Saturday, December 18, 2021

The Overrated Impact Of Party Preferencing Decisions

Advance Summary

1. For all the noise about preferencing strategies and preference flow changes, changes in the relative primary votes for the major parties are a much bigger factor in most recent federal election results.

2. The widespread claim that United Australia preferences caused the Coalition to win the 2019 election is false.

3. Labor is much more dependent on preferences than the Coalition and routinely wins many seats from behind.

4. No party's preferences will ever flow 100% to any other party and there is nothing anyone can do about that - whether it is Greens or Labor being excluded, some preferences will always go to the Coalition.

5. The preferences of climate-concerned independents in city seats tend to flow to Labor, although not quite as strongly as the other way around.

6. In many cases the preference flows from climate-concerned independents are irrelevant, since they will rarely be eliminated in seats that are closely contested between the majors.  

7. How to vote cards mainly exist to protect against informal voting.  Their impact on outcomes is minor.

-------------------------------------------------------------------------------------------------------


This piece is written to tie together a number of strands about preferences in House of Representatives (and to a much lesser extent Senate) elections, mainly in response to claims I commonly see on Twitter.  There is a persistent myth that United Australia preferences caused the Coalition's 2019 victory, but in recent weeks I have also seen an upswing in misleading claims about Green and independent preferences.

Preference flows in House of Reps elections are in one sense very important.  It's very rare for one side to get more than half of the national primary vote by itself (this last happened in 1975, though 1983 was close) and the major parties usually depend on some sort of preference flow to get across the line in enough seats for victory.  This is particularly so for Labor, which has received substantially more preferences than the Coalition from 1987 onwards, but in that time has only beaten the Coalition on primary votes three times.  

However, the preference flows of most of the parties that run for parliament are fairly stable, and for this reason changes in the primary votes for the major parties are a much bigger deal in deciding elections than preferencing.  For the elections between 1987 and 2019 I estimate that:

* the average impact of primary vote shifts between the major parties on the two-party preferred result has been 2.86%.

* the average impact of changes in the preference flow behaviour of established minor parties on the 2PP result has been only 0.28% (see how I calculate this here)

* the average impact of other factors (mainly changes in support for and breakdown of minor parties, but also the level of three cornered contests) has been only 0.44%.

The 2PP does not necessarily determine who wins the election (as seen in 1998) but it's generally a good proxy.  

Of these twelve elections, only three have seen substantial preference flow changes (1990, 2013, 2019) and only in 1990 did this have a major impact on the outcome, saving Labor from what would otherwise have been an extremely close seat result that could have about as easily gone either way.  In 2013 Labor's share of Greens preferences increased but this was cancelled out by the Green vote falling.  In 2019 the 2PP swing to the Coalition was largely explained by an increased flow of One Nation and United Australia preferences, but the lack of a primary vote swing to Labor and the Coalition's superior marginal seat performance meant the Coalition would have won the election anyway.

Minor party breakdowns had a major impact at the 2010 election, at which an increase in the raw Greens vote saved Labor from defeat without the flow of Greens preferences changing much.  There were also substantial minor party breakdown impacts in 2013 (see above), 1993 (collapse of Democrat vote hurt Labor but they won anyway) and 2016 (swing to Greens boosted the 2PP swing to Labor making the election much closer than it otherwise would have been.)

Overall though, the major game in House of Reps elections is major party primary votes, not the raw votes for each party as often misleadingly claimed (see debunk here) but where the major parties stand relative to each other.  At present minor parties get about a quarter of the vote and Labor's preference share is recently around the low 60s (dipping just below 60 in 2019).  This means that Labor can reach two-party preferred equality from about 5-6% behind on primary votes, but no further back.  A sufficient cause by itself of Labor's 2019 defeat was losing the primary vote by 8.1% - nothing that could have realistically happened with preferences was going to save them from that.    Though the breakdown of swings in the marginal seats left Labor needing a 2PP somewhere in the mid 51s to form government (depending on the will of the crossbench), so they would have actually needed their primary vote to be just a few points behind the Coalition's to win. 

The UAP Preference Furphy

The claim that Clive Palmer delivered government to the Coalition via United Australia Party preferences is a widespread myth on Twitter, spread mainly by accounts that seek to blame Palmer for the result rather than blame the basic fact that Labor's primary vote was way too low. 

In fact, the Coalition led on primary votes in every seat it won - it is generally Labor and independents who use preference flows to win seats where they trail on primaries, not the Coalition.  Furthermore, the UAP polled only 3.43% of the primary vote, and despite it preferencing the Coalition on its how-to-vote card recommendation in every seat, the flow of its preferences was still not all that strong (65.14% to Coalition).  This preference flow gave the Coalition a 0.52% two-party preferred legup compared to what would have happened if UAP preferences had split evenly, so even had UAP preferences not helped the Coalition, it would have won the 2PP 51-49 anyway.  

In terms of actual seats, there was one and only one seat that the Coalition won because UAP preferences favoured it rather than breaking evenly, and that was Bass.  And that doesn't mean UAP's decision to run or its preference allocations caused that result, since the UAP voters who preferenced the Coalition there may have preferenced the Coalition anyway.  There was one other seat (Chisholm) where the Coalition needed some preference flow from UAP voters, but an even share would have been enough and they were obviously always going to get that.  Even had the Coalition lost both Bass and Chisholm it would still have formed government, albeit with a one-seat minority.  So it is simply not true that UAP preferences decided the election.

Aside from the claim being false - and in effect a denial of the agency of the voters whose primary vote choices decided the election - it's also remarkable how people present the fact that one minor party's preferences helped the Coalition win an election it was winning anyway, while ignoring the far greater assistance Labor gets from Greens preferences.  In 2019 Labor's leg-up on Greens preferences improved Labor's 2PP by 3.35% compared to what would have happened if Greens preferences had split evenly.  Getting a favourable share of Greens preferences saved Labor from defeat in a massive fifteen seats (including some where Labor led all the way but would have lost on preferences had the Green votes specifically split 50-50).  Of course, Greens preferences will not split evenly anytime soon, so this is taken for granted, but when a smaller party's preferences flow more weakly to the Coalition and in smaller numbers, somehow many people on social media decide that the UAP's preference flows decided the election.  

A related Palmer preference claim I've seen a lot of is the claim that the UAP takes Labor votes and turns them into Coalition preferences via how to vote cards.  We don't actually know what percentage of UAP voters even follow Reps how-to-vote cards but I suspect from the Senate it is quite low, and similar to One Nation (10-12% in Queensland 2017, where a variety of preferencing choices made it possible to estimate the follow rate.)  The 2019 UAP flow can be explained just fine by (i) The UAP attacking Labor heavily and therefore being more attractive to voters who disliked Labor anyway (ii) The UAP's natural support base including outer-suburban demographics that swung against Labor - largely as a reversal of what they did in 2016 because they didn't relate to Malcolm Turnbull or his campaign.

None of this is to say Clive Palmer's megaspend (which wasn't all about attacking Labor) didn't  influence the election result, but that's a different story.  It may well have contributed greatly to Labor's disappointing performance on primary votes.  This, though, is easy to claim and more difficult to prove.  Palmer was far from the only source of viral scare attacks on Labor that were rampant on Facebook and other social media, many of which probably cost their creators very little.  Also, there are other examples (such as Queensland 2020) of the UAP throwing big bucks around with nothing to show for it.

100% Greens Preference Flows Will Never Happen

Largely following Labor's announcement of a 43% emissions reductions target, I have been seeing an increase in comments about Green preferences, especially around the fact that not all Greens preferences flow to Labor.  Sometimes these forms take the claim of outright electoral misinformation, in which the tweeter tries to claim that a vote for the Greens assists the Liberal Party, ignoring the fact that the voter decides their own preferences and hence who it will assist.  A recent moderately viral example read "If you've decided to vote green or independent without knowing where their preferences are going then you are potentially deciding to vote LNP."  (That ignores that the Greens have recommended preferences to Labor above the Coalition in all seats consistently in recent Reps elections.)

There is a lot of attention on the minority (17.8% as of 2019) of Greens voters who preference the Coalition ahead of Labor.  

In fact, we see a very similar thing when Labor is excluded.  The following shares of Labor voters preferenced the Coalition ahead of a left-of-Coalition alternative:

Warringah: 14.1% for Abbott (Lib) over Steggall (IND)
Wentworth: 15.9% for Sharma (Lib) over Phelps (IND)
Kooyong: 16.6% for Frydenberg (Lib) over Burnside (Green)
Indi: 18.4% for Martin (Lib) over Haines (IND)
Melbourne: 19.6% for Sherson (Lib) over Bandt (Green) (NB: ALP candidate disendorsed)
Mayo: 19.8% for Downer (Lib) over Sharkie (CA)
Cowper: 24.4% for Conaghan (Nat) over Oakeshott (IND)
New England: 25.3% for Joyce (Nat) over Blakester (IND)
Farrer: 28.5% for Ley (Lib) over Mack (IND)

(For completeness, in Kennedy 25.6% of preferences flowed to the LNP instead of Bob Katter and in Maranoa 65% flowed to the LNP in preference to One Nation.)

The Wentworth result is especially interesting because it was a very rare case in which the Coalition needed some Labor preferences to cross the line and got them.  In fact, Sharma only needed 4% of Labor preferences and nothing could have stopped him from getting that many, but that puts Wentworth in the same boat as seats where a perfect flow of Greens preferences would have won the seat for Labor but the flow required was unrealistic.   There was one other such case in 2016 when Labor issued an open HTV card in Grey, and NXT candidate Andrea Broadfoot did not get a strong enough flow to knock off Rowan Ramsay (Lib).  Broadfoot needed 81.6% of Labor preferences and got 72.5%; had she been preferenced on the card it's possible the Turnbull government would have started the term with 75/150 seats instead of 76.  (In fairness to Labor's otherwise senseless tit-for-tat non-preferencing of NXT, some of the other NXT candidates were flaky.) 

In 2019 Labor lost three seats it would have won with a perfect flow of Greens voter preferences: Bass, Chisholm and Boothby.  That number was down from seven in 2016 (Capricornia, Forde, Gilmore, Robertson, Chisholm, Dunkley and Hasluck - not Dickson as per another common false claim although Dickson was such a case in 2007).  

Left-leaning voters frequently don't understand why any voters for a left-of-centre party would preference the Coalition over a left or centre candidate.  However, there could be many reasons, each minor and unlikely but together adding up:

* Some of these votes are determined by random factors such as donkey voting or partial donkey voting.

* Some voters are not voting based on left-right politics but based on impressions of the candidates that may come from community connections (or in some rural seats simply on which candidates are from their area).  

* Some Labor voters strongly distrust independents and Greens and think that if the party they support doesn't win, the other party should get a majority and a clear run.

* Prominent figures will often acquire some level of local opposition.  For instance in Phelps' case this could come from her council career or perhaps even her views on medicine or other issues.  (Replies to my own viral tweet about this included some suggestions, though her views on COVID vaccines were not known in 2019 of course.)

* The gender of the candidates influences some voters.

* Some Greens voters are attracted to the party based on non-environmental policies, or have environmental views but also crosscutting economic views that favour the Coalition.

Many, many others could be added.  Greater harmony between Labor and the Greens could probably increase the preference flow between them by a few points but it will never be 100%.  Likewise it is never 100% from Liberals to Nationals and especially not from Nationals to Liberals.  Now and then the Coalition loses in a seat that would have been won with a 100% flow between the Coalition parties (most recently Eden-Monaro both in 2019 and the subsequent by-election.)

Making a big deal of candidates losing because the preference flow is not quite 100% has a long tradition, going back to at least 1961 when at least 93 Communist Party voters preferenced Jim Killen (some others would have done so indirectly).  Had no votes arrived from the Communist candidate Killen would ultimately have lost and the parliament would have been deadlocked, but by far the greater contributor to Killen's win was subsequent DLP preferences in his favour.

In the modern version of the game, the commonest misconception is that the Nationals or Greens candidate caused the preference "leakage" by running. Mostly, a person who votes 1 National 2 Labor wouldn't have voted 1 Liberal anyway, and a person who votes 1 Green 2 Liberal wouldn't have voted 1 Labor.  There will be the odd exception - for some voters it might be taboo in their social circle to vote 1 for a given major party, but preferencing that party is considered fine - but Labor would both gain and lose preferences as a result of this.  The main reason Coalition parties minimise three-cornered contests isn't preference leakage (though that can be an issue where a bit of tension develops between competing Coalition candidates or, eg, Nats voters want to stop Liberals occupying the seat).  The main reason is to avoid wasting resources.

For sure, the Greens are a pain in the neck for Labor.  Their presence compels Labor to defend otherwise safe seats and tempts Labor to throwing them sops in search of preferences or left-wing votes (with most such sops providing ammo for the Coalition).  The Greens' habit of declaring hung parliaments in advance and outlining what they will do with their power in them is especially toxic (and especially silly until the Greens manage to win more than one crossbench seat out of six, since there is a fair chance a party forming minority government will be able to ignore their less palatable demands and work with others.)  However, it should be recognised that the Greens do quite consistently recommend preferences to Labor over the Coalition in federal elections and do work hard on handing out cards to that effect. This is saving seats - such as Macquarie, where Labor needed a seemingly difficult 87.0% of Greens preferences to win and got 89.1%.

Indies Don't Send Preferences To The Coalition

In the last few weeks there have been various announcements of medium to high profile independent "Voices Of" and similar candidates running.  "Voices Of" is essentially a franchise-like movement that isn't a formal party but conducts preselections of candidates who agree with core policies. 

One thing I have seen is a rather obsessive focus on how independents might direct their preferences, the concern being that they will act as preference harvesters for the Coalition.  This is all rather silly though, because:

 (i) independents have relatively little impact over their own voters' preferencing decisions (unlike major parties) 

(ii) many independents - whatever their politics - issue open preference tickets anyway for branding reasons 

(iii) the preference flow from an independent only matters if the independent is excluded without making the top two and the seat is reasonably competitive between Labor and the Coalition.  Of the seats where "Voices Of" type candidates have been announced so far, only Boothby is very marginal, though Labor might also hope to be competitive in vacant Flinders.

(iv) independent preferences (especially those of what I call "climate independents") tend to flow to Labor, not Coalition 

A good example of all of these points together was what happened with Phelps in the 2018 Wentworth by-election.  Phelps actually preferenced the Coalition ahead of Labor on her card, but of those voters who ranked Phelps above both major parties, it can be determined that 58.5% put Labor above the Coalition anyway.  This flow was also irrelevant since Phelps was never excluded.

The following are 2019 preference flows to Labor (or in the case of Kooyong, Greens) vs Coalition from a bunch of independents or crossbenchers who made noises about climate or allied with other climate indies during the campaign (those featured in the "#IndependentsDay" video plus a few others).  A * means an estimated preference flow based on 3CP results.  An X means the candidate was excluded and their preferences thrown.

Yates (Kooyong) 79.8 X 
Steggall (Warringah) 77.6 
Banks (Flinders) 73.6 X
Haines (Indi) 73.5 
Wilkie (Clark) 71.8 
Phelps (Wentworth) 70.4* 
Sharkie (CA, Mayo) 69.7 
Oakeshott (Cowper) 68.1* 
Blakester (New England) 65.9* 
Kingston (Hume) 64.5 X
Thompson (Mackellar) 63.6 X
Modica (Mallee) 59.3 X
Stewart (Curtin) 52.4 X
Miller (Lyne) 49.7 X
Mack (Farrer) 45.9*
Kingston (Mallee) 32.0 X

For the excluded urban indies listed above the average flow was 66.8% to Labor or Greens.  For the elected urban indies it was very slightly higher (72.4%).  In rural seats the flows were much weaker and in some cases to the Coalition - eg Kingston (Mallee) is a farmer so it's not that surprising the Nationals did well on his voters' preferences.  The IND to Labor/Greens flows in urban seats are not as strong as the Labor to IND flows in those seats (typically low 80s) but they are not that far behind.

Many of these 2019 candidates made climate change only a relatively minor part of their campaign.  In 2022 there will be many "voices of" candidates (this Wikipedia page is listing them with 13 announced so far) running very similar campaigns with climate, an integrity commission and the treatment of women in politics as their main campaign planks.  Perhaps even more so than in 2019, such candidates will take votes from the Greens and Labor with these votes returning to the left as preferences if the candidate is excluded.  

There's a theory that these independents act as enablers for token protest votes (a climate-concerned Liberal voter votes 1 Independent 2 Liberal and has ticked the box of taking a stand on climate change without voting to kick the government out) but (i) we don't know that such voters would have voted 1 Labor or Green if there was no independent (ii) if the indie makes the final two, then the protest vote may help elect the independent.  

The serious uncertainty from the Coalition and Labor/Greens supporter perspectives alike is nothing to do with preferences - it's who "Voices Of" independents will support, and on what terms, if they do get the balance of power.  

The Main Purpose Of How To Vote Cards

How to vote preference "deals" are often the subject of more media attention than they deserve.  Even the term "deals" is often incorrect, since in many cases a party unilaterally decides to recommend preferences to another.  Even when deals exist, the preferencing decisions of obscure parties have virtually no impact: their vote is low and the proportion who follow the card is low too, so not much times not much equals more or less nothing.  For parties like United Australia and One Nation, it's possible their preferencing decisions might be worth a few tenths of a percent to the overall 2PP - not nothing, but highly unlikely to be decisive.  

The main purpose of how to vote cards is to keep the informal voting rate down.  Currently about 3% of voters make errors with the voting system (such as numbering only one box, or repeating or omitting numbers by mistake).  Without something for low-information voters to copy, this rate would be much higher and parties would miss out on votes.  (This is why calls to ban how-to-vote cards are naive and should be ignored.)

Major party voters are more loyal copiers of their own party's cards than minor party voters, but even so the proportion copying them is probably less than half.  We can see exact evidence of the differences in rates from South Australia, but it is also obvious from cases where a party has changed its order from election to election - for instance the Liberals switching from preferencing Greens to preferencing Labor - and from cases where minor parties have preferenced different parties in different seats.  

While the South Australian data suggests the proportion of Greens voters who copy the card is still in double digits, it also appears from Senate how to vote cards that voters will not follow cards where the party preferences an obscure candidate or one that is anathema to voters for that party.  So the proportion of Greens voters who would follow a card if the Greens chose to preference the Liberals is probably even lower - an exception coming in the 2020 Johnston (NT) by-election where, with absolutely nothing on the line, the Greens recommended a preference to the CLP as a protest about fracking.

Finally, Re The Senate ...

It wouldn't belong in this article except that it has so frequently come up in the Twitter discussions of various parties' preference flows.  There are a lot of people out there who think that to control your own preference flows in the Senate you need to vote below the line and if you vote above the line the parties will send your preferences somewhere.  This is no longer true.  Under the Senate reforms introduced in 2016 if you vote above the line, your vote can only go to the parties you have personally preferenced above the line, and never to any others.  Also, if you vote below the line and make an error in the first six numbers, your vote is informal.  For most voters, voting below the line is more onerous than voting above the line, so generally you should vote below the line only if one or more of the following apply:

* You wish to re-order the candidates within a given party and put them in a different order to that their party has listed them in.

* You wish to vote across party lines, eg putting some candidates for party X above party Y and other candidates for party X below party Y

* You wish to vote for or give a preference to an ungrouped candidate.  A special case of this is if you want to put a grouped candidate last, meaning that you need to vote below the line to preference all the ungrouped candidates ahead of them.  

The only elections where voting below the line remains necessary to avoid preference harvesting are those for the Victorian Legislative Council (state upper house) and Melbourne City Council.

3 comments:

  1. Kevin, I've only just seen this, but it is very good. I have shared and recommended it on my networks.

    ReplyDelete
  2. It is also my impression, based on some decades of interactions with voters at polling places, and discussions in various forums, that a significant number of Australians believe that preferential voting is a kind of "points" system similar to Eurovision Song Contest voting or Brownlow Medal voting, so that (to simplify) such people think that a vote that goes 1 Greens 2 Labor in a single member constituency only counts as two-thirds of a vote for Labor if the Greens preferences are distributed, rather than being transferred at full value. It is understandable that such a misunderstanding exists, as a non-trivial number of people have much more engagement with things like Eurovision, sports award voting, the tipsters' poll in the racing form guide, etc, than with electoral systems. However, I believe it is a misunderstanding that the AEC, political parties, and election commentators of various kinds should do more to try to correct.

    ReplyDelete
    Replies
    1. Yes I have encountered this too, albeit more with reference to the Senate.

      Delete