Unable to reproduce weights from CPS Voting supplement

CPS
1

I have been trying to replicate the margin of error reported for the CPS Voting and Registration Supplement. Specifically, I am trying to replicate the margin of error reported for racial/ethnic groups as reported in this table: https://www2.census.gov/programs-surv…

I am using the replicate weights from the Census websight and following the instructions for implementing these weights in Stata that are listed on this website. However, I am unable to produce standard errors that would result in the margin of error they report in the sheet.

I assume that the use of replicate weights is the correct way to try to reproduce these margins, but am I missing something?

Here is the stata code I used to try to reproduce the national figures:

svyset [iw=weight], sdrweight(repwgt1-repwgt160) vce(sdr) mse

* Create indicator for citizen voting age respondents
gen citva=1 if citizen==1 & voting_age==1
recode citva .=0

svy, subpop(citva): mean vote, over(race)

2

As far as I can tell, it looks like you are using replicate weights correctly. Do note that we do not generally expect to exactly replicate published statistics. (See more about this here.) Published statistics often use restricted access data that includes various levels of detail not included in the public use files. So, while we do expect to get relatively close and within some margin of error exact replication is not necessarily expected. Therefore, I am not sure how close your calculations of the margin of error are to those listed in the published table, but if you are close I wouldn’t be too concerned. Additionally, the MOEs in the published table produce the 90% confidence interval around point estimates. This is slightly different from the default 95% confidence interval.