How to create a sample of mothers and the father of her children and children for South Africa, 2001 and 2007?

I am working on creating a study sample using the South African 2001 Census and 2007 Community Survey data. The main variable of interest we are looking at is the number of children born alive who have died for each mother (CHBORN minus CHSURV). We’d like to construct information on maternal and paternal summary statistics. The sample we’d like to create would include all individuals identified as mothers, with "paternal’ (‘parental’ as in the men who are fathers to her children) and child variables added on to each maternal record. I am having some trouble figuring out how to do this using the ‘pointer’ variables.

First, there seems to be 2 variables that can be used to determine if a female is a mother:

CHBORN (universe is woman ages 12-50)

NCHILD (universe is all individuals)

These variables do not capture the same information. For example, I’m looking at one household in the sample. There is a father (age 57), a mother (age 58), and one “child” (age 26). Both the mother and father are coded as having 1 child of their own in the household, but the mother is coded as not in the universe for the children ever born variable.

Would it be best to determine if a woman is a mother based on CHBORN or NCHILD?

Second, I’m confused on how to merge the "paternal’ (‘parental’ as in the men who are fathers to her children) and child variables to each maternal record. I think that this can be done with the ‘pointer’ variables. Is there a tutorial or more detailed information available on how to do this with SAS?

Thank you for any help!

The two variables you have identified (CHBORN and NCHILD) have different universes, as you noted, but they are also different in two other important ways. CHBORN is directly taken from the source data where NCHILD is constructed by IPUMS-International based on MOMLOC (also constructed by IPUMS-International). This brings up the second important difference, MOMLOC identifies social as well as biological relationships, which will create inconsistencies between CHBORN and NCHILD.

It is also important to note that CHSURV uses the same universe as CHBORN and would thus have similar conflicts with NCHILD.

Finaly, sample SAS code for merging records based on family interrelationship is available in the first End Note of the Family Interrelationships page here.

I hope this helps.