Over on the right hand side of this site is a projection of how the current polls would translate into seats at a general election tomorrow, if there was a uniform swing. On twitter and suchlike I sometimes see if referred to as UKPR’s current prediction, but I’m afraid it isn’t. Polls don’t predict the next election, they measure support now, so the polling average here isn’t my best guess for the shares of the vote at the next election, it’s a measure of support in an election tomorrow. Of course, there isn’t an election tomorrow, and if there was, the polls probably wouldn’t be as they are – if there really was an election tomorrow then the last three weeks would have been full of manifestos, policy announces, campaigns and debates which may or may not have had some impact.

It’s also worth noting that while uniform national swing is not a bad guide by any means, it can certainly be bettered. To start with it’s definitely worth dealing with Scotland seperately based on Scottish polling figures, it might also be useful to include some assumptions about incumbency effects in seats with new MPs, and some degree of random variation at the margins.

I deliberately don’t make predictions this far out, given the huge amounts of unknowns. I tend to find most people who do predict this far out with any degree of confidence are – probably unconsciously – merely predicting what they would like to be the case. It’s rare to find someone confidently predicting a Labour victory who wouldn’t like a Labour victory (or who has an ideological axe to grind against the Tory leadership), or vice-versa on the Conservative side. Given the prominence of Nate Silver and other election prediction sites at the last US election I would expect a plethora of more academic and sensible election prediction models come the actual election (hell, I know for certain of several groups of academics working on various models), but so far virtually the only prediction I have seen that moves beyond wish-fulfillment to actually come up with a poll-based model is the attempt by Steve Fisher at Oxford here, with an explanation of the model here.

Steve’s model is a simple one – it is purely based upon voting intention polls and how they have tended to relate to the election result that follows*. We cannot assume that the polls will remain unchanged in the run up to the next election, given that in past Parliaments they have tended to change. Past change has not been a random walk, with equal likelihood of government’s gaining or losing in the polls – this is the key to Steve’s model. In the past the polls have rended to regress towards the result of the previous election (usually in the form of the government recovering). What Steve has done therefore is to take the current polls, and then factor in the sort of size and scale of changes that have typically happened to the polls over the last years of previous Parliaments, then based a prediction on that. At past elections this would have proven to be a more accurate predictor than just taking the current polls. That is not to say that that it is a particularly accurate prediction, only that in the past it would have been more accurate than assuming no change.

On that basis, if the polls over the next year behave like the polls in the last year of previous Parliaments the most likely result come the general election is a Conservative lead of 5 points over Labour, which would produce a hung Parliament with the Conservatives the largest party. The most important word in that sentence is probably the “if”, and perhaps the most important thing to note in Steve’s projection are the large prediction intervals around it. Steve’s model predicts the Conservative vote will be 37%, plus or minus 8.5 (so between 29% and 46%), the Labour vote at 32%, plus or minus 6.4 (so between 26 and 39). These are huge gaps. Of course, results towards the centre of those ranges are still considered more likely, but it underlines the imprecision of the projection, and the limitations on using current polling data to predict a general election a year away. Polls a year out from the election are not a very good prediction of the election. It would be wrong to say that anything could happen (Steve’s model, for example, suggests it is unlikely that Labour would get over 40, or that the Conservatives would fall below 29), but certainly a lot of different outcomes could happen.

It also reflects the sheer variety of elections. One criticism I’ve seen of Steve’s model is that this election will be different because of the coalition, the UKIP factor and the realignment of the Lib Dem vote. That may very well be true, but we could say the same about other elections – 1964 had two late changes of leader, 1966 wasn’t a whole term, 1974 was different because the Liberals started contesting all seats, 1979 was different because of the Lib-Lab pact, or the winter of discontent, 1983 was different because of the Falklands and the SDP split, 1992 was different because of Thatcher’s removal, 1997 was different because of the sheer scale of the landslide. 2001 was different because Labour never really had any mid-term blues to come back from. The infrequency of elections means that almost by definition each one has things that make it unique and different – yet Steve’s out-of-sample predictions shows the model would been a better tool at predicting those past elections from 20, 12 or 6 months out than just looking at what the polls 20, 12 or 6 months out were saying (it also underlines the difficulty for political scientists in coming up with any decent models at all – you only get 16 data points and they are all weird).

That doesn’t mean it would have been a particularly good prediction at those past points, just that it was better than the alternative of just looking at the polls 20, 12 or 6 months out. The polls now are a snapshot of public support now, they are not a prediction of what will happen in May 2015. If polls move in the sort of way they have in the run up to past elections we can expect the Conservatives to significantly recover. If they don’t, then they won’t, simple as that. Polls do not move by magic, drawn towards past election results by some invisible force. If they narrow, it will be because of the economy, because of changing attitudes to the parties, because, perhaps, of different factors weighing upon people’s political choices as an election becomes more imminent… that, however, is a post for another day.

(*I should also add that this is NOT Steve’s personal prediction of the election – it’s an attempt to see to what degree you can predict election results months in advance using just national poll data. I expect if Steve was making a personal prediction he probably would ponder what the impact of the economy, the party situation etc would be, but that would be a very different and more subjective model.)


253 Responses to “This is not a prediction”

1 2 3 6
  1. My far more scientific calculation produces a healthy Lab majority every time.

  2. You should submit a working paper Nick ;)

  3. Brilliant Anthony.

    Thank you.

  4. Fascinating post. Of course, despite the headline, it will be reported as predicting a 5% Con lead in 2015. That is a prediction.

    However, one factor that jumps out at me in very large, flashing red lights, is the change in election timing.

    Steve Fisher’s work is based on elections where the election date was, by and large, chosen by the incumbent. Based on the fact that incumbent turkeys will not vote for Christmas, we can be pretty certain that the timing of all previous elections were selected to the perceived advantage of incumbents. We are now in an election where the date is set by the calender, so assuming anything from the past might happen again introduces a further source of potential statistical incoherence.

    This doesn’t mean the same thing won’t happen – it may be that a lengthy 5 year parliament is, indeed, the most electorally beneficial option for the Tories. However, it’s also true to say that the inability to call a snap election at the most appropriate time will mean that it’s less likely that Tory support can be maximised.

    While I could foresee Tories getting 37% if everything goes well for them over the next 13 months, although technically perfectly possible, I think most people would intuitively struggle to honestly accept a forecast that Labour would only add 2% from a very poor 2010 result, thus questioning one of the central findings of this study. But it’s highly interesting nonetheless.

  5. Are not some of the pollsters weighting and reallocating according to previous voting patterns and therefore attempting to make a prediction for the General Election?

  6. The other huge problem with Fisher’s model, which I complained about when he first came out with it and which I still think is a fatal flaw, is it assumes there has been no change in the accuracy of polling since 1992.

    We already know this is wrong, and it’s going to bias any prediction he makes in favour of the Tories. Even if you set aside all the issues with the left recombining and the right splitting, he’s coming out of the gate with a big handicap.

    None of this means he should resign if Labour wins the election- academics come up with terrible models all the time; that’s the whole basis of the scientific method – but it does cast serious doubts on the predictive value of his model, or any other model that relies on raw polling data from pre-1992 elections.

  7. Alec – well, I meant more that my poll average and projection isn’t a prediction. Steve’s model does make a prediction (it’s not Steve’s prediction per se, but it is what his model predicts).

    Mr Beeswax – weighting doesn’t imply a prediction at all. Reallocating don’t knows does to some extent, it does cross that line from what respondents say, to the pollster predicting what they would actually do. Strictly speaking though, the pollster is still predicting what “don’t knows” would do in an election tomorrow, not making any assumptions about how those who do express an opinion might change their opinion in the months to come.

  8. Anthony, I feel you were being a little disingenuous in your description of the variety of elections. Most of your comments were about ordinary political events, not about ways in which the rules of the game changed systematically. 1983 and the SDP split is really the only one that saw a rule change on a par with what we see this year (coalition government, UKIP, fixed terms).

  9. Spearmint – I was going to mention that and forgot. It’s wrong to say he assumes no improvement since 1992 (that would imply assuming a 1992 level of error for the coming election) but he does factor in the average level of polling inaccuracy, which is obviously less than 1992 once you spread it over all elections (essentially, he has the final polls underestimating the Conservatives by a mean of 0.7%, overestimating Labour by 1.1%)

    I understand why he’s done that – it would be a bit presumptious to assume the long standing historical overestimation of Labour in polls has gone away just because it wasn’t there in one election 2010. As a pollster though, I’m rather more optimistic that the problem has been dealt with, so if I was doing the same as him I don’t think I’d include it.

    Presumably (and assuming a very crude back of a fag packet application, which I expect is wrong) without that factor his model would be predicting a Conservative lead of about 3 points.

  10. As someone working in the social sciences, I think it’s great that Fisher has produced this model and is using it and talking about it because the best way to get better models is to publish the findings of the ones you have, so that you can get feedback and improve them.

    This is all extremely interesting.

  11. It would be more accurate to predict the weather in 3 months time. I read that there is a 75% chance of a really hot summer. I also remember reading that 2012 would be a BBQ summer and it was pretty miserable, apart from when the Olympics was on.

    So much can change between now and May 2015. In 2010 we did not have UKIP on near to 10% and the Lib Dems also near 10%. UKIP will be taking votes off Labour and Tories. Lib Dems could well do better than predicted, but it would seem that one third of their 2010 votes could go to Labour.

    If I were to guess an outcome in 2015, it would be Labour between 33-36%, Lib Dems 14-17%, Tories 33-36% and UKIP 8-11%. I don’t see the Tories winning by a large enough percentage to win most seats. If the Tories could not win a majority in 2010 against Labour who had been in power for 13 years, then I find it difficult to think what would have changed voters minds.

  12. Interesting model, but if I had one I would not bet my tenure on its prediction. Now let’s look at the Beach Model.

    Sometime year ago (during the rise of the LD in Parliament) I read an argument concerning the relative stability of a two party system as compared to a three party system. Using the example of beach deckchair concessions competing on a single beach, the argument was that the third operator destabilised the system by moving his chairs a little to the left or right from the middle of beach there by squeezing the party on the right and opening up the beach to the party on the left or the other way round. Now with a fourth operator on the beach with his chairs firmly on the right, and the middle operator committed to the right side, the center right ground in under real pressure and while the left of beach in under one operator who can not lose in terms of sales or votes.
    What do you think guys?

  13. “If the Tories could not win a majority in 2010 against Labour who had been in power for 13 years, then I find it difficult to think what would have changed voters minds.”

    I agree with this & if I was to make any kind of prediction, I would say there will be a small Labour majority.

    Time will tell.

  14. “It’s rare to find someone confidently predicting a Labour victory who wouldn’t like a Labour victory”

    I’m not confident to put any money on it, so I can hardly be that confident, but I think that (a) a Labour majority is the most likely outcome and (b) a Labour majority is just about the worst of the many bad things that could probably happen in the 2015 election.

    My reasons are (1) that Labour aren’t doing badly in the polls, (2) that the current system and boundaries mean that Labour only have to avoid a total disaster in votes in order to win, (3) it’s very hard for a government to win during a time of austerity, and even economic recoveries don’t benefit governments as they used to, and (4) UKIP has split a lot of the anti-Labour vote, though how much of a difference this will make in Con-Lab marginal seats is yet to be seen.

    Obviously, this is just an extrapolation of some selected past trends, with no major causal architecture, so I wouldn’t bet a penny on it. Also, political betting seems to be a fairly efficient market, so the information is likely already in the odds (and I have no insider information, e.g. that David Cameron is a necromancer or Ed Miliband eats dolphin steaks for breakfast).

  15. Today’s Populus, looking a bit more normal but with Labour still on the low side:

    New Populus VI: Lab 36 (+1); Cons 32 (-2); LD 10 (=); UKIP 13 (=); Oth 9 (+1) Tables http://popu.lu/s_vi140317

  16. @Alec

    I agree that the likelihood of a five year parliament upsets the model.

    Fisher’s model, in my view, confuses cause and effect. When, at the start of the fourth year of a parliament, there’s been a recovery in the polls for the governing party, then the governing party often chooses to hold an election, in the knowledge that it might be risky to go the full term. And that selectively generates data for his model which of course points to a recovery for the governing party.

    But in our case, we’re considering a situation where a parliament goes as long as possible. The Fixed Term Parliaments Act has influenced that, but I nonetheless feel that if the Conservatives had recovered to above 40% over the past year, then they would now be looking to end the coalition and government and prompt a fresh election by frustrating the ability of any alternative government to carry a vote of confidence.

  17. @Hamish – That phenomenon is known as Hotelling’s Law, named for the great economic statistician Harold Hotelling.

  18. @ Chris Riley

    I’m not so sure. I think it is one of those things that can’t be predicted statistically (even though we have all had our opinions) and as such I don’t see any way that the model can be improved upon because there are a series of one off events that cannot be entered into a spreadsheet.

    Let’s say LD vote does not recover and the model is adjusted. It doesn’t mean it will be the same next time- new leader etc. Let’s say the UKIP vote holds and that too gets adjusted for a new 4 party system. Next time around you may find the UKIP vote does not hold because ex Tories faced with another term of Labour government then go back to the fold late on.

    It’s very different from how polling organisations have refined their weighting/Don’t know treatment etc over the years which seems to be valid, although I feel there is a big uncertainty over the LD don’t knows this time around.

    I think it would have more relevance in a standard parliament but this is a particularly unusual one with LD’s seen in government and UKIP rise. Other than the SDP split from Labour we have never before had a similar situation.

  19. I won’t rehearse all the arguments as to why the Fisher analysis might be flawed, and I’m certainly not one of those who have ruled out the possibility of the Tories winning the next election. There are only two parties who can win in 2015, and the Tories are one of them. Both horses in a two horse race usually have a chance despite one of them being the favourite!

    The one thing that I’d question is whether Fisher has paid enough attention to the rather unusual way that the opinion polls have behaved in this Parliament. I’m always nervous about historical precedent being trotted out as a reliable predictor of the future, and that’s why I’m ignoring the “no opposition that has enjoyed a consistent and enduring polling lead for over two years has gone on to lose the subsequent general election”, but has Fisher factored this into his model?

    There are so many different political circumstances at play now that I’m highly sceptical of the Fisher model, especially one with a + or – 8% margin of error!!

    Would we predict this year’s Cup Final result on the basis of how Arsenal got on against Hull City in 1935?

  20. @Phil Haines – “Fisher’s model, in my view, confuses cause and effect.”

    Very neatly put.

  21. @Shevii

    Social phenomena that involve a lot of human choice are probably not properly *predictable* but they can be *modelled*.

    In this case, Fisher recognises that his model is relatively rudimentary as it has a giant MOE that means it is not particularly useful as a predictor. If, however, he or someone else can get a model that has a much narrower MoE and, crucially, is both testable and passes those tests, well, then, now you’re talking.

    It would be very tough to manage, but I don’t believe it impossible.

  22. This site just gets better and better.

    I welcome the discussion of Steve Fisher’s paper even though, from what I understand of it, I don’t agree with it.

    The point was made previously that parties even die out sometimes, and Labour did displace the Liberals for a very long time. So quite apart from anything else there is no automatic bounce-back for a party. Perhaps this point is hidden because the Conservatives/Tories have been in existence since around 1700.

    I have been one of those predicting a Labour victory with a small but workable majority, based on the polls.

    However I think if the Martians were to invade the UK (though if they were landing at random they might be more likely to come across a large land-mass like Ukr- oh never mind) I think they would come to the conclusion that although the parties have different origins and traditions, their actual decisions in government do not vary all that much.

    Perhaps I will be proved wrong this time.

  23. Now with a fourth operator on the beach with his chairs firmly on the right, and the middle operator committed to the right side, the center right ground in under real pressure and while the left of beach in under one operator who can not lose in terms of sales or votes.
    What do you think guys?

    -Two 99’s and a Raspberry Ripple Please.

    I favour the Bookies (put you money where you mouth is) model which have a Labour majority Government as the most likely outcome

    We will know come 8th May 2015

  24. Psychohistory seems to be the great dream of the social sciences. Unfortunately I suspect that it is both in practice and in principle impossible.

    Not that what Fisher is doing is necessarily wrong, and there’s no harm in trying. But I suspect that election forecasting is like weather forecasting – the closer you get to the day the more accurate the prediction can be, but anything further out than a week (or likely a month or two in election terms) and its just going to be largely guesswork.

    That and the other issue is the prediction only holds if there is an assumption of relative stasis, which may not be the case (and seems like rather a large one).

  25. @Alec
    Agree that fixed term must change the poll/prediction calculation. The government can make a better go of choosing the election date to suit their poll rating than they can manipulate their popularity to peak on a specific date.

    As you may remember though, Callaghan didn’t choose his election date, it was thrust upon him when he lost a vote of no confidence. I’d also class governments that ran to the buffers as not quite “chosen”; usually these governments were waiting for something to turn up, rather than selecting 5 years as the best time to run – in effect they ran out of time.

    There are, unfrtunately, very few data points for “unchosen” elections, so it would hardly be worth analysing them.

  26. @Alec – there’s elections like 1987 and 2001 (arguably) that were chosen by the incumbent – it was more a case of getting it out of the way, the opposition being in such a state it probably mattered little when the election was held.

    I’d probably rule out elections beyond 30 years ago, maybe quite a bit sooner. The general situation of the country is simply too different to be applicable – for instance demographics, where the marginal seats are, who did what last they were in power, etc. Too much has happened.

    Agreed, looking for an election enough like the upcoming one is going to be tough. Take a sort of average of a bunch of them? No guarantee that will be better. Anyway, by the time the predictions get accurate enough, the election will be on us and the picture rather clearer.

    Who knows what effect the election campaign will have. Some of the time, nothing much happens, other times we get Cleggmania or 1992.

  27. @Postageincluded

    It’s worth remembering that if this was a typical post-war Parliament, then the general election campaign would be starting in about three weeks time!

  28. I hope after the election is over, we can consign fisher’s “model” to the dustbin of history.

    Nate silver was excellent in the US…he didn’t rely on polls from a generation ago to work out likely outcomes. He used “current data”.

  29. Has anyone looked at using individual constituencies for predictions of overall national vote share, to try and get more than 16 data points? You’d have to add extra parameters to the model to describe the quality of the constituency somehow (main party vote shares and turnout at the previous election might do it), but you’d have about 9000 data points for England so overfitting hopefully would not be a concern.

  30. Mr. Fisher was never going to be received with garlands here -well not with a 32% figure for Labour anyway.

    That says the LD defectors will go home ( though the interaction between Lab & LD isn’t apparent in Fisher)

    Not at all sure I believe that either.

    I can see 37% for Con-or even a tad more with a following wind-but 32% for Lab????

  31. Sam – not based on constituency polling, for obvious reasons, but one of the other forthcoming academic models I alluded to in the post is by Nick Vivyan and Chris Hanretty, and will be one built on the demographics and characteristics of individual seats. Watch this space

  32. There are far too many variables to attempt an election result, for instance.
    Our economy may well be growing but so too is our borrowing.
    The growth in house prices bubble has the potential to unwind
    The reaction to Europe and the USAs sanctions could well result in Russia reducing its gas supply or raising the price which will have a serious affect on the EU including the UK.
    We are going into the next election with 4 party participation, all of which makes the outcome far less predictable. The UKIP factor has already forced issues like referendums and immigration into the wider debate and these are factors that the major parties are struggling with. This to me is the most unpredictable election I can remember in the last 50 years.
    One of my reasons for saying this is that with the unprecedented growth of internet use and the accesable information available none of the political parties can now limmit or hide facts like they could in the past thus undermining their credibility.

  33. No way fully half the 2010 Lib Dem defectors will go back next year. A couple of percentage points maybe (which incidentally will probably help Labour since the most likely to return are probably the ABT tactical voters) but 32% is suspiciously low.

    I could buy 34% if Labour do particularly badly during the campaign or the Tories do very well before then. I’m of the opinion that Miliband would do rather well out of TV debates, so that scenario could happen if they aren’t held.

    We’ll see.

  34. Peter – I suspect Nate Silver constructed his models using analysis of past US elections and then put current inputs into them. That’s sort of how these things work (indeed, one of the factors in his model was indeed the tendency of historical races to narrow or widen depending on circumstance). Obviously this is a far simpler model – predicting one race, on just two variables (current polls and time till election), but the principles (analyse past results, construct model, test against past results, put in current data, predict) is just the same.

  35. I wouldn’t say Fisher’s model is complete bunkum, and he’s obviously a serious psephologist attempting to construct a scientific model, but the flaws, already highlighted in this discussion thread, are many and obvious.

    My analysis is much less scientific and more intuitive. The current political situation at this stage in the electoral just feels very different to anything I’ve experienced in my 40 or so years of election-watching. The way the opinion polls have behaved and are continuing to behave, what’s happened in the by-elections held to date, the emergence and durability of UKIP, the Lib Dem collapse etc etc.

    There’s very definitely something different in the political air and I think it may be a little delusional to expect it all to fall back into a comfortably predictable place over the next 13 months.

  36. @Chris

    Goodness, and here was me thinking it was named after that great railway reformer, Dr Beaching

  37. Anthony – As much as I like your poll of polls, there is one aspect of it that I would question… Because there is no adjustment for house effects (vs the average or vs past results), it seems like in the short run it could be sensitive to which polster conducted the most recent poll. I realise you weight down polsters with poor track records but have you considered adjusting each poll for its house effect? Looking at the two posts you

    As for the Fisher model, it looks very solid from a statistical point of view. I don’t think the fixed-term parliaments matter all that much, it’s more the principle of mean-reversion. The objections based on structural changes (eg UKIP) are much more important in my mind and clearly very hotly debated!

  38. Anthony – As much as I like your poll of polls, there is one aspect of it that I would question… Because there is no adjustment for house effects (vs the average or vs past results), it seems like in the short run it could be sensitive to which polster conducted the most recent poll. I realise you weight down polsters with poor track records but have you considered adjusting each poll for its house effect? Looking at the two posts you did on the subject, it looks like they are relatively stable over time.

    As for the Fisher model, it looks very solid from a statistical point of view. I don’t think the fixed-term parliaments matter all that much, it’s more the principle of mean-reversion. The objections based on structural changes (eg UKIP) are much more important in my mind and clearly very hotly debated!

  39. Anthony,

    I love this website, but one weakness is the total disengagement with betting probabilities, as expressed on betfair and other betting websites.

    The market is pretty indifferent as to who will win. We should assume that its views on outcomes are therefore not as biased as you suggest other pundits’ views are.

    By contrast, Nate Silver himself made bets and used the market to back up his own views. As someone else, in the States, has said, “a bet is a tax on b*llsh*t”. On that basis Fisher is either a genius, and the market is wrong, OR, more likely in my view, he is peddling BS.

    Your contention that

    “most people who do predict this far out with any degree of confidence are – probably unconsciously – merely predicting what they would like to be the case”

    doesn’t, or shouldn’t, really apply to matched betting sites betfair and, still less, to paddy power, unless you suppose that the bookies themselves are biased and their principal object is not to make money, but to support particular parties.

    Punters are making predictions on outcomes all the time, as they did in the US elections. They can of course be very wrong one year out, and yes markets can move, but I don’t recall significant moves ahead of the US presidential elections in ’08, and ’12, nor really in the UK elections in ’10. The market decided that the Tories would be part of the next government in 2008, though views about the scale and nature of labour’s defeat changed.

    Similarly from May 2011, the betting markets assumed Obama would win in November 2012.

    Smithson’s political betting site of course does all this admirably, but i think we could do more on this website to take account of market signals.

  40. crossbat11
    I quite agree and it will be very interesting to see the results of the EU elections and the local elections that take place at the same time. These may well give an indication on the importance of both the EU membership and immigration in assesing the public mood.

  41. Whilst there is a flaw with Fisher’s method when it comes to fixed term parliaments, I would think that on this occasion the fixed term will profit the Tories. What I think is the greater flaw is with his presumption that the right splitting and the left recombining will not affect polling any more than the Falklands, the Winter of Discontent, etc., etc.
    I agree with Crossbat but what I don’t know is whether the different political ‘air’ will benefit or disadvantage the Tories.

  42. Number Cruncher – it doesn’t, no (for various reasons – pollsters will all change methods to some degree after an election (or circumstances and their interrelation with polling results will change – a method that can produce a low Lib Dem undersome circumstances will produce a high score under other circumstances).

    The averages that Will Jennings, Rob Ford and Mark Pickup produce here do, however, adjust for house effects if that’s what you are after (the main difference, as you’d expect, is that they downgrade current Lib Dem support based on the overestimate of Lib Dem support in the polls in 2010.

  43. I tweeted Steve Fisher to ask why he thought the probability implied by the betting markets was so different to his own calculation.

    He said “I guess the bookies believe this time is different which is a common tendency”

    I call it BS. If he was truly certain of his own probabilistic determination he would now have a massive arbitrage advantage over the bookmakers and he would be betting his house (if he had one) on the outcome.

    He would be able to get on the Tories at 7/2 against and then bet on the other possible results including Labour majority as that price would inevitable drifts nearer to the day.

    It is clear that the punters, the ones who put their money where their mouth is, are not terribly convinced by Steve’s reasoning. The punters clearly do not believe, as Steve does, that the bookmakers have got these probabilities very wrong.

    If they did, you would be seeing those 7/2 odds being reduced substantially.

    I have just opened a Ladbrokes account as 9/4 against Labour in Bermondsey (v Simon Hughes) is massive value still.

  44. “If polls move in the sort of way they have in the run up to past elections we can expect the Conservatives to significantly recover. If they don’t, then they won’t, simple as that.”
    ————–

    Cliff notes of article: We dont know what the future holds :D

  45. Edit: “inevitably drift”

  46. AW
    I do hope you do not take Peter Crawford’s advice. It’s bad enough on here with CB11 on football and TOH on cricket (actually CB11 on both subjects). If you want to talk betting there is a site, as Peter informs us, and doubtless there are sites for other subjects too.

    I think the problem with models such as those of Fisher, is that they don’t actually contain any recognition of real factors in political developments, so can only survive, as a recognised fact, if their predictions are constantly achieved. It’s like flipping a coin, however, just because it came up five times heads, it says nothing about the sixth occasion. So it is, every time, including 2015. If it came ‘good’ in 2015, I would still not be convinced.

    One might as well go by the betting, as Peter recommends.

  47. Anthony – Thanks for your quick response, that site looks interesting but looks like it’s only updated monthly. Are you aware of any that adjust simply for house effects vs other polls (as you looked at in post 7744), rather than the vs last election? If not I’ll happily crunch (and share) the numbers myself…

  48. AW wrote: “Presumably (and assuming a very crude back of a fag packet application, which I expect is wrong) without that factor his model would be predicting a Conservative lead of about 3 points.”

    Which brings us slap bang back to where quite a number of us a couple of posts back were guestimating (rather than predicting I hasten to add!) that the result could plausibly be C36 L33 LD15 giving seats of circa L295 C290 LD35

  49. Basing house effect scores on a single election- especially an election like 2010 in which the Lib Dems gave all the pollsters so much trouble- is a recipe for disaster, I suspect.

    Nate Silver can do it because he has state Senate polls to work with, which give him 33 new data points for each pollster every four years. (This isn’t quite true because not every state is polled by every pollster, but the point is he’s working from a reasonably large data set.) He’s also working in a two party system, where house effects operate on a single axis instead of in four dimensions simultaneously.

    A British analyst trying to do the same calculation would have, at best, four data points to work with- the last general election, the Welsh and Scottish Parliament elections, and the European elections- and three of the four would be from elections in which voters behave very differently from general elections. I wouldn’t say it’s impossible to model, but it’s certainly far more difficult than it is in the United States, and I don’t have much confidence in the system Polling Observatory are using.

  50. The Conservatives haven’t led in any poll for more than two years, so Steve’s model predicts a Tory lead of 5% in the election? Steve’s model is broke.

1 2 3 6