• Authors: Paul Starr and Christina Pao (Princeton University)
• Title: The Multiracial Complication: The 2020 Census and the Fictitious Multiracial Boom
• Journal: Sociological Science

The 2020 census reported a surprising uptick in the number of people who are multiracial in America: The percentage jumped from 3.4 percent the year before to 10.2 percent. That boom was just a “statistical illusion,” Paul Starr and Christina Pao show in this study.

The U.S. Census Bureau “confounded ancestry with identity and mistakenly equated national origin with race,” Starr and Pao write, explaining that the Bureau used an algorithm to reclassify people as multiracial if they marked a single race but listed a country of origin that was not coded as belonging to that race.

Under this system, a person who self-declared as “white” and wrote Argentina as a country of origin was marked as multiracial, as was someone who identified as “Black” and listed the Dominican Republic as their origin.

“Identity and origins are not the same; people may write in an ‘origin,’ perhaps a distant ancestry, even a minute one according to a DNA test, which is not part of their identity,” they state. “Multiple origins do not necessarily multiply racial identities or indicate that people think of themselves, much less are regarded by others, as mixed race.”

Starr and Pao provide evidence that the increase in multiracial numbers is an artifact of changed methods. They compare the 2019 and 2021 American Community Surveys (ACS), the annual surveys conducted by the Census Bureau in between the decennial census. They did not use the 2020 ACS because of the data collection challenges presented by the COVID-19 pandemic.

Between 2019 and 2021, they note, the “white-alone” population among Hispanics was cut by two-thirds, and the “white and some other race” population jumped by a factor of eight. “This shift is highly consequential,” they write. “Much of the ‘shrinking’ white population seen in headlines has come from the algorithmic reclassification of the ‘white-alone’ Hispanic population as multiracial.”

They note a similar shift among people of African origin, as the algorithm conflated African origin with blackness. “If Elon Musk filled out the 2002 census and entered ‘South Africa’ (where he was born) as an origin, he would have been recorded as white and Black and therefore multiracial,” they state.

Finally, the 2021 ACS shows “white and Native” people surpassing “Native alone” identification. The algorithm re-coded people who wrote a Native tribal origin under white or Black as multiracial, “even if that does not align with the respondent’s stated racial self-identification,” Starr and Pao write.

The authors list several challenges that come with their findings. First, many social scientists are unaware that methodological changes were made by the Census Bureau, which could result in inflated estimates of subpopulations and reduce the comparability of data across years.

The enlarged multiracial count also raises the question of how to interpret a category with such a radical mix of races. “A category that includes both Asian-whites and multiple-race Blacks is not a coherent category for most analytical purposes,” they state.

Starr and Pao suggest that the change in census classifications will pose problems for the use of census data for civil rights enforcement. Since 2000, the federal government has deemed people who identify as part white, part minority to fall in the minority race population for legal purposes, such as legislative redistricting and enforcement of antidiscrimination laws. But for Musk and others now labeled as multiracial, this “double switch” — classifying them first as multiracial and then as minority — risks creating legally troublesome questions.”

The authors question the Census Bureau’s move away from its long-standing commitment to racial self-identification. When the Bureau adopted the mail-in census in 1970, it ended the practice of racial assignments by census takers, instead allowing people to self-classify their race “according to the race with which they most closely identify.” The authors note that the most recent change in 2020 “reverted to [the] earlier practice of making racial assignments, though this time using an algorithm.”

“The census should return to its tradition of respecting racial and ethnic self-identification, and it should keep that commitment in mind as it addresses the new challenges of a government-wide change in race-ethnic data,” they write.