Blogging Senate forecasts and results in the WA Senate re-election until officially declared.

Twitter: @AU_Truth_Seeker

Sunday, 29 September 2013

WA Senate - Below The Line (BTL) analysis

Tonight's update focuses on WA - the Senate race shaping up to be closer than Tasmania.

WA Senate (92.89% counted)
The most critical count in the WA Senate race is the Shooters vs Christians, as I'm sure you will recall. If the Shooters lead at the critical point, ALP & PUP are elected. If the Christians lead, GRN & SPORT are likely elected. In essence, this represents a case of who will be the lowest out of these two parties when a mammoth 10-party preferred contest is considered. So it's not really who's winning, it's who's not losing. (Ten parties are ALP, GRN, LIB, PUP, NAT, LDP, HEMP, SPORT, SFP, Christians)

The way to consider this is as follows: For each ballot paper, count the highest preference of the 10 parties. Eliminate candidate with lowest vote.

Accordingly, this critical count position can now be estimated with different methods.

Method 1: All votes are ticket votes
Historically, this is the way that online calculators have estimated Senate outcomes. It tends to give false hope to parties that have successfully negotiated a preference snowball to attempt to get elected.
Applying this method, the Shooters currently lead by 413 votes. This can be seen on Antony Green's Senate Calculator.

Method 2: Only count locked in votes
This method has been used extensively by myself, and intermittently by others including the ABC, Kevin Bonham and others. This count deducts BTL votes from ticket votes received, so hence only includes:
- Ticket votes for the party
- Ticket votes for parties preferencing the party in question
- All BTL votes for the #1 candidate (or other critical candidate, if required)
- Most BTL votes for the #2 candidate, I have assumed 90%
- Unassigned votes, since almost all of them will become ticket votes once assigned.

Applying Method 2, the Christians currently lead by 48 votes.

Method 3: Include Locked in votes, but apply a modifier for estimated BTL flows

By my calculations, 94,457 votes for other candidates need to be reassigned. Of these, 85,419 are ticket votes with known preferences and 9038 votes are BTL votes with unknown, but estimable, preferences. We can further categorise the BTL votes by party ideology:
- Left: 5207 votes (58% of eliminated party BTLs)
- Central: 717 votes (8%)
- Right: 2260 votes (25%)
- Religious: 854 votes (9%)

After undertaking analysis of the 2013 Tasmanian Senate distribution of preferences, I have estimated the percentage likelihood of each of the 10 remaining parties getting the BTL preferences of each eliminated party. This analysis was crazily tricky, owing to different parties and different orders of eliminations. But making just a few assumptions at a couple of points I was able to estimate the vote transfer from each party or ideology, to each of the 10 remaining parties.

For the wonkish types, the full table is as follows:

BTL votes
Smokers Rights
Socialist Equality Party
Australian Voice Party
Sex Party
Secular Party of Australia
Australian Independents
The Wikileaks Party
Katter's Australian Party
Family First Party
No Carbon Tax Climate Sceptics
Stable Population Party
Stop The Greens
Australian Democrats
Animal Justice Party
Australian Fishing and Lifestyle Party
Australian Motoring Enthusiast Party
Rise Up Australia Party
One Nation

To understand this, Smokers Rights has registered 599 BTL votes. Of these, an estimated 15% will flow to the Shooters, 5% to the Christians and 80% to all other parties combined. Given we are just after the Shooters v Christians gap, we can ignore the 80% that will scatter and focus on the votes that matter.

This means that for the current BTL vote, the Shooters will receive a net advantage of approx 184 votes, when compared to ticket votes.As far as I know, I think this numerical application of this method is unique. It takes the method 2 output (ticket only votes) and adds estimated net BTL flows between the two parties. This way, all votes are accommodated in a reasonable manner.

Calculation of the Method 3 modifier is as above, and this adds a net +184 vote lead to the Shooters, reversing the Method 2 lead of the Christians.

This is my first assessment involving a numerical and quantitative estimate of BTL votes. Currently, the Shooters hold a lead of 148 votes, now making it likely that ALP & PUP will capture the final two spots. 

(I have previously analysed two other critical points: LDP vs PUP and AJP vs HEMP. These have been trending away from a "flip" such that the margins are respectively 1774 votes and 947 votes. I won't spend time discussing these tonight).


  1. Hi there, nice work again. I guess the ACT and SA counts _might_ be done before WA so you might be able to add more data to your Left / right/central/ religion preference splits.

    The gap changes above meet the 'seems right' gut feel.

    I think ballot order might mean more Rise Up votes get locked up with the Libs who are in the second column to their right (and I think all Rise Up lower house candidates were preferencing Libs 2nd, - no idea how many Rise UP party workers were handing out HTVs in WA).

    I suspect Fish and Lifestyle + Motorists will also get more votes locked up with Sport who are also in a column just to their right - I guess that probably does little to the relative gap between the competing parties facing exclusion.

    I suspect people voting BTL have a few parties in mind that they like, to preference high, and a few that they really dislike, to preference low, and the rest just get a scattergun in the middle.


    David B

    1. Looking at the declaration schedules ACT is scheduled to be declared on Tuesday; Vic, SA and WA at various times on Wednesday. I don't know exact details of button-press times but it does not sound like there will be much if any time between Vic, SA and WA.

    2. @ David: the ACT Senate will not give me any worthwhile meaning for my speculative 10-party preferred Shooters v Christians lottery - this is because there are not enough parties to add to my datasets to increase my data quality.

      With the full SA BTL dataset, meaningful analysis could be gleaned. But it is unlikely this will be made available in reasonable time to feed into the WA analysis.

    3. Hiya, yes, for some reason I thought there were several days between the various State declarations. Won't have long to wait to see what comes out in the wash.

      David B

  2. Thanks Truthseeker. Just wondering why you would think that FF would only see 15% of the BTL votes go to the Christians? Or am I missing something on the Religious right? Perhaps they are more likely to just preference Liberal....

    1. Christian were a completely new party in Tasmania, so the FF first, Lib second would be what a lot of (religious right) people headed in to the booth planning to do.

    2. Hi Anon & IntReason

      I have tonight used two different sources for my estimate of WA BTLs. Tas FF was flowing very strongly to Lib and less so to Christians, but of a very small sample size. The WA 2010 FF BTLs flowed very strongly to CDP, which I am using as my best guess proxy for the 2013 Christians. So you will note I have revised this figure up.

      Also, my modelling does not (and will not) take into account ballot position - too tricky to incorporate...

  3. I am not sure you can ignore the 80* scatter. they must land somewhere and question what scenic rout will they take. if the pass though the LNP/PUP tickets then they are devalued disproportionally die to the calculation of the Surplus transfer value.

    I sue a system based on your model C. Loced in votes vrs flaotring percentage of BTL votes.

    BTL votes belong to the Greens and ALP are locked in.

    Those belonging to PUP/ ALP and LNP are also locked in for all intensive purposes.

    The Holly trinity also flow within their grouping,

    The aim of the modelling applying current count riles is to distributed the percentage of crucial BTL vote distributions stopping when the nominated drift produced=s a tipping point. IE the results of the count change and then ascertaining the likely hood of that percentage drift.

    Its like using the ABC calculator but drilling down two to three steps in resolution.

    I see the main point of change being Christian and Fishermen. And an outsidce contest between LDP and PUP

    The Greens can not be elected without Sports surviving the count. I do not think they can attract sufficient number of votes.
    Has the AEC published t6yhe BTL preference data file progressively we would be in a better position to determine the outcome. Without having to make wilds assumptions as to percentage of each groups BTL votes.

    The publication of the BTL preference data files would facilitate an open and transparent count

    Te quality of the AEC count sheets has little to desire and lacks crucial information.

    The Greens have more BTL votes likely to drift then the ALP