do or do not match?

On a less ambitious note, have you tried to identify households that match despite having no matching individuals? Have you attempted to see if the householder, the person who forms the central node of your family relationship variables, is different over time? Again, if so, is there anything added to the data that identifies such households or such householders?

If I want to check these things myself, it would be useful to have some estimate of the frequency with which various individual variables are misreported, or are inconsistently reported. Have you made any attempt to measure these error rates? Are you aware of anyone who has?

Are there particular variables beyond the obvious (age, race/Hispanic, sex) that you would recommend as being particularly valuable in making such individual identifications?

2

These are good questions. Just last summer we hosted a workshop that covered many of the details associated with linking CPS data across samples. You may find many of the resources, available here, from this workshop helpful. In particular, the CPSIDP variable links individuals in the CPS across samples. Both the CPSID and CPSIDP variables are constructed by only using the linking keys available in the CPS data. Therefore, no demographic information such as AGE, RACE, or SEX are used to generate links. This means that you can verify the validity of links by using the AGE, RACE, and SEX variables. More specific detail about validating links is described in this presentation from last summer’s workshop. Finally, more documentation about linking CPS data with IPUMS is available here.

3

Actually, I think all my questions here except the last two paragraphs, both concerning doing my own matching, are answered by your documentation on CPSIDP, which I somehow overlooked. Sorry.

But I am still interested in the last two.

Edit: And I wrote this before I saw your answer above. Thanks!