If the polls are so lop-sided, then why aren’t the betting markets?

Monday’s release of the latest ACNielsen poll highlights an interesting discrepancy between the national polls and the betting markets.

That Nielsen poll had Labor on 55% two-part preferred, on a sample size of 1,410. Under an assumption of uniform swing (implausible, but a great simplification for pre-election prognostication), Labor will win the election with 51.4% 2PP or better.

Given the poll result of 55%, and its sample size (and assuming the poll result is an unbiased estimate of voting intentions), the probability that the current level of 2PP support is greater than 51.4% is .997.

If one were to pool the relentlessly similar results coming from Morgan, Newspoll, Nielsen and Galaxy of late, the implied probability of a Labor win would be even higher. But the prices being offered by the betting markets have Labor’s chances of winning at just over .6.

An even bigger contrast is apparent when we look at last week’s Galaxy poll in Bennelong that had Labor’s Maxine McKew on 53% 2PP, on a sample size of 800.

Given the poll result and its sample size, the probability that the real level of Labor 2PP support in Bennelong is above 50% is .96, again assuming the poll is unbiased.

But the betting markets have Labor’s chances of winning Bennelong in a range from .36 to .41 as of Monday morning, some 24 hours after the release of the Galaxy poll.

Of course, there are several ways to reconcile the vastly differing probabilities implied by the poll result and the prices in the betting markets.

The possibilities include:

(a) punters believing that the polls are biased;

(b) perhaps the poll isn’t really as precise as implied by its stated sample size;

(c) perhaps most plausibly, there is going to be a move back to the government between now and election day;

(d) uncertainty about the 2PP number Labor actually needs to win (on average, 51.4% 2PP should win it for Labor but depending on how the swings are distributed, that number could be higher or lower), although this factor is irrelevant when we consider just a single electorate.

Notwithstanding this, the gap between the .997 (the polls) and .6 (the betting market) is huge.

If you more or less believe the poll, and don’t think a 4 percentage point move back to the government is quite on the cards, then you’d have to think that Labor at 1.50 is great value. What other investments out there are offering a 50% return over the next couple of months, with comparable levels of risk?

A win by Maxine McKew in Bennelong at 2.40 represents a 140% return on your wager, with a reputable polling outfit (Galaxy) and a decent sample putting the probability of this outcome at .96.

Which raises the obvious question: what does the betting market know about the outcome in Bennelong that 800 randomly selected Bennelong voters don’t?

The whole is not the sum of the parts? Portlandbet.com is offering prices on all 150 House of Representatives seats, as well as a book on the election outcome itself.

As of Monday afternoon, Portlandbet.com prices corresponded to a .63 chance that Labor would the election (1.50 to 2.55). But the seat-by-seat prices saw Labor the favourite in only 72 seats. That is, the betting markets have Labor coming up short of a majority by 4 seats.

The seat-by-seat markets suggest Labor will pick up 13 seats from the Coalition (listed on the graph), but lose Swan (where the implied probability of a Labor win is .48), for a net gain of 12 seats, which is insufficient to win government.

The four Coalition-held seats closest to crossing over into favouring Labor are Blair (Qld, probability of a Labor win of .496), Stirling (WA, .48), Deakin (VIC, .44) and Page (NSW, .44).

Presumably, relatively small market moves could flip these seats into the “tipping Labor” column, and/or move Swan into the “Labor retain” column.

But until that happens, we’re faced with a puzzle.

Punters in the national market favour Labor, but the seat-by-seat markets don’t quite have Labor over the line.

Again, we’re left to ask: what do punters in the seat-by-seat markets know that investors in the national markets don’t know?