Audit studies run on names. But a name whispers more than race — it hints at citizenship, schooling, and money. This dataset measures the whispers.
Since Bertrand and Mullainathan's landmark 2004 experiment, the workhorse of discrimination research has been the racialized name: send identical applications, change only the name, and measure who gets the call. White-signaling names received 50% more callbacks. Thousands of studies now rest on this design — and on its central assumption.
identical résumés · only the name differs · bertrand & mullainathan, aer 2004
The design assumes a name moves perceptions of race and nothing else. Does it? A name like the one below may signal Asian — and simultaneously “highly educated,” “high income,” or “probably not a U.S. citizen.” If those signals travel together, studies can mistake citizenship or class discrimination for racial discrimination.
one treatment, four signals — only the first one is supposed to be there
Three surveys collected 44,170 evaluations from 4,026 U.S. respondents (n = 1,004 / 1,989 / 1,033) for 600 names, scoring each on perceived race, citizenship, education, and income — the largest validated set of name perceptions ever assembled, free to any researcher designing a name-based experiment.
Respondents identified the race behind in-group names far better than out-group names. And the asymmetry has a direction: non-white respondents read white names better than white respondents read non-white ones — in a racial hierarchy, the oppressed study the dominant more closely than the reverse.
“the souls of black folk” anticipated the asymmetry in 1903: minorities must learn the dominant group; the reverse is optional.
With perceptions quantified, researchers can hold the extra signals constant — two Black-signaling names matched on perceived education and income, or names that vary citizenship within a race — and finally separate racial discrimination from the discriminations that travel with it.
Crabtree, Charles, Jae Yeon Kim, S. Michael Gaddis, John B. Holbein, Cameron Guage, and William M. Marx. “Validated Names for Experimental Studies on Race and Ethnicity.” Scientific Data 10:130, 2023.
read the paper →figures from the paper and its policy brief: 600 names (100 white · 300 asian · 100 black · 100 hispanic) · 44,170 evaluations · 4,026 respondents across three surveys · recognition rates 85/71/75/68% · the +50% callback figure is bertrand & mullainathan (2004)