Senate Preferencing Reform: Reply To Electoral Reform Australia

Advance Summary 

(Note: This article has had content added at the bottom, and one wording correction, following further debate.)

1. Electoral Reform Australia, the NSW branch of the Proportional Representation Society of Australia, has recently sharply criticised various psephologists and lawyers for their input into Australian Senate reform.

2. Some of these criticisms are invalid in that they suggest that psephologists did not provide reasons for proposals when in fact sound justifications were - in some cases - presented.

3. The critique proposes a version of full optional preferential voting (without above the line boxes) and a method of dealing with exhaust that is used, for instance, in NSW, the ACT and Ireland.

4. However the jurisdictions in which that method of dealing with exhausting votes is used differ from the group's Senate reform proposals in various ways, including (i) in NSW, having a very low BTL voting rate (ii) in the ACT, instructing voters to number a certain number of squares (iii) in the ACT and Ireland, having a long history of use of Hare-Clark in that system, as well as small enrolment sizes per electorate.

5. Criticising alternative reform proposals as "just plain wrong" when they are defensible is an unhelpful distraction from the consensus among serious electoral observers that exhaustive group ticket preferencing must go, and that any of a wide range of alternatives (including ERA's, despite its risks) would be better than it.

Warning: the rest of this article is long, and probably about Wonk Factor 3 4 out of 5. 

Update 3/9: ERA have now responded to this article.  Their response and my comments appear at the bottom of this piece.  And a very short comment was added on 4/9.
======================================================================



Not much has been heard on the Senate reform front for a few months.  The Joint Standing Committee on Electoral Matters (JSCEM) tabled its Interim Report in May this year, but the final report has not yet appeared, and no new legislation has been seen.  These things take time - just hopefully not too much of it, because another election under the current system would not be acceptable, especially not if it were a double dissolution.  

However, there has still been debate going on about the issue.  One such example is a current piece by Stephen Lesslie, President of Electoral Reform Australia, which turns out rather confusingly to be the NSW branch of the Proportional Representation Society of Australia.  The PRSA is, as its name suggests, an advocative society for PR.  The ERA page states support for national Hare-Clark with at least nine members per electorate, fully optional preferential voting, high deposits, and the Meek system for preference distribution.  

I was invited to join the Vic-Tas branch a while ago, but declined.  Firstly because I'm the sort of person who tends to sooner or later get asked to do things in voluntary societies (after all, what is the point of joining them otherwise), and I'm toxically overburdened with that sort of stuff as it is.  Secondly because while I study proportional representation, I like to keep my options open and my position independent as to advocating it (and in what form).  In Tasmania under Hare-Clark I've seen how the instability of a three-cornered lower house parliament, and the strategic vote-shifting swinging voters undertake to avoid it, mean that proportional seat allocation delivers neither truly proportional representation of opinions nor anything like proportional power.  That doesn't mean I support going to a single-seat system either, but it at least means I think we should keep alternative options open. Someday, one of them might be better.

The current "Largest Remainder", being the ERA (as apparently distinct from PRSA) newsletter carries the Lesslie piece I referred to above, which espouses fully optional preferential voting and attacks the interim JSCEM report.  It argues that federal parliament is going to get electoral reform wrong again, because "most of the advice given by psephologists and lawyers is just plain wrong".

The Lesslie piece, which now and then appears to speak on behalf of ERA, specifically attempts to debunk the submissions by Malcolm Mackerras, Antony Green, George Williams and yours truly.  I am not going to defend the Mackerras submission or the Williams threshhold proposal, and Antony is well capable of fending for himself if he so chooses.

The main portion of my comments with which Lesslie takes issue is my recommendation for OPV above the line with semi-optional preferencing below.  

Number of boxes to fill

In my submission I argued that four for a half-Senate election and eight for a double dissolution was a reasonable number of squares for voters to be required to number if voting below the line.  (Antony Green argues likewise in a recent reply to a reader comment). The ERA piece criticises these numbers as mathematically arbitrary, but they are not.  The reason for selecting such numbers is to strike a balance between (i) attempting to keep both the number of endorsed candidates and the informality rate down and (ii) maintaining a very high chance that a voter for a party below-the-line will preference all a party's electable candidates rather than, in ignorance, stopping too early and costing their party's last competitive candidate a vote.  

This justification was spelled out in my submission:

"On the matter of a specified minimum number of boxes, it is very unlikely that any single party will win more than four seats in a state at a half-Senate election. However, parties are likely to run as many candidates as a voter needs to vote for for a valid vote. Extra candidates slow the processes of counting. For this reason (and also to reduce informal voting) a requirement to vote for at least four candidates might be better than requiring six. "

In suggesting that such proposals as all just "random selections" of numbers the response suggests either that the author didn't read this part of my submission, or else is more interested in putting all the submissions he disagrees with in the same box. In fairness, there is actually supporting evidence for the first possibility, since all the material quoted as supposedly from it is actually (sigh) from the text of an article on this site, JSCEM Comes To Hobart , and not in fact from my submission at all. But even then, my submission still contained the following:

"But I think a maximum requirement to vote for four in the Senate is sufficient, since the Senate does not have intra-party contests, and since there is no risk of any party winning five out of six seats in a Senate contest."

Fully optional preferencing, exhaust, and options to deal with it

The ERA response goes on to quote and comment on this part of my article:

As to why I don't support fully optional preferential voting below the line, I believe that where major parties run especially strong candidates who have a cult-like popular appeal, they would be more likely to attract voters who just voted 1 for them and then stopped.  This would disadvantage the party since those votes would not flow to other candidates in the party, and this would create a perverse disincentive to the party fielding such candidates.  The situation in the Tasmanian House of Assembly with a required vote of 1-5 avoids this problem.

The response starts "Surely it is the choice of the party whether or not to run 'strong candidates who have a cult-like popular appeal' [..]"  Indeed so, and my point is that such a choice, if made, should not be penalised through greater exhaust risks.  Senate candidate choices are prone enough to be party hacks as it is.

The response then proposes a "simple" solution to the just-vote-one problem that I've seen discussed elsewhere too:

"Both the NSW Legislative Council and the ACT Legislative Assembly electoral systems have done so, simply by allowing the surplus to be carried by those votes that do have preferences – fewer votes transferring but each with a slightly higher transfer value." 

This deserves detailed discussion of those two systems (and other similar approaches) and the differences with ERA's proposal.  

The NSW system and how it is different from what ERA proposes:

The NSW Legislative Council system, like the proposed new Senate system, allows above the line voting for a single party (the vote runs through all that party's 15+ candidates and then exhausts), above the line voting for multiple parties, and below the line voting. The circumstance in which a 1 vote for a candidate for a party might then not be valid for a party's other candidates is a very rare one: the voter has voted below the line (about 2% do so), the voter has filled at least 15 squares, and the voter has repeated or omitted a number early in the sequence. In this case, making up the surplus from the votes that do have valid preferences expressed results in a trivially higher transfer value for those votes, thus preventing the party from suffering.

The number of votes involved is thus so trivial you'd hardly bother with the provision at all, except that some parties have much higher BTL rates (or higher voter error rates) than others and could be said to be disadvantaged without it.  Crucially this system does not apply to the preferences of excluded candidates - if an excluded candidate's preferences have run out of gas prematurely, the vote exhausts.  Indeed this is what happens to most of the votes when the last candidate from a party is excluded, giving the NSW system a first-past-the-post-like character for those parties short of a quota.  (That in part results from the large number of squares required for a valid below-the-line vote.)

Now consider this as applied to a Senate election with no above the line voting, and with an electorate used to just voting one and fully permitted to do so (this is what ERA are proposing).  Unlike in the NSW system where the number of votes that are effectively just 1 for a single candidate is trivial, in the proposed Senate system it could be very substantial.  The same method could be used for surpluses, but there are cases (say, a hugely popular candidate for a major party in a double dissolution) in which using only those votes with a continuing preference as a surplus might increase the continuing value of each of those votes to above one vote each. 

That's at least slightly counter-intuitive should it happen, but it gets worse when we consider the question of excluded candidates.  For an excluded candidate from a major party, there is no option of diluting the non-continuing votes by effectively placing them at full value in the candidate's retained quota.  Either those votes are lost as exhaust to the disadvantage of the party (the more so if the excluded candidate is charismatic or distinctive within their party, hence more likely to attract votes that do not flow as preferences unless forced to do so by the threat of informality), or else the votes with continuing preferences are increased in value to above one.

Right here is the issue that caused me to consider supporting fully optional preferential voting for candidates and pretty much dismiss it out of hand when framing my preferred option for Australia: if the votes from exhausted candidates can increase in value to above one, then this creates a delicious opportunity for voters to game the system by deliberately running their vote through as many candidates who will be excluded as possible.  At every such exclusion the value of their vote increases, so no wonder this system isn't used.  Alternatively, if exhausted ballots leave the system at full value, then parties may be harmed by voters who think they've done their bit for the party by voting for one of its candidates and stopping.

Under the current Senate system, the issue of exhaust from minor party candidates minor candidates for a given party who are excluded would not be a serious one, because above-the-line voting with a single party order means that the minor candidates for a party poll hardly any votes.  But ERA are proposing abolishing a single party order and introducing Robson Rotation, meaning that the votes for all candidates from a given party are likely to be substantial unless minor candidates run completely dead.  This is not an issue that can be avoided with comparisons to how the issue of within-party exhaust is dealt with in systems with completely different properties.  [Correction added to this paragraph using strikethrough and italics, 3/9/14].

Stephen Lesslie argued (in Largest Remainder 20) that the increase in exhausted votes under an optional preferential system will be outweighed by the decrease in informality.  I am very doubtful of this since accidental informal voting under the current system is at very low rates, and would also be at very low rates under the JSCEM proposal.  It is true that fully optional preferential voting with no above the line option would more substantially reduce informal rates when compared to semi-optional preferencing with no above the line option, but the latter isn't being taken that seriously as a national option anyway.  

The ACT and Tasmanian systems:

Tasmania has Hare-Clark with a compulsory requirement to preference at least five boxes once and once only; any vote that fails to get the first five boxes right becomes informal.  This means that a vote that is 1 for a candidate of a party can only be formal and leave that party through voter choice to leak the preference, or where there is no continuing candidate.

The ACT system is very similar but includes saving provisions. Voters are directed (see sample ballot) to number from 1 to the number of vacancies and reminded at the bottom of the paper "Remember, number at least [n] boxes from 1 to [n] in the order of your choice".  But in practice, if a voter numbers only one box or makes an error, their vote is still counted.  Furthermore, the surplus-exhaust prevention method referred to is used.  So, for instance, while 0.66% of Zed Seselja's primary vote papers exhausted in Brindabella in 2012, there was no leakage to exhaust in vote terms.  But when 7/314 (2.3%) of the primary vote papers from ungrouped candidate Mark Gibbins exhausted,  these actually departed from the count at full value.  As Gibbins was the first candidate excluded, these votes could not possibly have had sufficient squares numbered; voters must have either deliberately stopped (possibly knowing that the ballot paper's instructions were in fact advisory only) or else made errors.

From what I can determine the number of votes lost to exhaust from a candidate with a continuing ticketmate in the ACT system has always been low, meaning that it seems to be a non-issue.  But this is the case because preferencing in the ACT is quasi-compulsory: voters are told to number a certain number of squares, but the requirement is not actually enforced.  Also the ACT did not have a long history of a different system for its elections (only two elections with the modified d'Hondt mess before switching to Hare-Clark) and the ACT electorates (like Tasmania's) have a high number of candidates per voter, meaning that the likelihood of a voter having heard of multiple candidates is high.  I'd also suggest that ACT voters are more likely than most to be politically literate.

The ACT partial solution is thus not a viable model for what would happen under a national Senate system with group voting completely abolished, Robson Rotation, and an open requirement to number only one square.  It is likely that such a system would feature substantial unintended leakage to exhaust that would significantly disadvantage parties whose voters had lower voter education levels than others.  The costs of avoiding this problem by using semi-optional preferencing and retaining above the line voting while abolishing group ticket cross-party preferencing are: allowing parties to greatly influence chances within a party, preventing voters deliberately exhausting their ballot after one preference, and a very small increase in the informal rate (depending on the existence and nature of saving provisions).  I think those are acceptable prices to pay, especially since a likely cost of nationwide Hare-Clark with Robson Rotation would be a vast profile advantage in within-party contests to rich and celebrity candidates.

Fully optional preferential above and below the line without RR isn't even worth thinking about for too long (you get party preselections and within-ticket exhaust, the worst of both worlds.)  However, it is worth considering whether the current JSCEM proposal (with semi-optional preferencing below the line) should include unpublicised saving provisions, either for voters who despite the instructions number just one vote below the line, or who try to number the right number but make an error.  I don't see too much danger in such provisions, much as I dislike any situation in which an official process gives instructions but doesn't actually require them to be followed.

Ireland

In an earlier ERA article, the example of Ireland was mentioned as a reason not to be afraid of exhaust in an optional preferencing system.  Ireland openly allows voters to fill just one box if they wish (usually with exhaust rates from the very minor candidates of a few to in cases 10%, and with a similar system for surpluses to NSW and ACT).  However, that's within the context of small electorates (similar in turnout to House of Rep electorates) and a long familiarity with Hare-Clark.  But even if it was a valid model for introducing a new Senate system, it's notable that in the Irish system parties often attempt to minimise within-party voter choice, by running only as many candidates as they think might have a chance of winning.  A voter might, for instance, get a choice between two candidates from their party for what turns out to be one position (which could still be helpful in a Bullock/Pratt type situation) or they might get no such choice at all.

Finally, I have a philosophical issue with the NSW/ACT surplus idea when applied to cases where there would otherwise be large exhaust.  Here's an artificial example to illustrate: suppose that two candidates, C and D, are in a close race for a seat based largely on the roughly equal surpluses of candidates A and B. 90% of candidate A's supporters preference candidate C over D and 10% preference candidate D over C.  50% of candidate B's supporters preference candidate D and 50% of candidate B's supporters have no preference between C and D and exhaust their ballots.  Between the voters for candidates A and B, about 45% prefer candidate C, 30% prefer candidate D and 25% don't care.  But under the NSW/ACT system, the less-preferred candidate D is the winner, and in fact wins by the same amount (55% to 45%) as if they were preferred by all of candidate B's supporters.  The example is constructed more to illustrate the point than as something that's actually likely in that exact form, but it does show why I think that allowing just votes that do have preferences to determine the whole surplus is only safe when it is known that the number of papers without preferences is small.

The Irish-style system might be more appealling down the track with more time to gradually familiarise voters with the idea of voting for multiple candidates, including from the same party, and to acquaint them gradually with the idea that doing more than just voting 1 can be helpful.  The JSCEM-proposed new system will help do that.  But I don't think an Australian electorate so used to just voting 1 in the Senate can be made ready enough for full-scale OPV in time for the next election, and the immediate priority is to change the system now to something voters can be comfortably made familiar with, and that eliminates the problem of candidates cruising into office on dodgy preference deals and/or with minimal votes.  I'd rather a minimal-change model, that we know from NSW experience doesn't do anything disastrous (even if it probably will be drab on the cross-party preference-distribution front) than bringing in changes that are untried on a national scale in a system that is accustomed to something totally different, and that have the potential to create new problems.

Miscellaneous Issues

There are a few other points to round up in reply to the ERA response.

The response at various stages rebukes commentators for failing to cite any "academic or mathematical study" either in support of our proposals or in evidence that proposals will increase voter participation.  It is an argument from authority to argue that any conclusion is weakened by failure to reference anyone else's published work, and it is especially spurious when many of the systems being proposed cannot actually be formally studied given that they have never existed.  Whatever solution is adopted is going to be something of an experiment, with opportunities to refine it in the light of experience.

To criticise someone for failing to argue that their system will increase voter participation is especially pointless if that person doesn't consider "voter participation" to be a reliable measure of system success.  Indeed some psephologists, including me, believe that coercing voters to participate (call it compulsory voting or compulsory booth attendance as you will) is illiberal and should ideally be abandoned, which of course would reduce "participation" substantially.  I'd also say that if you allow a voter to, in error, express fewer preferences than they would had they known the impact of so doing on their preferred party, then that's actually reducing that voter's participation anyway.

Herding Psephologists

One of the more curious complaints in the ERA reply is that it slams psephologists for their lack of consensus on what should be done, apart from abolishing group voting tickets.  It even criticises those who flag multiple options for reform.  Quite aside from the fact that the ERA position is itself contributing to the lack of consensus on solutions it complains about, it is incorrect to rebuke someone for saying that system A is broken but systems B and C are both acceptable alternatives.  Such an opinion both narrows the field and highlights the range of acceptable solutions.  Government doesn't need to be told that there is only one right answer, especially not if that is not in fact true.

The ERA solution is clearly not my preferred solution, and in my view it would create some serious problems that do not exist within the current system.  But it would still be an improvement on the current situation, and just because I prefer something else does not mean I think that ERA's proposed alternative is indefensible.  I just think its flaws are far greater than they are acknowledging, and hence that the alternatives to their system are far less "wrong" than they insist.  Yet the ERA piece argues that alternative solutions to theirs are "just plain wrong", on a basis that ultimately boils down to not everybody else wanting to jump through their hoops.

Electoral system design is difficult and specialised and slightly different, defensible assumptions about goals produce different results.  To expect psephologists to agree fully on a specific alternative to the current system is like expecting to herd cats, and should be.  This is frustrating to those who would like to herd those cats into supporting their preferred model rather than another, and I suggest that this frustration lies at the base of this lash-out from ERA against prominent Australian psephologists.

What is significant then is when a group with such diverse views about the best solution as the Australian psephosphere produces an almost unanimous condemnation of an existing system and declares that it has to be changed to something better, never mind exactly what.  I really don't think getting too fussy about what it is changed to - as if only one of the many alternatives is valid - is at all helpful to recognising this important consensus.  And while I'd be happy to see any level of further investigation of options that led to a good solution, that applies only if the current system will definitely be fixed before the next election, and not if it creates any chance whatsoever of another election being held under the current system.

Encore: Now This Is What "Plain Wrong" Looks Like

While I have ERA on the phone, I'm reminded that if they are going to propose Hare-Clark on a national basis, they should at least be more aware of some of the difficulties it creates and less optimistic about solving them too readily.  The Dec 2011 issue of Largest Remainder included a lengthy (and in critique terms, strong - though I could add quite a bit more) discussion of the many flaws in the current Hare-Clark countback system (which only counts the votes held by the resigning or deceased member at the time of their election in filling a vacancy).  The suggested solution was to recount the whole election as if the resigning or deceased member never contested the election in the first place.

Quite aside from this solution being a cumbersome nuisance in manual counting, there's a very good reason why it hasn't been adopted.  That is that in some circumstances, this solution can result in one of the other candidates who was elected at the original election losing their seat mid-term because someone else has died or retired.  The usual case in which this arises is a close two-candidate battle within a party, at the end of which one candidate is narrowly excluded while the other goes on to be elected on their preferences.  But had some third candidate from another party (who was elected) never stood at the initial election, it's possible that enough of that third candidate's votes would have flowed to one of the competing candidates (or to some other candidate who thereby gained a larger surplus) to change the outcome.  For instance, had David Bartlett's (ALP) Denison seat from the 2010 election been recounted with all votes redistributed as if Bartlett hadn't contested, it's possible that the full recount would have unelected Elise Archer (Lib) and replaced her with fellow Liberal Richard Lowrie.

There is also the possibility of this proposal electing a candidate from a different party to the one actually elected.  For instance, one candidate has a massive surplus, some of which leaks.  If that candidate resigns, in a full recount the leaking surplus votes now leak at full value rather than reduced value.  This could change the outcome of the final seat by party, creating a disincentive for that member to resign even if they ideally should. This sort of thing becomes a more serious problem in cases where a party only has one or two low-profile candidates left from the original count (which in Tasmania happens quite a bit).

It's also philosophically unsound to argue that restaging the election as if the retiring member had never stood is a fair way to resolve a vacancy.  If a retiring member had actually never contested the election, then that would have resulted in differing decisions by other parties and candidates (including, for starters, that their own party would have run one more candidate.)  So why is the reconstruction of the whole election under a hypothetical candidate lineup that would not have happened even had that member never stood a fair way to model how a retiring member ought to be replaced?

Providing a trouble-free Hare-Clark recount system is actually an extremely difficult problem.

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





Update: Response received

Stephen Lesslie has sent a response, which I have uploaded to Google Docs:  ERA Response to Kevin Bonham reply.  It can also be viewed on their website in comments.

These are my responses to points raised in their response, roughly in the order raised but with some recombination:

1. OPV reducing candidate numbers: It is true that full OPV would further reduce the number of candidates parties need to run below my suggestion of a requirement to vote to four (or eight) for a valid BTL type vote.  However, I consider four to be the best tradeoff between increasing exhaust risks as the number of required squares reduces, and increasing count complexity (and informality risk) as the number of required squares increases.

2. Effect of #1 (only) for lead candidate under OPV: A single vote 1 for the lead candidate has "the same value for the party as if the voter had voted 1-4 down the ticket" if all the other candidates for that party are excluded, however this is not usually the case for the two major parties. In that case, even under the proposed system for dealing with surplus votes that would otherwise exhaust, the effect of a 1 with no other numbers is slightly different to the effect of a 1-4 down the ticket. Effectively, a 1 with no other numbers slightly magnifies the value of every other paper received by that party, whereas a 1-4 down the ticket vote does not. A party that attracts significant numbers of votes for its lead candidate that stop at 1 is disadvantaged compared to if all those voters had voted down the party ticket.

3. ACT Experience: The ACT experience does not demonstrate that the fear that voters will stop at 1 is misplaced. My article outlined some (unrebutted) reasons for doubt that the ACT experience will translate nationally, but it is notable that ERA only propose that "The voting instructions [..] will all encourage" the voter to express preferences. In the ACT experience voters are ordered to express full preferences, and votes exhaust prior to that only if the voter fails to follow the instruction (for instance through error, or by being one of the few who knows the instruction is not binding). If the ERA are proposing that the ACT system be copied and that Senate voters be ordered to express full preferences (although in practice preferencing would be totally optional) then they should say so.

4. Hare Clark Informal Rates: Lesslie's figures for the informality rate for Braddon and Lyons in the Tasmanian state elections are both slightly incorrect. He has divided informal ballot papers by formal ballot papers. The correct formula is informal divided by (formal plus informal), so Braddon was actually 5.12%, Lyons 5.45, Bass 4.63, Denison 4.79, Franklin 3.99 (statewide 4.8%). However, most of these were deliberate informals (blank, scribble etc) or votes that would be informal anyway (ticks and crosses, no first preference, multiple first preferences etc). The TEO's ballot paper surveys showed that only 15.7% of informals cast in 2010 (less than 1% of all votes cast) would have been formal under full OPV. I believe this year that figure was very similar.

We can argue about whether even that's too high and what should be done about it if so, but it's irrelevant to this debate, which is between full OPV without group voting boxes, and semi-OPV with group boxes retained (but their preferences exhausting at the end of the group). Because many voters will vote above the line under the latter (and those who vote below the line will be most likely to know what they are doing) the BTL informal rate caused by requiring a set number of boxes to be filled will be very low indeed, and can be reduced further compared to Tasmania if necessary by retaining the existing approach to cases of apparent honest error.

5. "Special Pleading": When accusing people of logical fallacies on the internet, it pays to be sure they are actually guilty. The term "special pleading" refers to making an exception to a sound principle in a specific case on an apparently spurious basis. It does not refer to making an exception in a hypothetical case in which there would be a genuine problem, just because you think that hypothetical case would never occur. Hence the claim of special pleading is an example of my favourite logical fallacy of all, the Fallacy Card Fallacy.

6. Risks of high exhaust rate from most popular candidate: This was the exhaust-based objection to full OPV I originally flagged; in my response above I indicate that the risk of exhaust from excluded candidates disadvantaging their ticket mates is a much bigger problem in my view (see below.) But the reply's attempts to knock over even the exhaust rate issue from popular candidates are not that strong.

For starters, we have no idea how votes might be distributed among Senate candidates within a given party in any given state without a fixed ticket order, so the argument that very popular candidates within a party don't seem to exist is spurious: they may exist; we wouldn't know. We do know from Tasmanian and ACT elections that sometimes a candidate will poll well over half their party's vote in an electorate. Normally it is a party leader (eg Will Hodgman's 70.5% of all Liberal votes in Franklin this year) but there have been cases of less prominent figures (eg second-term Opposition MHA Frank Madill, Liberal, Bass 1992) at least getting half their party's electorate votes. In a double dissolution even with rotation it's thus possible specific candidates will get three quotas or more in their own right. And, despite my best attempts to assume rotation in considering the ERA proposal, they don't think it will happen: "With the fixed order grouping currently in place, and highly likely to remain in place [..]"

In the pre-1984 Senate system the candidate placed at the top of the party list by their party would commonly get nearly all the party's votes, and frequently poll four or five quotas in a double dissolution. We might think we know that no such candidate is going to exhaust a whole quota out of that, but the electorate is very habituated to just voting 1 in Senate elections, and unless they are told they cannot do it, I don't think we can be sure.

7. Double Dissolutions: Some of the ERA's points rest on the premise of no double dissolution. System designers can't assume a double dissolution won't happen, especially not when the hopefully impending system reforms will render the votes lost by major parties to micros polling <2% of the vote irrelevant, by rendering such parties unelectable even in a DD. A plausible DD scenario under a new Senate system would be that an incoming ALP government (not that I think Labor will win the next election) clashes with the 2013 detritus and wants to be able to pass legislation with support from just the Greens and Xenophon Group.

8. Exhaust from lesser candidates: Unfortunately there was an ambiguity in my article above. The term "minor party candidates" was meant to be used in the context of the less prominent candidates from a given, not necessarily minor, party, but could have easily and naturally been taken as referring to any candidates from a minor party. I've corrected this in the article using strikethrough and hopefully my concern is now clearer. That said, I think the rest of the paragraph made it clear enough that I was talking about a party losing votes because of exhaust when its own candidates are excluded, a matter not at all addressed in the ERA response (unless the sidelining of Robson Rotation has something to do with that).

9. Cross-party exhaust: The issue of votes exhausting when a minor party's last candidate is excluded is something I actually didn't mean to canvass in my article, but it's come up because of the ambiguity mentioned above, and it is worth discussing in its own right as it is one of the limitations of keeping party boxes. So on that score, it is quite true that there will be a large rate of exhaust when micro and unsuccessful minor parties are excluded under a NSW-style system, as a consequence of voters just voting 1 for a party above the line. It shouldn't be quite as high as in NSW because the number of preferences required below the line for a formal vote will be smaller and the number of BTL boxes less intimidating, but to start with it may not be much smaller.

I agree that the rate of cross-party preferencing under full OPV has the potential to be much greater. I don't accept either the ACT or Tasmania or even Ireland as models for how high that rate would be, especially to start with (because of the differences discussed). I am more concerned about parties losing votes on within-party exclusions because low-information voters have stuck to just voting 1, than failing to gain votes when whole parties are excluded because they have done a bad job of convincing more savvy voters to use the system.

10. The Shelving Of Robson Rotation: Really I think that ditching Robson Rotation from a recommendation (because it's not likely to happen) removes the very best reason to consider full OPV, despite its risks. The biggest limitation of keeping party boxes is that parties will continue to be able to allocate high Senate positions through internal hackery, but pre-1984 experience suggests that even OPV with a single ticket order is likely to result in much the same thing.

11. Conclusion and Occam's Razor: Occam's razor cuts both ways (it usually does). We are dealing with a system with already multiplied entities, and many voters find the option to just vote 1 for a party of their choice, without having to worry about doing anything else, convenient. From that reason I've suggested that we should make the minimum set of changes necessary to eliminate the farcical outcomes of the last election, while retaining a range of ways in which voters can vote. There is not a clearcut case that we need to go further and remove more aspects of the existing system.

If it is "undemocratic" to require a very limited expression of preferences that need not even go beyond a single party, then it is very "undemocratic" to limit voter choice by using "high electoral deposits" (supported by ERA) to make it harder for many possibly genuine emergent parties to exist (after all, just because you have a good idea does not mean you will get 4% from scratch). I really think the ERA can get that beam out of their own eyes before they go dissing other solutions with such words.

Encore: Hare-Clark Recounts: In reply to my comments about Hare-Clark recounts, Lesslie refers to the possibility that a candidate can be replaced by a candidate from a different party at a recount under the existing system. Indeed so, but only if it is the will of the voters who elected the retiring candidate. This was the case when Norm Sanders' recount elected Bob Brown, because Sanders (a Democrat) and Brown (an independent) were both extremely high profile opponents of the Franklin Dam, who were often active on the issue together. Sanders ran on a ticket with a bunch of obscure Democrats, but needed help from Brown's preferences to be elected in the first place. Not only did votes from Brown that helped elect Sanders in the original election go back to Brown in the recount, but Sanders' own voters didn't stick to the ticket and "leaked" heavily to Brown. There was nothing wrong with all this. What I am talking about is the possibility of a recount by the ERA's proposed method replacing, say, a Labor member who won a very close contest with the Liberal who he defeated for the final seat, and who would most certainly not have been preferred by the voters who elected that Labor member. This is the ultimately fatal flaw in the ERA's suggestion and the fundamental reason why it can't be taken seriously.

I will think about whether it is possible to configure such a system so that it runs validly but never dis-elects anyone (through some form of quarantining so that a given candidate can't be excluded even when running last), but the answer is academic in view of the above. To give a concrete example of the above, in 2006 Kim Booth (Green) won his seat by 136 votes against Labor, off a primary vote of 5653. In a recount of Booth's seat conducted under the ERA proposal, enough of Booth's votes would leak that Labor would win his seat, rather than one of his fellow Greens, even if nearly all of his voters would have preferred one of his ticketmates to get it.

The argument that vote sorting into batches makes it easy to conduct a whole-election recount is incorrect because the flow-on effects of removing a candidate can often change the order of electoral events. Suppose a vote flowed from A to B after C was excluded in the original election, but in the new election A is excluded before C. Now that vote might flow to C first rather than B, and has to be found again to find out if it does, but because of the use of amalgamation of votes of equal transfer value to speed the original count, there is actually no way of knowing where it is.

The current system is indeed flawed but no-one has come up with a better one. At least it generally delivers a candidate of the same party as the retiring member, unless the voters who elected that member wish otherwise.

==========================================================================
And one more thing (Sep 4): I realised that I've written a very large number of words discussing various pros and cons of OPV and semi-OPV on a very specific, point by point basis, without bothering to bring up a very major problem with what ERA propose. That is its massive impracticality. At the last election, 13,413,019 Australians cast a Senate vote. Of these just 471,030 (3.5%) voted below the line, and the other 96.5% just voted 1 for a party, enabling their votes to be counted quickly with limited risk of error. 

The semi-OPV system I propose may well increase the number of ballots needing individual data entry to a few or several times its current level, but most people will just vote 1 unless their party makes a really stupid ticket order choice.  But under full OPV, every single vote must be entered manually and individually into a computer then double-checked by another operator.  Granted, the number of candidates and groups will be smaller, and each individual paper therefore faster, but the increase in staffing costs to handle a 28-fold increase in the number of ballot papers entered will be huge.  That is another of the reasons why full OPV is not my most preferred alternative.