Comparisons across survey modes reveal that support for Donald Trump is lower in phone surveys than online surveys, leading some to hypothesize that there are “Shy Trump Supporters” who withhold their true views from pollsters due to social desirability bias. List experiments offer one avenue for overcoming such biases. Using the standard direct question, a large national poll conducted by Reuters/IPSOS estimated Trump at 32.5% among all American adults. A list experiment conducted on the same respondents yields an estimate 31.5%. The difference between the question modes is not statistically significant, suggesting that Trump’s poll numbers are not artificially deflated by social desirability bias.
Link to draft here.