Last March, the U.S. Office of Management and Budget approved changes to the ethnic and racial self-identification questions used by all federal agencies, including the U.S. Census Bureau. The changes include merging the separate race and Hispanic ethnicity questions into a single combined question, along with adding a Middle Eastern and North African category.
New research, published in Sociological Science, co-authored by University of Chicago scholar René D. Flores, showed that the combined question reduces the percentage of Americans identifying as white or as “some other race,” a catchall category meant to account for those who self-identify as something other than the listed options.
The researchers, who also include UC Irvine sociologist Edward Telles and NORC research methodologist Ilana Ventura, found the key mechanism driving these effects was a tendency among Hispanics to decrease their identification in other categories when a Hispanic category is available in the combined question format. This leads to significant decreases in key minority populations, such as Afro-Latinos and indigenous Latinos.
“We need to think about the tools that we use to measure race and ethnicity more carefully because we're capturing subjectivities,” Flores said, “and our instruments may be affecting those subjectivities.”
Race and ethnicity
Since 1980, government documents have included separate questions for Hispanic ethnicity and racial self-identification. The idea was to allow respondents to identify ethnically as Hispanic and then choose a race in the separate race question—but many self-identified Hispanics did not know how to respond to the race question.
“Most identify as white, but a growing number identified as [some other race],” said Flores, an associate professor in the Department of Sociology. “They would just write their national origin. They would say, ‘I'm Mexican,’ or ‘I’m Argentinian’ and the like—so this was a data quality issue for the Census.”
The solution to the problem isn’t quite as simple as updating the questions. Survey methodologists have previously found that changes in the format and phrasing of identity questions can alter individuals’ responses. Because these changes affect how all federal data on race and ethnicity are collected, altered responses can have major effects on political redistricting, congressional representation, resource allocation, academic research and consumer marketing. Understanding this, the government called for “additional research, testing, (and) stakeholder engagement” to assess the impact of the approved changes on self-identification practices.
Among the scholars exploring the issue was the research team led by Flores. He and his co-authors focused on two questions: How do the proposed modifications affect Americans’ ethnoracial self-identification choices? And what are the likely mechanisms driving these effects?
“If you look across the last few census waves, more and more Americans are choosing [some other race], as opposed to any of the other available racial categories,” said Flores. “From the point of view of the census, this is a problem. [Some other race] was never meant to be an actual racial category; it was meant to be just a residual category to catch the very few people that would not select the other categories.”
New patterns
In the study, the team analyzed a nationally representative sample of 7,350 adult Americans to evaluate how the question changes impact responses. Participants were randomly assigned to answer either the existing separate race and ethnicity questions, or the newly proposed combined question.
They found that the combined question decreases the percentage of Americans identifying as white and as some other race. One of the driving factors for this effect was that, when given the combined self-identification question that lists Hispanic separately from categories such as white, African American, or American Indian, some Hispanics who previously identified with these other categories chose to identify solely as Hispanic. This reaction in turn led to statistically significant decreases in key minority populations, including Afro-Latinos and Indigenous Latinos.
Flores and his coauthors theorize that this effect could be caused by a general tendency to select only one option in multiple-choice questions. Hispanics may also feel a stronger connection to national origin labels (such as Mexican or Cuban), which are listed under “Hispanic,” rather than to ethnoracial labels like Black, which may carry more stigma.
“Anytime you introduce a big change to your ethnoracial identification questions, this raises concerns about comparability,” Flores said. “Significant changes, like those we document, make data over time less comparable as different people begin identifying with these labels.”
Flores believes there's been a push to roll out these survey question changes, but his team’s study shows that the changes are likely to affect identity data in ways that are still not fully understood.
“Our research is really calling attention to the fact that these changes are pretty impactful,” he said. “We need to take a second and understand what's driving these changes before we fully roll them out.”
—Prof. Jacqueline Stewart
