Opinion polls published in the run up to the general election are mostly based on questions that ask which –if any– of the parties one intends to vote for. Such questions are seemingly simple, and lend themselves to a straightforward projection of vote shares. Yet, they also pose some problems when trying to gauge the partisan preferences of the public, let alone the ways in which the outcome may still change. The limitation of these questions is that they implicitly assume that people have a preference for just a single political party. We know, of course, that this is not true, as some people are genuinely torn between two parties. Relatively little attention is given, however, to how many such people there are, between which parties they hesitate, and how easy it is for them to switch. In this blog post, and a few that follow in the next few days, we look at such multiple party preferences.
We use here data from the British Election Study 2015 that were collected immediately before the start of the election campaign. In addition to the traditional question that asks how people intend to vote in the general election, the survey also asked for each respondent to indicate for each party separately how likely or unlikely it is that they would vote for that party. The answers on this electoral preference scale were expressed in a score ranging from 0 (“very unlikely” to vote for this party) to 10 (“very likely”). In England this question was asked for each of the following: Conservatives, Labour, Liberal Democrats, UKIP and Greens; in Scotland and Wales also for, respectively, the SNP and Plaid Cymru.
Not surprisingly, we find that people give the highest electoral preference score to the party that they intend to vote for. But quite often there is not a single most preferred party. For some 21% of voters we find that two parties are tied in terms of highest electoral preference. For 14% of voters the difference in electoral preference for the two most preferred parties is 1 (on a scale from 0 to 10), and for yet another 14% this difference is 2.
These preference scores are, of course, subject to change, just as vote intention is. All kinds of events can have an impact: debates and other media events involving politicians, promises and policy pronouncements by the various parties, new information about parties’ record in office or in opposition, and so forth. From other research, however, we know that such factors generate only small changes in the electoral attractiveness of parties. If someone has a high preference for a party, new developments may strengthen or weaken that a little bit, but rarely by more than one or two points on this scale. The same holds for a party that someone really does not like at all: campaign developments may mitigate this, but not by much. Although such short term changes are generally small, they become important for party choice when they alter which of the parties occupies the top spot in someone’s preference ordering.
In other words, when the difference in electoral attractiveness of two parties is 0 or 1, there is ample scope for party choice to switch between these two as a consequence of the unfolding campaign. When the difference is 2, switching is still quite possible, but when the difference in electoral attractiveness between the two most preferred parties is 3 or more, it is very rare that the most preferred party will be overtaken by the runner up.
All in all, at the start of the campaign about half of the electorate was unlikely to change their intended party choice, as the strength of their preference for their second best party was too much below that of their best party. The other half, however, had clearly multiple preferences, with their top-two parties of approximately equal electoral attractiveness. This has some interesting consequences. One of these, discussed above, is that it is very easy for the voters in question to switch from their intended party choice. Another consequence, not surprisingly, is that many people with multiple preferences answer “don’t know” when asked about their intended choice for May 7th. It would be incorrect, however, to think that we are totally in the dark about how these people will vote. From these multiple preferences we know which parties they will almost certainly not vote for, while it may remain a toss-up for the remaining two. For some voters this virtual tie will be broken in the course of the election campaign, for quite a few it will only be solved at the time they enter the polling station. In that respect many parties have still a lot to play, even in the last week of the campaign. Yet other consequences of multiple electoral preferences will be discussed in additional posts on this blog in the next few days.
Cees van der Eijk is a Professor of comparative politics, research methods and electoral studies at the Methods and Data Institute at the University of Nottingham. Stuart Fox is a PHD candidate who research focuses on political participation, alienation and apathy of young people in Western democracies.