Something unusual has been taking­­­­­­ place with the United States Census: A minor category that has existed for more than 100 years is elbowing its way forward. “Some Other Race,” a category that first entered the form as simply “Other” in 1910, was the third-largest category after “White” and “Black” in 2010, alarming officials, who are concerned that if nothing is done ahead of the 2020 census, this non-categorizable category of people could become the second-largest racial group in the United States.

Among those officials is Roberto Ramirez, the assistant division chief of the Census Bureau’s special population statistics branch. Ramirez is familiar with the complexities of filling out the census form: He checks “White” and “Some Other Race” to reflect his Hispanic ethnicity. Ramirez joins a growing share of respondents who are selecting “Some Other Race.” “People are increasingly not answering the race question. They are not identifying with the current categories, so we are trying to come up with a (better) question,” Ramirez told me. Ramirez and his colleague, Nicholas Jones, the director of race and ethnic research and outreach at the Census Bureau, have been working on fine-tuning the form to extract detailed race and ethnic reporting, and subsequently drive down the number of people selecting “Some Other Race.”

The U.S. census form has evolved over 226 years. “Race is the oldest question we have in this country,” Ramirez said. “We asked it in our first census in 1790, and we have been asking it ever since, every 10 years in a different way and different shape, but consistently throughout.” “White” has been the only consistent racial term since August 1790, when marshals knocked on doors in the original 13 states and in the districts of Kentucky, Maine, Vermont, and the Southwest territory (Tennessee) to classify people as a “Free White Males” or “Free White Females,” “Slave,” or “All Other Free Persons.” The civil-rights era was a pivotal moment for how census data was used, Jones said. “Prior to that, the measurement of race and ethnicity in the census was often used, not for helping people, but to show how people can be differentiated,” he told me. “But from the 1960s onwards, the measurement was really used to address problems and concerns.” Today, it also serves to reapportion congressional seats and Electoral College votes.

A number of factors affect census results. Take, for example, an increase in ethno-racially mixed families. Among marriages in the United States, 15 percent are between people of different racial and ethnic origins, according to Richard Alba, a sociology professor at the City University of New York’s Graduate Center. Alba’s research also found that one in seven infants are born into an ethno-racially mixed family. “This is a really new and­­­ possibly important development because these are individuals who grow up in families that involve whites and minorities. They are truly straddling the dividing lines in American society,” he said. “We don’t really know enough about them to be able to say how they will identify themselves, how they will locate themselves within American society.”

In 1977, the Office of Management and Budget, which supervises the U.S. Census Bureau, issued a directive on racial and ethnic classification for federal statistics. Ethnicity—such as “Hispanic” and “not Hispanic”—was separate and distinct from the concept of race. As a result, the “Some Other Race” category captured a lot of Hispanics. Twenty years later, the OMB issued a fresh directive, allowing respondents to report more than one race on the 2000 census form. The racial categories available were: “White,” “Black, African-American or Negro,” “American Indian or Alaska Native,” “Asian,” and “Native Hawaiian or other Pacific Islander.” The “Latino” classification was also introduced as an alternative phrasing for the “Hispanic” ethnic category. But the “Some Other Race” category, long part of the census, was not mentioned in the OMB directive. Instead, the Census Bureau decided to keep it to capture respondents who didn’t identify with any of the other categories provided.

Still, to address the country’s changing demographic landscape, the Census Bureau frequently conducts tests in between decennial censuses. In August 2015, the bureau sent a questionnaire dubbed the National Content Test to 1.2 million households. It had two objectives, according to John Thompson, the director of the U.S. Census Bureau, “to evaluate and compare different versions of questions to ask in the 2020 census, such as those about race and origin,” and to encourage households to respond to the census online, the cheapest and most efficient option.

The questionnaire emulated another experiment also aimed at eliciting detailed race and ethnic reporting known as the Race and Hispanic Origin Alternative Questionnaire Experiment. It involved half a million households and was conducted close on the heels of the 2010 census. The results showed that the bureau would be better off scrapping the separate “Hispanic” ethnicity question on the census form and putting it alongside the list of categories asking for a person’s race. In doing so, the census results were more reliable—and valuable—since a larger share of people reported their race. In fact, in the experiment sample, “Some Other Race” declined from 7 percent of the population to less than 1 percent. But focus-group participants who ran alongside the AQE raised a series of questions: What was the census form really asking? Some felt “race” and “origin” were the same. Others believed “race” was defined as skin color, ancestry, or culture, while “origin” referred to where they or their parents were born. The takeaway: The terms were confusing and needed to be defined or eliminated altogether.

The bureau’s focus-group moderators went a step further, asking questions to try to understand participants’ “situational identity,” too, recognizing that respondents discussed and reported on their race differently depending on the context in which questions were asked. They explored themes of awareness and fluidity with questions such as, “When did you first become aware of your race?” to understand if and how racial identity changed over time. Jones noted that “the categories are not an attempt to define race biologically, anthropologically, or genetically, but we know that some people interpret it that way.”

Census officials also found that people were more likely to report their race as long as they had a way to express their self-identification. “If you look at the current way we ask the race and ethnicity questions, one of the issues you will see here is that we don’t have a write-in line for ‘White’ or ‘Black,’ so many groups went down to the ‘Some Other Race’ category,” Jones told me. When space was offered for people to write in their choices, respondents seldom checked the box that said “White” or “Black” and instead wrote in “Irish” or “Jamaican” or similar. “The proportions were very different, too. It went from 3 to 5 percent of the white or black population giving the bureau detailed responses, to over 50 percent of whites and 75 percent of blacks using the write-in lines,” he said.

Tweaks and additions to the form continue today. A new category dubbed “MENA” was tested during the 2015 NCT in an attempt to allow respondents who may have Middle Eastern, North African, or Arab roots to identify themselves. In the combined question format of the experiment sample, the “MENA” category was included as the seventh race, after “Hispanics.” “What we observed in the AQE and the focus groups were that the Middle Eastern and North African population saying that they didn’t see themselves in the current categories,” Jones said. Last year, the Census Bureau met with the Arab American Institute and leading Middle Eastern and Arab American scholars, activists, and organizations to discuss including it to the form in 2020.

This and other data collected from the NCT will be instrumental in shaping the 2020 census form. After the studies are conducted and the recommendations are made, the Census Bureau is required to submit its report to Congress followed by an outline of the questions in 2018 for approval.  “The objective of the NCT was to collect detailed information on ethnic and racial groups, which is what the public has been demanding,” Jones said. “It is more than just our quest to have identities validated, to be able to say, ‘This is who I am,’ but it is also to get data back for (government) programs—information that tells us who we are.”