Who is going to win the general election? On Sunday we suggested Ed Miliband is more likely to become PM after May 7 than David Cameron. We ran through various scenarios and argued that Cameron will struggle to cobble together 323 seats next month – the number he needs to survive a vote of no-confidence and remain in Downing Street.
When we wrote that piece Miliband and Cameron were equally favoured by the bookies to take power. Miliband is now a 4/6 favourite, and Cameron’s chances have fallen below 50 per cent; his odds are 5/4. (Election Forecast, our fellow forecasters, and the Telegraph have written follow-ups to our take, which you can find it in this week’s New Statesman.)
That piece focused on the challenge Cameron faces. This one is centered around Miliband. If we think he is the more likely PM, what is his path to 323 seats?
Since Sunday, the polls have not – at least as a group – moved in a significant way. Labour and the Tories are polling between 33 and 34 per cent, as they were last week. As far as we can tell, that is not going to be good enough for Cameron. If he is, instead, ahead by four points, as three polls suggested this week (only to be blunted by others showing Labour leads), he is in a much stronger position, but still may not hold power.
The brutal maths of this election mean that, while Labour and the Tories are headed for roughly the same number of seats, only Labour has a clear path to 323 seats; their potential partner, the SNP, are set to win nearly twice as many seats as the Tories’ best option, the Lib Dems, and the minor MPs on either side largely cancel each other out.
For Cameron to win, he needs to win considerably more seats than Labour.
So for Cameron to win, he needs to win more seats than Labour – considerably more. We think he needs to win around 20-25 seats more. But let’s be specific. The Tories, Lib Dems and Ukip collectively need to win 314 seats. The DUP will add 9, if they win Belfast East, and take a Tory-led bloc to 323 (should the Lib Dems and Ukip vote together; we are not envisioning a formal agreement).
In this scenario, where Cameron leads a group of exactly 323 MPs – the bare minimum he needs – he can survive with Labour and the SNP winning 312 between themselves (9 minor ‘anti-Tory’ MPs – made up of the SDLP, Plaid, Greens, Respect and Northern Ireland’s Sylvia Hermon – will take this Labour-led bloc to 321).
So 312 is the maximum number of seats that Cameron can let Miliband and the SNP win. But if Miliband and the SNP can collectively win 314, Cameron will be ‘locked out’, in Alex Salmond’s memorable phrase. Those 9 minor MPs will take an anti-Tory bloc to 323 seats, and leave Cameron’s centre-right bloc with 321.
323 is the magic number you need for a majority in a Sinn Fein-less house, assuming Northern Ireland’s nationalists don’t lose Fermanagh to the Tory-leaning Ulster Unionists (Sinn Fein won it by four votes in 2010, three of which were later invalidated) and Sinn Fein continue to not take their seats. So how can the Labour leader reach 323?
The problem for Cameron is that we can quite easily see how Miliband, the SNP and the anti-Tory minor parties collectively win 319 seats – without winning a single seat that’s very close.
Here’s how. First, Labour hold the 217 English and Welsh seats it won in 2010. The only seat they may not hold is Bradford West, which they lost to George Galloway in 2012; but Galloway – for all his idiosyncrasies – can be counted on to vote against the Tories should he hold off Labour.
If Miliband and the SNP can collectively win 314 seats, Cameron will be ‘locked out’.
It seems unlikely that they will lose any of the other 216 seats to the Tories. The national swing in polls has gone in the other direction (Labour are winning Tory seats, not the other way around), and Ashcroft’s polls confirmed this in 2014.
The only seat that has ever seemed in peril was Southampton Itchen, where the rise of Ukip and departure of a longtime Labour MP seemed as if it might let the Tories in. But a recent Ashcroft poll implied the race has turned in Labour’s favour, and the local party there is well-funded and well-organised.
So we have 217 Labour seats and 106 to go. First we can add the 8 minor MPs (or 9 to 216 if Galloway hangs on): three from Plaid Cymru, who may lose Arfon, but only to Labour; three from the SDLP, who have strong majorities in their seats and are holding up in the few Northern Irish polls that exist; and one from the Greens (Caroline Lucas in Brighton Pavillion, who is only being challenged by Labour). The final one is Sylvia Hermon, who left the UUP before the 2010 election when they threatened to ally with the Tories, and seems secure in her seat.
We have 225 seats. None of these seats are contentious. None of them are going to turn Tory in the next 13 days unless one of these parties – Labour or the others – completely collapses. These seats seem to be beyond the effects of any plausible shift in the polls before election day.
Now we can take a significant step towards 314: we can add 55 of Scotland’s 59 seats to this Labour-led bloc. There are only four seats in Scotland that have more than a one-in-five chance of voting SNP or Labour, according to Election Forecast, who power the ‘Newsnight Index’ and Nate Silver’s UK prediction (and whose forecast for the SNP is the most conservative of those we track).
The problem for Cameron is we can quite easily see how an anti-Tory bloc wins 319 seats.
The four seats we’ve excluded are Orkney & Shetland, which we expect the Lib Dems to hold onto despite what national polls imply (and our model therefore predicts, but we’re changing this); and Dumfriesshire; Dumfries; and Berwickshire – Scotland’s three big border seats, all of which may vote Tory.
The only seats among the other 55 which one might contest would be the three Lib Dem seats of Dunbartonshire East (Jo Swinson’s seat); Ross Skye (Charles Kennedy’s); and Edinburgh West. But national and Ashcroft polls suggest the coalition’s minor partner have little hope in these seats, despite the popularity of their incumbent MPs.
So we have reached 280 seats. How exactly the SNP and Labour split their seats is, for Cameron’s survival, irrelevant. Both parties will vote against him, and the former will surely be compelled to vote the latter into Downing Street once they dispose of one Prime Minister – if only to then abandon Labour. (Our concern here is purely whether Ed becomes PM, not whether he lasts for very long.)
So we are now within 43 seats and still haven’t assumed Labour win a close race. We can move within 35 by handing Labour eight Lib Dem seats: Brent Central; Manchester Withington; Bradford East; Burnley; Norwich South; Redcar, Hornsey & Wood Green; and Cardiff Central. The final seat may be more competitive than public polls suggest, but Ashcroft has handed Labour double-digit leads in all of these seats.
How exactly the SNP and Labour split their seats is, for Cameron’s survival, irrelevant.
The Lib Dems are defending single-digit majorities in the first five, and all eight will switch to Labour on anything greater than a 6.5 point swing. That’s less than the swing implied by national polls, which is around 8-9 points. An ‘incumbency effect’ could theoretically save a few Lib Dems, but Ashcroft found swings of 10-12 points in five of these seats when he alluded to the local Lib Dem MP, and of 19 points in the other three; all far greater than the 6.5 point swing Labour need.
The Lib Dems are fighting a bunker campaign on limited resources – they have more winnable seats to defend. Our theoretical Labour-led bloc now has 288 seats. We are 35 short of 323.
These final 35 seats are what the election comes down to, and they will almost all be in England, in seats the Tories took from Labour in 2010. Can Labour win enough of these seats back? That is the game.
To work this out, we have looked at the 80 or so most closely contested Tory victories in 2010. These are the seats that Labour lost by as little as 0.1 percentage points (Warwickshire North; by 54 votes) to as many as 12.9 points (Basildon South & East Thurrock; by 5,772 votes). Ashcroft has polled 68 of these seats since last summer.
Before giving Labour a single Tory-Labour marginal, we can get them within 35 seats of 323.
May2015’s election prediction takes all those polls and adjusts them every day as national polls change. If Ashcroft hasn’t polled a seat since last summer, that poll will now be slightly more favourable to the Tories, because the Tories have caught up with Labour in national polls since then.
If Ashcroft hasn’t polled a seat, we just apply a fairly crude – but historically quite accurate – national swing to that seat (we use a variation on uniform swing). But he has polled the 64 most closely contested seats in 2010, so we have seat polls for almost all the crucial contests.
So we have a prediction, based on Ashcroft and national polls, for each of these 80 or so seats. We then cross-reference those forecasts with Election Forecast’s probabilities of a Labour victory in each seat (recall that EF are Newnsight’s and Nate Silver’s forecasters, and are relatively pessimistic about Labour’s poll share – they predict a 1.5-point Tory poll win on election night).
And this is what we can then create – click on the graph for a bigger version and to zoom in.
This graph compares the size of the Labour victory implied by Ashcroft/national polls, with the probability of one as suggested by Election Forecast. If a seat is in the upper right quadrant, both the polls and Election Forecast agree that Labour are likely to win the seat.
If a seat is in the upmost upper right quadrant (marked by the thin red lines), they strongly agree; Forecast give Labour at least a two-in-three chance in these seats, and polls suggest a win of at least 4 points – beyond the margin of error.
There are 24 seats in this upmost upper right quadrant. Remember, Ed Miliband needs 35 seats to reach 323. So if we assume all this seats vote Labour in 13 days, Miliband is within 11 seats of a majority.
Miliband needs 35 seats. There are 24 in the upmost upper right quadrant.
This is when things finally become contentious. So far, we have handed this Labour-led, anti-Tory bloc of parties 312 seats, and none of them seem to be particularly close contests. The winning party in these seats has always been at least a two-to-one favourite.
Where can Miliband’s final 11 seats come from? According to all the publicly available data we have, his best hopes are the light red seats in our graph – the ones that both the polls and EF agree are likely to vote Labour, but aren’t quite as convinced as they are about the other 24 (the dark red ones in the upmost right corner).
There are 9 of these light red seats: Hove; Nuneaton; Northampton North; Ipswich; Ealing Central & Acton; Warrington South; Croydon Central; Lincoln; and Finchley & Golders Green. These seats are in play, but Labour are favoured in them. If we hand Labour these 9 seats, an anti-Tory bloc is now within two of a majority.
To get this far we still haven’t had to hand Labour a seat where they are not favoured. You have to make far more heroic assumptions about the way polls will move or are wrong if you want to get a Tory-led bloc to 321 seats. That is why we say Miliband has to be considered the most likely post-election PM. Cameron can hold on, but his path is clearly more difficult.
As for Labour’s final two seats, they have many options.
First, they could win these seats elsewhere. They or the SNP could win one of the three border seats that we have handed to the Tories: Dumfries (where both are very competitive), Dumfriesshire, or Berwickshire (the SNP are the biggest challenger to Cameron in these two).
Labour have many options to win their final two seats.
Or they could win one of the Lib Dem seats we haven’t given them. We have assumed Nick Clegg will win in Sheffield Hallam, but Labour have led in Ashcroft’s two most recent polls of the seat and it is clearly very competitive. Or Labour could win Bristol West, which national and Ashcroft polls imply they will win, but Ashcroft hasn’t yet polled.
There are other Lib Dem seats that could end up in the anti-Tory column, like Ceredigion. Election Forecast’s disaggregated national polling data, provided by YouGov, suggests Plaid Cymru are serious challengers here.
But if they can’t win any of these seats, they have many more options in the Tory-Labour marginals. They need to win just two of the dark green seats in our graphic.
There are 12 of these seats. 11 of them are seats in which polls suggest Labour are in front, although Election Forecast’s model disagrees (remember, their model assumes a 1.5-point Tory poll win on election day and the polls are closer than that at the moment). One – Amber Valley – is a seat where Labour just trail according to the polls, but Forecast think will turn red.
We can cast the net wider. We can add Thurrock, where Ashcroft showed Labour slightly ahead of the Tories but just behind Ukip last summer. Or we can add the four light green seats – Swindon South; Colne Valley; Peteborough; and Rossendale & Darwen – where Labour seem to be strong challengers, if not favourites.
(We haven’t included five other seats – the light blue ones in our graphic – where Labour seem to have a one-in-three chance of winning, but we have, in turn, assumed they will win those where the Tories only have a one-in-three chance.)
Labour have many paths to 323 seats. If the polls begin to show strong Tory leads over the next 13 days its number of paths will narrow. But even when we base our analysis on a model which predicts a definitive Tory win in the popular vote, Labour have a convincing route to power. This is Ed Miliband’s race to lose.