Maps, Swings and Trivial Things

The London Economic recently posted an article which showed an electoral map bathed in red, providing only young people were permitted to vote.

Despite being hailed as “new research”, the map was created by running a small sub-sample from a single survey through a projection tool.

This method could be inaccurate. The article has been shared over 25,000 times.

In short

Beware sub-samples: Data tables contain voting intention estimates by various segments. These estimates should only be considered indicative.

Uniform national swings: There are some large regional variations in public opinion, particularly within Scotland. A uniform swing could be misleading.

A bit of fun: When over 60% of specified voters back a single party, we can expect that party to take the vast number of seats. The map is trivial.

The Red Map

On 4th September 2018, the Twitter account ‘Election Maps UK’ posted an image showing what the British electoral map would look like if only those aged between 18 and 24 had voted.

The views of this age group were found by looking at a sub-sample from a recent YouGov poll, sponsored by The Times.

When you look at that survey, it shows that 66% of those expressing a voting intention in the 18–24 age group would vote for Labour.

This is a sub-sample of 186 respondents (167 in the unweighted sample).

More importantly, sub-samples are not ‘internally weighted’.

As an example, YouGov respondents are weighted to give the right number of young people and the right number of Londoners. However, these weightings may not give the right number of young Londoners.

To be exact, YouGov uses age interlocked with education and gender, and 2017 recalled vote interlocked with region, among its weighting procedures.

Swings by Region

The idea of a uniform national swing has serious limitations.

The assumption is that the changes in voting intention from the last election will be the same in every constituency. In reality, swings vary across the country.

This method could still provide an accurate seat count as long as the differences from the average are ‘symmetric’: for every seat that has an under-average swing, there is another seat with an above-average swing.

In Scotland, a party which stands nowhere else — the Scottish National Party — has a large lead over both the Conservatives and Labour.

In a YouGov survey conducted between 1st and 5th June 2018, sponsored by The Times, the SNP led the Conservatives by 14 points.

For 18–24 year-olds, the sub-sample had 45% intending to vote for the SNP.

Accounting for this regional variation would seriously dent the Labour seat count.

Triviality

It should be unsurprising that an electoral map which only considers the vote intentions of the youngest voters would be bathed in red.

In 2017, the British Election Study found over 60% of younger voters cast their ballots for the Labour candidate.

Ipsos MORI also estimated — based on pre-election voting intention surveys — that 62% of voters aged 18–24 backed Labour.

A well-shared map showing something well-known shows the power of data visualisation, even if the methods to derive that map are potentially (and inadvertently) misleading.

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

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