As mentioned in my previous post, New South Wales is the only Australian jurisdiction that uses optional rather than compulsory or full preferential voting for single member lower house elections. It is also the only state that data enters all lower house ballot papers and publishes the data for outside research.
Optional preferential voting (OPV) means that voters do not have to number a preference for every candidate on the ballot paper. A ballot paper requires only a first preference to be formal. All further preferences are optional.
If during the distribution of preferences, a ballot paper for distribution has no preference for a candidate remaining in the count, the the ballot paper is put aside as having “exhausted” its preferences. Exhausted ballot papers have no direct involvement in determining the winning candidate, but they have an indirect role in altering the winning post a candidate must pass to win. As explained below, OPV works in favour of leading candidates over trailing candidates.
This can be explained by comparing the maths of full versus optional preferential voting.
- Under full preferential voting, the winning candidate must achieve 50% of the formal vote after the distribution of preferences. The winning post of the votes needed to win is set at the start of the count and does not change.
- Under optional preferential voting, the winning candidate must receive a majority of the votes remaining in the count, that is the formal vote minus exhausted preferences. The winning post is lowered with each exhausted preference making it easier for the leading candidate to win by making it harder for the second placed candidate to catch and pass the leader.
The chart below shows the two-party preferred preference flows for minor parties and Independents at the 2019 NSW election.
I'll explain the political implications of the above graph inside the post. But if you want more detail of the preferences flows by electorate by candidate, or by party by electorate, you can find all the details in this pdf document I've prepared.
The Political Impact of OPV
As I explained above, OPV alters the winning equation because exhausted preferences lower the number of votes remaining in the count, in effect lowering the winning post. A candidate under OPV needs 50%+1 of the (formal vote - exhausted votes), not 50% of the formal vote as is the case with full preferential voting.
If you think of it in percentage terms, exhausted preferences re-weight the percentages, and the weighting favours the candidate with most votes. Let me explain by example.
Say we have a contest involving 1,000 votes. Candidate (A) has 450 votes, (B) has 350 and (C) has 200.
Say (C)'s preferences split 50:50. That's 100 preference transfers to both (A) and (B). (A) finishes with 550 votes or 55%, (B) with 450 or 45%.
Lets' repeat that exercise under OPV. If (C)'s preference still split evenly, BUT 50% of (C)'s preferences exhaust, then the transfers are 50 to (A), 50 to (B) and 100 set aside as exhausted. (A) finishes with 500 votes and (B) with 400 and 100 set aside as exhausted. The winning margin in votes is the same for both systems with this example.
There are now only 900 votes in the contest because of the exhausted preferences. With 900 votes as the divisor, (A)'s final percentage is 55.6%, (B)'s 44.4%. (A)'s percentage has been boosted from 55.0% to 55.6% by the exhausted preferences. (A mathematical brain fade had me originally saying 61% but the correct figure is 55.6%.) Exhausted preferences lower the votes in the count, lower the winning post, and boost the percentage of the winning candidate with the re-weighting of percentages caused by exhaustion.
Under full preferential voting, (B) needed 151 of (C)'s 200 votes to pass (A), but exhaustion made it impossible for (B) to catch and pass (A). Compared to Federal elections, OPV at NSW elections means far fewer seats result in a second placed candidates winning.
There have been two clear consequences of OPV on NSW elections.
- Three-cornered contests between competing Liberal and National Party candidates have disappeared at state elections. There has not been one since 1999 when preference exhaustion delivered Labor victory in the north coast seat of Clarence and made Monaro and Burrinjuck more marginal for the Coalition than they should have been.
- Labor struggles to win state seats from second place on Green preferences, something that occurs more often in NSW at Federal elections.
NSW will see two three-cornered contests at the 2023 election. In Port Macquarie, the defection of sitting MP Leslie Williams from the National to the Liberal Party has seen the National Party insist on its right to contest the seat. Labor's vote in the seat is so low that the seat should not be lost the Coalition. The only danger would be Labor slipping to third place and using its preferences to determine whether the Liberals and Nationals win. That last occurred at the 1996 Southern Highlands by-election when the second placed Liberal candidate passed the National on Labor preferences.
The other three-cornered contest is Wagga Wagga, where Independent Joe McGirr is so safe that the Coalition parties can engage in defining territory for a future contest when McGirr retires.
Green Preferences
As shown in the preference flow chart at the top of this post, Green preferences split 52.5% to Labor, 7.8% to the Coalition, and 39.7% exhausted.
Compare those numbers to contests in NSW at the 2022 Federal election. Under full preferential voting, Green preferences flowed 84.8% to Labor and 15.2% to the Coalition.
That means that Labor missed out on more than 30% of Green preferences. This does not matter in seats where Labor leads on first preferences, but does where Labor is second. Not only does OPV cause Labor to miss out on around 30% of Green preferences, but the 39.7% that exhausted would boost the chances of any Coalition leading on first preferences.
In short, OPV makes it harder for Labor to win electoral contests on Green preferences.
High Exhaustion Rates for Right-of-Centre Parties
One Nation and the Liberal Democrats contested the 2019 state election and last year's Federal election. Preference flows for both parties were very different under OPV compared to full preferential voting.
In 2019 One Nation preferences split 10.9% to Labor, 18.0% to Coalition and 71.1% exhausted. In NSW at the 2022 Federal election, One Nation Preferences split 64.3% to Coalition and 35.7%. Not only does the Coalition miss out on these preferences, but the high rate of One Nation exhaustion makes it tougher for any trailing candidate to win. That's exactly what happened at the first NSW election contested by One Nation in 1999.
I suspect the One Nation's exhaustion rate will hurt the Coalition in marginal seats. It will also boost Labor's winning percentage in safe Labor seats contested by One Nation.
Similar figures show up for the Liberal Democrats. At the 2019 state election the flows were 8.2% to Labor, 23.7% to Coalition and 68.1% exhausted. The 2022 Federal flows in NSW were 71.5% to the Coalition, 28.5% to Labor.
Is there a discernable difference in exhaustion rates in marginal vs safe seats? I can see some logic behind not preferencing a major party if the seat is safe, but making a choice in a marginal seat that may end up being close or deciding government.
COMMENT: The biggest correlate is based on how many preferences were on HTVs, and the more HTVs distributed, the bigger the impact. Hotly contested seats see the most HTVs handed out. Exhaustion rates were higher where One Nation ran, and in seats where Liberal preferences were distributed. It was lower in seats where Labor or the Greens were distributed and had distributed a complete preference list.
I don’t have 2019 HTVs for comparison. I did in 2015 and published a post comparing average preference flows to HTV recommendation. It showed that HTVs did have an impact on the number of preferences completed. (2015 NSW preference data)
Do you have views on how these systems would play out in the UK where the existing party landscape is different? I realise voluntary voting is a spanner in the works
COMMENT: OPV was rejected in the UK at a referendum in 2011. OPV would result in fewer seats being won by candidates with very low votes when support is split 3 or 4 ways. It would make strategic voting less important. For example, it would allow the Greens to attract votes without splitting Labour or Liberal Democrat support in an electorate as Green voters could give preferences. The most important consequence is ensuring the elected MP has a higher base of support from voters who prefer them over another candidate. Unfortunately, this benefit for local representation is constantly overlooked in favour of analysis of the overall impact of House of Commons numbers.
Thanks for an interesting and insightful article. The system moves the goals and changes the field but it seems a democratic innovation. Those extinguished votes reflect voter choices – ‘I’ve had my say and if it isn’t one of my candidates then I refuse to choose the least worst’. NSW has the social benefits of compulsory voting, without the detriment of squeezing preferences from unwilling voters. HTVs are more likely to say ‘stop here’ or ‘Vote 1 only’ than to engage in preference whispering.
Could you please explain the Donkey column.
It may be worth looking at how deep preferencing goes before votes are extinguished (no doubt heavily influenced by HTVs). Do LAB, GRN or IND voters fill more boxes than SFF, LIB, NAT?
COMMENT: The Donkey column has an entry when one of the candidates on the left appeared in ballot position 1. The entry indicates which of the final two candidate benefited from any donkey vote.
I notice a lot of ALP candidates have drawn the number one slot on the ballot paper, including Chris MInns. Any thoughts on how much this is worth in percentage terms? I seem to recall hearing it was about 1%.
COMMENT: About 1% seems the best estimate, and the affect is stronger if the order of parties has changed since the previous election. So if, for example, Labor is at the top two elections in a row, then the impact is built into the margin. If the Labor candidate moves to the top of the paper they are likely to see a bigger swing as the donkey vote from the previous election reverses, and reverse for dropping down the ballot paper.
Where can I find the default preferences for a candidate? The how to vote info I have received just tells me to vote 1 for the candidate with no indication where prefernces will flow. Or have I got it wrong and unless I number all candidates preferences don’t matter?
COMMENT: There are no default preferences. Candidates and parties cannot control your vote.
The only preferences that count are the ones that you write on the ballot paper. If you just vote ‘1’, as is allowed in NSW, that’s the only vote that counts. If you vote for a candidate that gets excluded in the count and has their preferences distributed, your vote ‘exhausts’ as it has no preferences for a candidate remaining in the count. If you vote for a candidate who finishes in the final two and doesn’t have their preferences distributed, then whether your vote has preferences or not does not matter and only your first preference counts.
If you vote for a candidate likely to be excluded and have preferences distributed, it is best to mark further preferences so your vote stays live in the count.
“(A) finishes with 500 votes and (B) with 400 and 100 set aside as exhausted. The winning margin in votes is the same for both systems with this example.
There are now only 900 votes in the contest because of the exhausted preferences. With 900 votes as the divisor, (A)’s final percentage is 61.1%, (B)’s 38.9%. (A)’s percentage has been boosted from 55.0% to 61.1% by the exhausted preferences.”
I hate to correct the legendary Antony Green, but mate, your maths are wrong. the impact of exhaustion is far smaller in percentage terms than stated here.
If (A) finishes with 500 votes, and (B) with 400, then (A) now ends up with 55.56% of the vote, and (B) with 44.44%. The difference is just over one-half of one percent, not 6.1%.
I was looking at that paragraph for a moment, thinking those numbers looked a bit dodgy – given it’s a Saturday morning and I’m only on my second coffee, I had to pull the calculator out, but those are the actual numbers.
COMENT: You’re right. I’ve divided full preferences 550 by 900, not OPV 500 by 900. I though that impact was bigger than I expected and that’s why. Sometimes the obvious error is looking at you in the face.
Is the negative for Labor’s loss of Green preferences significantly more or less than the loss to the Coalition of One Nation, etc preferences? And therefore what chance any future dumping of the optional in NSW’s optional preference voting.
COMMENT: It requires a referendum to abolish optional preferential voting. I would be surprised if any government would put such a referendum and be surprised if it passed if put.