The Truth About White America

The Census Bureau wanted to gather data about a changing nation, but ended up reinforcing old racial categories.

Illustration of a blurry census form fading into a black background
The Atlantic

If you paid attention to the news this summer about the release of 2020 census data, you probably heard that America’s white population is in free fall. Big, if true.

The statistic that launched a thousand hot takes and breathless voice-overs about racial change was a supposed 8.6 percent, or 19 million, drop in the number of white Americans since 2010. Headlines cast this decline as unprecedented in census history and signaled that the nation’s majority-minority future loomed even closer than previously forecast. Pundits spun it as a harbinger of policy change and partisan realignment, for better or worse. Some wisely cautioned against demography-as-destiny assumptions in a country where the definition and public understanding of race can change rapidly. But few observers questioned whether the reported differences between the 2010 and 2020 censuses reflected real demographic change or simply statistical noise.

Commentators should have read the fine print before rushing to trot out their favorite narratives. If they had, they would have discovered that the eye-popping figure at the center of this summer’s hoopla is an illusion. The apparent decline in the white population is a result of changes to the Census Bureau’s protocol for measuring and classifying racial identity. The changes aimed to more accurately gauge the expansion of the country’s mixed-race population through new and more sophisticated data collection and classification techniques that capture the nuances of Americans’ multifaceted racial and ethnic identities. But a combination of bureaucratic constraints and messaging failures paved the way to public confusion.

Ironically, a segment by the Fox News host Tucker Carlson inadvertently exposed the myth of massive white decline. During a rant about what he perceived as left-wing giddiness over the “extinction of white people,” he asked, “Where did all these people go?” The millions of missing white Americans did not, in fact, go anywhere. And they are not being replaced by minorities. Growing numbers of white Americans have multiracial children and grandchildren. Others were recategorized in 2020 as multiracial themselves, instead of single-race white.

How, then, did so many pundits and commentators come to the conclusion that the white population had dropped 8.6 percent? This calculation stems from two errors. The first is a failure to recognize that the degree, and even direction, of change in white population depends entirely on how one defines white.

Many white people who self-identify as white also identify as members of another race or as Latino. Thus, white can mean four different things: (1) non-Latino single-race white people, (2) non-Latino including multiracial people, (3) all single-race white people including Latinos, or (4) all white people including Latinos and multiracial individuals.

Under the first, and narrowest, definition, the white population did drop, but by 2.6 percent, not 8.6. Under the second, it increased by 1 percent. Under the last and broadest definition, it increased by an even greater 2.6 percent. Only under the third, rarely used, definition does one reach the precipitous 8.6 percent drop.

Which of these is most accurate? The standard definition of white in census data is the narrowest and excludes people who are both white and members of a racial or ethnic minority group. This follows the idea that everyone should fit in a single racial or ethnic category and hews to the ethno-racial specifications of the Office of Management and Budget, last updated in 1997.

But much has changed in the early 21st century. Categorically excluding Americans who have both white and minority parentage from the white classification makes little sense in a society in which plural identities and backgrounds are common and the great majority of mixed-race individuals have a white parent. As we previously wrote in The Atlantic, this practice harkens back to the racist one-drop rule and clashes with present-day sociological realities. Many Americans identify as multiracial, and they are concentrated in social and economic strata that more closely resemble those of white than minority Americans.

The second error driving the myth of white decline is an accounting snafu. Ignoring the Census Bureau’s caveats, analysts who touted the 8.6-percent-drop number conflated true demographic change with method changes in how the census form asked people to report their race and how the bureau used this information. To better assess the size and characteristics of America’s multiracial population, the bureau added a write-in space under the “white” and “black” race checkboxes, where respondents could specify their ethnicity. The instructions read: “Print, for example, German, Irish, English, Italian, Lebanese, Egyptian, etc.”

These changes appear to have produced a recategorization: Many people who identified only as white in 2010 were classed as multiracial in 2020, especially Latinos. Among Latinos, the number of people who were classified solely as white dropped from 53 percent in 2010 to 20 percent in 2020. A person of Latin American origin who self-identified as single-race white in 2010 would presumably have checked the “white” box again in 2020. But she might also have written in her own or her family’s country of origin in the new text box. This would result in her automatic reclassification as multiracial in 2020, white and “some other race,” instead of single-race white, as in 2010.

That the sharp rise in multiracial Latinos in 2020 is due to an accounting change, rather than a real demographic or social trend, is clear when we look at the 2019 American Community Survey, run annually by the Census Bureau. The ACS collected and classified race in the same way that the 2010 census had throughout the prior decade. The last of the 2019 ACS data were gathered just a few months before the census, and the reported results showed that the percentage of Latinos categorized as single-race white was unchanged since the ACS survey of 2011.

In short, the confusion over white decline occurred not because the population changed but because the nuances in many individuals’ plural identities became more visible in the 2020 census. This is a positive development. The Census Bureau is moving toward a more comprehensive scheme for understanding how mixing is changing America’s ethno-racial contours. But it’s not there yet.

Three crucial steps remain to provide a clear picture of the mixed-background population and avoid the sorts of misunderstandings that have plagued reporting and commentary on the 2020 census data.

First, the bureau should make available data on the responses that individuals gave to the census questions on race and ethnicity, including what they wrote in the text boxes on the race question and how they were classified. This would give journalists and academics the tools they need to distinguish real and illusory change in the 2020 results. It would permit researchers to assess precisely how much of the apparent shift toward multiracial identification resulted from reclassification. It would also give us a much-needed granular look at the way people specified their own origins and ancestry.

Second, the OMB needs to update its rules for ethnic- and racial-data collection so that the Census Bureau can allow individuals to report mixed Latino and non-Latino backgrounds. In the current system, these backgrounds are not recognized, and the census lacks the authorization to alter its race question so as to identify them. In 2015, the bureau tested an integrated race and ethnicity question that would overcome this flaw, and deemed its performance superior to that of the questions used in 2010 and 2020. In the integrated question, “Hispanic” is treated as a racial category. Individuals are able to check as many of the ethno-racial categories as they think appropriate. This would substantially change the picture Americans have of mixing, because we know from birth data that the largest mixed group by far has Latino and non-Latino white parentage. There is no reason to wait for the next census in 2030. An OMB rule change now would allow the Census Bureau to collect better data in its annual American Community Survey right away.

Third, the bureau should break from its long tradition of emphasizing mutually exclusive ethnic and racial categories, whereby every person appears in only one. Granted, such a scheme has the advantage of neatness—all the percentages add up to 100. But this neatness comes at a major cost because it suppresses important components of the identities and group affiliations that are socially and psychologically meaningful to Americans from mixed backgrounds.

These changes would make it possible to more accurately count America’s ethnic and racial groups and to thoroughly comprehend the societal changes that are coming about through increased mixing. For example, census data could then be used to compare and contrast the upward mobility and integration of mixed white-minority Americans and of racial minorities. We could discern more definitively which groups of mixed-race Americans face barriers to the mainstream, which encounter abundant opportunities, and why.

Long a multiracial nation, America is also becoming a nation of multiracial people. Much needs to be understood, and communicated to the public, about how this ethno-racial evolution is defining our present and shaping our future. This will require more accurate and nuanced data about mixing from the Census Bureau. It will also require the rest of us to unburden ourselves of the antiquated, zero-sum lens of white loss and minority gain so that we can see 21st-century racial change in all its kaleidoscopic dynamism.