Today via the Courier Mail comes a Federal poll for Qld showing the primaries running 46/39 to the Coalition, washing out into a two party preferred of 51/49 to the Coalition. This comes from a sample size of 800, giving us an MoE that maxes out around the 3.5% mark.

At the 2007 election, the primaries ran 44.5/42.9 to the Coalition for a two party preferred of 50.4/49.6 to Labor – so this poll shows a slight swing to the Coalition.

This is the second poll result we have for Labor in Qld this month at the federal level, with the most recent Nielsen poll also publishing their 250 Qld sub-sample. What we can do (and this is something we’ll be doing more often as we approach the election) is pool these poll results, run them through our monte carlo election simulation, and work out nominal implied probabilities for the ALP winning each seat based on the current polls.

For the pooling of our poll results we will use here, we’ll just weight by sample size in our aggregation. Once we take the Galaxy and Nielsen results, this is what we end up with.

We can then apply this pooled 50.7/49.3 result to the Qld component of our election simulation (which uses the new boundaries), and work out how many times in each 20,000 iteration election simulation any given seat would have been won by the ALP .

This is what the results looked like:(click to expand)

On the new boundaries, the ALP notionally holds 17 seats at the moment based on the 2007 election results (the dark red are notional Labor seats, the Blue notional Coalition seats). While there are 18 seats in Qld where the simulation based on the Galaxy and Nielsen polling gives a greater than 50% probability of Labor winning, the cumulative simulation results suggest that the most likely number of seats the ALP would have won in Qld was 16. Effectively, these polls suggest a net loss to Rudd of 1 seat in Qld at the moment compared to the 2007 election results applied to the new electoral boundaries.

The difference between the two figures (the 18 seats where Labor is >50% and 16 seat total forecast) comes about as a result of the distribution of the probabilities for each seat. Let’s say, for example, that there were 10 seats where the probability of a Coalition win was 30%. On any given election, we’d expect the Coalition to win around 3 of those seats give or take, even though they didn’t have a probability of 50% or more in any of those seats.

The cumulative probability of all seats combined is as important as the given probability on any particular seat.It’s actually more important when you’re looking at the most likely state wide outcome from a given polling result.