NSW Labor preselection ballot reveals unusual electoral system


The NSW Labor conference was held over the weekend, and as part of the conference there was a ballot to choose the ticket for the NSW Legislative Council at next March’s election.

Before this result, I didn’t really know how this sort of thing worked. I knew that usually the two factions (left and right) would alternate seats on the ticket, with the right having slightly more, but how it worked wasn’t clear to me. But a split within the factions saw Cameron Murphy, who had unsuccessfully sought a spot on the left faction’s ticket, preselected to a surprisingly high position.

The vote counts reveal something interesting about how the ballot is conducted, and I’m going to compare it to how this sort of thing is conducted in the Greens NSW.

There’s not much dispute amongst those who conduct ballots about the best practice when you’re electing a group and you want it to be representative – you use a single transferable vote ballot. But what do you do if you want to rank the winners? There are two ways you can go. Either you use the order that those candidates are elected, or you conduct multiple counts, electing a different number of seats in each count. The former is much simpler, but it can cause some serious disproportionality along the way.

So the winning order of candidates was:

  1. Courtney Houssos (right)
  2. Rose Jackson (left)
  3. Cameron Murphy (left – sort of)
  4. Emily Suvaal (right)
  5. Stephen Lawrence (right)
  6. John Graham (left)
  7. Khal Asfour (right)
  8. Sarah Kaine (right)

The left’s internal caucus had preselected a ticket of Jackson, Graham and fellow incumbent Mick Veitch, with Murphy unsuccessfully contesting preselection in the left faction.

It appears that the ETU (who have broken from the right faction) and the CFMEU (normally part of the left) both voted for Murphy.

It also sounds like the remaining delegates voted almost entirely for Houssos or Jackson, and then preferencing the remaining candidates in each faction.

I don’t have all the vote totals, but I know that the top three candidates’ votes were:

  • Houssos – 55.3%
  • Jackson – 26.3% (215 votes)
  • Murphy – 14.2% (116 votes)

That leaves another 4.2% split amongst the remaining candidates. The other two left faction candidates polled 23 votes between them, and I think there would’ve been about 11 other votes to split between any remaining candidates including the right faction candidates. I believe there were 2 votes for Shaoquett Moselmane, and 4 for right candidate Kaine, which leaves just 5 votes between Lawrence, Suvaal and Asfour.

So the vote is extremely concentrated between those leading candidates.

For the next stage, you need to understand how orders of election work. You start by electing those with a full quota. Once you’ve distributed their preferences, you elect any who have now reached their quota.

In this situation, the quota would be just over 1/9th of the vote (92 votes, or something like that).

So if there had been two clear factions, with all votes cast for the faction leader and following the ticket, you’d first elect the majority leader in first, the minority leader in second. Then their surplus would elect their second candidates 3rd and 4th, and then (assuming the minority faction has 3 quotas) their third candidates 5th and 6th candidates. And if the quotas split 5-3, the last two seats go to the majority factions. That’s right, left, right, left, right, left, right, right.

But when that order breaks down, it can have some weird outcomes.

In this case, Murphy polled more than the quota, so he is elected before the surplus from Houssos and Jackson flows on to their allies, which then fill most of the remaining seats without too much trouble.

Except the quota is set as a bar to qualify for the group of 8, not a smaller group. If they were only preselecting three people, the quota would be 25%, and Murphy would not have won the third seat. It would have gone to the second right candidate, Suvaal.

I’m also aware that a similar system is used for Tasmanian Labor to preselect their Senate tickets, and may well be used by the ALP elsewhere.

There is actually an alternative method of conducting this sort of count that creates proportionality throughout the ticket, and would work whether the vote is concentrated on factional tickets, or spread out in a loose pattern.

The Greens NSW hold full statewide ballots of members to preselect candidates for the Senate or the Legislative Council.

Voters number preferences on a ballot, and a number of counts are conducted. The first position is decided by a single-member election. A second count is conducted with a 33.3% quota, and so forth.

The Greens have only ever been able to win a single Senate ballot, so in one sense they could use a simpler ballot to elect a single candidate. In the case of the upper house, there is a variable level of winnability. The first seat is safe, the second seat is reasonably safe but could conceivably be lost. The third seat is usually not winnable, but they have won that seat once. There’s also an added complexity because they apply affirmative action at the end of the count, so if men are selected for both the first and second position, the second-placed candidate is demoted to third, but I won’t get into that any further.

You can see an example of a count using this process here. It’s the preselection of the Greens ticket for the 2007 state election, which was won by Lee Rhiannon, with John Kaye defeating Ben Oquist for the second seat.

This sort of sequential system is a lot more resilient and can work in a variety of situations.

Whereas using an order of exclusion is vulnerable to manipulation. If you just need a full quota for the top 8 spots to get advance to one of the top positions, there might be an advantage to one faction finding a different way to manage their votes.

I estimated what I thought the vote was for the right, left, and Murphy, and calculated quotas based on a magnitude of 1 to 8. Depending on how strongly preferences flow from the left to Murphy, he would’ve won between the 4th and 6th spot on the ticket.

Factions % 1 2 3 4 5 6 7 8
Right 56.4 1.098 1.689 2.249 2.811 3.365 3.907 4.476 5.011
Left 29.1 0.567 0.872 1.161 1.451 1.737 2.017 2.311 2.587
Murphy 14.2 0.276 0.425 0.566 0.707 0.847 0.983 1.126 1.261

It’s possible, indeed likely, that none of this will matter. Labor won 7 seats in 2019, and their vote appears to have grown since then. Any of the reshuffling would’ve likely affected the top six. But the ALP system relies on accurately forecasting the number of seats won – if there is a big shift in public support or a polling failure, this sort of thing could make a difference.

There are also some imperfections with the sequential count. Specifically the “Alabama paradox“, where an increase in the number of seats available reduces a state’s apportionment. That could happen in situations where a large number of candidates had a similar vote. I encountered such a situation during a preselection count for a local council a decade ago. A Greens local group was deciding the order of candidates, with candidates at the top getting first pick as to which ward they’d contest. The candidate who won the first count didn’t win on the second count. The second count tie was split in favour of the candidate who was elected first, and the candidate elected second on the count didn’t end up winning until fourth. But that is unusual.

Anyway, I thought it was an interesting opportunity to shed light on two different ways to take a single transferable vote proportional ballot and apply it to rankings, which becomes more important in party list systems like the NSW upper house, or the new Western Austrailan upper house.

Sidenote: a similar issue plays out when deciding which senators get long terms after a double dissolution. Federal law requires that the AEC conduct special counts of the Senate ballots with just half as many seats, which could then be used to decide which of the twelve senators get full terms. Yet there is an alternative approach which involves giving the long terms to the six senators who were elected first. This tends to favour the major parties since they tend to reach full quotas. In 2016, the recount method would have given long terms to Lee Rhiannon of the Greens and independent Derryn Hinch – both candidates who did not poll a full double dissolution quota but would have been elected if there were only six seats up for election. Instead the order of election method gave those seats to major party senators. And that is the method the Senate chose to implement, as they had done in the past.

Liked it? Take a second to support the Tally Room on Patreon!
Become a patron at Patreon!


  1. Hmmm. Been there, done that

    It can get complicated and challenges have arisen over disputed pre-selection results in the past. One such challenge ended up in a Supreme Court. Trying to explain the system to the Barristers is all but impossible – they are not very numeric. Neither are the judges sometimes … it gets too hard and the status quo is sometimes the easy way out.

  2. The result is due to the complex sub-factional arrangements within the left faction of the Labor party. Murphy is from the soft left (which gets its support from the CFMEU) whilst the other three left candidates were from the hard left subfaction. The CFMEU had some delegates excluded from the left conference, which led to the hard left being able to win all three seats at the preselection. The result from the Labor conference is what would have eventuated from the left ticket had the CFMEU delegates not been excluded from the left conference in the first place.

  3. All over now.. the effort the hard left put into this became pointless they had 2 quotas not 3.. Cameron Murphy got 116 votes when the quota was 90. There are issues as to why there was no sharing of positions and why one of 3 large left wing unions were excluded from left decision making. The result of this ballot will hopefully rectify those 2 problems

Comments are closed.