Published:

17 comments

Quentin Langley studied politics at Plymouth under Colin Rallings and Michael Thrasher. He now teaches public relations at London Metropolitan University.

We need a new way of projecting opinion poll results into election results. In the past there have been two: the uniform swing hypothesis (USH) and proportional loss hypothesis (PLH). I have written about my preference for the latter in the past. Today, I suspect, that both these methods are wholly out of date.
Both are based on the proposition that, in one way or another, most constituencies will behave in similar ways. Despite the occasional importance of local factors, such projections have often proved broadly accurate in the past. The USH supposes that, if 10% of the electorate have moved from the LibDems to Labour then in each constituency the Lib Dem support will be reduced by 10% of the electorate and Labour’s support will be increased by the same amount. The PLH supposes that if Lib Dem support is down by half, then in each seat the party’s vote will fall by half, and Labour’s vote will increase by the same amount. But what if different constituencies behave differently?

Let us look at current polls, which suggest that LibDem support  is running around half the 22% of 2010. But we need to ask who has switched from the LibDems since 2010. There are many different types of LibDem voters.  The most obvious ones to have swung away are those who would have preferred the party to have done a deal with Labour. These might be thought of as falling into two categories: people broadly of the left who preferred the LibDems to Labour for any one of several reasons, including tactical factors; and partisan Labour Party supporters who chose to vote LibDem on tactical grounds. It is the second group I want to examine.

Let us begin by playing a thought experiment. What would happen if the only voters to change affiliation at the next election are Labour supporters who tactically voted LibDem in 2010? Obviously, this is ridiculous. Many people may change their votes for many reasons and in all sorts of directions, but let us still play this game.


If tactical votes for the LibDems unravel this will show as being very significant in opinion polls. We don’t know – at least not without some tremendously detailed polling – how big this group is, but there are good reasons for supposing it is significant. For one thing, there is a clear tendency for the support for parties to fall once it is clear they are in third place, even in the face of the party doing well overall. Right across the south of England there are many seats where Labour is clearly in third place.

So, let us suppose that some 7% of the electorate – about a third of LibDem voters – fall into this category, and let us stick by our supposition that these are the only voters who change their vote in 2015. What does that mean? Well, the Lib Dem vote is well down, and the Labour vote has surged. In fact Labour moves up to 36%, tied for first place with the Conservatives. But Labour does not win a single extra seat. By definition, all of these tactical votes were in seats where Labour was already in third place, prior to the tactical defections. The unravelling of the tactical defections will leave Labour in third place, albeit a slightly stronger one than before. In the majority of these seats, the Conservatives are in first place. Since our thought experiment posits no change in Conservative support, this remains the case. In all these seats the LibDems are second, and the party’s vote falls, leaving the sitting Conservative members with bigger majorities.

In some of these seats, however, the LibDem candidates were elected in 2010. In those seats, a fall in the LibDem vote could well deliver the seat to the Conservatives. In our thought experiment, therefore, Labour gains no seats at all, despite a significantly higher national vote. The Conservatives – with a stable national vote – gain some seats from the LibDems.

Even though there are a great many urban seats where the real battle is between Labour and the LibDems, these are not the seats where there is a group of Labour partisans supporting LibDems.  Indeed, insofar as the existence of the Coalition will affect these seats, it might lead to an increase in the number of Conservatives casting tactical votes for LibDems. The paradoxical result could be that, although the Labour Party makes major gains in votes from the LibDems, the LibDems could win seats off Labour in those areas where the two are most closely embattled with each other. There are far fewer Lab-LibDem battles than there are Con-LibDem battles. A major unravelling of tactical votes for the LibDems by Labour supporters combined with a small growth of tactical voting for the LibDems by Conservatives would still mean Labour having a greatly increased national vote, but still winning fewer seats.

We do need to put this in context. Many people will change their vote at the next election, and not all of the growth in Labour’s aggregate polling position comes from tactical voters returning home. Indeed, people currently telling pollsters that they will vote Labour, may well end up casting the same tactical vote for the LibDems, albeit with much less enthusiasm than before. After all, those people voted LibDem because they preferred that party to the Conservatives, and may still do so. This might particularly apply in seats the LibDems hold. Labour supporters may well want to punish the LibDems, but perhaps not strongly enough to hand a seat to the Conservatives.

So, not all of these tactical votes will, in the end, unravel. And not all of the extra support for Labour in current polls represents the unravelling of tactical votes. But much of it probably does, and insofar as that is the case it will deliver extra votes for Labour exclusively in seats the party cannot win.

So we need better polls, and much more sophisticated ways of projecting from those polls. We need to find out not just if someone has switched support, but why they have done so. And we need to find out to what extent swings are different in different types of seat. Without this, any national projections will be bunk and we will literally need to poll each constituency individually.

17 comments for: Quentin Langley: National opinion polls can’t be trusted to predict election results

Leave a Reply

You must be logged in to post a comment.