Tasmania Senate 2016 Button Press And Analysis

Result

Order 1 Abetz (Lib) 2 Urquhart (ALP) 3 Whish-Wilson (Green) 4 Lambie (JLN) 5 Parry (Lib) 6 Polley (ALP) 7 Duniam (Lib) 8 Brown (ALP) 9 Bushby (Lib) 10 Singh (ALP) 11 Bilyk (ALP) 12 McKim (Green) 
Colbeck (Lib) was passed by McKim and McCulloch and excluded
5 Labor 4 Liberal 2 Green 1 Lambie

I would like to thank all the AEC staff in Hobart for all their help with information on the count and congratulate them on their efficient and punctual delivery of the Tasmanian outcomes.


------------------------------------------------------------------------------
Another great preferencing tool

William Bowe has prepared a spreadsheet that shows which parties' voters most often placed which other parties in their top six.

Just Vote 1

As already noted, hardly anyone just voted 1 for a party in the Tasmanian Senate race, despite the hysteria about people doing so before the election.  Just over 1% of all voters (this includes those who voted below the line) cast a single 1 for a party above the line.

For both major parties the share of all their ATL votes that were just-1 was the same: 1.5%, the same as the state average for all ATLs.  Parties with conspicuously high (compared to the state average) rates of just-1 were Citizens Electoral Council 3.8, Australian Liberty Alliance 3.5, Sex/HEMP 3.5, Liberal Democrats 3.3, Shooters Fishers and Farmers 2.8, Family First 2.5, Palmer United 2.3.  Parties with low rates of just-1 were the Greens 0.6, Nick Xenophon Team 0.7, Christian Democrats 1.1, Renewable Energy 1.1.  All others were in the range 1.5-2%.

Donkey Votes

A large "semi-donkey" component was obvious in scrutineering the Family First vote in Tasmania.  Some voters had picked six parties and then voted for their chosen six in order across the ballot, without realising that the order matters (or without caring about it).  But the fully-fledged above the line donkey-vote (at least 1-6 for the first six parties across the line) was very rare: 0.1% of all Tasmanian voters did this.

Section 282 Ambiguity On 3 and 6 Year Terms

There is a detailed post up by Dean Ashley (Alaric) dealing with the ambiguity of the Section 282 rules to prepare a recommendation for which Senators get 3 and 6 year terms.  The drafting of these rules is quite a mess and this was a hot topic of discussions I had with a few people following the declaration of the polls today.  There are ambiguities in the wording of Section 282 concerning, for example:

* whether votes that were informal in the original count (eg they had two 1s) can become formal in the S 282 recount (I would think not)
* whether votes that were partially formal in the original count (because of duplications of numbers) can become more formal in the recount if numbers are no longer duplicated (again I would think not but this has never been clarified)
* whether votes that were fully formal in the original count can become informal in the S 282 recount because they do not have enough preferences (again I would think not, hardly the voter's fault)
* whether a vote that has zero usable preferences in the recount because it is entirely for excluded candidates counts towards quota although it in fact exhausts immediately (this is largely a quirk of applying S282 the new system, although it could have happened in the old system too; the AEC's view as of 2011 was no; Antony Green also believes such a vote is excluded).

(PS Another one has come up, and it's a cracker:

* whether a vote that was originally formal both above and below the line, but has zero usable preferences in the recount below the line can have its above the line preferences used in the recount instead.  I think the legislation has not anticipated this application of the new system at all!)

The AEC made recommendations to JSCEM to fix this mess up in 2011 but it didn't happen.

The interesting thing is that simulations by Ashley and Grahame Bowland have found that different answers to these questions produce different 3 and 6 year winners in the case of Tasmania.  As there is no prospect of the parliament fixing this in advance of the allocation we'll just have to see what the AEC comes up with and what the Senate makes of it.

Exact Stats On HTV Follow Rates

Thanks again to David Barry's brilliant explorer function, we can find the portion of each party's above the line voters who followed each party's how-to-vote card as far as six. (Or as far as it went).  I should point out that many parties had their HTV cards online, and preference recommendations were collated by the ABC and here, so they were not that hard to find for people wanting to know what their party suggested.

These are the rates:

Liberal 11.9%, Green 7.3%, Christian Democrats 4.1%, Australian Liberty Alliance 3.0%, Labor 2.8%, Rec Fishers 2.3%, Animal Justice Party 2.0% (card preferenced 1-5 only), Shooters Fishers and Farmers 1.6%, Liberal Democrats 0.8% (card preferenced 1-4 only), Science Party 0.5%, One Nation 0.4%, Family First 0.3%, Renewable Energy 0.2%, Sex/HEMP 0.1%.

Also, 19.1% of NXT voters followed the recommendation to preference Lambie Network second.

As especially stark evidence of what I noticed in scrutineering that the gun-toting and bible-bashing parties didn't mix, both One Nation and Shooters, Fishers and Farmers preferenced the Christian Democrats second on their card.  But only 1.3% of One Nation voters and 2.3% of Shooters voters gave their second preference to the CDP!

So if anyone is doing any analysis of other states that relies on HTV cards in any way, my firm suggestion to you is to put it in the bin.  And if anyone is plotting preference deals between micro-parties for a future Senate election under this system, I suggest they give it up and take up racing model tanks.

Preference Flows: Harvesting Is Dead, Good Riddance

(Note: There is a better version of the data below at David Barry's site , with a neat explorer function, that eliminates the problem of flow between parties. But the conclusions are the same.)

The table below shows the grouped parties that left the count in order of them doing so, and the three biggest preference flows from each party.  (Exh = exhaust).  If the party issued a how-to-vote card (including online), then I show where the vote would have gone next if the voter followed that card, based on the actual count.  This is not necessarily the party shown 2 on the card, since it may by that stage have been excluded.  (The Sex Party preferenced five micros, all of which were excluded before it).  If the party shown on the card was not among the top three preference destinations then I show the percentage that went to that party in brackets.

Bear in mind that as we go down the list, the number of remaining parties decreases, so the share of preferences for each remaining party (and exhaust) goes up.  The high exhaust rates at the end are atypical; for the whole count only 2.8% of all vote-values exhausted.

Also, not all the preferences are above-the-line votes, with BTL votes a significant contaminant (especially for the Lambie Network exclusion).  And the preferences include votes that have come from other parties. (A breakdown by originating party could be retrieved from the original data to get a clear figure for HTV card follow rates but it would be a lot more work).

But even so, if voters were following how-to-vote cards in any great numbers, we would see it.  They're not.

In Tasmania, Senate reform has killed the preference harvesting model of Glenn Druery and co stone cold, spectacularly dead.  Of the bottom five parties that issued cards, in no case was the preferenced party one of the top three actual preference getters!  In one case the preferenced party received only 3.6% of preferences from the recommending party, which they surely would have got even with an open ticket.  There is no evidence preferences flowed more strongly when a micro-party had a card to when it didn't (if anything, the contrary).

In the cases where the preferenced party did receive the most preferences, this could well be explained by it being a similar party ideologically rather than by people following the cards.  Indeed David Barry has now determined that precisely two voters followed the Renewable Energy Party card from 1 to 6, whereas the other 99.8% of REP voters made up their own minds.  The funny thing is that at the booth I voted at, the REP were actually handing out, as were the Sex Party, whose votes would have all exhausted had their voters been dumb enough to follow the Sex Party card.  (Turns out they weren't.)

We can see here that the near-100% preference flows from party to party created by the old group ticket voting system were nothing but a monstrous lie, maintained only by forcing voters above the line by making below the line voting too onerous.  At least, this is so in Tasmania, and we await the verdict from the other states.

What's especially amusing here is the extremely weak flow between the Renewable Energy Party and the Recreational Fishers, since these parties swapped preferences, both ran in the Reps in an apparent attempt to boost their Senate vote, and were predicted by Glenn Druery to be a chance of fighting out the final seat.  Both got nowhere near.  Druery was obviously in cahoots with the REP folk, and though I am aware of no direct evidence in the case of the Rec Fishers, I suspect he was involved there too.

Who People Put Last

Voting all the way below the line (1-58 in this case) is the best way to make your vote as powerful as possible if you have strong views about particular candidates as opposed to just parties, but not that many voters do it.  I have done a quick sample of the top 5% of the Tasmanian data-entry voting and find that in this sample 1.8% go all the way to put someone 58th.  

Last time the honour of most last places was fiercely contested by The Greens and Stop The Greens but I had a sneaking suspicion there would be a runaway winner this time around.  It looks from the first 5% of the data-entry voting that of the few voters going the distance, a massive 35.3% have put Eric Abetz (Liberal) last.  He is followed distantly in this sample by John Tucker (bottom of the Liberal ticket) on  9%, Anna Reynolds (bottom of the Green ticket) 6%, Nick McKim (Green) 5.3% and then the bottom Christian Democrat and One Nation candidates.  For everyone but Abetz the sample size is inconclusive so I may check the full list as time allows.  (My computer is a bit old for processing spreadsheets with over 300,000 rows quickly.)

Poster Alaric, who I owe an enormous debt of thanks to for his help with the data, has since crunched the lot, and it gets worse.  Of all Tasmanian Senate voters who went all the way (2.3% of the total), Abetz tops the list with 40.3%, followed by Reynolds on 5.5% and Tucker 4.7%.  100 of the 7951 voters going all the way put Richard Colbeck first and Abetz last.

And another informant has told me that if you weed out those who didn't get from 1 to 58 without an error, Abetz goes up to 41.4%! (Some of the 58s for other candidates may be from voters who put Abetz 59th.  Some voters even voted 1-12 with a symbolic 58 for Abetz).

Twist Of Fate

I was just responding to a query on Twitter when I discovered something amusing.  It was notable while scrutineering that a lot of Richard Colbeck votes went to Lisa Singh and vice-versa; the classic up-yours-party-machines vote.

Well, strangely, those votes have ended up saving Nick McKim.  Only 74% of Colbeck's vote stayed in the Liberal ticket, with 13% "leaking" to Lisa Singh.

If we imagine instead that all that 13% had gone to David Bushby, then that would have increased the size of Bushby's surplus (which was dominated by Liberal ticket votes because of Inclusive Gregory distortion) to the point that McCulloch would have beaten McKim by about 50 votes.

Recount?

There has been some speculation about the possibility of a recount in view of the 141-vote margin.

I hesitate to refer to the AEC's recount policy on Senate elections when the AEC has abandoned its own recount policy for Reps elections by fast-tracking Herbert, so who knows what might be decided?  But according to it:

* there is no specific margin of error that causes an automatic Senate recount - this differs from the House where any margin under 100 triggers an automatic recount.
* a close margin does not, by itself, establish a need for a recount.
* there must be specific reason to believe a recount could change the margin.

I would expect that unless One Nation can demonstrate something concrete and unusual, any request for a recount would be refused.  The margin is tiny, but the amount of double-checking in the system makes it deeply unlikely there would be enough errors in the same direction to overturn it.

Six Year vs Three Year terms

Commenter Alaric has very kindly posted a comment regarding three vs six year terms for the Senators elected.  The suggestion is if either the Section 282 method (which has never been actually used) or the order-of-election method are used, then the Senators receiving six-year terms will be Abetz, Urquhart, Whish-Wilson, Lambie, Polley and Parry, and the rest (3 Labor, 2 Liberal, 1 Green) will be up for re-election in 2019.  Of course the Senate may come up with some completely new way to allocate seats.  

Preference Distribution

The preference distribution is online and the key events of the count are as follows.  Richard Colbeck was eliminated after the exclusion of Peter Madden, with 16918 votes to McCulloch's 18136 and McKim's 19431.  To compare the quota scores for the contenders at this point with my model (actual with modelled in brackets):

Bushby .934 (.947)
Bilyk .843 (.904)
Singh .917 (.910)
McKim .741 (.779)
McCulloch .695 (.634)
Colbeck .648 (.632)

So both Singh and Colbeck did slightly better on below-the-line preferences than I expected, while Labor and the Greens did worse on preferences than my model and One Nation quite a lot better.  It is notable that Lisa Singh crossed the line entirely on below the line votes, with a surplus of four after Colbeck's exclusion (had Singh had five fewer votes, Bilyk may have finished ahead of her.)

After the preferences of Colbeck and the surpluses of Bushby, Singh (all four votes of it) and Bilyk, McCulluch closed rapidly but missed out by 141 votes (21247 to 21106).  The count came down to the final Bilyk surplus worth 593 votes with McKim leading by 43 before it, but the superior flow of Labor ticket votes to the Greens compared to One Nation made him safe.  The total exhaust rate was just 2.75%, way lower than many predictions from those who opposed reform. (My model had 3.2%).

From the Colbeck exclusion point my model didn't give projections since at the time I computed it Colbeck looked like outlasting McCulloch.  But McCulloch then got a 21% share of the Bushby surplus compared to McKim's 12%, which combined with her generally stronger performance on micro-party preferences than expected was almost enough to put her over the line but not quite!  So a massive upset was averted by a tiny margin indeed, but you get that.

It will be difficult in my view for One Nation to mount a convincing case for a recount, unless any errors are found.

I will be compiling more figures from the distribution and also from the download of formal votes (if my computer can open it!) later, but the first thing to know is this: everyone's preferences go everywhere!

Button Press and Outcome

The button press happened exceptionally quickly, with two draws for tied candidates early in the piece and then a list of names scrolling as candidates were elected and excluded.  Richard Colbeck was excluded behind One Nation (which according to my projection as of last night was fractionally more likely than not) and the only mild surprise in the ordering is that Lisa Singh has been elected ahead of Catryna Bilyk.  

Singh is the first candidate to buck a ticket order and win election to the Senate since Liberal Robert Wardlaw in 1955.  Australia first adopted a proportional-representation system for the Senate (without above the line votes) at the 1949 federal election, and at the first three such elections Tasmanian candidates (sometimes on both sides) bucked the ticket order - in some cases because their parties deliberately selected them down-list for strategic reasons.  However after the first few runnings of the system, this stopped happening, and under the "group ticket voting" system brought in by Labor under Bob Hawke this became effectively impossible.  The new system gives the power back to voters to overturn bad party preselections. 

The result is a complete disaster for the Liberal Party federally since Tasmania has sent an 8-4 voting block to Canberra on most core government issues (Lambie is pretty left-leaning on health and education).  Much more to say on that later.

Pre-Button Waffle

Time for a new thread for the Tasmanian Senate race because today from 2 pm it is finally upon us ... The Button! The AEC, having entered all the numbers into their computers, will conduct an automatic preference distribution that will tell us the winner (pending challenges) of the 12 Tasmanian Senate places.  The provisional result should be known over the next 30 minutes or so, and a preference distribution will follow.  Based on this distribution it will be possible to check the margins and see if there is any risk of challenge.  Full vote details will also be released to make it possible for those with a lot of time and a very good computer to verify the result.

For the second time in the last two Senate races (see 2013 button press thread) we go to this hallowed occasion without being able to know for certain who will win at least the final seat.  I've put a very large amount of work into scrutineering sampling for this count and into trying to project the outcome (see thread here) but in the end it could all be in vain; the result for the final seat is too close for my model to call with complete confidence, and scrutineering sampling for this election has been very difficult.  It being a new system, something totally surprising could even happen, and then we learn something for next time!

I will try to post the result of the button press at the top of the article as soon as I have it - this will depend on whether I'm able to update this page live from where I am while (probably) watching this event in person.

We know from the primary figures that four Senators will be re-elected by having quota in their own right: 1. Eric Abetz (Liberal), 2. Anne Urquhart (Labor), 3. Peter Whish-Wilson (Greens), 4. Jacqui Lambie (Jacqui Lambie Network).  The surpluses of Abetz and Urquhart will elect 5. Stephen Parry (Liberal), 6. Helen Polley (Labor) and their surpluses will elect 7. Jonathan Duniam (Liberal), 8.Carol Brown (Labor).  (There is a remote chance some of these orders will be reversed).

There is then a race between remaining Senators Catryna Bilyk and Lisa Singh (Labor), David Bushby and Richard Colbeck (Liberal) and Nick McKim (Green) plus all the micro-party candidates for the remaining four seats, which will be determined by the cutup of unsuccessful candidate preferences (micro-party, ungrouped and minor ALP, Liberal and Green candidates).  Based on my sampling and model I expect all the micro-parties to be excluded without rising much above their present positions on the ladder, with the exception of One Nation, whose candidate is Kate McCulloch.

The starting order of these six candidates after all the surpluses will be Singh, Colbeck and Bushby close together, Bilyk, McKim, McCulloch.  However Singh and Colbeck will move very slowly on preferences because they can only receive below-the-line preferences until Bilyk and Bushby (respectively) are elected, which takes a very long time.  Singh's primary is so high that a trickle of BTL preferences should put her in a position where she cannot be caught because of votes exhausting (indeed it's not clear that she needs any preferences at all), but Colbeck will be overtaken by Bushby and Bilyk and is defeated if either McKim or McCulloch also passes him and stays ahead.

The tipping point in the count should come after the exclusion of all micro-parties other than One Nation.  At this point I expect one of Colbeck, McCulloch and McKim to be in last place and to be excluded.  My sampling had about 21% of micro-party (excluding One Nation) preferences going to the Greens, but if they only get around say 12%, McKim may be excluded here.  It also had 19% of preferences going to One Nation, which if correct makes it impossible to say whether Colbeck or McCulloch will be excluded first.

Provided McKim gets over this hurdle, he should win, though I didn't have time to sample the scenario of a battle between McKim and McCulloch for the final seat on Liberal preferences.

In general the flow of micro-party preferences has seemed stronger for Labor than the Liberals, and not that much stronger for the Liberals than the Greens, in my sample.  This seems really counter-intuitive given how "right-wing" many of the micro-parties are, but it should be kept in mind that the Liberals have polled a two-party Reps vote below 43% in the state despite having the personal votes of three new sitting members to help them along.  Also, while people stereotype One Nation as hard-right, their preferencing behaviour normally leans only weakly to the Coalition, and in this case a lot of their supporters are just exhausting their votes.  Finally, a lot of the Family First votes are donkeys.

So my take: probably 9 and 10 Bushby and Bilyk in some order, 11 Singh and for the 12th seat lukewarm favourite McKim, realistic chance Colbeck, outside chance McCulloch, and let's not completely rule out something else happening, just in case.

All will be revealed in a couple of hours and I will post the result here as fast as I can, to be followed by analysis of the preference cut-up this afternoon or tonight.