ConHome, Political Betting and now David Blackburn at CoffeeHouse have all picked up on a paragraph from Jackie Ashley’s column this morning which says “Some Labour people may think I’m sounding too gloomy, but those who have been privy to recent private polling are a lot more than gloomy. This suggests that Labour could return to the Commons with just 120 MPs or thereabouts”.

Rumours of private polling from the political parties are often given far more credence than they deserve, as if it’s some special secret knowledge that trumps the stuff that is fully available for us to pore over.

Private polling isn’t more accurate, pollsters are not like NHS consultants. They don’t do cheap bog-standard polls for the papers, then bugger off down the road to plush, oak-panelled private polling offices to do much better polls for private clients. The voting intention polls that are carried out for the papers are the best that the pollsters can do and if some of the pollsters have private polls from the main UK pollsters, their voting intentions will in all likelihood be carried out in the same way with the same results (in fact, on principle I’d advise due scepticism of anyone showing anything wildly different).

For Labour to be reduced to only 120 seats on a uniform swing would require a Conservative lead of around about 28 points. With the main pollsters all showing leads in the mid-teens, I would treat any poll showing a Conservative lead of 28 points with extreme scepticism.

Of course, what political parties can do as part of their private polling is commission polls in specific seats, or specific groups of seats. It’s just possible that is what is what is behind Jackie Ashley’s claim. For Labour to actually be reduced to 120 seats, we’d have to see them lose seats like Vauxhall, Hackney North & Stoke Newington, Scunthorpe, Worsley & Eccles South, Alyn & Deeside, Stoke on Trent South – that sort of place, as opposed to Labour’s absolute safe areas like former mining villages and Northern inner cities. Perhaps Labour have done some private polling in that sort of area and seen horrid results. More likely, some minister has just said something off the cuff to demonstrate how horrible their polling position is, and it doesn’t reflect some great poll finding at all.

One of the issues at the time the British Polling Council was set up was political parties making outlandish claims about their private polling showing something different from the published polls. When Conservative morale flagged Lord Saatchi would wheel out some “private polling” allegedly showing how well the Conservatives were doing. Whether they actually showed that, or were slanting reporting or interpretation the public and media couldn’t tell, since the tables and questions were not made public.

These days under BPC rules if parties release stuff like this, they are obliged to release the tables that the research is based upon and observers can see for themselves what they really say, and whether the parties are trying to hoodwink them. In practice though, it doesn’t work like that, the only difference it makes is that when they used to mention figures, now parties just say that their private polling “shows” something, without mentioning any actual figures or questions that interested parties could demand the release of under BPC disclosure rules. Hence we are even less able to tell if they mean anything at all.

My advice would be to ignore what political parties claim, or tell friendly journalists, their private polling says unless there are actual figures (and consequently tables) to check.


15 Responses to “The mystique of private polling”

  1. Anthony – what is the smallest reasonable sample size for a seat-specific poll? I’m assuming that it is a lot less than 1,000?

  2. Tam –

    A rather counter-intuitive thing about polling is that above a certain figure the size of the population you are polling doesn’t make much difference.

    So if you are polling Great Britain, with an electorate of 40,000,000 or so, a poll of 1000 will give you a margin of error of 3.1%

    If you were polling the USA, with it’s much larger population, a poll of 1000 would still give you a margin of error of 3.1%

    If you were polling Glasgow North East, with only 60,000 or so people, a poll of 1000 would give you are margin of error of an almost identical 3.07%

    Obviously if you are polling a population of only 1000 people in total, it would make a difference, but once you get above 40,000 or so the size of the population you are polling really doesn’t make much difference at all to the sample size you need.

    So – for a seat specific poll, if you wanted figures as robust as in a national poll you’s still need 1000 people. Of course, that doesn’t mean that parties might not find smaller, less robust polls of use.

  3. HOW VERY INTERESTING. I READ MS ASHLEYS ARTICLE AND DASHED STRAIGHT HERE, HOPING YOU WOULD CONFIRM IT AS GOSPEL. NEVER MIND, ALL VERY GOOD STUFF.
    BY THE WAY WHEN IS IPSOS MORI COMING OUT?

  4. I guess this could be on the basis that:

    Conservatives are 14% ahead
    They normally do 3- 5 % better than the opinion polls
    Cs are likely to be doing better in marginal seats (labour vote can’t be -ve in the home counties and Cs still won’t get many votes in the inner cities)

    If you add this together the effective lead could be in the 20s. If you then add in concerns about L turn out and their continued drift in the polls – you could get to a worst case of a 28% deficit. This is what they might be scaring themselves with.

    But surely….

  5. Does that mean that politicians, who are managing the economy, arn’t very good with statistics?

  6. @ JOHN B Dick

    Or that Journalists reporting on it aren;’t either. Or the Tories gloating over these figures over on ConHome aren’t either!

  7. The old 10DM note celebrated Carl Freidrich Gauss and I kept one in my wallet for years in case I needed to explain the basics to a politician.

  8. Doesn’t Labour always play this trick before an election – downplay expectations so when the result comes in slightly better they can say how well they’ve done?

  9. Anthony , the Yougov website gives data tables for a poll on Afghanistan for Sky News which you don’t appear to have covered . It was taken immediately following the Channel 4 poll – 5th/6th November but the questions are phrased slightly differently . The sample size was virtually the same as the Channel 4 poll . There is also a GE voting question but no published voting prediction . By comparison with Channel 4’s figures the voting intention would have been Con 40 Lab 26 LibDem 19 .

  10. One wonders, with the possible exception of the Tories, how much money the parties currently have to commission private polls, except perhaps for critical by-elections.

    It should be said that several of the seats Anthony mentions are actually ones which I and others have already discussed on this site as likely to swing more than the average against Labour, e.g. Scunthorpe and Stoke on Trent South. Reasons includethe high number of (male particularly?) working class voters who may be deserting Labour in particular, and also the concentration of such seats in regions such as the West Midlands which may be swinging particularly hard against Labour, as Anthony’s excellent large-scale poll suggested recently, and for understandable reasons, e.g. particular losses of manufacturing industry in such seats.

    We will have to see what happens in seats such as Scunthorpe that are particulalry caught up in expenses problems.

    London seats have historically tended to swing more than average in whichever ay the political wind is blowing, which may affect Vauxhall and Hackney North and South Newington. And London seats can be affected rapidly by demographic change. I don’t know about the specific cirucmstances of these two seats, but it is

  11. One wonders, with the possible exception of the Tories, how much money the parties currently have to commission private polls, except perhaps for critical by-elections.

    It should be said that several of the seats Anthony mentions are actually ones which I and others have already discussed on this site as likely to swing more than the average against Labour, e.g. Scunthorpe and Stoke on Trent South. Reasons include the high number of (male particularly?) working class voters who may be deserting Labour in large numbers, and also the concentration of such seats in regions such as the West Midlands which may be swinging particularly hard against Labour, as Anthony’s excellent large-scale poll suggested recentl. There are understandable reasons which can be used to explain such voting behaviour plausibly, e.g. particular losses of manufacturing industry in such seats.

    We will have to see what happens in seats such as Scunthorpe that are particularly caught up in expenses problems – there is a lack of precedent from a psephologist’s point of view.

    London seats have historically tended to swing more than average in whichever ay the political wind is blowing, which may affect Vauxhall and Hackney North and South Newington. And London seats can be affected rapidly by demographic change. I don’t know about the specific circumstances of these two seats, but it is quite possible they are in jeopardy. Labour could have problems in Vauxhall over the Olympic Games.

    Actually, there are seats even in the areas Anthony describes as “absolutely safe” for Labour about which people posting on this site are not totally certain. These include some mining seats, such as Ashfield and possibly even some Yorkshire seats (see e.g. discussion on the Hemsworth thread), although it is more likely that Labour candidates in these seats will be returned with, for them, shockingly reduced majorities.

    Further, it is difficult to disentangle from the opinion polls how far the anti-Labour swing will go to the LibDems in “safe” Northern seats, for instance Newcastle Upon Tyne North, where the Tories are nowhere. If the anti-Labour vote does get its collective act together, Labour does stand to lose a number of Northern English city seats – indeed the LibDems may well get win such seats whilst losing in notionally much better prospects such as three or four way Scottish marginals. There has been relevant discussion on this site on the Liverpool Wavertree thread.

    Like Anthony, I don’t see a national Tory lead over Labour of 28% when the election comes as at all likely. But there are a number of seats where Labour were 25-30% ahead of the next party in 2005 which, as indicated in my discussion above, look vulnerable for particular reasons, and it is not so easy to identify which these are. I think it not unreasonable to suggest that there are only 120 Labour seats, perhaps even less, where the Government is for all practical purposes certain to retain the seat.

    To come round full circle, one reason for this state of affairs may be that the Tories currently have much better resources than Labour to gain intelligencewith which to target their campaigns.

    P.S. for Anthony, I accidentally posted half this contribution, which has come up “awaiting moderation”. You may care to tidy things up.

  12. Labour could have problems in Vauxhall over the Olympic Games.

    Why would they? Kate Hoey (is inexplicably) liked here. Polling seems to suggest that the swing has been lower in London than in other parts of the country. I think! Maybe somebody could clear this up?

  13. I’d say Scunthorpe’s quite likely to go since Mr Morley’s due before the beak.

    Worsley & Eccles South also seems to be likely to leave the loving embrace of the party.

    Stoke on Trent South could see Labour’s vote being sappered by the P N B, facilitating a Tory smash n grab.

    I don’t see the sense in assuming that the swing will be largely monopolar and very 20th century; I think it’s up for grabs.

  14. Sadly many journalists are statistically illiterate. But note what the DG of the IoD says: “The whole conference, even after Brown’s speech, was very subdued. Everyone seems to be expecting defeat at the coming election, with no real belief that anything can be done about it.” Bit like Trafalgar really.

  15. I am amused to see different opinion pollsters being compared to consultations with a hospital consultant. You imply that you are getting a different or better opinion if you pay for it. This is, I assure you, not the case. A consultant will give you the same opinion wherever you see them, in public or in private.

    The trick in the public sector (NHS) is to actually see the consultant. If you don’t then your simile may stand up.