When results of the 2020 Census were released over the summer, nearly 50 million U.S. residents had checked a box for “Some Other Race” — an increase of 129 percent from a decade ago. The nation’s population is becoming more diverse, and demographers say people are marking that box because government categories for race are increasingly out of step with the ways people identify themselves.

Part of the problem lies in the fact that based on 1997-era standards, the form asks about race and Hispanic origin in two separate questions. The race question offers choices that pinpoint some identities with granular specificity while lumping others into broad categories that span continents and skin tones.

For Hispanics, identifying themselves on the census questionnaire can be especially bewildering. Hispanic is not one of the race choices, as it is not considered a racial category, but many respondents miss that nuance.

That prompted the Census Bureau to ask for an update to the classifications in advance of the 2020 Census. But the proposal was never acted upon by the Trump administration.

In 2020, over 15 percent of all respondents marked Some Other Race, either alone or in combination with another race; the vast majority of them were Hispanic. That made Some Other Race the second-largest “alone or in combination” group in the country, edging out “Black or African American,” and leaving demographers with increasingly imprecise information about who lives in America.

“It’s very confusing,” said William Frey, a Brookings Institution senior demographer who studies diversity. “If you say you’re Hispanic, they [then] ask you to fill out a race question, and many people say, ‘I’ve already filled it out — I’m Hispanic.’ The distinction between ethnicity and race is something that most people don’t make.”

Questions about race and ethnicity are asked by agencies across the federal government and are used by policymakers, researchers, advocacy groups and others. The classification system is periodically updated by the Office of Management and Budget, and those updates determine the changes the Census Bureau may make in its surveys. Since 2000, for example, people have been able to check more than one box in the decennial survey’s race question.

Some of its race categories are broad — it lists European, Lebanese and Egyptian as examples of White, and Black examples include African American, sub-Saharan African and Jamaican. Other race categories are defined by specific nationalities or geographic locations, such as Chinese, Filipino, Japanese, Korean, Vietnamese, Samoan or Native Hawaiian. Respondents can also check American Indian or Alaska Native, or “other” Asian or Pacific Islander.

And then there is the increasingly popular Some Other Race, which census officials say was originally intended to be a “small, residual” category. In the mid-20th century, only a fraction of a percent of respondents marked it, but in 2020, 8.4 percent, or nearly 28 million people, marked Some Other Race and no other box.

One reason it is so hard to classify people by race and ethnicity is that the categories are subjective and not scientific, said Jeffrey Passel, a senior demographer at Pew Research Center. “These are social constructs. There’s nothing inherently biological about race,” he said. “What is viewed as Black at one point in time may not line up with what is viewed as Black at another point in time.”

Who is doing the viewing has also changed. In the early part of the nation’s history, census enumerators looked at respondents and decided how to identify them. Since 1960, respondents have chosen how to identify themselves.

Interracial marriage, immigration, DNA testing, and increased awareness and pride around ethnicity and race complicate those choices. “The country’s gotten more diverse … and that [has] led to it being a lot more difficult to put people in these little boxes that the census has, or for people to figure out which box they go into,” Passel said. “So the Census Bureau is tasked with collecting data in a way that people can answer the question, coding the data in ways that line up with the directives they have, and producing data that people can use.”

One way it does that is by scrutinizing and sometimes overriding Some Other Race responses, based on additional information respondents can put in a write-in section. “If the response can be classified within the OMB race category definitions, then the Some Other Race check box is removed, and the write-in response is classified with its corresponding OMB race category,” said Nicholas Jones, director and senior adviser of race and ethnic research and outreach in the Census Bureau’s population division.

For example, he said, if a respondent checks the Some Other Race box and writes in “Iranian” in the Some Other Race response area, the bureau removes the Some Other Race check box and codes, and then tabulates “Iranian” as part of the “White” racial category, per the OMB standards’ definition of White. Similarly, if someone writes “Jamaican” in the Some Other Race response area, they will be coded and tabulated as part of the “Black or African American” racial category, per OMB definitions.

When the government started taking a census in 1790, it counted the number of free White men and women, other free people, and enslaved people. As immigration patterns changed and the country diversified, race and ethnicity categories shifted, often taking on political overtones.

“It used to be that people felt that they had to be in the White category to be an American,” Frey said.

Often, they did: Until 1952, federal law restricted routine naturalization to Whites. Mexicans had been classified as White until the 1930 Census, when the bureau added Mexican as a race category, eliciting protests from Mexicans leery of being classified as something other than White at a time when the United States was restricting immigration and racial restrictions were codified in many state and local laws.

But in the postwar period, that changed. “Because of the fascism of World War II, identifying people as inferior races became a very bad thing to do,” said census historian Margo Anderson. Then the civil rights movement in the 1960s “flip[ped] all the meanings of categories to … say, ‘No, no, no, we’re going to reject that long racist history and now use all the data to facilitate inclusion.’”

Civil rights advocates debated over whether to eliminate or double down on and institutionalize race categories. “It turns out the federal government did not have a standardized classification,” Anderson said. “And given the 1930 experience with Mexicans, Hispanics and Latinos had no intention of ever allowing their community to be integrated into the race classification. … They had to call Hispanic something, and they called it ‘ethnicity.’”

The bureau has tried to keep up with the evolving ways people identify themselves, including adding the Hispanic ethnicity question starting with the 1980 Census. But to many respondents, particularly Hispanics, the survey increasingly did not reflect how they saw themselves.

“For a growing number of people for the past few decades [the classifications] were confusing and problematic … and they wanted to find a better way to express their identity,” Jones said.

An interagency task force spent several years devising a new format that collapsed race and ethnicity into one question. It also added a new category, MENA, for people of Middle Eastern and North African descent, who have long been counted as White or Black. The changes were supported by demographers and advocacy groups, and the bureau tested the new format in 2010 and 2015.

In a 2017 video presentation describing the results, Jones noted that Some Other Race had in 2000 and 2010 been the third-largest group and could become the second-largest group if no changes were made. But when a single combined question was tested, the number of people checking the Some Other Race box plummeted to half a percent. The percentage of people reporting as White went down slightly; other groups did not change.

Roughly 70 percent of people who marked themselves as Hispanic on the combined question reported themselves only as Hispanic, the researchers found. “A combined question is more in line with how Hispanic respondents view themselves,” Jones said in the video.

But those changes were never adopted. Nancy Potok, OMB’s chief statistician at the time, said that when she sent the proposal to the administrator of the agency’s Office of Information and Regulatory Affairs, “It never got signed off on.”

Every time she asked about its status, “I was rebuffed,” she recalled. “I was purposely excluded from conversations that were taking place among appointees.” (The administrator appointed by President Donald Trump, Neomi Rao, was later appointed to the U.S. Court of Appeals for the District of Columbia Circuit, filling the space left by Brett M. Kavanaugh’s Supreme Court appointment.) The OMB and Rao did not respond to queries about the proposal.

According to John Thompson, who directed the Census Bureau until 2017, “The political leadership at OMB basically didn’t want to collect accurate race and ethnicity data … so it died at OMB, and the Census Bureau had to move on, and they couldn’t implement what all had agreed was the best way.

“It was very disappointing to the Census Bureau staff,” he added. “They had spent a lot of time putting a lot of work into it. … They were used to seeing science prevail over political decision-making. To me, it had looked like politics had won over science.”

Now, with a new administration and another decennial census to plan for, demographers and census experts say they would like to see the proposed changes revisited.

“The Census Bureau’s research and recommendations made a strong case that the combined question produced ‘better data,’ in the sense of allowing people to pick a category that they felt best identified them,” Passel said.

Brookings’s Frey agreed, saying implementing the changes in 2020 would have led to “a real improvement.”

If OMB approves the policy changes this time around, the new format could be implemented earlier than 2030, appearing in smaller surveys such as the annual American Community Survey, Jones said.

For now, the bureau plans to resume the conversation with stakeholders and advisory groups, he said, adding, “We expect that this will begin to develop organically.”

Ted Mellnik contributed to this report.