I am using the revised IPUMS-CPS data. There are a fair number of cases (hundreds in 2017) with cohabiting couples (with values>0 for PECOHAB) but with SPLOC=0 and no attached ‘spouse’ characteristics.

Is there a reason these couples wouldnt have been linked, or is it just a data error (and if it’s the latter, do you all happen to have an easy fix for it)?

thanks for any help, and for your wonderful data!


The reason for this discrepancy is PECOHAB is the self-reported cohabitation identification variable. It reports the respondent’s answer to the question: “Do you have a boyfriend, girlfriend, or partner in this household?” The SPLOC variable, on the other hand, is an IPUMS defined variable that identifies probabilistic spouses and partners within households. The method used to define SPLOC values is to combine information found in RELATE (relationship to household head), AGE, and MARST (marital status). In particular, the first priority in assigning SPLOC values is links within a household based on the clarity of the relationship between two people using RELATE. (More details about how these probable links are made can be found here.)

So, it looks like cases where PECOHAB>0 and SPLOC==0 ambiguous relationships are defined via RELATE. One way to get around this issue is the “re-code” SPLOC values in cases where PECOHAB>0 and SPLOC==0 to equal the PECOHAB values. This will not, however, allow you to “fix” the assigning of “characteristics” using the Attach Characteristics tool, since this feature uses the existing values of SPLOC.

Hmmm. Most of these cases seem pretty unambiguous–unmarried people of a similar age, who both point to each other on PECOHAB. and RELATE=housemate, other non-relative, etc.

a few examples: serial=42, 542, 2492

i’ve read thru the documentation but didnt see the detailed info on when cohabitors would be linked. is there some other reason they might not be linked?

thanks again!

I understand the suspicion. I think looking into the variable description in SPRULE is clarifying. I’m not able to find an example in the 2017 sample that have a zero value in SPLOC, but should according to the programming rules noted in SPRULE. In particular, serial 42 has a relationship between a “householder” and a “roommate”, serial 542 a “child” and an “other nonrelative”, and serial 2492 a “sibling” and an “other nonrelative”. None of these types of relationships fall under any of the linking rules found in SPRULE.


So how would you recommend identifying cohabitors, since it sounds like PECOHAB>0 is not the best way to do so, given the recent data updates?

Would you recommend doing (if marst>1 and sploc>0) or something like that?

or is there another variable that i’m overlooking that would work better to ID these cases?

Well, SPLOC already tries to identify cohabitors. However, as you’ve seen, there are instances where this designation is difficult to assign. The emphasis of SPLOC, and the other IPUMS family interrelationship variables, is on consistency and in some cases this may conflict with comparabiltiy. So, my best suggestion is still to “re-code” SPLOC values in cases where PECOHAB>0 and SPLOC==0 to equal the PECOHAB values.