NOTE 19/2/16: This post is not yet complete. I will do my best to complete it soon as possible. Since I started writing this post, additional data has been made available and I now think that 400 out of 650 seats voted Leave. You can see an outline of my older model in my YouTube clip published in November 2016.
For those used to staying up for election night, the BBC coverage of the EU Referendum on 23rd June 2016 must have been disconcerting. Where were the figures showing how many seats Leave & Remain had won? Unlike a general election where the winner is the party with the largest number of seats, the referendum was decided by a popular vote with Leave winning with 17,410,742 votes to Remain’s 16,141,241 votes.
Also different was that the results were declared for the 399 counting areas (CA) used in EU elections rather than the more familiar 650 parliamentary constituencies. Of the 399 CAs, Leave won a majority in 270 CAs as shown in figure 1. However, the counting areas vary enormously in size from 1,799 eligible voters for the Isles of Scilly to 707,293 eligible voters for the city of Birmingham which makes it difficult to compare CAs. The apparently overwhelming victory for Leave in terms of CAs could be a statistical mirage with Leave winning small CAs and Remain winning large CAs.
With a possible General election looming with Brexit at the heart of the debate, politicians and voters will want to know if their seats voted Remain or Leave. The only way to answer this question is to build a statistical model that extrapolates the results for the 399 CAs to the 650 parliamentary seats. I have done this using a method called principal components regression and I estimate that Leave won by a landslide with 403 seats out of 650.
I am not the only person to make such an estimate. The first person to do so was Chris Hanretty of the University of East Anglia and he estimates that 410 out of 650 seats voted Leave. His model uses the 2011 census whereas my model uses the 2015 general election results to estimate the Leave vote in each seat. Specifically, I used two derived variables for each of the 650 seats to predict the Leave vote share:
- GREEN KIPPER = UKIP vote share in 2015 minus the Green vote share in 2015.
- ENGAGEMENT = average of % turnout in 2010 & 2015 minus the BNP vote share in 2010.
TO BE CONTINUED – Please see my YouTube clip of November 2016 for an outline of the basics.