The day is here! We have seen three large sample polls from Newspoll, Nielsen and Roy Morgan, and the odds have shifted slightly, so I thought it was worth publishing one more update.
We have seen:
- 54/46 in the Coalition’s favour from Newspoll with >2500 respondents
- 54/46 in the Coalition’s favour from Nielsen with 1431 respondents
- 53.5/46.5 in the Coalition’s favour from Roy Morgan with 4,937 respondents
As of 12:10pm, Centrebet’s odds are 1.01 for the Coalition and 15.00 for Labor.
A few keen readers noted that some of our numbers previously reported implied that our prediction was for only one independent. Our “most likely” case had actually been reporting the distribution mode for each of the ALP and Coalition, which is not quite the same as the mode of the entire distribution, which is probably what is more interesting to people. For the technically minded, mode[P(A,B,C)] is not necessarily mode[P(A)].
To be clear, our model predicts that the most likely outcome for the number of “others” (defined as anyone other than ALP or Coalition) is 2.
Updating the model, and now being clear to express our “most likely” prediction as the mode of the full distribution, our final predictions are:
- Most likely result: 51 seats for Labor, 97 for the Coalition and 2 for others
- Median result: 51 seats for Labor, 97 for the Coalition and 2 for others
Our estimate for the 2PP vote share is 44.7% (95% CI: 43.8% to 46.2%) for Labor, and 55.3% for the Coalition (95% CI: 53.8% to 56.2%).
As noted in our latest full update, our model bias is calibrated off the last election, where pollsters were on average too pro-Labor. Our predicted 2PP thus comes in below the polls. If pollsters turn out to be on average party-neutral then we will see a better result for Labor than predicted here.
Betting markets model
Our model for the betting markets predicts a 2PP vote share of 44.4% (95% CI: 41.0% to 47.0%). Conditioning this on historic elections gives 44.9% (95% CI 42.2% to 47.6%).