Saturday, October 19, 2013

Senate Reform: Change This System, But To What?

As noted at the top of the previous post, the pace of new material will be slow for the next few weeks because of work commitments, but there will still be new posts from time to time. Aggregate, Fairfax and WA Senate updates may also be a bit slow - I'll be aiming for daily, in the evenings (when there is actually anything to report), but don't guarantee to meet that target.

The Senate post-count isn't over yet as we've barely started the laborious WA Senate recount, which may well be followed by a court case depending on the margins and the outcome.  News on that process will be posted here.  But the results are final enough to make some observations on what the elections told us about the faults of the current Senate electoral system.  There will be a standard Joint Standing Committee on Electoral Matters review of the election and this is likely to pave the way for a process leading to reform including opportunities for public submissions and hearings.  Now is a good time for those interested in reform to be discussing ideas and priorities for alternatives to the current system.

In my view, the overwhelming priority is the abolition of group-ticket voting.  The various alternatives have their strengths and weaknesses, but all of them pale into insignificance compared to the importance of having preferences directed (or perhaps in cases exhausted) through the actions (or inactions) of voters, rather than by preference deals between parties.

What's wrong with the current system?

The following are in my view the main things that this election demonstrated to be wrong with the current Senate system.  They are all major, are each fatal to the system's credibility as a democratic electoral process, and are not in any specific order.

1. Snowballing from very low primaries

One candidate (Ricky Muir, AMEP) was elected on much less than one percent of the primary vote.  Another (Wayne Dropulich, Sports) was very nearly elected (and may yet be elected) on less than a quarter of a percent.  (Update: Dropulich will be declared elected following a recount, but this is not a result likely to stand.) Various other candidates polling considerably less than one percent came reasonably close to election via preference snowball in various states.  It does not matter what the calibre of such candidates is or whether they turn out to be better or worse Senators than the usual mob.  What matters is that no proportional representation system that did not include group ticket voting would have elected them, since they have negligible primary support and a natural preferencing flow would not have got them over the line even if everyone directed full preferences.

2. Dependence on irrelevant events

The Western Australian preference distribution especially highlighted this problem.  The makeup of the final two seats was eventually determined by the question of which out of the Australian Christians and the Shooters and Fishers party was excluded at an early stage of the count.  Victory as a result of such a contest, between two parties that were neither in the race for a position themselves nor ideologically related to the parties contesting the final spots, had nothing to do with electoral merit.  Furthermore, other such contests at earlier stages, such as HEMP vs Animal Justice, were also resolved by small margins and could have gone the other way had certain parties made different preference decisions at seemingly arbitrary points.

In normal preferential multi-candidate elections this problem very rarely occurs, firstly because all parties with low primary votes will usually be eliminated sooner or later anyway, and secondly because even if there existed a network of similar parties that strongly preferenced each other, the exclusion of any would favour any other member of the network, so roughly the same sort of candidate would still be competitive.  In the Senate system the idea that preferences flow between ideologically similar parties is often violated by preference deals.

3. Frequent perverse outcomes

In many scenarios at this election, parties might have won seats had their own position on various votes been worse, or lost them had they been better.  For instance the Liberals would have won in Tasmania had more of their voters mistaken the Liberal Democrats for them and voted Liberal Democrat instead of Liberal.  The Sports Party would have won in Western Australia had the Wikileaks Party not preferenced the Sports Party at a certain point in the count and instead preferenced the Animal Justice Party.  The Greens would have won in Western Australia had a handful of their own voters instead voted for the ideologically opposed Australian Christians.  The Palmer United Party would have been at risk of losing the seat they eventually won in Tasmania had enough voters who voted informal instead voted for PUP.  The technical term for this sort of thing is non-monotonicity.

This is an occupational hazard in preferential elections - to a degree.  Some examples include the 2009 Frome by-election in South Australia (the Liberals would have won had a handful of their voters instead voted Labor) and the 2010 Denison federal result (Labor would have beaten Andrew Wilkie had enough of their voters instead voted Liberal.)  However, in such elections it is reasonably rare.  In this Senate election it was normal, happening in several states, often multiple times in the same state.  It is virtually impossible for parties to strategise effectively when there is such a complex relationship between their vote and the result.  Again, this is largely a product of group ticket voting; it is strongly connected to the problem above.

4. Oversized ballot papers

The system encourages large numbers of candidates.  Large numbers of candidates resulted in oversized ballot papers with very small print.  This had its maximum impact in New South Wales, where the perfect storm of:

(i) the Liberal Democrats having a similar name to the Liberals
(ii) the Liberal Democrats drawing first place on the ballot
(iii) the Liberal Party drawing well away from the Liberal Democrats
(iv) the sheer number of parties

... resulted in the Liberal Democrats winning a seat.  Again, this has nothing to do with how good or bad a Senator David Leyonhjelm will be; it is just a fact that he was astonishingly lucky to be elected as had any of (i) to (iv) not happened he probably would not have won. 

5. Absurd preference deals and strategies

Pragmatic games of attempting to gain advantage saw many parties highly preference other parties that their supporters would oppose, or demote parties or candidates their supporters would view favourably.  There were even cases of parties down-preferencing others not for pragmatic advantage at this election but out of spite.  The most influential example of dodgy preferencing involved Labor and the Greens preferencing Family First above less ideologically incompatible forces, resulting in the election of free-market social reactionary Bob Day (SA) and nearly the election of anti-gay-rights extremist Peter Madden (Tas).  Labor seemed to have some incentive to do this at least in the case of South Australia, but the Greens' decisions to preference FF appeared to be founded in spite (SA) and extreme cluelessness (Tasmania).

Another notorious case involved the Animal Justice Party preferencing the Liberals ahead of the Greens in the ACT, though in the end this did not cause Simon Sheikh's defeat, as the Liberals would have won the seat thanks to below-the-line preferences anyway.  In an interesting (though in places inaccurate) article about minor party preference deals here, the Sex Party's Robbie Swan attributes this decision to the AJP punishing the Greens for "agreeing to a kangaroo cull earlier this year [..] Animal Justice felt they had been betrayed by their old allies."

There were many contentious cases involving micro parties, such as the Sex Party's still-unexplained high preference for One Nation in New South Wales, the Wikileaks Party's preferencing of the Nationals in Western Australia, and the LDP cluster of micro-parties all failing to lodge their group tickets in Victoria (thus disadvantaging the parties they had dealt with).

Ultimately, the idea of group ticket voting works well if it is assumed that parties direct their preferences in a way that resembles what their supporters would do.  But this is not the case; parties game the system for immediate or long-term tactical advantage or even direct preferences incompetently.  Voters not wishing to spare the time to vote below-the-line have their preferences sent somewhere they had no idea they would go, and in many cases are not aware that their vote may be live at full value for a party they'd never willingly support.

Having mentioned those five major issues with the system I will also throw in a sixth that is not so major but that nonetheless rankles, as it is undemocratic and there is no benefit to it.  This quibble is a little bit specialised and technical, so please feel free to skip it:

6. "Inclusive Gregory"

Inclusive Gregory is the system that determines how votes are allocated when a candidate exceeds their quota and has a surplus.  What happens is that the total by which the candidate is over their quota is decided by the total votes held by the candidate.  A notional quota remains with a candidate, and all their votes are redistributed at reduced value, being collectively worth one quota less.

So far, so good, but this system ignores the value that votes previously had as a result of being reduced in value at previous surpluses, and instead determines their value in the surplus based on the number of ballot papers showing particular values.

So, for instance, a vote for the ALP in a given state might help elect two Labor Senators out of a total of 2.3 quotas. The vote flows on and its new value is 0.3/2.3 = 0.1304 of a vote, with 0.8696 staying with the elected candidates.  Some time later, the third Labor candidate is excluded, and their votes flow to a different candidate, say a Green, who crosses by 0.15 of a quota (0.85 of a quota in primaries plus 0.3 from the ALP surplus).

In a fair system the continuing votes flowing from the ALP surplus would now be worth 0.1304*(0.15/1.15) = 0.017 of a vote.  Of the original one vote, 0.8696 stayed with the Labor Senators, and .1134 was used electing the Green.  Most of the Green surplus would now come from the remainder of their votes when they crossed the line.

But in the Inclusive Gregory system, the Green candidate has 2.3 quotas of ALP ballots (by number of papers) and 0.85 quotas worth of ballots from other sources.  This causes the on-flowing ALP ballots to be worth 73% of the Green surplus, and the other votes to be worth 27%, even though the value of those votes before the surplus was 26% and 74% of the Green total respectively.  The effect is that the Labor ballots make a total contribution to the count of more than one vote each, and the Green ballots and other ballots reaching them make a total contribution of less than one vote each.

We go out of our way to eliminate malapportionment (eg through differently sized electorates) in the Lower House because it unfairly means some people's votes are worth more than others.  Yet we tolerate a Senate system for calculating surpluses that does exactly the same thing, and we do so completely unnecessarily because trivial coding changes in computer calculations would fix it.  Indeed, a superior system is used in National Union of Students elections around the country (although NUS ruins it by having a single down-the-page ballot order in elections very prone to ballot order effects.)

What's right about the current system?

I believe points 1-5 at least provide an overwhelming case that the current system is wrong and needs to be ditched.  But to avoid repeating the mistakes of the past and perhaps making new ones, we need to understand where we're coming from and what was done wrong then.

The 1949-1983 System

The current system was introduced by the Hawke Government following its victory in 1983 and first used in the 1984 election.  It replaced a system that had been in place since 1949.

The 1949 election was Australia's first Senate election to use proportional representation.  Prior to that the Senate used a system very prone to electing slates of Senators from the winning party, and which often led to near-total domination by one party or other.  The system that ran from 1949-1983 was much more proportional.  It did allow for victories by minor party candidates, but they had to poll a substantial portion of a quota to have a chance.  I have found no instance of a minor party candidate winning from less than half a quota - in the current situation, this would mean minor candidates would become competitive at around 7% of the state vote, or 3.8% for a double dissolution.

From 1949-1983 voters were required to vote in modern below-the-line style, numbering effectively all boxes.  There's a view that towards the end of this system's life, increasing minor party diversity was driving the informal rate upwards and that ballot papers were being flooded with fake micros for this reason, but looking at the most populous state, NSW, it seems that there was not that strong a corellation between candidate numbers and informal rate over time.  The highest informal rate in NSW was in 1961, 12.7% with just 25 candidates; in 1983 it was 11.1% with 61.

Looking at the patterns in candidate numbers I infer that the Whitlam government made changes that allowed for greater candidate numbers, as there was a massive jump in 1974.  Perhaps altered saving provisions curbed the increase in informal rate that happened at the same time?  I haven't researched this. [Edit: apparently the massive jump represented an orchestrated stack of spurious parties designed to damage Labor.]

What is clear is that the system always had an unacceptably high national informal rate.  This averaged 9.7% in elections held concurrently with House of Representatives elections and 6.6% in those not. 

The other problem with the system was that the distribution of individual preferences was extremely tedious.  Even running a similar system by computer, as was done nowadays, with all votes distributed would involve a massive increase in data entry load.  The impracticality of individually distributing millions of papers (often repeatedly) in the larger states meant that papers to be distributed had to be selected randomly.  In the case of very close counts, like the 1980 WA count, this introduced the risk that the winner of the last seat was winning because of which votes had been randomly chosen and that a different random selection might create a different result.

The central problem

The central problem of Senate system design is how to trade off three goals: 

A. High formality: Excluding those voters who deliberately vote informal (eg by submitting a blank ballot) we would like to capture as much voter intention as possible, since if there is a significant rate of unintentional informal voting, there is the possibility that the election result will be skewed.  Especially, Labor supporters are in principle more likely to vote informally if the system is complex or onerous than Coalition voters. 

B. Full preferencing: Ideally we would like to encourage voters to direct preferences all the way down the ticket and not to exhaust their votes.  If there is a high exhaust rate then competing parties with similar ideas effectively "split the vote" because there is only a weak preference flow from one to the other when one is eliminated (this is seen with the Greens and Labor under Hare-Clark with semi-optional preferencing in Tasmania, and is one reason why the close federal 2PP vote in the state is not a portent of a similarly even State election).

C. Voter control over preferences: Ideally we would like the flow of preferences to reflect what voters actually want, rather than an assignment of preferences that they are not aware of, would not understand if they were aware of it, and would not accept if they understood it. 

The current Senate system delivers on A and B but disastrously fails on C.  It replaces a system that delivered on B and C but seriously fell down on A.

Some proposed solutions

Ideally the Australian Senate would be elected nationwide using a multi-member proportional system similar to those in European nations.  This would almost never result in a party winning a Senate majority but this does not matter as the Senate is an upper house and it very rarely happens anyway.  Party lists could be employed and voters granted some ability to direct preferences between parties that would assist in mopping up the scraps of the few quotas not won on primary vote.   The obstacle is that the Constitution requires that each state have Senators and an equal number thereof, although this is a gross violation of one vote, one value.  It's very unlikely that voters in those states affected by a more proportional system would agree to change the system by referendum, and support from a majority of states is required for a referendum to carry.  So we are stuck with the current system of electing twelve Senators from each state, in alternating tranches of six except where there is a double-dissolution.

Some examples of proposed solutions are:

i. Threshhold for election

Under this system the current system would largely remain but it would be necessary for a party to poll 4% (or some other figure) of the primary vote in a state to be eligible for election in that state.  All candidates for parties not polling at least that much would be bulk-excluded and their preferences distributed early in the count, probably after the surpluses of those candidates elected on primaries and those of any eligible candidates elected on their surpluses. 

This system would remove the possibility of parties snowballing to victory on tiny percentages of the vote.  Possibly, this alone would deter some of the micro-parties from competing.  However, it would not stop horse-trading between those parties capable of getting 4%, and the number of such parties would be likely to increase as some of the micro-parties either did not run or merged to avoid splitting the primary vote.  Candidates like Muir and Dropulich would no longer win, but the Bob Day scenario in SA could easily repeat (since Day polled just under 4% in the current system).

Furthermore, while micro-parties would no longer win (or would be encouraged to merge into broader niche parties that were more competitive, eg a broad libertarian right party, a broad left-libertarian non-Green party, a broad Christian-right party) they could still use their preferencing power to influence political goals.  So it's not clear how much this would really cull the candidate list.

Of the five main problems I listed at this election, this reform would stop 1, but would have a limited effect on 2, 3, 4 and 5. Personally, I'm not keen on it, although it would be some improvement.  I don't think it's likely a party with a primary vote of 3.5% would ever win by genuine voter intention at a half-Senate election, but I'd rather that they had the chance to do so if that were ever the case.  And for a double dissolution, a 4% cutoff might well eliminate candidates who would otherwise win fair and square, since the quota for election is only 7.7% and may well go down should population increases cause an increase in the size of Parliament in future.

Note: there is discussion about the possible unconstitutionality of threshholds in comments.

ii. Full Optional Preferential

This option works the same way as lower house state voting in NSW and Queensland.  Voters can vote for as few or as many candidates or groups as they like.  When every candidate they have voted for or preferenced is elected or eliminated, their vote goes to exhaust and plays no further role.

This system would necessarily maintain an above-the-line option, and most likely the voter would be free to preference to their heart's content either above or below the line.  An alternative version is to make preferencing optional above the line only, and to retain the current below-the-line rules.  A further alternative option is to abolish below-the-line voting for candidates and just have the voter preference as few or as many parties as they liked; votes for parties would automatically follow the party ticket.  (In practice, this happens a lot of the time anyway, and candidates who are low down on the party ticket do not beat those higher up on it.)

Parties might try to direct preferences using how-to-vote cards, but most likely those that directed too many preferences would discover that their voters didn't bother following the card and just voted 1. 

Whatever version of this system was used, it would completely eliminate problem 1 and more or less eliminate problem 5, and greatly reduce problems 2, 3 and 4.  However, while scoring highly on goals A and C, it would do so by flunking on goal B, resulting in high exhaust rates and a degree of similarity to first-past-the-post outcomes.

(I note in passing that there is a kind of misconstrued reverence for the quota that comes up when some people talk about OPV.  They consider that it causes unacceptable distortion because the proportion of exhausted votes makes quota harder to reach or causes candidates to win without quota.   It's really not a big deal.  Quota is not some magical sacred figure that must be cherished at all costs - it's basically a shortcut to quickly elect those candidates who were certain of election anyway and to allow them to pass on votes to others if they have way more votes than needed to achieve this goal.  If you're really bothered by it, you can use a progressively reducing quota as votes exhaust, which is dead easy to implement.)

A problem with full OPV is that to avoid voter confusion,  there would be a lot of pressure to introduce it at House level as well to prevent informal voting from confusion between the systems.  That would be bad news for the ALP, which won twelve seats from behind on preferences at the 2013 election, and would have lost six of those with an across-the-board exhaust rate of 50%.  So I suspect Labor would be reluctant to support full OPV.

A note that applies to any form of optional preferencing is that it is important to build voter awareness that while stopping early is convenient, it also limits the power of your vote.  That said, a substantial proportion of voters will always stop as soon as they can, because they are only voting because they are forced to.

iii. Semi-Optional Preferencing

The idea of semi-optional preferencing is that a vote must number a certain number of squares to be formal but can stop from that point on.  An example was the one repeatedly introduced by the Greens in recent years: a voter may select three or more parties above the line, or six or more candidates below.  This was rejected by the other parties with JSCEM finding that voters liked the above the line system because voters were increasingly using it.  In reality, ATL is popular by default, because BTL is increasingly onerous as the number of candidates increases.

A consequence of the Greens' policy would be that major parties would deliberately run teams of six candidates so that a voter could vote below the line for just that one party without their vote becoming informal.   Indeed parties might well direct their voters to do this.  An alternative is to allow semi-optional preferencing above the line only (with only full preferencing if voting below the line, which hardly anyone would do.)  Another is to scrap BTL voting and have semi-optional voting for parties only. (Note: It's been suggested this might be unconstitutional - see discussion in comments.)

There is also the option that mirrors Hare-Clark in Tasmania and the ACT - scrap ATL and require a voter to number only a number of candidates equal to the number of vacancies (they can number more if they want to).  Because parties will preselect enough candidates for a voter to just vote for their party and no others, this system then has a similar exhaust issue to fully optional ATL.  But it does have the benefit of forcing candidates to compete for their own party's vote, meaning that unpopular major party Senators may lose within-ticket contests.

Semi-OPV also removes all of the problems 1-5 that I mentioned with this election.  Of the three objectives, it increases the rate of preferencing, but a tradeoff is that the informal rate goes up.  Another is that counting becomes more complex and expensive, since every vote has to be manually entered in the way in which all BTL votes are entered now.

If semi-OPV was introduced there would also be pressure to bring it in at House level.  In my view this would not be such a bad thing as it would reduce the informal rate in electorates with long candidate lists, while having much less impact on the ALP than full OPV.  Tasmania uses semi-OPV in the Legislative Council (voters must number 1-3) and while the system seems a bit quaint when other systems do not copy it, I think it is not a bad tradeoff.

iv. Compulsory Preferencing With Above or Below Option

I've seen this suggested in a few places.  The idea is that voters could vote either above or below the line and would have to number all squares whichever they did.  The idea is that voters would mostly choose to number above the line and would number boxes for all parties.  If they made errors their votes would exhaust at the error point, as current Senate votes do.

This system has some of the same problems as the old Senate system.  Even when required to number just, say, 20 boxes, voters may make more errors than the saving provisions allow, so informality may still be high.  I also think that even if the below the line option is removed, the data-entry costs for this system would be very high.  I also suspect this system would cause donkey voting, which is not currently a big issue in the Senate.

v. Limited GTV

This interesting idea is canvassed by Antony Green here (in comments).  Voters would retain the above-the-line just-vote-1 option but an above the line vote would only be distributed to the chosen party plus a small number of others (say, 3-4 other parties) and then exhaust. 

The attraction of this idea is that it maintains high formality while also keeping exhaust rates relatively low, a combination that no other system that does not allow for rampant snowballing can do.  It would seem to make it very difficult for parties with trivial vote share to snowball up via a network of deals with other micros, since they can only offer their own preferences to a small number of other parties.  However, micros could circumvent this by forming cross-state networks - for an extreme example, 24 micro-parties might each run in every state with all those in NSW preferencing parties 1,2,3,4 from their group, all of those in WA preferencing parties 5,6,7,8 from the group and so on.

The other tricky aspect of this solution is that parties that preferenced parties with no chance to get elected would be prone to the accusation that a vote for them was a wasted vote.  While this actually might be a selling point for a small proportion of them (since some voters quite like being able to deprive both the majors of support) there would be a lot of pressure on micros to preference their ideological wing's bigger parties, and this might actually distort the outcome in favour of big parties.  

I think a more limited form of GTV is an idea prone to unintended consequences that would need to be thought through very carefully.

Other Matters

These are matters, some of them minor, that are largely independent of overall system choice.

Fix Inclusive Gregory: This is a no-brainer.  The Inclusive Gregory system must be replaced by the Weighted Inclusive Gregory system, the Wright System, or some other system that does not allow a vote that has been reduced in value to later unjustly increase its value.  Any politician who does not support its replacement should also be replaced by the voters at the next opportunity.

Improve Saving Provisions: Often examination of the AEC BTL data reveals that voters have had a vote exhaust from the system where, for instance, the vote appears to have two fours, one for a party for which the voter also assigned 3 and 5, one for a party for which a voter also assigned 39 and 41.  Even if it looks for all the world like there are two 4s rather than a 4 and a 40, I think the saving provisions could be improved to recognise the overwhelmingly likely intent of the voter, and keep their vote from exhausting.

Parties With Confusing Names: Some people think the Liberal Democrats should be banned from running under that name again as it causes confusion with the Liberal Party.  But I agree with the court finding that words like liberal, democrat, Australian, Labor (etc) are generic words that parties should be able to compete for.  To allow the Liberal Party to hog the word "liberal" under a trolls-who-were-there-first style provision means that the Liberal Party can pass itself off as a liberal party without any pressure to actually be one, and that no genuinely liberal party can compete for the label.  To me that's unfair, and on most issues the LDP (mostly for better, in some cases for worse) is a more "classically liberal" party than the Liberal Party.  Perhaps if the Liberal Party is so worried about being confused with a liberal party, it should try changing its name to something more accurately describing its nature - for instance rebadging itself as what it is, a Conservative Party.

Instead, I like the suggestion that parties that receive a certain share of the state Senate vote in a given election - say 4% - are placed ahead of those that do not, on the Senate ballot at the next election.  The unfair advantage accruing from this is negligible given that there is virtually no donkey voting in the Senate, and it means that if two parties do have similar names it will be easy for voters for the larger such party to find them on the ballot.  (Note added Nov 4: Truth Seeker suggests a simple poll-position type system in which parties would qualify for ballot spots in order of their performance at the previous election.)

Increase Registration Difficulties: I'm not mad keen on the push for greatly increased deposits, membership numbers to register and so on.  I like the idea that a small party can run for an election cheaply, and if it does so successfully can build up its following.  But I do think it should be required that all the electors who register a new party are not registering electors of any other extant party, to stop overlapping sets of electors registering multiple front parties.

How To Vote Card Reform: This is mainly a Reps issue but also one that may be more significant for the Senate depending on the reform adopted.  I would like to see very strict legislation concerning misleading claims on HTV cards.  For instance, ALP cards that say "you must number every square as shown" are designed to dupe voters with poor English levels into fearing that their vote will be informal if they don't.  Apparently it doesn't work anyway, but I'd like to see all HTV cards be required to carry a prominent statement that the card is a suggestion only and that a voter for a party can distribute their votes as they wish (according to the system adopted) with their vote remaining formal.

Concluding Comments

It is difficult to find an ideal solution to the problem of what to change the Australian Senate system to.  At this stage, I prefer some kind of OPV or semi-OPV approach over the threshhold system or the option of above-the-line compulsory preferencing.   But I don't want to fly the flag for a specific solution just yet as there will be a lot of debate to come.

What I do want to indicate is that in my view virtually anything is more democratic than the farce we have endured at this election.  The most important thing is that the current system is reformed in a way that will eliminate or greatly reduce the problems mentioned.  There needs to be cooperation between parties to achieve this and a willingness to compromise on preferred systems to ensure we get a better system, even if it may or may not be the best available. 

I welcome comments on other ideas to improve the Senate system.  I do suggest that the constraint of AEC data entry effort and costs be kept in mind when considering alternatives.

This post may be updated with further comments, but when there are major developments in the process towards Senate reform I intend to start new threads to keep the issue prominent on this site whenever something relevant to it is actually happening.


Moderation policy comment: As this issue is one that often attracts the attention of a particular well-known commentator on pseph blogs, I should advise in advance that I'll reject all comments containing the following:

1. Advocacy, implied or direct, of the "Hare quota".  Totally silly idea!
2. Continual imploring of the AEC to publish BTL files during counting.
3. Anything implying, without proving, deliberate AEC malpractice, or canvassing the possibility of it.

Excessive banging on about issues I consider to be minor *might* also eventually get the chop.


Nov 1:  Antony Green discusses the issue of solutions further here.  Also not a big fan of the threshhold system.

Nov 4: In light of the WA Senate fracas I think it's time to add one more to the list of things this election has shown to be wrong with the Senate system:

7. Greatly increased risk of a void result

Because outcomes can be sensitively dependent on early exclusions (see problem 2) and because early exclusions by definition involve candidates with few votes, there is an increased risk of elections that are determined by very close differences.  We saw this in this election with the perhaps irresolvably close S+F/AC elimination in WA (which potentially decides two seats), and also the very close Sex Party elimination in Tasmania, which eventually decided a seat for PUP rather than the Sex Party. 

Very close eliminations increase the chance that even small irregularities in the count will be enough to invalidate the result., causing a by-election at vast expense and under politically unrepresentative conditions.  Such irregularities can include mundane issues like admission decisions for absent votes, double voting, votes not received in time through no fault of the voter, and so on, as well as more serious errors like the loss of ballot papers in this instance.

Eliminating group ticket voting would greatly reduce the chance of micro-close results that fall within the margin of successful litigation.  A threshhold system, by contrast, will not necessarily do this, and may encourage the running of fake parties to spoil the threshhold chances of legitimate minor parties.

Nov 15: The Sainte-Lague divisor method (proportional representation based on primary votes alone, without preferences) is getting attention because it has been proposed for the South Australian upper house. See Antony Green on it here.  Sainte-Lague has the advantages of being incredibly simple to count and explain. It eliminates snowballing and preference dealing and ensures that if there is a close outcome it is one that could be reasonably argued either way and that is determined by primary vote. It guarantees high formality, but the solution to the problems of preferencing control and full preferencing vs exhaust (namely, getting rid of preferences) seems an extreme one. 

With a relatively small number of seats per state there would probably be some big objections to it in the Australian Senate context.  The most significant one is that without preferences, two likeminded small parties might split the vote between them and both miss seats, so the system is likely to force parties to coalesce.

Still, it might be interesting to have a look at what sort of results it would produce.  From the actual primaries in the 2013 Senate election, and assuming the Liberals and Nationals ran as a joint ticket in all states, I've prepared the following table:

Most of the table shows the hypothetical results of a Sainte-Lague Senate election based on the actual primaries.  This is then compared with the actual results of the recent Senate election (I've counted the disputed seats in WA as a half) and on the far right, just for interest, is the result of the national Senate count under a national Sainte-Lague with 40 seats, not split into states. The last option would require constitutional change. (The Others row contains one seat for the Sex Party, with Family First just missing out, although most likely if such a system existed the religious-right parties would merge and win one or two seats.)

The state-based system has the benefit of getting rid of seats for parties that poll next to nothing like AMEP and Sports, none of which would get elected under any realistic system. It also removes Bob Day's undeserved seat in South Australia. It does however diddle PUP slightly (although it's quite debatable whether they'd "deserve" to have won in Tas and been in the mix in WA anyway, so this is really about the problem of segmentation into states, and likewise with the two seats for Xenophon).

There are two problems with this simulation for those who may find its results more attractive than what actually happened.  Firstly this model is based on primary votes polled under the current system, but parties would adapt to Sainte-Lague and adopt different strategies.  That would include micro-right parties merging (since they could not win otherwise).  It's also possible that micro-right voters would gravitate to the Coalition.  And it's very unlikely the Liberal Democrats would be winning any seats based on lucky ticket draws, giant ballots and a confusing name.

Secondly we shouldn't just look at a single election, and indeed looking at 2010 shows up some serious potential flaws in the method.  In 2010 Sainte-Lague implemented on a state-by-state basis using the actual election primaries would have returned identical results to the actual election with two exceptions.  In Victoria, the 2-2-1-1 outcome with John Madigan of the DLP winning by preference snowball, would be replaced by 3-2-1 in Labor's favour. Given that the House of Reps 2PP was 55-45 to Labor this is a little bit harsh on the Coalition, but not that much.  More contentiously, in South Australia the actual 3-2-1 to Coalition would flip to 3-2-1 to Labor.   In SA, the 2PP was only 53-47 to Labor in the House, so a 3-3 left-right split seems much fairer than 4-2.

This underlines what seems to be a serious flaw in Sainte-Lague at a Senate level with state-based electorates.  In a case where the Greens come third (as happened in every state in 2010) they don't need anything like a normal Senate quota to get elected.  All they would need to do would be to beat 20% of the vote of the worst performing of the two major parties.  Eight or nine percent wins them a seat.  This then creates a situation in which Labor and the Greens combined can win four out of six with less than half the primary vote.

Indeed if we merge the 2010 and 2013 state-based Sainte-Lague projections we get 33 Coalition, 29 Labor, 10 Green and 4 Others, which would give Labor and the Greens a blocking majority.  Is this really a fair outcome for a Senate representing the 2010 election (which was a left-right tie) and the 2013 election (which the right won decisively)?  It's debatable.  Labor and the Greens would have 51.3% of the seats off an average 43.5% of the vote, while the Coalition would get 43.4% of the seats off 38.2%

I add to the list of possible Senate solutions that in my view do not require Constitutional changes vi. Abolish preferencing and switch to a divisor system, but I am not convinced it is a good idea.  At least, not in the Sainte-Lague version.


  1. Yes to a weighted Transfer system'
    But missing from the above list is the need to remove the segmented distribution of preferences from excluded candidates.

    The system we have was designed to facilitate a manual counting process.

    With the aid of computers and a weighted surplus we should adopt a reiterative counting system where the vote is reset and restarted on every exclusion.

    "Distributing the vote as if the excluded candidate had not stood"

    remainders should stay with the value of each vote. a single transaction per candidate. Surpluses only distributed on any iteration. The count progressing until all vacancies are elected in a single iteration.

    Try counting the QLD 2007 Senate Ballot by excluding all candidates except the last seven standing. (3 ALP, 2 LNP and 1 Grn)

    Parties should also be able to allocate the percentage of vote allocated to each registered Group distribution ticket and allowed to change the order of preferences within the group.

    if optional preferential voting is adopted then the Electoral Commission MUST include a statement that says

    "In order to maximise your vote you should preference EVERY candidate in order of your Preference"

    In 2010 Western Victoria the VEC promoted Optional Preferential and a group of voters preferences Family First and DLP and stopped after 5 candidates. This had the unintended consequence of electing the Greens to the last position in the count. Had they completed their preferences allocation the Greens would not have been elected.

  2. Western Australia Count

    (Subject to review following the WA Senate recount)

    1. if you had a weighted surplus then Palmer United and the Greens would have been elected.
    2. if you removed segmentation and adopted a reiterative count then Palmer United and the ALP would have won the last two seats
    3. If you adopted a reiterative count and weighted surplus with the major groups able to proportion out the group ticket and removed the Droop quota vote the LNP would elect 2, ALP 2, Palmer United 1 and Greens 1

    Its all in the order of exclusion. At present the system favours minor parties. Segmentation and non weighted surpluses must go.

    The last thing would should adopt is the Tasmanian/ACT Hare Clark "last bundle" segmented system. It is inherently undemocratic.

    A reiterative count best reflects the voters intention and mans that a primary vote will always be allocated to the first candidate in the count at full value. A reiterative count also allows for adjustments in the calculation of the quota with the adoption of Optional preferential voting.

    Above the line voting is akin to using a HTV card.

    Given that most people vote for groups/parties then preferential voting above the line makes sense but should include all parties or a savings provision

  3. Kevin, that's a pretty good run through on a complex issue. The only concept I think it misses to some extent is commentary on the purpose of the senate and how that should influence the decision.

    The nominal purpose of the senate in Australia is to prevent domination of the smaller states by the larger ones. This to a reasonable extent can be generalised as an effort to provide protection on a geographical basic, so that urban areas don't completely dominate rural areas in lawmaking.

    To be blunt, it hasn't worked.

    This doesn't mean it isn't a worthwhile goal - for long term stability, a federal democratic system needs both geographic protection (to make federation viable for a state) and economic protection (so that it doesn't abuse those who pay the costs of the state). Given the original intent, and the continuing need, the system chosen for the senate needs to function in such a way as to preserve and strengthen regionalism.

    As such, I think the concept of a national senate is counter-productive, as it only worsens the shortcomings of the senate.

    Assessing the options proposed against a regionalism criteria:

    Thresholds: - make the thresholds state based rather than national. There were a number of people standing for the Tasmanian Senate who have never even been to Tasmania as far as I can tell.

    Full option preferential would strengthen the dominant party paradigm, and thus weaken regionalism.

    Semi optional preferential, especially with some of the Hare Clarke style provisions regarding competition within each party could be beneficial in terms of regionalism.

    Compulsory A/B would just be a disaster from a count perspective. Compulsory Above may work well, but is anti-regionalism in nature.

    The Limited GTV would be interesting to see in practice. I suspect based upon current practice, that the minor bandings would preference other minors then exhaust without preferencing a major in many cases, leading to significant rates of exhaustion.

    I can't see the regionalism issue being resolved fully without dividing the existing states up further. Possibly the best outcome would involve partitioning off each of the state capital and surrounds into their own new city states, and then considering the remainder in terms of lasting geo-political divisions.

  4. While the regionalism issue is obviously the root of the current Senate system, I've never been convinced that states (or any other regional category) actually deserve special representation and protection any more than any number of other categories. It's the way it has been but that doesn't mean it has to be the way it should be.

    Firstly I believe that no matter what system we had, state-of-residence would not be one of the most powerful indicators of advantage or disadvantage for a particular citizen. I cannot see any a priori reason why there should be 12 Senators reserved for Tasmanian voters and yet no Senators reserved, say, for unemployed voters nationwide. If anything, I think the arguments for specific representation for unemployed voters are stronger (ignoring the demarcation and other practical problems involved), since they have unusual difficulty mobilising as a group. But it's all irrelevant because we are stuck with equal Senator numbers per state as a legacy of Federation; I cannot see that changing any time soon.

    Secondly I think that with the two-party systems that democracies tend to develop, even with a disproportionate number of seats, a state could still wield power if it had to. So for instance if there was a national Senate PR, even two Tasmania Party Senators would have a fair chance of holding the balance of power. Most likely if there were multiple states or regions being downtrodden a national-PR Senate would be flooded with Harradines in the blink of an eye. At present, it's difficult for disenchanted rural voters to band together because there are just not enough of them to get over the state quotas to get elected.

    I didn't canvass Robson Rotation in my article because of the logistic difficulties in printing and counting when done on a massive scale, but it's worth having in the mix if it can somehow be practical. It would be very nice to have voters and not parties deciding who is dropped to third on the Senate ticket. Parties might also become less inclined to preselect bigots or low-profile hacks to high positions, as they would not attract preferences.

  5. It wouldn't be my first pick as a 'category' either; it's far more important to have a house where those whose primary source of income is the government aren't eligible to vote. But we don't have that, and we do have a Federated structure. Even in it's current dysfunctional form, it is what we have, and in so many ways its is an easier task to restore the intent of federation than to introduce something altogether new.

    The simple reason we have 12 senators is that without them we wouldn't have federated. For better or worse, the state is our formal unit of government, with the federation as an overlay.

    Your third and fourth paragraphs align fairly well with my thoughts/expectations given those circumstances.

  6. Thanks Dr Kevin Bonham for your valuable insights and suggestions. We at Australian Senate Ballot Reform agree with the majority of your recommendations. But it's worth your while having a look at Attorney-General (ex rel McKinlay) v. The Commonwealth (1975) 135 CLR 1 ("McKinlay's Case") as to why abolishing or restricting Below The Line voting, or using a threshold against minor party candidates but not the weak candidates of the major parties in the same way, would be unconstitutional.

  7. Thanks for that comment. I could not directly access the Facebook page, possibly because I am not a Facebook member (for reasons related to personal aversion to public "friending" decisions and associated social pressures.) I did manage to access a cache of the front page of the site.

    I also managed to find what seems to be the full text of McKinlay's Case at!article=66561

    Given the lack of direct or even apparent indirect relevance of much of what is discussed there and given that it predated ATL voting I did wonder if you actually meant the similarly named post-ATL case of McKenzie v C of A and others ( (Point 6 may be relevant). If I'm wrong to wonder that I'd be interested to know how McKinlay's Case is relevant.

    Abolishing BTL voting would entail that some candidates would be running for election while being incapable of receiving primary votes, but only as a consequence of their decision to stand as secondary candidates of a party and not as a consequence of their decision to stand as candidates per se. I can see that this could be problematic.

    As for the threshhold issue I wonder if any implication of unconstitutionality in the different treatment of candidates with the same primary votes could be avoided by the following method: Don't set a primary vote threshhold per candidate. Instead set a threshhold for the total vote a candidate must have following the distribution of all initial surpluses and at the point at which exclusion of the candidate with least votes would normally commence. (At this point any major party candidate with any hope will be above the threshhold.)

    Alternatively if the issue is consistency with requirements in the House, perhaps threshholds could be introduced in both at once?

    I'd be quite happy if threshholds were unconstitutional as that would get one of the less good reform options off the table. But it would be good to be clear on it if the concept is really unsalvageable - and I add a general request: if readers here ever want to say something is unconstitutional (and it isn't obvious or general knowledge), please spell out why.

  8. The systems in force prior to 1983 were quite iniquitous and undemocratic. The '49 - '82 rules only survived, I suspect, because almost all voters were unaware of the random nature of the preference distribution.

    In regard to HTVs I should like to see the distribution of these banned on polling days and within (say) 50 metres of pre-polling locations.

    Having handed the things out since 1966 I am confident that only a very tiny percentage of people do not know how they are going to vote prior to going to the polling place and that the influence of the sheets thrust into electors hands is minuscule. The only useful purpose handing them out to voters on polling day serves is to keep partisans moderately sane prior to the commencement of the count.