A confession

Thanks to all, who gave the answers to the “Guess the election” competition. I’ll be doing the results this evening, so if you want to enter, you’ve got till 5pm.

But first, a confession.

The reason I did the competition was to make the point Anthony Wells makes  here, which is that given the polling situation, the only really accurate answer to the question “Who will win in 2015?” is “Reply Hazy, Try Again Later“. Except of course, that Anthony makes this point with far more detail and elegance than I can.1

Lacking Anthony’s precision, I thought I’d ask people to use past poll results to predict past elections, hoping to show that while there are identifiable trends, (such as governments recovering between mid-term and the General Election) they are only trends. Nothing is for certain in any specific election, because every election is different, and the events that precede each election are very different.

To make this point, one of the polls I included (not saying which one) was a recent poll from the last couple of months. Sorry. Couldn’t resist.

Thus far, the poll from the current parliament has been identified as a poll conducted a couple of years before the elections of 1959, 1964, 1979, 1983, 1987, 1992, 1997, 2005 and 2010.

That’s everything from a landslide government defeat to a landslide government re-election, with a hung parliament bang in the middle.

Why so varied? because two and a half years is a long, long time, and anything could happen, from boom to bust, from leadership coup to vote of confidence.

General Election 2015? Pretty much anything could happen.

How can a party make their desired outcome more likely? Now, that’s a much more interesting, and useful, question.

  1. He also puts it a bit less kindly: “With some honorable exceptions, I suspect in many cases people’s predictions this early say a lot more about their own personal preferences or what political axes they have to grind against their party leaderships than what is likely to happen at the next election.” []

7 Responses to “A confession”

  1. James Gray

    But surely the whole point (and fun) of the polls is to look at trends? The problem with your challenge (as interesting and enjoyable as it was to try and do) is that they were random, one off polls. No trends were evident, thus making the task next to impossible.

    If you looked at one poll after Brown’s “no time for a novice” speech, or a poll during the petrol protests in the first Blair govt. you would draw incorrect conclusions about a future election result. However, if you arm yourself with more information, you can start to make better guesses (though they are just guesses, granted).

    Hence, knowing that the Tories need a 7 point lead, and a 40%+ poll at the next election means you can begin to make some more meaningful predictions. Add to that they only managed 36% in 2010 and have failed to poll anywhere near 40% since then, adds a bit more information.

    You can then look at Miliband’s leadership ratings and Ball’s economic ratings and temper your predictions against them.

    All elections are different, but you can know a fair bit about each election before it has happened, their uniqueness does not become apparent completely in hindsight.

    That said, you can’t get a full picture until after the event, I agree. No one can say 100% the Tories won’t get an overall majority, but something is going to have radically change between now and 2015 if they are. Looking at the polling trends and influencing factors to make predictions is all part of the fun of political polling!

    • hopisen


      But do we “know” what we think we know. For example, you say the Tories have failed to poll over 40% since the GE. But they were regularly on 40/41 in Dec Jan last year, and were scoring 42/43 towards the end of 2010. Admittedly that latter was a while ago now, but it does suggest that such a score isn’t de facto *impossible*.

      Equally- do we *know* the Tories need a lead of 7%? Imagine a scenario in which there a decent incumbency factor for new 2010 MPs. (remember there were a lot of these!), or that the TV scenario is very different, or that Labour loses c10 seats in Scotland to the SNP (which obv wouldn’t impact majority, but would impact largest party/coalition) or that a disproportionate no. of Lab voters in LD/Tory marginal seats go “home” to hand those seats to the Conservatives.

      I’m not saying any of these things are likely, but they are possible, and given that I can just about construct a scenario where the Tories win a majority on a 39-36-14 lead even without incumbent effects, a deal with Unionists, or breaking the 40% barrier (though this is reliant on a fair degree of pro coalition tactical voting, and should be taken with a hefty pinch of salt!) we should be very wary of saying we “know” things about the next election which it turns out we don’t actually know!

  2. Mark Pack

    Good points, especially as with a peacetime coalition government we are almost by definition in unchartered territory.

    Another good reason for prediction modesty comes from looking back at previous predictions, and how very wrong even predictions from the most reputable sources have often been.

    A good political example is all the talk about how Labour might never win again after the 1959 election (won to some degree 4 of the next 5) and then after the 1992 election (won next 3 in a row).

    We never did turn Japanese (http://www.markpack.org.uk/18483/predicting-the-future-we-didn%E2%80%99t-turn-japanese/)

  3. Andy Cooke

    Ooh, that’s evil. And brilliant.

    Do you have the answers? I’m dying to know which ones were which.

  4. James Gray

    Hopi – I agree, no one can know for sure what the result will be. My point was we can know a little and then base predictions on that and then put a probability on each prediction. For example, at the current state of play, a Tory majority is looking less likely than a hung Parliament and probably even a Lab majority. Of course events can change everything – what if Boris became leader? What if D Miliband became opp. Leader? And so on.

    To say, look at 10 random polls and then conclude it’s impossible to guess an election outcome does not prove anything. We have far more information that a random poll two and half years out from the next election. We have longer trends and know the political narrative.

    Anyone who thinks its a done deal for Labour is in cloud cuckoo land; but we can say the Tories have a steeper mountain to climb.

  5. James Gray

    I take your point about knowing though – my lazy “knowing” the Tories need a 7 point lead is a case in point! Getting the assumptions right is important to making the predictions!!


Leave a Reply