Using person and replication weights in a pooled 2003-2020 analysis

Hello IPUMS time use user forum,

I noticed that the new 2020 ATUS file has a specific person weight variable (WT20), which is designed to represent only person-days in which data collection actually occurred during 2020 and not the whole year (excluding the period between March 18, 2020 and May 9, 2020). In addition, ATUS 2020 also has a separate set of replication weight variables for this year.
If I am only interested in the inverse probability aspect, rather than inflation to person-days population size estimates, when doing a pooled 2003-2020 analysis, is it correct to use WT06 for years 2003-2019 and WT20 for 2020? Are these weights comparable?
Same question goes for replication weights. Is it correct to combine them?
If not, any suggestions on how to correctly weight such a pooled analysis?

Your reply will be much appreciated,
Thank you very much for your time.
Amit.

You may be interested in this recent working paper (part of the IPUMS working paper series) on this issue of what to do with 2003-2019/2020 ATUS data. The authors’ recommendation is to use WT20 for 2020 and WT06 for 2003-2019. Please note that the paper is currently under review; I encourage you to read it and determine whether or not you agree with the authors’ assessments.