I am interested in obtaining estimates from IPUMS-USA for child poverty by race/ethnicity at the 200% FPL. I have created an extract with race variables, poverty, age, cluster, strata. It seems that cluster and strata are for households rather than person. Is it appropriate to incorporate these into my weighting if I am doing a person level analysis?
Since the US Census and ACS data use households in the sampling process, rather than individuals within households, the use of CLUSTER and STRATA may be useful in calculating standard errors around individual-level population estimates. These population estimates, however, should be “correct” with the appropriate use of sampling weights. Applying CLUSTER and STRATA should not make a difference for point estimates. For more documentation on this topic see the Variance Estimation page and this IPUMS user note.