a1 Stanford University
a2 University of California at Berkeley
Candidates often make ambiguous statements about the policies they intend to pursue. In theory, ambiguity affects how voters make choices and who wins elections. In practice, measurement and endogeneity problems have impeded empirical research about the consequences of ambiguity. We conducted survey experiments that overcame these obstacles by manipulating a common form of ambiguity: the imprecision of candidate positions. Our data show that, on average, ambiguity does not repel and may, in fact, attract voters. In nonpartisan settings, voters who have neutral or positive attitudes toward risk, or who feel uncertain about their own policy preferences, tend to embrace ambiguity. In partisan settings, voters respond even more positively to ambiguity; they optimistically perceive the locations of ambiguous candidates from their own party without pessimistically perceiving the locations of vague candidates from the opposition. We further find, through analysis of two additional new data sets, that candidates often take—and voters frequently perceive—ambiguous positions like the ones in our experiments. The pervasive use of ambiguity in campaigns fits with our experimental finding that ambiguity can be a winning strategy, especially in partisan elections.
We are grateful to Paul Sniderman for his advice at various stages of this project. We also appreciate helpful comments from Scott Basinger, David Brady, Henry Brady, Steve Callander, Jack Citrin, Bill Clark, Jim Fearon, Morris Fiorina, Sean Gailmard, Laurel Harbridge, Rick Hall, Simon Jackman, Cindy Kam, David Karol, Ken Kollman, Jon Krosnick, Gabriel Lenz, Phillip Lipscy, Rob MacCoun, Neil Malhotra, Norman Nie, Jonathan Rodden, Eric Schickler, Ken Schultz, Jas Sekhon, Ken Shepsle, the APSR editorial board, the anonymous reviewers, seminar participants at Stanford, the University of Michigan, Waseda University, the Institute for Governmental Studies and the Goldman School of Public Policy at University of California at Berkeley, and panel participants at the annual meetings of the American Political Science Association and the Midwest Political Science Association. We thank Mike Dennis, Joe Hadfield, and Sergei Rodkin of Knowledge Networks for their superb work in fielding the survey experiments, and Reagan Thompson for research assistance. Finally, we gratefully acknowledge financial support from the National Science Foundation (SES-0550844 and SES-0549504) and the Vice Provost for Undergraduate Education at Stanford.