Using ACS data in decades-spanning longitudinal analysis

Hi all,

I am working with yearly disaster management plan panel data at the place and county subdivision levels, and I am hoping to bring ACS data to the mix. My data runs from 2000-2023, and I am hoping to cover as many of these years as possible with ACS data. I am planning to use 5-year estimates because the bulk of the places and county subdivisions in the data have small populations.

I understand that the Census advises using non-overlapping data, so I am planning on using the 2017-2021 five year estimates (because they are the most recent), which will mean that 2022 and 2023 will not be covered... Working back from there, I'm thinking it would be best to use the 2012-2016 estimates, and the 2007-2011 estimates. This would all give me data coverage from 2007-2021.

I'm wondering if there is another avenue, or way of thinking about this, that would allow me to have data coverage that extends back to 2005 (the beginning of the first available ACS estimates?), without running into the overlapping data issue. Any thoughts or advice from anyone who has worked with ACS data in a longitudinal context would be greatly appreciated!



  • I'm not really recommending this, but if you can accept a one-year overlap between 5-year data, you could do 2005-2009, 2009-2013, 2013-2017, and 2017-2021. you would have to wait until 2026 for the next drop in the series, 2021-2025.