I’m calculating survey-weighted childcare cost estimates using SPMCHXPNS, weighted by SPMWT (SPM unit/family-level weight), and I need to compute standard errors that account for the CPS complex sample design. Should I pair SPMWT with REPWTP (person-level replicate weights), or is there a different replicate weight set intended for SPM-unit-level estimates? Any guidance on the correct varmethod for this pairing in SAS would also help.
The Census Bureau recommends using the ASEC replicate weights for analyses of SPM variables. You can read their recommendations, including sample code, beginning on page 8 of the SPM read me file. IPUMS provides the ASEC replicate weights and sample code: Replicate Weights in the Current Population Survey. In the extract system, find the replicate weights under person → technical variables, or household → technical variables.
Since the SPM variables are constructed for supplemental poverty units, which are slightly different from the concept of households, the household-level weights (whether replicate weights or simply sampling weights) are not perfectly aligned with the unit of analysis. Nonetheless, the Census Bureau recommends using these weights with the SPM variables, so you should feel empowered to do so.