In preparation for the 2020 decennial census, the U.S. Census Bureau has been conducting experiments testing different ways to ask about race and ethnicity, in part to address high shares reporting “some other race” and high refusal rates for the race item, especially among Hispanics. While no final decision has been made, results from these experiments suggest that race and Hispanic origin could be combined into a single question (from its current two-question format) without lowering the population estimates of different race groups or Hispanics. In anticipation of this possible change, Pew Research Center conducted an experiment to test how to effectively implement the potential new race item on a telephone survey. This change will be essential for telephone surveys so they can collect comparable data on race/ethnicity to be used in weighting samples to national parameters established by Census Bureau surveys. Using the American Trends Panel, we separated respondents into three groups: phone respondents with response categories read aloud for all seven race and origin options in the census question, phone respondents with an abridged list of response options read aloud that more closely mirrors our current telephone survey race question, and web respondents with a question and mode comparable to what would be included on a Census Bureau survey. Across major racial and ethnic groups—Hispanics and non-Hispanic whites, blacks and Asians—the shares for the phone survey more closely matched the web version when the long list of response options was read. As the Census Bureau found in their experiments, the new question tested on the phone resulted in a much lower share reporting “some other race” (6% vs. <0.5%) and lower refusals (1.9% vs. <1%) compared with the standard two-question race format.