So, what's Britain Elects?

Founded in 2013, initially as an archive for council by-elections, we are now the UK’s leading poll aggregator. Our linear moving average trackers are weighted to reduce volatility and provide the most accurate representation of public opinion on key political questions.

Our poll trackers are updated as data is released and cover voting intentions for the main political parties, leaders and policies at local and national level, as well as Brexit. We also publish polls in other subject areas that are not included in our trackers. We are a non-partisan organisation and only use polls, for the purposes of transparency, from pollsters who are members of the British Polling Council.

Polling is both an art and a science so Britain Elects also offers a rolling commentary both on the latest poll results and in-depth analysis on the wider polling industry and its challenges.

At present we are an entirely voluntary organisation, set up by those with an interest and experience in consultancy and politics and funded by your kind donations.

Who runs Britain Elects?

Britain Elects is run by Ben Walker, the site’s Co-Founder, and is frequently helped by friends and colleagues at election time: Dayle Taylor, Thérèse Martina, and Chris Dickinson.

Ben founded Britain Elects alongside his late best friend Lily at the age of 18, and has been running and professionalising it ever since. Now also Data Journalist with the New Statesman, Ben has a keen interest in the state of both British and world public opinion and voter psychology.

Ben is also the one responsible for sarcastically responding to trolls on Twitter.

Methodology:poll trackers

The various Britain Elects poll trackers you see on our site and elsewhere are designed to cut through the noise and provide an understandable summary of where British public opinion stands. The trackers start as simple linear moving averages but are modified with additional weightings so as to account for factors such as sample size and publication date.

  • We weight according to sample size. There is little difference between a reasonable 1,000 person sample size or something larger, but because with a larger sample you can be more confident in viewing representative data in the cross-breaks, we weight them higher.
  • Prolific pollsters are weighted down. Our tracker seeks to measure all pollsters. We weight down pollsters which publish more regularly than others so as to limit a potential over-saturation.
  • We weight by publication date. Newly published polls have a higher weighting in our tracker than older polls. 
  • We weight by historical pollster accuracy.

Methodology:election forecasting

The Britain Elects National Forecast is a projection of what ‘probably’ would happen if a general election were held today. It factors all Westminster voting intentions into its formula alongside weightings based on general demographics, Leave/Remain vote, regional variation, anticipated turnout, and local historical performance.

The headline data we use is initially drawn from our very own poll tracker which features all Westminster voting intentions that pose the standard voting intention question. Pollsters which afford equal prompting to parties that have demonstrated a certain sustained level of national support are given greater weighting than pollsters which don’t. What reliable regional and devolved data we know are also included in the calculation.

Using just the above would produce a forecast with uniform swings across all constituencies. Uniform National Swing is not usually a reliable indicator of a general election result, particularly when we have such a polarised electorate, so we must gather further data. We must, as the nerds say, stratify.

We go on to calculate a party’s performance in a constituency relative to its national changes weighted according to previous local election results in the area as well as the party’s general performance with certain types of voters. This requires us to use data from the Office of National Statistics: census data. An example would be thus: were our poll tracker to put the Liberal Democrats 20pts up on its performance in a previous election, such swings in their favour would most likely be depressed in constituencies which voted heavily for Leave in 2016. Another example would be parties with poor relations with voters of certain backgrounds, be they faith-based, academic, or certain living standards. We base our weightings for these off general polling taken over the course of the past five years. Assumptions on our part are made as well.

We would then go on to produce the forecast, accounting for a margin of error and allowing our system to produce multiple simulations within the bounds of that range, offering a probability of how a seat might go based on all that has been included and accounted for.

This is the most thorough the Britain Elects National Forecast has been to date. Our model in 2017 followed through on regional variation, but wasn’t as thorough when examining demographic data. Our 2017 forecast (Con 356, Lab 219, SNP 43, LDem 9) performed poorly in projecting the national numbers, but was correct in forecasting a swing to Labour in the South of England, albeit not to the degree it was in actuality.