Longitudinal cohort in TUS 2014/2015?

I’ve been working on a repeated cross-sectional analysis using CPS-TUS data spanning 2001-2015, and I would like to exclude longitudinal cohort observations in order to preserve the cross-sectional analysis. It seems that there are individuals with the same CPSIDP values in the July 2014 and May 2015 samples, but I wouldn’t expect a longitudinal cohort based on this explanation: https://cps.ipums.org/cps/tobacco_sam…

Could you clarify whether there are longitudinal cohort observations linking July 2014 and May 2015 data in the TUS?

I’ve replicated your observation that the 2014-2015 wave seems to be a longitudinal cohort. I am also unable to find any documentation from the Bureau of Labor Statistics, the Center for Disease Control, or the National Cancer Institute noting that the 2014-2015 wave includes a longitudinal component. We will be reaching out to our contacts for more clarification. In the meantime, treat the 2014-2015 wave as including a longitudinal cohort. If your aim is to pool these samples for cross-sectional analysis, then I’d suggest excluding the second TUS supplement response. We will be sure to update our documentation accordingly once we have further clarification.

I’m curious as to whether or not there has been any follow-up to this question. I have, in my own work, identified the cohort (we are only using data from self-responders and I have identified a sample of 5,642 respondents matched on sex, age, and race) and we are going to proceed with an analysis. I have done a cursory search of the epidemiological literature and can’t find any references to analyses that have used this cohort longitudinally, so one of the things I’m wondering about is how one would cite it or describe it when writing it up, absent any documentation from official sources that the cohort does indeed exist.

We have not heard back from anyone about this topic. Regarding describing these data in your reporting, I’d suggest describing the structure of the data as much as possible and empirically showing the validity of the cohort longitudinal links. Once we have any documentation to share, we will post it on our website.