Two polls over the weekend show Labour 17 points ahead of the Liberal Democrats in the Oldham East and Saddleworth by-election. So is Labour’s Debbie Abraham a shoo-in, and Nick Clegg heading for a bloody nose, as much of the commentariat seem to think?
I am not so sure. I expect the margin of victory on Thursday night to be narrower, and do not completely rule out a victory for Elwyn Watkins, the Liberal Democrat candidate.
To see why, consider three by-elections in Labour seats in the last Parliament – the only three, as far as I am aware, in which opinion polls were conducted.
First, Crewe and Nantwich in May 2008. Three polls (two by ICM, one by ComRes) all showed the Tories ahead, by margins of 4, 8 and 13 points. In the event the Conservatives won by 19 points. The three polls overstated Labour’s eventual support by between four and eight points.
Second, Glasgow East in July 2008. Two polls were conducted in this traditionally very safe Labour seat. Both showed Labour well ahead, with the Scottish National Party in second place. One, by ICM, conducted two weeks before polling day, reported a 14 point Labour lead (Labour 47%, SNP 33%). The second, by Progressive Scottish Opinion the following week, reported a 17 point lead (52-35%).
However, on the day, the SNP narrowly won the seat, with 43% to Labour’s 42%.
Third, Norwich North in July 2009. The only survey was conducted by local students just before the start of the campaign. It pointed to a narrow Tory gain: Conservatives 34%, Labour 30%. In the event the Tories’ Chloe Smith won easily, achieving 39%, more than double Labour’s 18%.
If one delves further back into history, we find that overstatements of Labour support in by-elections are commonplace. In Chesterfield in 1984 – the election that saw Tony Benn return to the Commons – I reported a Harris poll for the Observer that showed Labour fending off the Liberals by 53-26% with a week to go. On the day, Benn won by the somewhat narrower margin of 47-35%.
In Brecon and Radnor in 1985, one poll, published on polling day, said Labour would capture the seat from the Tories with an 18-point lead over the Liberals. The poll was right about the Tories slumping from first to third, but hopeless wrong about the winner – the Liberals captured the seat.
The problem of overstating Labour support is especially acute in seats that Labour is defending. I cannot recall a single example of any campaign poll understating Labour’s eventual vote share; and the polls from the last parliament are broadly consistent with polls in previous parliaments in overstating Labour support.
I say this not to denigrate by-election polls. Subject to normal considerations of margin of error, they are usually broadly accurate in what they measure. But what they measure is the mood of a constituency some days before election day. Often two things happen: many voters – more than in a general election – tend to make up their minds at the last moment; and Labour often has great difficulty in persuading their supporters actually to vote. Turnout is generally much lower in by-elections than general elections, and polls often understate Labour’s difficulty in motivating its supporters.
One other factor could make the race closer this week. The two weekend polls both show the Conservatives trailing a distant third. By Thursday, the Lib Dems will ensure that every voter in the constituency knows that the best way to defeat Labour will be to vote Lib Dem. If this late campaigning proves effective, Kashif Ali, the Conservative candidate, will underperform his weekend poll rating on Thursday, while Watkins will do better than those polls suggest. (I recall reporting an NOP survey in the Ribble Valley by-election in 1991 which showed the Tories well ahead of the Lib Dems and Labour a distant third. But when NOP asked how people would vote if they knew Labour couldn’t win, many Labour AND CONSERVATIVE supporters said they would switch to the Lib Dems, who would gain enough to win – which is exactly what happened.)
Of course, this week could be different. We know from its performance in marginal seats in last May’s general election that Labour has sharpened its act in getting its supporters to turn out when it matters. The Populus poll in Oldham East and Saddleworth, commissioned by Lord Ashcroft, shows that Labour activists are speaking to as many local voters – in person or by phone – as the Liberal Democrats. (The same poll, incidentally, suggests that the Conservatives are making little attempt to contact voters in person. This bears out the view that David Cameron would prefer supporters of the coalition voters to vote Lib Dem rather than Tory.)
Moreover, the specific circumstances of this by-election mean that past precedents might be misleading. And the fact that Labour is in opposition and the Lib Dem in government could well make a difference. By-elections are generally better for oppositions that exploit voter-disenchantment with the government of the day than governing parties struggling to overcome the disappointments of their supporters.
Ed Miliband must be hoping, and Nick Clegg fearing, that Oldham East and Saddleworth will not follow the pattern of past by-elections in Labour seats. Later this week we shall know whether those hopes and fears are justified, or whether the pattern of past by-election polls in Labour seats has persisted.
