- Originally published May 08, 2017
UK 2017 General Election Poll Tracker
We’re tracking public opinion polls of the United Kingdom’s 2017 General Election here until election day, June 8. As of now, the Conservative Party is headed for a 6 percent margin of victory, most likely garnering a majority of seats in UK Parliament. Can Labour make a surprise comeback? Is UKIP doomed? Check back here for updates to our polling average and blog feed on #UK17.
Shift in Popular Vote Percentage Since the Last Election
- How do you average the polls?
Polls used in this average are compiled from Anthony Wells' UKPOLLINGREPORT. After collecting the individual polls I average them together using a common technique called pooling, which separates polls into individual observations for each day they are in the field. This approach was used in this research article on the 2016 United States election, which I coauthored with Dr. Christopher Wleizen of the University of Texas at Austin and Stuart Soroka of the University of Michigan. Importantly, this pooling creates an average that allows us to analyze respondents on the day they actually professed their vote intention, placing emphasis on those polls with a lot of respondents (up to 1900).
Once the polls are split into individual "pools," I give extra weight to observations conducted more recently. I do so by using to a weighted moving average with an exponential decay function. It's similar to what I've used in past election forecasts in the United States, Netherlands, and France. This weighting towards recency is of unique advantage in UK elections, where vote intentions move rapidly in the final weeks/days of the campaign.
- Is this a seat projection?
No. this polling average is not a forecast of seats in Parliament, just the national vote. You can find a more sophisticated, probabilistic forecast of seat shares at Chris Hanretty's ElectionForecast.co.uk. If you're wondering why I'm not computing a full election forecast, it's because Theresa May's Tories are almost certain to win a majority in parliament, and the minor parties are, well, minor. There are also some other items consuming my time this summer, and I really don't think I could produce a forecast as good as those from the pros!
- How do you compute a confidence interval?
Similar to what I've done in the past (which, of course, borrows from other modeling work), I compute a confidence interval based on historical polling error in UK elections. According to data complied by Christopher Wlezien and Will Jennings, the average error for UK election polls (as measured by the average error for each party's prediction in the 9 elections since 1979) is in the neighborhood of 3.5%. Accounting for the small sample size of UK elections, the 95% confidence interval for error in UK election polls is close to 9 percentage points. A 90% interval is 6%.
Of course, error is larger the farther out from election day we are — polls in the UK taken 2 months before an election are roughly one percentage point less accurate that polls taken 1 month before. The average takes this into account when computing confidence intervals.