Thursday, July 23, 2020

How Should We Solve The Problem Of Unintended Informal Voting?

Advance Summary

1. In single-seat elections using compulsory preferential voting, high rates of unintended informal voting occur.

2. Informal voting is especially high where there are many candidates, where there is confusion between voting systems, and where electoral and/or English language literacy are low.

3. It is unclear whether unintended informal voting creates a significant two-party preferred advantage for one side of politics, although it appears to deflate Labor's primary vote.

4. There are many ways to reduce the number of votes that are disqualified without having to adopt Optional Preferential Voting.


Among the many things that didn't change much in the recent Eden-Monaro by-election was the informal vote.  In fact it declined very slightly, from an unacceptably high 6.80% to an unacceptably high 6.71%.  The number of candidates had risen from eight to 14, but the lack of a simultaneous Senate election would have probably reduced the number of confused voters voting 1 to 6 and then stopping.

Nonetheless when the dust settles it will probably be found that most of the informal votes in this by-election were by mistake.  Unlike in the Senate, where savings provisions preserve anything with a unique 1 above the line or a unique 1,2,3,4,5,6 below it (whatever the errors in any other numbers), the Reps formality rules have no margin for error.  If a voter numbered 1-12 and two 13s on the 14-candidate Eden-Monaro ballot, their vote would not count, even if the two candidates numbered 13 were the first and second to be eliminated.  If a voter voted 1-13 and left one box blank, that would be saved by a savings provision, but even if a voter voted 1-13 and then put the last candidate 15, their vote would be informal.  (Oddly, in a two candidate race if a voter puts a 1 for one candidate and, say, a 3 for the other, that is fine, presumably to eliminate arguments about whether a given 2 is actually a 3).  Overly strict formality rules are excluding votes from the count for minor clerical errors.

Some further attention on the informal voting question and alternatives to compulsory preferential voting comes from South Australia, where the government intends to introduce optional preferential voting at state level, though this may not get through the state's Upper House.

Many informal votes in Reps elections are deliberate, and the AEC's analysis of the 2016 count found that more than half of the c. 5% informal rate in 2016 was deliberate, for the first time since 2001.  The rate of assumed unintentional informal voting in 2016 fell from 3.46% to 2.41% of all votes, as a result of Senate reforms that made it less likely that voters would get confused between the Reps and the Senate and attempt to vote 1 only in the Reps.  This was an impressive improvement, but the total rate of informal votes rose by half a point to 5.54% in 2019, and the numbers are not yet in on whether the rise was caused by a rise in deliberate informals, a rise in accidental informals, or both.  

There were some especially stark cases in 2019.  Here's the historical informal rate graph for Mallee (Vic):

Mallee is not normally a high scorer on informal voting but in 2019 it had the largest number of candidates (13), and a redistribution had moved new areas into the electorate.  

Overall, unintended informal votes tend to be driven by the following:

number of candidates (the more candidates, the more voters will make a mistake in trying to order them thus invalidating their vote).  There may, given the new Senate system, be a particular increase when there are eight or more candidates.  (Eight is the smallest number for which a vote is informal if the voter numbers exactly six squares).

optional preferencing at state elections (currently in NSW only) - this results in more voters in federal elections voting just 1.
education and English language literacy, thus certain Western Sydney electorates persistently appear near the top of the informal vote rate lists.

Why Are Strict Formality Rules A Problem?

Strict formality rules coupled with compulsory preferencing are a problem because voters who may have expressed a clear, and partly or entirely useable, preference, have their vote discounted because they have made a mistake in one or more places.   Often the mistake is irrelevant to the outcome and the vote could be counted (to the point to which it was clear) without changing the results.  Ignoring whatever advantage this may give some candidates over others, it is above all a problem because it excludes voters from our democracy just because they have made a clerical error. Many voters have clear political views but are simply not good with numbers.  Furthermore, as language skills are one of the drivers of formal vs informal voting, this exclusion is bound to break against some ethnicities more than others.  Strict formality rules are elitist and implicitly racist.  Are we really a great democracy if we go to election day knowing that the rules are such that double-digit percentages of voters in Western Sydney will have their votes discounted, mostly because of unintentional errors or not understanding our voting system, and do nothing about it?

In my view leaving the situation as it is should not be considered as an option.

Does Unintended Informal Voting Actually Advantage One Side?

Whether either side of politics is actually advantaged by strict formality rules is a complex question.

The only state, to my knowledge, that provides a breakdown of informal votes by party of intended vote is Tasmania.  Tasmania has the Hare-Clark system with semi-optional preferences.  So long as the voter gets the numbers 1 to 5 correct, they can stop at that point or make as many mistakes with higher numbers as might happen, and their vote still counts.  This leads to low rates of accidental informal voting.  Despite this,  Labor voters are more prone to unintentional informal voting than the Liberals and Greens.  (Small parties are the most prone, often because they run fewer than five candidates.)  In recent Tasmanian elections, the unintended informal rate was 0.91% among intending Labor voters, 0.74% among intending Liberal voters, and 0.59% among intending Greens voters.   Assuming a similar pattern recurs at federal level, Labor's disadvantage would be muted by the fact that Greens voters, who strongly preference Labor, are very likely to vote formally. 

Most of the divisions with very high rates of unintended informal voting are usually or always safe Labor seats in NSW, especially western Sydney.  The Coalition has never won Blaxland, Fowler, Watson, McMahon, Werriwa or Chifley under their current names (though it did hold St George, precursor to Watson, in the late 1970s) and has only won Barton once since 1983.  But one exception to this is Lindsay, the outer-Sydney swing seat that has been won by the government of the day at every election bar one (2016) since its creation in 1984.  

Labor seats having high informal votes alone proves nothing.  It may be that each division has its own underlying Reps informal voting rate that affects all parties in that division equally.  This would cost Labor more 2PP votes nationwide than the Coalition, but would not affect its chances in any seat.  

However, a similar pattern can occur at booth levels in particular divisions.  In this case, even if each booth had its own level of informality affecting all parties, a higher informal rate in booths with more Labor voters would harm Labor's chances in a seat.  

One seat that displayed a strong pattern in informal voting by booth in 2019 is indeed Lindsay.  In 2016, 7.5% of Lindsay votes were unintentionally informal, the highest in the nation.  Here is the 2019 Lindsay informal vote by booth compared to the Labor Party primary:

... which flows through, a little less strongly, to the 2PP:

However, looking at any pattern like that requires a strong caution re the ecological fallacy.   We don't know that just because some booths have a lot of Labor voters and a lot of informal voters, that the informal voters are ones who intended to vote Labor but had their votes disallowed.  It's quite probably the case that areas of high Labor 2PP voting in Lindsay also have high deliberate informal voting rates, given that deliberate informal voting is associated with identifying as working class or with not identifying with any social class.

Even if the pattern in Lindsay is driven exclusively by variation in unintended informals, that doesn't show that intending Labor voters are so much more likely to vote informally as is implied.  More likely factors exist in specific areas that cause both parties' 2PP voters to be more likely to vote informally in those booths.  It's even possible that in the most Labor-y booths, the unintended informal voters could be mainly intending Liberal voters for some reason, but that isn't likely.  In theory the graph above would be consistent with Labor losing so much to the informal pile as to disadvantage them by nearly 2% across the seat, but it's probably nothing like that, in terms of unintended informals.

Also Lindsay 2019 was unusual in showing this pattern.    Ten NSW seats in 2019 had eight or more candidates (the number most likely to lead to high rates of unintended informal voting) and of these only Hunter and Lindsay displayed a significant relationship between Labor's 2PP at booths and the informal rate at booths.  The remainder (Cowper, Eden-Monaro, New England, Richmond, Robertson, Berowra, Fowler and Lyne) did not display such a pattern.  Both Hunter and Lindsay had specific local factors that could have driven Labor-y booths to deliberately vote informal.

Across NSW as a whole in 2019 there was a correlation between the Labor 2PP and the informal vote by booth after controlling for 2PP and informal vote rate differences between particular seats, but the Labor 2PP explained only 5% of variation in the informal rate.  That's still significant - because the number of data points is so large - but there could be many explanations beside Labor voters tending to get caught out by mistakes.

The weakness of that pattern might be specific to this election but it's been notable in recent polling that there is very little relationship between education levels and a person's 2PP vote at the moment.  Overall I don't think there's convincing evidence that unintended informals hurt Labor more on a 2PP basis, though it might be true.  I think if they did so by a lot it would have been obvious to (and highlighted by) scrutineers and the party by now.    I think the arguments from inclusiveness are the stronger reasons why informal voting needs to be fixed.

Options for fixing

A major problem with debates about unintended informal voting is that any call to relax the current rules tends to get mistaken for a call for entirely optional preferencing, which leads to reflex opposition from the left especially.  But as I show below there are many alternatives and partial solutions:

1. Allow breaks in sequence

This has been proposed by Antony Green, and exists in WA (see comments).  A vote that contains all unique numbers with no repeats would be accepted even if it had skips, thus if a voter votes 1,2,3,5,6 on a five candidate ballot their vote is converted to 1,2,3,4,5.  There might have to be some clause to eliminate vexatiously formal votes (for instance still disallow any vote with numbers more than three greater than the number of candidates).  In an 11 candidate race for a Tasmanian Legislative Council seat (where voting 1-3 is required) a voter voted down the paper 1,2,3,10,20,30,40,50,60,70,80.  

2. Reallocate tied values arbitrarily

This is my idea and I don't know how workable it is legally, but it is another way to put more votes in play while maintaining full preferencing on every ballot.  Where a vote contains duplicates (perhaps up to a certain limit) the ties would be broken in favour of the higher candidate on the ballot in booths with even booth numbers, and in favour of the lower candidate on the ballot in booths with odd booth numbers.  (Some assignment method would have to be devised for postals).  This would be used in conjunction with 1 above.

3. SA Savings Provision Ticket

In South Australia currently, parties can lodge a savings provision ticket vote, and informal votes that are 1 for their party are converted to the ticket vote provided that these votes were following that ticket to the point of the first error.  This is effective in reducing informal voting, such that at the last state election, lower house informals in SA were 4.1%, compared to 4.34% in Queensland, 4.54% in Western Australia (which allows breaks in sequence), 5.54% federally and 5.83% in Victoria at their most recent elections.  Even the 4.1% rate was unusually high for SA.  A problem with this system, however, is that it does not offer equal protection for informal votes that disagree with a party ticket.  That form of discrimination shouldn't be accepted, but there are ways around the problem (such as simply having the vote follow the supplied ticket for remaining candidates from the point of the first error irrespective of any prior deviations.)

4. Save Votes That Have Irrelevant Errors

This is an informal suggestion by Antony Green.  A vote is saved if it contains an error but the vote would never be distributed to the point of the error.  So, for instance, a 1 Labor vote in a seat where Labor finishes in the top two is saved even if it contains errors.  However a 1 Greens vote with the major parties tied, in a seat where the Greens are excluded, stays informal.

There are some minor issues with this idea because the major parties would benefit in funding terms from the inflation of their vote share compared to other parties.  The most significant issues with the proposal are:

* major parties might become lazy about getting their voters to number all boxes only for it to matter when their candidate is unexpectedly eliminated.

* in a contest with a close and critical exclusion (such as the Labor/Greens race to make the final two in Melbourne Ports 2016), it is impossible to know whether some votes will need to be distributed to the point where they contain an error or not until preferences are thrown.  This would result in a small number of votes being of uncertain formality until a late stage.  In a series of close exclusions (such as Mallee 2019), it is possible that a candidate could outlast another candidate using votes that at that stage were potentially formal but later became informal, meaning that arguably the exclusion order would have been different.

Depending on how these matters were resolved, the system could be very complex to administer.  A possible crude solution is to throw preferences until two candidates remain and at that point admit otherwise informal votes that start with or flow to one of these two candidates before the first error, even if including these at an earlier stage would have altered the order of exclusion.

5. OPV-As-A-Savings-Provision

Rather than openly allow optional preferential voting, this method asks the voter to number all the boxes but in practice accepts any vote with a unique 1, up to the point of the first error.  This used to be the method used in federal elections but became contentious because of so-called "Langer voting".  Albert Langer, a far-left activist, had campaigned for voters to number all the boxes but use tied values for the major parties.  This was partly connected to a desire to bring down the Labor government and partly to a groundless theory that this could cause no candidate to be elected in cases where nobody passed 50% of all votes cast after preferences.  Tens of thousands of votes were exhausted in the 1990 and 1996 elections, and several thousand in the 1993 election, following Langer-voting campaigns.  Over time Langer's campaigns aroused the ire of both the ALP and the AEC.

Such a system has to either allow people to publicise "Langer voting", which might lead to a rate of high Langer-voting by people who think it will acheive something that it won't, or else it has to ban advocacy of Langer voting.  The latter carries significant free speech implications because it becomes illegal to advocate a method of voting that is itself legal and valid.  This method, in common with all OPV systems, also has the downside that even a small rate of exhaustion makes a lot of calculations done in election-watching less efficient.  For instance one can no longer say exactly how many voters in Clark put the top three contenders in the order Wilkie, Labor, Liberal as opposed to Wilkie, Liberal, Labor.

Variants of this idea might or might not save votes with several squares not filled.

I think the most satisfactory version would be to allow Langer voting and to allow people to advocate it, but to prosecute people making false claims that Langer voting could lead to a count being voided on account of no candidate reaching 50%.  These could be taken as misleading electors in relation to their vote.

6. Semi-OPV

A semi-optional system is used in Tasmanian Legislative Council elections.  Seats with four or fewer candidates have effectively compulsory preferencing, but for those with five or more, the voters only have to number three candidates and can then stop.  In the last six-year cycle the informal rate in these elections has averaged 4.00%, and the higher rates of informal voting have often been in the less contested races rather than those with lots of candidates, suggesting that much of the informal voting is deliberate.

A version of semi-OPV that could be used at federal elections would be to save votes with at least a unique 1 to 6, thus protecting the vote if the voter has become confused with the Senate.  Semi-OPV has the same issue as mentioned for OPV-As-A-Savings-Provision in terms of electoral calculations but if one is willing to allow that issue it is a possible solution that lacks the high exhaust rates and other negatives of full OPV.

7. Optional Preferential Voting

And finally, the current SA proposal, which also exists in NSW, and formerly existed in Queensland and the Northern Territory but was abolished by Labor governments in both.  Under fully fledged OPV any vote with a unique 1 is valid, and the voter gives as few or as many preferences as they like.

There's a strong freedom-based argument for fully fledged OPV - why make a voter express a preference they don't want to distribute, especially since a mistake made in the process might invalidate their vote?   However, OPV has some undesirable aspects.

One of them is that parties sometimes try to encourage voters to "just vote 1" - often doing this using signs that are arguably or clearly mistakable for official election signs.  Sometimes this message is used to make it easy for a party's voters or to portray other parties as a rabble, but sometimes it is also done for tactical reasons - a major party wants to weaken the flow of preferences from an independent to the other major party for instance.   My own dislike of any form of signage implying that it is in any way a good idea for a voter to stop at 1 - let alone even the remotest misleading hint of it being an official instruction - is such that I can't support OPV unless this kind of thing is severely banned.  I would want to at least see parties responsible for misleading "just vote 1" signs bankrupted by fines and deregistered.

Another problem with OPV is that major party voters can get extremely used to just voting 1 because their preferences are not distributed.  When a minor party sneaks into second - as with the Greens in some of the northern NSW seats - the minor party may have great difficulty getting a preference flow if major party voters did not consider the possibility that their preferences could matter.

I'm not convinced OPV is as much a benefit for the right as both the right and left think, or even in the current environment a benefit at all.  In the Queensland 2017 election the switch from optional to compulsory preferencing seemed to have very little direct impact on the results.  Labor is more dependent on preferences overall than the Coalition, but Greens voters are more likely to distribute preferences anyway, while Coalition-leaning minor parties like One Nation and (usually) Shooters, Fishers and Farmers tend to have higher exhaust rates.

All the same Labor especially dislikes OPV because it needs to enthuse Greens voters to get their preferences under it, instead of just taking them for granted.  This gives the Greens more policy leverage over Labor, as if Labor's desire to compete with the Greens was not enough of a problem already.

For South Australia specifically, in 2014 thanks to the existing ticket savings provisions, only 0.19% of all votes cast were declared informal but would have been saved under OPV.  However, in 2018, this figure rose to 0.68%, possibly because of an increase in candidate numbers.  It is worth noting here that all of the votes that would have been saved under OPV would also have been saved by using OPV-as-a-savings-provision (option 5).

8. Don't Have Single Seat Electorates

I mention this one only for completeness as people are bound to suggest it anyway.  I'm assuming for the sake of this article that switching to non-single-seat-electorates is considered a non-starter.

I hope the options listed above show that there are many ways to make progress on reducing the unacceptably high level of informal voting in federal Reps elections, without needing to go all the way to full OPV.


  1. Since 1996 breaks in sequence have been allowed in Western Australian Legislative Assembly elections. Neither side of politics were silly enough to worry about Langer-style campaigns and the tally of exhausted votes in preference distributions remains small. We have avoided the injustice of invalidating votes with clear intention but with irrelevant breaks in sequence.
    There is also concern that some electors of Chinese heritage dislike writing the number 4 and have sometimes invalidated their votes through this.

    1. Thanks Jeremy; wasn't aware of that and have added notes to that effect. Quite a contrast to the harsh rules for voting BTL in the WA Legislative Council!

  2. yes look at voter intention a vote is formal to the extent voter intention can be identified..... all ones informal, one one and nothing else formal. a 1 23333 is formal up to the second preference a tick or a cross in isolation is formal... the voter choice must be paramount..... no games

  3. "There's a strong freedom-based argument for fully fledged OPV". Indeed - it's a mild form of totalitarianism to say "Unless you number all squares, or X squares, we'll toss your vote aside". But the ballot paper could have some *persuasive* text on it rather than dogmatic instructions - something like "Place the number 1 in the square for the candidate of your choice. Your vote may be more effective if you place further numbers (2,3,4,etc) showing who you would prefer if your first choice is not elected." Then translations of the instructions into the languages commonly used in each electorate could be placed in the polling stattions.

    And yes, the display of "Just Vote 1" posters, especially in colours resembling those used by the electoral authorities, should be subject to heavy penalties.

  4. SA's parliament is about the right size to do 10 5 seat electorates based on the federal seats (similar to Tasmania). Those seats are more likely to stick than the ever changing SA seats. I'm not sure if it's constitutionally possible, but SA has had multi member electorates before. Hopefully it isn't a non starter - seems to be a viable solution to SA's electoral quirks.

  5. Ummm, John - total of 50 seats? What happens when leftish parties get 25 seats and rightish also 25? Libs bribe a Green to be Speaker or Labor bribes an ON? Because if a major provides the Speaker then any vote on left/right lines will go 24-25 against that party. In any of those scenarios the Speaker's casting vote won't arise because that only happens if the votes are equal.

    I like the idea of a smallish number of districts each electing 5 or so members by PR, but I'd suggest 9 or 11 districts, and forget the idea of having fed and state boundaries the same. Until, perhaps, SA's federal entitlement rises back to 11 or, more likely, falls to 9.

    1. Tasmania had 30 seats from 1909 to 1956 and often had parliaments where one party had exactly half the seats. Towards the end there was a rule that the major party that lost the popular vote had to provide the Speaker. Eventually it just got unbearable and the House was expanded to 35.

    2. Somewhat late to the conversation here, but you have issues either way.

      If you have an odd-number of members of the assembly and elect one of the members to serve as the speaker, the assembly now has an even number of members without the speaker. Then there’s the question of whether the speaker still has a vote on most legislation or not (most assemblies tend to not, except as a casting vote), and then the question of, when the vote is tied, the speaker casts the vote according to their own desires (including potentially bringing the government down) or casts the vote according to neutral principles (which essentially deprive their constituents of any vote on any issue, as is currently the case in the Tasmanian Legislative Council). There is, at least, the potential advantage of the latter that the larger side of the assembly has provided the speaker and that, if ruling to preserve the government, the speaker is most likely voting in line with the largest side of the assembly.

      If you have an even number and have to provide a speaker from their number, and the speaker does not vote on normal legislation, you have the potential situation where legislation loses that would have tied with the speaker’s vote.

      It seems to me the ideal might be to have an odd-numbered assembly and to then elect the speaker from outside the assembly, as is common in some Caribbean Westminster systems and as is allowed, but has never happened, in the US. This way, there is no deadlock in electing the speaker, no constituency is deprived of their voice and simultaneously the need for a casting vote is decreased because the assembly has an odd number of members.

  6. "Eventually it just got unbearable" Yes, exactly! The "rule" depended so much on the forebearance of the "losing" party. Our current pollies aren't like that. Not even in Tasmania, I suspect.

  7. Even numbers of seats total aren't ideal but aren't unworkable. The strength of centrist parties in SA would help here. Even without xenophon, from federal senate results Mayo looks like it would have elected 2 Liberals, 1 ALP, 1 Green and one Centre Alliance. The Tasmania situation seems to have been caused by having 6 member electorates making a tie the likely outcome in each of them. With 5 member electorates you would need to get a tie statewide. It's possible but not as likely as Tasmania from 1909-1956

  8. Looking at last year's federal LH results. (Lnp/Alp/Grn/Other)

    Adelaide: 2/2/1
    Barker: 4/1
    Boothby: 2/2/1 (close to 3/2 but ALP excess should break strongly enough to Greens to hit a quota)
    Grey: 3/1/0/1 (Lib 4 gets eliminated early and they pick up 13% over the count that could easily put another right winger, probably PHON, to a quota before Labor)
    Hindmarsh: 2/2/1 (ALP 3 vs Green is very close with GRN slightly ahead at exclusion)
    Kingston: 2/3
    Makin: 2/3
    Mayo: Hard to say due to Sharkie. I looked at the senate results for this to give 2/1/1/1
    Spence: 1/3/1 (Greens would hit quota on ALP excess and LNP never quite hit 2 quotas 2PP
    Sturt: 3/2 (close to 3/1/1 but ALP 2 maintains a lead over GRN throughout)

    LNP - 23
    ALP - 20
    GRN - 5
    Centre Alliance - 1
    PHON - 1

    Result - CA would probably choose Labor over deadlock and working with PHON to form government

    It's alarmingly close to a tie, but there's enough variability to have a workable system even with even numbers of seats. Passing bills would be tough however.

  9. "I'm not convinced OPV is as much a benefit for the right as both the right and left think, or even in the current environment a benefit at all."

    What do you think the impact of OPV would be in seat terms at a federal election? So far I've seen three analyses online:

    An academic study done by the Conversation, who say it would devastate Labor:

    A simpler analysis by Dr Peter Brent, who comes up with similar but less damaging results for Labor:

    And an analysis by a group called Armarium Interreta who find it would have a small effect compared to the above two:

    Personally it seems like the first source is most credible (being done by academics and published in Conversation), but as you say, it might be that the Greens don't exhaust as much as other parties and that mutes the effects on TPP. What do you think?

    1. I have done my own analysis piece here: My estimate was that OPV would be worth a few seats per election for the Coalition. The difficulty is that preference flows under OPV are more variable between elections and indeed in Qld 2015 OPV delivered very strong preference flows to Labor.

      Unfortunately something being by an academic (even a political science academic) and published in The Conversation is not a reliable indicator of quality. The major problem with that particular analysis in The Conversation is that it uses NSW 2015 - a single OPV election where Labor performed badly - as a baseline without considering the wider range of OPV elections. Also, the piece contains at least one error (Wayne Swan would not have lost Lilley in 2010 under OPV as he trailed on the primary vote by only a fraction of a point). So my own views are closer to those of Brent and AI. I was as usual very impressed by the level of modelling detail in the AI analysis.

      All of us find that Labor would have lost the 2010 election under OPV.


The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. Comments will be published under the name the email is sent from unless an alias is clearly requested and stated. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.