Bandwagon and Underdog Effects

Background

Bandwagon and underdog effects were a concern of the earliest critics of public polls, and the founders of polling had to defend themselves against such effects from the start. The use of straw polls was common by the 1920s, and by 1935 a member of Congress had introduced an unsuccessful piece of legislation to limit them by constraining the use of the mails for surveys. A second piece of legislation was introduced in the U.S. Senate after the 1936 election, following on the heels of an editorial in The New York Times that raised concerns about bandwagon effects among the public as well as among legislators who saw poll results on new issues (even while the Times acknowledged such effects could not have been present in the 1936 election). A subsequent letter to the editor decried an “underdog” effect instead, and the debate was off and running.

In 1937, a scholarly article by Claude E. Robinson presented a defense of the polls that focused on two claims that he disputed empirically. One claim was that the release of the polling data depressed turnout; Robinson argued that turnout had steadily increased from 1924, when the straw polls came to prominence, until the 1936 election. And the second claim concerned the bandwagon effect. Robinson argued that it was too soon to judge that such an effect occurs, because the data did not show any clear demonstration of it; among the multiple instances he cited was the fact that in 1936 Republican candidate Alf Landon's support actually dropped after the release of the 1936 Literary Digest results showing Landon in the lead.

George Gallup and S. F. Rae, in 1940, addressed the issue just before the next presidential election, again citing empirical data from multiple states and discussing reactions to presidential candidates and issues in national surveys. They concluded that there were no demonstrable effects while holding out the possibility that additional research might produce evidence in the future. Their approach is interesting in that it discusses alternative research designs that could shed light on the phenomenon. One was the possibility of panel designs for surveys, and the other was the use of experiments, although they warned against using college students as subjects and of issues of external validity associated with unrealistic settings or issues to be evaluated.

The concepts themselves require some definition and specification in order to understand why research on their existence was limited and inconclusive for such a long time, allowing the public pollsters to defend themselves so well. Even when research designs became more refined, the magnitude of effects that could be demonstrated appeared to be relatively small, not enough to affect most elections but with the potential for an impact on close ones. In one sense, both bandwagon and underdog effects reflect a simple stimulus-response model. A potential voter has an initial predisposition, either toward a candidate or to abstain. After exposure to polling information disseminated through the media (newspapers and radio in the 1930s and all kinds of media now), the individual's preference shifts toward one or another candidate, based upon whether the candidate is leading or trailing in the polls. So the first implication of assessing such effects with a survey design is that there should be measurements of preferences over [Page 50]time, preferably with a panel design as suggested by Gallup and Rae. But such panel designs have rarely been present in survey research on underdog and bandwagon effects.

...