**Principal investigator:** **Brendan Nyhan** Dartmouth College Email: nyhan@dartmouth.edu Homepage: http://www.brendan-nyhan.com/ **Sample size:** 1387 **Field period:** 11/10/2010-08/09/2012 **Abstract:** Previous studies have found that race of interviewer can affect the survey response, but do these effects vary based on other survey elements or cues given to respondents? We tested whether the effectiveness of corrections of the rumor that Barack Obama is a Muslim varied among white participants depending on the race of the researcher to whom the study was attributed. In a previous study conducted on Mechanical Turk, Republicans who were exposed to corrections were significantly more likely to report reduced misperceptions when the study was attributed to a nonwhite researcher, which suggests a social desirability effect created by the interaction between reviewer race and the cue provided by the corrective information. Our TESS study replicated the interaction between researcher race and corrective information, but only among less educated members of a nationally representative sample of white non-Hispanic Republicans. **Hypotheses:** An extensive literature examines how social desirability concerns can influence the way survey or experimental participants answer questions, but in some cases respondents may only give socially desirable responses to people from other backgrounds when prompted by cues suggesting which responses are (un)desirable. We examine this effect in the context of corrections of the widespread and persistent belief that Barack Obama is a Muslim. Previous studies found that reported misperceptions decreased more (especially among Republicans) when a nonwhite researcher was present, suggesting that the context in which a correction on a sensitive issue is delivered may affect how people react to it. In particular, when context makes racial or ethnic difference salient, respondents may be more likely to give socially desirable responses to messages that provide cues about what such a response might be (i.e., that Obama is not a Muslim). To assess this hypothesis, we explicitly manipulated both researcher race and exposure to corrective information, allowing us to test whether researcher race will moderate the effects of the experimental manipulations that are intended to correct misperceptions about Obama’s religion. (Note: The study we ultimately fielded differed from our initial TESS application based on the results of a Mechanical Turk study we fielded before TESS. We originally intended to test the semantic format of the corrective information but found that this manipulation made little difference in our results.) **Experimental Manipulations:** Participants were randomly assigned to one of two different conditions that varied the race of the supposed PI. In the white researcher condition, the PI was said to be Brad Kelly, Ph.D and was depicted as a white man using an image purchased from a stock photo website. In the black researcher condition, the PI was said to be Rasheed Jackson, Ph.D. and the survey included a closely matched stock photo of an African American man. (See Supporting Information for the photos and other study materials.) The names of the fictional PIs were selected from a list of stereotypically white and black first and last names. Subjects saw the name and picture of the fictional PI at the top of every survey page. Participants first saw a video clip of Senator McCain discussing his religious faith to help disguise the purpose of the study. After the McCain clip, participants were independently randomly assigned to watch an unrelated control video or one in which Obama corrected the Muslim myth. Each clip lasted about 10 seconds and was followed by on-screen text displaying a key quote from the clip. **Key Dependent Variables:** We measured perceptions of Obama's religion in two ways. Each measure was coded such that higher values indicate stronger beliefs that Obama is a Muslim. ObamaReligion records responses to a question asking "Do you happen to know what Barack Obama's religion is?" where 1=Christian, 2=DK/other, and 3=Muslim. Muslim is a seven-point scale for whether participants agree that Obama is a Muslim (where 1="Strongly disagree" and 7= "Strongly agree"). **Summary of Findings:** Unlike our previous study, the correction successfully reduced misperceptions and the effect did not appear to vary by researcher race for the sample as a whole. However, we do find a dramatically different result that mirrors our previous study among respondents who did not graduate from high school (n=111). For this subgroup, stated misperceptions decreased more in response to a correction when the study was attributed to a nonwhite researcher, suggesting a social desirability effect similar to our previous study. The fact that this effect was concentrated among a low education subgroup, though unanticipated, is consistent with the fact that less educated people tend to be both the most likely to hold misperceptions about Obama's religion and the least racially tolerant. **References** [Brendan Nyhan, Jason Reifler, Christopher Edelman, William Passo, Ashley Banks, Emma Boston, Andrew Brown, Robert Carlson, KayAnne Gummersall, Elizabeth Hawkins, Lucy McKinstry, Jonathan Mikkelson, Emily Roesing, Vikram Srinivasan, Sarah Wakeman, Lindsey Wallace, and Rose Yan. "Answering on cue? How corrective information can produce social desirability bias when racial differences are salient." Working paper.][1] [1]: http://tessexperiments.org/data/nyhan%20obama-muslim.pdf