Saturday, August 31, 2019

Why I Don't Prefer Abolishing Above The Line Voting

This week I sent a submission (not yet posted) to the Victorian Electoral Matters committee, concerning the 2018 Victorian election.  Primarily, my submission called for the abolition of Group Ticket Voting in the Victorian Legislative Council and its replacement with a Senate-style system or similar.  This follows a farcical, gamed-to-death 2018 election in which ten micro-party MLCs were elected from primary vote shares eight of them would not have won from under any other system, including two from less than 1% of the vote.

In the event that Victoria won't abolish Group Ticket Voting completely, I suggested the state at least clip its wings a little by:

* allowing an above-the-line preferencing option, so that votes that were just-1 above the line would still be distributed by Group Ticket, but voters could choose to distribute their own party preferences as in the Senate.

* banning preference trading and a range of related consultant activities

* bulk-excluding all parties that fail to clear a primary vote threshold of 4% at the start of the count

I'm not sure how much impact the first would have or that the second could be made loophole-free, and I actually strongly dislike thresholds.  But I wanted to put options on the table that might at least mean Victoria would never again see candidates elected off less than 1% of the primary vote - no matter how good or bad they might be as MPs.

As ever, the Proportional Representation Society of Australia (Vic-Tas) and Malcolm Mackerras (submission 12 with numerous appendices - read at own risk) have called for abolishing above-the-line boxes and having a system more like the Hare-Clark with Robson Rotation methods practised in Tasmanian state and ACT territory elections.  Mackerras has devoted a page of his increasingly strange ramblings to attacking and attempting to mind-read my position.  I'll dispose of his attack at the bottom.

I should stress that getting rid of above the line voting entirely, whatever problems it raised and whatever it cost, would still be better than keeping Group Ticket Voting.  Even scrapping preferences entirely would be better than Group Ticket Voting!  But as the question comes up fairly often, I am going to explain here why I do not, at least not at this time, have scrapping above-the-line boxes in the Senate or the upper houses of NSW, Victoria, SA or WA as my preferred solution.  For all these cases I prefer either the Senate system or the NSW variant of it.

In this article I will refer to systems without ATL boxes as "Hare-Clark", simply because Hare-Clark is the model used in Tasmania and the ACT.  However, different models could be used, especially as concerns Hare-Clark's last-bundle approach to surplus transfers.  The differences between those models and Hare-Clark generally don't affect my argument.

The arguments that follow are in no particular order and you might find some more persuasive then others, or spot easy system patches to avoid some of them.  But I believe in totality they provide good reason to be cautious about getting rid of ATL boxes, at least for now.

Hare-Clark works best as a standalone

Tasmanian and ACT elections are standalone elections where the house of government (in Tasmania's case) and the lone house (in the ACT's) go to the people at a time without other elections.  Candidates can compete for attention with other candidates from the same party without also competing against the candidate-based attention for candidates running for single seats.  In all the other jurisdictions listed, Lower House and Upper House elections are held together.  Most of the personality politics will be devoted to the Lower House where candidate popularity can be a big advantage in a marginal seat.  Within-party contests in the Upper House may interest some voters but are not likely to get the same amount of attention as they do when they are the primary election being held.  That's especially so when there will be many candidates competing for candidate-based attention, not just the odd one here and there as in the Lisa Singh and Jim Molan cases.

In Tasmania, the focus on candidates means that major parties have to pick diverse candidates.  But with less of a candidate focus, candidate outcomes under Robson Rotation might be dependent on a small number of voters who care about candidate factors, with candidates within a party closely packed.  Far from encouraging diversity, this could have the reverse effect - parties might be reluctant to pick interesting candidates who might defeat more conventional factional types.


Many voters have an attitude to voting in which they will go to the ballot box, do the minimum necessary to comply with the instructions, and leave.  Most Senate ATL votes stop at 6 and most Senate BTL votes stop at 12.  Many voters either don't know or don't care that their vote would be more powerful in preventing outcomes they did not want if they numbered a lot more boxes.  Even in Tasmania this is seen, with around a third of Greens voters voting just for Greens candidates then exhausting their votes, and close to half the major party voters doing so.

This actually means votes for individual candidates that comply with Hare-Clark instructions (5 candidates) are often less informative in preference terms than above-the-line votes in the Senate, which cover multiple parties rather than just one.  It makes it harder for a party that is closely trailing another party to catch up on preferences even if, with a little prodding, voters for excluded parties would like it to do so.

Low-information voters

Some voters just aren't very good at understanding our electoral system.  They may come from countries where voting is nothing more than an X in the box and may not have the language skills to understand the instructions, or they may just be politically illiterate.  The same seats in Western Sydney persistently crop up near the top of all seats for informal voting - but also in the Senate for just voting 1.  (The NSW system doesn't help there.)

With above the line voting, a 1 for a party flows through all its candidates and helps the party's key candidate to fight for a seat.  But imagine this voter going to vote 1 for a party and seeing only a list of candidates, and voting 1 for just one candidate.  Even if that vote is allowed at all (it would be informal in Tasmania but saved in the ACT), maybe that candidate gets excluded and the vote just exhausts.

So just solve it with How To Vote cards? Well, How To Vote cards are normally banned at polling booths in Hare-Clark to encourage candidate contests, but it won't work anyway.  This sort of voter already isn't following the how-to-vote card in the current system.  How are they going to do so when they have to match an instruction in a language that they don't understand with a Robson-rotated Hare-Clark ballot where the names don't even appear in matching order?

I am forever bemused by how many PR purists assume everyone votes as easily, as willingly and with as much interest in politics as they do.  Whatever the intentions behind such thinking, in its actual impact on electoral design it is both elitist and racist.  Some of the people who run this line actually are out and out voting elitists.  I now and then encounter people on Twitter who really do think that anyone unwilling to research and number 50+ candidates doesn't deserve a system in which they can simply and safely direct their own party preferences.


It's basically impossible to get electoral reform passed anywhere without at least one of the major parties supporting it (in plenty of cases both) so as a result of this, anything that raises a risk factor for major parties is going to make it harder for them to support a reform.  I see no point supporting reforms that are needlessly dead in the water.

A big concern for parties in Hare-Clark is managing leakage out of their ticket to other parties.  Under the Senate system leakage is low - most voters vote above the line, and most below the line votes stick with the ticket.  Leakage occurs when voters vote across party lines and only pick some of the candidates from their most preferred party.  When it's under 10% and all parties are leaking (as in Tasmania), it's not such a big deal.  But if all voters are voting for candidates they know little about there may be more spray - especially based on geographic factors. Voters at one end of a state or large electorate may not know anything about some of a party's candidates, but may know about candidates in their area from multiple parties.  How this could affect elections, and whether the impacts on different parties would cancel out or not, isn't easy to predict.

Design And Cost Issues

In the Senate system, most voters vote 1-6 and stop.  Those who keep going above the line are only numbering as many boxes as there are parties.  Very few voters vote below the line and number every box, even in Tasmania.  But under Hare-Clark, the only way to effectively distribute preferences between all parties is to number pretty much every box.  Because Hare-Clark reduces the efficiency with which a voter can express their preference for a party, it takes longer and costs more to count.

But Hare-Clark is also more complex to print (because you don't have a single ballot design, but thousands of rotated variants).  And it's more complex to program data entry for (because not every ballot will look the same, so there have to be systems to match a ballot paper number to a ballot layout and check that the matching has been performed correctly.)  All of this means cost increases needlessly (and good electoral systems don't come cheap to begin with).

The system also means increased costs for candidates and/or parties.  Even in Tasmania there have been cases of candidates spending over $100,000 to successfully market themselves to a smallish seat with 70,000 or so voters.  Imagine trying to campaign effectively as one of, say, five Labor candidates fighting each other for two or maybe three Senate spots across the whole of New South Wales.

Hare-Clark Has A PR Weak Spot

This is a point that gets very mathsy - wonk factor 5/5, beware.

It really does surprise me that "proportional representation" societies and PR purists hold up Hare-Clark style systems as perfect PR when they are actually impure examples of it that can in some cases unfairly advantage major parties over significant minor parties.  Consider the following: the quota in a seat with 59,999 voters and 5 vacancies is 10,000 votes.  In the race for the final two seats, 17,000 votes will end up with Party A, 8,000 with Party B, and the rest will either end up with other parties' winning candidates or else exhaust.  Which result is fairest and most proportional? Two seats to Party A or one seat apiece?

Here are two ways to look at this:

* Closeness to quota - Party B (0.8 quotas) is closer to the target for one seat than Party A (1.7 quotas) is to two.  Party B is also further from zero quotas than Party A is from one.

* Percentage of seats won - Party B finished up with 13.33% of the vote after preferences and Party A with 28.33%.  If each party gets one fifth of the seats (20%) that's a difference between vote share and seat share of 6.67% for Party B and 8.33% for Party A.  If Party A gets two and Party B gets none then the difference for both parties is larger.  (13.33% for A and 11.67% for Party B.) Naturally the sum-of-squares difference, used in many proportionality indices, will be greater too.

Under the Senate system Party A's first candidate will be elected with quota, leaving 7,000-ish votes for the second Party A candidate, who Party B will defeat.  But in Hare-Clark thanks to the Ginninderra Effect (warning: that is one of the wonkiest links on this site) it's possible for Party A to get two candidates up at Party B's expense.  It's a rare event, but not super-rare, having happened for instance in Ginninderra (ACT) in 2012 and 2016 and Braddon (Tas) in 2014.  In the latter case, the Liberal Party with 61.8% of the vote after preferences won 4/5 seats (80% of seats!) while Labor with 29.5% won only one.

If anything, applying Hare-Clark to lower profile state Upper House contests will only increase the problem.  With little knowledge of candidates, a lot of voters will just "donkey" whoever their party has listed.  That is likely to make for a more even spread of primary votes, which makes it easier to get multiple candidates ahead of a minor party candidate and dubiously beat them in this way.

This issue is one reason why we still have Hare-Clark in Tasmania.  Any support for more regimented voting orders has been snuffed out by arguments like this brilliant one by the late George Howatt that points out that big parties can use Hare-Clark to their benefit by using the spreading of the vote to eliminate minor party candidates.  There's just one problem.  The paper argues that the minor party candidate losing is "more equitable" and proportional because the average vote per candidate for the other parties is higher.  But that's got nothing to do with proportionality, which is about the relationship of seat share to vote share.  (Usually expressed in terms of primary vote share, but we do have a preferential system.)

(And yes, this whole Ginninderra quirk is only because Hare-Clark doesn't use a progressively reducing quota.  But in a case like this, the effect of using a progressively reducing quota is to assign the preferences of the voters who exhausted their vote more heavily to the party with two candidates.  Generally I like the idea, but in this case that seems rather unfair.)

Other arguments may be added later.  That ends the serious part of this article.


Wrong About Psephology,
Wrong About Psychology,
Wrong For America ...

Now we come to the popcorn section where I dispose of the latest erratic Mackerras attack (enter at own risk of being scarred for life).  You only have to get a few paragraphs in to that PDF, where he starts judging the supposed beauty of electoral systems as if they were women, to see that this is going to be weird stuff even by young Malcolm's recent standards.  Normally I love taking a bad opposing analogy and rewriting it to how it really works, but in this case, no, I'm just not going there ... no really ...

I haven't even read that whole attachment yet, because my eyes glazed over by the end of page 1, but the bit about me is on page 9.  The first error is in the first line, where he calls me an academic.  I'm a consultant.  I haven't worked for pay in academia beyond a number of shortish tech-officer type contracts (specimen sorting and so on) since finishing my PhD sixteen years ago.  I used to have an honorary academic post but (i) it had nothing to do with elections and (ii) I held it purely for convenience.  I let it lapse when it ceased to be convenient.

The second error is in the second and third lines where he claims Antony Green has seemed to accept the decision of parliament not to fix Group Tickets.  Um, yeah?  Let's see what Antony actually said about that:


Mackerras' position is that above the line voting is a "contrivance" that gets in the way of people voting directly for candidates, which he erroneously connects to the "directly chosen by the people" bit in the Constitution (in fact an antidote to indirect election by state legislatures and so on).  The PRSA (Vic-Tas) submission makes a vaguely similar but clearer argument - they say that since above-the-line voting was introduced to cure high informal voting rates in the days when Senate voters had to number every box, and high informal voting rates can also be cured simply by allowing optional preferencing, therefore ATL voting is an unnecessary solution.  However, their argument makes the mistake of confusing the rationale for introducing ATL voting in a different time with its merits today - in an age with a larger electorate, more parties and an increasingly multicultural society.

Mackerras quotes from a comment where I replied to someone asking why we don't just get rid of above the line boxes and gave mainly the "leakage" argument above.  However in my 2014 JSCEM submission I'd given most of the other reasons too.  His feeble gotcha attempt in reply to the "leakage" argument is:

"The error in that lies in the simple fact that "how to vote" material would still be allowed under my reforms where it isn't allowed in Tasmania".

It seems that although Mackerras had previously compared Tasmanian Hare-Clark to a "gorgeous" woman he now doesn't like the way she dresses, but in any case Mackerras should know that voters largely ignore Upper House how to votes anywayThat's even under the current Senate system which doesn't have rotating ballots.   And also, when I said that in Tasmania "people know who the candidates are", I was talking about a higher level of familiarity than the mere knowledge of their names and nothing else about them.  I was talking about the sort that acts as a bulwark against "nah, never heard of that name, I'll preference somebody else".

The rest is mostly empty posturing so I'll move to the hack psychoanalysis at the end.  Mackerras ludicrously claims that all my reasons for not supporting scrapping ATV boxes are a smokescreen that conceals my real motive - to stop a shift in power from party machines towards those who vote for a party!

Why on earth would I have such a motive?  This is me:

* never joined a party
* has not closely identified with any party since age 20
* frequently votes independent
* frequently attacks the machines of parties large and small for various reasons
* called from the start for below-the-line Senate voting to be moved to optional preferencing to give voters some chance to overturn dud preselections
* likes Hare-Clark for Tasmanian state elections and defends it against alternative proposals such as MMP or single-member seats

If I thought exporting Tasmania's system to the nation would be even remotely likely to work I would love to support it!  I just don't.  Unlike those who want to impose the same simple solutions everywhere in an unseemly haste to look principled, I just want to find whatever solution is workable in a given area that will get rid of rubbish systems where people can win seats off less than 1% of the vote.


  1. Kevin congratulations for your sensible defence of above-the-line PR voting and your repudiation of voting elitists who consider BTL voting is somehow superior. We should respect the right of voters to be party-focused rather than candidate-focused, and their subsequent right to make that choice with convenience and not be bamboozled by such artificial devices as Robson rotation.

  2. The "Ginninderra Effect" is often seen in both Northern Ireland and the Republic of Ireland where the Hare-Clark (or STV) experience is very different.

    Candidates aren't grouped by party on the ballot paper and parties generally only nominate as many candidates as they would reasonably hope to get elected on a good day (and parties that overnominate are often condemned for playing with fire). Candidates will often run high localised campaigns, carving up the constituency with running mates to keep out of each other's way and openly urging specific orders of preferences designed to keep them all in the count for as long as possible. Leakage rates are much higher in both jurisdictions than in Tasmania/ACT and there's also a strong phenomenon of voters transferring between the candidates local to their area rather than along party lines.

    Not every party is always perfect at nominating and/or keeping campaign discipline - some of the most interesting contests occur when candidates realise their only path to election is to defeat their own running mate, whilst some parties have really struggled to accept they're in a weaker position than they'd like and so put up too many candidates then lose seats on leakages. However many parties work the system well and often "balance" their vote such that it's evenly spread across their candidates, keeping them in the count long enough to receive transfers and thus win a great number of seats than the first preference quota % implies.

    (Casual vacancies are filled either by a constituency wide by-election or by co-option of either the nominee of the outgoing member's party or someone from a pre-submitted list of replacements, the latter mechanism being relatively Independent friendly.)

    It has long surprised me that Australian parties have never tried to exploit this effect - are/were the multiple ATL lists locked against this? I suppose some of the separate Liberal and National tickets may have delivered an extra Coalition Senator.

  3. Yes there was at least one case where separate Liberal and National tickets delivered an extra Senator, in Queensland 2004.

    In Group Ticket Voting systems (now only surviving in WA and Vic upper houses and Vic councils) the group submitting a ticket is required to keep its own candidates in the order listed - they're not allowed to use the multiple-ticket option to spread their own vote. But they can juggle the order of candidates on other tickets (both within tickets and between tickets) as much as they want to. Some preference harvester groups would exploit this by chopping back and forth between tickets from different parties in their preference order to make it even harder for anyone to understand which candidate would actually benefit.

    Thanks for the comment; interesting.

  4. Thank you. I agree with all of this other than the threshold bit. I think thresholds are so dangerous, even though they would be advantageous to me personally as a Green, I'm not sure I would prefer them to GVTs.

    BTW, I have a lot of experience with the system you describe at point 1. It was used at the Melbourne University Student Union elections for more than a decade, and for much of that time I was the deputy returning officer and counted many committees under that system.

    Obviously the voting habits of university students are very different from those of the general public, but my experience there has led me to believe that if implemented it would diminish the power of preference harvesting dramatically. In principle I think it would be much better to adopt the Senate system, but as a compromise I think this would in practice work quite well.