Peer’s warning about ‘statistical junk’ needs to be checked for errors

 David Lipsey is a member of the same Constituency Labour Party as This Writer and is, in general, a very interesting gentleman. His discussion of Tory plans to gerrymander constituency boundaries was fascinating.

He has a point here, regarding the sampled 1,900 emails used to generate the claim that 75 per cent of respondents to its emailed poll were against bombing Syria.

However, I had this argument with Chris Bryant on Twitter a couple of evenings ago. He was referring to the people who had contacted him directly, begging him to oppose air strikes on Syria (he didn’t). He reckoned, as David Lipsey suggests, that opponents of bombing would be more likely to express their preference.

In the words of Lord Lipsey, there isn’t a scrap of evidence for such a proposition. In fact, it simply seems more likely that those with a strong opinion – either way – would respond. If I had supported air strikes (I didn’t) but I thought Labour was going to oppose, then I would certainly have written in to express my view that this would be a mistake – as, we are led to understand, at least 16,000 Labour members did. The actual numbers would certainly be welcome.

Therefore I argued that Chris should have heeded the will of his electors in the case of this important vote.

I am unconvinced by Lipsey’s queries over representation, as I fail to see what difference it makes to know how many Labour members aged, say, 25-34 voted for air strikes, or how many with surnames beginning with ‘M’ were against the bombings. Margin for error would likewise be unnecessary in a result that states simply that a proportion of the number of people who responded voted in a particular way.

I could use the same argument as Lipsey to call our electoral system into question. Is the result drawn from a full alphabetical list of UK citizens? No it isn’t. Is it drawn from a list of those who are eligible to vote? No it isn’t. Is it drawn from a list of those who were eligible to vote and who then expressed an opinion by exercising their right to do so? Yes. What checks have been made to see if the result of the election was representative of the whole electorate, by age and region? None.

You see? The results of all our elections are based on samples of non-samples – which is perhaps why Labour did not win the majority the Party expected in May!

[Lord Lipsey] said: “If the party is to hold such consultations in future, as Corbyn has suggested it will, they need to be independently designed and verified by neutral and professional third parties. Otherwise they too, like this poll, will produce one thing only: statistical junk.”

In a press release on 30 November, Labour presented the outcome of a survey in which 75% of the 64,711 party members who responded were opposed to bombing.

In his note Lipsey commented: “I do not know if this is true or not. What I do know is that the consultation does not provide a shred of evidence for such a proposition.” He pointed out “there is no way of knowing if the 64,711 that responded were a representative sample of party membership. Indeed, prima facie, it seems likely that opponents of bombing would be more likely to reply.”

He added: “The poll was based on a sample of the non-sample, comprising some 1,900 members. The party claims that this sample was random. But nothing is said about how it was drawn. Was it done from a full alphabetical list of members, or just the first 1,900 read? What allowance has been made for the statistical margin of error on such samples? What checks have been made to see if it was representative of the whole membership, by age and region?”

Source: Corbyn must prevent ‘statistical junk’ as seen in Syria survey, warns peer | Politics | The Guardian

Join the Vox Political Facebook page.

If you have appreciated this article, don’t forget to share it using the buttons at the bottom of this page. Politics is about everybody – so let’s try to get everybody involved!

Vox Political needs your help!
If you want to support this site
but don’t want to give your money to advertisers)
you can make a one-off donation here:

Donate Button with Credit Cards

Buy Vox Political books so we can continue
fighting for the facts.

Health Warning: Government! is now available
in either print or eBook format here:

HWG PrintHWG eBook

The first collection, Strong Words and Hard Times,
is still available in either print or eBook format here:


2 thoughts on “Peer’s warning about ‘statistical junk’ needs to be checked for errors

  1. John Gaines.

    Cameroon’s –‘thing’– Gardy Loo Gideon, has been spewing Statistical BS from the self admitted unfit for purpose OBR since the Tory gangsters slid into 10 Dirt St.

Comments are closed.