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

Twitter: @AU_Truth_Seeker

Thursday, 12 September 2013

Senate results - 12-9-13 - 10pm update

Welcome to my latest daily 10pm update of Monte Carlo simulated Senate forecast.

The forecast summary is as follows:

Note that my model has now calculated a new winner in the 6th seat in Tasmania. However, please note that this is not taking account of Below The Line (BTL) votes which will leak to the PUP candidate.
I have downloaded the full 2010 BTL preference set for Tasmania and attempted to analyse them. But my fruits have been without success tonight, as I have successfully crashed Excel several times. But I will persist, and hopefully will have time to complete this over the weekend so please keep looking back here!

Daily changes in vote:
Total: 84.67% to 85.59%, the highest percentage counted out of any state Senate
ALP:  -0.13%
LNP: +0.04%
PUP: +0.02%, stopping the steep decline
SXP: +0.01% - edging closer

The model is now saying that the LNP is most likely to win the 6th seat, but this may not be accurate. The PUP candidate, Jacqui Lambie, is also under threat from the SXP who now sit behind by just 233 votes (0.08% at count 21). If Robbie Swan (SXP) gets ahead of the ALP at this point, it is very likely that he will win the final seat, and give Tasmania a remarkable 4-2 Left-Right split. My model is giving this likelihood as 20%, although this may be overestimating Swan’s real chances.

Note that Kevin Bonham's blog represents the best quantification of the impact of below the line votes in Tasmania.

Daily changes in vote
Total: 71.62% to 74.15% - a relatively large increase in counted vote
ALP: -0.11%
GRN: -0.06%
LNP: +0.04%
LDP: +0.04%

The model keeps reducing the likelihood of the 3rd liberal candidate (Arthur Sinodinos), now down to 77%. However, it still appears likely that he will retain his seat as my model is using relatively large variations (each party’s vote numbers vary by within +/-3% in each Monte Carlo simulation, implying a range of 34.6% +/-1% in the LNP vote).

Alternatively elected candidates, as per my model, are either the Democrats with 0.21% of the vote and a 10% likelihood of winning, and the Shooters and Fishers (also a 10% likelihood of winning). I wouldn’t be holding your breath about the 2% chance of the Greens winning!

Daily changes in vote:
Total: 69.3% to 71.0% - a non-insubstantial increase
LNP: +0.11%
ALP: -0.09%
GRN +0.06%, against the trend
Apart from a sub 1% chance of the third coalition candidate (Helen Kroger) and Family First upsetting the AMEP, the six senators appear locked in in Victoria.

Daily changes in vote:
Total: 67.2% to 68.52%, a mediocre daily increase in count
LNP: +0.05%
GRN: -0.04%
KAP: +0.04%, possibly reflecting a late swing in vote away from KAP to PUP, with today’s counted votes being cast possibly 1-2 weeks prior to the election.
PUP: -0.05%, opposite of above.

HEMP’s likelihood of election is now up to 1.2%, but this is still too small to be seriously considering it. If it gets much higher, I’ll undertake a full analysis.

Daily changes in vote:
Total: 64.08% to 64.83%, little extra counted today.
No party’s vote changed by more than +/-0.02%

The Sports Party is the interesting conundrum in this state. My model is not giving Wayne Dropulich 100% chance as he sailing close to the wind in early stages of elimination. At count 12, Dropulich leads the RUA candidate by just 250 votes, including 820 votes from the Australian Voice Party. So, as long as there’s less than 30% of BTL votes for Voice, Dropulich should be safe. Looking at historical data, 10-20% would be my maximum BTL expenditure of Voice. So this may mean that the Sports Party is safe, although it would reduce its margin at this point to a paper thin 300-600 90 - 170 votes (UPDATED) (out of 1.5 million!!!). Oh, our Democracy just keeps electing the most popular candidates!

Daily changes in vote:
Tocal counted: 75.17% to 75.30% (minuscule)
No Carbon Tax: 0.08% to 0.11%.
This is highly unusual, it implies the NCT got approximately 17% of all vote counted today. Or there was
- a prior error that was corrected, unlikely given the stability of NCT at 0.08% each day until today.
- a new error that was made today.

The reduction came at the expense of all major parties, which lost 0.01% of the vote each, and the only increase was NCT. This is bordering on bizarre, however the total number of votes counted today increased by just 1000… Odd

With this tiny increase in its vote, my model is now giving NCT a whopping 18% chance of election as NCT can avoid early elimination. The second hurdle for NCT is count 29. If FF are eliminated before LDP, then this gives NCT a free run to victory. The margin now is 0.26% or 2100 votes, but with both sitting on the same proportion of primaries. For the mathematically minded, the calculation is:

FF + AI + SPP > LDP + Smokers + Group L 

If this equation is True, FF win a Senate spot in SA. If false, NCT win a Senate spot in SA.

Next update: Approx 24 hours, 11pm AEST ! See you there!!


  1. As updated on my site I have pulled back from considering Tasmania a likely to very likely PUP win and now think it's unclear. There are two reasons for this:

    1. Labor going backwards, apparently off a c.500 vote hit on ordinaries.

    2. I was previously modelling early out-of-electorate prepolls off out-of-electorate PPVCs. There is no real alternative to this for PUP, but assuming it was valid for Labor was a very bad move - in 2010 PPVCs were better for Labor than out of electorate prepolls by 5.5 points! The difference is worth about 600 votes.

    Both of these hit PUP's chances at both the ALP/Sex Party exclusion point and the LDP/PUP point, because Labor's surplus goes via the Greens to PUP so any vote Labor loses PUP loses too.

    There is also oddness going on with vote totals appearing then disappearing on the AEC site in specific electorates.

    1. Thanks for the update and your continued observations of this Senate contest.

      Part of me likes the notion that 2-3 States could come down to btls. Would be ironic given the best efforts of the obfuscators to make it damn hard to avoid their harvesting skulduggery, by running a million candidates.

  2. From where can one download the full 2010 BTL preferences for TAS? I'd be very interested in seeing them.


      Please share any results you may come up with, I'd be most interested and I suspect Kevin would too!

  3. We had a BTL-sensitive situation in Tasmania in 2004 (win to Family First on calculator, actually a win to the Greens) but it was an absolute sitter to model compared to this one since it was a straight race in two. It was clearly over well before the button was pressed since the Greens only needed a 21% split of BTLs from a wide range of parties.

    Modelling the BTLs is going to be painful this time because a lot of the parties being excluded with significant BTLs have no track record in the state. Likewise for parties receiving them. And it will be necessary to look at one party needing to gain X votes over another out of Y BTLs in a crowded field, not a two party contest. But once the votes are all assigned as ATL/BTL it will be possible to start on this for the two key points in the count. (At least, I hope there are only two!)

    Labor are going to suffer on the pre-polls but none have been counted yet.

    1. Three key points? FF vs SXP, SXP vs ALP, LDP vs PUP!

      Ironic that the BTLs are likely to be so important given micro party apparatchiks went out of their way to nominate so many candidates to increase our % of voting ATL!

    2. Yes and the FF vs SXP key point has an FF vs ALP key point if FF get over it.

      PUP seem to have slipped backwards on the ordinary count again, so even although they did well on absents today they still fell further behind LDP on the calculator. Nothing has gone right for them in the last day and a half. They'd still win on BTLs if the button was pushed right now but I can see that LDP lead blowing out towards 2000 and then it gets really dicey.

      An interesting thing is that the Sex Party did massively better on the day than on postals. Hope I don't flatter myself too much by suspecting my campaign to convince people to not vote ATL for parties that fed Family First (which was everyone except Sex and Pirates) had something to do with that. :)

    3. Perhaps post-election, you can test this theory by looking at the difference in SXP votes in each state by vote type! Throw in 2010 election for a sample size of 12, and you've got a bare minimum reasonable data set.

  4. IN Qld, at the point where HEMP gets excluded (Count 38 on the ABC Calculator) they are fairly close to Fishing & Lifestyle. I don't know if F&L dropping out before HEMP changes the likely outcome much

    1. Hi Andrew, Thanks for your interest in my analysis.

      I have analysed three different possible "nexus points"
      1. Count 37 - does HEMP overtake AFLP?
      2. Count 36 - does PUP get to quota?
      3. Count 35 - does KAP exceed AFLP?

      All 3 make no difference, if we assume 100% ATL votes. The only ways my Monte Carlo simulations find a win for HEMP is if ALP dips below 2 quotas. At the moment, they are above quota by about 0.27%. ALP vote has been dropping recently, but in reality any figure above 28.3% is likely to get ALP over the line, after BTL leakage.

  5. Re: Andrew Bartlett's comment. My measure is that HEMP needs about 24% BTL leakage on current numbers to overtake AFLP. But it is also likely that PUP passes quota later in the count than currently projected, which will eliminate the GRN's in PUP's favour with the excess going to the LNP.

    In WA, it is extremely unlikely SPORT will pass quota before PUP is excluded (less than 1% BTL leakage) which would mean the GRN being elected 5th on PUP preferences and the bulk of the excess going to the ALP. That puts the ALP is a much stronger position than the calculator would suggest.

    The other thing I'd add, in the booth I voted in (Southbank, Melbourne Ports) they only counted GRN, ALP and LIB bundles and put the rest under informal. If that is a common practice for a first count in the Senate then it will understate the minor party vote by quite a bit.

    1. Russ - re southbank, I have looked at the Melbourne Ports electorate Senate votes. Believe it or not, only Southbank appears to have done this and no other booths in Melb Ports that I have analysed display this happening.

      Re the Qld election, I will look at it tonight.

    2. Russ Re WA: Yes I've noticed the Sports Party may fall below a quota, however all it seems to do is change the order of election, ie Greens get 5th Senate Spot, Sports get 6th Senate spot.

    3. Yes, exactly. I doubt Sport will get to 14.53 where indicated, I'd expect leakage of about 0.3% leaving them just a tad short at count 29 but Sport are most vulnerable at count 11. My update today shows they lead by an estimated 83-166. So if I was a WA PUP scrutineer, I know whose piles of votes I'd be looking at VERY closely!

    4. Note that I didn't say the ALP would win, only that it is much closer than implied by the calculator. Possibly as little as 2000 votes. Postal and pre-poll votes are trending strongly away from PUP, ALP and GRN towards the LNP though, which will ultimately help SPORT get over the line.

      Incidentally, how did you come up with 0.3%? If you model ~10% of non-first-preference votes as BTL (which is typical for minor parties even on big ballots, use ~2% for ALP/LIB/GRN), then randomly redistribute those vote to whoever is left, how does that effect the various counts?

    5. Russ - my rough calc there was 15% x 2% = 0.3%. You're right in that it might be a bit higher, possibly double. My assumption at all "point in time" decisions is to ignore the BTL's, where these are preferences, by multiplying by the estimated BTL%. Only for SA, WA and TAS would BTLs make a difference.

  6. Russ, even if HEMP overtake AFLP, it will likely not change the result of the election. I can only find a path to HEMP winning if ALP falls short of 2 quotas, as explained above.