The interactive maps and tables below illustrate low, medium, and high estimates of potential losses of federal funding to states for five programs: the Medical Assistance Program (Medicaid, children only), the Children’s Health Insurance Program (CHIP), Title IV-E Foster Care, Title IV-E Adoption Assistance, and the Child Care and Development Block Grant (CCDBG). The low-estimate scenario is based on published research by the Urban Institute, and assumes a Census count that proceeds as planned by the U.S. Census Bureau.[13] The medium and high estimates (based on research published by the Census Bureau[14] and Harvard University researchers, respectively[15]) assume that participation will be reduced due to data-privacy and other concerns resulting from federal efforts to determine the citizenship status of Census respondents. For more on our analysis, see the Methodology section.

Under existing federal funding formulas, a total of 37 states will forfeit a portion of federal funds for the five aforementioned child and family programs as a result of a Hispanic undercount in the 2020 Census.[16] Some states and the District of Columbia already receive the statutory minimum federal funding share, and likely will not see direct fiscal effects of an undercount on these programs.  This includes several states with substantial Hispanic populations, such as California, New Jersey, and New York.

Key Findings

Our analysis indicates that:

• Under each of our three scenarios, Texas (low-to-high range: $339 million–$1.4 billion), Florida ($139–$555 million), Arizona ($70–$279 million), Illinois ($61–$243 million), Pennsylvania ($55–$218 million), and Colorado ($48–$193 million) stand to lose the most in annual federal funding as a result of a Hispanic undercount (download Table 1 (Excel)).
• Across all states, the median annual loss of federal funds is about $5 million in the low-range scenario and about $20 million in the high-range scenario. For many states, the loss in federal funding will be considerably greater.
  • Thirteen states, including North Carolina, New Mexico, and Georgia, may lose more than three times the median amount, and five states—Arizona, Florida, Illinois, Pennsylvania, and Texas—are estimated to be at risk of losing more than 10 times the median amount.
• The loss to states over the 10-year period from the 2020 Census to the 2030 Census would be considerable. For example, Texas stands to lose $3.4–$14 billion, depending on the undercount scenario, while Pennsylvania is estimated to lose $550 million to $2.2 billion, assuming no changes in the federal funding formula or other congressional action.

Among the five programs considered, Medicaid for children accounts for, by far, the largest share (range, by scenario: $930 million–$3.7 billion) of federal funding lost to states (download Table 2 (Excel)). Additionally, we find that:

• States that have not expanded Medicaid eligibility under the Affordable Care Act (e.g., Texas, Florida, Georgia, North Carolina) are among those that will experience the greatest projected reductions in federal funding, but some Medicaid-expansion states (e.g., Illinois, New Mexico, Pennsylvania) are also vulnerable.
• After Medicaid, the next-largest losses in federal funding are associated with the Children’s Health Insurance Program (CHIP) (range: $24–$96 million) and the Child Care and Development Block Grant (CCDBG) program ($12–$49 million), which funds the child care subsidy. Title IV-E Foster Care and Title IV-E Adoption Assistance each account for about $7–$30 million in annual funding losses.

---

*  States without estimates will not experience any loss in federal funding as a result of an undercount based on their Federal Medical Assistance Percentage.

^a Estimates assume no changes in current federal funding formula, state per capita income, or other congressional actions.

Sources: Reamer, A. (2018); Ruggles, S., et al. (2019); Baum, M.A., et al. (2019); Brown, J.D., et al. (2019); Elliott, D., et al. (2019); and American Community Survey (2018).

---

Why focus on the Hispanic undercount in the 2020 Census?

A number of Census experts and scholars have expressed concerns that the 2020 Census may result in a more extensive undercount of Hispanics than in prior years, even without a citizenship question.[17],[18],[19],[20] While estimates vary widely, some suggest the Hispanic undercount could be 3 percent (the figure used in our low-estimate scenario) or more.[21] A number of factors, many of which are prevalent among the U.S. Hispanic/Latino population,[22],[23],[24] may contribute to an undercount:

• An ever-increasing diversity of language and culture
• Widely varying levels of literacy
• Inequities in access to digital platforms increasingly used for secure data collection
• Heightened public concerns about privacy and the misuse of personal information
• More complex family living arrangements, which may cause some households or household members to be uncounted, particularly children
• Suspicion of government-sponsored activity in general
• Fear of exposure among those who may have an unauthorized-immigrant status, as well as those who may wish to conceal their presence at an address for other reasons
• Underfunding of the Census outreach effort, as well as cutbacks in pre-2020 testing

Hispanic children are particularly at risk for being undercounted, and accounted for more than 36 percent (a disproportionate share) of the 2010 total net undercount of all children under age 5.[25]

Conclusion

About 300 federal assistance programs accounting for more than $800 billion annually depend on the Decennial Census count for their allocation of funds.[26] Although Hispanics are especially likely to be undercounted, the impact of any Census undercount will be felt in state budgets and communities throughout the country. At stake is federal funding for programs that help states improve the well-being of their residents, and their children especially.

---

Methodology

Federal funding to states based on Census data

Data collected through the Decennial Census are used directly or indirectly to determine the equitable distribution of federal resources to states, including funding for many programs affecting children and their families.[27],[28] While roughly 300 programs will be impacted by an undercount,[29] only a relative few rely directly on the headcount information from the Census. These include the following five U.S. Department of Health and Human Services (HHS)-administered programs explored in this brief[30]:

• Medical Assistance Program (Medicaid, children only)
• Children’s Health Insurance Program (CHIP)
• Title IV-E Foster Care
• Title IV-E Adoption Assistance
• Child Care and Development Block Grant (CCDBG)

We estimated the federal funds each state may lose for every Hispanic person missed in the 2020 Census count. Our approach is based on prior work by Andrew Reamer of the George Washington Institute of Public Policy, who has published estimates of the Fiscal Year 2015 (FY2015) funds foregone by states, for each of these five programs, per person missed in the 2010 Census.[31]

Funds for these programs are allocated by the federal government to states based on their Federal Medical Assistance Percentage (FMAP). The FMAP defines, for each state, the proportion of federal and state spending, respectively, available for the programs. The FMAP, in turn, depends on a state’s per capita income: The higher the per capita income, the lower the federal share, and vice versa. The per capita income is tied to the Census headcount because total state income in a given fiscal year is constant; dividing it by a smaller population estimate raises the per capita income and, in turn, lowers the FMAP. The federal share under the FMAP can vary from 50 to 83 percent; because all states receive at least a 50-percent federal share, those with an FMAP at 50 are not affected by an undercount.[32]

For example, the 2010 Census estimated the population of Texas at 25,145,561. That figure was used to calculate a per capita income of $37,104 for fiscal years 2010–2012. (The state per capita income amounts used in the FMAP formula are based on the average of the three most recent calendars years of data available from the Department of Commerce, which are often three years prior.) For FY2015 (the latest year for which data are available), the FMAP for Texas was 58.05, meaning that the federal government contributed 58.05 percent of funding for these assistance programs. A one-percent undercount in 2010 would have missed 251,456 people, resulting in a lower population estimate, and thus a higher per capita income ($37,475). That, in turn, would have changed the state’s FMAP to 57.28, leading Texas to lose $285,151,104 in federal funds. The federal funds lost per person missed in the 2010 Census is that figure divided by 251,456, or $1,134 in the case of Texas.[33]

For FY2015, 37 states had FMAPs greater than 50 and thus were affected by the 2010 undercount.[34] Another 13 states and the District of Columbia had FMAPs of 50 percent and are not affected by an undercount, because they were already at the minimum. Notably, California (with more than 15 million Hispanics in 2018) faces a potential Hispanic undercount ranging from more than half a million to as many as 2 million. However, as a state with a relatively high per capita income, its FMAP (and corresponding federal receipts) would be unchanged from the “floor” of 50 percent.

How we developed our estimates of the Hispanic undercount and the potential impact on federal funding to states

To predict the likely funding impact on states, we considered three scenarios: low (a national Hispanic undercount of 3 percent, assuming a Census survey without the citizenship question, based on the Urban Institute estimate[35]); medium (a 6-percent undercount that assumes the inclusion of the citizenship question and is based on the Census Bureau’s calculation[36]); and high (a 12-percent undercount that also assumes inclusion of the citizenship question and is based on a randomized controlled study[37]). These estimates provide benchmarks for estimating the impact of a Hispanic undercount on the 2020 Census. Estimates based on the inclusion of the citizenship question have continued relevance, for three reasons: 1) the recent controversy may already have affected the willingness of individuals to complete the survey; 2) the administration has made ongoing efforts to pursue the linking of citizenship data to the 2020 Census data; and 3) the Executive Order mentioned above explicitly directs federal officials to consider how to include a citizenship question in the 2030 Census.

We adjusted each of the national Hispanic undercount estimates to account for state-level variation in the Hispanic undercount. This involved two steps. First, we accounted for state-level differences in response to a citizenship question; such differences likely reflect a range of factors, including the geographic dispersion of Hispanic residents, local or state policies affecting immigrants, and public attitudes toward immigration. For this adjustment, we utilized data, by state, on Hispanic non-response to a citizenship question included in the 2017 American Community Survey (ACS).[38]**

Second, we integrated the state-level differences in non-response into our low-, medium-, and high-range projections of the 2020 Census undercount of Hispanics. Specifically, we weighted each of the national Hispanic undercount estimates (3, 6, and 12 percent, respectively) by the state-specific non-response rates to the citizenship question included in the 2017 ACS, such that the ratio between each state’s non-response rate and the national non-response rate remained the same as in the ACS.

We made three additional assumptions:

• We applied the appropriate state-level undercount rate to the estimated population of Hispanics in 2018[39] to estimate the number of Hispanics projected to be missed in the 2020 Census. This is a conservative estimate, given that it assumes no growth in the Hispanic population from 2018 to 2020.
• The dollar amount of funding lost per person undercounted is unchanged from Reamer’s estimates for FY2015, based on state per capita income amounts from fiscal years 2010–2012. In actuality, it is almost certainly greater, even if we consider inflation only. Moreover, estimates of fiscal losses are for a single year; cumulative losses over 10 years (until the 2030 Census) would almost certainly be greater, although specifics will depend on federal policies and on economic conditions that are subject to change.

We applied Reamer’s estimates of the fiscal loss per person missed by undercounting, based on a total 1-percent undercount in the 2010 Census. A 2020 total undercount larger than 1 percent would result in a slightly different (+/- 1–2 percent) per person loss.[40]

Tables

Table 1. Estimated Annual Loss in Federal Dollars (in millions) by States from a Hispanic Undercount in the 2020 Census, for Five Federal Programs, by Scenario

Table 2. Estimated Annual Loss in Federal Dollars (in millions) by States from a Hispanic Undercount in the 2020 Census, by Program and Scenario

---

** These data are from a 1 percent sample—about 3.19 million respondents—from all 50 states and the District of Columbia, which includes information on respondents who failed to answer specific questions (one of which was a question on citizenship). This survey has been used in several publications to draw inferences about potential non-response to a Census citizenship question, including a study by the Georgetown University School of Law, Center on Poverty and Inequality, and a judicial memo by the Census Bureau’s chief scientist (see http://www.georgetownpoverty.org/wp-content/uploads/2018/09/GCPI-ESOI-Demographic-Profile-of-People-Who-Do-Not-Respond-to-the-Citizenship-Question-20180906.pdf and https://miracoalition.org/images/Documents/Census-Bureau-Jan2018memo-citizenship-question.pdf). The primary reason we use these non-response rates is that, in some states, we expect the propensity to not respond to citizenship questions to be much higher than in others. We calculated the percentage of Hispanic respondents who did not answer the citizenship question, which ranged from 3.56 percent in West Virginia to 12.68 percent in Indiana.

---

References

[1] Reamer, A. (2018). Counting for dollars 2020: The role of the decennial census in the geographic distribution of federal funds. Report #2: Estimating fiscal costs of a census undercount to states. Washington, DC: George Washington University, Institute of Public Policy. Retrieved from https://gwipp.gwu.edu/sites/g/files/zaxdzs2181/f/downloads/
GWIPP%20Reamer%20Fiscal%20Impacts%20of%20Census
%20Undercount%20on%20FMAP-based%20Programs%2003-19-18.pdf.

[2] Vespa, J., Armstrong, D. M., & Medina, L. (2018). Demographic turning points for the United States: Population projections for 2020 to 2060. Current Population Reports, P25-1144. US Census Bureau: Washington, DC.

[3] Mule, T. (2012). Census coverage measurement estimation report. Washington, DC: U.S. Census Bureau, Decennial Statistical Studies Division. Retrieved from https://www.census.gov/coverage_measurement/pdfs/g01.pdf.

[4] O’Hare, W. P., Mayol-Garcia, Y., Wildsmith, E., & Torres, A. (2016). The invisible ones: How Latino children are left out of our nation’s census count. Child Trends Hispanic Institute & National Association of Latino Elected Officials. Bethesda, MD: Child Trends.

[5] Baum, M. A., Dietrich, B. J., Goldstein, R., & Sen, M. (2019). Estimating the effect of asking about citizenship on the U.S. Census: Results from a randomized controlled trial. Cambridge, MA: Shorenstein Center on Media, Politics, and Public Policy, Harvard Kennedy School. Retrieved from https://shorensteincenter.org/estimating-effect-asking-citizenship-u-s-census/.

[6] Kissam, E., Mines, R., Quezada, C., Intili, J. A., & Wadsworth, G. (2019). San Joaquin Valley Latino immigrants: Implications of survey findings for Census 2020. Sacramento, CA: San Joaquin Valley Census Research Project, San Joaquin Valley Health Fund. Retrieved from https://www.shfcenter.org/assets/SJVHF/SJVCRP_Survey
_Findings_Report_011819_Web.pdf.

[7] Office of the President of the United States. (2019). Executive Order on Collecting Information about Citizenship Status in Connection with the Decennial Census. Issued July 11, 2019. Retrieved from https://www.whitehouse.gov/presidential-actions/executive-order-collecting-information-citizenship-status-connection-decennial-census/.

[8] NALEO Educational Fund. (2019). The community speaks: A report of the National Latino Commission on Census 2020. Los Angeles, CA: Retrieved from https://d3n8a8pro7vhmx.cloudfront.net/naleo/pages/
1489/attachments/original/1558496505/
1.TheCommunitySpeaks-Report_1.pdf.

[9] McGeeney, K., et al. (2019). 2020 Census barriers, attitudes, and motivators study survey report. Washington, DC: U.S. Census Bureau, U. S. Department of Commerce. Retrieved from https://www2.census.gov/programs-surveys/decennial/2020/program-management/final-analysis-reports/2020-report-cbams-study-survey.pdf.

[10] Brown, J. D., Heggeness, M. L., Dorinski, S. M., Warren, L., & Yi, M. (2018). Understanding the quality of alternative citizenship data sources for the 2020 census (No. 18-38). Discussion Paper, U.S. Census Bureau, Center for Economic Studies. Washington, DC: U.S. Census Bureau. Retrieved from https://www2.census.gov/ces/wp/2018/CES-WP-18-38.pdf.

[11] Vargas, A. (2017). Respondent confidentiality: Concerns and possible effects on response rates and data quality for the 2020 Census. Presented at the National Advisory Committee on Racial, Ethnic, and Other Populations Fall Meeting, November 2017. Retrieved from https://www2.census.gov/cac/nac/meetings/2017-11/vargas-respondent-confidentiality.pdf.

[12] Kissam, E. et al. (2019). Op. cit.

[13] Elliott, D., Santos, R., Martin, S., & Runes, C. (2019). Assessing miscounts in the 2020 Census. Research Report. Washington, DC: Urban Institute. Retrieved from https://www.urban.org/sites/default/files/publication/100324/
assessing_miscounts_in_the_2020_census.pdf.

[14] Brown, J. D. (2018). Op. cit.

[15] Baum, M. A., et al. (2019). Op. cit.

[16] Reamer, A. (2018). Op. cit.

[17] McGeeney, K. (2019). Op. cit.

[18] Brown, J.D., et al. (2018). Op cit.

[19] Vargas, A. (2017). Op. cit.

[20] Kissam, E., et al. (2019). Op. cit.

[21] Elliott, D., et al. (2019). Op. cit.

[22] Jarmin, R. (2018). Counting everyone once, only once, and in the right place. Census Blogs. Washington, DC: U.S. Census Bureau. Retrieved from https://census.gov/newsroom/blogs/director/2018/11/counting
_everyoneon.html.

[23] O’Hare, W. P., et al. (2016). Op. cit.

[24] Vargas, A. (2017). Op. cit.

[25] O’Hare, W. P., et al. (2016). Op. cit.

[26] Reamer, A. (2018). Op. cit.

[27] Reamer, A. (2017). Counting for dollars: The role of the decennial census in the geographic distribution of federal funds. Initial analysis: 16 Large Census-guided programs. Washington, DC: George Washington University, Institute of Public Policy. Retrieved from https://gwipp.gwu.edu/sites/g/files/zaxdzs2181/f/
downloads/Counting%20for%20Dollars
%2016%20Programs%2006-04-17.pdf.

[28] Murphey, D., & Harper, K. (2018). The decennial census is essential for federal programs that support children. Child Trends Blog. Bethesda, MD: Child Trends. Retrieved from https://www.childtrends.org/decennial-census-essential.

[29] Reamer, A. (2018). Op. cit.

[30] Reamer, A. (2018). Op. cit.

[31] Reamer, A. (2018). Op. cit.

[32] Reamer, A. (2018). Op. cit.

[33] Based on material provided in Reamer, A. (2018). Op. cit.

[34] Reamer, A. (2018). Op. cit.

[35] Elliott, D., et al. (2019). Op. cit.

[36] Brown, J. D., et al., (2018). Op. cit.

[37] Baum, M. A., et al. (2019). Op. cit.

[38] Ruggles, S., Flood, S., Goeken, R., Grover, J., Meyer, E., Pacas, J., et al. (2019). IPUMS USA: Version 9.0. Minneapolis, MN: IPUMS. Retrieved from https://doi.org/10.18128/D010.V9.0. 

[39] U.S. Census Bureau. (2018). American Community Survey, American Factfinder Table PEPASR5H: Annual Estimates of the Resident Population by Sex, Age, Race Alone or in Combination, and Hispanic Origin. Washington, DC: U.S. Census Bureau.

[40] Reamer, A. (2018). Op. cit.