The McGowan government in Western Australia has appointed a Ministerial Expert Committee to recommend changes to the electoral system for the state’s Legislative Council. (You can find the Committee’s website here.)
The Committee has a number of issues to examine. Some are controversial, such as whether to change the state’s zonal electoral system. I wrote on the zonal electoral system and its unequal enrolments two weeks ago.
The proposal that has attracted least criticism is the abolition of group voting tickets (GVTs). GVTs were first introduced for Senate elections in 1984. They were introduced as a solution to a chronic high rate of informal voting and designed to make voting easier while retaining full preferential voting.
What has not been fully appreciated is that the tickets sped up voting and also simplified the counting process. GVTs meant that less than 10% of ballot papers needed to be examined for formality and re-examined for preferences during the count. The rest of the ballot papers were ticket votes, and all ticket votes for a party being the same, could be treated as block votes.
These benefits have since been outweighed by the manipulation of results produced by GVTs giving parties almost total control over between-party preferences.
For major parties, GVTs strengthened the strong flow of preferences that parties had previously achieved through influencing voters with how-to-votes. But GVTs gave the same power to small parties that previously struggled to influence preferences due to lack of members distributing how-to-votes. Even the smallest micro-parties that didn’t bother to campaign suddenly had total control over their preferences. Over time, as participants learnt how to use GVTs strategically, the system began to elect candidate from parties with tiny votes who would never been elected had voters controlled preferences.
Three jurisdictions have now abolished GVTs. In this process, great attention was paid to ensuring voters did not have to revert to the pre-1984 situation of completing vast numbers of preferences. But a price of abolishing GVTs has been to make counting more complex. It has required a switch to scanning rather than hand counting and data entering ballot papers. Complexity has also been increased by changes to formulas calculating transfer values for surplus to quota preferences.
As the Ministerial Expert Committee searches for a replacement Legislative Council electoral system, it has the advantage of being able to draw on experience with abolishing GVTs for elections to the NSW and South Australian Legislative Councils and the Commonwealth Senate.
Two models for electing the WA Legislative Council are being discussed. One retains regions, a four region model with each electing nine members the most discussed. The second is a switch to a single state-wide model.
The state-wide model in particular requires careful design. Careful thought needs to be given to ballot paper design, voter instructions, and the counting method.
Without careful design, using a single electorate to elect the WA Legislative Council could end up producing a ballot paper that is unprintable or uncountable.
The Case Against GVTs
The victory in March of the Daylight Saving Party’s Wilson Tucker in Mining and Pastoral Region has been the catalyst for change. As I explained in a post last week, Tucker polled just 98 votes or 0.2% of the vote, GVTs electing him by delivering preferences from 12 other groups. Ten of those groups polled more votes than Mr Tucker, his votes making up just 1.3% of his vote on passing the quota for election. The National Party polled 50 times as many votes as Mr Tucker but was beaten by him in the GVT engineered race for the final seat.
Using GVTs to elect candidates from parties with very low votes has become known as ‘preference harvesting’. If a range of small parties agree to swap preferences amongst each other, then GVTs allow the parties to stack their votes on top of each other. The luck of a good ballot draw, the choice of an attractive party name, or a name that can be confused with another party, adds to the chances of an individual party being elected, and also adds to the pool of preferences shared amongst the participating preference harvesters.
In a state with regions, such as Western Australia, parties are also able to arrange preference deals across regions. The so-called ‘micro party alliance’ at the 2021 WA election split the regions between the participating parties. In Mining and Pastoral Region, the Daylight Saving Party was the beneficiary of all other micro-party preferences, and returned the favour by directing preferences to the designated micro-party in the other five regions.
GVTs were introduced for Senate elections in 1984 as a solution to the then scandalous level of informal voting. The system was subsequently copied by the states that elected or introduced proportional representation to elect their Legislative Councils.
The table below sets out the timeline for introducing proportional representation, introducing GVTs, and the more recent abolition of GVTs.
Introduction of Proportional Representation and the Rise and Decline of Group Voting Tickets
|New South Wales||1978||1988||2003|
Note – Years indicate first election under changed system. Tasmania elects its Legislative Council from single member electorates, and Queensland, the Northern Territory and ACT do not have upper houses.
That GVTs can be manipulated has been known since the 1999 NSW election, the infamous tablecloth ballot paper election that led to the abolition of GVTs in NSW. Similar manipulation of preferences led to the abolition of GVTs in South Australia and the Commonwealth.
The three jurisdictions that have abandoned GVTS adopted a similar solution. The divided ballot paper was retained, but GVTs were abolished, ending party control over between party preferences. Voters could indicate preferences between parties using an ‘above-the-line’ (ATL) vote with preferences, or choose between candidates ‘below-the-line’ (BTL) using optional preferential voting beyond a minimum number of preferences.
All three jurisdictions ended party control over between-party preferences, but the retention of ATL voting left parties largely in control of within-party preferences. Even there, the easier method of BTL voting allowed voters an option to interfere with a party’s ordering of candidates. The Tasmanian Senate race in 2016, the first under the new Senate system, saw Labor’s Lisa Singh re-elected despite having been demoted to sixth on the party ticket. Singh’s election would have been impossible under GVTs.
But there is one important difference between the approach adopted for Senate elections and the method used in NSW and South Australia. It is a small difference in the instructions for voters to complete their ballot papers, and it has major implications for the conduct of the count.
That difference concerns ballot paper instructions. For NSW and SA Legislative Council elections, the ATL instructions are that a voter should number one square, and if they wish, number further squares. Senate ballot paper instructions state that a minimum six preferences must be shown with preferences beyond six optional, though a strong savings provision ensures that any vote with a single-ATL vote that would have been valid under the old system remained formal under the new.
This difference on the number of preferences sounds small, but it has major implications for the process of counting. For Senate elections, apart from an election night tally of first preferences, hand counting has had to be replaced by scanning. Hand counting is still used for the majority of ballot papers in NSW and South Australia.
Proportionality versus Preferences
Australia is one of the very few countries in the world to use preferences as part of proportional representation. The system is known generically as proportional representation by single transferrable vote (PR-STV).
No other country uses such large electoral districts for PR-STV, or elects so many members per district. As I understand it, the NSW Legislative Council is the largest PR-STV electorate in the world, and in electing 21 members is also the largest in number of members elected. Everywhere else in the world would use some form of list proportional representation to elect 21 members, with the only debate being whether choice of candidate in party lists would be open or closed.
So the Australian experience of using PR-STV in large electorates is unique and largely outside the normal literature on electoral systems. Usually PR-STV is put forward as an alternative to single member electorates, and usually in districts of five or fewer members. Most countries that already use proportional representation elect many more members per district, so the non-proportional nature of preferences and the extra cost of counting mean that PR-STV is not considered.
Electoral system textbooks (outside of Australia) treat PR-STV as a peculiar sub-type of proportional representation. Electoral system textbooks in Australia treat PR-STV as almost the only form of proportional representation.
From my experience in working with Australia’s PR-STV systems, there is a trade off to be considered between district magnitude (number of members elected) and the minimum number of preferences that are required to be completed.
My observations are –
- The move from full to optional preferential voting means that non-GVT systems give more weight to first preferences and less weight to further preferences in determining proportionality and who is elected.
- The fewer the number of members to be elected, the harder it is to achieve proportionality on first preferences. In this case there is benefit in increasing the number of ballot papers with preferences to help in filling the final vacancies.
- Conversely, the more members you elect, the more that the outcome will be proportional to first preference votes, so preferences can tend to downplay the proportionality of the result.
- As the proportion of ballot papers with ATL preferences increases, the harder it is to rely on hand counting and the more that ballot papers need to be scanned. As outlined below, NSW and SA still make extensive use of hand counting, where Senate counts have moved to 100% scanning with all the additional costs.
Let me explain these points by looking at the experience of abandoning GVTs for NSW, South Australia and .Commonwealth upper house elections before moving on the the implications for Western Australia.
NSW and the infamous ‘tablecloth’ ballot paper
NSW elects half of its Legislative Council every four years from a state-wide electorate. 21 members are elected at each election producing a quota of 4.55%, though the final vacancies can be filled with fewer votes. The low quota was already attractive to minor parties before GVTs added another incentive to the mix.
GVTs and loose party registration rules collided at the 1999 NSW Legislative Council election to produce the infamous ‘tablecloth’ ballot paper (See picture below). 264 candidates nominated across 80 groups and one ungrouped column. Above and below the line voting options had to be triple decked to accommodate the nominations. The Electoral Commission was required to widen voting partitions, hire stronger fork lifts, use larger planes to deliver ballot papers to distant areas, widen slots on ballot boxes and put stronger glue on declaration envelopes. Candidates were elected from as low as 0.2% of the vote.
Group voting ticket were abolished before the next NSW election in 2003. A new form of above the line (ATL) voting was introduced allowing voters to number preferences in the group voting squares above the line. Only a single ATL preference was required for a formal vote. NSW already had a form of optional preferential voting below the line with only 15 preferences required. NSW has also used optional preferential voting for lower house elections since 1980.
One of the problems with reforming the NSW Legislative Council ballot paper goes back to the way that the Wran government replaced indirect election of the Legislative Council with public election in 1978. At the time the Labor Party was determined to avoid the experience of the 1974 NSW Senate election when the ballot paper was stacked with candidates, 73 in all, to drive up the informal vote at Labor’s expense. There were no party names on the ballot paper in 1974 and all 73 preferences were required, slowing down both voting and counting. (See ballot paper example below taken from a Labor how-to-vote)
For NSW, Labor chose to adopt the then in use Senate ballot paper and counting system but with optional preferential voting. The first Legislative Council election was held in 1978. Piecemeal changes then followed, The split ballot paper and group voting tickets were introduced in 1988 but without party names. Party names were introduced at the 1991 election. Fifteen members were elected until 1991 with MLC terms lasting three lower house terms. A referendum in 1991 removed three MLCs and resulted in half of the Council facing the electorate at each state election. The first election for 21 members took place in 1995. Only 10 BTL preferences were required for a formal vote until 1991, 15 since 1995.
With the exception of the split ballot paper, most of the detail of the electoral system is entrenched in Schedule Six of the NSW Constitution, including a minimum number of preferences. Schedule Six can only be amended by referendum. The reforms introduced after the 1999 election had to work around the entrenched provisions, so parties were required to stand 15 candidates if they wanted an ATL voting square. This has produced the ballot paper pictured below that in my opinion is the worst in Australia. I describe it as looking like a mutant lotto form. It is confronting to voters, especially those with poor skills in written English. It has a very high rate of informal voting, the largest category of which are blank ballots.
If Western Australia were to adopt a single state wide electorate for its Legislative Council, the above ballot paper is a warning of what not to do. Admittedly, WA does not have the minimum 15 candidates restriction, but electing the 36 member Legislative Council would see the major parties nominate more than 20 candidates each. A ballot paper that lists the candidates risks producing a monster like the 1999 tablecloth ballot paper. Perhaps there could be two ballot papers, one listing only groups and the other listing candidates, but this has its own logistical problems.
For all the romance attached to the Hare-Clark variant of PR-STV used in Tasmania and the ACT, Hare-Clark has a fundamental problem with scale. A proportional system based on votes for and choice of candidate falls apart when the overwhelming majority of voters know the party but not the candidates. The problem is even worse when the election is for a second chamber at an election on the same day as the lower house race for government.
And then there is the counting problem. The chart below shows the three categories of ballot papers at NSW Legislative Council elections since the divided ballot paper was introduced in 1988. ATL votes with preferences were not allowed until 2003, but data on ATL votes with preferences is not available for the 2003 and 2007 elections. Preference distributions at the 2003 and 2007 elections suggest the rate of single-1 ATL votes was around the 80% level shown for 2011 and 2015.
The high rate of exhausted preferences means that the final vacancies have generally been filled by candidates short of a quota, even after preferences. At each election since 2003, 17 vacancies have effectively filled on first preferences, then final four vacancies filled by the highest polling candidates at the end of the preference distribution. At the last three elections, 20 of the 21 vacancies could be picked from the tally of first preference votes, with only the identity of the final candidate elected being changed by preferences. That is a lot of preferences counted to identify one final candidate.
Under the new system after 2003, 80% of ballot papers could still be counted by hand. Only one-in-five votes had to be data entered for counting. The introduction of the new Senate system in 2016 clearly influenced how voters completed their ballot papers at the 2019 state election with a near doubling in the rate of ATL votes with preferences.
NSW still counted more than two-thirds of ballot papers by hand in 2019, but the doubling in data entry load was significant. It is proposed to introduce scanning for the 2023 election, though this will have some logistical difficulties. The current NSW Legislative Council ballot paper is deeper the A4 (294mm) standard depth for scanning and may require some tweaking of its structure. Another proposal to add party logos to the ballot paper would, in my opinion, make it extremely difficult to scan the ballot papers.
On a side note, it may be time that NSW looked at a referendum to amend the archaic parts of Schedule 6 that are getting in the way of voting and counting. The minimum 15 preferences required has been overtaken by the split ballot paper, and the current random sampling method mandated for distributing preferences should be replaced, as it has been for local government. While broader reforms of the Council may not get bi-partisan agreement, change that produces a sensible ballot paper that can be understood and counted should be able to attract all party agreement.
The problem spreads to the Senate
At the 2004 Victorian Senate election, Steve Fielding of Family First polled 2% but leap frogged numerous candidates to eventually beat the higher polling Green candidate in the race for the final seat. Fielding won thanks to a preference swap with Labor, but within a decade, parties were winning election from even lower votes without the need for major party preferences.
A profusion of so-called ‘micro’ parties entered the fray. By the 2013 election, the Australian Electoral Commission was forced to shrink the font size on Senate ballot papers and issue magnifying sheets so voters could read the ballot paper. (See example right) Voters were increasingly finding it difficult to locate the parties they knew with as many as 50 parties appearing on some state’s Senate ballot papers.
Under Commonwealth electoral law, the Australian Electoral Commission (AEC) must print all candidates and groups on a single row. The AEC cannot split candidates across multiple rows, and with a fixed width one metre ballot paper determined by printing technology, the only solution to printing the ballot paper was to shrink the font, resulting in the AEC making magnifying sheets available in voting partitions.
Which created the farcical situation where voters were forced to manipulate a ballot paper one-metre long in a 600mm voting partition while wielding a magnifying sheet to read the names on the ballot paper, searching amongst a bewildering array of candidates and parties voters had never heard of in the search for the candidates and parties they knew. With a below-the-line vote requiring a complete sequence of preferences, the physical difficulties created by the farcical ballot paper herded voters into voting above the line for controlled party tickets, exactly the aim of those engaging in preference harvesting.
At the 2013 NSW Senate election, the little known Liberal Democrats shocked by polling 9.3% with the advantage of drawing column A on the ballot paper. A significant number of voters selected the Liberal Democrats out of confusion on the giant ballot paper. With column widths reduced, the name Liberal Democrats was split across two lines giving greater prominence to the word ‘Liberal’. Many voters reading left to right looking for the Liberal Party were clearly confused by the similar looking name.
That the Liberal Democrats poll higher when in a column to the left of the Liberal Party has been seen at several elections since, including in South Metropolitan Region at the 2017 Western Australian election.
The 2013 election also saw Ricky Muir of the Australian Motoring Enthusiasts Party elected from 0.5% in Victoria, and Wayne Dropulich of the Australian Sports Party elected from just 0.2% in Western Australia. (Dropulich’s victory was later cancelled after lost ballot papers resulted in the WA Senate election having to be re-run.)
After an inquiry, the Commonwealth Parliament’s Joint Standing Committee on Electoral Matters recommended adopting the NSW system, abolishing GVTs, allowing optional preferential ATL preferences, and allowing optional preferential voting with a minimum number below the line. Labor, the Coalition and Greens backed the proposal.
By the time it came to legislate the recommendations, Labor had withdrawn its support. Forced to negotiate with the crossbench to pass the legislation, the Coalition agreed that the ballot paper instructions would direct to complete six ATL preferences, with the savings provision that at least a valid first preference ATL vote would count.
The new system was used for the first time at the 2016 double dissolution and again at the 2019 half-Senate election. It ended the election of parties from very low initial quotas, and the 2019 election provided some evidence of fewer parties contesting election.
(For more information on how the new Senate system worked in 2019, see my blog post comparing half-Senate elections under the old Senate system in 2013 and the new system in 2019.)
The rate of ATL voting at Senate elections rose from the system’s introduction in 1984 to its last use at a national election in 2013. The rise corresponded with an increase in the number of groups nominating. But the new system used in 2016 and 2019 has created a counting problem much greater than in NSW, as shown by the chart below.
The AEC has had to switch to full scanning of ballot papers. Where previously 95% of ballot papers were 1-only ATL vote, now 95% have preferences. Apart from the initial tally of first preferences, the AEC can no longer count the votes by hand. All ballot papers must be scanned.
South Australia Follows
Like NSW, South Australia elects half of the Legislative Council every four years from a single state-wide electorate. In South Australia’s case, 11 members are elected with a quota of 8.33%. GVTs were introduced as part of the electoral system when PR-STV was introduced in 1985. In the early 2000s there was an explosion of nominations, made worse by Independents in South Australia being able to nominate and create their own pseudo-party name on the ballot paper.
In response to the 2013 Senate election, South Australia made several administrative changes to nomination. There was a huge increase in the nomination deposit, and registered parties were given priority in the allocation of ballot paper columns.
Like NSW, South Australia is permitted to split groups over several rows on the ballot paper. South Australia avoided the single-row 1-metre ballot paper, instead choosing a 600mm width so that the ballot paper fitted in the polling partition. It produced ballot papers like the one below from 2010.
As will be discussed below, while the above ballot paper is designed to make voting physically easier, it has a major problem when it comes to counting. The above ballot paper is to tall to scan. It can only be hand counted. (See discussion of proposed ballot paper below.)
Ahead of the 2018 state election, the Weatherill Labor government adopted the NSW optional preferential model. GVTs were abolished, only one ATL preference was required for a formal vote with further preferences optional. A minimum 12 preferences were required below the line, relaxing full preferential voting rules. Nomination deposits were restored to previous levels and the Senate’s Ungrouped column was adopted for Independent candidates. There were only 13 groups on the ballot paper creating a single row of candidates on a manageable ballot paper.
The chart below shows how voters responded to the new system.
As with NSW, the 1-vote with further preferences optional created a similar pattern in voters choosing not to give preferences. More than half of ballot papers could be counted by hand. But now 40% rather than 5% of ballot papers needed to be data entered to conduct the count. As was the case with the AEC, the SA Electoral Commission (ECSA) moved to scan the ballot papers.
Which means the ballot paper must be designed to cope with the physical demands of scanning. All of the high speed scanning equipment available in Australia has an A4 height (294mm) standard. If ECSA wants to scan ballot papers, it can’t do the layering it has used in the past. ECSA has come up with the design shown below for a double decked ballot paper that can cope with up to 46 groups.
Legislation will have to be changed to prevent parties nominating more than eight candidates for the 11 vacancies. If parties were permitted to nominate more than 8 candidates, then the double decked ballot paper would be too high to scan. If there are more than 46 groups then the ballot papers can’t be scanned.
For all the democratic gain in ending party control over preferences, the consequence has been to make the count more complex and pushed Electoral Commissions into scanning ballot papers. But scanning has physical restrictions if it is to be undertaken at the speed required to complete a count in reasonable time. Larger flatbed scanner are available, usually used to scan maps and plans. But they do not exist in the numbers, or operate at the speed, required to conduct an election count.
It also has to be remembered that scanning ballot papers does not operate like scanning in an office. Voting scanners have to be set up and aligned to the boxes on the ballot paper. The optical character recognition is designed to read what’s in the squares and nothing else. If what’s in a square can’t be recognised, human intervention is required.
As a result, aligning the ballot papers to enable the scanning is one of the major tasks of counting staff. And if you’ve ever tried to bulk scan a series of documents that have been folded in strange ways, you’ll know some of the mechanical challenges required before a ballot paper even gets into a scanner.
The Lessons for Western Australia
As mentioned at the start of this post, there seem to be two models in circulation for a reformed Legislative Council.
- State-wide election as in NSW and South Australia. Both states have staggered terms with half of the MLCs elected every four years. Western Australia currently elects the entire council of 36 members every four years. Whether state-wide election would be for the entire Council or the Council would be split back into staggered terms is not clear.
- Retain a region based system, the most commonly discussed proposal being four regions of nine members.
A Constitutional requirement in WA is that the Legislative Council cannot be reduced in size without a referendum. So five 7-member or seven 5-member regions are not options without resorting to a referendum.
My discussion on abolishing GVTs in NSW, South Australia and for Senate elections points me to making the following observations on whether to retain Legislative Council regions in Western Australia or move to a state-wide model.
The following observations can be made about abolishing GVTs while retaining regions as a basis for election. Most are positive observations.
- The current ballot paper could be retained with the advantage that voters are familiar with it from past state and federal elections .
- Regions have a higher quota for election, which combined with the end of GVTs, could discourage nominations from parties and candidates with little chance of election.
- A regional system could allow a measure of over-representation for areas outside of Perth and the state’s south-west. Presumably this would be reduced from the current 3-1 weighting against Perth.
- With proper thought given to ballot paper layout, the increased number of ballot papers with preferences could be scanned as part of counting.
- A formula needs to be devised to align Council regions with Assembly districts.
A state-wide electorate would produce a fully proportional result but has a number of drawbacks.
- A state-wide electorate is strict one-vote one-value without even the large district allowance that permits remote lower house electorates to have lower enrolment.
- The quota for election would be 2.7% for a 36 member election, or 5.3% for a 18 member half-Council election. That is a very low quota and would encourage many parties to run, hoping to benefit from a lucky ballot draw and the low quota.
- If the low quota encourages more nominations, problems of controlling ballot paper size arise. It may require additional measures concerning minimum party membership, minimum nomination numbers and higher nomination deposits to control the size of the ballot paper.
- The experience in NSW is that a high rate of exhausted preferences means that the final vacancies can be filled by parties with less than half of a quota even after preferences.
- How can you have a ballot paper that list candidate names for a huge state wide electorate, given physical limitations on printing and scanning ballot papers? Will there be a party and a candidate version of the ballot paper?
- No other country in the world would attempt to elect 36 members by PR-STV. There are numerous other forms of proportional representation that could be used, but all would probably require the abolition of candidate choice within party tickets.
- Using any electoral system other than PR-STV risks having a ballot paper with which voters are unfamiliar.
In my view, the abolition of GVTs will require careful consideration of the following issues –
- Minimum requirements for party registration and candidate nominations.
- Consideration of increases in nomination deposits.
- As noted concerning South Australia, limitations on the number of candidates per groups to control ballot paper size.
- Consideration of how many preferences should be required for an ATL vote. In my view, the NSW and SA options of a 1 with preferences is the better option.