A couple of people are getting flustered about the BNP trying to infiltrate the YouGov panel. Over on Political Betting Mike Smithson has rightly dismissed it, but I’ve now got the chapter and verse from Peter Kellner (see the bottom of this post).

Things like this come along occassionally – basically they could only work if YouGov didn’t manage or monitor its panel and took no measures to prevent it, when in reality, carrying out accurate online polls is all about carefully recruiting and managing a representative panel. Any attempt to infiltrate the panel to the extent that it could actually have an effect upon results would be easy to detect and very simple to counter.

You can do the maths yourself. YouGov has in the region of 250,000 on their panel, so to actually shift the result of an opinion poll by 1 point would require about 2,500 people. More to the point, it would require 2,500 joining in a way that several full time professional panel management people did not notice. A sudden surge of 2500 new members, who all joined through the open website (most people are recruited pro-actively) and who all answered questions in a particular way would be pretty bloody obvious. (In this case, secrecy probably wouldn’t be helped by them all using the same referral code and openly informing people of the “plot” on their website ;) )

All that aside, YouGov have reanalysed the poll, looking to see if recent recruits are any different to long standing members of the panel, and more specifically, were more likely to say they were backing the BNP. They weren’t and newer recruits made no difference to the result.

Anyway, here’s Peter’s take in full


1. There have been reports in parts of the blogosphere that certain YouGov members have been attempting to recruit BNP members to the YouGov panel, in order to influence the results of polls and generate revenue for the BNP. Here is YouGov’s response to these reports.

2. YouGov actively recruits the majority of our panel using a variety of techniques, although self-signup and referrals from other members are also possible. We constantly monitor the profile of new panel members, and track differences in survey results, to ensure that our panel is representative, and to protect the quality and integrity of our data. Moreover, YouGov’s sampling methods ensure that new members who sign themselves up cannot have a statistically significant impact on any YouGov polling results.

3. As a further test, YouGov has examined the results of the survey conducted after BBC Question Time poll. The survey, of 1314 electors, included 156 who had joined our panel since May 2009. This covers the period when, it is claimed, BNP bloggers advised party members to join our panel. Of these 156, just one respondent said they would vote BNP in a general election. Any attempts to inflitrate YouGov’s panel with the aim of increasing the BNP’s reported support have plainly failed. We are not surprised: the number and characteristics of people joining the panel since May have been no different from normal.

4. Nevertheless, to put the issue beyond doubt, and in line with our practice at the last general election, we had already started a “close”period, during which no new self-signups or member referrals to YouGov will be invited to take part in political polls. This “close” period started on September 1 and will last until after the election.

5. Any panel member who acts, or entices others to act, in a way that seeks to distort our data violates our rules. We apply various techniques to detect such actions and remove offenders from our panel. In practice the impact of this is statistically insignificant; but we consider it vital to take all possible steps to protect the quality and integrity of our data, and so maintain our record as Britain’s most accurate survey research agency.


9 Responses to “The BNP and YouGov entryism”

  1. I wonder why Yougov panel members weren’t told about this change to referral rules. A bit unfair if you are close to your £50.00 cheque (like me!)

  2. Anyone who thought this was possible was a bit of a dunce.

  3. It’s unfair. I joined since since Sept, and would not vote BNP in a million years. Don’t throw the baby out with the bathwater.

  4. I don’t think there is much of a change to referral rules – people can still refer people and get paid, just for the next 6 months or so those people won’t get any political surveys. 95% of surveys aren’t political, so it waon’t make a huge difference.

  5. Andy,

    They probably thought they could influence YouGov polls the way they would typical voodoo polls in newspapers etc. As you say, dunces. But then, somewhat representative of the intellectual profile that is attracted to BNP’s simplistic message.

  6. I don’t know where your getting the “95 % of surveys aren’t political”. I would say that this figure goes up significantly when we discount polls that don’t pay out any credits. But anyway i’m not that bothered, but would have been nice to be told.

    BTW AW, what are they couting as “political polls”, just the ones that ask who you would vote for, or the ones that ask “are you satisfied with how things are going in the world/your country/your neighbourhood type ones?

    x

  7. Jack – 95% is a very rough figure! I’ve no idea, I expect it’s ones that use a political sample (which really just leads to the question of which ones use a political sample. Matter of Peter’s judgement, but generally anything with party political stuff).

  8. Anthony,

    Sorry to correct you. However, even if 2,500 BNP supporters joined, they would still have to meet some criteria. For starters, not all 250,000 take part in a single survey (approx 1,000 do). Likewise, since your location and demographic details are given (presumably in weightings also) – the 2,500 would have to select the right demographics (like, women aged 65+, ethnic minorities, etc). Even if the BNP managed to get 2,500 members joining inputting in false data, they would all have to represent a cross-section of society. The easiest way to manipulate yougov would be to encourage current panel members to give up their position on the panel.

  9. Beans

    Surely the easiest way to manipulate yougov would be to identify all 250k current panel members and persuade a large number of them to respond in certain ways, even if that does not represent their true view. Question is, how easy is that compared to generally persuading joe public ?