No change in the polling average this week (George Osborne’s Autumn Statement appears to have had little impact on the polls), so there’s very little change in our forecast since last week.
It means Labour retain the lead they gained last week, with a 54% chance of winning the most seats (up a touch from 53%). It’s only the sixth week we’ve had Labour as favourites since launching our model last October, and, significantly, the first time they’ve kept a lead from one week to the next.
The most likely outcome next May remains a Hung Parliament, with a 59% chance. Our model now gives Ed Miliband’s party a 23% chance of winning a majority, and Cameron’s an 18% chance.
Our central forecast is for a Hung Parliament with Labour the largest party, on 299 seats to the Tories’ 291.
However, our seats forecast doesn’t currently take into account the variation of swings between constituency, which looks like it will be very important for the outcome of the election. Steve has written about his research into this using British Election Study data, here. We therefore need a better method, which is currently under development and will hopefully be published early in the new year.
Date of forecast: 12 December 2014
Days till the election: 146
Inputted current average poll shares
Others (inc. UKIP): 28%
– UKIP: 16%
Forecast Election Day Shares (with 95% Prediction Intervals)
Con: 33.5% (±6.7, i.e. 27% – 40%)
Lab: 31.0% (±4.9, i.e. 26% – 36%)
LD: 10.8% (±6.5, i.e. 4% – 17%)
Implied point estimate shares for:
– Others (inc. UKIP): 24.7%
– UKIP: 14.1%
Forecast Election Day Seats (with approximate 95% Prediction Intervals)
Con: 291 (221 – 372)
Lab: 299 (222 – 365)
LD: 29 (24 – 35)
(Prediction intervals assume LD & others shares at central forecast, Con & Lab shares vary as per prediction intervals)
Central forecast: Lab largest party, but short of a majority by 27
Approximate probabilities of key outcomes
Con largest: 46%
… with a majority: 18%
Lab largest: 54%
… with a majority: 23%
Hung Parliament: 59%
… with Con largest: 28%
… with Lab largest: 30%
(probabilities may not sum due to rounding)