Optional Senate Preferencing: Not An ALP/Liberal/Green Stitch-Up

Advance Summary

1. This article welcomes the recent interim Joint Standing Committee on Electoral Matters report into the Senate voting system and generally supports its recommendations.

2. One important issue in Senate voting reform not yet adequately addressed is the calculation of transfer values for surplus votes. 

3. A recent article by Malcolm Mackerras claims that nearly all Senators elected in 2013 were elected by informed voter choice and that major parties are changing the system opportunistically while their vote is falling.

4. However, the Coalition's Senate vote only fell in 2013 because of defects in the current Senate system.

5. At least four and possibly as many as seven Senate outcomes in 2013 did not fairly reflect the will of the voters.

6. The idea that micro-parties combined should be entitled to seats in proportion to their total vote share assumes that micro-party voters very strongly prefer other micro-parties generally to the bigger parties.

7. Analysis of actual Lower House micro-party preference flows shows that any such assumption is false.



(Note: Comments about surplus calculation methods have been revised following feedback.)

=============================================================

I've been a bit slow off the mark on this one (mainly because of a high current workload) but thought I'd belatedly add my comments on the fairly recently released interim JSCEM report on the Senate voting system - and also respond to a recent attack on the JSCEM proposal by Malcolm Mackerras.  This is not the whole of the Joint Standing Committee's work on this issue and there will be further hearings, reporting on matters other than Senate voting, and with any luck the demise of Inclusive Gregory, later.

In a previous article Senate Reform: Change this system (but to what?) I pointed out that there is basically no perfect Senate system.  Because many voters are incapable of filling in large numbers of votes without making a mistake, any given system can only strongly meet two of the three goals of high formality, full preferencing by most voters, and voter control over preferences.  The Senate system from 1949-1983 had an excessive rate of informal voting caused by mistakes.  It was then replaced by the current Group Ticket Voting preference system in which formality was high, and most votes expressed a full preference, but most voters had no idea where their preferences were going or what the meaning of them going somewhere was.  At the 2013 Senate election, this system was finally gamed to death.

In another article on all this, JSCEM Comes To Hobart, I gave a link to my submission to the process and stated my support for some version of semi-optional preferencing including optional preferencing above the line.  (A link to the Hobart hearing transcript is also available from that page.)

So, how did we go?

Basic System

Firstly, the committee has indeed proposed optional preferencing above the line with semi-optional below.  As expected it has recommended six squares below the line for a half-Senate election and twelve for a full-Senate election.  There is also this: "The Committee further recommends that appropriate formality and savings provisions continue in order to support voter intent within the new system."  At this stage the saving provision hinted at is that if the voter numbers the minimum number of squares, then their vote is formal even if they made some mistakes in that process (such as doubling or omitting a few numbers).
I did have some concerns that the system would encourage parties to submit needlessly long lists of candidates so that voters could vote formally for just one party in order of choice without their vote becoming informal - hence my suggestion that a requirement of only four squares would be fine - but I don't think this is actually a major issue.  The sorts of voters likely to want to buck their party's ordering are also likely to be politically aware enough to vote across columns for a formal vote.  The rest can vote above the line.

Party Registration

I was not supportive of large increases in party registration targets, since I saw them as inhibiting the ability of a small startup party to make an impact.  My suggestion was a required nomination target to establish that there was enough support for a party in each state.

The committee has found a different middle path between the current registration system and increasing registration targets.  It has proposed that a party would need an increased tally of 1500 unique members (not shared with any other party) to be registered nationally.  But it has also proposed "the provision to register a federal party, that can only run in a nominated state or territory, with a suitable lower membership 
number residing in that state or territory, as provided on a proportionate population or electorate number basis; "

Quite aside from this being necessary for a Tasmania-only party like the former Harradine Group to have a chance of getting registered, I think this solution also well accommodates my concern about having a way for a small party to spring up and make an immediate impact.  Such a party could register in one state with relatively few members, and perform well enough in the state it registered in to attract interest in other states.  Thus, while this approach is a quite different one structurally to mine, I think it meets the same goal, and I support it.

State of Residence

JSCEM has recommended as follows:

"The Committee recommends that the Government determine the best mechanism to seek to require candidates to be resident in the state or territory in which they are seeking election."

While it may seem that non-resident candidates are not an issue if the above proposals are adopted, there is still the potential for a relatively large number of parties to flood a ballot in a given state with candidates not resident in the state, for the purpose of confusing voters for a given party.  I was interested to see that a version of my proposal for candidates to be nominated by residents of a state makes an unexpected appearance here:

"Another option would be to require any non-resident candidates who is nominating for Senate election to be required to provide a list of nominators from the state/territory (similar to the requirement for independent candidates in section 166 of the Electoral Act). "

This is proposed as an alternative to outright banning a candidate from running outside their own state, which may be unconstitutional.

Threshholds

It was pleasing to see that proposals to impose a threshhold for election were not accepted.  (Nor, incidentally, was a thought bubble about requiring a party to get 1.4% of the vote before its preferences could be distributed - a terrible idea that appeared fleetingly as an option in the Liberal Party submission but was then misreported by Dennis Shanahan as the output of both major parties.)

Surplus Calculation Methods

Of the issues not covered in the report and of interest to me, one is the method of calculating the values of ballot papers in a surplus that arises from votes of multiple value.  Again see Senate Reform: Change this system (but to what?)  for my previous discussion of this.  Perhaps this will be explored later.

A (possibly over-trumped-up) concern hinted at is that whatever system is used must be easily replicable by hand.  The initial version of this article argued that the difference between the Inclusive Gregory system and the vastly superior Weighted Inverse Gregory system in this regard isn't massive.  After all, if you can, by hand, calculate and write down an eight-digit decimal and multiply it by integers as required, then you can probably handle multiplying two eight-digit decimals by each other and writing down the result if you need to.

However, I've since been alerted to this problem: in WIG the number of transfer values (and small parcels of votes with significantly distinct histories) does increase much more rapidly - whereas in IG the number of transfer values operating at any given time is one plus the number of surpluses to that point, in WIG in the case of a double dissolution there could well be votes at, say, a hundred different values by the time all the surpluses are out of the way.  I'd suggest that if replicability by hand is really such a big problem, then even switching to the Hare-Clark style last-bundle method would be better than keeping IG.  It is true that that method is incredibly arbitrary in determining which votes that preference a given candidate can continue and which cannot, but at least you don't get votes being worth more than a vote.

I did also receive an email about Meek after describing Meek as a "very complicated" system, from someone familiar with it who insisted it wasn't complicated at all.  My comment was based largely on having had no idea what was going on the previous two times I read about the system and tried to grasp it, but on seeing a fully worked example I found the principle behind it is quite easy to grasp.  The principle is simply that as votes exhaust from the system, not only does the quota required for election decrease, but this reduction is back-applied to the candidates who have already been elected.  So for instance, if the quota is 1000 votes, and A is elected with 1100, initially A's votes are transferred with a combined value of 100.  But later in the count it might turn out that had A started with 950 votes, there would have been no possibility of enough candidates overhauling that total to stop A from being elected.  Increasing the transfer value of A's papers means that the full value of votes not required by A is able to flow to other candidates.  An elegant solution, but an extreme nuisance to hand-calculate since transfer values not only proliferate but also constantly change after exclusions.

In Summary

I think the recommendations of the committee so far - on the issues covered - are excellent and well worth supporting.

The Mackerras Stitch-Up Claim

A counter to this position comes from a piece by Malcolm Mackerras in The Australian.  The argument run by Mackerras is that the ALP, Liberals and Greens have seized upon concerns about preference snowballing and will have "remoulded the situation nearer their hearts desire" to hide the fact that their votes are all declining.

Mackerras starts with the claim that "Both Labor and Liberal seat losses are a true reflection of their lost votes - which tells us much about the fairness of the [current] Senate electoral system". But here there is something to disagree with right away, since the apparent fall in the Coalition Senate vote from 39.9% in 2007 to 38.6% in 2010 and 37.7% in 2013 is actually artificial and under a proper electoral system the Coalition's vote would not have gone down.

In New South Wales, the swing against the Coalition in the Senate was 4.75%.  But in that state, the Liberal Democrats polled 9.5% after they drew first place on the massive ballot paper, while the Liberal column was difficult for voters to find.  Probably only about three points of the NSW Liberal Democrat vote were genuine voter intention, and if the remaining 6.5 points are reallocated to the Coalition then their vote nationwide in 2013 would have been higher not only than 2010 (clearly) but about the same as 2007.  The ballot paper was only so massive because there were so many parties on it, and there were only so many parties on it because Group Ticket Voting either gave them a chance of fluking a seat or allowed them to exist as fronts for attempts by others to do so.  So while the Coalition's seat loss in the Senate does reflect their lost votes, what you're not being told there is that the Coalition lost votes not on merit but because of the defects of this system.

And while it was to be expected that Labor lost a pack of seats for dropping from 40.3% in 2007 to 30.1% in 2013, the proportion of seats they lost (a third) was even greater than the proportion of votes dropped (a quarter).

Reflecting Informed Decisions?

Mackerras characterises the JCSEM report as saying to the Ricky Muirs of this world "You should not be here. To make that quite clear we intend to enact a radical reform of the system.  Your type will never be allowed to enter the Parliament again."  In response to the claim that the incoming Senate will reflect "deal-making and preference-swapping" Mackerras writes "My take is to assert that 36 Senators will be sworn in next month, of whom 35 clearly reflect the informed decisions of the electorate.  We can agree to disagree about Muir."

Disposing of this claim is like shooting fish in an especially full barrel.   State by state:

In NSW, the Coalition (2.39 quotas), Labor (2.20 quotas), the LDP (.665 quotas) and the Greens (.545 quotas) were the leading scorers. PUP had .237 quotas and the tiddlers combined almost a quota between them.  However, the LDP fleeced about .46 of a quota from the Coalition because of the size of the ballot paper and the freak draw helping them to votes based on name confusion.  With this considered the correct result for the state without GTVs would have been three Coalition, two Labor and one Green, since it's very unlikely any of the other parties would have closed down such a massive gap to the Greens on the preferences of the micros.  Instead the LDP got the seat the Greens should have won.

Victoria is already semi-conceded by Mackerras, but for the record the Coalition (2.81 quotas), Labor (2.20) and the Greens (0.76) again were massively ahead of any other forces and should have split the seats 3-2-1.  Instead, Helen Kroger's seat was taken by Ricky Muir.

In Queensland, the LNP polled nearly three quotas and won three seats, Labor polled almost exactly two and won two, and PUP polled .69 and beat the Greens who managed only .42.  Preferences would not have changed the last seat (if anything they would have stretched PUP's margin).  This is the only state the current system got clearly right!

In South Australia, the leading forces were the Liberals (1.92 quotas, +.02 quotas Nationals), Nick Xenephon Group (1.74), Labor (1.586), Green (.4962), Family First (.2631) and Liberal Democrats (.247).  While it might seem that the size of the vote for the various tailenders would give one of the minor non-left parties a chance of catching the Greens, even if they did so in a fair system the Green preferences would just favour Labor (it's also notable that there's about another .24 Q in preferences from left-leaning micros in the mix).  The fair result in SA was probably 2-2-2; instead, both Labor and NXG got done by the group-ticket preferences gathered by the Greens (most notably from PUP) and Family First (most notably from Labor!) It's quite contestable that Sarah Hanson-Young's re-election represents informed voter choice, while Bob Day's certainly doesn't.

In Tasmania, the leading forces were the Liberals (2.62 quotas), Labor (2.30), Green (.816) and PUP (.461).  PUP's Jacqui Lambie beat the Liberals on preference snowball.  I think that under a fair system there's a good chance PUP would have overhauled the Liberals to win the last seat anyway, but it's not exactly clear that this is so.

I have left Western Australia until last, because while Mackerras' claim that the six Senators elected from WA reflected voter choice at the by-election is true, the problem with it is that the by-election happened only because the original election was void.  The original election was void because the loss of a modest number of ballots meant that a tipping point between irrelevant parties could not be resolved, and the Group Ticket system is especially vulnerable to creating this problem because it is innately likely to create irrelevant tipping points.  Then, the fact of the by-election being what it was meant that the vote shares cast were very different to the original poll.  So the true yardstick is not how well the by-election results reflected voter intent at the by-election, but whether the will of the voters the first time around was reflected.

And that will was 2.74 quotas to the Liberals, 1.86 to Labor, .664 Green, .355 National, .350 PUP, .240 LDP.  The outcome of this one under a voter-directed preference system would be not so clear, since left-leaning preferences would have to help put Labor over the line as well as helping the Greens (unlike NSW where Labor's surplus could help the Greens against the right micros).  So just maybe PUP would close down the Green lead (I rather doubt it.)  Anyway, the Liberals would get three seats and Labor two without much trouble, most likely leaving the Greens and PUP to fight over the last one.   Instead we ended up firstly with the absurd election of the Sports Party from a negligible vote, and then a by-election that elected both the Greens and PUP and reduced Labor to a single seat.

In summary, while Mackerras asserts that only one of the seat outcomes did not clearly reflect voter choice, I find that there were seven that were not clear-cut, including at least four results that would surely have gone the other way under a system where preferences flowed by voter choice.  And that applies even if we could somehow, miraculously, get all voters to provide a fully expressed choice of their preferences.

Should Major Party Vote Shares Match Seats?

It's all very well to look at the overall seat results and say that Labor won 30% of the state seats with 30.1% of the votes and that this is some miracle of proportionality. In fact, the Coalition (41.7% of state seats with 37.7%) and the Greens (11.1% with 8.65%) did slightly better than proportionally.  However, if it is desirable for the big three to get the same proportion of seats as votes, then the question is what is left over.  Parties other than Labor, the Liberals, the Greens and PUP won a combined 18.6% of the Senate vote, which would in theory be good for seven seats.  However, even assuming the 36 state seats were elected proportionally by a nationwide system, only one other party group - the Liberal Democrats  - would have enough votes to be assured of at least one seat based on primary support (and that because of the special issues in NSW).  With a quota of 2.7%, the Nick Xenephon Group (1.93%) would be the only other party with much more than half a quota; of the remaining parties only the Sex Party (1.36%) and Family First (1.11%) polled over 1%.  Enough of the Senate vote for five seats if it went to a single party, went to parties that polled below 1% each.

In my first Senate Reform article I mentioned that if we had a nationwide multi-member non-preferencing system along the lines of Sainte-Lague, using a single electorate, the micro-parties (excluding the special case of the LDP) would win only a single seat between them.  Even a nationwide multi-member non-preferencing system with a known inherent bias towards smaller parties, like the largest remainder method with the dreaded Hare quota, would not get parties up from half a percent of a vote (those with around 1% could still make it, though most likely the system would force amalgamations that would raise the barrier.)

The only basis for arguing that it would be proportionally accurate to have some micro-parties winning Senate seats with fractions of a percent of the vote even on a national basis, is to argue that the supporters of each of these micro-parties would prefer the other micro-parties to win if their own party did not.

It's easy to "refute" the idea that this is so by looking at how those  who vote below the line for micro-parties voted, but it can be counter-argued that those voters are not typical because they are such a small proportion of the total and represent the ones who have deliberately chosen to buck their party's group ticket, often in disagreement with it.

I think a more useful approach is to look at House of Reps preference distributions.  When all we have is a genuine expression of voter preferences, do the voters for micro-parties see their party and the other micros as some kind of Oppressed Brothers' League against the Coles and Woolworthses of the political world?

To examine this question I had a look at preferencing data from every electorate beginning with a letter from A to C and examined every case in which:

* a micro-party candidate (excludes PUP) was excluded
* at least one other micro-party candidate (excludes PUP) was still in the contest for the seat
* the excluded candidate was not the recipient of the donkey-vote, including indirectly

For each such exclusion I found the percentage of preferences that flowed to the Coalition, Labor and the Greens combined:

	Cases	Big-3 %
Rise Up Australia	14	35
Sex Party	2	70.5
Katter	6	44.6
DLP	10	59.8
Family First	9	62.7
One Nation	3	48.5
Country Alliance	3	49.4
Aus Christians	6	32.7
Secular Party	3	64.2
Other	11	59.5

In general micro-party voters are about as likely to distribute their votes to one of the "big 3" as they are to distribute them to the remaining micro-parties, independents or PUP.  This is probably not quite the case on the initial ballot papers, because some of the micro-parties that may be preferenced by the elector will be excluded before their micro-party is cut out.  A partial reason for the Rise Up Australia preferences flowing as weakly as 35% to the "big 3" is that RUA are usually first to be eliminated.  However, left micros tend to flow strongly to the Greens and Labor no matter when they are cut out, and even voters for the most right-wing micro-parties will still send some portion of their preferences to big parties.

Some of these figures have special explanations (for instance often by the time Family First are excluded, the only other micro left is the Sex Party) but on the whole, micro-party voters don't have a strong mindset of putting the "big 3" last.

For this reason, the idea that the Senate system is working best if the "big 3" are represented in as close as possible to proportion to their primary votes is not correct.  A lot of lukewarm support for the bigger parties is hidden in the primary votes of the left and right micros and should be expected in a fair system to flow to the bigger parties as those micros are excluded.

I am not even sure that the bigger parties will be all that much advantaged under OPV compared to how they would fare under a system of compulsory voter-directed full preferencing.  The idea that OPV is an advantage entails that the preferences of the micro-parties would normally assist other micro-parties to catch up to the bigger players.  However, with voters for left micros and right micros frequently regarding at least one major player as better than the other side's micros, it doesn't seem that this is the case.

What The Senate Is And Is Not

I'll put this out there: the Senate is just not a proper Proportional Representation system, and cannot be made into one.  The main reason it automatically fails PR criteria is the extreme malapportionment between states that sees a Tasmanian voter with fourteen times the influence on the composition of the Senate to one living in New South Wales.  Even with Tasmania grossly failing to take advantage of that by taking a leaf out of the book of the Bloc Québécois, this can still have some major impacts.  Labor and the Greens held the majority in the Senate in the last Parliament solely because this malapportionment resulted in them winning 21 seats in both 2007 and 2010 off the back of 4-2 left-right splits in Tasmania.  Without this, they would have had to negotiate with Nick Xenophon to pass legislation.  

Indeed, that the Australian Motoring Enthusiasts Party has just won a seat with 0.5% of the national vote is something of an affront to PR, but a bigger one occurred in 1998 when the Brian Harradine Group had its late sole Senator re-elected with 0.22% of the national primary vote.  The difference is that Harradine would probably have been elected by the choice of Tasmanian voters in a fair system while Ricky Muir would not have won under any system but the current one.

The Senate is therefore first and foremost designed as a states' house, though whether it really serves that purpose all that much is debatable, and whether it is right to have such gross variation in voter power is not even worth thinking about.  Rather, what we should expect from the Senate is that the system makes it fairly difficult for either side of politics to win a majority, that the system does not unduly advantage the right over the left or vice versa, and that the system gives a genuine chance to anyone who actually represents a substantial minority in their state.  We should expect that if the Senate electoral system does run a risk of electing unrepresentative candidates who may be extremists or lunatics, that the extremists and lunatics at least belong to parties big enough to care if we vote to punish them for doing so next time around.  Any PR system that elects members from electorates with as few as six positions up for grabs at a time should be expected to display some proportional bias towards large players in its outcomes.  If it doesn't display that bias, that's probably because it isn't working.  The farce of 2013-4 is not the only blot on the current system's copybook, as those who remember the efforts of the original preference-snowballer, Steven Fielding, would be quite likely to agree.

Anyone who wants to see true proportional representation and fairness in the Senate shouldn't be concerned about the major parties increasing their share of the spoils at the expense of weighted-lottery-style winners of the scraps available to dealing micro-parties.  Rather, that fight has to start with the abolition of state boundaries in the Senate electoral system and the elimination of state-based malapportionment.  This would require all of WA, SA and Tasmania to agree by referendum to accept a reduction in their proportional share of Senate seats and it's not looking likely any time soon. (NB This originally said at least one, but I've been alerted to S 128 of the Constitution.)

Where To For Micro-Parties?

Reading through the above after writing it I realise it doesn't offer too much by way of electoral hope to those who (including me!) often prefer the right kind of micro-parties to the big 3 or, for that matter, PUP.  But I'm much more concerned about stopping obnoxious micros winning, than about encouraging the micros I like to do deals with the micros I don't, in pursuit of a nominal improvement in their chances.

Ultimately, those who support parties that command below 5% of the vote in a given state just need to accept that this does not deserve to be a winning score.  The new system will remove the point of running for most of the existing micros and is likely to lead to a rapid thinning of the current grove of political saplings.  This, however, leaves gaps of several percent on both left and right, consisting of voters who are very accustomed to not automatically voting Labor, Liberal or Green.  It may well be that the sorting-out of the micro-market in the first few elections under the new system will provide a great opportunity for micro-parties that have their act together and that have a purpose beyond just doing preference deals, to increase their vote shares sharply.