Election polls in the 2020 U.S. presidential election had the worst performance in decades, overestimating support for Democrat Joe Biden and underestimating support for Republican Donald Trump. Although scholars and journalists have speculated about the cause of polling errors, empirical evidence has been lacking. The American National Election Study (ANES) design offered unique leverage for diagnosing the cause of polling errors. In this paper, we evaluate the source of error in the 2020 ANES and discuss the implications for the future of survey research.

D. Sunshine Hillygus is Professor of Political Science and Public Policy at Duke University.  She serves as director of the Duke Initiative on Survey Methodology, Co-Director of the Polarization Lab, and Associate PI of the American National Election Study. She holds a PhD from Stanford University and a BA from the University of Arkansas.