Can Social Desirability Be Manipulated in Online Surveys?

Social scientists have long debated the extent to which social desirability bias affects online survey responses. We argue that the most promising way to measure it is through a global manipulation at the survey's outset. If effective, this approach allows researchers to (1) isolate social desirability from confounders, (2) verify manipulation success, and (3) assess effects across multiple outcomes. Using well-powered survey experiments in the United States and Denmark, both on online access panels and a student convenience sample, we provide the most comprehensive test of this approach to date. Our findings show this method is unreliable: some treatments fail to induce perceptions of being observed, while others succeed but do not alter responses known to be affected by social desirability bias. These results challenge existing approaches and suggest researchers should prioritize specific inferential goals over a universal solution. We conclude with recommendations for alternative methods.