How can I use Census data to estimate the number of children that should be attributed to CPS households, 1962-67?

I am trying to estimate the number of children under 14, by race and Hispanic status, for the years for which the CPS does not count such children, 1962-1967. Specifically, I want to use Census population estimates for this age range or IPUMS-USA data for 1960 and 1970 to impute values in the CPS. Is there a straightforward way of estimating the population of such CPS counterfactually eligible children from Census estimates – some correction factor for the different scope of eligible universe in the two data sets?

Will the number of such children – children covered by the CPS but not the Census, or visa versa – be substantial? I am thinking that this would be children in group quarters plus the children of members of the armed forces. I am not sure how the CPS treats employed spouses of armed forces members in terms of eligibility. Are there other groups of children where the difference in coverage is important?

In principle, it should be possible to construct projections based on age, race, and state. IPUMS USA has 1960 and 1970 census data with data on children under the age of 14. Additionally, information on children under the age of 14 exists for IPUMS CPS samples from 1968 onward, and this data could be used to validate the projections. The one potentially tricky part, as you indicate, is making sure the universes (e.g. GQ status) lines up. The ASEC samples includes only non-institutional group quarters. So, when using the Census data you’ll want to be sure to restrict the sample to all those excluding GQ==3 “Institutions”. Note that most of the restrictions on armed forces individuals relate to labor force questions and not to these basic demographic questions. I think that should cover it. Good luck!

Thanks, Jeff! That’s very helpful.

I think for my purposes, which mainly relate to income distribution, I would want to exclude children from households outside of the CPS income question universe in both samples. Do you know of any instances besides GQ households? What about military personnel living off-base?

Later I may be attempting to combine these sources. We really ought to include military personel, both on and off-base, in looking at overall income distribution trends. People in prison too. But that is a task for another day.

For basic questions about demographics in the CPS, I think the main difference in the universe (in comparison with the Census) relates to GQ status. Since the restrictions on armed forces individuals only really relate to the labor force questions, questions relating to demographics and even income are asked of all people. You can always check the Universe Tab on the IPUMS variable descriptions to varify these details.