Which counties have sample sizes large enough for accurate aggregation of CPS microdata?

I am working on a research project where I need to determine county-level union density in a single year. Because the sample size for particular counties in a single year are so low, I am pooling three years’ worth of data on union membership to increase the number of observations. Nevertheless, I know that certain counties have too small of a sample size to have accurate estimates at the county level.

Does IPUMS or CPS provide a list of counties whose microdata are unreliable for aggregating to the county level because the sampling is incomprehensive or too small?

Our apologies, this post seems to have slipped past our radar. Though you have likely found an answer to your question by now, we want to provide a response for you and for future users who may have a similar question. Due to the fact that the CPS is stratified at the state level (as outlined in the Census Technical Paper 66 on sample design and methodology), it is not designed to generate sub-state level estimates, making county-level estimates less reliable (as you indicated). However, according to the March 2018 ASEC codebook (page 2-8), smaller metropolitan areas should not be used though large metropolitan areas such as NYC or LA are considered acceptable. The documentation provides general guidelines, but does not include a detailed list of counties. The decision about whether a county is reliable may be determined by the analyst according to their specific use case or if they pool enough adjacent months or years of data together.

1 Like