It seems that due to the nature of the electoral calendars in SA and Tasmania, I always end up having to spend a brief window of time on my home state due to focussing on the Apple Isle, and this year will be no different. I guess I could be more proactive and start posting a little bit earlier, but:
- it always pays to use the most up to date polling just before an election [citation needed].
- the candidates for SA were only finalised this week, and a rundown of who is sitting is probably the only unique effort I will put into the SA elections
- procrastination is fun
So, first, a quick summary of how the Tasmanian Legislative Assembly is elected.
A quick summary of how the Tasmanian Legislative Assembly is elected.
Tasmania’s lower house contains 25 seats. These seats are elected by the Hare-Clark method from 5 districts. To keep this quick summary a quick summary, imagine that Tasmania is just a smaller Australia, the five divisions of Tasmania are the states, and they are electing 5 MLAs instead of 6 federal senators but with some different maths.
This means there are five divisions (Bass, Braddon, Denison, Franklin and Lyon) which will elect 5 MLAs each by proportional representation.
Polling so far
There has been much made of the impressive spike in polling for the Liberal party in this election lead-up, and rightly so. This is claimed to be the largest polling jump on record for EMRS, including hesitantly by Tasmania’s own Dr Kevin Bonham. This is something I have not personally verified, but this does not appear to be an erroneous outlier or issue with methodology, as two polls have now confirmed this trend:
Predictions
The Hare-Clark system used in Tasmania uses the Droop Quota to calculate how many votes are needed for a seat to be won. This works out to a seat being won by exceeding 16.666…% (1/6th) of the vote. If we applied the latest state-wide polling:
|
Liberal
|
46%
|
|
ALP
|
34%
|
|
Greens
|
12%
|
|
Jacqui Lambie Network
|
4%
|
This would result in two automatic seats for the Liberals and ALP each, with the following remainder (to the nearest whole percentage point):
|
Liberal
|
13%
|
|
ALP
|
1%
|
|
Greens
|
12%
|
|
Jacqui Lambie Network
|
4%
|
Although numerous provisions are in place to promote voter preference flows over party preferences (e.g. banning how-to-vote cards and no above-the-line voting), it is probably a sound assumption that the ALP votes will predominantly flow to the Greens, making it a close contest as to who wins the 5th seat in each division between the Libs and Greens.
Fortunately, we have division-specific polling from the February 24 ReachTel Poll:
Using this data, we can first look at which seats are filled by a quota on the primary vote, and the remaining support:
In all cases, the ‘Others’ votes will be redistributed first (except possibly Denison, and even then, only by a highly unlikely distribution of the entire ‘Others’ vote to a single candidate). Since ‘Others’ is a diverse and unpredictable group, I will redistribute these according to the popularity of the other parties (though I have a hunch the actual distribution may favour the Greens and JLN disproportionately over the major parties, since these are likely to be voters engaged beyond the Lib vs Lab headlines in the media and discontent with both major parties). The remaining vote then looks like this:
And then assign one extra seat in Lyon is awarded where this pushes the Libs over the quota.
This leaves one seat in Bass, Braddon, Franklin and Lyon to assign.
Obviously, the EMRS polling is only three days more recent, so the changes in the polling numbers are more likely to be random variation than genuine changes in polling (all are within the usual 2-3% margin of error.) However, for the sake of capturing the full range of possible data, I have replicated the process after adjusting* the ReachTel data.
(*To adjust these figures, I would normally look at the proportion of the vote that shifted Tasmania-wide and apply this to the data, e.g. the drop in the liberal vote from 48% to 46% would result in the liberal scores across the board dropping 2/48ths of their total. However, as the ReachTel data will remain the most accurate division-level data we have and this is just to check for any differences that might occur, I took the more extreme approach of dropping all liberal scores 2 percentage points. If there is any change of note, this should definitely reveal it.)
From this starting data:
We get the following results:
The only practical differences here are putting one Liberal seat in Denison in doubt and awarding one ALP seat in Lyon. (Note, although the ‘Others’ outpolled the JLN in some cases, these were eliminated first as they are likely to be split between multiple candidates and would not necessarily flow between themselves).
Bass
In both analyses, Bass has 3 Liberal and 1 ALP members elected, though in different orders. In both cases, the Liberal party will be the next to drop out. In the EMRS-modified data, the ALP narrowly leads the race for the final seat; in the directly sourced ReachTel data, the Greens have a narrow margin ahead. In either case, short of an implausibly homogenous preference flow, the ALP or Greens will need preferences form the Libs and JLN. Additionally, given the nature of the parties, a solid preference flow from the Liberals may be enough to push the JLN into second place, outlast the third-place candidate and win. This last seat becomes a three-way race, and would be a prime candidate for a tossup if I still gave those out.
IF LABOR LEADS THE 4-WAY RACE FOR THE LAST SEAT (EMRS-(POORLY-)ADJUSTED DATA): JLN would need near 100% flow from Liberals to not drop out, leaving the contest between the ALP and Greens. 100% preference flow seems unlikely. JLN is a ‘big tent’ party, and its preference flows are unpredictable. It is also hard to tell if Liberals will favour the Greens to spite their main rivals in the ALP or favour the ALP over the even-more-left-wing Greens. This remains a mess, difficult to call and is based on hastily manipulated data, but if Jacqui Lambie’s network was to win any seat it would be this one. My true prediction for Bass, though, is as follows:
IF THE GREENS LEAD THE 4-WAY RACE FOR THE LAST SEAT (REACHTEL PREDICTION): Liberal preferences will strongly favour the JLN, possibly enough to push it into first-place (this would require ~50% flow to JLN, which seems highly plausible to me). Labour will drop out next unless there is a strong flow from the Libs to ALP over Greens, and while the ALP vote will bolster the Greens, some ex-Lib votes will go to JLN. Greens will breach the quota from the ALP preferences and take the final seat
Braddon
In both analyses, Braddon has 3 Liberal and 1 ALP members elected, though in different orders. In both cases, the Greens will be the next party to drop out. In the EMRS-modified data, the ALP leads the race for the final seat; in the directly sourced ReachTel data, the Liberals have a two percentage point margin. Greens preferences should aid the ALP to overtake the Libs in this latter case, and the final decision will fall to the preferences from the JLN. Being a big-tent party this is hard to predict, but I would expect almost all of the Greens votes to land with Labour directly or via Lambie’s network, so that anything over a 1:2 flow from JLN primary voters should see the ALP take this seat
Denison
On the ReachTel data, this electorate is filled by 2 Libs, 2 ALP candidates and 1 Green. The dubiously modified data to accommodate the otherwise trustworthy EMRS poll puts one of these seats in doubt, but only just below the quota with 16.3% of the vote in their pocket, 16.7% needed and a two percentage point lead on the ALP. Due to the rough way the ReachTel data was beaten into a vaguely EMRS shape, the JLN is almost non-existent—no doubt the reality will see more than 0.1% of the 4-way vote fall to this party and possibly overflow to the Libs. It seems a pretty safe bet that the Liberal party will take out a second seat in Denison.
Franklin
Both analyses agree that Franklin should elect its first four seats as Liberal, Labor, Liberal and Labor, with the Greens holding patiently somewhere in the mid 15% out of a needed 16.7%. Liberals will be next in contention but a long way behind. Even with 100% flow from the JLN, the Liberals will remain in second place and the ALP will presumably favour the Greens to give them the final seat.
Lyon
On the ReachTel data, three Libs and an ALP MLA will be elected. The adjusted data suggests another ALP may be added to this, and the ReachTel data has the ALP ahead of the Greens by 3 percentage points. The Liberal overflow is unpredictable beyond falling to the JLN briefly. Ideologically, the ALP is closer to the Liberals, but also commonly perceived as their greater adversary. Fortunately for our predictions, the Liberal preference flow is largely irrelevant; the JLN will play kingmaker in Lyon. Despite some similar policies on, for example, medical marijuana, I feel the JLN will scoop up a lot of disenfranchised right-wing voters who will hate the ALP but despise the Greens; I am giving this seat to the ALP.
Conclusion
I am predicting 13 Liberal seats—just enough to form a majority on their own—9 ALP candidates elected and three Greens MLAs this Saturday. If the Jacqui Lambie Network is to win any seats at all, it will be from the Greens in Bass and not affect the narrow Liberal majority.
