Over the course of this increasingly interesting Australian election, I have followed the betting markets. I want a quick proxy of the likely election outcome without the requirement of monitoring the news. I try to avoid overdosing on short term news, preferring to read books and articles that someone has put some thought into. However, being unable to completely detach myself, I have had a quick peak at the betting markets each day. Everyone else can absorb and collate the events of the day for me.
During the course of the campaign leading up to the election (and this is from memory), Labour was generally ahead, with the Coalition paying over $3 at some points. On the night of the election, the polls in Eastern Australia closed at 6pm EST. At 9.30pm EST after some of the results had come in, the odds on a coalition victory had blown out to $3.80 on Betfair. When I returned two hours later after watching a movie, the Coalition was favourite (paying around $1.40) with the most likely outcome a hung parliament and the balance of power in the hands of a few rural independents.
That result has played out for the last two weeks, with the three rural independents possibly announcing who they will support tomorrow. Over those two weeks, the Coalition has slipped out of favouritism and is now paying $3.20 on Betfair (they have been as high as $3.40 today).
This raises the interesting question of the accuracy of the betting market in predicting an election outcome. On the one hand, it is wrong to say that the market was “wrong”. The implied odds of a Coalition victory during most of the campaign was around 25 per cent. If those odds are accurate, the Coalition should win one in four elections in those circumstances. Unfortunately, since our sample size is one, this election result does not help us in assessing the accuracy of those particular odds. On the other hand, I am not sure that the markets have provided a useful guide, particularly in the last two weeks. I might have just been better off completely detaching myself until the new Prime Minister is announced.
This highlights a question worth exploring. Over the last few years, I have seen a lot of comments and articles about how the betting markets are generally right. In an article on election markets, Andrew Leigh and Justin Wolfers noted that in 2001, the Centrebet favourite won in 43 out of 47 marginal seats, while they won 24 out of 32 in 2004. The implied Centrebet odds for 11 of the seats was between 50 and 60 percent for the favourite (that is from a visual inspection of Figure 6 in another article by Wolfers and Leigh). Another dozen or so seats were between 60 and 70 percent. To have seats in those probability ranges and get 43 out of 47 right suggests that the odds were far too conservative. The favourite should have been more favoured and a larger payout available for the underdog.