My forecast for the Australian 2019 federal election on 18 May 2019 (with the latest polling data available at 17 May 2019) is a narrow ALP win in their own right (41% chance) or with the presumed support of one or two Green members (54% chance in total). However, a range of other options are very much in play. An 80% prediction interval for the ALP in the House of Representatives is 69 to 80 seats.
The Liberal-National Coalition is likely to end up with between 66 and 77 seats. An outright win is a distinct possibility, and a win with the help of independents or minor parties more so. The Greens are most likely to end up retaining their single seat, but have a fair chance of picking up a second. In total, parties other than the ALP and Coalition are likely to end up with three to six seats.
| Party | Lower estimate | Upper estimate |
|---|---|---|
| ALP | 69 | 80 |
| Lib/Nat | 66 | 77 |
| Ind | 1 | 2 |
| Grn | 1 | 2 |
| KAP | 1 | 1 |
| NXT | 0 | 1 |
The full distributions are shown in the chart below. This is based on 10,000 division-level simulations that take into account uncertainty in the overall swing as well as random division-level variation in swings. Division level swings are assumed to be normally distributed with a standard deviation of 3.2 percentage points, which is a fairly stable figure historically.
For contests other than ALP versus Lib/Nat coalition, trends in the two-party preferred vote are ignored and the simulation is done as straight out random variation from the most recent result. In my view, there isn't enough publicly-available (or probably, any) polling to do anything else reliable in these situations.
Several unusual situations in individual divisions are not taken account of in this approach; notably the Warringah division in Sydney. My model shows this as safe for the Coalition based on previous margin; but unreliable reports of polls with small numbers suggest that independent candidate Zali Steggall is giving sitting member Tony Abbott a run for his money. I've opted to leave this sort of thing out of my calculations rather than make a series of ad hoc adjustments.
For seats where the main contest is ALP versus the Coalition, the simulations above include a nation-wide swing that is forecast with a Bayesian state space model drawing on all publicly available national level polls, taking into account the house effects which show the systematic under- or over-estimates of the polling firms of two party preferred vote in previous elections. That model is summarised in this next chart:
The "house effects" of the different polling firms can be seen below. We can't be sure of the exact average bias of each firm, so we have a distribution of likely results. As an example interpretation of this chart, Roy Morgan is estimated to over-estimate the two-party preferred vote for the ALP by somewhere between 1 and 2.5 percentage points.
This page will be updated at least weekly as new polls become available. Code for all this modelling is available on GitHub.