Wednesday, 27 April 2011
YouGov! What, MeGov?
If opinion polls are to be believed, Labour is heading for an embarrassing dip in votes and seats at the forthcoming Holyrood elections. Some more excitable commentators have even been moved to predict a ‘crushing defeat’*. However, as most activists and politicians know, polls are more useful as weapons than as sources of clear information. They may (as these recent ones undoubtedly will) be used to galvanise Labour’s notoriously sluggish turnout. They will be used to glorify the name of Salmond. But in all cases polls are deceptive. They were when predicting a Labour lead of up to 10 points early in this campaign, they were when predicting a Lib Dem surge in 2010’s Westminster election, and they are now.
Anyone remember last year, the incredulity with which both media and politicians dismissed the TV exit poll, because it gave the lie to the polls’ prediction? For the record that exit poll forecast Con - between 303-306 seats; Lab - between 251 and 262 seats; LibDem - between 69 and 55 seats (the figures were revised as the night went on) The actual result? Con - 306; Lab - 258; LibDem - 57.)
There are at least three reasons why opinion polls need to be treated with caution. Pollsters themselves recognise that, unless the sample is huge, a margin of error factor of plus or minus 3%. So a poll predicting a 6% lead could equally suggest a neck-and-neck race. But this is well-known (although not well-reported). Less well considered are methodology and sample taking. YouGov for example, selects its respondents from an internet panel of people who have chosen to volunteer for this work in exchange for cash. The company, says it uses demographic information to balance its results, but there are two or three chances for skewing the sample here, and their methodology remains hotly debated.
Less obvious recently has been the introduction into Scotland of differential swings in local areas. It has been very obvious in Westminster elections that swings vary considerably in different parts of the UK - the increasing Westminster vote for Labour last time being a glaring example . It has been less so within Scotland. I suggest that this is changing, and will make this election virtually impossible to call by extrapolating from opinion polls. It is not just where the disenchanted LibDem votes go. The SNP will surely do best in rural constituencies if that vote does implode, but it may well be different in places like Dunfermline and possibly in Edinburgh. And in the second vote, any shift away from the LibDems will surely benefit the Greens. It seems also clear that the Tory vote will largely hold up - even if only because it has been bumping along the bottom for some time.
It isn’t simply the potential melt down of the LD vote (and in any case I suspect that the polls overestimate this likelihood too.) but the impact of local party politics via local council control. For the first time we have a Scottish Election being fought with councils under a variety of party and coalition controls. This may play out differently in different areas. Again Edinburgh, with the LibDems and SNP in coalition presiding over the debacle of the tram project, and proposals to outsource huge rafts of council services, will LD votes go to their SNP partners? Has the recent controversy over education cuts in Renfrewshire, and the SNP council leader’s climbdown scuppered his chances in Renfrewshire North? What will be the impact of the Aberdeenshire Council LibDem’s public fallout over the Trump development - particularly in the list where Cllr Martin Ford now tops the list for the Greens?
Of course the final reason that opinion polls never tell the whole story is that they are never told the whole story! The recent YouGov poll for example - even now finds that around 30% of those asked are either uncertain whether they will vote, or certain not to. That is a large undecided area to exclude. (YouGov also has a very low incidence of ‘won’t say’s - almost certainly because their sampling comes from the self-selected panel referred to in para 1).
So, while the trends are important and justify the Labour Party’s attempts to rejuvenate their campaign, the details of recent polls are no more (or less) significant than the earlier ones.
As ever, oppositions don’t win elections, governments lose them. Only May 5 will tell whether the Scottish electorate is sufficiently annoyed with the SNP government for them to have ‘lost’ this one. I suspect that it is going to be a much closer call than opinion pollsters tell us.
*Peter Kellner in a commentary on his own organisation’s poll