With the Newspoll release today of a slightly smaller than usual demographic breakdown (n=3437 rather than the usual n=5500+), we can combine them up with the last 2 Nielsen polls and run them through our election simulation – representing the broad period of July 22nd through to August 2nd, with a pooled sample of n=6193.
Update: Ordinarily this would give us us state level sub-samples between 650 and 1850, but Newspoll tells us that their breakdowns are using samples of between 657 to 681, which would make NSW about 2/3rds of its expected sample size in a total sample of 3437 and SA about 1.5 times larger than expected. A bit unusual – but nothing dramatic. Newspoll would have weighted it accordingly.
Usual caveats apply here – if an election were held somewhere between those days and where the election result was compatible with the polling over that period etc etc.
The most likely single result would have been a Parliament with 76 seats going to Labor, 71 to the Libs and 3 to the Independents with an assumption of Melbourne being ALP vs Coalition. If we replace that assumption in the simulation with the betting market odds for the ALP vs Greens head to head there, it doesn’t actually change the most likely point result (instead, it increases the likelihood of 75 seats occurring, but it still doesn’t increase it enough to replace 76 as the most likely result).
If we look at the cumulative probability, the most likely general result, the one which goes into our sidebar, would have been for Labor to win at least 77 seats – mostly because of how the relative probabilities wash out between 75 and 79 seats.
The full sim results for the number of ALP seats in Parliament and the number of seats which would have changed hands from the current Parliament come in like this (click to expand):
As we can see from the bottom of the first chart, the probability that the ALP would have won at least 75 seats came in at 71.4%. It’s also worth noting that the probability of the ALP winning at least 74 seats (the result most likely to give the ALP a minority government) jumped up strongly to 79%.
If we rejig those numbers to show the implied probability of the ALP winning at least any given number of seats, we get:
To read it, simply pick a number of seats from the bottom axis and the left hand axis value tells you the implied probability of the ALP winning at least that many seats. As you can see, the 50% mark cuts in at 77 seats (52.3%). As a point of interest, if we just use the Newspoll results, it cuts in at 78 seats – nearly 79 seats.
While the polls have had a fairly variable spread through the campaign so far, the combined results broken down best we can, along the lines of state results (which are the best predictor of the result of any given seat), show that Labor is still in front – but not by a particularly large margin.
UPDATE: I got asked a question about the implied probability of a hung Parliament in these results. If we assume the Greens win Melbourne, giving us 4 non-major party members in the new House (1 Green plus Oakeshott, Windsor and Katter), the Parliament would be hung if Labor got either 72, 73, 74 or 75 seats. The implied probability that such a thing would have occurred with this simulation was 28.6%: