CHBORN in 1950 and 1960


I am having issues analyzing the demographic variable CHBORN in the 1950 1% and 1960 1% US censuses. For my work I focus on women born between 1893 and 1903 (so aged 47-57 in 1950 and 57-67 in 1960).

In the 1950 1% data, I weighted each CHBORN value using the PERWT variable. I find that about 63% of women aged 47-57 had 0 children. This is obviously wrong, the number should be 7-8%. However the distribution of CHBORN omitting childless women is comparable to that from the 1960 data. I am trying to figure out why there is such an excess of CHBORN=0 in that sample.

I also ran into a problem with the 1960 data: no women declared more than 13 children in that census (CHBORN is 13 or less). This is odd because in 1950 about 0.6% of women who had children had more than 13. It seems odd that the 1960 data would be truncated at 13.

I would appreciate any resources or advice that might help solving these two issues.

Many thanks,

As noted on the universe tab for CHBORN, in 1950, CHBORN is only available for ever-married sample-line females, age 14+. These analyses should be weighted with the variable SLWT instead of PERWT. You can learn more about sample line persons and weighting here. In contrast, in 1960, the universe is ever-married females, age 14+; using PERWT for this sample is appropriate.

Regarding the truncation of CHBORN in 1960, as described on the codes tab for this variable, in 1960-1990, responses to CHBORN are top-coded at 12 children. This means that women who had more than 12 children are assigned values of 12 in these years of data.

For both issues you raised, it is also worth noting that the code of 0 indicates not applicable, and the code of 1 indicates no children. After removing the “not applicable” category, to get the actual number of children ever born, users must subtract 1 from the value of CHBORN.