Count Question Resolutions and ACS data

I'd appreciate any experience or knowledge anyone might have around Count Question Resolutions and any impact they might have on historical ACS estimates.

I ask because a colleague told me today that a state demographer told him that “the vintage estimates could change in the future if Count Question Resolutions are accepted by the Census.”  Has anyone had experience with this?

Even if the vintage estimate were updated, could there be an impact on ACS estimates?  We have serious discontinuities between ACS population estimates and the decadal number for a number of cities in Michigan, so this could really impact a number of areas.  Thanks.

Parents
  • hi Jon--

    Your posting says their were some big discrepancies or "discontinuities" between 2020 counts versus your expectations coming out of ACS 2019 (or population estimates v2019). Population estimates can be wrong. Decennial census counts can also be wrong, though for different reasons.

    But about the ACS... Once ACS products are published, they are published and final. ACS Office does not go back and revise past products (unless there was a huge error the likes of which we have not seen since ACS began).

    Count Question Resolution errata changes could affect the Census PEP estimates series though. This affects ACS later on (not until after 2023), as ACS receives its population control totals from Census PEP.

    While Census does not go backward and revise ACS products... corrections are made to the PEP "base" or starting point -- and for all years, really. Census PEP allows revision of its entire time-series of county population estimates with each annual cycle. Data users may not realize this because changes are typically small, and typically due to minor updates in demographic inputs.

    Having said all that, I expect local population changes out of Count Question Resolution will be limited. Census Bureau will make many small changes due to CQR cleanup of geog boundaries errors (city and town boundaries). But CQR operation has been extremely frugal about "coverage" errata corrections; these coverage corrections would be where they entirely missed housing and population, and there was evidence of error, and the Census Bureau was willing to concede it. I have in my notes: there were only 116 successful "coverage" cases nationwide following 2010 Census = extremely rare.

    hope that helps

    --Todd Graham
    Metropolitan Council Research

Reply
  • hi Jon--

    Your posting says their were some big discrepancies or "discontinuities" between 2020 counts versus your expectations coming out of ACS 2019 (or population estimates v2019). Population estimates can be wrong. Decennial census counts can also be wrong, though for different reasons.

    But about the ACS... Once ACS products are published, they are published and final. ACS Office does not go back and revise past products (unless there was a huge error the likes of which we have not seen since ACS began).

    Count Question Resolution errata changes could affect the Census PEP estimates series though. This affects ACS later on (not until after 2023), as ACS receives its population control totals from Census PEP.

    While Census does not go backward and revise ACS products... corrections are made to the PEP "base" or starting point -- and for all years, really. Census PEP allows revision of its entire time-series of county population estimates with each annual cycle. Data users may not realize this because changes are typically small, and typically due to minor updates in demographic inputs.

    Having said all that, I expect local population changes out of Count Question Resolution will be limited. Census Bureau will make many small changes due to CQR cleanup of geog boundaries errors (city and town boundaries). But CQR operation has been extremely frugal about "coverage" errata corrections; these coverage corrections would be where they entirely missed housing and population, and there was evidence of error, and the Census Bureau was willing to concede it. I have in my notes: there were only 116 successful "coverage" cases nationwide following 2010 Census = extremely rare.

    hope that helps

    --Todd Graham
    Metropolitan Council Research

Children