# Calculating standard errors with ACS replicate weights for PUMA-level aggregated data

**URL:** <https://forum.ipums.org/t/calculating-standard-errors-with-acs-replicate-weights-for-puma-level-aggregated-data/7242>\
**Category:** USA\
**Created:** [November 3, 2026, 3:28pm UTC](https://forum.ipums.org/t/calculating-standard-errors-with-acs-replicate-weights-for-puma-level-aggregated-data/7242 "2026-11-03T15:28:49Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Estefanie\_Aguilar\_Pa](https://avatars.discourse-cdn.com/v4/letter/e/97f17d/32.png) [@Estefanie\_Aguilar\_Pa](https://forum.ipums.org/u/Estefanie_Aguilar_Pa)\
**Post date:** [November 3, 2026, 3:28pm UTC](https://forum.ipums.org/t/calculating-standard-errors-with-acs-replicate-weights-for-puma-level-aggregated-data/7242/1 "2026-11-03T15:28:49Z")

</div>

Hello. I am requesting advice on how to handle the replicate weights included in the ACS microdata when calculating standard errors for my analysis.

I aggregated 1-year ACS person-level microdata for 2012–2024 to the PUMA level to create a PUMA × year panel for six states. To do this, I summed person-weighted (PERWT) numerators and denominators within each PUMA-year and harmonized 2010- and 2020-vintage PUMAs using crosswalk allocation factors before computing rates. I’m estimating the effect of a state-level policy in PUMAs with high incarceration rates using a triple-difference (DDD) design with fixed effects and standard errors clustered at the state level.

Since I aggregated the data to the PUMA level, I am unsure how to handle the replicate weights to calculate appropriate standard errors. Should I aggregate each of the 80 replicate weights the same way I aggregated the person-level data, re-estimate the model with each set of replicate rates, and apply the successive difference replication formula? Or are clustered standard errors from the regression sufficient in this setting? Any advice here would be greatly appreciated.