On Elections

How people elect parliaments

Improving the way Australian Senators are elected

April 2014

Since 2013 Australia has seen an extensive debate on the defects in the design of the ballot papers used since 1984 for electing Australian senators.

This debate has not questioned the underlying voting method – the single transferable vote (STV) system.

Nor does the recent debate reopen the issue of the numbers of senators each state elects. The six Australian states vary greatly in population, yet each one elects exactly 6 senators at every ordinary election. Each of the two territories elections 2 senators, regardless of their populations. The result is very serious malapportionment, but that problem is unlikely to be corrected. The state equality rule dates from the establishment of the Commonwealth in 1901, and is firmly entrenched in the Australian Constitution.

The recent debate has instead been about the design of the ballot paper, and the options which that design gives to voters.

There are actually three distinct problems with the current ballot design. These are, in increasing order or importance, the problem of unfairness and distortion caused by having only one variant ballot paper, the problem of having parties dictate the order in which their candidates appear on the ballot paper, and the major problem of the existence of group voting tickets, which allow parties to dictate a voters’ entire preference sequence.

The fourth problem is the use of compulsory preferencing, which is not strictly a ballot paper design issue, but is tied up in how the ballot papers are used and counted..

A fifth issue, even less related to ballot design, is that some commentators have criticised the phenomenon of candidates (or parties) with a low number of first preference votes occasionally winning seats after receiving votes transferred from other voters. There are proposals for new rules to prohibit such results, based on by imposing a threshold of vote share for eligibility to win a seat.


The first problem: only a single ballot-paper order of presentation of parties candidates

The Commonwealth Electoral Act currently specifies that in each state or territory election of senators just one variant of the ballot paper presented to all voters. The order of party columns is drawn randomly, while the order of candidate names within each column is specified by each party.

In any election having only one variant paper makes possible the donkey vote effect. We can safely assume that this is very low in Senate voting with the current voting options, because ‘above-the-line’ voting requires only one preference be marked, and anyone minded to donkey vote would not vote ‘below-the-line’ (for non-Australian readers these terms will be explained in a moment).

Possibly a more significant consequence of using a single ballot paper variant is the problem of ‘mis-voting’ caused by similar party names. Of course such mistakes can occur with any name-based ballot paper design, but the likelihood of such mis-voting is magnified if the beneficiary of such confusion – which is typically a minor/micro-party with a name similar to a major party – appear before the latter on the ballot paper, especially if they appear in the first position on the paper. This was what occurred in relation to the Liberal Democrat party in the 2013 election of NSW senators.

The solution to this problem is simple and is already standard practice in state or territory elections in Tasmania and the ACT, and that is to create multiple variants of the ballot paper with different party/candidate orders. This is known in Australia as ‘Robson rotation’. The better options use variation both horizontally between parties and vertically between each party’s candidates, and ideally (although this is not done in Tasmania or the ACT) the ‘ungrouped’ candidates’ column should form part of the rotation. The more variants the better, although even a small number of variants will quickly diminish the impact of ballot position distortions.

This is an easy fix, and modern printing capabilities mean that there is no meaningful additional cost involved in printing variant papers.


The second problem: allowing parties to set candidate order within their party distorts choice

Any mechanism by which parties can control the order in which their candidates are elected is a breach of the principle of choice, and is a partial step towards ‘party list’ approaches which destroy the direct relationship between electors and the elected representatives.

Under current arrangements parties can dictate candidate order in a minor form by ordering their presentation on the ballot papers).

There is no need to be especially worried about letting parties merely specify the ballot paper order of their candidates. Presentation of a preferred order of candidates on a ballot paper does not force voters to mark preferences in that order, even if it perhaps provides a prompt. The desire by parties to publicly identify lead candidates and urge that they be given preference is really quite legitimate. With a more open ballot paper, if voters want to reject the presented order of candidates, they may do, and indeed they should take some responsibility for the result if they choose not to do so.

That said, randomised presentation would still be preferable in principle, and this is used in state/territory elections in Tasmania and the Australian Capital Territory without any great difficulty.

The more serious form of party control occurs when above-the-line voting (“ATL”) is used on the ballot. ATL is a device used in the Australian elections for senators and for some other Australian state assemblies. The ballot paper is divided horizontally into two zones by the ‘line’, (normally a thick black bar running across the paper). Parties and their candidates will be presented in columns which run extend both above and below the line; above the line will be a single box for the party, while below the line (“BTL”) will be a distinct row and box for each individual candidate. The instructions on the ballot tell voters to use the boxes either above the line or below it, but not both (if they do mark preferences in both areas, only the above the line markings will be counted).

In the national Senate form of this ballot design, voters using the ATL area are only allowed to make a single “1” vote, choosing just a single party. By contrast the NSW Legislative Council ballots allow voters to use all the ATL boxes preferentially between all the parties.

When such systems are widely used, they create safe automatic appointments for any party which can be confident of securing one or more quotas of votes. As such, candidates effectively achieve effecting appointment to Parliament simply by winning the necessary slot on the ticket at party preselection.

A brief look at possible variations of the ATL method is relevant to dealing with the issues. Most ATL voting uses an ‘autofill’ approach, where marking the ATL box for a party is taken to constitute preference voting 1, 2, 3, etc for the candidates of the party in the order specified by the party. But there are at least four distinct levels of control which ATL voting may impose on how preferences are directed:

  • ‘autofill’ within each party together with full/compulsory preferences for all parties (this is the current Senate system)
  • ‘autofill’ within each party, but where preference between parties are optional; if desired a minimum number of parties that must be preferenced
  • ‘autofill’ within each party, where preferences between parties are fully optional above the line at the voters discretion
  • Finally an approach without autofilling, but instead where an equal distribution of votes at each vote counting stage is allotted to each of the remaining un-elected and the un-eliminated candidates of a party.

All versions of ATL party boxes – except the last in the above list – create the phenomenon of ‘automatic appointments’ for larger parties.

The ideal solution is to omit ATL voting in all forms. But if not, it is preferable to at least use the ‘equal distribution’ variant, which would leave the fate of individual candidates in the hands of those voters who were motivated to express a specific preference for or against the individuals on each party’s slate.

The justification for introducing ATL in 1984 was that it would reduce informal ballots. The informality in question was entirely an artefact of compulsory preferencing requirements. If such requirements were abandoned then much of the need for ATL voting as an option simply disappears.

The third problem: allowing parties to direct preference flows

Party control of the full flow of preferences outside their own party is an even more serious problem than control within their party. It is a major departure from the principle of direct election through the choices of the electors. Specifically, the device of group voting tickets (“GVT”s) used with the current Senate election ballot papers breaks the direct relationship between electors and their representatives that is required by the Constitution.

While marking GVT boxes on ballot papers may constitute choosing a box, it cannot possibly constitute choosing a candidate, because the sequence of eliminations in the preference flow process cannot possibly be predicted in advance by any of the participants in the process: not the elector, nor the candidates, nor even the party officials who determine the contents of the GVTs. Parties can negotiate their GVT deals, and some deals will increase the odds of certain outcomes, but no-one can know which of millions of possible outcomes will ultimately occur. That being so, how can the selection of the successful candidates be described as an act of choosing?

While the GVT device was tolerated by the Australian High Court in the McKenzie judgement in 1984, but that judicial conclusion deserves to be reviewed. The decision in that case was required in haste and was given by the Chief Justice sitting alone. The reasons for decision are very brief and do not conform to the more careful process developed by the Court in more recent decades. The Court now clearly holds that constitutional law requires that the representatives be “chosen” by the electors.

The solution is simple: abandon the use of GVTs.


The fourth problem: invalidation of votes resulting from compulsory preferencing

In a system that is fundamentally premised on free expression of voter choice, compulsory preferencing requirements are conceptually unsound. They create apparent preferences which can no longer reliably be understood to be positive indications of support for the candidates, they intrude upon the free choice of voters, and they are enforced by the disentitlement of many voters from effective participation in the election.

To impose the penalty of vote invalidation on electors who refuse to adhere to compulsory preferencing, or who make an error in completing their ballot, is too severe a rule, and it is difficult to see how it would pass scrutiny against constitutional principle as outlined in the Lange test.

The solution is simple: allow optional preferencing. In the wake of the 2013 election results, most all commentators on the Senate electoral system in recent months are argued for optional preferencing both above and below the line for the Senate. After an inquiry into the matter the Australian Parliament’s Joint Standing Committee on Electoral Matters also recommended that preferencing become optional for future elections.


The election of candidates with small starting first preferences

Some commentators have suggested rule changes so that candidates with low first preference votes should be actively prevented from winning a seat.

The proposed new barrier to such election are usually defined in terms of a minimum first preference vote share for each candidate, normally termed ‘thresholds’. Candidates below that level of first preference support would all be eliminated simultaneously.

Such proposals are unsound in principle. The call for thresholds involves a fundamental misunderstanding of the nature of preferencing.

Thresholds are basically a declaration that the marketplace is severely culled after only one stage of the preference sorting. This clearly damages the freedom of choice of all voters, and involves a discrimination between voters on the basis of whom they vote for. All thresholds are essentially a rule that some votes don’t get counted as their voters intended. Thresholds make votes unequal in the effectiveness of their votes based on their political preferences.

The threshold concept is often expressed as being applied not only to candidates but also to parties, that is to the whole column of a party’s candidates where the aggregate of the votes for all such candidates is below a threshold. The submission to JSCEM of Michael Maley (Submission no. 19) addresses this specific question:

“11. It would also be arguable that the summing of first preference votes of grouped candidates for such a purpose would be of a fundamentally different character to the use of group vote totals to determine the return of deposits, as mentioned by Professor Williams, since the latter exercise does not impinge on the directness with which senators are chosen by the people.

12. Whether it is defensible in principle to deem first preference votes for a candidate to be votes for his or her group is also questionable. On this, it could be noted that a voter whose intention is, for example, to give higher preferences to female candidates than to male candidates will still have to give a first preference to some candidate; but to treat that vote in such a way as to benefit the other candidates of that most preferred candidate’s party would seem dubious: they might in fact be among the voter’s least preferred candidates.”

Surely Maley is correct. Using aggregate party totals as a threshold device is not something that the voters can be assumed to have intended or permitted, and is a misuse of the aggregate totals of their individual votes. The potential impact on voters – forcing tactical calculations to be made and distorting the true expression of their preference ­– is a derogation from freedom of choice and, potentially, is in conflict with the Australian constitutional imperative that the representatives be ‘chosen’.

In truth, the whole matter as seen in the 2013 Senate elections is yet another artefact of the use of group voting tickets. Maley, again:

“14 … the phenomenon of preference harvesting and the consequent election of micro-parties is basically a consequence of the use of ticket voting.”

In 2013 the election of some Senate candidates with very low first preference votes was artificially brought about the use of GVTs. But note carefully that it is not correct to criticise the election of such candidates in general, but only that such instances should be generated solely by the GVT device.

The whole matter can be well illustrated by the case of the election of Ricky Muir as a Senator for Victoria. The quota for election to a Victorian Senate seat was 483,076 votes. Mr Muir’s party aggregate of votes was 17,122, just 0.51% of the formal votes, representing 0.0354 of a whole quota. But Mr Muir received all the votes he needed to reach the quota in the preferences of 482,000 other voters. By the last significant round of counting, shown below, there were 7 candidates remaining un-eliminated.

Count of Election of Victorian Senators 2013: counting stage 291: 7 candidates remaining



FIFIELD, Mitch (Liberal) (ELECTED)


RYAN, Scott (Liberal) (ELECTED)






RICE, Janet (Greens) (ELECTED)


MUIR, Ricky (Motorists)


KROGER, Helen (Liberal)






Lost in fraction




At this final stage of counting five candidates – M Fifield and S Ryan (Liberal), G Marshall and J Collins (ALP) and J Rice (Greens) – had already received the quota, and as a result 5 quotas of votes (totalling 2,415,380 votes) were allocated against those 5 candidates. As between the remaining two candidates a tally of 489,652 votes preferred Mr Muir, while 437,894 preferred Senator Kroger. (The gap between them is now greater than the 36,526 votes for Fiona Patten that remain to be distributed, so at this point the count is concluded.)

Quite correctly, Mr Muir was awarded the final seat. In short, Mr Muir is legitimately elected because he was the preferred choice of a pool of 489,652 voters – a number larger than the quota – against any of the dozens of unsuccessful other candidates, including Senator Kroger.

Now, it may be argued that that pool of voters might have actually settled on someone other than Mr Muir in different circumstances. For example, 148,281 voters whose first preference was for the Palmer United Party (or its individual candidates) formed the largest component of that pool of 489,652 votes – nearly a third of them and a number nearly 10 times that of Mr Muir’s party.

Summation of the first preferences of voters whose votes contributed to election of Ricky MUIR

Candidate Votes
Elimination of Barry MICHAEL (PUP) 161,252
Elimination of Fiona PATTEN (Sex Party) 143,118
Elimination of A FENN (Family First Party) 50,160
Elimination of M FARRELL (DLP) 36,160
Elimination of T MALONEY (ASP) 27,497
Elimination of M RILEY (Hemp Party) 19,104
Ricky MUIR (Motorists Party) 17,083
Elimination of J ZAMMIT (AFLP) 15,909
All others ballots 19,369
Total 489,652

Mr Muir’s own votes have contributed only around 4% of the total needed to win the senate seat.

More importantly, why is Barry Michael, the lead candidate of the Palmer United Party’s group voting ticket, not the most obvious winner among the preferences of the voters of this pool? He would most likely have been preferred over the Sex Party’s Fiona Patten in the preferences of voters who supported must of the smaller party groups.

It might have been that Mr Muir coming through this pack to win is simply his good fortune in a highly particular flow of preferences. But in truth we know the reason that he and not someone else prevailed – it was driven by the strict preference sequence created artificially by the use of group voting tickets.

To conclude: it is not illegitimate for a candidate to win a seat even from a low basis of first preferences, if that happens to be the genuine pattern of preferences as recorded on our votes. But it is highly questionable if such a flow of preferences is manufactured by a group voting ticket. In the controversy about the legitimacy of the election of Mr Muir, and similarly with the initial election of Mr Dropulich in the election of senators for Western Australia, and the possible impact of preference sequences on some other elections, it is the use of the automated GVT device, not the starting tally of any candidate’s first preference votes, which undermines the sense that the result is a genuine choice by the voters.

Once again, the correct change to the rules is to abolish the use of GVTs, not to invent eligibility thresholds.

Given the High Court’s tendency to protect voter choice by scrutinising legislative amendments which have a clear partisan impact, deliberately contrived levels of threshold aimed at excluding certain categories of parties or candidates are quite likely be held invalid if the constitutional validity of such a measure were challenged.

Finally, note that these ‘low-start’ events do not appear to occur in elections in the ACT and Tasmania or in the Victorian or Western Australian Legislative Councils. The mere absence of GVTs seems to be enough to prevent them.


This article is based on extracts from my Submission 181 to the Joint Standing Committee on Electoral Matters inquiry into the 2013 election which took place during 2014.

%d bloggers like this: