Stephan Shakespeare: The myth of the campaign swing
In the first of a series of five myths and truths of polling Stephan Shakespeare examines the idea of The Big Swing. Follow Stephan on Twitter.
As we start getting serious about the next general election, we should remind ourselves of some of the basic political realities measured or illustrated by polling. First up, and perhaps the most important for the campaign planner who actually expects to win, is that magical ‘Big Swing’ which at some point lights up even the limpest horse-race, but which is almost always a pure phantom, an in-built illusory effect of our desire for narrative structure.
Remember all that noise about how Romney was closing the gap – nay, winning! Panic in the Obama camp - all hands to the pump! But the race ended exactly as it began, with the President a few points ahead. YouGov’s contention was that Obama was ahead by about the same margin every single day of the election (every one of our many polls had him between 1% and 4% ahead). The only real excitement came from observer over-reaction to error in the polling.
Let’s focus on what was supposed to be the critical moment: the first presidential debate. The polls told us that Romney outperformed Obama. A swing duly followed – Gallup even showed Romney 6% ahead. Republican pundits stoked up the excitement and almost all observers were taking it seriously. But by polling day it was all gone. Two swings? Or none?
But there was a twist: in the second poll, not everyone could be bothered to take part: only 80% of the original sample allowed themselves to be re-interviewed, and the group of poll-refuseniks was differentially slanted to Obama voters, 22% of Obama supporters v 18% of Romney supporters, hence the apparent ‘swing’. The best-fit explanation: there was no real change, but Obama voters, demotivated by perceptions of a poor Obama performance, were less likely to make the effort to respond to a pollster. (By the way, for the record, YouGov’s final poll got every state right except for Florida, which was after all a statistical dead-heat - and that’s just using our own data, Nate!)
Or take an example nearer to home: the first Prime Ministerial Debate on May 15th 2010. I’m pleased to say I made my warning before all the excitement began: “Beware of the immediate post-debate polls” I wrote, and outlined my ‘Theory of Froth’, that the people who are most likely to ‘change’ are those least likely to actually vote: those who, when confronted by a pollster, don’t really care but also don’t want to admit it; and they are most likely to be nudged one way or another by whatever recent news they noticed. So a huge PR victory for Nick Clegg - dancing in the streets, LibDem surge – all meant nothing and was followed by a hugely disappointing actual LibDem vote. In the case of the Obama v Romney debate, where there was no great difference between them, so we saw that random and equal effect of 7% ‘switching’ one way and 8% ‘switching’ another. Does anyone believe that those dull, guarded, scripted TV confrontations made so many people rethink their Presidential choice?
Most of the academics who spend years poring over the data agree: there are only the smallest changes in real voting intentions within campaigns. That doesn’t mean you can’t make a difference; every genuine half-percent can matter to the result of a genuinely close race; but if we don’t recognize how tiny actual campaigning effects usually are, you might forget that elections are really won and lost long before the official campaign ever starts. The time to get your strategy in place is pretty much right now.