I’ve mentioned before that the fact that most bookies have Labor odds-on to win, while the Liberals are favoured in about half the individual seat markets (they were ahead in a majority until recently) seemed to create arbitrage opportunities. A nice piece by Tim Colebatch resolves the apparent paradox to my satisfaction at any rate.
The paradox is explained by the fact that of the 29 seats in which the odds are closest, the bookies have the Coalition as favourite in 23 and Labor in just six. The law of probabilities suggests that if Labor gets anything like the vote being recorded in the polls, it will win a lot more than six of those seats.
The only blemish in this explanation is that there’s no need to refer to the Labor vote in the polls. Given 29 seats that are nearly even-money, it’s reasonable to expect Labor to pick up at least 10, just on the basis of the betting odds.
Going to the more general question of prediction, 2007 has been a win for the polls in their contest with pundits and punters as to who provides the best prediction of election outcomes. Labor jumped to a winning lead in the polls as soon as Rudd replaced Beazley and has held that lead, with only marginal erosion ever since. The pundits and the betting markets have gradually come into line, but it’s hard to believe that they would have done so in the absence of the information provided by the polls.
Of course, it’s still possible that Howard will come back, as he has done in the past, but that will just prove all predictions wrong.