Seat predictions for the English local elections: sophisticated modelling or back of fag packet?

Every year Colin Rallings and Michael Thrasher at Plymouth University reveal their seat predictions for the local elections in England. This is based on a model which uses ward-level by-election performance to calculate a ‘national equivalent vote’ which is then converted into seats likely to be gained or lost.

However, a bit like the ‘sophisticated’ computer models Cornwall Council uses when forecasting population and household growth, there’s only one small problem, The predictions aren’t necessarily that accurate. The media focus on the predictions but rarely if ever ask the obvious question – how well did they turn out in the past? As we can see from the table below the Rallings/Thrasher model usually gets the direction of gains or losses right, but the actual number of seats won or lost are sometimes well off.

Rallings and Thrasher’s predictions compared to outcomes, 2013-2016

2013 predicted    actual 2014 predicted    actual
Con -310 -335 Con -220 -236
Lab +350 +291 Lab +490 +324
Lib Dem -130 -124 Lib Dem -350 -310
Ukip +40 +139 Ukip +80 +163
2015 predicted actual 2016 predicted actual
Con -450 +541 Con +50 -48
Lab -50 -203 Lab -150 -18
Lib Dem -80 -411 Lib Dem +40 +45
Ukip +400 +176 Ukip +40 +25

Their worst performance was 2015 when they completely missed the rise in Tory seats at the expense of the Lib Dems and badly over-estimated Ukip’s performance. Last year too, they forecast a gain for the Tories, which turned out to be a loss. In 2015 the local elections were combined with a general election, when the polls missed a swing to the Tories, and a higher than usual turnout, which made predicting the results of the locals more precarious. Of course, this year we have the novel factor of local elections taking place while a general election has already kicked off, which may also affect turnout, but to a lesser degree.

Here’s what they are projecting this year.

Rallings and Thrasher prediction 2017

Con       +115
Lab -75
Lib Dem +85
Ukip -105

There are around 2,300 seats up for grabs in the county and unitary elections in England. On past performance we might expect Rallings and Thrasher to be around 200-220 seats adrift over the four parties. This could be critical for party morale as their predictions are often used as benchmarks by both media pundits and political parties. If a party does better than predicted, then morale is boosted, if worse it’s dampened.

There’s no evidence of any systematic party bias in Ralling and Thrasher’s model. However, in three of the last four years they’ve over-estimated Labour’s performance. Doing so again this year will only stoke the feeding frenzy of the Tory press. So is their prediction of a 75 seat loss for Labour in England feasible?

Four years ago, when these seats were last fought, Labour made 288 gains, although it was only an average year for them as in the previous set of elections in 2009 (under Gordon Brown’s leadership note) they had performed abysmally, losing 313 seats. Labour are now running at 25-27% in the polls, compared with 38-41% in 2013.This is much lower so some losses must be expected. So a predicted 75 seat loss looks to be on the low side and is surely over-estimating Labour’s performance based on current polling. This is particularly the case as the rural shire counties are hardly the best ground for Labour.

I would suggest a more realistic forecast would be for Labour to suffer a much higher loss, of around 170 seats,(which is still a couple of hundred better than 2009). With Ukip likely to lose over 100 of the 147 seats it won in 2013, the gainers will be the Tories and Lib Dems. The Lib Dems had a bad year in 2013, losing 130 seats, They may well claw back the majority of those. Which leaves a gain of around 150-160 for the Tories, with the Greens, Independents and others taking the balance.

So here’s my alternative prediction, drawn up on the back of a fag packet. By this time next week we’ll know which method has worked best.

Kernowpolitico prediction 2017

Con      +150
Lab -170
Lib Dem +110
Ukip -120
Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s