HEALTH INSURANCE IN TRANSITION: INSIGHTS FROM ACS QUESTION UPDATE

The following information was released by the United States Census Bureau:

Written by:

Caitlin Carter, Survey Statistician, Health and Disability Statistics Branch

Estimated Read Time: 10 minutes

Annual estimates of health insurance coverage for 2025 from the American Community Survey (ACS) were released today and highlighted in the Health Insurance Coverage by State: 2025 brief. These estimates are the first to use an updated health insurance survey question and processing system, following the 2022 ACS Content Test. As such, the Census Bureau recommends that data users compare estimates from 2025 to previous years with caution.

The ACS is the largest household survey in the United States with an annual sample size of about 3.5 million housing unit addresses.1 The ACS collects information about health insurance coverage at the time of the survey by asking respondents if they currently have health insurance coverage and asking respondents to select all the type(s) of health insurance coverage they currently have.

The ACS broadly classifies health insurance coverage as private or public. It defines private health insurance as a plan provided through an employer or a union, coverage purchased directly by an individual from an insurance company or through an exchange (such as healthcare.gov) or coverage through TRICARE. Public insurance coverage includes federal programs (such as Medicare, Medicaid and the Children's Health Insurance Program [CHIP]), individual state health plans and the Civilian Health and Medical Program at the Department of Veterans Affairs (CHAMPVA), as well as care provided by the Department of Veterans Affairs.

In the ACS, people are considered insured if they were covered by any of these types of health insurance at the time of interview. People are considered uninsured if they were not covered by any of these types of health insurance at the time of interview or if they only had coverage through Indian Health Services (IHS), as IHS coverage is not considered comprehensive. Comprehensive health insurance covers basic health care needs and excludes single-service plans such as dental, vision or prescription medicine plans.

In 2022, the Census Bureau conducted a Content Test, which aimed to test the wording, format and placement of various ACS questions. As part of the test, revised health insurance questions were examined. Prior research had found that since the ACS began asking about health insurance in 2008, Medicaid and other means-tested programs were underreported and direct-purchase coverage was overreported in the survey. This was in part due to misreporting of noncomprehensive health plans and reporting multiple coverage types for the same plan (Mach and O'Hara, 2011; Lynch et al., 2011; Boudreaux et al., 2014; O'Hara, 2010; Boudreaux et al., 2011; and Boudreaux et al., 2013).2,3

To improve question reliability and validity, the Census Bureau tested a control version of the question against a test version and compared two test question versions against each other as part of the 2022 ACS Content Test.4

Following the ACS Content Test, the 2022 American Community Survey Content Test Evaluation Report: Health Insurance was released, detailing differences in the questionnaires. The evaluation report recommended adopting the second test version of the questionnaire, with the "check all that apply" response option and an explicit response category for those who are uninsured. These decisions were communicated to the public in several Federal Register notices in 2023 and 2024.5

The 2022 ACS Content Test recommendations were implemented beginning in the 2025 ACS. In addition to the changes detailed above, the entire processing routine was updated in part to prevent over-imputation of multiple coverage types. Additionally, the coding of write-in responses was changed so that write-ins were no longer coded if a respondent selected another valid insurance type.

The 2022 ACS Content Test flagged a concern that Medicaid receipt rates declined in the test versions of the instrument and recommended additional research into the quality of Medicaid responses. Following the 2022 Content Test, the Census Bureau launched a Medicaid match study to link Medicaid administrative records from the Center for Medicare and Medicaid Services (CMS) to survey responses to evaluate how accurately respondents reported their Medicaid coverage. The study found the accuracy of reporting was not significantly different between the test and control question.6

In this blog, we expand upon the ACS Content Test analyses and compare the control question to the second test version of the question. When comparing estimates for the control question against the second test version of the question, direct-purchase coverage decreased from 12.5% to 8.3%, Medicaid coverage decreased from 17.1% to 15.1%, and multiple coverage decreased from 16.1% to 11.2%.

Direct Purchase Coverage Rates: 2021 to 2025

Direct purchase coverage is any health plan purchased directly by an individual from an insurance company or through a marketplace exchange. This coverage type was known to be overreported in the ACS compared to other surveys.7 Figure 1 details the direct coverage rates at the national level among all ages from 2021 to 2025. In 2025, the direct purchase coverage rate was 8.8% among the total population, while estimates from 2021 to 2024 ranged from 13.7% to 14.1%.

Medicaid Coverage Rates: 2021 to 2025

While Medicaid is underreported in most household surveys and the intent of the questionnaire redesign was to improve measurement of Medicaid, it was one of the coverage types that had previously been reported in multiple coverage rates. Therefore, by reducing erroneous reports of multiple coverage, Medicaid rates could potentially decline. To disentangle the impact of decreasing multiple forms of coverage, we look at both total Medicaid receipt and Medicaid alone receipt. For the purposes of this analysis, Medicaid alone was defined as respondents selecting Medicaid as their only type of comprehensive health insurance coverage. Medicaid overall was defined as respondents selecting Medicaid either alone or in addition to any other type of comprehensive health insurance coverage.

From 2021 to 2024, Medicaid alone was between 14.5% and 15.3%, and Medicaid overall was between 20.5% and 21.3% (Figure 2). In 2025, Medicaid alone was 14.3% and Medicaid overall was 16.6%. This is consistent with our analysis of the ACS Content test data, which found that Medicaid overall receipt fell from 17.1% in the control to 15.1% in the test two instrument, while Medicaid alone was not significantly different between the test and control.

Multiple Coverage Rates: 2021 to 2025

Like direct purchase, research has found that the ACS reports higher rates of multiple types of health insurance coverage than rates found in other surveys.8 Multiple coverage is defined as respondents reporting at least two types of any comprehensive form of health insurance. In 2025, 9.6% of the population had multiple coverage (Figure 3). The percentage of people with multiple coverage between 2021 and 2024 ranged from 17.2% to 17.3%.

Health Insurance Coverage Rates: ACS vs. CPS ASEC

The Current Population Survey Annual Social and Economic Supplement (CPS ASEC) is a Census Bureau survey that has been collecting health insurance statistics since 1987 and is the recommended source for official national health insurance coverage rates.9 The CPS ASEC collects national and state-level data but is not recommended for subnational estimates due to its smaller sample size when compared with the ACS.10 The CPS ASEC collects data on approximately 89,000 households. The CPS ASEC health insurance coverage questions collect coverage information at the time of the interview and ask whether a person had a specific type of coverage at any time during the previous calendar year reference period. For this blog, we are using data on coverage during the previous calendar year from the 2025 and 2026 CPS ASEC.

Like the ACS, the CPS ASEC classifies health insurance coverage as private or public. The definitions of each type between the ACS and CPS ASEC are the same. While the ACS and CPS ASEC are different surveys that collect health insurance coverage in different ways, they generally show similar health insurance coverage rates.11 However, caution should be used when comparing these estimates directly, as their reference periods and collection methods differ.12

Prior to the ACS questionnaire update, in 2024, ACS rates for direct purchase, Medicaid (both overall and alone), and multiple coverage were higher than the corresponding CPS ASEC rates. In 2025, ACS coverage was lower than CPS ASEC coverage for 3 out of 4 of these categories. The only exception was for Medicaid alone, which remained higher in ACS than in CPS ASEC in 2025.

Discussion

The 2025 ACS introduced changes to both the questionnaire and data processing, seeking to improve the quality of the health insurance estimates. Estimates from the 2025 ACS reflect similar findings to the ACS content test data, showing expected changes to direct purchase, Medicaid overall, and multiple coverage rates.

Direct comparisons between 2024 and 2025 ACS data for health insurance coverage rates require caution, because the changes in measures may be an artifact of the questionnaire and processing change rather than a true change in population coverage rates.

As a result, in our ACS brief, we did not directly compare health insurance coverage in 2024 to 2025. Data users should also be aware, the estimates for health insurance coverage in the comparison profiles, which compare 2025 health insurance estimates to the previous years' estimates, will not appear on data.census.gov. Comparison guidance and user notes were released today alongside the 1-year data. Additionally, the health insurance variable will be excluded from all 5-year products until the research to bridge the estimates is completed.

Bridging the estimates will involve creating four demonstration tables that provide unofficial 5-year estimates of insured and uninsured rates overall, as well as by age and sex, for the levels of geography covered by the standard ACS 5-year tables. These tables will not bridge the change in methodology; rather, they assume that high-level reports of insured vs. uninsured are consistent across the 5-year period, which contains data before and after the health insurance question change.

1 For more information about the ACS refer to https://www.census.gov/programs-surveys/acs/library/handbooks/general.html.

2 For more information on Medicaid measurement error in the ACS please refer to https://onlinelibrary.wiley.com/doi/10.1111/1475-6773.12308.

3 For more information on direct purchase and Medicaid misreporting in the ACS please refer to www.shadac.org/publications/estimates-direct-purchase-acs-and-medicaid-misreporting-there-link.

4 For more information about the content test and questionnaire changes refer to https://www.census.gov/library/working-papers/2023/acs/2023_Hernandez-Viver_01.html.

5 Federal Register Notices include 10/10/2023 60-day Federal Register Notice for the 2022 ACS Content Test

and 03/25/202430 day Federal Register Notice of Submission to the Office of Management and Budget .

6 For more information on the 2022 ACS Content Test and on the Medicaid Match Study, refer to https://www.census.gov/newsroom/blogs/random-samplings/2026/10/2025-acs-health-insurance-question-changes.html.

7 Refer to https://www.shadac.org/publications/estimates-direct-purchase-acs-and-medicaid-misreporting-there-link and https://www.ncbi.nlm.nih.gov/books/NBK209610/ for more information.

8 Refer to https://www.census.gov/library/working-papers/2011/demo/SEHSD-WP2011-28.html for information about how and why the ACS overreports multiple coverage.

9 For more information about the CPS ASEC health insurance estimates please refer to the latest CPS ASEC health insurance report at https://www.census.gov/library/publications/2026/demo/p60-291.html .

10 For more information on using ACS and CPS ASEC health insurance estimates please refer to the guidance for health insurance data users https://www.census.gov/topics/health/health-insurance/guidance.html.

11 You can find more information about how the ACS and CPS ASEC health insurance estimates at https://www.census.gov/newsroom/blogs/random-samplings/2024/09/health-insurance-coverage-measurement.html.

12 For more information on using ACS and CPS ASEC health insurance estimates refer to the guidance for health insurance data users at https://www.census.gov/topics/health/health-insurance/guidance.html.