The poll's a longshot

Peter Brent in February:

Centrebet is paying $1.90 for a Labor win and $1.80 for a Coalition one. These (I think) are shortest Labor odds (ie longest Coalition ones) since around September/October 2001. Andrew Leigh has a lot to answer for. This “the betting public is better at predicting election results than opinion polls!” shtick has become a standard roll-out for journos and commentators; SMH today two pieces on or alluding to it. (See also Simon Jackman’s take.) You know, the punters, who had the Coalition at over $3.00 in early 2001, but changed their minds when the government took the lead in opinion polls later in the year. Not too bright in early 2001 (or for most of the 2005 WA campaign), were they? Or the dills who on New Zealand’s election day in 2005 had the Nationals winning? All these positions were based, mainly, on whatever the polls said at the time. The punters are sheep who reflect the current received wisdom, which is usually correct, but sometimes wrong. B-a-a-a-a-h. I’ll follow my own judgement, thanks.

Peter Brent in Crikey today:

Since Monday’s now famous Galaxy poll in News Ltd tabloids, Centrebet’s odds on who will win the next election have moved a little in the government’s direction. On Monday morning you would only get $1.69 for a Labor win, now it’s $1.72. That sound you can hear is sheep dawdling after the latest poll. B-a-a-a-h. Over the last few years there’s been an explosion of articles and academic papers spruiking the superiority of betting markets over opinion polls. Its genesis is economists who see Adam Smith’s invisible hand at work: a wondrous distillation, motivated by the dollar, of our collective knowledge about the upcoming election. What rot. What you get is just distilled current received wisdom. And that received wisdom consists of some truths but also lots of woolly nonsense, all overlayed by the latest opinion polls. If you had bet $100 in early 2001 on Howard surviving that year’s election you’d have got back more than $300. On the morning of New Zealand’s 2005 election, Centrebet was paying $2.10 for a Labour government. Punters not too clever then, were they? They were, of course, following the opinion polls. Proponents of the “betters know best” narrative often cite 2004 as their evidence. That was when the opinion polls usually had Latham and Labor ahead, while betting markets always favoured the government. But that rests on a straw man: that opinion polls are supposed to be literally predictive. No-one, least of all the pollsters themselves, claims they are. Polls are instead imprecise dips in the public ocean (that get more precise as election day approaches). Even before this week’s Galaxy, no-one seriously predicted Labor would win the next election by the 57 to 43 the polls were averaging. Here’s something you can bet on. If the next polls show Labor’s gap increasing again, the betting money will follow. If they show a tighter election, money will move that way. As I said, b-a-a-a-h. Those current Centrebet odds imply a 54% chance of a Labor win and 46% chance of a Coalition one. This is probably about how the commentariat sees it also. Personally, I reckon it’s closer to 65 to 35, but apparently I’m supposed to adopt the view of “the mob” instead. Heeding the will of the majority is one thing, but that’s ridiculous.

Well, if Peter’s just going to repeat himself, I guess I’ll take the liberty of doing the same.

Justin Wolfers and I have two papers now in which we show that in Australia, the betting markets are better predictors of election outcomes than the polls. Internationally, others have published maybe two dozen papers showing the same thing. The betting favourite doesn’t always win, particularly when the two parties are close (the answer to the NZ 2005 example is simple: a party with a 45% chance still wins about half the time). But I am aware of no paper that finds that the polls outperform the betting markets as a prediction tool (which is what you’d find if betting markets just followed polls). 

I’m grateful that smart political scientists like Peter are interested in this debate, but I find it a bit odd that he’s continuing to make the claim that betting markets are nothing but a reflection of the polls, given that all the academic evidence points the other way.

Advertisements
This entry was posted in Australian Politics. Bookmark the permalink.

31 Responses to The poll's a longshot

  1. Peter Brent says:

    Ouch.

    Andrew, I suppose we’re talking at cross purposes. For example, you wrote in Feb (http://andrewleigh.com/?p=1345#comment-59040) that “JW and I (and others) have done formal tests of this. Polls do affect betting markets a bit, but not much. ”

    I find that so preposterous – presumably if the polls today had Howard ahead 57 to 43 today’s Centrebet odds would still be favouring Labor – that we’re really talking at cross purposes.

    Those 12 papers can’t can’t “show” it to be so, they can produce evidence for or against. But anything that finds the polls make “not much” difference is employing what is, in my opinion, absurdly narrow focus.

    The polls effect everyone’s view of the contest. They are ridiculously over-reported and feed into Michelle and Paul etc’s reporting. This is what the punters read, as well as the polls.

    One day when I’ve mastered the art of writing academic papers relatively quickly (I’m just a PhD student!) I’ll address this in that forum.

    But in the meantime, in popular forums like Crikey, referring to academic papers “showing” such and such isn’t very useful. To me, anyway.

    Cheers!

    PB

  2. Leopold says:

    Well he’s not making that claim, he’s claiming they are a reflection of the general ‘consensus’, which the polls feed into.

    And frankly Andrew, the way the betting markets have jerked back and forth with the polls in the last 3 months has been crystal clear to anyone with their eyes on both sets of numbers. And Peter has actually made a small error – Centrebet had $1.63 before the Galaxy was published, and had moved Labor’s way for nearly two weeks before that poll was published. I suppose the Galaxy poll had nothing to do with the reversal to $1.72?

    Further: there was a ‘budget bounce’ in the betting market, which unravelled with the emergence of polls showing no such thing. All this just coincidence in your view? The punters initially thought the budget was great for the government, then, coincidentally, reversed their view at the same time as the polls suggested otherwise?

    That betting markets and polls are not consistently correlated, I grant; but that the punters have been responding to the polls over the last 3 months has been crystal clear IMO.

  3. Andrew Leigh says:

    Peter, you’re one of the most sharpest political scientists I know, and a frequently cited commentator in the media. So I’m holding you to a higher standard than I’d typically expect of a PhD student.

    What worries me about your attitude to betting markets and polls is that I get the sense that you’re not engaging seriously with the theory and empirical evidence about betting markets. A lot of people have done work in economics and political science on this question, and pub putdowns (b-a-a-a-h) imply to your readers that this literature doesn’t exist. I’m not saying you should cite articles in your Crikey posts – merely that your public commentary should be informed by research. (Once again – a party that is 55% in the betting markets should lose 45% of the time.)

    We’re not at cross purposes: the research speaks directly to your question. If betting markets just followed polls, then polls would be a better predictor of election outcomes than markets. Every paper that I know of finds the opposite to be true.

    Those of us who have come to the view that betting markets have superior predictive power to polls didn’t do so because we’re dogmatic gamblers, it’s because we looked carefully at the available evidence and let it shape our views. All I’m asking is that you do the same.

  4. Andrew, I find your logic very hard to follow. “If betting markets just followed polls, then polls would be a better predictor of election outcomes than markets.” Why? Firstly, the logic within this sentence does not follow. It is quite possible for betting markets to follow polls but end up being a better predictor of the final result. But as Peter says, polls do not ‘predict’ election results, they are no more than what they are, a cross sample of how 1,000 people respond to question on voting intention, just as betting markets reflect the expectations those willing to put their money on an election.

    I suggest if commentators want to read any more from these measures and give them predictive powers, it is because they cannot form an independent view themselves of political developments.

  5. Verdurous says:

    Andrew’s position if justified here. As an objective observer, he has a duty to take empirical evidence and make conclusions from there. I’m no particular fan of “the wisdom of markets” type thinking (I remember well seeing Justin Wolfers predictions about the likely stock market reaction to the invasion of Iraq go 180 degrees wrong), however, it is not sufficient to rely on our gut instinct or anecdotes here. If we feel betting markets are no better predictors than polls then we must show superior literature/research to support this. Until then, Andrew’s findings must be respected.

  6. I certainly have no wish to disrespect Andrew’s findings but to question its basis. There has been an increasing obsession with opinion polls to the point where they are given predictive powers they simply do not have. Rather than looking for some other ‘predictor’, it strikes me as more useful to try and question our understanding of the political trends themselves.

    Peter Brent’s style can be flip, but enjoyable to read.

  7. EconoMan says:

    I think the cross purposes here is that Peter and Leo are arguing that the polls reflect the betting market, believing Andrew doesn’t agree. Andrew can speak for himself, but my guess is he (like me) accepts that polling results are one of the sources of information that can affect betting markets. I’m sure he (like me) also agrees that the betting markets aren’t perfect — hence the budget betting bounce, which I thought at the time had no justification. But hey, I could be wrong too.

    If I’m not mistaken, Andrew’s fundamental point, which I completely agree with based on the evidence, is that betting markets are a better predictor of elections than the polls. As to the strength of the effect of polls on betting markets, I don’t know one way or the other on this.

    Peter, you point out examples of where betting markets got the election wrong (or were wrong in the lead up to an election), presumably as evidence that betting markets are wrong or simply follow polls. (I note your objection Leo, but the Crikey article says “They were, of course, following the opinion polls”.) A price of $2.10 implies odds of about a 45% chance of Labour winning in NZ. Not exactly a long shot and hardly a solid refutation.

    On the $300 for Howard winning early in 2001, the betting market would, according to the efficient market hypothesis, reflect information available at the time. Just like you claim polls reflect current voting intention, not an election prediction. While I’m on it: If polls aren’t intended to be used as a predictor of election results, why are they taken. (One obvious and true answer is: to sell papers.) Despite all of the protests from the pollsters along the lines of “this reflects current intention only”, the polls are always used to predict. And even if they aren’t intended to predict elections, it is perfectly legitimate to test whether polls or betting markets provide the better predictors of elections.

    As Andrew repeatedly argues, looking at all the available evidence, not just selected elections, it shows the betting market is a substantially better predictor of election outcomes. If you disagree, show us the evidence.

  8. Peter Brent says:

    Andrew,

    Studies have shown I respond to flattery at least as much as the next person.

    But I wrote “just a PhD student” re your implication/suggestion that I address the academic material. Well, I’ve only read one of those papers, and it didn’t convince me. To me it contained unsustainable assumptions. As I said, one day hopefully I’ll put my arguments in that academic forum.

    I’ve read other papers (by others) purporting to “show” electoral things that are just ridiculous.

    EconoMan’s excuse the times punters were wrong, that they just “reflect[ed] information available at the time” pretty well says it all. An argument insulated with stuff like that could go on self-nourishing forever and “show” just about anything.

    Your argument is correct only against the straw man proposition that opinion polls are literal predictors of election results. But wins against straw men don’t contain much power.

    Rather than cite papers, I suggest you economists address Leopold’s simple proposition: that one Galaxy poll seems to have produced a dramatic turnaround in punter activity.

  9. Andrew Leigh says:

    Peter, three quick responses.

    1. “Your argument is correct only against the straw man proposition that opinion polls are literal predictors of election results.”

    I’m puzzled as to why you think this is a straw man. Why are we looking at polls if not to predict the next election result? Does anyone seriously think the actual quesion asked by the pollsters (‘if an election were held tomorrow…’) is of any inherent interest?

    2. “EconoMan’s excuse the times punters were wrong”

    Let’s go back to probabilities. If the betting markets are perfectly accurate, then when they say that a candidate is a 55% chance, that candidate should lose 45% of the time. By contrast, you seem to think that the betting markets are perfect if 55% candidates always win, and 45% candidates always lose. That’s simply a misunderstanding of what a probability means.

    3. “Rather than cite papers, I suggest you economists address Leopold’s simple proposition: that one Galaxy poll seems to have produced a dramatic turnaround in punter activity”

    Economists prefer to deal with statistical evidence and large sample sizes, rather than anecdotes and single cases. The simple answer is that if you’re interested in learning about the systematic relationship between polls and betting markets, you should prefer evidence based on a regression using hundreds of poll movements to evidence based on one poll movement.

  10. Bring Back CL's blog says:

    seems to me Andrew is saying the betting markets are behaving simialr to the financial markets re-prediction of interest rates.

    Opinion polls never predict election, they merely show current opinion.
    This may or may not change.

    My experience of last years US elections and watching the betting markets this year is punters change very quickly on opinion polls and in Australia incorrectly predict opinion polls

  11. Verdurous says:

    One could propose a number of explanations for betting markets out-performing polling.

    a) maybe betters have access to more information than poll sample population or seek out that information more avidly.

    b) maybe polling may be subject to bias that is not evident in betting markets e.g. “please the interviewer” type bias.

    c) Perhaps betters are selected from a more politically influential group of people (e.g they are more often from marginal seats – or maybe they are wealthier and exert political power through wealth).

    The way I see it, either polling is somehow a less powerful way of making predictions than markets OR there are practical issues with polling that need refining (such as sample size, randomisation etc., questionaire design, anonymity, bias etc).

    I tend to think that the latter is the true story. Ultimately, the actual election is a poll (the main differnce being that the sample size is exactly the same as the target population!)

    Polls and betting markets are good in that they scare politicians but potentially bad in that they might set-up postive feedbacks in voting intention.

  12. Richard Green says:

    I’m yet to be convinced of the efficiency of the Australian betting markets in regard to elections since the number of punters is unlikely to be high.
    Afterall, why would any rational punter (taking intto account no rational person would bet with a bookie due to the latter’s profit premium) put money on now for a 60c-$1 return on the dollar in 4 months when they can get similar odds on the track this afternoon. Without punters, there is no market, and a non-existent market can’t be efficient. As it stands, the money being laid on is likely only from political junkies who suffer from the same problems that make betting markets (when they exist) better than “experts” in the first place.

    I’m completely convinced by the American markets though, since the betting population is large enough to overcome this, and moreoever, the empirical studies on the topic are dealing with a great deal more elections (what with 50 legislature elections, 50 governors, a congress and a president) and have a much more satisfying sample size than the piddling 9 available in Australia.

    Whilst the theory is utterly sound, we can’t overlook the caveat that an efficient market necessarily has to be, well, a market.

  13. EconoMan says:

    Andrew has covered most of it, but I’m happy to adress your and Leo’s simple question. The betting market isn’t perfect. It is the aggregation of decisions from boundedly rational individuals. Some of those people over-reacted (IMO) to a single poll. Nevertheless, as betting volume increases, the market tends to efficient prices, and hence a fairly accurate probability of each side winning. Lets recall that in NZ, as slight outsiders, Labour had to give the Foreign Affairs ministry to someone who is pretty much racist in order to form government.

    As indicated or at least implied in my previous post, I also completely disagree with the ‘straw man’ proposition. Despite all the protestations, polls are used to predict elections. If they weren’t no one would care about them.

    Bring Back CL: “opinion polls never predict election, they merely show current opinion. This may or may not change”

    While it is strictly true that the betting market simply reflects current opinions (that may or may not change — my point above Peter), we can still use betting markets as an election predictor. Just like we can use polls to try predict elections. Then compare the two predictors…

    I’ll also end with a simple proposition (again): Show us the systematic evidence that betting markets simply reflect opinion polls.

  14. Leopold says:

    If I (or Peter) were arguing the case in your ‘simple proposition’ we’d clearly be wrong. We’re not making that argument.

    Our case (I think) is simply that the markets generally reflect the existing consensus and DO respond to opinion polls. I emphasise: this is not a systematic or exclusive relationship, and as such, might well not show up strongly in a regression. Many poll movements one way or the other (1-2 points) do not generate a reaction. And there is no visible response to Morgan polls (meaning: punters aren’t complete dills and any regression including Morgan will be diluted). And betting markets clearly move in response to other things which would further mess up a statistical relationship. But there have been very clear moments in the past 3 months when the betting market view has responded dramatically (and even reversed itself) to one or a couple of opinion polls. Particularly: the Monday in March when ACN said 61-39, the period following the budget, and the response to this Galaxy. It is not a systematic relationship, but a relationship exists.

    A further point: that polls are ‘used’ to predict elections does not make them a predictive instrument. They are not. They are a statistical sample of a population at a moment in time. Betting markets are by their nature a predictive instrument. Various studies have argued that the predictive instrument is a better ‘predictor’ than something which is not a predictive instrument. Well, my giddy aunt. Whoda thunk it? You can’t argue that because newspaper headline writers use polls in that way that they ARE a predictive instrument; that’s just absurd. Polls are polls, however they are used. People who use betting markets to try and predict the number of seats are silly; so are people who use a poll in June as a prediction of the vote in November.

    One final point: my recollection is that AL’s paper criticised the polls in 2004 because Morgan and Newspoll showed 51-49 and 50-50 respectively in the week before the election. But Morgan is rubbish (as the punters clearly know) and Newspoll was using a highly dubious preference allocation. The primary votes of the three credible pollsters converged on 47-38, a thumping Coalition win in the vicinity of 53-47, which is what eventuated. Polls worked beautifully for those prepared to read them critically.

  15. Verdurous says:

    Economan,

    Your proposition that:

    “The betting market…is the aggregation of decisions from boundedly rational individuals….”

    is likely not true. The boundedly rational individual is not influenced by friends and other punters. That’s partly why markets of all types often get it wrong. So the flow-on assumption that….

    “…the market tends to efficient prices, and hence a fairly accurate probability of each side winning.”

    is also not necessarily true. Just as stock markets and currency markets are subject to raging internal instabilities and feedbacks, it is probable that betting markets are too. The warm, fuzzy idea of markets tending to equilbrium is simply not born out in the real world. Otherwise, markets would get it right %100 of the time.

  16. Stephen L says:

    I think the problem here is the question of what is meant by using polls to predict outcomes.

    I accept Andrew’s claim that the betting markets out do a simple reading of polls, but I think what Peter is saying (certainly what I would say) is that they don’t necessarily beat a more sophisticated interpretation.

    To expand on Leopold’s example. Many of the polling companies use frankly stupid ways of allocating preferences from smaller parties, and more used to in the past. Lets imagine for the sake of things that all polls did this. In this case betting markets would find it easy to outdo the two party preferred estimates produced by the pollsters. On the other hand any half decent psephologist could produce a more reliable estimate of how preferences would flow based on those primary votes. Consequently I’d be inclined to take trust the assessment of at least a group of capable psephologists over the markets.

    I’m using the preferences issue as an example, but there are other factors. For example in an electorate where there is a gerrymander the polls might be able to predict the votes, but not the result in seats. Either the betting markets or commentators can look at polls showing 52/48 and think “well the gerrymander means party A will need 53% to win, so on this basis I’ll predict party B will win.

    The question is whether the betting markets do a better job of interpreting a range of factors – preference allocations, gerrymanders (even unintentional ones), trend estimates etc – than intelligent commentators. I’d be surprised, and I’d also be surprised if papers such as Andrew’s one actually compare the markets with these sophisticated interpretations of the polls rather than the raw figures.

  17. Michael says:

    Aren’t polls and betting markets asking totally different questions?

    As neither an academic, nor a mere PHD student, I would have thought that:

    1. Polls assess how people say they would have voted had an election been held at the time of the poll (which doesn’t mean that they actually would have voted that way, of course).

    2. The betting markets indicate who people think will win when an election is called.

    I reckon if I were intending to bet money on election results, I’d be keeping an eye on the polls.

    So, if I had been thinking of betting on this year’s election a month ago, I have thought Labor was a pretty good chance based on the polls, but I’d also have thought of other factors – like the fact that Howard has come from behind before.

    On the basis of the other factors I might have placed a bet on Howard, but if the pollsters rang me, I’d be giving my vote to Labor.

    The strange thing that I think would be more worthy of study is why the betting market is so much closer in its predictions than the polls which ask people who they think will win the next election.

    Cheers
    michael

  18. barney says:

    I guess i’m left asking the question:

    Upon what evidence/information do punters make their decisions?
    Unless they are psychic or intuitive, then they are going to be relying upon signposts like opinion polls and commentary. Even if it’s inside information (party apparatchiks etc) surely they are going to be getting their info from polling/qualitative research etc.

    It just occurred to me that it might be interesting to see if there is an initial “plunge” by smart money followed by a “baaah” reaction. The punters see which way the market is going and get on board. Anyone who has ever been to the races will have seen this phenomenon.

    The other interesting thing for me is that the betting market is being seen as a homogenous whole when it is (one assumes) a whole lot of disparate and differently motivated punters.

    I have to admit that I find Leopold’s point about the “budget bounce” that wasn’t, a compelling one.

  19. The nice thing about this election cycle so far is that we’ve got quite a bit of data now with which to look at the relationship between polls and markets, with plenty of turning points and 50/50 crossings etc, I think more than in previous cycles (but I am willing to be corrected on that last point). There are lots of fun (?) things one could do with the data, even up to this point in the cycle. Testing hypotheses of the sort that

    p(y_t | y_{t-1}, x_{t-1}) = p(y_t | y_{t-1})

    i.e., x provides no more information in forecasting y beyond the information already in y may not make for compelling reading for a lay audience, but part of our job is to find ways to make that intepretable to one another, to journos, and the public.

    Incidentally, 53-47 would be Labor’s biggest win since its record 53.1 2PP in 1983. I took it as an interesting signal of the state of play in the cycle that such a poll result could be “interpreted” the way it was. LOL…as they say.

    and greetings from London, en route to a conference on election polling at Essex.

  20. Panadawn says:

    I don’t think the entire betting odd turnaround is just because of the Polls.

    I think a lot of value-seeking punters out there see the coalition at $2.20 and think thats a pretty good return based on past performance, so they put money on.

    I put $50 on myself as soon as I saw it at $2.20 for that very reason.

  21. Peter Brent says:

    I agree with all those who have agreed with me.

    btw Here, in more readable format, is the original offending article.

  22. Andrew Leigh says:

    Barney, a good primer on prediction markets is this one by Justin Wolfers and Eric Zitzewitz. They also have a few other articles on prediction markets, collated here.

    Simon, I agree that this would be a fun cycle on which to test some of these hypotheses. More data are always better (unless you prefer storytelling to empirical analysis, that is). Oh, and to answer your question, in 2001, there were 3 crossings in the Centrebet odds (27 Sept, 26 Oct & 29 Oct). Hope the conference goes well.

  23. I just wanted to return to the original premise of this thread. Peter Brent’s concern seems to be that the markets move in respond to the polls.

    I would have thought that the only thing more distressing than markets responding to the polls would be if markets didn’t respond to the polls.

    (Of course it is possible for this to be rationalized, but one would need to argue either that the markets were omniscient and already knew how a random sample of 1000 Australians were feeling, or that the polls were worthless, neither of which strike me as defensible.) All told, Peter’s observation is an observation entirely consistent with market efficient (but not with omniscience).

  24. Peter Brent says:

    Omnicient (adj): having complete or unlimited knowledge, awareness, or understanding; perceiving all things.

    Pompous (adj): characterized by an ostentatious display of dignity or importance.

    Centrebet’s price for a Labor win has now moved from $1.62 to $1.78, all because of one opinion poll and, more importantly, the effect of that pollon general reporting of the political situation.

    I think we would all agree that if Galaxy staff had stayed at home last weekend it’d still be around $1.62. (There would also be – as there was before Monday – lots of talk about Peter Costello for PM.)

    Sadly, that’s the world we live in.

    As I originally wrote: the betting markets give you “distilled current received wisdom. And that received wisdom consists of some truths but also lots of woolly nonsense, all overlayed by the latest opinion polls.”

  25. Peter Brent says:

    Spelling (verb): to name, write, or otherwise give the letters, in order, of (a word, syllable, etc

  26. jack strocchi says:

    The polls have consistently predicted an ALP landslide. Whats the bet that this wont happen?

    If it does happen then mumble is right. mumble crow.

    If it does not happen then punters know something pollsters dont. mumble eat humble pie.

  27. Peter Brent says:

    Oh. Upon re-reading JW I realise he was referring to the market’s lack of omniscience, not mine.

    Oh.

    Please ignore “pompous” reference above. Apologies.

    Bye for now.

  28. Graeme says:

    This debate seems of limited consequence. At best those promoting markets as predictors can claim that some businesses/lobbies would like a crystal ball giving the likelihood of a change of government.

    Polls measure disgruntlement with governing parties/leaders (either overall or on specific issues). Modern governments – and oppositions – respond to these, both substantively and with spin to address perceptions. Whether this is good government/leadership is moot.

    1. It is undoubtedly democratic.
    2. Electors seem (consciously and collectively, unconsciously) to be aware of the power these polls have to shape the political process that we are otherwise largely locked out of in a mass, rep’ve democracy.

    Hence it would be regrettable if polls disappeared in favour of the reporting of markets. Parties would continue to rely on polls, not markets, to shape their policies/tactics, but a narrower (and hence less reliable) range of polls, and these polls would not be publicised.

    What we need is not attention to betting markets, but a saner polling environment – less frequent polls, better reported both as to MoE and that they measure current dissatisfactions, not electoral predictablity.

  29. Rambler says:

    Interesting article on Way2Bet on this topic, for those with an interest in the subject. Not definitive in any sense of course, and only reviews one bookie, but in light of the above, thought some of you might find it of interest.

    http://www.way2bet.com.au/news/article/analysis-of-election-markets-12-june-2007

    Cheers.

  30. David says:

    I did as Andrew suggested and read the paper “Competing Approaches to Forecasting Elections: Economic Models, Opinion Polling and Prediction
    Markets’ for me the two of the most interesting numbers in the paper were 2005, the date of publication, and 2004, the date of prediction.

    I would love to read Andrew and Justin’s predictions for the coming Federal election. Who will win and by how many seats? Their insights into the coming political contest are just too valuable to wait until a 2008 to find out who was going to win in 2007.

  31. Andrew Leigh says:

    David, there’s a difference between assessing the various forecasting tools (which by definition can only be done after the election), and making a judgment call on who’s going to win. The first is a quintessentially academic exercise – the second not so much. I’m afraid that I’m only really comfortable saying what we said prior to the 2001 election and prior to the 2004 election: watch the bookies.

Comments are closed.