Gambling on election polls

The Galaxy Poll forecasts a two-party preferred vote in the forthcoming election of 53% for Labor and 47% for the Coalition. It is only a single poll and in inconsistent with the other polls so I see it only as very weak evidence of an improvement in the Coalition’s prospects. The poll itself did do well in the leadup to the 2004 election. As of today (June 6) Sporting bet was paying $1-70 for a $1 wager on a Labor win and $2-05 for the Coalition. The odds have moved a few cents against Labor in the few days since the Galaxy Poll was announced.

The question is whether I should put a $1000 bet on Labor? Suppose I did:

If Labor did win I would net about $700 which would buy me 4 dozen 2004 Kalimna shiraz and would yield me a-once-weekly decent bottle of plonk for almost a year. (Expected utility theory suggests that reasonable decision makers should focus on the expected utility of the prizes they gain from participating in a lottery). The probability weight I should attach to this is about 0.55.

If Labor lost I would lose $1000 but gain the utility of having the party I support electorally in power for another term and being able to avoid several years of comradely triumphalism in the blogosphere. While this might be a somewhat muted victory it would compensate a little for the lost $1000. The probability of Labor losing seems from Sportingbet to be about 0.45.

I might back Labor who are favourites at this stage – but I’ll wait for a few weeks to see if I can get a better price on them.