Friday, May 3, 2019

Senate Voting: Another Poor Article By Richard Denniss

It's a shame to have to be distracted from more important things (like counting all the candidate malfunctions this election) in order to deal with an article by Richard Denniss in the Guardian.  I wasn't going to do this, having already done so on Twitter, but unfortunately an email has been seen by me suggesting that a voter took it seriously.  I can scarcely believe this is so, and wonder if I should refer the email to the AFP as a hoax, but the thought of people being Wrong on the Internet about this needs addressing.  Rant warning applies.

To start with, let's declare on Richard's behalf what he failed to admit in his latest article.  Richard Denniss is a former sceptic of Senate voting reform who defended the coercive, corrupted and dangerous Group Ticket Voting system.  In that system, abolished prior to the 2016 election (but still extant in the WA and Victorian upper houses) voters' preferences were being sent they knew not where, and parties could be elected off tiny vote shares based on a weighted lottery of preference deals and effectively random events.  In defence of this old system,  Denniss, of the left-wing Australia Institute, made the ludicrous claim that "the vast majority of Australian voters trust parties to allocate preferences and opt to vote above the line."  Really, Australian voters trust political parties when politicians are among the least trusted occupations?  (OK, the link is only Morgan, but with effect sizes that large, even Morgan can be trusted here!)  We saw exactly how much Australian voters really trusted parties in 2016 when only about 30% of Coalition above-the-line voters, 14% of Labor ATL voters, and 10% of Green ATL voters (and single figures for most minor parties) followed their party's Senate how-to-vote card.  In Tasmania, voters didn't even trust the Labor Party to pick its own candidates properly, and overturned its preselection.

The worst part of Denniss's 2015 article was a part that tried to explain how preferential voting worked via an analogy about ordering dinner.  The analogy would have been a good explanation of how House of Reps voting worked and how Senate voting should work (and now does) but that as a description of how the old Senate system worked was an absolute joke.  I had far too much fun with that one in If Ordering Dinner Was Like Senate Voting.  Now, Denniss is back for second helpings.  He's even used a food analogy again, oh dear.

The main thrust of the new Denniss article is that Senate voters should number as many boxes as possible.  That I generally agree with (see my more detailed piece on how to vote wisely in the Senate).  I particularly agree that it's very important that AEC staff tell all voters to number "at least" 1-6 above the line or 1-12 below, and it's disappointing to have seen two reports to the contrary already, as well as a transcript where the Australian Electoral Commissioner appears to have missed out the magic words.

However in the course of defending these concerns Denniss has made many misleading arguments, and put a suspect survey ahead of actual evidence of how voters voted in 2016.  Mainly, he is yet another left-wing commentator trying to blame the new electoral system for One Nation winning Senate seats, instead of accepting that One Nation won seats fair and square as a result of voter choice that was accurately captured by the system.  Indeed, if One Nation had any unfair advantage in 2016, it was the amount of garbage put about by Labor figures and parts of the "independent" left about how Senate reform was a conspiracy to exclude little parties from the Senate completely.

Denniss starts by diminishing the case for Senate reform as having been just about Ricky Muir, elected to the Senate by preference-harvesting in 2013 with 0.5% of the primary vote.  It's easy to use Muir as an example because he was actually fairly well-regarded as a Senator, though Victorian voters didn't think enough of him in two years to go anywhere near re-electing him.  But Denniss doesn't mention that the 2013 farce also saw the near-election of the completely unknown Wayne Dropulich of the equally obscure Sports Party in WA off 0.2% of the vote, nor that a combination of counting error and inherent risk in the Group Ticket system led to the entire WA contest being voided and re-run at vast expense.   Nor does Denniss mention the extent to which spiralling party group numbers created monster ballot papers, something that is only now starting to abate.

A Dubious Survey

Denniss goes on to report a poll commissioned by the Australia Institute on Senate voting instructions .  Given that Denniss as a major TAI figure was criticising Senate reform in 2013, this is a bit like a Liberal Party commissioned poll "showing" that voters strongly oppose negative gearing changes - it needs to be treated with extra caution because the party pushing it is the party benefiting.   I'm not doubting that voter education about the new system could still be better.  But I don't accept the view of it painted by this TAI survey.

Good reason to cast serious doubt on the care taken by TAI's respondents, and hence to consider the whole survey as dodgy, comes from contradictory results to two questions.  Two questions asked were "You are allowed to number every box above the line" and "If you number beyond 6 your vote is disqualified".  If the first one is true, the second one is false, yet 57% agreed with the first statement and only 37% disagreed with the second.  If the second is true, the first is false, yet 32% agreed with the second and only 25% disagreed with the first.  (Of course, the first one is true and the second is false).  These discrepancies show that at least 27% of respondents (probably many more) were giving self-contradictory answers.  Were they rushing too fast to notice this or was it too difficult to go back and change?  The respondents were also told that "On the Senate election ballot paper, you can vote 'above the line' by numbering at least 6 of the boxes in the order of your choice" yet six questions later only 37% remembered or knew that numbering beyond 6 was OK, having just been more or less told it?

The most intriguing thing in the survey is a result claiming that 47% of voters agreed with the claim "You should give a '6' to the party you dislike more than any other party on the ballot paper."  Respondents in online surveys frequently flick through questions rather quickly and might easily jump at a novel suggestion like this and say yes, whereas had they not been prompted it's unlikely they would even have considered it let alone thought it through.  In discussions around the 2016 Senate system leading into the election, I never saw or heard of this claim anywhere "in the wild".

TAI seem to be keen to blame One Nation's victory of two seats in Queensland 2016 on this finding.  Denniss mentions it after long discussion of Anning-vs-ALP or One Nation-vs-Greens hypotheticals and immediately before claiming:

No voting system is perfect. The first-past-the-post system gave the world George W Bush, our old Senate system elected Muir on a primary vote of 0.5% and our new system gave us Malcolm Roberts and Anning. 

The argument by insinuation is all too obvious.  But rather than assume One Nation won by confused voters who hated them putting them 6th, one could check actual election results to see if it was actually true.  If voters in 2016, in significant numbers, were giving One Nation a 6 to put them last, then what we would see is spikes in One Nation preferencing at number 6 compared to number 5 among voters for parties that hate One Nation - especially the minor left-wing parties.

In fact (and a big plug again for David Barry's superb ATL explorer) this wasn't the case at all.  Of the 37 ATL groups on the Queensland ballot, voters for 28 groups were less likely to vote One Nation 6th than 5th.  The nine more likely to put One Nation 6th included two parties with less than fifty 5th/6th One Nation votes, four parties with less than 1000 such votes and statistically insignificant skews towards 6th (Rise Up, Christian Democrats, Health Australia and Renewable Energy) and only three parties with statistically significant skews towards putting One Nation 6th.

The three parties whose Queensland voters tended to put One Nation 6th more often than 5th to a meaningful degree were Family First (2.41 times more likely), Katters Australian Party (1.45 times) and Democratic Labour (1.18 times).  The first two had One Nation 6th on their how-to-vote cards.

Considering the Greens, whose voters are very anti-One Nation, they had a nearly even split between 5th and 6th place One Nation votes.  But they also had the lowest rate of putting One Nation 6th (just 1.1%) of any party on the ballot.  Labor voters (and LNP voters too) were 20% less likely to put One Nation 6th than 5th.

So there's no evidence any significant number of voters, especially not left voters, had this idea of putting parties 6th until The Australia Institute popped up with it in a survey and fooled 681 voters, who may have then passed this foolishness on to their friends.  There's no indication as to whether the survey even told the respondents the correct answers afterwards, but if it didn't, that's a serious concern.

The Exhaust Issue

Denniss is right to point out to voters that if they don't number all of the boxes they risk their vote exhausting and electing someone they don't like.  But as with many making this point he exaggerates the extent of exhaust in 2016.  Yes, 7.5% of vote-values technically exhausted, but this is a very misleading statistic, as the figure is inflated by the AEC continuing to throw some preferences past the point where any of the outcomes could change.  Only 5.1% of vote-values exhausted while a contest was still live.

This issue is even more acute when it comes to the claim that "half of all ballot papers had some share of their vote “wasted” by not continuing to number to the end of the ballot paper."  Leaving aside that this may have been the deliberate choice of some voters who were well aware of the consequences (this is commoner than some might think), a ballot paper is not in any sense "wasted" if it only hits the exhaust pile after the contest is determined.  The number of formal ballots that lost some part of their value to exhaust while contests were still live was 26.1% in NSW,  15.3% in Victoria, 13.3% in Queensland, 13.4% in SA, 12.4% in WA, 32.0% in Tasmania, 0.0004% in ACT, zero in NT.  And while Tasmania had a high rate of partial exhaust, the average ballot paper that exhausted in Tasmania had already used over 91% of its effective value helping one or more winning candidates to get elected.  I haven't yet calculated the national percentage of votes that exhausted at full value (ie the vote did not help anyone get elected, nor take any part in the final seat contest) but it would have been much less than 5.1%.

When it came to One Nation winning two seats in Queensland, the preferencing decisions of Labor, LNP and Green voters and whether or not they exhausted their votes had almost nothing to do with it.  The LNP had a smallish surplus of around 9000 votes, the Greens a negligible one, and Labor didn't cross quota for their final candidate until after the contest was decided.  Rather, the last seat in Queensland was decided mainly by the preferences of a huge gaggle of little parties between other little parties.  Voters for the little parties collectively preferred One Nation to the other little parties and would have done so no matter how many squares they had to fill out.  Most of the exhaust in Queensland came from preference throws between the micro-parties.

What we see here is really a moveable feast of complaint.  Denniss seems to think the voters whose ballots exhausted would have had a preference between the Greens and One Nation if asked.  But in the dinner article he admitted that  "most voters do not really have strong preferences about the hundreds of candidates that can populate a Senate ballot paper."  He then went on to comment on the "choice" voters had under the old system, but if you're arguing that most voters don't really care about every party, then the obvious choice that goes with that is to stop.  And, in the case of what went down in Queensland in 2016, it's irrelevant.  The Greens were elected easily, leaving voters to choose between One Nation, the Liberal Democrats and Family First.  Try going back to 2016 and doing a poll to see what preferences voters would have between those. 

Dubya, Muir and Roberts/Anning

The passage that I most dislike in Denniss' article is the bit already quoted above:

No voting system is perfect. The first-past-the-post system gave the world George W Bush, our old Senate system elected Muir on a primary vote of 0.5% and our new system gave us Malcolm Roberts and Anning. 

In fact, Bush winning Florida because potential Al Gore supporters in a two-way contest voted for Ralph Nader was a result of a system defect that denied voters the ability to freely express the preference 1 Nader 2 Gore.  Muir winning in Victoria Senate off a primary vote of 0.5% was a result of a system defect that coerced voters into sending their preferences to preference-harvesters by denying them the ability to choose their own preferences in a practical way.  (When they had that choice they used it in gloriously diverse ways, unlike the near-100% fake preference flows of Group Ticket Voting.)  But Denniss has not shown that Malcolm Roberts winning as the second One Nation candidate was the result of any system defect, or of anything that was unfair to voters, and nor has he provided any evidence to credibly link Roberts' victory (and subsequent replacement by Fraser Anning) to vote exhaust issues or misunderstanding of the system in any way.   In fact, in 2016 the new system worked brilliantly at translating voter intention to seat share.

It's fine to encourage left-wing voters to preference more fully in the hope of warding off a potential future situation where they might lose a seat to the xenophobic right through exhaust (though in fact as the NSW Upper House election just showed, the opposite is rather more likely.)  It's just as fine for right-wing sources to encourage right-wing voters to do the same, and I think the right should do more of this instead of whinging about the existence of preferencing.  But I will not have comparisons of the new system to first-past-the-post or Group Ticket Voting, both of which are not even fit to be considered voting systems.

No comments:

Post a Comment