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

Twitter: @AU_Truth_Seeker

Thursday, 5 September 2013

SA Senate - TruthSeeker forecasts

Here’s the second state in my senate series – South Australia.

Most likely elected Senate (SA):

1 GRN, 2 ALP, 1 XEN, 2 LNP

As with Victoria, I have updated my calculation of primary votes to allow for the latest polls, and slightly tweaked my method for calculating minor party support to allow for more national consistency. Variation is also slightly higher to represent more uncertainty than was previously allowed for.

Primary votes:
LNP:       33.0% (+/-3.3%)
ALP:      29.2% (+/-2.9%)
XEN:      15.4% (+/-3.9%)
GRN:      9.2% (+/-1.8%)
NAT:      0.5% (+/- 0.2%)
Minor parties:       12.5% comprising selected parties
FF:      2.7% (+/-1.3%)
PUP:      2.7% (+/-1.3%)
KAP:      1.0% (+/-0.5%)
No Carbon Tax:      0.14% (+/-0.07%)

Likelihood of election:
LNP:      2 elected, +1% chance of 3rd senator
ALP:      2 elected, with 99% probability
XEN:      1 elected
GRN:      65% likelihood
FF:      4%
PUP:      2%
NCT:      29%
KAP:      0.1%
Everyone else: 0! (Sorry)

Ideological splits:
3 Left, 2 Right, XEN: 63% likely
2 Left, 3 Right, XEN: 37% likely

- Two ALP, Two LNP, almost certainly
- SA appears set to reelect Xenophon
- Greens likely to retain the last spot but...
- No Carbon Tax Climate Sceptics are a good chance to win the final spot, with less than 1500 votes statewide. Some would argue that for this to occur, the system must be broken.

As the final spot in SA has a reasonable chance of going to either the Greens or the anti-Greens, it could play in important part in a future senate. Perhaps, these parties may have different views on whether to repeal the Carbon Tax? J

In both the states analysed, there appears no chance for the coalition to make up the numbers to attain a Senate majority.

Over the coming days, I will publish equivalent summaries for other states. I won’t talk about the methodology – this is largely covered in prior posts. On election day, I will republish my full final results, updated for latest polling data.


  1. Great blog! Can you please do analysis for the ACT and Western Australia!

    1. Thanks. All states are coming. And hopefully territories too. ACT will be easy 1-1, NT may be different.

    2. You don't think the Greens have a chance of upsetting the Liberal candidate in the ACT? Would require the LIberals getting less than 33% but still seems a possibility what with proposed public service cuts in Canberra. Haven't seen the recent poll data tho.

    3. I don't think so. In order to get a swing we'd need a scare factor that is bigger this time than last time. I think the 2010 scare campaign was much more effective, and the general trend is for a swing to the Coalition. Accordingly, I don't see any chance that more people will be more scared about public service jobs in Canberra this time, especially when KR is anecdotally not universally loved by public servants!

    4. Also, we will never get a poll as specific as ACT senate voting intention. I doubt we'd even see a HoR reachtel poll, given the safeness of both Canberra and Fraser. So, we're just guessing, but my guess is an easy 1-1.

  2. You seem to have left out XEN in the likelihood of election list - could you add them in? I assume the probability is close to 100%

    1. Whoops. Yes, XEN is 100% - he got elected in ALL 1000 simulations I did. Will amend the blog post later this afternoon.

  3. IMHO XEN did poorly in preference deals so if he gets much below 13% he doesn't get elected.

    1. I applied a wide variance to XEN to account for uncertainty. The range over which I modelled XEN was 15.4+/-3.9%. So over a range of 11.5% to 18.3% he gets elected 100% of the time

  4. I'm surprised you've got such a low probability for the Libs getting three seats. I must have been using a higher primary vote for them. But I guess the polls have narrowed (and a wider range of right minors taking from Lib), so now the ALP/Green primary is well above the Libs.

    But interestingly, you've got NCT winning most of the time there's a 3rd right seat.


    1. I find the Liberal primary implausibly low. They got 35% in 2007 when there was a 8% "other" vote, Xenophon was running, AND they lost government. Yes, there are more minor parties now, but is it really likely that they will under-perform that result when the swing to their side in the last 6 years has been so substantial?

      I'm also with Adam (below), FF is well established in SA while PUP does not really have much presence, nor any candidate with a profile to match Glenn Lazarus. I would expect that most of their 4% primary in the national polls is being run up in the more populous eastern states, especially QLD and NSW. Of course that's all conjecture and it makes sense to model possible outcomes if their vote is higher.

      Thanks again for all your great work TS. :)

    2. Thanks again for your comments and continued support!
      In 2007, LNP got 43% in the HoR, Current BludgerTrack graphs put the LNP primary at 40%-41%, consistent with a -2-3% swing in HoR.
      In the senate, from 2007, there are more minor parties and Xen will poll a little higher, removing possible vote from the LNP, so a -2-3% swing to LNP from 2007 is reasonable IMHO.

      I acknowledge there is more uncertainty in SA, so I have applied a higher variance than other states. Also, note the NATs didn't run in Senate 2010 in SA, so there's an extra arbitrary -0.5% that my model deducts from the the Lib vote and gives to the Nats. (Perhaps, in my summary above, I should have relabelled LNP to LIB to make it more clear that the NATs are a separate (and more independent) entity in SA (&WA).

      My model assumes that LIB polls 36%-36.3% approx 10% of the time, and this means that even with this higher primary they are still unlikely to get their 3rd member up.

  5. Are there any polls of PUP popularity in SA. Anecdotally, I don't see or hear much about PUP. I have seen a few Clive ads in recent days, but their posters are weak and their candidates have no personal appeal or following I can see. I would have expected a higher vote for FF (something like 3.5% +/- 1.5%) and lower for PUP. I can't imagine KAP doing better than 1% on any view. I think a more realistic range would be 0.6 (+/-0.3).

    1. PollBludger has some fascinating graphs that show that "Other" vote in HoR is approx 12% - assuming it is true, it needs to be split widely.

      I appreciate your estimates and your knowledge of ranges, My estimates are just that which is why I apply a high range - yours may well be right! :-) But in the absence of any reasonable polling it's anyone's guess.

    2. I am very interested in your methodology. Is it automated? Would it be possible to scale it up to, say, 10K simulations over double the ranges. (I am not suggesting you actually do this) I don't know what is involved - I am just speculating that a probability analysis with so many variables [parties] needs to be both broader and finer grained to give a true picture. While your ranges are plausible (even likely), there are other plausible ranges and I would be interested to know how that changes the results. Obviously, every 0.01 counts. Thanks for your work.

    3. Sorry, I have gone back and read your methodology post. I now understand, I think. Do you think there is any merit in increasing the ranges for micro and minor parties?

    4. Thanks. My theory is that larger parties can be estimated more precisely than minor ones. So a larger percentage variation should apply to minor parties. Ideally, there should be a normal distribution with an appropriate Standard Distribution - this would give a more realistic set of results.