Hi there,
I am wondering if the algorithm to create MOMLOC, POPLOC, SPLOC,… is available. I have the complete census for Colombia for 1973 I and would like to use those variables so I can also use your sample of the 2005 census. I checked the documentation that you have available, but before trying to do it myself, I thought it would be worth it to ask.
Thank you very much for ay help.
Best,
Juliana.
The algorithm we use to create MOMLOC, POPLOC, and SPLOC is not publicly available. This has less to do with our willingness to share the algorithm and more to do with the practical challenges of implementing the algorithm outside of the internal IPUMS system. The IPUMS data conversion program does not use a standard statistical software and we do not currently have a way to share the family interrelationships algorithm. You can, however, use the information in the documentation for these variables. Specifically, the information included in PARRULE, SPRULE, and this working paper will be helpful.
Thanks very much Jeff. I’ll ask for your advice if I find any problems generating the variables.
Best,
Juliana
Dear Jeff,
I am trying to implement the algorithm to construct SPLOC and MOMLOC/POPLOC following the rules mentioned in the working paper you suggested. However, I find a bit confusing the age difference that appears in column 3 of the table Appendix 3, p.33. In particular, I find it confusing that the range for the age difference between parents and children is 10-69, shouldn’t this be 15-64? I also don’t understand why in some cases the range changes to 15-44. Could you please clarify this for me?
Thanks,
Juliana
These age differences are defined in a way so that direct reports and IPUMS probabilistic estimates largely match each other. In some cases, this requires that the age difference rule is adjusted as defined in Appendix table 3. Outside of the overall match rate, these definitions are largely arbitrary and you can change them to fit your context however you see fit.
Hi there,
I am trying to replicate MOMLOC in the 1973 sample following all the rules mentioned in the working paper. However, I find some differences between my MOMLOC and the one IPUMS generated, and I would like to understand better why.
The following are examples in which our MOMLOC differ:
SAMPLE 170197301 SERIAL 17115000: why are children (Person 4 and 5) assigned to Person 2 (instead of person 3) despite the reduced age difference range that comes with polygamous relationships?
SAMPLE 170197301 SERIAL 260653000: I think person 4 should have MOMLOC of 2 but IPUMS doesn’t generate a link for this person.
SAMPLE 170197301 SERIAL 54655000:
3.1 I think person 4 should have MOMLOC of 1 but IPUMS doesn’t generate a link for this person.
3.2 Why is PARRULE of person 3 ==12 (ambiguous)?
Can you please help me understand why IPUMS generate the links the way they appear?
Best,
J.
I’ll try to address each of these cases one at a time.
(1) In this case, the information in the RELATE variable determines the value of MOMLOC. Both person 4 and 5 are listed as a child of the head of the household. Additionally, the head of the household is identified as the spouse of person 2 via SPLOC. Therefore, MOMLOC points to person 2.
(2) This is an ambiguous case. Although person 4 is listed as a child of the head of the household via RELATE, they are also listed as married via MARST. Therefore, they have a weak link via SPLOC and are not linked via MOMLOC or POPLOC.
(3) This ambiguity comes from the same type of case as the previous case. Although person 4 is listed as a child of the head of the household via RELATE, they are also listed as married via MARST. Therefore, they have a weak link via SPLOC and are not linked via MOMLOC or POPLOC. This is the case for both person 3 and 4.
Of course, in all of these cases, you can edit the outcome of the IPUMS family interrelationship algorithm. Particularly in the case of IPUMS International data, our algorithm sometimes does not pick up all of the nuance found in a specific sample.