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

Twitter: @AU_Truth_Seeker

Tuesday, 17 September 2013

Senate results - 16-9-13 - 10pm update

Hello again - welcome to my latest 10pm update. This time, it's slightly later than 10pm... Also, I have chosen to focus on the State with the closest Senate election.

Here's the forecast senate outcome:

In order to keep providing interesting yet truthful and meaningful stories, tonight I will focus on the state with the closest Senate race: Western Australia.

I have mentioned previously that the Sports Party Candidate, Wayne Dropulich, may be in trouble. According to the ABC Senate calculator, he currently leads by 195 votes at count 11/12. At this stage of elimination, his vote includes preferences from Australian Voice, and he is very close to being eliminated by Rise Up Australia (RUA). For Dropulich to remain in the count, the following equation needs to hold:

Sports + Voice > RUA

Furthermore, you can consider the surplus by vote type:
Ordinary votes: +200
Absent votes: +9
Provisional votes: (none counted, yet)
Pre-poll votes: -1
Postal votes: -13
Total lead: +195

However, it is worth noting that this is before preferences. Helpfully, this calculation is rather simple, as all we need to consider is the preferences of the Voice Party. It is likely that WA will have a BTL% of approx 3%. But analysis shows that BTL% is higher for micro parties such as Voice. So it is not unreasonable to suggest a range of 10-20% BTL for voice. What happens when we factor this in to the calculation? Let's assume a midpoint 15%.

Ordinary votes: +58
Absent votes: +7
Provisional votes: (none counted, yet)
Pre-poll votes: -2
Postal votes: -15
Total lead: +48


Next we need to extrapolate for as-yet-uncounted Declaration votes. All we can do is assume that they break the same way as other declaration votes, and I cannot find on the AEC website a breakdown of uncounted Senate Declaration votes. This is unfortunate. (If you can find this, please send me the link!)

So we take the adjustment in the Sports lead due to declarations (+7 - 2 - 15 = -10) and we then multiply this by the ratio of total declarations remaining (ratio 5.9 to 1, at time of counting). So this means that after all the declarations are done, the declaration will contribute an additional 60 votes away from the Sports Party. In summary this means that Wayne Dropulich will lose the Senate by 12 votes! :-)

Obviously, to make this statement with 1.5million WA voters and several hundred thousand votes are uncounted is outrageous. But, what this does imply is that it will effectively be a coin-toss as to whether Dropulich gets eliminated at count 11 or 12, or whether he remains in. If he remains in, he gets elected. Simple. But if
-Postals continue to be unkind to Dropulich
-RUA vote is higher in Postals
-Voice vote drops away
-Voice BTLs are higher than 10-15%,

1. Wayne Dropulich will drop out of contention.
2. Zhenya Wong will win a third Senate Seat for Palmer United
3. Louise Pratt (ALP) will almost certainly shave Scott Ludlam (GRN) from the final seat.

So it appears a 50-50 shot that a PUP and an ALP senator will replace the SPORT and GRN senators that everyone thinks has already been elected...

I have set up a model auto-linking to AEC data feeds. Although I cannot easily write a long post during the day owing to work commitments, I will monitor regularly and tweet (follow me @AU_Truth_Seeker) updates if this scenario changes at any time during the day. If so, I will also do a short post here too.

I ran my Monte Carlo analysis to look at the percentages of election:
PUP: 32%
ALP: 32%
GRN: 68%
LDP: 8%

When we have 1,500,000 potential voters, and the current margin is +/- a dozen, this is amazingly close.

Anyone care to make a prediction?


  1. We can already see that of the votes that have been apportioned for Aust Voice:

    516 Ticket votes
    16 BTL votes

    That's just on 3% there, with 447 left unapportioned.

    Why do you think 10-20% will vote BTL? I would think with a metre-long ballot paper and a record number of candidates the BTL vote should be lower than ever...?

    1. These figures are incomplete. Is it true that votes are first assigned to tickets, or when they are BTL they are set aside until they undergo data entry? Until data entry confirms them as formal do they stay in unapportioned? Look at for example. Do you really believe only 3 green votes out of 2000 are btl?

      10%, possibly higher is where id expect Voice BTL to be.

      Truth Seeker

    2. No, but that was a ridiculously early example to cherry-pick. If you look at today's update about 50% of the ordinary votes have been apportioned and the BTL rate for the Greens (quick eyeball) is about 2500/47500 - a bit more believable. Over 50% of the total Voice vote counted to date has already been apportioned. Okay, fine, maybe a higher proportion of the votes left to enter were BTL, but surely a random sample of ~ 50% has a pretty small margin of error (Dr Bonham's point below about sample bias notwithstanding). The BTL fraction of the remaining half would have to be much higher to balance out the 3% already counted.

  2. The Ticket/BTLs at the moment aren't in proportion to where they will end up. For instance looking at Tas where I know there is a c. 10.5% BTL rate, the BTL rate in the current figures is only about 5%. There must be a higher BTL %age in the unapportioned vote for Tas. I'd expect the same for other states too.

    Regarding declaration votes: what I've been doing for Tas is assuming that any declaration vote processed for the House is also a Senate declaration vote from the same elector, and then adding up the declaration vote totals for the five Tas electorates and comparing to the totals already processed in each category for the Senate. However this is much more tedious for a big state.

    A bit of a trap is that in considering samples of already processed declaration votes it is worth keeping an eye on where they come from. For instance the 1100-odd EPPVs processed for Tas Senate come from disproportionately good ALP/Green areas, which makes them disproportionately good for PUP. Could scale them off the electorate data but probably not worth the time when the base samples are small and I'm confident a valid projection of the LNP/PUP point is going to land in the grey area where I'll really need to know the rate of BTLs for the feeder parties.

  3. Anotther way of looking at it - on the current count (75.79%) Sport + Voice are 195 votes ahead of Rise Up. Voice have 979 votes. 195/979 is about 20%. So, unless there is a substantial change in trend for the late counts, the Voice BTL% needs to be 20% or over for the Sports party to lose their kickstart.

    2010 WA BTL figures for religious right parties:

    CDP ~ 8%
    FF ~ 6%
    DLP ~ 2%

    I take Kevin's point about sample bias, but:

    a) Over 50% of the counted votes for Voice have already been apportioned, and only 3% of those are BTL
    b) How do you pick what is likely to be a good or bad area for Voice, Rise Up or Sports? They are such low numbers that they'll either be effectively constant or effectively random

    You are of course right that the unapportioned count could contain a higher BTL fraction (staying in the hard basket a little longer) so I could be wrong, but there are enough "comfort factors" above for me to favour Dropulich + Ludlam over Wang + Pratt.

    1. Firstly there is a strong negative correlation between log(vote) and BTL%, and this is a negative correlation. Because Voice will poll such a small number, the BTL% will be higher than equivalent parties. Also, in 2010, there were 23 groups in WA. This time, there are 27 groups, so not significantly more.

      10% may well end up being higher than actual, but my point is we won't know as the figures on the AEC website are not representative.

      Again, I am not saying that Pratt/Wang will get elected, I'm saying it's on the knife edge and effectively a borderline call, with less than 100 votes the difference out of ~1.5 million votes total.

      (Today's new numbers have slightly favoured Sports, for what it's worth. So they're up an additional handful or so). But still very tight.

  4. OK. I agree that the number of groups hasn't gone up a lot. And I just plotted the BTL% vs total votes and can see the very clear trend on a semi-log chart (though FF and DLP do sit on the low side of the cloud, they both have an order of magnitude more votes than Voice).

    Nonetheless I don't see it as a "100 out of 1,500,000" thing. The total number of votes cast by Sport + Rise Up + Voice is 6385 so far and not going to get a lot bigger with the last fraction of the count. The 195 vote current margin is about 20% of the current Voice vote and about 8.5% of the current Sports vote.

    There could be some other close cuts further down the track beyond this first important step as well so you are absolutely right in not wanting to call it either way.

    1. Yes, I did the same analysis on the weekend: semi-log vs BTL, broken down by state and "ideology" (Left, Right, Christian, Central, Labor, Liberal, Green)

      Also consider that probably each and every BTL vote will get lost, and go to other parties instead. So these should be deducted from the Voice total. In practice, as many BTLs will come back through Sport as through RUA, so we can just ignore BTLs in our calcs.

      Note that postals are going backwards at the rate of 1 in 1500 for sports, and absents are going forwards at 1 in 1500.

      The other hurdle for Sport is the previous elimination. Voice votes go first to No Carbon Tax so if Sports are knocked out here they never see the Voice votes. NCT is polling very well in absents, and if they poll much better Sport will not even make it past NCT.

      At later points, I don't see much getting in the way of a Sports victory...

    2. Sports vs NCT+Voice is only slightly closer than Sports+Voice vs Rise Up, 185 votes vs 195 (going on close of play last night). Same issue with the possibility of BTL leakage from Voice (except in this case every vote that leaks reduces NCT's chance of overtaking Sports).

  5. Oh, and to eat up today's margin of 195 votes by Voice BTL leakage:

    Ticket 531
    BTL 16
    Unapportioned 434
    Margin - current BTL = 195-16 = 179
    [Margin - current BTL] / unapportioned = 179/434 = 41.24%

    So - that's the fraction of the current unapportioned Voice vote that needs to go BTL to change the result.

    Not allowing for any of the other (perfectly valid) factors mentioned by yourself and Kevin.