We are investigating changes in fertility behaviour in the USA between 1976 and 2007. We use CPS data to recover fertility but we observed some inconsistency with USA census data, specially between 1980 and 1990. (CPS show a fertility increase and the Census a decrese). My question is if its correct to use the Census and what problems could be with that.
Differences in the rates would potentially arise for two reasons: (1) different questions are asked on different surveys and (2) the CPS is a much smaller sample drawn from the civilian non-institutionalized population whereas the ACS is meant to represent the entire U.S. population.
Furthermore, it should be noted that in years prior to 1990, there is little consistency in the universe for fertility and marriage history variables. You should consult the comparability and universe tabs for the variables that you are using for additional details. Here is additional documentation about the fertility and marriage supplement. One resource that you may find valuable is a comparison of the available birth data which is provided by the Census Bureau. Ultimately, which you decide to use is up to you.