I am wondering if it is appropriate to calculate state estimates with the linked ATUS-CPS data. For example, I want to look at the leave use module outcomes from the ATUS in a single state. I have weighted the data with the LVWT weight and 160 replicate weights to get national level estimates but am unsure if I can make conclusions about state-level estimates.
According to the ATUS User’s guide, section 3, the ATUS respondents are a nationally representative sample based on the sampling frame of CPS respondents. They are stratified by a number of demographic characteristics, but not by state. It doesn’t appear that state-level estimates will be properly estimated or weighted. However, we are communicating with BLS about this and should have a more complete answer soon.
We asked our contacts at BLS and they recommend pooling 5 years of ATUS data to generate state-level estimates. There aren’t 5 years of Leave Module data, so state-level estimates for that module are not recommended.
Thank you for finding this out!