Friday, June 14, 2013

Prospects for the Tasmanian Senate Race

Note: This is a pre-election article.  Of all the wild scenarios entered into below, the one that never came into consideration was that one of the apparent micro-parties would poll seven percent!  This is what has happened with the last-week rise of Palmer United.  A post-count thread dealing with the Tasmanian Senate spot will be added soon.

Updated Summary 24 August:

* While Julia Gillard was Prime Minister, an outcome of 3 Liberal, 2 Labor, 1 Green for the Tasmanian Senate was overwhelmingly likely.  This is still the most likely outcome, but at times since the return of Kevin Rudd, other outcomes have seemed reasonably plausible.

* Polling since the return of Kevin Rudd has suggested the Green vote is softening and a Green seat is no longer guaranteed (though in my view likely).

* A very good above-the-line preference allocation for Family First gives them a chance of taking the third Liberal seat should the Liberals fail to poll three quotas (42.8%) in their own right.  Some other micro-parties are also in remote contention. 

* With declining national polling the chance of Labor retaining three Senate seats in Tasmania appears to be very low.

New Update Text (August 24)

History Of Tasmanian Senate Contests

This text is taken from the original version of this article.

Until recently, the composition of the 12 Tasmanian Senate seats has been extremely stable.  Norm Sanders won the Australian Democrats' first Tasmanian Senate seat in 1984 and the balance at that time was five Liberal, five Labor, Independent Brian Harradine, and one Democrat.  This balance survived the 1987 double dissolution, and for the next five half-senate elections after that, one cycle would return 3 Liberals, 2 Labor, 1 Democrat (or from 1996 onwards, Green) and the alternate cycle would return 2 Liberal, 3 Labor and Harradine.

Twenty years of 5-5-1-1 ended in 2004 when Harradine retired.  With Harradine no longer competing for the Liberal vote, Tasmania returned 3 Liberal, 2 Labor and 1 Green for the second election in a row, putting the numbers at 6-4-2.

Counting Harradine as loosely of the "right" and Labor as in some sense "left", this meant there had been a 6-6 left-right balance for a very long time, but in both 2007 and 2010 the ALP won three seats, the Liberals two and the Greens one.  These two unusual 4-2 left-right splits have been very significant outcomes nationally.  Had the state returned three Liberals in both 2007 and 2010, then (i) the Coalition could have blocked legislation alone during most of Kevin Rudd's Prime Ministership and (ii) Labor in its current term would have needed support from the Greens and one of Nick Xenephon or John Madigan to pass legislation.

This is what the half-Senate blocks for Tasmania since the last double dissolution look like as a graph:


The Greens (or in 1990, Democrats) have won a seat in every half-Senate election except 1993 and 1998, which were Harradine years (though whether that caused them to not win in either year is quite debatable).  In both 1993 and 1998 the Greens were reasonably, but not very, close to winning the final seat at the expense of Labor.

This is what the patterns in primary vote quota total for Liberal, Labor, Green and Others look like for the last four elections:

 In 2001 the Liberal Party polled only slightly higher than Labor and well short of a third quota, but they were put over three quotas by the preferences of One Nation and the Australian Democrats' practice of splitting preferences between the two major parties.  In 2007 and 2010 the third Labor candidate was put over quota as a result of the Greens' surplus.  The 2010 Liberal results in the state were generally very bad, with the party making relatively little campaign effort and being seen as neglecting the state on broadband policy.

The most interesting election was 2004 in which a strong flow of above the line preferences to Family First gave the Greens, who had just fallen short of a quota, a bit of a potential scare.  The 2004 count has been the subject of a persistent myth that the result between the Greens and Family First was especially close; it was not, and Milne's margin of victory had all preferences been distributed would have been about 6000 votes.  It just looked very close based on figures on the night before changes in the balance of the parties in post-counting made it less so.

Current Projection: August 24

The legacy text sections below followed the ups and downs of the national vote over the last few months and its impact on the Tasmanian Senate picture.

In this most recent update below I attempted to model the Tasmanian senate vote using comparisons to the last federal election and the ReachTEL polling then available for some seats.

I have plugged the current ReachTEL poll (see ReachTEL Says Tas Labor Still Losing Three) into a model similar to the one discussed below.  The raw Senate projection of the model is Liberal 45.1 Labor 30.6 Green 15.3 Others 9.  I think this more likely overstates the Liberal vote and understates Labor than the other way around.  A neat thing about this model, however, is that it projects an increase in the Others senate vote (which was 5% last time) via the polled increase in the Others (excluding Wilkie) vote in the Lower House, as a result of the increased number of minor Lower House candidates.  A possible flaw in the model is that it assumes the increase in the Andrew Wilkie vote between 2010 and the current poll is drawn equally from Labor, Liberal and Green.  Perhaps it's actually coming slightly more from Labor in which case the figures above are harsh on the ALP and they would probably do better than that.

I've been using Antony Green's Senate calculator (note: still in development, works well for me in Firefox but may not work well in IE) to test some scenarios.  An important limitation of the calculator is that it assumes 100% above-the-line preference flow but in practice, even with 54 candidates, flows will be weaker than that, and this will favour tickets that have lots of votes (after the distribution of surpluses) over those that have to snowball their way up from the bottom.

Thus, on election night you may well see Antony's calculator say that some silly party you have never heard of is winning: be aware that (i) that its projections are hugely dependent on the exact primary figures (which change after the night as postals are added) (ii) below-the-line votes will always disadvantage any party with a very small primary vote.   Such projections might be correct but should be treated with very great caution.

If this model is anywhere near accurate then we are back to a straight 3-2-1 scenario (3 Liberal) and the tiddlers have nowhere to snowball from because the three main parties have their quotas locked down and are elected automatically.  However it could be inaccurate in various ways:

1. ReachTEL polling in Tasmania could be skewed to the Coalition by a large amount.  The model above is already adjusted for a possible ReachTEL house effect of one point.   Suppose instead that I "morganise" ReachTELs data by shifting it to match the results of Roy Morgan Research, as adjusted for house effects.  This is done by moving six points from the Liberals to Labor.  If I do not increase the Others vote much, 3-2-1 is still the outcome.  Preferences from the right micros go to the Liberal Party when the right micros run out of steam.

2. The Green vote could be too high.  Knocking the Greens down below a quota doesn't seem to make a lot of difference.  The Labor third candidate is eventually eliminated, putting them over the line.  Even if Labor and the Greens combined are reduced to below three quotas (which involves taking over 3% from the model) they tend to be elected on PUP preferences.  I have to take five points off the combined ALP/Green vote and assume Palmer United do quite badly before I can get the Greens to lose - as it turns out, to Family First.  The Greens were projected by the calculator to lose to FF in 2004 but did not do so because of below the line leakage.  This will be less of a factor this time but would probably still save them.  So the prospects of a four right - two left scenario don't look at all strong.

3. The combined Others vote could be much larger than nine points, mainly at the expense of the Liberals, knocking them below 42.8%.  This is where things get interesting.  If the Liberals are reduced to around 41% with Others on about 12, in some scenarios a preference snowball "elects" Family First (even off only 1% of the vote) after Labor is eliminated, producing a 2-2-1-1 result, although this might not happen from the same vote range in practice.  In most such scenarios I ran the Liberals narrowly beat Family First, most commonly with the preferences of the Australian Sex Party (the "anti Family First") the difference. 

Amusingly, in only the second run of this assumption that I tried, the obscure Australian Independents group snowballed from an assumed 0.47% of the vote and "won" the Liberal seat; I was later able to refine that set of primaries to put AI as low as 0.31% and have them still "winning".

In practice such outcomes are artificial because of BTL voting and sensitivity to the exact votes of the tiddler parties, and most likely a Liberal ticket with 41%+ of the vote would hold off any preference snowball.  Also the current model suggests the Liberals have a very good chance of making three quotas anyway.  But it isn't cut and dried.

By combining multiple assumptions from the list above it is possible in theory to generate almost any outcome you like (though two Greens, an often requested scenario, is a number of bridges too far.)

I now think 3 Liberal, 2 Labor, 1 Green is again the most likely outcome by far.  I give it at least a 75% chance of occurring, with the Family First snowball for 2-2-1-1 perhaps the most realistic of the remaining unlikely outcomes.  The Greens' seat is not completely secure but the threat to it noticed in my July model appears to have faded.

All that follows is legacy text from previous versions of this article that is no longer current and is left up just for any remaining interest it may have.  And for the benefit of readers who were looking for the ferret, but haven't found it yet.


Legacy Update Text (mostly written 29 July)

Below the many August updates at the bottom of this section, you can find the legacy text of an article previewing the Tasmanian Senate race, written when Julia Gillard was Prime Minister.  At that time, Labor's polling was very poor and 3 Liberal, 2 Labor, 1 Green was the overwhelming favourite.

My current article ReachTEL - Tas Federal BaByLon Still Falling? (and yes that ? is staying in the title now) covered the new ReachTEL poll of Lyons, Braddon and Bass, and showed a swing back to Labor since the June ReachTEL that is about comparable to the national swing back (about six points).

Using that poll I have attempted to project the Tasmanian Senate race via the following method:

1. Find the difference between the Senate primaries in Bass, Braddon and Lyons and the ReachTEL electorate primaries this time, and apply that difference statewide to obtain notional state Senate primaries assuming a uniform swing.

2. We know the state swing isn't uniform, so adjust it for the difference between the state swings in the June ReachTEL and the Bass, Braddon and Lyons average swings in the July ReachTEL.

2a. Point 2 is complicated because of Andrew Wilkie taking votes from other parties in Denison, which is not an issue for this Senate projection.  Therefore, those votes are considered as returned.  We don't know who he's getting them from, and there's at least a case he's getting them more from the Greens (for their size) than other parties, so I've treated them as being returned to other parties equally.  When I do this, the massive swing against the Greens in Franklin in the June sample doesn't show up in Denison.

3. Adjust Labor and Liberal totals for the apparent house effect of ReachTEL polling.

The net result of this very rough projection was a Senate estimate that comes in at Liberal 46 Labor 39.8 Green 11.2.  If that was repeated at an election, the Liberal Party would get 3.23 quotas, Labor 2.78, the Greens 0.78 and others the rest.  The last seat would then be determined between Labor and the Greens based on the preferences of the Liberals and others, and indeed on those figures Labor would win.  Most likely there will be an increased showing from others at this election and therefore the Liberal excess would be lower, if indeed the Liberals get three quotas.  The key point is that Labor's vote seems to have recovered to the point that if the Greens cannot get very close to a quota in their own right, they might not survive on preferences.

This does seem a rather extreme result for the Greens, and it's possible they'd do significantly better and put the result beyond doubt.  The bad ReachTEL Franklin result for the Greens from June has a large impact on the projection.  But what other state breakdowns are available (most of them Morgan, for what it's worth) suggest the party is struggling to hit double digits in Tasmania at House of Reps level, and would therefore not get a quota in its own right at Senate level even after adjustment for the Wilkie masking effect.  The steady erosion of the Green vote nationally appears to be biting in Tasmania too and if there is going to be a refugee-issue bounce, we are so far not seeing that much by way of signs of it.

3 Liberal, 2 Labor, 1 Green can therefore no longer be considered the overwhelmingly likely outcome of the Tasmanian Senate race.  3-3 Liberal-Labor now seems to be a contender.  (This bit originally said "an equally serious contender" but that was a bit of an overreaction on my part - it's an extrapolation based on one poll against a picture that had otherwise very solid support, so 3-2-1 is probably still the more likely result.)

The key point is that Scenario 2 from my original article is now in play because it does appear possible that the Greens vote could fall low enough.

Update 7 August: Lib Seat Isn't Locked Down Either?

The Examiner yesterday carried comments about internal polling for this race by the Greens ("their polling shows the party right on the quota") and Labor ("a "line-ball'' contest for the sixth seat between Senator Thorp and Liberal newcomer Sally Chandler.")  Indeed, there is now more than one piece of recent internal polling out there that purports to show the gap between the major parties as relatively small. Furthermore, Labor's internal polling would probably be showing the Greens with a quota or nearly so and "safe" for the above comment to be made.

Senate polling is a curious practice with a rather poor accuracy record at past elections, and given the ease of over-polling the Green vote, even public polls showing them at or just above the quota wouldn't indicate certain safety.  But given the likelihood of help from micro parties, a vote of around 12.5-13% would probably be enough. 

In 2010, the Liberals gained 3.95 points on Labor through the count as a result of the preferences of micro right parties.  Something like this is likely to happen again this time, so any poll showing the two parties level on primary votes is really showing the Liberals ahead, and as such the Labor claim would not be that convincing if it was published polling they were referring to (even less so for internal).

But ultimately, the Liberals do still need to find an extra 5.8 points from Labor and/or the Greens compared to 2010 to win.  That's below the swing that is generally being recorded for them in polling (except by Morgan) but not so far below as to be certain that we'll stay there.  Broadly, we've ticked the first two boxes for Scenario Five below- Rudd is back and the 2PP is 50-50ish (though there's no guarantee it will stay there).  So if the 2PP does stay close, that just leaves us with the question of whether the Coalition's improvements on their 2010 shocker, plus the state factors draining the Labor vote, are worth six points on their own.

Time to put some rough probability estimates on the current situation.  I reckon (UPDATED 21 Aug): 3-2-1 to Liberal 55%, 3-3-0 20%, 2-2-1-1 (a micro, probably Family First wins) 15%, 3-2-1 to Labor 5%.  A lot less clear-cut than it used to be.

Reversed ferret image (original source)
(The ferret's estimates are not on record, which is a shame; they might be more useful than mine.)

Senate Preferences: Major Parties (17 August): Above-the-line preference allocations are taking shape with the Liberals' announcement that they will preference Labor above the Greens and Labor's announcement that it will preference the Greens second in all states except Queensland, where it has done a deal with Katter's Australian Party. 

There has been a lot of speculation about what this deal means for Tasmania, but apart from the possibility of the Labor vote suffering from more "A Vote For Labor Is A Vote For The Greens" claims (which aren't likely to do well in fact-checking in the Tasmanian context), the answer is very little.  It's not telling us anything that would not have been generally assumed anyway.  Assuming Thorp (ALP) and Whish-Wilson (Grn) are not both successful, the likelihood is they will be competing with each other for the final seat and hence Labor preferences will not be distributed.  Labor preferences will only be thrown if they have done so badly that a candidate from a fourth party is ahead of them and competing with the Greens and/or Liberals for the final place.

What will be more intriguing is whether Labor preferences some of the more significant right-wing micro parties above the Libs, and I have heard that they might. (Update: Correct. Labor has preferenced Shooters and Fishers after the Greens.) With the massive number of parties contesting the ballot (by Tasmanian standards) that would then put Scenario Four below in play.  We will know more when the group voting tickets are released very shortly and this article will be updated again.  But I'd say now that the chance of 2-2-1-1 is increased somewhat above my initial 5% estimate, and probably that 3-2-1 to Labor is becoming more unlikely as Labor's national vote slips.

Update 21 August: The Group tickets are quite friendly to Labor and unfriendly to the Greens, who have been preferenced only by PUP (oddly) and some (but not all) of the left-wing micros. Family First have done extremely well in the group ticket dealing for Tasmania and appear to be the most competitive micro assuming a decent (1%+) primary vote.  They pool votes from five other micros that don't seem to have a serious chance, and if they are ahead of PUP and/or KAP at that stage they can pick up the preferences of those parties.  They can also pick up the preferences of the DLP and Shooters+Fishers if they can outlast them.  The most likely scenario for a FF snowball involves the Liberals with 2.8 quotas, Labor with, say, 2.5. the Greens with 1 and Others the remaining 0.7 or so.  Because of the unexpected sheer number of Tasmanian micros at this year's election, this is quite a realistic possibility.  Some other micros are competitive, including Shooters+Fishers if they can poll a high primary vote and snag Labor preferences. (This requires Labor to do very badly and the Greens well.)  I am waiting for Antony Green's Senate calculator model to go live to model this in more detail. 

Original article follows below the line.


Original Article:  Advance Summary (Ancient! Dates from Gillard Era!)
This article models possible outcomes for the six Tasmanian Senate seats up for election this year and concludes that an outcome of three Liberal, two Labor and one Green is by far the most likely.

Antony Green has written a preview for the five Tasmanian House of Representatives seats at the
upcoming federal election that is so comprehensive that there is not a lot to add until there is more detailed polling publicly available on statewide trends or individual seats.  Projections for these seats will be posted here much closer to the election.

However, I think it is already possible to say useful things about the Tasmanian Senate race.  This is most likely at this stage to produce a result of 3 Liberal, 2 Labor, 1 Green (a net change of one seat from Labor to the Liberals) but various other results are possible.  In particular, state polling showing the Greens at below the Senate quota level has some asking whether the seat held by Peter Whish-Wilson is safe.

[History section moved to top of article.]

The Case For 3-2-1

Looking at the 2010 results (Liberal 33%, ALP 41.4, Greens 20.3) the case for a 3-2-1 Liberal-Labor-Green result is rather easily made.

Firstly, given general (if scanty) state evidence that the swing to the Coalition at House of Reps level is likely to be at least 10% on a two-party preferred basis, it is quite plausible the Liberals will gain a sufficient swing to reach the 42.8% bar for three quotas directly on primaries.  Even if this does not quite happen, the Liberals frequently receive preferences from right-wing minor parties.  In the 2010 count they gained around 4% during the count by this method.  So the swing effectively required is about 6%.

The swing against the ALP's primary needed to even create a chance of them missing a second seat is 12.8%.  That's very unlikely to happen (discussed at length in Scenario 6 below).

A swing of 6% against the Greens is needed to knock them below a quota on primaries.  Even then the Greens are likely to receive some preferences from feeder micro-parties (in 2010 just over 1%) and should also benefit from any ALP surplus over a second seat.

If the Liberals reach 43%, the ALP 28% and the Greens 14% on primaries, then the election is simply over, and it doesn't matter whether soemone else has even, say, 10% of the vote. 

Will The Greens Get Quota?

The Greens have a long history of usually not polling as well in Senate elections (and for that matter, House of Reps elections) as they do in state politics:

* In 2001 the party polled 13.8%; at the subsequent year's state poll the party polled 18.1%.

* In 2004 the party polled 13.3% midway between state elections in which it polled 18.1% and 16.6%.

* In 2007 the party polled 18.1%, a result which is just ahead of a line between the party's state results of 16.6% in 2006 and 21.6% in 2010.

* In 2010 the party polled 20.3%, slightly down on the state vote of 21.6%.

The 2001 result was affected significantly by competition from Democrats and in 2004 ex-Labor forestry dissident Independent Senator Shayne Murphy may have taken votes from them.  Competition with micro-parties is always a factor, but based on 2007 and 2010 there doesn't seem much reason to believe the party's Senate vote now would be worse than its state result now.

That said, the party's state result now would be hideous.  The last three EMRS state polls show the party with state votes of 15, 18 and 14%, for an average 6% swing since the 2010 result. That is from a pollster that habitually overstates the Green vote because of historically false assumptions about the voting behaviour of "undecided" voters; a current state vote around 12.5% for a 9% swing away seems a more realistic reading unless EMRS have fixed the problems they have had with polling the Green vote in the past. [Update: Since this paragraph was written the ReachTEL poll has pointed to a better state result for the Greens than the recent EMRS polls have suggested.  There would still, however, be a substantial swing away from the party at state level.]

Another possible line of modelling is to look at evidence concerning the federal decline in the Green vote. The Pollbludger Bludgertrack aggregate presently projects an 8.5% vote for the Greens in the House of Representatives nationwide, a loss of 28% of the party's 2010 vote.  The same decline applied proportionally to the state Senate vote cuts it to fractionally above a quota (14.6%).

It is possible Green voters will take a different view of the federal Greens to the state Greens, since the latter are seen as selling out on the forestry peace deal, while Christine Milne has strongly criticised it.  Even so it's clearly possible at this stage that the Greens will not get 14% on primaries, and it's plausible that, as in 2004, they will be far enough short of a quota to have to work for their victory, rather than being tipped straight over by feeder groups.

Scenario 1: The Right Gets Four

Rather than just looking at individual party totals it is often worth splitting the parties into "left" and "right" on the assumption that, with rare exceptions, the right wing parties will feed to the Liberals and the left-wing to the Greens and then to Labor.  In 2010 there was about a 37-63 right-left split in Tasmanian Senate voting.  A 6% swing brings the right above three quotas and the left below four, and I've treated this as a done deal.  However, for the Right to get four, there has to be either a 20-point swing, or a preference defection at some point.  An example of preference defection was Labor preferencing Family First ahead of the Greens in Victoria in 2004, causing Steve Fielding to get elected.  (Had Labor outlasted Fielding in the count it would have been a defection in the other direction, albeit a smaller one.) 

The size of swing required is so large that even given what we know about possible double-figure swings in the state, I don't think this is realistic unless we see a further meltdown in Labor's federal vote.

It is worth looking at the dire Federal ReachTEL for the electorate of Bass for some insight into the difficulties of finding four Senate seats for the Liberals.  That poll (of House of Reps voting intention) showed the Liberals on 54.7%, Labor on 26.7 and the Greens on 8.7 with "other" getting most of the rest.  In that case the Liberals are much closer to their 57% than Labor and the Greens are to their 43%.  However, in 2010 the Liberals outperformed their state Senate average by 2.7 points in Bass, and Labor and the Greens combined underperformed by 2.5 points.  On that basis (assuming the Reps vote was a reasonable model for the Senate) the two sides would be about equally distant from their quota before we considered the impact of ReachTEL's "Others" (who are not all that conservative in my view).  And that's even before we consider that the Bass ReachTEL showed a seat swing that is far higher than the (albeit large) swings believed to be occurring across the state (on the other hand, those may have increased a bit since January).  That in turn is not surprising because at state level swings appear likely to be larger in the north than in the south, the north of the state always being more vulnerable to economic downturns. 

We did get fairly close to a 4-2 right-left result in 2004 (as discussed above).  However, that was only the case because the ALP, the Democrats and Shayne Murphy were all directing preferences to Family First ahead of the Greens.  In other words, a lot of preference defection from parties that would have attracted centre-left votes, an unusual scenario and not too likely to be repeated.

Scenario 2: Three-All And The Greens Miss Out

In this scenario the Liberals poll in the mid 40s, Labor in the mid-30s, and the Greens a bit below 13, with an increased fourth-party vote.  If the fourth-party vote consists mostly of right parties, and the left parties include some that preference Labor ahead of the Greens, it's just possible at a stretch that Labor's third candidate stays ahead of the Liberals' fourth.  Then if the Liberals have decided to direct preferences to Labor ahead of the Greens, Labor wins a third seat.

I think this scenario is artificial and very unlikely - it relies on too many conditions that must all be met, and even then it only just works. Assuming that the Green vote really crashes badly (to, say, 11) makes it easier to make it fly but it's doubtful the Green vote would go that low

Scenario 3: 3-2-1 And The One is Not The Greens

In this scenario, the Liberals poll in the low to mid 40s, Labor in the low to the mid 30s and the Greens 13 or less.  A fourth-party candidate with a moderate orientation (or two such candidates with each feeding the other) gets about 6% statewide and right feeder groups get most of the rest.  In this scenario, the Liberals have their three quotas, the right preferences pool to the moderate and put them over Labor, and the Labor preferences pool to the moderate and put them over the Greens. 

Even given the vote levels in question, this is not as easy as it might sound, because of Tasmania's high rate of below-the-line voting, which makes it hard for a candidate to catch another from well behind (as we saw in 2004).  But the biggest obstacle is that at this stage there is no such candidate openly running.  Capturing a hefty Senate vote statewide usually requires resources, organisation and hard work on profile building over a much longer period.

Scenario 4: The Right Gets 3, but the Libs Get Only 2

In this scenario, the Liberals get about 41, Labor 34, the Greens about 15, and minor mostly right parties ten points with substantial showings for new parties like KAP and PUP plus the usual stuff for Shooters and Fishers, FF and so on.  The minor right preferences pool strongly enough to one of them to just get them over Labor, and then they beat the Liberals on Labor preferences.

I don't think this is that realistic for Tasmania - there is just not the evidence there that the minor right parties have that sort of vote-pulling power between them in this state.  At the high-water mark of minor right party populism in Australia, 1998, the minor right parties polled only about 5% combined in Tasmania (and slightly less in 2001).  Furthermore the ALP will frequently preference the more controversial right-wing parties below the Liberals anyway.  So this scenario too seems very remote.

Scenario 5: The Status Quo

In this scenario, Julia Gillard resigns fifteen minutes after you read this, and Kevin Rudd is installed.  There is a lasting swing back to Labor resulting in a close election and altering expected swings everywhere.  The swing in Tasmania is still substantial but Labor and the Greens cut their losses to six points between them and Labor is pushed over a third quota on Greens preferences.

I think this is even less realistic than some of the above, because it depends on heroic assumptions about the party's federal performance from a situation that is obviously a deep crisis.  Even assuming that the national picture did lift to the region of 50-50, Tasmanian state and economic factors could well keep the left-to-right swing too high for the status quo to be possible anyway.

Candidates? What Candidates?

I haven't talked about candidate factors in this election because there is just so much evidence that they simply do not matter.  Most voters whack a 1 in the party box and may not even read the names underneath; many of the remainder follow the party ticket down the line but just want to allocate their own preferences between other parties.

So, unless something extraordinary happens, when the next Senate takes its position, then Carol Brown and Catryna Bilyk will be there for Labor, as will Liberals Richard Colbeck and David Bushby.  The at-risk incumbents are Lin Thorp for Labor and Peter Whish-Wilson for the Greens, and the most likely new Senator would be Sally Chandler for the Liberals.

In Conclusion

Assuming that the final federal nationwide voting pattern is anywhere near what current polling is showing, I believe that three Liberal, two Labor and one Green is overwhelmingly the most likely Senate result in Tasmania.  The ALP is highly likely to poll two quotas on primary votes alone, while the Liberals and Greens may either poll their targets on primaries or cross the line assisted by feeder groups.

It is easy to play up the chances of the Greens missing out by pointing to the real chance that they will poll below a quota.  However, to argue that they are at serious risk of losing, it is not enough to just project their vote as landing below a quota.  It is necessary to also construct a credible scenario in which that causes them to lose, in spite of the likelihood of Labor preferences.

I have tried, above, to come up with a credible way in which an outcome other than 3-2-1 might occur.  I think these alternative scenarios are at best remote, and more likely very remote, possibilities.

This article will be updated closer to the election, especially as contesting parties and preference flows are known.

Update 15 June: ReachTEL

As reader intuitivereason has already pointed out in comments, as dire as the ReachTEL polling is for "the left" it still points to a 3-2-1 Senate result.  Labor has its two quotas with the undecided redistributed and Labor + Green are much closer to 3 than the Libs are to 4. Indeed, the Greens House of Reps vote is strongly deflated by competition from Andrew Wilkie in Denison and it would be much closer to a quota, if not over, in the Senate.  The good news for the Greens when it comes to competition from Wilkie is that there is so far only one of him and he cannot run against them everywhere at once.  If Wilkie clones start hatching out all over the place, they'll have big problems.


  1. With the ReachTel poll now out showing ~3.4/2.0/0.8 on primaries, 3/2/1 certainly does remain the most likely result.

    The Greens primary vote wouldn't want to drop much more between now and the election though.

  2. Kevin,

    i have give this article some publicity as one of the better of thev week.

    I hope you get more visitors here

  3. Hi Kevin

    At the last election the Greens got just on 1.5 Quota's which in the event of a double dissolution is 3 seats, how likely do you believe it is that the greens could pull all the way up to say 1.6-1.7 and gain a 2nd seat on preferences?

  4. The Greens actually polled 1.42 quotas in Tasmania in 2010, which would have been 2.64 quotas at a double dissolution. Nonetheless if 2010 had been a double dissolution they would indeed have won three seats (the third on Labor preferences). (Note that the quota at a double dissolution is not actually exactly half the normal quota. Normal quota is 1/7th of total (plus one vote then rounded down), double dissolution quota is about 1/13th (ditto).)

    I give the Greens no chance of increasing their Tasmanian Senate vote this year unless something extremely unusual happens. The only scenario I can think of in which I would even consider it possible at the moment is if Kevin Rudd is expelled from the ALP and becomes Greens leader.

  5. Dear Dr Bonham,

    I was wondering if you had any thoughts as to what would happen to the outcome IF in the unlikely event Andrew Wilkie were to run for the Tasmanian Senate rather than the Seat of Denison he currently represents?

    Kind Regards, ~Ian..:)

  6. If Wilkie ran for the Senate I think the most likely outcome would be three Liberal, one Labor and a three-way fight between Wilkie, the second Labor candidate and the Greens for the last two seats. Wilkie would probably get about half his quota in Denison alone (unless there was a backlash against him switching) and would need to average about 9% in the remaining electorates to get a quota outright. I'm not sure if he'd do that or not, but I don't think he'd be too far short. I think there is enough evidence at least that parts of Franklin would give Wilkie votes that were well into double figures.

    A Wilkie Senate candidacy would make it very hard for the Greens to get quota statewide and I think a fairly likely scenario would be that the Liberals plus minor right parties got, say, 3.5 quotas, Labor a bit less than two, the Greens a bit less than one, and Wilkie a bit less than one. In this circumstance the fate of Labor 2/Wilkie/Greens would be determined by preferences from the minor right parties plus the Liberal surplus. The Greens would not want to be needing too many votes in such a case as it's likely the Libs and minor right parties would generally preference Labor or Wilkie, if not both, ahead of them.

    A difficulty for Wilkie if he ran for the Senate would be that to avoid making his voters vote below the line, he would need to run a group ticket and lodge a preference allocation. He'd need to either compromise his independence or else split his group-ticket preferences three or more ways. If he chose the latter then, of course, opponents could say that an above the line vote for Wilkie is a third of a vote for the Liberals.

    There would also be a risk that the eventual vote was less right-leaning than the current polling patterns suggest, in which case it's possible to imagine that the Liberals would get three quotas, Labor and the Greens together would have three quotas (and preference-swap) and Wilkie would end up left with 0.8 of a quota or so and nowhere to go.

    1. Dear Kevin, thanks so for your prompt & interesting reply, it is much appreciated! I also asked @Poliquant this question some months ago, but because his website is now closed and I've been as yet unable to locate the copies of my query & his reply on my FB Timeline, I can't copy those to you. I will attempt to find aforementioned over forthcoming days. With Kind Regards, ~Ian..:)

  7. The Australian Senate Proportional Representation system needs is seriously flawed and needs urgent reform. The system of counting the votes is not "proportional" but semi proportional. It was designed to facilitate a manual count. With the use of computer aided counting we no longer need to prop up an outdated counting system. The Tasmanian "Last Bundle" system is also in need of review.


    There is no overriding justification to maintain the Droop quota where the cake is divided by seven and one slice thrown way. Instead we should be adopting a full pure proportional count with the quota for a States half senate election to be one sixth not on seventh + 1 Why should a seventh of the community go unrepresented disproportionally. The argument of that it provides majority representation is also flawed.

    The Droop quota distorts the proportionality of the vote.


    Currently the surplus transfer value is calculated by dividing the surplus by the number of ballot papers that make up the surplus. This means that votes worth a fraction of value and transferred at the same value of full valued primary votes.

    The Surplus Transfer value should use a weighted calculation based ion the value of the vote not the number of ballot papers


    The currently system of segmentation was implemented to facilitate a manual counting process and does not reflect the voters intentions.

    With a computerised counting system there is no need to segementation the count. All ballot papers can and should be transferred in a single transaction. So as to reflect the voters intention it would be best if the ballot count is reset and all ballots redistributed according to the fist available voters preference as if the candidate(s) excluded had not stood. One transaction per candidate with surpluses distributed on each iteration of the count until all vacancies have been filled following the distribution of any surplus votes

    In 2007 Queensland Senate we elected the wrong Senate candidate as a result of the method of segmented counting where votes skip candidates second and third preferences and are allocated at full value to a candidate of lower preference.

    A reiterative cont where the vote count is reset and restarted afresh would reflect the voters intentions

    To reiterative count by a manual process would have been too time consuming. With the use of computer aided distribution count a reiterative count would be completed within to 2 to 3 hours,

    These changes are relatively easy to implement. The method of voting and data entry would remain the same.

    The outcome being more accurate and proportional to the distorted outdated system currently used

    1. Unfortunately the method requires computerisation for viability (which introduces significant issues with verifiability), and the requisite paper checks would take forever. To throw away a time honed, integral, respected voting process - which is what has to be done to gain the nominal advantages of computerisation - for the sake of a esoteric difference in result is simply foolishness.

      Even so, the suggestion that one method selected the wrong candidate is false, unless it was incorrect by the method elected for use. Now you may not like the method, you may not like the candidate elected, but that doesn't mean the wrong candidate was elected. That another method would elect a different candidate is also irrelevant.

      The additional work required to further reduce the number of exhausted votes would be a considerable impost, for little gain.

      Tasmania already has one of the best systems going for electing representatives for the lower house. If you are fussed about electoral systems, there are a lot better places to invest your time advocating for change.

    2. For the record I agree with d@W about the surplus transfer issue in Senate counting and I could hardly believe it when I found out what the existing system was. It's one case where the current system is unfair (tending to overweight small surpluses containing lots of ballot papers - which tend to come from voters for big parties) and easily fixed.

      I can't really get excited about the rest until much bigger fish, starting with the above-the-line party preference ticketing scam, are fried.

  8. If the predicted results are 45-31-15-9 like you indicate here, then my runs do start to show AIP getting up on sparse occasions - 4/50 runs (twice of those at the expense of one Labor seat). FF wins are also very sparse at these numbers, only 6/50 and all at the expense of the third Liberal.

    1. Thanks, that's a useful test of how the model goes when allowed to vary by small amounts, which I didn't experiment with very much; I was more interested in testing what happened if it was totally wrong for an obvious reason. The whole thing is very difficult because we don't have a clue which micros might do well on primaries but the pattern of FF and AI having the best preference flows is quite a strong one. The issue with AI is that all else being equal they would be lucky to get a few hundred votes in this field unless people were suckered by their sneaky name. So it will be interesting to see where they are if the Liberals don't smack it out of the ball park on primaries.

      One reason the Green vote is much higher in my model in the Senate than in the House is that a lot of the Green vote in Denison is soaked by Wilkie. Another reason is that the Greens tend to outperform their House vote in the Senate anyway - a few voters vote for a major in the House but Green in the Senate because they do not believe either major should have a Senate majority.

      I tend to agree that the Others vote is more likely to be higher than my 9 points than lower. Might look for some historical comparisons from WA or SA to see if I can find matchable examples with three prominent parties and a similar total number of candidates. Another option might be to just try a general regression of Others vote against number of groups for whatever cases with three significant parties exist.

      Something else I've noticed is that having a sitting crossbencher can boost the Others vote in the Senate in that electorate. That would happen partly because there was no incumbent party to get an advantage on postal balloting and partly because when voters are accustomed to voting indie in the Reps they may be more likely to do so in the Senate. Kennedy and New England displayed this strongly last election although regional factors also influence it. Possible there will be a greatly increased Others vote in Denison in the Senate. (I have accounted for Wilkie suppressing the House Others vote to a degree.)

    2. I also suspect the 'pox on all your houses' - which now includes the Greens - attitude that is out there will force up the 'other' votes, more than would be expected by just the increase in number of parties. Mind you, it could be argued that the higher number of other parties is a response to this.

      I set the minor parties with large variance in my model so that over a series of runs it would throw up a whole variety of flows. It really doesn't seem to make much difference.

  9. Ok, a little more on conditions for certain outcomes, mostly concentrating on the oddities where FF gets up.

    2-2-1-1 : If the Liberals get less than three full quota (allowing for leakage through the count), FF are almost certain to get a seat under anything like current polling conditions. Unless Labor makes above 38%, the Greens get the last seat.

    If the Libs get their 3 quota, apart from the circumstances below, it will be 3-2-1-0. 3-3-0-0 is looking unlikely.

    For 3-2-0-1 to occur:

    Lib + FF > 49.5%
    The Five (Lab + KAP + SEX + PP + GRN) < 43.5%
    FF + Other > 8.5%

    You then probably need to add a margin of another 1% or so for leakage during the count. So really, for this to be a likely outcome, you would have to see a drop of another 2-3% in the 2PP.

    Libs only start getting 4 seats if they have a 52% first preference, along with The Five getting < 40% between them. Not happening this time, black swan aside.


The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. Comments will be published under the name the email is sent from unless an alias is clearly requested and stated. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.