The Bronze Medal

The Greens are unlikely to be third for vote intention share.

After two polls, there were claims the Greens are the “third largest political party”.

Surveys provide estimates, subject to many sources of potential error. A difference of one point between parties could be due to sampling error alone.

The two polls

A headline on The London Economic website asserts:

Greens replace Lib Dems as UK’s third largest political party

The article on 12th October begins:

The Greens are now the third most popular party in the UK, new polling has found, as support for the Liberal Democrats dwindles.

YouGov interviewed 1,673 GB adults via their internet panel, on 6–7th October. The Times and its sister Sunday paper sponsors that polling series.

The central vote intention estimate was the Greens had 6% share. In this survey, the Liberal Democrats had a share of 5%.

Later YouGov polls did not estimate the Greens ahead of the Liberal Democrats.

There were similar claims after a poll from Number Cruncher Politics. The online polls were for ITV’s Peston, of 2,088 UK adults on 9-17th October. Again, the survey estimated vote intentions of the Greens at 6%. The Liberal Democrat point estimate was 5%.

That Number Cruncher Politics poll led to similar claims on Twitter:

Errors and estimates

Up to 26th October 2020, those surveys are the only two polls estimating the Greens were third.

Polling is an imperfect instrument. Surveys provides estimates, subject to many sources of potential error.

There can be sampling error. This is the cost of not asking everyone. Beyond sampling error, there could be:

  1. Specification error (validity): the question is not measuring what the researcher intended.
  2. Coverage error (frame error): units that should be in the frame are missing or duplicated. Some units in the frame are ineligible.
  3. Non-response error: respondents have major differences with those who did not respond.
  4. Measurement error: how researchers did the survey affected recorded values. Examples include socially desirable answers to interviewers, or mistaken clicks.
  5. Processing error: after collecting responses, researchers make mistakes with imputation, coding or weights.

Researchers make choices. There are choices about questions and response options. There are choices about sampling methods and survey mode. There are choices about weights and turnout models.

Those choices mean some companies overestimate parties compared to the average company. House effects are these systematic overestimates and underestimates.

In a simple sample of 1,500 people, the margin of sampling error for a true proportion of 5% would be around one point. The margin of sampling error for the difference of two parties (at 6% and 5%) would be closer to two points. Design factors could mean the possible sampling errors are higher.

There are many other polls that do not imply the Greens rank third for vote intention. Since the year began, vote intentions for the Liberal Democrats have waned. It appears close between the Liberal Democrats and Greens.

This is a Bayesian regression model. (Image: Jack Bailey)

Due to house effects and sampling error, we would expect some polls to show the Greens in third.

Based on recent polling, the Greens are unlikely to be third for vote intention. That could change in future.

Readers should focus on polling averages and trends. The British Polling Council wrote a guide to help journalists writing about polls.

This blog looks at the use of statistics in Britain and beyond. It is written by RSS Statistical Ambassador and Chartered Statistician @anthonybmasters.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store