Should you use WTSUPP and the replicate weights for state level analysis using the CPS?

Hello, I am working on a project where we are looking at person-level pension data on a state-by-state basis from a pooled IPUMS CPS file for 2011-2014. We’re looking for a little help on making sure we choose the correct weighting for this analysis.

If we just had a national-level, single year analysis it looks like we could use the WTSUPP weight (since the variable is from the March Supplement) and we could also employ replicate weights. However, does anyone know whether we could use the same weights for a state-level, pooled analysis? Or would we have to make additional adjustments? We’ve looked around a bit but have not found anything definitive.

I would appreciate any input anyone could share. Thank you!

CPS data is intended to be representative at both the national and state levels; thus, you can simply use WTSUPP to weight your individual observations in order to obtain parameter estimates. Replicate weights (REPWTP) should be used only to increase the accuracy of your standard error estimates. The general method for using WTSUPP and REPWTP is the same whether you are using a national sample or state subsamples.

Hope this helps.

1 Like