To fight discrimination, the U.S. census needs a different race question

Wendy Roth has been arguing for years that the U.S. Census Bureau should ask about race in a different way. The race box that people check for themselves on the census doesn’t always match the box someone else might have checked for them. And that, Roth says, is a problem.

Roth, a sociologist at the University of Pennsylvania, began researching that mismatch in racial identification in the early 2000s. She recruited 60 New Yorkers who had been born in Puerto Rico or the Dominican Republic, showed them the census race question and asked them how they would answer. The responses surprised her.

Consider the case of Salvador, a kitchen worker in the Bronx. “Many Americans observing him would consider him to be black,” Roth wrote in December 2010 in Social Science Quarterly. But Salvador told Roth that he checks “white.”

While attitudes in the mainland United States have been shaped by the long legacy of the “one-drop rule,” in which a single drop of “black blood” conferred “blackness,” Puerto Ricans believe the opposite — that even dark-skinned people can’t be black if they have “white blood.” Puerto Ricans use terms like mulatto or trigueño to describe those falling somewhere between white and black. But when presented with race checkboxes that offer no intermediate options, Salvador simply goes by what he knows.

People from the Dominican Republic, who have long viewed themselves as “whiter” than their Haitian neighbors, likewise avoided checking “black.” Roth found that Dominicans selected “white” or “some other race” and then wrote in descriptors, such as Latino or Hispanic.  

“There are a lot of other people who don’t understand how to complete the U.S. census … because it doesn’t match their way of understanding race,” Roth says. “Sometimes they will identify in ways that are the complete opposite of what the U.S. census is trying to capture.”

The census race question is trying to capture the changing demographic composition of the country from the federal and state levels down to neighborhood blocks. The stated aim — at least for the last half century — is to help policy makers and demographers assess whether members of different racial groups have equal access to housing, education, employment and other services, as mandated by law.

Population counts from the census, which has run every 10 years since 1790, are used to determine a state’s electoral vote count and how many seats a state receives in the House of Representatives. Those data also serve as the denominator for different demographics by age, sex and race. Data gathered by other government agencies provide the numerator, such as the number of children by race in schools or the racial makeup of the prison population, says Katherine Wallman, former chief statistician of the Office of Management and Budget. The OMB makes sure every federal government agency uses the same racial and ethnic categories in data collection.

Imagine, for instance, that the National Center for Education Statistics identifies a high school as having a student body that is 90 percent black and 10 percent white. But the census data show that the school serves a region that is 50 percent black and 50 percent white. That suggests the school is more racially segregated than the surrounding region. Accurately capturing that sort of discrimination in schools requires reliable race data, both for the numerator and the denominator.

The census, though, operates under the premise that people will identify themselves in the same way as those in their society see them. For instance, a person like Salvador will check “black.” When a person’s view of their own race aligns with that of the broader society, the race data can point to areas of inequality and potential discrimination.

But people who don’t identify with the census race boxes may check a box that doesn’t reflect how society sees them. Or they may skip the question or fail to return the form, resulting in undercounts, and the race data stop working as intended.

“Discrimination is more about how you’re seen by others than how you see yourself,” Roth says.

Missing groups

That capture problem is expected to crop up again with the 2020 census. For the race question, a representative from every household, answering online for the first time, will see options within five racial categories: White; Black or African-American; American Indian or Alaska Native; Asian; and Native Hawaiian or other Pacific Islander. The OMB established those categories in 1997. Federal agencies are allowed to ask for more detail within those categories. That’s why Asians answering the census can select from Chinese, Filipino, Asian Indian, Vietnamese, Korean, Japanese and other Asian. Native Hawaiians and other Pacific Islanders can choose Native Hawaiian, Samoan, Chamorro or other Pacific Islander. Respondents can also check “some other race” or multiple race boxes. 

Missing from that list are two groups that had hoped to find their own race checkbox in the 2020 census: Hispanics and people from the Middle East and North Africa. As in past censuses, members of both groups will mostly select “white” or “some other race,” which can make it harder to locate where there may be a need for local bilingual services in schools or during elections, for example.

Plus, there’s a new wrinkle brought on by the growing popularity of genetic ancestry tests: Roth’s recent work shows that white people have been checking nonwhite boxes on the census as an expression of their newfound genetic history. “They are searching for an identity that is more specific, more interesting than being just white,” Roth says.

That murkiness in the race data has prompted Roth and other sociologists to call for more nuanced race questions. There are ways, the researchers say, to get at both how respondents view themselves — which can help build a group’s social and political power — and how they are viewed by society, a clearer metric for measuring discrimination against people like Salvador.

“I would love it if [the Census Bureau] actually measured the thing that they were trying to measure,” Roth says.

Equity meets frugality

Up until relatively recently, the race question’s raison d’être was not to undo racial segregation but to justify practices that separated racial groups, such as the South’s Jim Crow laws. But starting about 50 years ago, following the civil rights movement, the purpose of the race question was flipped on its head. 

The civil rights movement led to the gradual dismantling of laws and practices that blocked minorities from certain schools, jobs or neighborhoods and led to the passage of new laws, such as the Civil Rights Act of 1964, the Voting Rights Act of 1965 and the Fair Housing Act of 1968, in an attempt to provide equal opportunity for all.

Meanwhile, up through the 1950 census, specially trained enumerators who visited people’s homes filled out the form for each family, including the section on race. Asking directly about race was considered rude, says sociologist Carolyn Liebler of the University of Minnesota in Minneapolis. So enumerators eyeballed the person at the door and made a decision.

Assuming that most people they came across were black or white, enumerators seldom reported American Indian as a race, and they often didn’t consider that other family members may be of a different race. In the 1950 census, for instance, enumerators were instructed to “assume that the race of related persons living in the household is the same as the race of your respondent, unless you learn otherwise.”

Starting in 1960, though, officials made a money-saving decision with far-reaching consequences: They started cutting back on enumerators. By the 1980 census, almost all households received the census in the mail, and enumerators visited only those households that failed to respond to the questionnaire in a timely manner. Crucially, for the last half century, the vast majority of respondents have been identifying their own race.

So just as the Census Bureau started relying on the race question to quantify discrimination, the agency switched away from using nonfamily members to check the race box. Those seemingly unrelated changes converged to obscure the new-and-improved purpose of the race question.

Most people answering the census don’t see the race question as a way for federal agencies to track discrimination, they assume the question is asking them to express their racial identity, says sociologist Ann Morning of New York University who served from 2013 to 2019 on the National Advisory Committee on Racial, Ethnic and Other Populations, a group appointed by the director of the Census Bureau. 

Where’s my box? 

The enumerators who checked boxes for other people had little reason to take offense at the available race categories. But when respondents started filling out the census themselves, the absence of an appropriate checkbox felt like an affront. Consider individuals from Middle Eastern or North African, called MENA, countries, such as Algeria, Egypt, Iran and Lebanon. The OMB defines such individuals as white. But since the 1980s, people who trace their roots to those countries have been calling for a separate MENA checkbox.

“We believe the MENA classification should be its own separate classification divorced from the white classification,” says Khaled Beydoun, a law professor and author at the University of Arkansas in Fayetteville.

With the government doling out more than $880 billion in 2016 alone toward census-guided federal programs, a separate designation would enable policy makers to funnel some of that money to MENA communities seeking the American dream. While such dollars are rarely allocated solely on race, except in the case of American Indians, many federal granting agencies will put applications aimed at minority and underserved racial and ethnic populations at the top of the pile, says Andrew Reamer, an economic and statistical policy expert at George Washington University in Washington, D.C. When groups like MENA are folded into the white category, their applications stay buried.  

While some worry that checking “MENA” instead of “white” could lead to increased government surveillance of MENA communities, Beydoun and others say the opposite. A separate MENA checkbox would signify that the United States acknowledges such groups outside the lens of counterterrorism and surveillance, Beydoun says. “For the community, it would be symbolic.”

Hispanic case study

Hispanics also don’t have a race checkbox, but the Census Bureau has been counting that population since 1980. That’s when every household in the United States began answering an ethnicity question alongside the race question: It asks if the person is of Hispanic, Latino or Spanish origin.

That data revealed that 14.6 million Hispanics were living in this country in 1980 — accounting for 6.4 percent of the U.S. population. By the last census in 2010, that number had risen to 50.5 million, or 16 percent of the population. As Hispanics’ numbers and visibility increased — in large part thanks to their census designation — bilingual Spanish-English schools became more popular, television programs aimed at Spanish-speaking audiences emerged and politicians realized that winning elections often required support from Hispanic voters.

With their new social and political clout, two-thirds of Hispanic adults have come to view their Hispanic origin as part of their race, according to a 2014 survey by the D.C.-based Pew Research Center. Hispanics “don’t see themselves as white, black or any of the other races that have boxes on the census race question,” Roth says.

In both the 2000 and 2010 censuses, over a third of those who selected a Hispanic ethnicity put themselves in the “some other race” category, making that ambiguous group the nation’s third largest for both decades. Absent changes to how the race and ethnicity questions are asked, growth in the country’s Hispanic population means “some other race” may become the nation’s second largest racial group in 2020, the Census Bureau noted in a 2017 report.  

For Hispanic people, at least, the ethnicity data can still reveal where such individuals might be experiencing discrimination — though confusion over the two questions could discourage people from completing the form, further increasing Hispanic undercounts, says sociologist Julie Dowling of the University of Illinois at Urbana-Champaign and chair of the Census Bureau’s National Advisory Committee on Racial, Ethnic and Other Populations.

Projections by the Urban Institute, a D.C.-based think tank, made while the current administration was considering adding a citizenship question to the census, showed that the 2020 count could miss 3.6 percent of Hispanics, or over 2.2 million people. Lingering fears over that question could still depress Hispanic participation, says the institute vice president, Rob Santos. 

Undercounts translate to lost dollars. For instance, Child Trends, a nonprofit research organization, estimates that 37 states will lose out on funding from five federal child and family programs if there is an undercount of Hispanics on the 2020 census. Texas alone stands to lose $339 million to $1.4 billion annually.

Nonetheless, money still flows into Hispanic communities thanks to the ethnicity question. But for MENA communities — estimated at 3.7 million people according to the Arab American Institute — there is no equivalent channel. In recognition of those difficulties, the Census Bureau recommended adding race checkboxes for both Hispanic and MENA groups to the 2020 census. But the OMB, which must approve such changes, did not do so. (The OMB did not reply to a request for an interview and the Census Bureau did not make anyone available for an interview in the three months it took to report this story.)

One tweak made to the 2020 census could provide some information on MENA and other minority populations hidden within the broader categories, Dowling says. For the first time, those checking “black” or “white” can opt to write in their nationality of origin. But without a checkbox, the information will be in a less accessible form. “Somebody has to comb through that data,” Dowling says.

A slippery sense of self

As minority groups fight for greater visibility, and the race question gets wound up in ideas about self-affirmation and group empowerment, the census data have been getting more difficult to decipher since the 1960 shift to self-identification. 

With the power to check their own race box, many people previously identified as white have embraced a nonwhite or mixed-race identity. That’s evident in the American Indian numbers. From 1890 to 1960, the American Indian population grew from 248,000 to 524,000, with an average annual growth rate of just 1.1 percent. But over the next several decades, and coinciding with the shift to self-identification, that population grew to almost 2 million by 1990 — with an average annual growth rate of 4.3 percent. That meteoric growth extends well beyond what is possible through births alone, Liebler says.

Working with collaborators at the Census Bureau, Liebler looked at changes to the American Indian category between the 2000 and 2010 censuses. Reporting in the April 2016 Demography, the team identified three “types” of American Indians — fewer than one-third reported American Indian on both censuses. The rest either alternated between single-race and mixed-race American Indian or added or dropped American Indian altogether from one census to another. That last category of people was less likely to report tribal affiliation or live in an American Indian area than those in the other two groups.

People with no connection to tribal groups who decide to check “American Indian” can have real-life implications. For instance, American Indians suffer the lowest high school graduation rates of any group in the country, while the education profile of those “new” American Indians resembles that of whites. When education statistics coupled with census race data make it appear as if American Indians in many regions have become more educated, it implies that American Indians no longer require as much government support for education, Liebler says.

Genetic ancestry tests may be contributing to white people adopting American Indian and other minority identities, Roth’s recent research suggests. From 2009 to 2010, Roth interviewed 100 individuals who purchased genetic tests and then re-interviewed 89 of those respondents after the 2010 census.

Reporting in the July 2018 American Journal of Sociology, Roth showed that 14 percent of participants changed their race in 2010 from that used in 2000 based on the test results. Though the study is small, the exploding popularity of such tests over the last decade (SN: 6/23/18, p. 14) means such tests could dramatically alter how people respond to the census race question.    

But extensive interviews with the participants revealed striking group differences. While just 25 percent of black respondents chose a new racial or ethnic identity, about 40 percent of white respondents did so. What’s more, white people often chose what new racial information to accept and what to discard. The ability to cherry-pick race is a uniquely white privilege, Roth says. Whites “can claim to be Native American … or whatever they didn’t know about before, and they can go to festivals and they can eat the food, but it’s costless. It doesn’t bring with it the same kind of discrimination that identity brings for somebody who is seen [by others] as a member of that group.”

A more nuanced approach

What Roth is getting at is the idea that discrimination is more about how a person is seen by others than how they see themselves. A white-presenting individual who receives a genetic ancestry test result that reveals American Indian ancestral ties could check “white” and “American Indian” on the 2020 census, but that decision says nothing about the level of discrimination that person has experienced, Liebler says. And that decision makes it harder for policy makers to identify communities in need.

What’s more, research over the last few decades has made clear that even for individuals within the same race, light-skinned individuals fare better in American society, receiving more lenient prison sentences and better job opportunities, than their dark-skinned counterparts. So even if “Hispanic,” “MENA” or other race checkboxes were added, identifying that sort of discrimination within a group will remain impossible with the current race question.

“We need more than one measure of race,” says sociologist Nancy López of the University of New Mexico in Albuquerque. One question should ask respondents to identify their own race so policy makers can continue to identify the country’s changing social and political blocs. And another question should ask respondents to gauge how they appear to an outsider, such as an enumerator standing on their doorstep, census form in hand.

But don’t expect to see changes to the race question anytime soon, Liebler says. “The push is entirely within the academic world. We’ve got no traction at all within the Census [Bureau].” Perhaps the place to tack on multiple race questions is on smaller federal surveys, such as those administered by health, education and housing agencies, Liebler says. “That seems plausible.”