ACS Data Users Group

I am doing trend analysis on ratio of income to poverty table B17001G for Travis County Texas

In 2024 5-year ACS the estimate for the total was: 259,918 and in 2019 5-year the total was: 43,531. This jump in the population seemed very suspicious to me. So, I dug into table B03002 (Hispanic or Latino Origin by Race) and in 2024 Not Hispanic Two or More Races estimate was 56,128 and in 2019 was 29,401. These estimates seem probable but, when I compared 2024 Hispanic or Latino Two or More race estimate (208,443) to 2019 estimate (14,719) of the same category the difference is staggering.

The data dictionary doesn't seem to indicate a change in methodology. Does anyone have any insight as to what may be happening with the data? Is this just an error?

Hi Rochelle,

This is likely due to a methodological change in how the Census Bureau codes and tabulates race responses, rather than a data error. Beginning with the 2020 Census/ACS processing changes, the Census Bureau redesigned aspects of the race question and overhauled how write-in responses are coded. The largest impacts were among Hispanic/Latino respondents, where write-ins such as "Mexican" or "Salvadoran" became more likely to be classified into multiracial categories rather than "Some Other Race" alone.

Your comparison is also between two non-overlapping ACS periods: the 2019 5-year file (2015–2019), which is entirely pre-change, and the 2024 5-year file (2020–2024), which is entirely post-change. So you're seeing the full effect of the change.

The Census Bureau provides a note explaining the improvements to the race question here: https://www.census.gov/programs-surveys/acs/technical-documentation/user-notes/2021-03.html

Possible workarounds:
(a) Combine "Two or More Races" with "Some Other Race" so the redistribution effect is minimized, or
(b) Focus trend analysis on broader categories such as NH White, NH Black, NH Asian, and Hispanic/Latino total, which are more stable over time.