Read this piece on May2015.com, our election site, for all the graphs that come with it.
On Monday an ICM poll suggested the Tories will win the election by more than a million votes. Today a TNS poll indicated Labour will win by 2 million. That’s quite a change.
In reality, little has changed. We’ve had six polls since Monday morning, but May2015’s 5-day Poll of Polls hasn’t moved much: Labour’s lead has shortened from 2.3 to 1.5 points.
Over the weekend we suggested the race has scarcely changed since the party conferences in early October. Labour have lost support in Scotland, which is why they’ve dipped since September, when they had a 3-4 point lead. But the party may not have slipped much at all in England & Wales.
The race is alternating between a small Labour lead (1-2 points) and a tie. The Tories have only twice led May2015’s 5-day Poll of Polls in 2015.
Outliers are being given air time by punters tired of a lifeless campaign.
That analysis looked at more than 150 polls over the past five months. But this week has shown how a lone poll can excite pundits into thinking everything has suddenly changed.
Outliers are being given air time by punters tired of a lifeless campaign. That’s all that’s really happening. The TNS poll should be taken with a pinch of salt: they’ve put Labour ahead like this before (by 6 two weeks ago, 7 in early January and 7 in mid-December). The ICM poll was more surprising, but it’s still just one poll with a 3-point margin of error.
So what would actually make news in polling? Here are three things we wouldn’t brush over so quickly.
1. YouGov hand one party a series of significant leads
YouGov poll five times a week. They are the highest-profile pollster, publish in the Sun andSunday Times, and have UK Polling Report’s Anthony Wells manning their numbers (not to mention Peter Kellner, Nigel Farage’s favourite pollster).
So far in 2015 they’ve published 33 polls. The leads they’ve given each party have all been within the margin of error (± 3 per cent). They swing between showing ties and small Labour leads, with the occasional small Tory lead.
If they had put Labour ahead by 7 today, it would have been news (their little noticed overnight poll put Labour up 1). It would still have been an outlier. But if they were to consistently put either party ahead, by say 3-4 points, something will have changed.
2. Populus start handing the Tories’ leads
Populus, the UK’s second most-prolific pollster, have given Labour a lead in every one of their thirteen 2015 polls. On average, they give them a 2.3-point lead. That’s in-line with the 2-point leads they gave the party throughout late 2014.
A series of polls showing the reverse, with the Tories in front by a few points, would be news.
3. Ashcroft … becomes more consistent
The Ashcroft polls that make news – Labour up five in late November, Tories ahead by six in mid-January – have proven to be outliers. Eight of his past twelve polls have shown a tie or put the parties within a point of each other.
It would news if he started putting one party ahead regularly, or stopped occasionally putting either ahead by a bunch. Not that a series of ties will win much coverage.
*
YouGov, Populus and Ashcroft are the most regular pollsters. But there are six others. Here are three general rules for when any poll is released:
1. Compare it to all the other polls
ICM and Ipsos MORI are the oldest pollsters, and they publish once a month (in the Guardianand Evening Standard respectively). The rarity of their polls and platform these papers give them – along with their name recognition in Westminster – makes their numbers newsworthy.
But they should always be compared to other polls at the time. If we look at how ICM’s numbers have compared to our Poll of Polls each month, we can see how this week’s ICM poll was more different than any other in the past year.
On average, they had been different by 1.9 points across the eleven polls before this week’s. So on Monday we should have expected anything from a 2.9-point Labour lead to a 1-point Tory lead.
That’s why the 4-point Tory lead was news. It suggested a shift for ICM. But because they only poll one a month, we can’t put too much stock in the number. They put the Tories ahead twice last year (in May and July) before reverting to a Labour lead.
2. Compare it to that pollster’s previous polls
We can’t put too much stock in individual polls, but we should track the trend. The trend among each pollster’s numbers is what really tells a story, not the specific lead they’ve shown that week or month.
Plugging every poll into a ‘Poll of Polls’, as we do, is helpful, but it irons out the way each pollster is shifting over time.
Ipsos MORI’s monthly polls are a showcase for tracking the trend. Since January 2014, they’ve gone from showing a consistent and comfortable Labour lead to showing a mix of small Labour leads, Tory leads and ties.
In other words, we’ve gone from a 6-7 point Labour lead to a 1-2 point Lab lead, which is what our Poll of Polls shows.
The specific monthly numbers are really approximations. MORI are saying “This is the state of the race, ±3 points”. [1]
3. Be aware of each pollsters’ past ‘outliers’
We treated TNS’s 7-point Labour lead lightly today because they’ve put Labour ahead by at least 6 points in three of their last four polls.
If Opinium, the Observer’s pollster, published a similar poll we’d have a fairly similar reaction. They have put Labour ahead by 5 or more three times in the past three months, although recently they’ve had fewer outliers.
The precedents of ComRes and Survation, two of the industry’s newest firms, shape the way we think about them too.
When Survation put Ukip on 24 per cent again in a few weeks, it won’t mean much.
Survation, who poll irregularly for theSunday Mirror and privately for Ukip, think Ukip are polling 23-25 per cent. These aren’t occasional outliers so much as systematic differences. Our Poll of Polls puts Ukip on 14-15 per cent. So when they put Ukip on 24 per cent again in a few weeks, it won’t mean much.
Correspondingly, they give the Greens 2-4, rather than 5-6, per cent. It’s the same story with ComRes’ monthly-ish poll for the Independent on Sunday. [2] They put the Greens low and Ukip high (close to 20 per cent).
If Survation or ComRes started putting Ukip on 10-12 per cent and the Greens on 8-9 per cent, that would be news.
For a broader look at why pollsters differ, check out Anthony Wells’ recent take.
Explore May2015.com.
Content from our partners
Related
