Quentin Langley read Politics under Colin Rallings and Michael Thrasher
at the University of Plymouth and now teaches Public Relations and Political
Communications at Cardiff University.
There is a widespread belief that the Conservatives must achieve an enormous swing at the next election to win a majority – much bigger than at any election other than 1997. It has been widely reported in the media that the Conservatives would need to be 10 points ahead of Labour to win a majority, whereas Labour won a comfortable majority in 2005 with a three point lead.
Such reports are taken seriously, as though there was some precise relationship between the aggregate vote a party wins and the number of seats it wins: plug in the overall numbers and you immediately get the results for most constituencies, all except a handful that are too close to call. This is all myth.
The relationship between aggregate votes and seats won under the first past the post system is tenuous at best, and degrades rapidly as more parties enter the equation.
Cast your minds back to 1987 – if, that is, you are old enough. The BBC predicted – oh, sorry, projected – that the Conservatives would win with a majority of 26 when it turned out to be 102. This was based, recall, not on poor opinion poll data, but on actual early election results. They got the votes right, but the seats badly wrong. The margin of 11 points delivered a very healthy Parliamentary majority, though much smaller than the margin of 12 points earned for Tony Blair a decade later.
The media explanation for the difference between the ’87 and ’97 outturns is that the system is biased in favour of Labour. It is a neat narrative. Both sides can take some comfort: Labour supporters are reassured and Conservatives get to feel hard done by. But why is it, if the system is inherently biased towards Labour, that the BBC got the result of the 1987 election so badly wrong? Certainly some things – including constituency boundaries – have changed since then. But to switch from an unanticipated bias to the Conservatives to an even bigger bias to Labour in so short a period is astonishing. There is also no evidence at all for the proposition.
There is a number of reasons why projections go wrong. The first is the inherent weakness of the Butler-swing. This was developed to predict election results in the 1950s. It coped reasonably well with a two-party system, but was already failing by the 1970s. That it continues to be used when a superior system was developed more than three decades ago is a testament to incompetence in the media.
Let’s say the next election produces a fall in the Labour vote from 35% to 25%. For the moment, let’s simplify things and assume that all these votes go to the Conservatives. The Butler swing, or uniform swing hypothesis, assumes that in each constituency 10% of the electorate would move from voting Labour to voting Conservative. In Labour–Conservative marginals, where the votes may broadly reflect the national aggregates, this may be roughly right. But it will get other seats badly wrong. What would happen in a seat where Labour only got 10% of the vote in 2005? Are we to assume that every single Labour voter would defect to the Conservatives? And how about seats where Labour got less than 10%?
In a two-party system, such arguments are nit-picking. Who wants to predict the result in a seat where one party is earning under 10%? Surely, a foregone conclusion if ever there was one?
But how about, North Cornwall? The Labour vote came in last time at just under 12%. If five sixths of these voters switched to the Conservatives, the Party would easily take the seat from the Liberal Democrats? But is that really credible? Fortunately, Sian Flynn does not need to do that well to win. If she were to pick up just six percent from Labour and UKIP or just three percent from the Lib Dems she would capture the seat.
The Proportional Loss Hypothesis (PLH) developed by Dr Gordon Reece of Bristol University is a much more useful and accurate. Dr Reece is not a political “scientist” but, if I recall correctly, a mathematical engineer. His model suggests that if Labour was to lose 10% of its national vote then it would lose ten thirty fifths of its vote (about 28%) in each constituency. That means that actual number of votes Labour would lose in North Cornwall would be lower than the number it would lose in Camberwell and Peckham. This is, of course, to be expected, as Harriet Harman has far more votes to lose than the unfortunate Labour candidate in North Cornwall.
This model is based on reallocating actual votes which Labour won at the last election to other parties. This is certainly better than the Butler swing. Some opinion polls have Labour down by more than 12%, which Butler would predict giving them a negative vote in several Cornish constituencies. But, like any model, the PLH is a simplification. And, of necessity, things get more complicated when there are three or more parties involved.
I think there are at least two other factors at work which are degrading the value of projections for the next election.
The first is tactical voting. At the last four elections there has been a great deal of anti-Conservative tactical voting. By 1992 most people who not voting Conservative were decidedly hostile to the Party. For most Conservatives, the 1992 result came as a relief, but from a psephological point of view it is a real oddity. Given a lead of eight points over Labour, why was the Conservative majority so small? John Major’s margin in the popular vote was two to three times higher than Blair’s was in 2005. Labour bias? But the 1992 result was just five years after 1987, when the Conservatives had done unexpectedly well, in terms of converting votes into seats. Tactical voting seems to be the major factor here, with the anti-Conservative vote focusing on whoever was best placed to win. This continued in 1997 and 2001 and even, to a lesser extent, 2005. But the brand has been de-toxified. If there is any tactical voting next year it will almost certainly be against Labour.
The second is the Blair-factor. We overlook this at our peril. From pretty much the moment Tony Blair became Leader of the Labour Party in 1994 there was a sudden de-coupling of opinion polls and the economy. The economy kept improving, but Conservative fortunes did not. To make matters worse he parlayed a decent lead of 12 points in aggregate votes into a huge Parliamentary majority. How did he do it? Quite simply, Blair was more popular among swing voters than he was among hard-line Labour supporters, many of whom had to hold their noses to vote for him, even in 1997. Given the support Labour won in places like Croydon and Birmingham, computer models would have predicted huge votes in the Labour heartlands of Tyneside and the Rhondda Valley, but actually turnout was greatly depressed in such areas. Blair won Labour a lot of votes, but more than anything he won them votes where they needed them most. A depressed turnout in solidly Labour seats simply didn’t matter.
Peter Mandelson may be back in the government, but Gordon Brown is not Tony Blair. He is much more immersed in Labour and its traditions. Labour took to Blair in desperation after four defeats, but it never really loved or trusted him. There is a reason they called him TORY Blair. It seems strange to think of Gordon Brown as being popular anywhere, but in Labour’s heartlands he is much more loved than Blair ever was. It is perfectly possible to imagine Labour winning exactly the same number of votes at the next election, but if that is based on higher turnout on the banks of the Mersey and the Clyde and a collapse in their support in the West Midlands and West London, Labour will lose.
So these, then, are the reasons why the computer models underestimate Conservative prospects at the next election:
- There will be little or no anti-Conservative tactical voting.
- There may well be widespread anti-Labour tactical voting.
- Gordon Brown will be better at winning votes where Labour doesn’t need them than where the Party does.
