As the Trump administration considers delaying the 2020 census to improve its chances of including a citizenship question, researchers, immigrant advocates and others have openly denounced the plan’s potential to discourage some U.S. residents from participating in the nation’s decennial population count.
Critics argue the move specifically targets Latinos and could have serious financial and political ramifications for many years to come.
The U.S. Constitution requires the federal government to count the American population every decade to determine how the 435 seats in the U.S. House of Representatives should be divided among the 50 states. If the tally is wrong, some states will get more seats than they should and others will get fewer. An accurate census also is important because population information is used to adjust the Electoral College and make decisions about how to distribute funds for hundreds of federal programs — from Medicaid and food stamps to highway construction and special education.
Despite the significant time and money spent conducting each decennial census, certain groups of residents consistently are “undercounted,” or have low response rates. A report the U.S. Census Bureau released two years after the most recent census in 2010 offers a detailed look at problems with its population estimates. While the net counts for the last three censuses were generally accurate, according to the report, a number of subgroups were miscounted. For example, the 2010 census undercounted the number of people who rent their homes by 1.1%; overcounted white, non-Hispanic residents by 0.8% and undercounted 2.1% of the black population.
“One way to think about the hard-to-count problem is to consider what groups are hard to locate, contact, persuade, and interview for the census,” officials from the U.S. Government Accountability Office write in a July 2018 report outlining ways the Census Bureau can overcome challenges to enumerating hard-to-count groups.
Another major reason for miscounts: confusion over how to answer questions about race and ethnicity. For instance, the Census Bureau considers anyone “having origins in any of the original peoples of Europe, the Middle East, or North Africa” to be white. But many people within this group don’t self-identify as white, research finds. In addition, scholars stress that Hispanics often are uncertain which racial category applies to them.
As the country prepares for next year’s census, journalists should familiarize themselves with research on who’s most likely to be missed by census surveys or counted more than once — and why. Below, we’ve summarized a sampling of academic research and government reports that examine these issues as well as studies that consider how faulty population estimates can harm communities.
Effects of an inaccurate census
Inaccuracies in the 2020 Census Enumeration Could Create a Misalignment Between States’ Needs
Strane, Douglas; Griffis, Heather M. American Journal of Public Health, 2018.
In this opinion piece published in a top medical journal, researchers explain how miscounts in the 2020 census could influence public health and health equity over the next decade.
“The results of the 2020 Census will be used to allocate trillions of dollars in federal funding to states, including support for programs vital to public health such as Medicaid and the Special Supplemental Nutrition Program for Women, Infants, and Children,” they write. “Inaccuracies in the census enumeration could create a misalignment between states’ needs and allocation of federal resources. Also, a census miscount of the population could create challenges for public health surveillance and research activities that inform public health policies and interventions.”
Given the importance of the next census, the authors express concern about the level of funding allocated for the project as well as field testing issues and leadership changes within the Census Bureau. “An estimated funding shortfall of $3.3 billion and leadership gaps currently leave the US Census Bureau without the resources it needs to prepare for the 2020 Census,” they explain. “These concerns have led to placement of the 2020 Census on the Government Accountability Office’s high-risk list, which calls attention to federal agencies and programs at high risk as a result of mismanagement.”
Effects of Census Accuracy on Apportionment of Congress and Allocations of Federal Funds
Seeskin, Zachary H.; Spencer, Bruce D. Working paper from the Institute for Policy Research at Northwestern University, 2015.
In this draft working paper, researchers estimate how much an inaccurate census tally could affect the apportionment of the U.S. House of Representatives and the distribution of federal funds.
The researchers used computer simulations to examine the relationship between variations in estimates for state populations and the expected number of House seats that go to the wrong states in apportionment. The authors also looked at how an inaccurate census would affect the distribution of funds for 140 federal programs that rely on census data to determine eligibility.
The findings suggest that between 10 and 15 House seats would go to the wrong state if the average percentage error for state population data is 4%. For comparison purposes, the most recent census had a 0.6% error, on average, across the states. If the upcoming census has an error of 4%, an estimated $60 billion to $80 billion in federal grant money would be misallocated over the following 10 years. A handful of federal aid programs, including Temporary Assistance for Needy Families and Highway Planning and Construction, would not be affected by errors in the new census, the authors note.
Balancing 2020 Census Cost and Accuracy: Consequences for Congressional Apportionment and Fund Allocations
Seeskin, Zachary; Spencer, Bruce. Working paper from the Institute for Policy Research at Northwestern University, 2018.
This draft working paper by the same researchers examines some of the same issues, but using more detailed census data. It also provides estimates based on a wider range of scenarios. This time, the researchers look at shifts in House seats assuming the 2020 census results in population errors ranging from an average of 0.6% — the amount of error in the 2010 census — to 2%. They also consider this range of errors in calculating potential misallocations among federal aid funds.
If the 2020 census error for state populations is the same as in 2010, the researchers predict that Texas will lose and Minnesota will gain a House seat. If the error for state populations is slightly higher in 2020 — say, 0.71%, on average — two additional states would be affected. Florida would also lose a seat to the benefit of Ohio. If the error increases to an average of 1.7% across the states, Texas would lose a second seat and Rhode Island would join the states that gain a seat.
The authors note that their projections might be conservative considering the country’s changing demographics. “For example, Hispanics comprise a larger proportion of Florida’s population now than in 2010, and Hispanics tend to be undercounted relative to non-Hispanic Whites,” they write.
The researchers estimate that if the error is as large as 2%, $40 billion to $50 billion in federal funding would be misallocated over the following decade. They find that “expected distortions in fund allocations increase about $9-$13 billion for each 0.5% increase in average error.” They write: “We hope the average error is much smaller than 2% or 4%, as appears to be the case for previous censuses, but the reality will strongly depend on the level of census funding.”
The Net Undercount of Children in the 2010 U.S. Decennial Census
O’Hare, William P. Chapter 4 of the book Emerging Techniques in Applied Demography, 2015.
Census counts tend to miss children, especially young children, William O’Hare, a demographer and data analyst, explains in this book chapter. The 2010 census resulted in a net undercount rate of 1.7% for U.S. residents under age 18. That means 1.3 million kids — more than three-quarters of whom were 4 years old or younger — were left out of the count, O’Hare writes.
The proportion of children omitted varies, however, according to their race and ethnicity. “The net undercount rate for Black [children] alone in this age range was 4.6% … and the net undercount rate for Hispanic children in this age range was 7.5%,” O’Hare explains. “Young Black and Hispanic children account for about two-thirds of the net undercount in this age group even though they only account for about 40% of the population in this age range.”
The author offers a few possible reasons for undercounts. He points out, for example, that census questionnaires provide room for information on six household members. Because household members often are listed from oldest to youngest, children sometimes get left off the form. “Understanding why young children experience such a high net undercount in the decennial census will require a major research effort,” the author writes.
Census Coverage Measurement Estimation Report: Summary of Estimates of Coverage for Persons in the United States
Mule, Thomas. U.S Census Bureau report, 2012.
This 39-page report from the U.S. Census Bureau assesses the accuracy of the 2010 census, offering a detailed analysis of where the census fell short. In total, the census missed an estimated 16 million people. About 8.5 million people were counted more than once.
The population of black, non-Hispanic residents was undercounted in each of the last three decennial censuses, according to the report. In the most recent census, black residents were undercounted by 2.07%. Hispanic residents and Native Americans who live on tribal reservations were undercounted in 2010 and 1990, but not 2000. Hispanics were undercounted by 1.54% in 2010 while Native Americans on reservations were undercounted by 4.88%.
People who rent their homes tend to be undercounted and homeowners tend to be overcounted. The analysis finds that the 2010 census omitted 8.5% of renters and 3.7% of owners. Vermont, West Virginia, Oklahoma and Texas were among the states with the largest net undercounts in 2010.
In this study, Matheu Kaneshiro of the RAND Corporation examines census undercounts among foreign-born residents of the U.S. in 1990 — a few years after passage of the federal Immigration Reform and Control Act (IRCA), which offered amnesty to many undocumented immigrants who arrived in the U.S. before 1982, but attempted to reduce the number of new immigrants by making it illegal for employers to hire anyone known to be undocumented.
Kaneshiro finds that the foreign-born residents most likely to be missing from the census count in 1990 “fit the stereotypical image of the ‘undocumented immigrant’ — namely, cohorts from Africa, Central/South America, Mexico, and the Caribbean (excluding Cuba), who also entered the country in periods that made them ineligible for amnesty through IRCA.” Kaneshiro writes that in 1990, “undocumented persons were more likely to feel threatened by the political environment because the recent passage of IRCA arguably created a climate of fear and distrust for undocumented populations. IRCA may have led undocumented persons to feel threatened by the government (and thereby the census) as well as by xenophobic populations who had become more exposed to anti-illegal-immigrant rhetoric.”
After analyzing data on population counts, deaths and emigration rates, the researcher estimates that foreign-born residents aged 15 to 44 had an 8.76% higher undercount in 1990 compared with 2000. Meanwhile, males tended to have a 9.07% higher undercount than females in 1990. Kaneshiro finds that arriving in the U.S. after the IRCA was passed — when unauthorized immigrants were not be eligible for amnesty — “has an even stronger effect (9.16%) on relative undercount.”
The “Other Hispanics”— What Are Their National Origins?: Estimating the Latino-Origin Populations in the United States
Chun, Sung-Chang. Hispanic Journal of Behavioral Sciences, 2007.
This study examines a “serious” underestimation of Latino residents during the 2000 census. It also provides new estimates for certain subgroups, including Mexicans, Puerto Ricans, Dominicans and Venezuelans.
Sung-Chang Chun, a research scientist at Mercy College of Ohio and former fellow of the Institute for Latino Studies at the University of Notre Dame, explains that Latinos were not undercounted so much as misclassified, possibly because of confusion over how to answer census questions about their identity. Nearly half — 46% — of the Latino population did not provide specific information about their national origin, place of birth or ancestry. Instead, they self-reported as “Other Hispanic or Latino.” Chun finds that most of the 5.7 million residents who did this were Mexican. “This seems to indicate that many Mexicans choose to identify with a general Hispanic or Latino term rather than a specific one, despite having a Mexican check box on the Latino-origin question,” he writes.
To better estimate the size of various Latino subgroups, Chun analyzed data from the U.S. Census Bureau’s 2000 Public Use Microdata Sample, which contains a range of data collected through the federal government’s American Community Survey. The survey provides more detailed information about the U.S. population than the decennial census.
Using this second set of data, Chun conducted a “statistical reshuffling” of the 2000 census data, re-categorizing more than 3 million of the residents who had selected the box for “Other Hispanic or Latino.” He could not trace the national origin of the remaining 2.6 million residents placed in this category “largely because of missing data on ancestry and place of birth and because many of them also reported non-Latino ancestry and were born in non-Latino countries, such as Germany and France,” he writes.
Chun estimates that the U.S. population of Mexican residents in 2000 was almost 7% higher than the 2000 census count of 20.9 million. His estimate for Puerto Rican residents is 4% higher than the official tally. The difference was much larger among smaller Latino subgroups. For example, according to the study’s estimates, the census should have counted 999,561 Dominicans instead of 799,768 — a difference of 25%.
Factors influencing census counts
Researching the Attitudes of Households Reporting Young Children — A Summary of Results from the 2020 Census Barriers, Attitudes, and Motivators Study (CBAMS) Survey
Walejko, Gina; et al. U.S. Census Bureau Final Report, 2019.
Officials from the U.S. Census Bureau sought to understand why young children are more likely to be undercounted in the nation’s decennial census than any other age group. For this study, researchers analyzed data from the federal government’s 2020 Census Barriers, Attitudes and Motivators Study, a survey administered in 2018 to a national sample of 50,000 households. A total of 17,283 households responded to the survey, which asked questions about the public’s knowledge and attitudes about the census and plans for participation.
Researchers found that households with young children — those age 5 and younger — were less likely to be familiar with and complete the census form than households without young children. They also were less likely to think their participation matters and that determining congressional representation was an “extremely” or “very important” use of census data. Among households with young children, 60% said there is a strong likelihood they will respond to the census, 40% reported using the internet “almost constantly” and 50% indicated a strong preference for online census forms. In comparison, 68% of households without young children said there is a strong likelihood they will participate in the census, 25% reported using the internet almost constantly and 38% preferred online reporting.
Researchers also discovered that higher-income, better educated families with young children differ in many ways from lower-income, less educated ones. For example, “only about 18% of respondents in households with young children, with incomes of $75,000 or more, were extremely concerned or very concerned about the confidentiality of answers they provide to the census,” write the authors of the report. “This rate was significantly lower than the rates for households with young children with incomes below $50,000 …” Respondents’ demographics also played a role in these differences. Those “who were not English proficient expressed greater concerns than those who were English proficient (42% versus 23%).”
Potential Explanations for Why People Are Missed in the U.S. Census
O’Hare, William P. Chapter 13 from the book Differential Undercounts in the U.S. Census, 2019.
This book chapter, written by demographer William O’Hare, examines possible reasons why the census misses some residents. O’Hare partially blames the census-taking process, including processing errors and the design of the questionnaire, which can create confusion about who is considered part of a household. For example, residents might be uncertain about concepts such as household and family. In some cases, there might not be enough space on the form to list every person living in the largest households. “The Census Bureau’s data collection methods have not kept pace with the rapidly changing American family,” O’Hare writes.
Some other reasons discussed in the chapter: People who want to conceal themselves from the federal government or don’t trust the Census Bureau to treat their data confidentially are less likely to respond to the census. People who move around a lot may not receive a census questionnaire. Also, residents who are poor, unemployed or not well educated are less likely to understand the importance of participation, the author explains. In addition, residents might be missed if their addresses are not included on the federal government’s Master Address file, O’Hare explains. Examples of addresses that might not be on the master list are illegal apartments within a multi-unit structure and homes that are located far off a road and hidden.
“Perhaps the most fundamental conclusion from the material reviewed in this Chapter is that there are many different reasons why people are missed in the Census,” O’Hare writes.
Race Counts: Racial and Ethnic Data on the U.S. Census and the Implications for Tracking Inequality
Strmic-Pawl, Hephzibah V.; Jackson, Brandon A.; Garner, Steve. Sociology of Race and Ethnicity, 2018.
This paper examines the racial and ethnic categories used since the nation’s decennial census was introduced in 1790 and how these categories can cause confusion, resulting in inaccurate counts of certain subgroups. The authors focus on Hispanics, Middle Easterners and North Africans — groups that tend to be undercounted. The authors explain that for Hispanics, “confusion sometimes arises about what race to choose since ‘Hispanic’ is listed separately as an ethnicity question, while the MENA [Middle Eastern and North African] population is currently grouped with Whites on the census, but often do not identify as such.”
The authors cite research that suggests the census count of Hispanic residents might be more accurate if race and ethnicity questions were combined. “Significant heterogeneity exists within the Hispanic population, and varying emphases on and understandings of race, ethnicity, and nationality can lead to confusion with the two-part ethnicity-race question,” they write.
The researchers write that it’s problematic to categorize members of the MENA population as white. “Relying on 67 focus groups with 768 participants, the Census Bureau found many people felt that including members of the MENA population (Egyptian and Lebanese, for example) within the ‘White’ racial category was ‘inaccurate’ or ‘wrong.’”
The authors also cite research that finds that millions of U.S. residents change how they report their racial and ethnic identity from one census to the next, further confounding census results. “Changes in racial identification along with undercounting and overcounting are some symptoms of the fallibility of census racial categories,” they write. They suggest sociologists be more involved in how racial and ethnic categories are selected and presented. “The census categories dictate not only the racial data we receive via this outlet, but also the data from a number of other federal and state surveys,” the researchers write. “The racial categories that appear on the census, therefore, both enable and constrain our analyses, conclusions, and recommendations.”
The Devil Knows Best: Experimental Effects of a Televised Soap Opera on Latino Attitudes Toward Government and Support for the 2010 U.S. Census
Trujillo, Matthew D.; Paluck, Elizabeth Levy. Analyses of Social Issues and Public Policy, 2012.
This study suggests a Spanish-language soap opera can affect Latino viewers’ attitudes about the U.S. census. Researchers find that Latinos living in Arizona, Texas and New Jersey who watched a scene from a soap opera featuring a positive message about the census expressed more support for the U.S. government and census than Latinos in those states who did not watch the clip.
In 2009, the cable television network Telemundo included information about the upcoming 2010 census in the storyline for its popular weeknight soap opera “Mas Sabe El Diablo,” or “The Devil Knows Best.” Researchers showed scenes from the show to 121 adults, 46% of whom were female and 88% of whom were born outside the U.S. Nearly all participants were familiar with the soap opera.
Researchers randomly assigned participants to watch either a scene showing one of the main characters, Perla, discussing and encouraging people to participate in the census or a scene showing Perla discussing other topics and acting on behalf of other causes. Overall, “participants who watched Perla promote and discuss the census rated the U.S. government more positively, expressed less negative affect toward government, reported greater awareness of and belief in facts regarding the census, and sought information about and publicly demonstrated support for the census,” the authors write.
When researchers looked specifically at how the 42 Arizona Latinos responded, they found that their “attitudes toward government did not differ according to which clip they viewed,” the authors write. “The census clip caused participants in Arizona to report significantly greater negative affect toward the government compared to census clip viewers in other states.” The researchers explain that Latinos from Arizona might have responded differently because the state legislature had recently passed a controversial bill requiring law enforcement to check people’s immigration status to prevent illegal immigration.
For more research on the census, please check out our write-up on how adding a citizenship question to the 2020 census could result in significant undercounts and our tip sheet on how to use census data to better cover health disparities.
This image was obtained from the U.S. Department of Agriculture’s Flickr account and is being used under a Creative Commons license. No changes were made.