question about household matching


Hi Joe,

me again. Sorry to bother a lot.

I just thought over on my problem. Even each panel for each household is short, I still want to try to merge and match data to explore a little bit. I just care about the monthly data after 2008. I have seen the paper of Julia you recommended. I just want to double check and make it clear here: for 2008 to 2016 monthly data, CPSID is OK and enough for me to get household match longitudinal data, right? Although there are a lot of concerns, but this could still be a valid and simple way, right?





Correct. The CPSID variable is primarily built from the original CPS variables discussed in that paper. For person-level matching (which is usually more prone to error in the CPS) we always recommend that users try to validate matches using variables like AGE, SEX, and RACE. Household links are largely considered more stable, but as you mention there are concerns there as well.