In Defence of Liberty

Driven by data; ridden with liberty.

Your mates are not a representative sample

One of the more frustrating aspects of the referendum on the UK’s membership of the European Union is the treatment of polling.

Polling is a common usage of inferential statistics, where inferences and generalisations are made about a whole population from properties of a sample [1]. Inferential statistics is used to estimate parameters, and to test statistical hypotheses, such as what proportion of the public support or oppose a particular policy.

‘Everyone I know is voting [Remain/Leave]’

A common complaint is that the polls must be wrong because ‘everyone I know is voting [Remain/Leave]’.

The people you know will typically be geographically concentrated, over-representing social grades and age groups, and share a social connection: you.

All these factors affect the propensity to vote Remain or Leave [2].


Unsurprisingly, the desire to vote Remain or Leave varies by geographic location. (Source: YouGov)

Your mates are not a representative sample.

“Don’t believe the mainstream polling”

On Twitter, a pro-Leave campaigner wrote:

Don’t believe the mainstream polling on the EU referendum. Go to any town centre and do your own. [Brexit]

A survey conducted by a man wearing a green ‘Grassroots OUT’ jacket, where people self-select in a town centre, is not equivalent to a survey conducted through random selection: on either an online panel or computer-assisted telephone interviews.

“Ignore mainstream polls”

Alternately, people point towards open-access polls, which suffer the same problem of self-selection [3].

UKIP Bournemouth decided to reveal their terrific insights into polling:

Ignore mainstream polls (IPSOS/MORI, Gallup, YouGov, etc.); their methodology is seriously flawed, and the EU referendum is little more than an exercise for them to see if they can turn around their results after failing disastrously at the last election.

Have a look at the electronic polls online.

Even the lefty Mirror poll had 67% leave.

A neutral one I saw has over 70% leave.

The Express, Telegraph, Mail and Times polls all had over 70% leave.

Look at comments on forums. Incredibly, even the lefty, middle class papers’ comments are predominantly leave: Guardian, Independent, Observer.

The BBC forums even more so.

As for the right-wing papers, they are just the same as this one.

Ultimately, your own empirical evidence will be in stark contrast to what the media is proclaiming, the mainstream polls are claiming, and what the insidious bias of BBC reportage is promoting.


Hold up everyone, UKIP Bournemouth have written one of the most stupid screeds I have ever read. (Source: UKIP Bournemouth/Facebook)

It is never explained exactly how the polling methodology of “mainstream polls” is “seriously flawed”, or why the “electronic polls online” are better.

People rushing to click a button on a website is not representative of the country as a whole.

Online open-access polls suffer from the difficulty of users attempting to perturb the final results in their preferred direction, through regular sharing on social media. Having an active, loud digital presence is not equivalent to actual votes in the polling booth.

Moreover, “your own empirical evidence” will be local, and is not necessarily indicative of the national reality.

If I had one wish for the betterment of this referendum, it would be for people to no longer pretend a self-selecting survey is somehow equivalent to scientific polling.


[1] Leard Statistics, 2013. Understanding Descriptive and Inferential Statistics. Available from: [Accessed: 3rd June 2016]

[2] Dahlgreen, W., 2016. The Eurosceptic map of Britain. Available from: [Accessed: 3rd June 2016]

[3] Masters, A., 2016. Voodoo polls. In Defence of Liberty. Available from: [Accessed: 3rd June 2016]



This entry was posted on June 14, 2016 by in European Politics and tagged , , .
%d bloggers like this: