Known Errors in the Census Bureau’s Vintage 2022 Population Estimates: Guidance for Data Users

In May, 2023, the Census Bureau’s Population Estimates Program released its vintage 2022 population estimates for Connecticut towns (see our prior blog post on the data release). In these estimates, some towns showed a surprisingly large population increase from 2020 to 2021, which was maintained in 2022. The largest percentage increase was in Mansfield, which showed a 24% increase in its estimated population from 2020 to 2021. We have worked with our colleagues at the Connecticut Department of Public Health (DPH), who are tasked with certifying the official population estimates for the state, to uncover the reason for this increase.

DPH has determined that the Census Bureau’s vintage 2022 population estimates are artificially high for certain Connecticut towns in 2021 and 2022 due to pandemic-related issues with the counts of residents in group quarters facilities, such as college and university student housing. DPH is working with the Census Bureau’s Population Estimates Program to correct this error for the vintage 2023 population estimates which will be released next year.

This blog post:

  1. discusses what happened;

  2. explains the impact on population estimates for specific towns, planning regions and the state;

  3. delineates which population estimates the state will be using for official purposes; and

  4. provides some recommendations for data users on which version of the population estimates to use, and where to find these estimates.

 

What Happened?

During the first year of the COVID-19 pandemic, the population living in college group quarters facilities (student dormitories) dropped steeply as students left college residential facilities to return home and continue their studies virtually. Prisons and nursing homes, also classified as group quarters in Census data, also saw decreases in their resident populations during the pandemic. To understand how this impacted the population estimates, we must first discuss how the Census Bureau incorporates group quarters population data into their annual estimates.

Each year, CT DPH submits a “Group Quarters Report” to the Census Bureau’s Population Estimates Program as part of the Federal-State Cooperative for Population Estimates. This report contains yearly counts of the population residing in each group quarters facility in the state, compiled by DPH from prison statistics, the state’s annual nursing home survey, and direct communications with colleges and other institutional facilities. For the yearly postcensal population estimates, the Census Bureau uses change over time in the counts from the Group Quarters Reports to make annual adjustments to the 2020 Decennial Census group quarters population counts. This process is illustrated below.

Note: GQ = Group Quarters. GQR = Group Quarters Report.

Importantly, the Population Estimates Program does not use the 2020 counts in the Group Quarters Report to adjust the 2020 Decennial Census base count; they only use the change in the group quarters population from one year to the next to estimate how the group quarters population has changed since the decennial census. Due to the steep pandemic-related drop in the group quarters population over the summer of 2020, for many facilities the April 1st, 2020 Decennial Census count was much higher than the 2020 count provided in the Group Quarters Report due to facilities accurately reporting a steep drop in residents during the early months of the pandemic. Additionally, many facilities also accurately reported a large increase in their population from 2020 to 2021 in the Group Quarters Report as their resident population rebounded. Consequently, for towns with group quarters facilities that had a pandemic-related drop in residents in 2020 followed by an increase in 2021, the Census Bureau’s approach of adding year-to-year change from the Group Quarters Report to the 2020 Decennial Census group quarters population base count (which was much higher than the 2020 counts in the Group Quarters Report) resulted in artificially high population estimates beginning in 2021.

This situation is illustrated below for the town of Mansfield, where UConn’s Storrs campus housing is located. Since UConn’s campus housing accounted for nearly half of Mansfield’s total population in 2019, the effect of the pandemic-related group quarters count issue is particularly dramatic here.

UConn’s Storrs (Mansfield) campus population was 12,312 in April of 2020 but had dropped by 7,730 residents by the time the counts for the 2020 Group Quarters Report were collected. When UConn’s on-campus housing population rebounded by 6,119 from 2020 to 2021, the Census Bureau simply added this population increase to the 2020 Decennial Census base count without accounting for the 7,730 drop between April of 2020 and the date of the 2020 Group Quarters count. This resulted in an artificial boost of 7,730 in Mansfield’s population in both 2021 and 2022.

Although Mansfield was not the only town affected, the relative impact of pandemic-related group quarters count inconsistencies is much smaller in other Connecticut towns.

 

Why Wasn’t this a Problem for Vintage 2021?

The 2021 Group Quarters Report, which provides counts for 2020 and 2021, was not requested by the Census Bureau’s Population Estimates Program for the vintage 2021 population estimates. Due to delays in the availability of detailed demographic data from the 2020 Decennial Census, the PEP had developed what they called a “blended base” approach to estimating the Census base count for the vintage 2021 population estimates. The blended base estimated the total resident population in 2020 and was not broken out by the household population versus group quarters residents. Therefore, for vintage 2021, the Census Bureau’s Population Estimates Program did not utilize a group quarters base population count to which they could apply change over time in the group quarters population from the Group Quarters Report.

In contrast, for vintage 2022, the Population Estimates Program changed their methodology for calculating the “blended base” to provide separate estimates for the household population and group quarters residents in 2020. Therefore, the vintage 2022 estimates were the first post-2020 population estimates to incorporate information about change over time from the Group Quarters Report.

 

Which Towns Are Affected?

Connecticut’s Department of Public Health has determined that, for eleven towns, their 2022 population is overestimated by at least 1% due to the pandemic-related group quarters count issue. See the table below for details. Fifty-nine other towns had overestimates of less than 1%.

 

How are Population Estimates for Planning Regions and the State Affected?

In the Census Bureau’s methodology for generating population estimates, group quarters population counts are added to the household population counts at all geographic levels. Consequently, the town-level overestimates also result in overestimates of the planning region and state level populations.

The Department of Public Health has concluded that, statewide, Connecticut’s 2022 population is overestimated by approximately 20,000 to 25,000 people (0.6% to 0.7%) due to the group quarters count issue. This roughly equals the estimated statewide population increase of 20,608 (0.57%) since 2021, using the vintage 2021 estimate as the base. Thus, Connecticut’s population likely remained fairly stable from 2021 to 2022.

The population overestimation by Planning Regions is provided in the table below. The Capitol Planning Region, which contains Mansfield, had the largest percentage overestimation of 1.2%. This was followed by the South Central Planning Region, which includes New Haven, with a 0.9% overestimation.

 

Which Population Estimates will the State be Using for Official Purposes?

To avoid discrepancies with the American Community Survey and other Census Bureau data products, the Department of Public Health is certifying the Census Bureau’s vintage 2022 population estimates as the official population estimates for the state of Connecticut, while noting that there are some known errors in the estimates for certain towns.  

However, given the known errors in the official 2022 estimates for some towns, the state will be using the vintage 2021 population estimates for fiscal allocation purposes.

 

How Will This be Corrected for Next Year’s Estimates?

Connecticut’s Department of Public Health has submitted the 2023 Group Quarters Report to the Census Bureau with the special addition of group quarters counts for April 1st, 2020, as well as the usual annual counts for 2020, 2021, 2022, and 2023. By providing the April 1st, 2020, counts, the Census Bureau will now be able to subtract the population loss between the Decennial Census date and the 2020 group quarters count. The vintage 2023 estimates released in 2024 will incorporate the revised group quarters inputs and more accurately reflect the town populations in 2022.

 

Guidance for Data Users

If you need population estimates that align with the state’s fiscal allocations, you should use the Census Bureau’s vintage 2021 population estimates until the vintage 2023 estimates are released.

For other use cases, we provide some suggestions below based on our calculations of the relative accuracy of the vintage 2021 versus vintage 2022 estimates for different geographies, after adjusting for the group quarters count issues. Regardless of which population estimates you decide to use, we strongly advise updating your population statistics for 2021 and 2022 when the Census Bureau’s vintage 2023 population estimates are released.

Towns:

  • For the towns of Mansfield and New Haven, we suggest that you use the Census Bureau’s vintage 2021 population estimates. Our analyses suggest that these more accurately estimate the 2022 population adjusting for the known group quarters count issues in these towns (an estimated 8.8% vs. 31.9% overestimation of the 2022 population in Mansfield, and 0.3% vs. 3.1% overestimation in New Haven).

  • For all other towns, the vintage 2022 estimates should provide a similar or more accurate estimate of the 2022 population.

Planning Regions:

  • For the Capitol and South Central Connecticut Planning Regions, we suggest using the vintage 2021 population estimates. Our analyses suggest that these more accurately estimate the 2022 population adjusting for the known group quarters count issue (an estimated 0.5% vs. 1.2% overestimation of the 2022 population in the Capitol Region, and 0.4% vs. 0.9% overestimation for the South Central Planning Region).

    • Unfortunately, vintage 2021 demographic characteristics of the population are only published at the county level and cannot be directly calculated for planning regions. For planning region demographic characteristics, vintage 2022 estimates will need to be used.

  • For all other planning regions, the vintage 2022 estimates should provide a more accurate estimate of the 2022 population.

State:

  • We suggest using the vintage 2021 population estimate for the state (3,605,597), since the statewide population increase from the vintage 2021 to vintage 2022 estimates can be entirely attributed to the group quarters count issue.

 

Where to Find the Population Estimates

  • Vintage 2021 population estimates were not released for planning regions, but the town-level estimates can be aggregated to planning regions using our town-to-planning-region crosswalk. If you’re looking for vintage 2021 planning region total population estimates, we’ve done the work for you; see the table to the right.

 

For More Information

See here for DPH’s report on the state’s official 2022 population estimates. For details on how the Census Bureau generates their population estimates, see the Census Bureau’s Population Estimates Methodology page.

Check out our Event Calendar for census-related events.

If you are interested in learning more about CTData, check out our mission and values and the services we provide. For training and tips on how to use data to inform your personal and professional life, register for one of our CTData Academy workshops or browse our blog. You can keep up with us by subscribing to the CTData newsletter and following us on Twitter, Facebook, and LinkedIn.

Census & ACS, Demographics & Population, Town & RegionalCynthia Willnerpopulation estimates