There is no consensus as to whether exposure to election opinion poll results influences citizens’ vote choice on election day. Some academics reject the suggestion that election polls cause a bandwagon effect, induce an underdog effect, or encourage strategic voting, while others believe these influences are not only a possibility, but a reality. This second blog post is the sccond in a series of five guest posts by University of Otago Politics student Michelle Nicol, who has recently carried out research on opinion poll influences. In this theory-based post she examines key overseas literature and the theoretical justifications for each of the three possible election poll influences. In doing so, a framework for analysis of the New Zealand case is developed. This evaluation shows that each of the three election poll influences exist: opinion polls do indeed cause a bandwagon effect, induce an underdog effect, and encourage strategic voting. Election polls influence voters in the US and UK. [Read more below]
The Bandwagon Effect
The bandwagon effect describes the situation in which a fashionable and popular political party or candidate receives extra votes purely as a result of that popularity. Discourse on the bandwagon effect generally falls into one of four groups: the first holds that election polls do not cause a bandwagon effect; the second believes the bandwagon effect exists, but does not cause any shift in votes due to being of equal strength to the underdog effect; the third holds the bandwagon effect exists and is stronger than the underdog effect; and the fourth that the bandwagon effect exists, but the underdog effect is much stronger. This section concludes that literature supports the idea that the bandwagon effect exists, and that it is stronger than the underdog effect.
Three main theoretical explanations are used to explain how election polls cause a bandwagon effect: momentum theory, cognitive dissonance theory, and reference group theory. The most straightforward of these is momentum theory. Momentum theory includes three specific elements, the first of which is an individual’s desire to conform to the majority view. The second element of momentum theory lies in individuals’ general preference for “winners over losers”. A third element is the proxy effect – the public may use poll results as a proxy for information that the poll’s respondents are presumed to have taken the trouble to acquire. In other words, poll results are treated as a crude indicator of other voters’ judgments about the candidate or party’s strengths, weaknesses, and ability.
Momentum theory is borne out by the evidence. Lang and Lang, and Henshel and Johnston take the conformity cause as given; accepting that some individuals vote against their interests purely to conform with, “the group”. Individuals’ preference to identify with “winners over losers” has also been shown to exist in the electoral context. Robinson reports that when Americans are asked who they voted for in past elections the winning candidates always receive more support than actually received on election day. Likewise, Lang and Lang describe the phenomenon in which post-election surveys regularly turn up a number of respondents claiming to have voted for the winner, yet who are not in fact registered voters. While these examples specifically describe a post-election bandwagon effect, election polls allow such “backing a winner” behaviour to occur during the course of the campaign as well.
Ansolabehere and Iyengar conducted experiments exposing participants to television newscasts of a current election campaign, identical aside from varying election poll information. Groups received either no poll information in the newscast, poll information showing the race was one-sided, or poll information showing a close race. They found mere exposure to poll results favourable to a candidate was, “sufficient to generate electoral support” for that candidate. The support was said to stem, “from the simple fact that people like a winner”.  Ansolabehere and Iyengar also support the possibility that individuals use election poll results as a proxy for information its respondents are presumed to have taken the time to acquire.
Lavrakas, Holley and Miller found that young adults who do not identify with the (US) Democratic Party and individuals in employment are most prone to show momentum theory bandwagon effects. Also likely to jump on the momentum theory bandwagon are the elderly, individuals with only peripheral interest and participation in politics, and those who make up their minds late in the campaign.
Somewhat more complex theoretical explanations for the bandwagon effect can be found in the work of Morwitz and Pluzinski, and Navazio. Morwitz and Pluzinski approach the bandwagon effect through hypotheses based on cognitive dissonance theory. The theory holds that individuals’ beliefs, attitudes and behaviours tend to exist in harmony, and if in disharmony, cognitive change to restore harmony will occur. In their study, subjects were questioned about individual preferences and expectations for the outcome of a current election. Some subjects received election poll information alongside the questions, while others did not. It was found that if an individual preferred a candidate they also expected to win, election poll results had no effect whether they confirmed or disconfirmed the expectation. If an individual preferred a candidate they did not expect to win and poll results confirmed that expectation, however, individuals would change their preference to be consistent with expectations. Put simply, individuals jumped on the bandwagon of the leading candidate when in a state of cognitive dissonance, or when preferences did not align with the predicted winner in election polls.
Navazio uses reference group theory to explain the bandwagon effect. This account differs from momentum theory in that, “it is not sheer numbers that produce an impression of majority opinion to which people conform but the character of the majority in relation to the individual’s psychological group membership”. An individual should only be affected by election polls that show a result contrary to their preferences when the poll is of a “like group of subjects”. For instance, students would be affected by a poll of other students, but not by a poll of retirees or cleaners. As such, Navazio hypothesised there would be no significant bandwagon movements between experimental and control groups, rather, bandwagon movements would only be obvious when comparing blue collar and white collar occupational groups. Context requires it be noted that at the time of the experiment, blue collar workers were seen as generally supportive of Nixon’s performance, while white collar workers were regarded as highly critical.
Questionnaires assessing President Nixon’s performance were mailed to subjects, identical aside from the addition of recent poll results for the experimental group. Poll results were highly critical of Nixon’s performance. Navazio found a bandwagon effect based on reference group theory. Clerical and white collar workers in the experimental group were more critical of Nixon than white collar workers in the control group. Blue collar workers in the experimental group were not. Exposure to poll results caused white collar workers to jump on the “anti-Nixon” bandwagon – where reference group theory indicated they should be. Group membership affects the influence of election polls.
Momentum theory, cognitive dissonance, and reference group theory literature is supportive of the existence of the bandwagon effect. Exposure to election polls causes some individuals to jump on the bandwagon and support the frontrunner. Whether these individuals are those identified in the momentum theory profile, those in a state of cognitive dissonance, or those individuals ensuring they are voting according to their reference group’s preferences (or perhaps a combination of the three), the bandwagon effect is very real. Of the seven other articles found on the bandwagon effect via internet search, a further six support these contentions. The one article questioning the existence of the bandwagon effect is Sanders’ study of British General elections between 1950 and 1997. Sanders examined the results of election polls, and compared them to the actual election result. He found only an underdog effect. This study, however, suffers from a number of flaws (outlined below) and can thus be disregarded.
Given the conclusion that the bandwagon effect exists, exactly how much influence the bandwagon effect has in numerical terms is of interest. The contested nature of the bandwagon effect, however, means there is some discomfort among academics in stating the bandwagon effect in numerical terms. Two articles covered nevertheless reject this trend, and show that the numerical influences of the bandwagon effect is heavily context dependent. In the first of these articles, Ansolabehere and Iyengar found, “the more favourable the poll information, the more significant the surge in electoral support for the candidate leading in the polls”. When subjects were given election poll information showing a close race (Clinton with a 5% lead over Bush), the bandwagon effect was responsible for a 3% move toward Clinton. When election poll information showed the race was one-sided (Clinton with a 17% lead over Bush), the bandwagon effect was responsible for a 10% movement. The second article to provide numerical indications of the strength of the bandwagon effect is by Mehrabian. Mehrabian’s study gave subjects election poll information showing a 28% gap in rating between the top candidates for nomination in Republican primaries. He found that, “6% of the variance on votes was due to conformity to peer-group preferences”. The figures are significant: a 10%, 6% or even 3% movement could realistically win an election.
It must also be noted that the bandwagon effect has the potential to grow throughout a campaign. High election poll ratings can lead to secondary bandwagon effects such as, “an outpouring of media coverage, endorsements, and campaign contributions”. This enables the recipient candidate or party to run a more effective campaign and improve their position in the next election poll, thereby enhancing the bandwagon effect. The bandwagon effect plays a very important role in elections.
The Underdog Effect
Like the bandwagon effect, the existence of the underdog effect is highly contested. Some academics question its existence, others say it is the most influential effect of election polls, while yet others agree an underdog effect exists, but say it is of substantially less influence than the bandwagon effect. This section summarises pertinent studies from each of the above categories. Through evaluation and comparison of study methods, it is concluded that the underdog effect exists but that it is of less influence than the bandwagon effect.
McAllister and Studlar found no evidence of an underdog effect in their examination of British elections between 1979 and 1987. The pair used exit poll data which collected information on whether subjects had heard the results of election polls, what the results of those polls were, and whether their vote had been influenced by this knowledge. McAllister and Studlar justified their reliance on self-reporting of poll influences by saying rooting for the underdog was a conscious decision.
The reliance on self-reporting, however, compromises the results of their study. Marsh points out that “the evidence of people’s beliefs about causes of their behaviour is [not] very convincing evidence for poll effects.” Beliefs about causes are far more likely to be incorrectly reported than beliefs themselves, for reasons of embarrassment, lack of understanding, or the tailoring of answers to fit what the subject perceives the surveyor wants to hear. McAllister and Studlar’s rejection of the underdog effect is therefore not entirely convincing.
Sanders, in contrast, found the underdog effect to be of greater strength than the bandwagon effect. In an examination of British elections from 1950 to 1997, Sanders found the general trend was for the “loser” of election polls to gain support, and the “winner” to lose support. He found this underdog effect by comparing the final results of general elections with results of election polls for the two main parties taken four months prior to the election.
As mentioned above, however, there are problems in using this study to support the predominance of the underdog effect. First, the experiment did not take into account the real “underdogs” of the British system, the Liberal Party. Second, a much more significant trend was found – that of the, “tendency for British voters to switch away from Labour in the last few months of each parliament” – which has happened with only one exception since 1964. As such, while Sanders’ study may be treated as evidence of the existence of the underdog effect, it would be wrong to say it categorically supports the proposition that the underdog effect is more influential than the bandwagon effect.
The remaining studies fall somewhere between McAllister and Studlar’s and Sanders’ positions. Mehrabian found an underdog effect of less influence than the bandwagon effect in his 1998 experiments based around a Republican primary election. He gave subjects bogus yet plausible election poll data showing one or the other Republican candidate leading the race, and shortly afterwards asked subjects who they would vote for.
Fleitas, while coming to the same conclusion as Mehrabian, added a fascinating dynamic to experimental underdog research. He was interested in whether subjects were influenced by the quantitative stimulus of election poll results, the qualitative information that accompanies election poll results (for example campaign events and personal circumstances), or both. Fleitas utilised a before-after experiment, consisting of the delivery of four pieces of information each followed by a poll. The information consisted of election poll results on two occasions, while subjects were provided with qualitative information on the other two occasions. The most salient piece of qualitative information was received last, and informed some subjects that the frontrunner in the polls came from a wealthy and prestigious family and had the support of high ranking members of the community. The underdog, in contrast, lacked financial resources and had been forced to conduct a door-to-door campaign. A statistically significant underdog movement occurred in the group that heard this information – the only statistically significant finding of the experiment. Mere poll information was said to have no effect – the underdog effect occurred only as subjects responded to the qualitative information.
The highly emotive nature of Fleitas’ qualitative information seems to have been designed heavily in favour of invoking an underdog effect. The impact of the qualitative information was also likely compounded by the artificial nature of the situation – fake candidates in a fake election – as voting for the underdog lacked the consequences present in a real election. Despite this, the importance of qualitative information in invoking an underdog effect should not be dismissed. In real elections, or experiments based on real elections such as Mehrabian’s, election poll information is not provided in a vacuum. Election campaigns are widely covered by the media and election poll results are regularly accompanied by commentary. Citizens use accompanying information to interpret election polls – perhaps making their own judgements as to why the “loser” is behind. This judgement may then determine whether that individual succumbs to the underdog effect.
As with the bandwagon effect, a profile of individuals most likely to be influenced by the underdog effect has been developed. Lavrakas, Holley and Miller concluded that women, liberals, those not affiliated with a political party, the unemployed, and those with lower incomes are more likely than others to succumb to an underdog effect. Likewise, those who make up their minds late in the campaign are more likely to exhibit an underdog effect than early deciders,  as are those with only peripheral levels of political interest and participation.
It can be concluded that the underdog effect does exist, and that it is of less influence than the bandwagon effect. The general spread of articles in this subject area would tend to agree; of the remaining four articles found via internet search, all support this proposition. Given this conclusion, exactly how much influence the underdog effect has in numerical terms is of interest. The contested nature of the underdog effect again means there is some discomfort among academics in making definitive statements on the percentage movements actually attributable to the underdog effect. Fleitas, however, records finding a 15.8% movement towards the underdog after the provision of qualitative information. In the absence of Fleitas’ artificial election and emotive language, however, the figure would likely be much lower. Mehrabian’s finding of the underdog effect being responsible for somewhat less than a 6% movement in votes is likely more realistic.
Unlike the bandwagon and underdog effects, the proposition that election polls encourage strategic voting is generally agreed upon. Strategic voting involves the interaction of an individual’s preferences and expectations. Votes are cast not only according to an individual’s partisan preferences, but also the individual’s expectations of how others will vote.
Any examination of strategic voting as an election poll influence must begin by showing that individuals form expectations on the basis of election poll results. Irwin and van Holsteyn did so in their study of the 1994 Dutch parliamentary election. Subjects completed a questionnaire designed to determine the main cause of expectations of election outcomes: election polls, past performance or wishful thinking. It was found conclusively that, “attention to polls... had the strongest influence on... [expectations of] Dutch voters in 1994.” Thus, election polls are central to forming expectations about election outcomes.
The type of expectations that election polls cause individuals to form, however, is in need of elaboration. Three main theories of strategic voting exist: third party squeeze, the wasted vote phenomenon, and coalition insurance strategy. Generally these theories exist in distinct electoral systems. There is, however, some crossover.
Third party squeeze, “refers to the flow of third party support to one of the other two parties when the race is close between those two but not the third”. Individuals may prefer the party shown by election polls to be coming third, but detest one of the two parties that are leading the polls. As such, the individual will vote for the other leading party. Cain confirmed the existence of third party squeeze in the British three-party system, where Liberal success depends upon a one-sided race between Labour and the Conservatives. Using survey data taken prior to the election, he found that subjects were, “more likely to vote for their second preference when they perceive that their first choice has little chance of winning”. This effect was enhanced when the individual’s next best party was within a 5% margin of another, less preferred party.
The wasted vote phenomenon is similar, in that individuals have a desire to avoid casting a vote that has no influence on the election result. A vote is wasted when cast for a party or candidate who, on election day, does not satisfy the minimum requirements to gain representation in parliament. Election poll results help citizens to avoid wasting their vote by giving insights into the likelihood of party or candidate success.
The third theoretical explanation for strategic voting lies in coalition insurance strategy. As the name suggests, this theory applies in electoral systems that generally require the formation of coalitions to govern. Unlike the third party squeeze or wasted vote phenomenon, coalition insurance strategy explains strategic voting on the part of both major and minor party supporters. For example, minor party supporters will often recognise that “their” candidate has little hope of winning the electorate vote. As such, they vote for the representative of their most preferred coalition partner, while giving their list or party vote to their most preferred minor party. Major party supporters, on the other hand, may realise that their most preferred junior coalition partner is in danger of falling below the national threshold for representation imposed by the electoral system. They then give their candidacy vote to the candidate of their most preferred major party, but their list or party vote to their most preferred junior coalition partner. Examining German National Election Study data, Gschwend found that not only did the theory apply, but that it resulted in clearly observable patterns in those favoured and deserted by strategic voters. Candidates of major parties are beneficiaries of strategic voting, while minor party candidates are deserted. On the other hand, minor parties benefit hugely from strategic list or party votes at the expense of major parties.
As with the bandwagon and underdog effects, a profile of the strategic voter has been constructed. Individuals with weak partisan ties and political sophisticates are those most likely to vote strategically. Gschwend and Meffert also found that strategic voters accounted for a full 5% of their experimental subject population, and anticipated that 15% of citizens vote strategically in actual elections.
There is general agreement over the existence of strategic voting, and the use of election polls to do so. All of the articles found via internet search support this position. The only question is over the degree of crossover between third party squeeze, the wasted vote phenomenon and coalition insurance strategy in different electoral systems.
Election polls have an influence on citizens. This analysis of literature primarily concerned with the US and UK situation shows election polls cause a bandwagon effect, induce an underdog effect, and encourage strategic voting. Exposure to election poll results influences citizens’ vote choice on election day.
Of course, these conclusions cannot be directly applied to the New Zealand situation. The differences between US and UK electoral systems on one hand, and the New Zealand electoral system of the other are simply too great. While both the US and UK electoral systems are based upon plurality, New Zealand voted to depart from such a system in 1993. New Zealand elections now take place under Mixed Member Proportional (MMP), encompassing both plurality and proportionality.
These differences in electoral systems translate into altered political landscapes. Three key areas require mention. First, while the USA and the UK are two and three party systems respectively, New Zealand has developed a multiparty system under MMP. Not only do more than two parties exist, eight are currently represented in our parliament. Secondly, plurality systems have no equivalent to the representation threshold under MMP; candidates either win constituency seats and thus representation, or lose constituency seats and gain no representation. In contrast, MMP allows candidates who do not win electorate seats to still be represented in parliament by virtue of the party vote. Third, while coalition governments are not unheard of in the UK, government without coalition has been impossible in New Zealand since MMP was introduced.
These diversities in electoral system and political landscape provide citizens with altered electoral incentives, resulting in different voting behaviours and outcomes. While the conclusions of this blog post therefore cannot be directly applied to the New Zealand situation, a detailed framework for such analysis now exists.
 M Henn, Opinion Polls and Volatile Electorates: Problems and Issues in Polling European Societies (Aldershot: Ashgate, 1998), 18; Henshel and Johnston, “The Emergence of Bandwagon Effects,” 497.
 Ansolabehere and Iyengar, “Of Horseshoes and Horse Races,” 414-15.
 K Lang and G Lang, “The Impact of Polls on Public Opinion,” Annals of the American Academy of Political and Social Science 472 (1984): 129-142, 133; Henshel and Johnston, “The Emergence of Bandwagon Effects,” 497.
 R Robinson, Mobocracy: How the media’s obsession with polling twists the news, alters elections and undermines democracy (California: Forum, 2002), 124.
 Lang and Lang, “The Impact of Polls on Public Opinion,” 134.
 Ansolabehere and Iyengar, “Of Horseshoes and Horse Races,” 415-6.
 Ibid., 427.
 Ibid., 414-5.
 P Lavrakas, J Holley and P Miller, “Public reactions to polling news during the 1988 presidential election campaign,” in Polling and Presidential Election Coverage, ed. P Lavrakas and J Holley (California: Sage, 1991).
 D Fleitas, “Bandwagon and Underdog Effects in Minimal-Information Elections” The American Political Science Review 65, no. 2 (1971): 434-438, 438.
 Lavrakas, Holley and Miller, “Public reactions to polling news during the 1988 presidential election campaign”.
 V Morwitz and C Pluzinski, “Do Polls Reflect Opinions or Do Opinions Reflect Polls? The Impact of Political Polling on Voters’ Expectations, Preferences, and Behaviour,” The Journal of Consumer Research 23, no.2 (1996): 53-67.
 R Navazio, “An Experimental Approach to Bandwagon Research,” The Public Opinion Quarterly 41, no.2 (1977): 217-225.
 Morwitz and Pluzinski, “Do Polls Reflect Opinions or Do Opinions Reflect Polls?,” 54.
 Ibid., 63.
 Ibid., 61.
 Navazio, “An Experimental Approach to Bandwagon Research,” 219.
 Ibid., 220.
 Ibid., 217.
 Ibid., 222.
 D Sanders, “Pre-election polling in Britain, 1950-1997,” Electoral Studies 22 (2003): 1-20.
 Ansolabehere and Iyengar, “Of Horseshoes and Horse Races,” 425.
 A Mehrabian, “Effects of Poll Reports on Voter Preferences,” Journal of Applied Social Psychology 28, no. 23 (1998): 2119-2130, 2122-23.
 Ansolabehere and Iyengar, “Of Horseshoes and Horse Races,” 415. See also: Henshel and Johnston, “The Emergence of Bandwagon Effects: A Theory.”
 Mehrabian, “Effects of Poll Reports on Voter Preferences.”
 I McAllister and D Studlar, “Bandwagon, Underdog, or Projection? Opinion Polls and Electoral Choice in Britain, 1979-1987,” The Journal of Politics 53, no.3 (1991): 720-741.
 Ibid., 723.
 McAllister and Studlar, “Bandwagon, Underdog, or Projection?,” 720.
 C Marsh, “Back on the Bandwagon: The Effect of Opinion Polls on Public Opinion,” British Journal of Political Science 15, no.1 (1985): 51-74, 53.
 Sanders, “Pre-election polling in Britain.”
 Ibid., 9-10.
 Ibid., 8-9.
 Ibid., 19.
 Mehrabian, “Effects of Poll Reports on Voter Preferences,” 2122.
 Ibid., 2121-2122.
 Fleitas, “Bandwagon and Underdog Effects in Minimal-Information Elections,” 435-6.
 Ibid., 436.
 Lavrakas, Holley and Miller, “Public reactions to polling news during the 1988 presidential election campaign.”
 Fleitas, “Bandwagon and Underdog Effects in Minimal-Information Elections,” 438.
 Ibid., 436.
 Mehrabian, “Effects of Poll Reports on Voter Preferences,” 2123.
 Note: The focus of this section is strategic voting, rather than split ticket voting. Previous studies into split ticket voting in New Zealand have confirmed that the majority of split tickets cast are strategic votes. See: J Karp, J Vowles, and T Donovan, “Strategic voting, party activity and candidate effects: testing explanations for split voting in New Zealand’s new mixed system,” Electoral Studies 21 (2002): 1-22, 1; H Pearse, “Making the most of a two vote ballot: Voter adaptation to a more complex electoral system” (paper presented to the Annual Meeting of the Canadian Political Science Association, Ontario, June 2. 2005), 2.
 G Irwin and J van Holsteyn, “According to the Polls: The Influence of Opinion Polls on Expectations,” The Public Opinion Quarterly 66, no.1 (2002): 92-104, 92.
 Ibid., 93.
 Ibid., 101.
 Cain, “Strategic Voting in Britain.”
 T Gschwend, “Ticket-splitting and strategic voting under mixed electoral rules: Evidence from Germany,” European Journal of Political Research 46 (2007): 1-23.
 Cain, “Strategic Voting in Britain,” 639.
 Ibid., 639-40.
 Ibid., 639.
 Gschwend, “Ticket-splitting and strategic voting under mixed electoral rules,” 4-5.
 Ibid., 4.
 Ibid., 5.
 Ibid., 19.
 Ibid., 17.
 M Meffert and T Gschwend, “Polls, coalition signals and strategic voting: An experimental investigation of perceptions and effects,” European Journal of Political Research 50, no.5 (2011): 636-667, 660.
 Ibid., 639.