Preferences, Donkey Votes and the Canning By-Election

With less than a fortnight to go until the Canning by-election, it is worth clarifying a few points on whether preferences will be important in the result.

Will Canning be Decided by Preferences?

That depends on your definition of decided. A broad definition defines any seat where no candidate polls 50% of the first preference vote, and where preferences need to be counted, as being 'decided' on preferences.

I prefer a narrower definition, where a seat decided on preferences is one where the winning candidates differs from the candidate with the first preference majority - that is the winning candidate differs from the first-past-the-post winner.

At the 2013 Federal election, only 53 of the 150 electorates saw a candidate win with more than 50% of the first preference vote. Another 82 saw the candidate that led on first preferences win after preferences were distributed, while there were 15 seats where preferences definitely determined the result by electing a second placed candidate.

At the nine elections for Canning since 1990, two were won by Don Randall with more than 50% of the first preferences vote, another six were won by the candidate that led on first preferences, while Labor came from second place to win on preferences in 1990.

All polls indicate that the Liberal first preference vote is down since the 2013 election, so a victory by Liberal Andrew Hastie on first preferences seems unlikely.

A victory by Labor candidate Matt Keogh on first preferences would require the Labor vote to double and seem extremely unlikely. Labor leading on first preferences also seems unlikely based on opinion polls.

Using the narrower definition of the result being decided on preferences, the chances of preferences deciding Canning depend on Labor being close enough to the Liberal Party on first preferences to come from behind and win on flows of preferences.

The last time Labor came from behind to win Canning was in 1990, when Labor trailed by only 0.3 percentage points, the Australian Democrats polled 7.3% and the WA Greens 7.0%.

The Greens have only polled above 8% once since 1990, polling 8.3% in 2010. Without a significant increase in the Green vote, it seems unlikely that Labor could close a large gap on first preferences. The size of the gap would tend to increase as the Green vote rises, as increases in Green support tend to come at the expense of Labor.

After preferences Labor closed the first preference gap by 3.8% on coming from behind in 1990, by 2.8% in 2001, but under 2% at every other election and by only 0.8% in 2013.

Realistically on past results, Labor would have to be within 3% of the Liberal Party on first preferences if it hoped to win Canning from second place on preferences. Labor could win with a larger gap if there was a rise in the Green vote, but Canning is not prime Green territory. It recorded the third lowest Green vote in WA in both the House and the Senate at the 2013 election.

What are the past preference flows ?

The table below sets out the minor party preference flows by party in Canning and in Western Australia as a whole at Federal elections since 2004. The distribution of National Party preferences have been excluded from the table.

Clearly the two key parties in this table are the Greens and the Australian Christians, previously known in Western Australia as the Christian Democrats.

In both Canning and across Western Australia, Green preferences have been consistently around the 75% mark in their flow to Labor. That means that if the Greens were to poll around 10%, 7.5% would flow to Labor and 2.5% to Liberal, narrowing a first preference gap by five percentage points. It is possible with a more intense local campaign, Green preferences could be a few percentage points higher at a by-election.

What is remarkable to me in Canning is the consistently high flow of Australian Christian preferences to the Liberal Party. Preference flows above 80% are highly unusual for minor parties with only 3% of the first preference vote.

Either the Australian Christians are very good at handing out how-to-votes in Canning, or they have tapped into local church networks, alerting supporters to how they should direct preferences.

Two of the newer parties contesting Canning also contested the Griffith by-election in early 2014. The Pirate Party finished fourth with 1.5% of the vote and saw a preference flow of 71.3% to Labor whose candidate was one above the Pirate Party candidate on the ballot paper.

The Sustainable Population Party polled just 0.9% of the vote but delivered a remarkable 72.5% of preferences to Labor, this high flow from a low vote largely due to the donkey vote.

Who Controls Preferences?

The answer is voters themselves. Parties and candidates have no control over preferences. They can try to influence preferences by distributing how-to-vote material, but the only preferences that count are those that are filled in by voters themselves on their ballot papers.

Ballot papers are counted out to preferences according to the numbers completed on each ballot paper by the voter. Parties and candidates have no say whatsoever over the actual distribution of preferences takes place..

What How-To-Votes Are being distributed in Canning?

I have now added how-to-vote material for Canning to my ABC elections page. The material is linked down the page in the candidate profiles. (If you want to send me HTV material, use the e-mail address at this link.)

The Labor and Liberal Parties do not distribute how-to-vote material to influence preferences. In a seat like Canning the two major parties will finish first and second and never have their preferences distributed. For both parties, how-to-vote material is designed to insure that voters for each party cast a formal vote with a correct sequence from 1 to 12.

A ballot paper with a single '1' is informal. If a voter misses any number between 1 and 11 their vote is informal. If a voter doubles up any number their vote is informal. If a voter goes 1 to 10 then 99, 100, their vote is informal. Formality rules for the House of Representatives are ridiculously strict.

The ballot paper draw has greatly assisted Labor's how-to-vote. Labor's candidate drew spot eight with the Greens at nine. This has allowed Labor to issue a how-to-vote that numbers 1 Labor, 2 Green and then a simple ordering down the ballot, back to the top then straight down again. There are no breaks in sequence.

The Liberal Party has had to break sequences and hop around the ballot paper to produce an ideologically consistent set of preferences. It will require greater effort for voters to follow the Liberal how-to-vote compared to Labor's.

The Australian Christians and the Greens have also published complex sequences of preferences favouring like-minded minor parties first, before the Australian Christians reach the Liberal Party and the Greens reach Labor.

Do people follow how-to-votes?

To get a strong preference flow, party workers must deliver a how-to-vote into the hands of every voter. The chances of someone guessing exactly the same preference sequence as a party how-to-vote is low, especially in a 12 candidate contest like Canning.

Data published by the Electoral Commission of South Australia shows that the proportion of people following a registered how-to-vote decreases as the number of candidates increases. At the 2014 state election, the number exactly following a how to vote in the 17 electorates examined ranged from 34.5% 58.3%, with the higher figure occurring in a four candidate contest.

South Australia held two by-elections last summer, an eight candidate contest in Fisher and five candidate contest in Davenport. With fewer candidates, how-to-vote conformity was much higher in Davenport. Liberal how-to-vote conformity was 49.5% in Fisher and 67.2% in Davenport, while for Labor it was 38.2% in Fisher and 53.3% in Davenport.

An important caveat in using this data is that South Australia displays registered how-to-vote material on voting partitions. Every voter can see the sample how-to-votes on the cardboard screen in front of them.

A similar analysis carried out in eight electorates by the Victorian Electoral Commission after the 2010 election found a clear relation between a party's level of vote and the proportion of ballot papers conforming to how-to-votes.

For Labor it was 41.0%, but much lower in country electorates, and 44.8% for the Liberal Party, but much lower in safe Labor electorates. Green how-to-vote conformity was only 24.5% and Family First 12.7%. The rates of conformity were clearly lower than under the different how-to-vote arrangements in South Australia.

Despite Green voters not conforming to the how-to-votes in the Victorian sample, the party still delivered 75-80% flows of preferences to Labor. It seems that many Green voters end up with their preference at the same destination, but prefer to use a different route to get there.

How Important is the Donkey Vote?

The top position on the ballot paper was drawn by the Palmer United Party candidate, and the party has published a preference sequence running straight down the ballot paper from 1 to 12. If many voters follow this recommendation, it will favour the Liberal candidate Andrew Hastie.

Obtaining the top spot on the ballot paper has been measured in the past as being worth on average about 1% of the vote. This may give a boost to Palmer United.

However, a paper by David Peetz in the 1980s used a different approach in measuring a 2-party preferred advantage between Labor and the Coalition based on which party appeared higher on the ballot paper. This analysis produced the estimate that if the order of the Coalition and Labor changed in an electorate between elections, this would produce a change of around one percentage point in the 2-party preferred swing.

The fact that Liberal Andrew Hastie is higher on the ballot paper than Labor's Matt Keogh gives Hastie a narrow advantage by the donkey vote, but the advantage is unchanged on the 2013 election when Don Randall drew top spot on the ballot.

So the donkey vote favours the Liberal Party in Canning, but it is neutral in its impact on the swing needed compared to the 2013 election because the Liberal Party had the donkey vote advantage in 2013 and again in 2015.

Will Voters be Confused by the Liberal Democrat candidate?

It seems pretty clear that voters confused the Liberal Party and Liberal Democrats in Senate races at the 2013 election, but the structure of the ballot paper means that this should be less important at the Canning by-election.

At 12 the field of candidates in Canning is unusually large, but the ballot paper is still manageable compared to some of the giant Senate ballot papers in 2013.

The Liberal Democrats polled 9.5% in the NSW Senate contest in 2013 having drawn Column A on the ballot papers. The Liberal and Nationals were buried away on the right of the ballot paper in Column Y, polling the lowest Coalition Senate vote in NSW since 1943 at the same election where the Coalition recorded its highest House vote since 1975.

It appears that, reading left to right, many potential Liberal voters marked '1' with the Liberal Democrats and stopped reading.

In Canning there are 12 candidate names with party names in smaller font underneath. Voters will have to number all squares which means reading all candidate names. Even though the Liberal Democrat candidate has drawn the helpful position two, full preferential voting means that more voters will spot the presence of both a Liberal Democrat and a Liberal Party candidates. Hopefully fewer will be confused.

Comments

Don't forget the #ReverseDonkeyVote

COMMENT: It is negligible.

- Angela Smith September 08, 2015 at 10:31 PM

Given that it is a virtual certainty that someone voting 1 for Liberal in Canning will never have their preferences distributed, why on earth did the Liberals design such a complex sequence of numbers on their how to vote card, thus increasing the risk of votes being lost due to informal votes? All they needed to do was to make it nice and simple, showing a "1" next to the Liberal candidate and numbering the boxes sequentially as you go down the voting paper, then continuing the sequence as you wrap up to the top. Sure, it would have meant Labor would get a high preference number of 3 and greens preference 4, but Liberal voter preferences would never be distributed anyway in that seat.

COMMENT: Since the 1998 promise by major parties to put One Nation last, parties have found their how-to-votes under greater scrutiny in the past.

At the 2013 election, Andrew Leigh in Fraser took your approach with his how-to-vote, which resulted in Rise Up Australia being placed ahead of the Greens. This did not matter as Labor preferences would not be distributed and the Greens would also poll more than Rise Up Australia.However, Leigh was still criticised in the local newspapers. (See article here http://www.canberratimes.com.au/act-news/canberra-mp-andrew-leigh-defends-howtovote-card-rejects-antiislam-agenda-20130903-2t3kc.html )

I presume the Liberal Party have chosen the path of avoiding controversy.

- Jamie September 08, 2015 at 11:44 PM

You know that the 1-12 on PUPs how to vote is NOT a list of how to vote. The comment at the top clearly states:

"You must number every box.

The example below shows you how to vote for Palmer United Party.

Providing you Vote 1 for Palmer United

allocate the preferences as you wish"

COMMENT: Anyone who votes for Palmer United will at some point have to choose an ordering between the Labor and Liberal candidates. A donkey vote will favour the Liberal Party. The sample shown will also favour the Liberal Party. I expect Palmer United preferences to favour the Liberal Party and I believe the published how-to-vote will play a part in that being the case.

- Catriona Thoolen September 09, 2015 at 06:02 PM

At the last election, 11% of eligible voters did not vote. Therefore, for any candidate to win, they would need 50% of the electorate but that would have to come from the 89%. 50% of 89% is 44.5%, not 50%. So assuming a potential 100 votes, but only 89 cast, the winner would need 50/89 which is 56%. How many were elected on less than 56%?

Why are abstentions not included in the count? They are, after all, valid expressions of opinion and choice.

COMMENT: The winner in Australia requires 50% of the votes cast, not of voters on the roll.

- Tony Sadgrove September 09, 2015 at 09:19 PM

How u can say Australian Christians are key and Palmer United are not shows that you just can't count or u don't go in real election results but media hype Palmer United increased their vote from 6.9 to 26.5 in the senate election yet u dismiss them like u did before last federal election when Clive Palmer need up with over 50 per of vote disappointing I am sure once again the result in canning will for the third time show strong support for Palmer United but u continue not to recognise it as a factor

COMMENT: Strong preference flows can help to change the order candidates finish on first preferences. The Australian Christians and Greens have exhibited strong preference flows in the past. Palmer United Party have not. Palmer United preference flows are therefore less likely to alter the ordering of candidates than Australian Christian and Green preferences. I do not expect Palmer United to poll anything like the vote at the 2014 WA Senate re-election.

- Mike thompson September 10, 2015 at 09:00 AM

Mike Thompson wants to know why the vote of Australian Christians will be more important than PUP. Antony has given one reason. The other is that last election, when PUP polled more strongly than opinion polls now predict, PUP preferences didn't overwhelmingly favour either party. In Canning, as can be seen above, it was pretty line ball. On the other hand, Australian Christians preferences flow to the Liberals with an extraordinary level of discipline. Therefore, a high PUP vote is good for PUP, but does little to alter the fate of the by-election. A high Australian Christians vote, however, helps the Libs and therefore influences the result.

COMMENT: Actually, I think the most important factor in the result will be the change in first preference support for the two major parties. The result of the by-election will be determined by the change on 2013 results. If the primary votes and preference flows for the minor parties are roughly the same as in 2013, then it is the primary vote change that is important. It may be that one or other minor party preferences put one candidate or the other over the line, but I expect the change in first preference support for Labor and Liberal to be the determining issues in the result.

- Greg Platt September 10, 2015 at 01:46 PM

Hi Antony, following up on Tony's question above... I have had a quick loot at the relevant section of the Electoral Act and can not see any defined threshold for informal votes to invalidate an election.

Taken to extremes, if an election was held with 100% turnout, 97% informal votes, 2% for Candidate A and 1% for Candidate B - Candidate A can be considered duly elected?

I'm not advocating a position on whether this should or should not be the case, merely testing what is currently coded into the Act.

COMMENT: Informal votes do not invalidate an election. A candidate wins by obtaining the required majority of the number of votes in the count. In the case of full preferential voting, that is 50% + 1 of the formal vote. Under optional preferential voting it is 50% + 1 of the vote remaining in the count.

As was the case with the Franklin Dam referendum in Tasmania in 1981, a high informal vote might de-legitimise a result but in law it can't alter the result.

- Scott September 10, 2015 at 02:36 PM

Antony, you credit the strong flow of preferences from Australian Christians to Liberal as being due to their ability to communicate to or direct their supporters; but isn't a better explanation simply that their supporters naturally prefer the Liberals anyway? (Support which only likely to intensify now that the the ALP has voted to force its MPs to support gay marriage.)

COMMENT: Yes, but the preference flow in Canning has been consistently extremely high for four elections. It is as strong a preference as you see for the Greens in their strongest seats, and as you might see for Liberal flows to the National Party. It is much stronger for the Australian Christians than in other WA seats, much stronger than for the same party in other states, stronger than the long established Christian Democrats in NSW, and much stronger than either Family First or Rise Up Australia.

- Alan September 10, 2015 at 03:22 PM

To add to my comments above. I am arguing that abstaining is voting, given that both voting and abstaining are expressions of opinion. A non- opinion is still an opinion. If a blank or spoiled paper is lodged, that, to me, is "casting a vote". Hence if no candidate gains 50%+1 on 2PP then "no election" should be declared. We have a compulsory system. It follows that it should be compulsory to include the full 100% in the count, not just the 89%. Election results would then be very different.

COMMENT: As I said, the law has always been otherwise. There are referendum that have been held under these sort of rules, such as the 1979 Scottish devolution referendum that required both a majority of the vote and a minimum vote representing 40% of the total enrolled voters. The ACT allows electoral provisions to be entrenched based on getting 50% of the electoral roll.

There have been some presidential elections around the world that have required a minimum turn out for validity, but this has sometimes been abused by candidates that can't win engineering non-turnout to invalidate an election.

- Tony Sadgrove September 10, 2015 at 05:17 PM

Tony Sadgrove: It's entirely academic anyway under full prefential voting, since as long as the valid votes cast are more than 50% of the roll the winner of "50%+1" of the valid votes will be able to achieve 50%+1 of the roll as well (eventually *all* the valid votes cast would end up with one candidate, since the valid votes must express a full slate of preferences).

COMMENT: In the same way the Electoral Act has never specified a majority be of electors, neither has it ever been reduced to a single candidate.

- kme September 11, 2015 at 12:44 PM

In the 2014 WA Senate Election the Greens polled 11.65% in Canning compared to 6.83% in the 2013 Senate Election.

- Vincent Mahon September 11, 2015 at 01:44 PM

Your narrow definition of "decided by preferences" precludes any possibility of saying that the preferences decided the election in favour of the candidate who was leading after first preferences. Perhaps it would be better to say it was decided by preferences if the preference distribution is significantly different to the 1st preference proportions.

COMMENT: You want somewhere in between the two clear definitions, which is seats that went to preferences, or seats changed by preferences. You need to come up with a set of criteria then.

- Scott D September 11, 2015 at 05:50 PM

My point precisely. Therefore, every voter must vote for every candidate. The only way not to vote for those you oppose is to abstain. The voting and counting system is a farce and cannot reflect the wishes of the electorate. It is not a democratic system when you are forced to vote against your wishes.

- Tony Sadgrove September 11, 2015 at 05:55 PM

Hi Antony,

If the Liberals do badly next week, and the Liberal caucus use that as a trigger to roll Tony Abbott, would that be a first? Has another PM lost office as a result of a bad by-election?

COMMENT: In this post (http://blogs.abc.net.au/antonygreen/2015/02/how-australian-prime-ministers-left-office.html) I list the methods of departure for Australian Prime Ministers. Six were deposed by their party. None related to by-elections.

If deposed, it would be because of concern about the polls and about the direction of the government. The Canning by-election would be the catalyst rather than the cause of a challenge.

- Damian Douglas-Meyer September 12, 2015 at 09:13 PM

Antony, What effect do you think the uninterested will have on the election, given that around 3-4 percent are chronically apathetic and that 16 percent really don't care (AES)

I am writing a PhD on interest in politics

Greg Kramer

COMMENT: One of the mysteries to me is that people claiming a disinterest in politics find the time and effort to fill in the long and involved AES survey. I presume the disinterested vote overwhelmingly on habit, but really the data on this is very patchy. The AES data is one of the few sources, but you always have to ask whether such a long survey is good at catching the views of the disinterested. Does its form introduce odd biases.

- greg Kramer September 13, 2015 at 02:42 PM

Meandering away for a moment from the result in this electorate, I would be interested in your views as to the relevance or otherwise of the first preference swing (if any / much) to the national debate covering leadership, policies and of course 'image'. Is a large swing - beyond the norm for a by-election - anticipated in this electorate, or perhaps just being talked-up by media? Thank you for your as always insightful comments.

COMMENT: The polls are predicting a swing of 10% which is large compared to past by-election results. I go through the last three decades of swings at by-elections in this post Background on Swings at Federal By-elections

- Lois September 13, 2015 at 07:22 PM

A bit off the track Antony, but there is talk about Turnbull challenging for the liberal leadership this week before the Canning bi-election.

So my question is if Abbott goes to the GG and ask for and gets a DD called this week, does the bi-election still go ahead or would it be cancelled?

COMMENT: Whether it be a double dissolution or a House only election, it involves a dissolution of the House and the issue of a writ naming polling day for all House electorates. The Canning by-election writ would be withdrawn and the by-election would not go ahead.

- John S September 14, 2015 at 08:02 AM

Antony, there is rumblings of a Double Dissolution at the moment - but does the Abbot government have the appropriate trigger yet? Last rumblings there was the issue that nothing had yet been rejected twice by the same Senate, as the members had changed in 1 July 2014, so anything rejected prior to that was under the old senate, and thus the Gov General would be unlikely to allow a DD.

Is there now any legislation that has been rejected twice by the current Senate? I seem to recall the Higher Education changes might count, but I am unsure.

Additionally, I've always been curious - if the proposed legislation has been altered significantly can it count as a trigger any longer?

COMMENT: In August 2015 the Fair Work (Registered Organisations) Amendment Bill 2014 met the requirements of Section 57 and is a valid trigger to request a double dissolution.

On your second question, that would require a legal dissertation on the meaning of the second paragraph of Section 57.

- Connell Wood September 14, 2015 at 01:44 PM

Antony, if I may ask, as early voters in Canning cast their votes prior to the change in Liberal leadership, when Abbot's personal unpopularity was still in play, how many 'grains of salt' does that add to the use of Canning as an indicator for the electoral chances of Labor and the Liberals? Or is early voting in Canning on too small a scale to matter in the byelection?

COMMENT: Not enough votes to make a difference.

- James September 19, 2015 at 11:53 AM

Re the donkey vote: I live in the ACT, and years ago I lived in Tasmania. As you will be aware, these 2 jurisdictions use the Robson rotation, effectively eliminating the risk of an election outcome being influenced by votes straight down the card. Do you think there would be advantages in using the system nation-wide, and does it have any drawbacks? Also, how (if at all) does it affect preference allocations in those 2 jurisdictions?

COMMENT: However both states fix the order of columns, which is the equivalent to the ballot paper order in other states.

- Alvin Hopper September 19, 2015 at 09:27 PM

And to return to the original topic of this post, Antony, the close-of-the-night count in Canning shows that after distribition of the preferences of the 13000-odd votes cast for Greens and minors, the ALP only caught up by about 500 votes. Sure ain't Queensland!

- Brisbane Jack September 20, 2015 at 03:50 PM

Hi Antony

With the change in Prime Minister in the week leading up to the by-election last weekend, I would be keen to know what the pre-poll votes say about whether there would have been a different outcome had there been no change in leader.

Thanks!

COMMENT: The pre-poll centres in Mandurah and Armadale displayed swings roughly in line with those that occurred in the nearby polling places. Postal votes were more conservative but it is impossible to know when they were cast. I can't see anything in the data that says very much.

- Martin September 22, 2015 at 01:36 PM