The American Community Survey (ACS) makes use of the Population Estimates Program’s (PEP) official population estimates as survey controls in order to improve the ACS estimates for those characteristics which overlap with the PEP. This results in ACS estimates being equal or similar to the PEP estimates for total population as well as having similar age, sex, Hispanic, and non-Hispanic by race distributions.

We do this for two reasons. The first is to improve the consistency of estimates between the ACS and the official population estimates produced by the Census Bureau. The second is to mitigate coverage error by demographics, where certain demographic groups may be under or over-represented in the estimates. By using independent estimates taken from the PEP, we accomplish both goals of making the ACS data more accurate as well as more consistent.

Original Plan for Use of Population Estimates in 2021

In Population Controls for the 2020 ACS (census.gov) we outlined our standard use of population estimates as controls for years that follow a census year including when the ACS 5-year estimates straddle the census year. For the 2021 data year, we will use the July 1, 2021 population estimates, Vintage 2021, as controls for the ACS 1-year estimates and the average of the July 1, 2017 – July 1, 2021 population estimates as controls for the ACS 5-year estimates. Of particular concern is the version or “Vintage” of population estimates that are used as controls. Our standard methodology is to use the latest and most consistent set of population estimates available from the PEP. Our plan, pre-2020, was that the July 1 estimates for 2017 through 2019 would be sourced from the 2010–2020 intercensal estimates and the July 1 estimates for 2020 and 2021 would be sourced from the Vintage 2021 estimates. Overall, both the intercensal estimates and the Vintage 2021 estimates were expected to be fully informed by the 2020 Census.

Revised Plan for Use of Population Estimates in 2021 and Going Forward

Due to delays in census processing, the production of the intercensals was delayed as was the full incorporation of the 2020 Census into the Vintage 2021 population estimates. Because of this, the 2020 ACS data products made use of the Vintage 2020 Evaluation Population Estimates which were informed solely by the 2010 Census and not the 2020 Census. The 2021 ACS data products will use the Vintage 2021 “Blended Base” population estimates that incorporate the 2020 Census, Vintage 2020 population estimates, and the 2020 Demographic Analysis (DA). In general, these estimates can be described as follows:

In the Blended Base at the national level, population totals come from the decennial census, age and sex detail comes from DA, and race and Hispanic origin detail comes from the Vintage 2020 estimates.

For lower levels of geography, the population totals from the census and the demographic distribution detail of the Vintage 2020 estimates will tend to dominate. The age and sex detail from the DA generally filters down to lower levels but weakens relative to the Vintage 2020 distribution since the DA distribution is only applied at the national level. Once the estimates for the total population are formed, the estimates for the group quarters population are informed by the proportion of the total resident population that resides in group quarters as estimated in the Vintage 2020 population estimates.

The Vintage 2021 population estimates provide the basis of the controls for the 2021 ACS 1-year estimates as well as the 2020 and 2021 portion of the five years of population estimates used as controls for the 2017–2021 ACS 5-year estimates. We created population estimates for the remaining years, 2017, 2018, and 2019, by adjusting the Vintage 2020 population estimates so that the revised estimates are consistent with Vintage 2021. These adjusted estimates were created only for the universes and demographic categories that are used by the ACS weighting (for example, age was collapsed into 13 categories and population estimates by demographics were produced only for total resident population and not for household and group quarters separately).

To accomplish this, the published methodology for creating the 2000 to 2010 intercensal estimates was adapted to adjust the July 1 estimates for 2010 through 2020 Vintage 2020, treating the July 1, 2020 Vintage 2021 estimate as the anchor point (normally this anchor would be the census data for the final year). The intercensal estimates methodology then distributes the difference between the Vintage 2020 and Vintage 2021 estimate for July 1, 2020 across each year in the decade. This eliminates the otherwise abrupt increase in total population of approximately 2 million people between the original Vintage 2020 and the Vintage 2021 population estimates. Instead, the difference is spread across the decade and the adjusted Vintage 2020 estimates form a smooth time series from 2010 to 2020 and match the Vintage 2021 estimates for July 1, 2020. In addition to total population, the adjusted Vintage 2020 estimates bridge the differences between the demographic distribution of the original Vintage 2020 estimates to match the Vintage 2021 blended base estimates.

These adjusted Vintage 2020 estimates will be used for creating the controls for the 2017-2021 ACS 5-year estimates and for future 5-year estimates until the official intercensal estimates produced by the PEP are available.

Page Last Revised - August 29, 2022

Is this page helpful?
Thumbs Up Image

Yes

NO THANKS

255 characters maximum

255 characters maximum reached

Thank you for your feedback.
Comments or suggestions?

Top