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

Twitter: @AU_Truth_Seeker


Friday 27 September 2013

And you thought the current Senate was chaotic...

After utilising my specialised Monte Carlo models to successfully predict the "barnyard" nature of the new Senate, I have remodelled outcomes that would be likely under a hypothetical double dissolution.




Firstly, I used actual votes and actual preference tickets of parties at the 7 September election to predict "point estimates" of likely Senate outcomes under a hypothetical double dissolution. Secondly, I then undertake Monte Carlo analysis by applying a small degree of variation to each party's vote in each state to test the sensitivity of outcomes to slightly modified votes of minor parties. This attempts to simulate the sensitivity of "point estimates" to potential, and as yet unknown, Below The Line (BTL) votes.

In almost all instances, a "half Senate" election is held concurrently with a House of Representative election. This involves the election of six senators in each state and two in each of the two territories, for a total of 76 Senators. But in instances where the Senate repeatedly blocks legislation, the Government may petition the Governor General to call a double dissolution. The most recent double dissolution was in 1987.

Under a double dissolution, each state elects 12 Senators. This means that the "quota" is 7.69%, instead of the 14.29% of a standard half-senate election. It is clearly easier for minor parties to conjure up 7.69% of the vote, including preferences, than it is to reach 14.29%. However, it is also easier for a popular government to achieve a majority in the Senate - to get 4 out of 6 Senators in a state takes 57.1%, whereas to get 7/12 Senators in a state requires a more achievable 53.8% of the vote.

Regular readers, including those following me on twitter, would be aware of how I have used financial modelling techniques (my day job!) to forecast potential outcomes in Senate elections. Tonight, I reveal the application of my model to forecast the likely makeup of the Senate, applying the assumption that all parties votes and group tickets would remain the same as the recent election. While not perfect, it is the best possible assumption we could make.

The "point estimates" of likely Senate outcomes under a double dissolution are as follows:

Double Dissolution likely Senate outcomes, based on 7/9/2013 election results:
(XEN = Xenophon Group, AMEP = Motoring Enthusiasts, ONP = One Nation, SFP = Shooters & Fishers, NCT = No Carbon Tax Climate Sceptics, AFLP = Fishing and Lifestyle Party, HEMP = Help End Marijuana Prohibition)

Regular readers will recall my statement that the final WA Senate spots are "too close to call". Presently, I credit the Green's Scott Ludlam and Sport's Wayne Dropulich with a 37 vote lead over ALP/PUP. But for interests of consistency, the "current" line above represents an ALP/PUP election. Note that the current vote difference represents a margin of just 0.003%.

In all bar one case, the senators elected in the 7 September half Senate election would also be elected in a full Senate election. But such is the absurdity of the current system that AMEP's Ricky Muir would not actually be elected in a full Senate election where twice as many candidates are declared elected. And in SA, the "No Carbon Tax Climate Sceptics" Party would win the new record for the party elected to the Senate with the lowest vote - 0.11% (approx 1100 out of South Australia's 1,000,000 voters!)

It is striking to note that all three major parties will go backwards under a Double Dissolution, with the "winners" to be a whole host of minor parties with their own single issues and agendas. No less than 12 parties would get a comfy seat on the Red couches at Parliament House.

Monte Carlo Analysis

In order to test the sensitivity of these outcomes, I additionally applied Monte Carlo Analysis techniques to determine likely different outcomes if parties' votes were to change minutely. In this scenario, I have applied a small +/-1.5% variation to each party's vote. This is large enough to produce different percentage likelihoods of election where the election is close, but small enough to output "100% likely" outcomes where the election is not close. This variation is not large - a party with 40% of the vote would vary in the range of (39.4%, 40.6%) and a party with a vote of 2% would vary in the range of (1.97%, 2.03%). The purpose of testing variation in votes is to consider alternative election scenarios in which a minuscule variation in vote may lead to a significantly different election outcome.

SA, TAS and the territories produce 100% likely results. But four states produce outcomes that may vary with just a very small change in primary vote:
NSW: Shooters & Fishers, and the Sex Party both have a 20% likelihood of election
VIC: AMEP has an 18% likelihood of election
QLD: HEMP has a 19% chance of election
WA: The state with highest uncertainty - the Nationals have a 37% chance of election whereas Sport Party has a 7% chance of election.

The AEC has now counted almost all the votes in the Senate - but not all. It has not released BTL preferences for most states either. Accordingly, these outcomes will be subject to minor variation as:
- final votes are counted; and
- final Below the Line vote data outputs are released by the AEC.

Summary
If we thought the current Senate was "unrepresentative swill", well we ain't seen nothing yet! A double dissolution held before significant Senate reform is enacted will result in a kaleidoscope of single issue and opportunistic senators and micro-parties.

23 comments:

  1. Very cool! Thanks so much.

    ReplyDelete
    Replies
    1. I think it's cool :-) I'm glad at least someone else does. Thanks!

      Delete
  2. John Madigan is out, too. Huh.

    ReplyDelete
    Replies
    1. DLP didn't do so well from prefs this time round - and less confusion with ALP name than previously.

      Delete
  3. Disagree. The ALP received 30% of the vote and deserve 76*0.3 = 23 Senators, LNP deserve 28 or 29. The double dissolution senate is more representative, not less.

    ReplyDelete
    Replies
    1. Yes, I think it ends up looking like a fair outcome. You have to remember the vote for 'none of the above' was huge this election.

      There are some aspects of the process that need to be fixed, but it looks like we may end up with a workable senate with balanced options on the cross this time around.

      Delete
  4. We may need to wait until the very slow Queensland count is finished to get a completely accurate fix on national proportionality. But on current national figures a proportional outcome would be Coalition 28-29, ALP 23 Green 6-7, PUP 3-4, LDP 3 (but mainly because of NSW ballot draw fluke), NXG 1-2, SXP 1, FF probably 1 and each of DLP, KAP, AJP, WLP, S+F, HEMP has enough vote to count themselves unlucky if they miss out.

    The projection is actually quite close to pure proportionality, but for state-based distortions like three for NXG, and FF presumably getting a bonus from preference-dealing. But a purely proportional national Senate would be a very messy anyway.

    ReplyDelete
    Replies
    1. When you say "ballot draw fluke"... I think that's a serious understatement, and only part of the issue.

      Of the 5 states where LDP had a ticket, compared to the Liberals:

      - One was ideal (LDP in first column, Libs buried in the middle of a long ticket) - NSW
      - Three were at a distinct advantage to the Libs - WA, SA, Tas
      - One was the opposite of ideal (Libs in 2nd column, LDP buried in the middle of a long ticket - QLD

      The LDP vote was:
      - Nearly 9% with the ideal ticket position (cf Lib)
      - 2-3% with the good ticket positions (cf Lib)
      - 0.7% when the Libs scored the ideal ticket position

      Never in history has the Donkey vote premium been 8%.

      It is almost ALL in the name, and I am certain they know it. They swindled about 8% of NSW and maybe 2-3% of the other states where they had an advantage. In Queensland they were simply one of the nobodies - much closer to where their baseline should be.

      Delete
    2. It's quite a crazy situation - 8% of NSWers clearly accidentally voted for not the party of their choice. I would be interested to see if this high distribution of LDP votes was concentrated in areas with high rates of ppl born overseas, or in low socioeconomic areas...

      I wonder what the outcomes would have been had the LDP vote instead been the simple 0.7% in each state... This can be fact checked by Truth Seeker! :-)

      As much as I would love to model what VIC would have looked like with an LDP preferencing SXP, unfortunately, it is too hard to tell...

      Delete
    3. Oh sure. Perfect storm of confusing name, best possible ballot paper draw and enormous ballot paper. Probably needed all those things to work. Ballot draw premium without a confusing name is negligible.

      That said, I don't think the Liberal Party should be allowed to monopolise the term "liberal" without substantively being a liberal party.

      Delete
    4. But to what extent should we actually ensure that parties represent the traditional dictionary meaning of their name? And what if the LDP could mount a stronger case that they were more liberal than the liberals?

      I think the easiest way is to stop parties registering confusing names, defined more broadly than previously. Secondly, parties should be listed on the Senate ballot Left to Right based on their vote in the state in the last election. This would minimise the opportunity people have to confuse names and make it easier for people to find the name of the party that they want to vote for.

      Delete
  5. The idea of the senate as the state's house hasn't existed in reality for 50 years or more, bar a couple of state based independents (Harradine/Xenophon). National proportional representation designed similar to the NSW or South Australia upper house systems is the way to go.

    ReplyDelete
    Replies
    1. National PR... Could be interesting. I think you'd also need a threshold though. Also, I would like to see a reform that places the most popular parties from last time towards the left of the ballot. It is undemocratic to work through 40+ columns to find a simple short word that conveys your party.

      Delete
    2. That just means it needs to be fixed to address that, not that it needs to be less so.

      Best option would be to form new states for each of the capital cities and their urban surrounds, and give them 6 senators each (3 each election).

      Other than that, making senators subject to recall elections by the upper house of their state would also be useful.

      Delete
    3. The recall stuff is interesting. I think I prefer my state and federal politics separate. My suggestion is to "fix" the ballot positions in, say, 2016, based on 2013 vote in the particular state.

      For example, in Tasmania in 2016, this would see:
      Group A: Liberal
      Group B: Labor
      Group C: Green
      Group D: Palmer
      etc...
      This would ensure that 88%+ of voters can easily find their ticket, and would eliminate the issue of voters spending too much time and still not being able to find what they were arfer

      Delete
  6. Do you know when the AEC will have the finalised data up? When they do, I might do some analysis and mapping of the per polling place figures.

    ReplyDelete
    Replies
    1. Hi Eric, I note in 2010 the AEC has made available the full Senate 1st preferences by polling place in an easy format. However, the best the AEC has done for the Senate in 2013 is breaking out the results by electorate. Obviously, this is unsatisfactory, so short of devising code for 150 electorates, you may struggle to get the data you need in an easy format I reckon.

      Delete
  7. Tom the first and Best10:58 am, September 30, 2013

    The Nick Xenophon Group only ran 2 candidates. Had that been done at a DD then they would have had over a quota of surplus to distribute, so your result does not show the outcome of this election held as a DD.

    ReplyDelete
    Replies
    1. Do you not think that, just perhaps, in a DD the Xenophon group may have run more candidates? I'd guess that they would have run at least five.

      Delete
    2. My assumption was that parties would run as many candidates as seats they were likely to win. Perhaps it is implausible for Nick X to achieve this share of the vote again in a DD? But I had to assume votes and pref tickets remained the same.

      The same situation also happened in Tasmania with LIB running 4 candidates in 2013 even though they would win 5 in a DD election based on 2013 votes.

      I had not stated this assumption up front - I thought it would have been obvious given the output.

      Delete
  8. Isn't another fly in the ointment, the various shonky or erroneous preference deals, that presumably would not take place in a Double Dissolution? Especially so soon after the experience in the2013 election!

    Pete

    ReplyDelete
  9. Pete I agree.

    If you need replicate the 2013 results and you will soon see i is advantageous for the ALP to hold, a DD at the next election. The main party to lose out on a DD is the Greens.
    .

    ReplyDelete
  10. The ALP would secure 5 seats in Victoria in a DD. Not 4 as outlined above. The Greens would secure only one.

    The System used is not "proportional" is has distortions in the way the vote is counted. Namely the method of calculating the surplus transfer value and the segmented distribution of excluded candidates preferences.

    if we recount the WA Senate election using a weighted Surplus Transfer PUP and Lumlam are elected to the last two seats. If we remove the segmented distribution and introduce a reiterative counting system where the count is reset and restarted on each round then PUP and the ALP is elected. (Try counting the QLD 2007 Senate count excluding all but the last seven candidates (3 ALP, 2 LNP and 1 Grn) and Larisa Waters should gave been elected, the extent of the distortion in the count is equivalent to over 10,000 votes.

    The other fact that prevents 30% of the vote securing 30% of the positions is the Droop quota.

    ReplyDelete