Validate ACS data compared to 2020 Decennial

I am trying to evaluate my community's population to help us plan a local community center.  I am a complete novice at this, but have accessed both the 2020 ACS and the P1 and H1 tables for PL94-171 for our two census tracts: 2010.07 and 2010.08.

In tract 2010.07, the ACS S01001 table shows an MOE +/- 571 of an estimated population 1601.  The decennial actual count in the P1 table for that tract is 2024.  Tract 2010.08 has a similar large MOE and higher decennial count than the estimate.

 Can I still use the sub categories from ACS when the actual recent census count is so much larger than the ACS?  Can I make an assumption that all age, sex, etc populations are increased in the same way and use the MOE + value proportionally to the higher actual count?

It's a small area data problem for sure.. but critical to us understanding how to serve our community.

  • Hi Jennifer,
    The usual guidance from the Census Bureau is to use ACS for share of population or rates (i.e. percent of the population who are children, poverty rates, etc.) and the 2020 Census for population totals. The 2020 Census does not have a margin of error in the traditional sense like the ACS, so I would not translate the margin of error from the ACS onto the Census count. So if you need to know the total population and basic demographics (age, sex, households) of an area, use Census; if you need to know the characteristics of the population, use ACS.
    There is the matter of differential privacy injecting noise (variance) into the 2020 Census, but I would not worry too much about it when using tract data.

    Hope this helps.
    Best,
    Howard Shih
    AAPI Data
  • Thanks I appreciate your guidance.  I figured the actual 2020 count was of course more accurate than a survey - no need for any MOE when you've counted every person.  But the gap in population characteristics is quite large between ACS and the actual 2020 count - race data being the most obvious to the quick observer.  Just wondered how the ACS for 2021 will (or won't?) incorporate 2020 actual count data to adjust the MOE or survey methodology for these census tracts given the disparity?  

  • The Census Bureau is incorporating 2020 Census data into the population estimates for Vintage 2021. Here's the methodology document: https://www2.census.gov/programs-surveys/popest/technical-documentation/methodology/2020-2021/methods-statement-v2021.pdf

    I don't think the ACS MOE will change as that would be a function of the survey design rather than any input from the 2020 Census.