The history of survey research is inseparably entwined with the development of election polling. As early as 1904, social scientists completed the first quantitative study focusing on the constituency of the Social Democratic Party in Germany. The breakthrough of the representative survey method in the year 1936 was brought about by the spectacularly successful election forecasts published by researchers George Gallup, Elmo Roper, and Archibald Crossley prior to the US presidential elections of that same year. From the start, election forecasts contributed significantly to both public acceptance of modern survey research and new methodological advances, since they offered researchers the rare opportunity to prove the efficacy of the representative survey method.
Over the course of the past few decades, the methodological approaches employed in election polling have diverged strongly. Today, election forecasts are based on face-to-face surveys, or on telephone surveys or, most recently, on online surveys. In this regard, we must fundamentally distinguish between exit polls – in other words, postelection surveys conducted among voters immediately upon leaving the polling station – and election surveys that are completed prior to an election and that, as a rule, employ the same interviewing methods as are used in representative surveys on other issues.
Almost right from the start, it was suspected that election forecasts might have an influence on voting behavior – a suspicion that often went hand in hand with the concern that election polls actually represented an illegitimate way of manipulating the electoral decision-making process. In the US, this seems likely because voting is time-lagged from region to region. When polling stations on the East Coast close and the initial results are announced, the polling stations on the West Coast are still open so that it is possible that the publication of early results or exit polls has an effect on voters who are still going to cast their vote.
Published election polls and forecasts are supposed to shape voters’ preferences and prompt tactical voting. A bandwagon effect might occur when published poll results are demonstrating that a particular party or candidate is clearly leading in the race. In that case, voters might change their vote intentions in favor of the leader. However, there is also the possibility that the polls are stimulating an underdog effect, i.e., changing voting intentions in favor of the parties or candidates that are trailing behind. Published poll results are also supposed to affect voter turnout and thus increase mobilization or abstention, depending on the political circumstances of the election.
A number of studies have examined whether publishing election surveys affect voter behavior. Based on a thorough meta-analysis of relevant empirical research, Hardmeier (in press) concludes that the impact of polls on voter turnout is rather marginal and that there is also no strong effect on voting intentions. The empirical evidence speaks for a weak bandwagon effect rather than for an underdog effect. If poll effects occur, they are more likely on undecided voters and on people with weak or no predispositions. In the light of these results, fears expressed in public debate about the manipulation potential of polls are exaggerated in most cases.
The reason why election polls have less impact on voter behavior than is often assumed is illustrated by experiments examining the persuasive influence of survey statistics as compared to illustrative single cases. As the findings show, exemplifying illustrations – e.g., remarks of passers-by interviewed on the street – have much stronger effects on people’s opinion formation than statements based on representative statistics (Brosius & Bathelt 1994; Daschmann 2000). Many people appear to have great difficulties in grasping the significance of statistical data in day-to-day life. Concrete single cases seem considerably more accessible and thus more convincing than abstract numbers.
Nonetheless, parties and candidates are frequently suspected of launching polls in order to improve their chances in the election campaign race. Similarly, mass media are sometimes falsely accused of publishing poll results that could be instrumental in furthering the political interests of their sponsors. In any event, election polls and forecasts, because of their high news value, are a preferred topic of campaign coverage and serve media interest insofar as they attract audience attention.
Although most voters are unimpressed by election polls, this does not mean that their influence on the course of an election campaign is generally negligible. Even if survey results have only weak direct effects on voters, there are still indications – albeit no certain proof – that election polls could, under certain circumstances, have a considerable indirect impact. In contrast to the majority of the population, journalists are used to dealing with survey data. Poll results may thus prompt journalists to change their definition of the political situation and ultimately alter the basic tenor of their reporting. Hence, the population’s voting behavior could be influenced by the overall tenor of media campaign reporting – i.e., not just reporting on survey findings, but on the political situation as a whole.
In most democratic countries, therefore, the question of whether it ought to be forbidden to publish the results of election polls in the final weeks prior to an election has occasionally been the subject of heated public debate. Almost half of the 78 countries covered by a worldwide survey have embargoes on the publication of poll results on or prior to election day (Spangenberg 2003). In many countries, however, calling for such bans is legally problematic – and also seems dubious from a democratic viewpoint. On the one hand, it does not contradict the principle of democracy when tactically minded voters obtain information on the strength of the parties or candidates and then make their decision in light of this information. On the other hand, survey findings represent only one of many different kinds of public statements made on the presumable outcome of an election. Forbidding their publication would mean suppressing the only scientifically based information and allowing less reliable statements, speculations, and contentions based on personal opinion and hearsay to go uncontradicted.
- Brosius, H.-B., & Bathelt, A. (1994). The utility of exemplars in persuasive communication. Communication Research, 21, 48 –78.
- Daschmann, G. (2000). Vox pop and polls: The impact of statements in the media on the perception of a climate of opinion. International Journal of Public Opinion Research, 12, 160 –181.
- Donsbach, W., & Traugott, M. W. (eds.) (in press). Handbook of public opinion research. London: Sage.
- Hardmeier, S. (in press). The effects of published polls on citizens, In W. Donsbach & M. W. Traugott (eds.), Handbook of public opinion research. London: Sage.
- Spangenberg, F. (2003). The freedom to publish opinion poll results. Report on a worldwide update. Amsterdam: ESOMAR/WAPOR. At www.unl.edu/WAPOR/Opinion%20polls%202003%20final%20version.pdf, accessed January 31, 2007.