I wish to identify the number of children under 14 of each mother in a household, and the number of children without a mother present and their parent or guardian, if any, for the years 1968 through 1973. My purpose is to back-cast children for the years 1962-1967, by regressing number of own children on a vector of economic and demographic variables. I don’t expect to do so with any precision, but I believe this technique, possibly with some added forcing to census totals, can make those years comparable to later years on some dimensions I wish to study, such as adult-equivalent household income and poverty rate.
For this purpose I do not care if children are biological, adopted, or fostered.
I see how to identify children under 14, with AGE, and I see how to identify number of own children present, with NCHILD. And I believe I can identify number of children with no mother present by subtraction. But for households with more than one mother present and one or more children who are over 14, the total number of children of mothers will exceed the total number of children under 14, so sum rule to allocate the under-14 children to the proper mother s is needed. I have not found a good method to do this. Can you suggest an approach using variables available furring these years?