Local Odds

Posted on August 12, 2007 by andrewleigh
This entry was posted in Australian Politics. Bookmark the permalink.

11 Responses to Local Odds

  1. Michael Potter says:
    August 12, 2007 at 10:47 pm

    Andrew – FYI, the coalition is ahead in 76 seats, labor in 73 seats and other in 2.

    To me it appears that either the betting for the overall election is wrong or the betting for the individual seats is wrong.

  2. Kevin Cox says:
    August 13, 2007 at 5:56 am

    They both could be “right”. e.g. if say 100 seats for the coalition were 51 49 and 50 for labor were 100 to 0 then labor would be expected to win.

  3. derrida derider says:
    August 13, 2007 at 10:08 am

    True, Kevin, but if you do a weighted average of the probabilities (deleting the “Other” column) you still get the coalition winning 50.7%, so there’s still a gap between the pattern of the individual odds and the overall betting market for Lib/Lab.

  4. Andrew Leigh says:
    August 13, 2007 at 11:19 am

    MP/DD/KC, one view is that the arbitrage opportunity is Bennelong. According to the Mackerras pendulum, it’s the seat Labor must win to take government; yet Portlandbet (and Sportingbet) have Howard as favourite to retain his own seat.

  5. Seneca says:
    August 13, 2007 at 12:58 pm

    Strictly speaking, to derive an election winning probability from individual seat probabilities, don’t you need to do it as a multinomial – that is, look at all the possible combinations that would yield, say, a Labor win and then calculate the probability of those combinations occurring? For example one outcome would be Labor winning all the seats and the probability of that is .72X.08X…X.91; another is where they win the first 77 seats alphabetically (that is up to and including Huner): probabilility of that is .72X.08X.73X…X.08X.90X.90X.35X…X.06 (subject to fiddling with the probabilities when we want to rule out an “other”win). There should be some way of approximating it to make it easier to do but I can’t remember it!

  6. Seneca says:
    August 13, 2007 at 1:39 pm

    PS: The thing that seems a bit odd is that the ALP is not making a huge comeback in Queensland (on the basis of simple comparison of probabilities, ALP could be 8 or 9 seats, Coalition 19 or 20): this seems to be the basis for the divergence between the seat count here and the impression from betting and polls.

  7. christine says:
    August 14, 2007 at 12:08 am

    The Canadian National Post reported (very briefly) today that recent polls show Howard well behind Maxine McKew (the Post has been a monster Howard supporter, running full page articles explaining how he’s the most brilliant conservative leader ever). I found that intriguing, and as I’d sort of expected, the betting numbers show the reverse, and I see Andrew’s highlighting it as an arbitrage opportunity. Any thoughts on what’s going on there?

  8. EconoMan says:
    August 14, 2007 at 10:37 am

    All I know is that I’m glad I got my money on when Maxine was $4.75! I have to decide now whether I want to lay off…

    I tend to agree with McKerras and Antony Green, that if Labor win the election they will win Bennelong. Not only because it’s the so-called ’16th seat’ on the pendulum, but if Labor’s going to win, there’s no point voting to have the ex-PM in your seat, (which would surely be worth a few points if they think the coalition are going to win). The coalition are more likely to hold Wentworth than Bennolong IMO.

    Other excellent betting opportunities IMO:
    – Boothby and Sturt. Despite needing ~5.4% and ~6.8% swings, aggregated polling for SA suggests the Labor swing is >10% at the moment.
    – Paterson and Robertson – though I think the ‘overs’ is smaller here.

  9. Spiros says:
    August 15, 2007 at 10:50 am

    Odds for individual electorates are just silly because the amount of money bet on any of them is likely to be tiny, except in elecorates of special interest like Bennelong. The markets are just too thin for the prices to be meaningful.

    Odds for the election as a whole have more meaning because insiders with access to the parties’ private polling are known to place bets based on that inside information. But not even they have information about every single electorate.

  10. Andrew Leigh says:
    August 15, 2007 at 3:28 pm

    Spiros, there’s some good evidence that isn’t true, and indeed some have suggested that prices can be accurate with as few as one trader. Intuitively, it must be true that these prices are at least approximately right, or plenty of people would be jumping in for a bargain.

  11. Sacha says:
    August 17, 2007 at 11:29 am

    Seneca, you could just write a little program to calculate those numbers.

Comments are closed.