Political

How accurate are the Thrasher and Rallings local election predictions for Lib Dems?

As I warned in advance, the Thrasher and Rallings prediction for the Lib Dem vote share* in the 2016 local elections was indeed on the high side.

In fact at a 2 point over-estimate, the error was in line with their previous pattern of over-predicting the Lib Dem share by on average 2 points in the last Parliament.

It is a good example of the value of a little bit of data crunching, yet their predictions rarely get a test against their past form when they come out each year.

Here then is the overall record of their vote share predictions for the Liberal Democrats, based on by-election results in advance of the big May round of elections.

Forecast Lib Dem % Actual Lib Dem % Error in forecast
2006 29% 25% 4%
2007 29% 24% 5%
2008 n/a 23%
2009 27% 25% 2%
2010 n/a 23%
2011 17% 16% 1%
2012 17% 15% 2%
2013 16% 13% 3%
2014 14% 11% 3%
2015 n/a 10%
2016 16% 14% 2%
2017 21% 18% 3%
2018 18% 14% 4%

A set of numbers to remember for next spring…

To see the full context of the predictions and results each year, see my PollBase collection of local election results.

* Different rounds of seats are up for election each year, so Thrasher and Rallings calculate a national equivalent vote, which adjusts for this and therefore make possible like-for-like comparisons between different years. The intent is the same as for the separate BBC / Curtice and Fisher calculations, though following a different methodology and hence also usually producing slightly different results. The fact that however they both, with their own different approaches to the issue, consistently produce the same trends and overall picture – e.g. this time Lib Dems moving back above Ukip – is a good sign that the pictures they paint are right.

Advertisements

There are 3 comments

Share your views

Your email address will not be published. Required fields are marked *

All comments and data you submit with them will be handled in line with the privacy and moderation policies.