After a wave of favourable polls, May2015 now projects Labour will win 279 seats in May, with the Tories forecast to take just 264.
That prediction is made by taking a two-week average of the national polls (rather than the 5-day one you’ll find on the May2015 homepage), and combining them with Ashcroft’s 130 constituency polls of key marginals.
This weekend’s polls mark four and a half months since the party conferences. Since then, as we noted last week, many pundits have been expecting the Tories to displace Labour atop the polls. More importantly, a pair of academic election forecasts – Elections Etc and Election Forecast – predict this will happen by election day.
Initially, it seemed like we were heading in that direction. After Labour consistently led the polls by 3-4 points in September, their lead slipped in October. The Tories closed to within a point of Labour just after the Tory party conference, and then did so regularly: in late October, mid November, late November, mid December, early January and mid January.
In late January the Tories even took their first substantive lead in May2015’s 5-day Poll of Polls. It appeared the familiar arguments of many pundits – voters will shy away from Ed Miliband as the election approaches, especially as the economy keeps improving – were coming true.
The polls are really swinging between a slight Labour lead and a tie, rather than an actual tie.
But two things rewrote that narrative in the past week. First, nine of this week’s 11 national polls put Labour on 33-35 per cent. Eight of 11 put the Tories on 31-33 per cent. Labour now lead May2015’s 5-day Poll of Polls by 2.3 points – they have re-taken their lead.
More broadly, this week was the latest confirmation that the polls are really swinging between a slight Labour lead and a tie, rather than an actual tie. If we were in a real tie we would expect the Tories to be ahead in as many polls as Labour are. Instead, Labour have led in nearly two-thirds of the 63 polls released so far in 2015.
Yet the race is clearly narrower than it was in September, when Labour led in almost every poll and often by 3-4 points. Labour’s lead is now more like 1-2 points. What’s happened?
A rarer type of poll – released by ComRes for ITV News on Friday – shows that Labour may not have actually lost any support in England & Wales since September, despite the party’s supposedly torrid end to 2014.
ComRes polled voters across the 40 most closely fought Tory/Labour seats in 2010. 25 are Tory seats won on very narrow margins, 15 are Labour seats on similarly narrow margins. Polls like these give us a sense of whether Labour are likely to win Tory seats in May, or the Tories are likely to win Labour seats.
ComRes’ poll—combined with national polls—suggests Labour’s standing in England & Wales hasn’t really dipped.
The poll showed Labour nine points ahead, which sparked more than 100 retweets. But that kind of lead is exactly what the national polls actually imply.
How do they imply this if Labour only lead by 1-2 points in national polls? Because if Labour lead by 1-2 points, that is a 5-point swing from 2010, when they lost the GB-wide vote by about 8 points. The aggregate Tory lead in the 40 marginals ComRes polled was necessarily tiny in 2010. Therefore a 5-point swing nationally implies an 8-9 point Labour lead in these 40 marginals.
That is what ComRes showed, and it is exactly what Lord Ashcroft’s constituency polls suggested over the summer. But confirmation that these leads are still in tact is helpful. It’s the latest evidence that a Tory majority is highly unlikely. 
But, more importantly than that, the ComRes’ poll – combined with national polls – suggests that Labour’s standing in England & Wales hasn’t really dipped since the conferences. The academics behind Polling Observatory argued as much last week on May2015.
Labour’s lead has dipped because of Scotland.
Labour’s lead has dipped from 3-4 points to 1-2 points because of Scotland.
Over the summer Labour were still polling in the high 30s in Scotland (they won 42 per cent north of the border in 2010). The referendum changed that. Their Scottish support has nearly halved since September. An average of the sixteen Scotland-wide polls published since the referendum puts the party in the mid-20s, which Lord Ashcroft’s Scottish seat polls confirmed ten days ago.
A fall from 40 from 25 per cent in Scotland would mean Labour losing 1.3 points in national polls (Scotland is nearly 9 per cent of the electorate). A fall from 37 to 27 per cent would mean an 0.9 point fall.
In other words, Scotland have lost at least a point because of Scotland, and maybe closer to 1.5 percentage points. That accounts for most of Labour’s dip in national polls since summer’s end.
Of course when the polls dip towards showing a tie, rather than a 1-2 point Labour lead, then Labour have lost a fraction of support in England & Wales, but Labour’s popularity clearly hasn’t faded much.
David Cameron has very little to be happy about.
Ed Miliband’s party is still set to win dozens of Tories seats (Lord Ashcroft has put Labour ahead in more than 40 Tory-held seats). Labour won’t easily won the most seats in May – as this site predicted in September – because their support has collapsed in Scotland.
But until the Tories actually start leading in a majority of polls, or Ashcroft returns to the closest Tory-Labour marginals and shows things have changed, or a marginal group poll like ComRes’ doesn’t conform to what we expect, David Cameron has very little to be happy about.
Cameron may be more popular than Miliband and his party may be more trusted to steer the economy (although that question is too blunt), but he’s the party leader set to lose seats in May.
 We would say impossible, but as Nate Silver points out in his The Signal and the Noise, one study has shown events that forecasters said were impossible actually happened 15 per cent of the time.