"Reignite Democracy" Cooks Up Some Senate Preference Myths

Yesterday my attention was drawn to a flier that had been circulated by Reignite Democracy Australia, a fringe right anti-mandates group, and also to a course they are offering.  The website is promoting online sessions in "how the electoral system works", how it supposedly favours the major parties and the Greens, how it supposedly works against "freedom loving and independent candidates" and how voters can "turbocharge their vote".  There's a bit of a push around for voters for parties like Liberal Democrats, One Nation and United Australia to preference each other, with supporters of the idea labelling all those involved "freedom parties" or "freedom friendly", even though some of the parties involved hardly have a libertarian bone in their bodies on any issue that isn't COVID-related.

I believe that everyone, whatever their politics, should have access to the facts about the electoral system and how they can make best use of their vote to represent their views.  Unfortunately what we have with this flier and may get with the presentations is some eccentric claims about the Senate specifically by one Peter Newland (apparently a veteran JSCEM correspondent, though not one I have noticed before). The claims sound plausible because of the detailed discussion of electoral systems, and indeed received a fairly friendly response on Twitter yesterday when tweeted, perhaps because they were more interesting than the usual RDA paranoia about voting in pen, stylising 1s and 7s and so on.  However, they still contain serious errors and unsound strategic voting advice.  It appears that RDA may be about to give a platform to junk psephology and recommend it to children.  

For sounder advice on how to make best use of your vote in the Senate here's my 2019 article on the subject .

The exhaust myth - who micro-party voters really preference more

A once-common myth about Senate voting in the new system is the claim that small parties are disadvantaged in comparison to big parties because voters exhaust their vote instead of filling in all the boxes.  While the brochure does not directly make this claim, it is implied by the claim that the current system advantages the Coalition, Labor and the Greens, coupled by a heavy focus on the issue of exhaust.  

The suggestion is that little parties would do better if their voters numbered all the boxes and preferenced other little parties with majors near the bottom, instead of exhausting their votes.  Well, they would do a little better if only those voters who hated the majors did that, but that still wouldn't get them elected at half-Senate elections.  

If voters for little parties all preferenced other little parties that would be one thing, but voters for left-wing micro parties often dislike right-wing micro parties and vice versa.  Ever-fashionable dismay with the majors aside, the voter who thinks the Labor and Liberal parties are genuinely the worst two parties on the ballot is a pretty unusual beast.  (Newland should understand this well since in 2019 he issued voting recommendations with the Coalition well up in the top half.)  Also, even within the main flanks, there is a reluctance to cross-preference on the right: libertarian right, religious right and right-populist minor parties are classified together but the three strands have often had very little in common and their supporters have each been reluctant to preference the other two.  

The fact that Senate preferences scatter without very strong flows from anyone much to anyone much is alone enough to mean that even if no voter ever exhausted their preferences, micro parties on <2% primary would never win in half-Senate elections.  But not only do the preferences of micro-party voters scatter, they also tend to favour the larger parties.  If all voters chose to fill in all their Senate preferences, it is likely that big parties would beat the smaller parties (except One Nation) by more, not by less.

This is not (as the brochure implies) because voters are specifically including big parties in their top six to avoid exhausting their vote.  Most voters have never heard of such strategies and are just voting for the parties they want to see elected, or in a minority of cases for those listed on a how to vote card.  If micro-party voters were preferencing Labor or Liberal in their six just to stop their votes exhausting we would expect to see relatively few second preferences go to those parties and a flood at five or six, but this is not the case: major parties do well on micro-party second preferences too.

I had a look at NSW Senate 2019 to get a figure (for above the line votes) of the share of all non-big-three voters that include each other party in their top six, excluding each non-big-three party's own votes from the vote pool when considering that party.  The leaders were:

Coalition 37.8%
Labor 30.9
One Nation 29.7 
Greens 26.3
Shooters Fishers and Farmers 25.6
United Australia Party 21.9
HEMP 20.6
Animal Justice 18.4
Liberal Democrats 16.4
Small Business Party 16.4
Australian Women's Party 16.1
Affordable Housing Party 15.7
Australian Workers Party 15.2

(Raw data via David Barry's Senate Preference Explorer but with further spreadsheet processing by me.)  This list is already showing that the majors are outperforming all the minor and micro-parties on non-big-three preferences, though the Greens are slightly outperformed in this field by One Nation (and One Nation would be further up the list here if I instead looked at Queensland.)  However, the above includes only the better performed end of the micro-party list, including a number of now-deregistered parties that would have got some preferences by having feel-good sounding names. Another 22 micros were each preferenced by less than 15% of fellow above the line non-big-three voters.  And virtually no fellow non-big-three voters included Voteflux (1.4%), IMOP (1.8%), Citizens Electoral Council (2.1%), Socialist Equality (3.2%) or Socialist Alliance (3.3%) in their top six. Mostly, the bigger a party the more preferences it will get at any exclusion.

The other important point here is that Senate exhaust is low.  The total value of effective exhaust (exhaust before the winners were mathematically certain) at the last two elections was 5.1% in 2016 and 4.8% in 2019.  In the latter, the margins between winners and losers were so large that in many states one could give all of the exhausting vote values to pretty much any non-winning minor party, and that party still wouldn't win.  This is a result of so many of the micro-party votes that are distributed sooner or later reaching the big three or One Nation.  

Concern about exhaust in the post-2016 Senate system is a throwback to an earlier proposal which was going to copy the NSW method of full optional preferential voting (with its much higher exhaust rate).  Hands were thrown up because of concerns that voters would no longer band together to shut down One Nation as had been possible using Group Ticket Voting, ignoring that all-party GTV gang-ups against One Nation had long ago ceased anyway.  In the current system, the extent to which preferences do flow  has ended up helping the same parties who critics claimed would be helped by preferences not flowing: the majors, the Greens and One Nation.

Does including a major party prevent your vote exhausting?

The flier says "Some ignorantly claim that including a major party in a 1-6 Above-The-Line vote prevents your vote exhausting.  In fact, such votes could still be partly or totally wasted."

It doesn't say who these "some" are.  The truth is that while including at least one of the big three in a 1-6 ATL indeed doesn't guarantee a vote won't even partially exhaust, it's a darn sight better than voting 1-6 ATL for a bunch of no-hope parties and then stopping, which absolutely guarantees that the vote will exhaust.  So if someone was going to stop at 6 then it makes complete sense to include one or two heavy hitters in the list.  Of course it makes more sense to number beyond 6 - numbering more boxes can never reduce the power of your vote. 

The other point here is that a vote exhausting (in part or full) does not necessarily mean it is "wasted".  A common reason that a vote that includes a major party might exhaust is that the party might have elected all the candidates it was going to elect, with all others excluded, before the vote could reach it.  If there are a few parties left in the count at that point but the voter really doesn't care between those, they're hardly in the same boat as a voter under First Past The Post who votes for their favourite party but cannot preference their second favourite.  "Wasted vote" is an imported concept from First Past The Post systems and mostly doesn't apply in Australia.  

Shoot That Poison Arrow (Through Your Foot)

The pamphlet refers to Arrow's theorem and says that it shows that our system is incapable of guaranteeing fair vote-counting.  What the theorem actually shows is that no ranked voting system can always satisfy a specified set of fairness criteria.  The author mentions that a candidate who is put last by 49% of the electorate can win.  This is not a valid example as preventing that from happening is not one of the fairness criteria.  It turns out that the author actually doesn't agree with Arrow's theorem criteria anyway and supports a home-cooked counting system which he claims to be perfectly fair, Arrow notwithstanding.  It's claimed to be a variant of Borda count (which is bad enough, and a common magnet for electoral wheel-reinventers) but it's actually much, much worse.  In Newland's system, candidates are ranked based on the proportion of voters who rank them in the top half of candidates.  Aside from this system throwing away information about where in each half the candidates were ranked, for single member systems the folly of this can be shown as follows:

Election 1: There are two candidates, A and B.  90% of voters rank A ahead of B.  10% rank B ahead of A.  A is ranked in the top half by 90% and of course wins.

Election 2: The same as election 1 with the same voters, but this time there are four candidates: A, B and two Nazis who both voted for B in election 1.  All A's voters hate Nazis and they vote 1 A 2 B and 3 and 4 for the Nazis.  All B's remaining voters hate Nazis and they vote 1 B 2 A and 3 and 4 for the Nazis.  The two Nazis hate each other, and each votes for themselves first, then B, then A, then the other Nazi.  B is now ranked in the top half by every voter.  A is ranked in the top half by every voter except for the two Nazis.  Despite being the preferred candidate for 90% of voters, A loses the election because two Nazis who hate each other and got one vote each decided to be candidates and voted for themselves.   (Even if the Nazis didn't hate each other, their presence would convert a 90-10 win for one candidate into a tie.)

That such systems are unfit for adversarial elections because ballot flooding with candidates with no support can alter the outcome is precisely the point Arrow's Theorem is making.  And indeed this is a worse problem than cases in our current system where an otherwise worthy winner loses because they don't have enough primary vote support to get into the final two.  The author seems not to have considered this problem because in another promo for his system he says it can only defeat a candidate preferred by the majority when it is fair that that occurs.  

The author then claims "This flaw is magnified in Senate Voting, where the last Senator to be elected in each state can be elected despite voters' preferences proving that one elected candidate would have won a one-on-one runoff against the alleged winner".  This criticism completely contradicts his opposition to bare majorities winning, because if a party running for the Senate attracts 51% of the vote, all flowing down their ticket, then every candidate for that ticket would win a one-on-one runoff against every other candidate, even if the other 49% of voters put it last.  The author similarly claims in the link above that 3/25 ACT winners in 2020 were beaten by excluded candidates on pairwise comparison, and I cannot imagine how anyone could go to the effort to calculate that without considering that they might be missing the point of a proportional counting system.  

Does not voting 1 for majors help them?

The author argues that the best way a voter can help their preferred small party win is to vote 1 for it, then a bunch of other small parties, with the majors close to last, followed only by the small parties they think are worse.  Here's a screenshot of their argument:

Firstly, if you put your favourite small party first, whatever you do after voting for them has absolutely no impact on their chances of election, and claiming that it does is simply misinformation.  The reason is that your vote will not leave your favourite small party until all its candidates have been elected or excluded.  

The author then claims that voting 1 for a small party could also be better for your favourite major party and that this vote could indeed be "more than twice as effective" for the latter.  This is also misinformation.  The error is in the author's understanding of how surpluses work.  The author suggests that if you vote 1 for a candidate who polls 1.7 quotas, only 0.7/1.7 = 41.2% of your vote is available to the big party's #2 candidate, whereas had you instead run your vote through some minor candidate first, your vote would reach big party's #2 at full value.  That's true, but what he doesn't realise is that by increasing the size of the big party's top candidate's surplus by 1 vote, you increase the combined value of all that candidate's remaining votes by 58.8% of a vote.  And virtually all of those votes will also be flowing to big party's #2 (bar a few below the line votes across party lines), so in fact your vote is only very slightly less effective for the second candidate.  Where this choice makes a bigger difference is if the second candidate is later excluded, meaning that your vote flows on to other parties at reduced value, while you have increased the value of other votes that may go all over the place.

It's also possible the recommended strategy could harm one's preferred major party.  Because major parties tend to outperform micro parties on preferences, it's possible that a major party candidate with a small fraction of a quota could snowball their way past several minor parties and win.  (One Nation's Malcolm Roberts overtook six slower-moving micro-parties in this way in 2016, and clearly there's potential for a major party to do it too).  There could be a tipping point where a voter engaging in what I call "preference running" unwittingly knocks out their preferred major party from a winning position, in favour of a smaller party that loses.  

It is sound advice for those who do wish to vote tactically to not vote 1 for any candidate who will be elected on the first ballot (see tactical voting section of my 2019 article on how to make best use of your Senate vote.)  However, a major party voter can achieve this by voting below the line and putting the bottom candidate for their party first; there is no need to vote 1 for a different party.  Also while putting an uncompetitive party 1 above the line then voting 2 for the preferred major party is a safe way to avoid getting one's intentions diluted, running one's vote through a lot of minor parties including some who may command a decent vote may be tactically risky as discussed above.

Updates may be added if I have time to keep an eye on RDA's output through the week.  

Footnote: Above I asserted that "the voter who thinks the Labor and Liberal parties are genuinely the worst two parties on the ballot is a pretty unusual beast."  I had a look at the 2019 Tasmanian ATLs, and of about 15,000 voters who numbered at least 14 out of 16 ATL boxes, only 40, a trivial number, put both major parties last.  A further 44 left the Garland group box (a common source of voter confusion) blank and otherwise put both major parties last.  The majority of these voters voted either for One Nation or Jacqui Lambie Network.