Firing Blanks: The Victorian Teal Open How-To-Vote Cards Dispute

 Update Nov 17: Teals Win (for now)!  It is being reported that teals have won their VCAT appeal against the VEC's refusal to register various cards.  This follows events yesterday where the VEC ordered various candidates to desist from distributing these forms of cards, an order it may now turn out the VEC had no business making.  A link to the judgement will be posted when available.  The VEC can appeal to the Supreme Court if it wishes.

Disallowed proposed Frederico how-to-vote card

Yesterday there was significant publicity about the status of some proposed open how to vote cards for various Victorian election teal independents including Felicity Frederico (Brighton), Mellissa Lowe (Hawthorn), Sophie Torney (Kew), Nomi Kaltmann (Caulfield) and Kate Lardner (Mornington).  The Victorian Electoral Commission has disallowed the proposed card above on the grounds that it shows blank boxes.  Despite the card twice saying the voter needs to number all the boxes, the VEC is concerned that the imagery may result in a voter voting 1 for Frederico and then stopping (which is informal in Victoria).  The VEC points to the 2018 VCAT decision in Sheed v Victorian Electoral Commission while the teals and their supporters point to the lack of any problems (in either law or formality) with similar cards for Monique Ryan at the federal election.  Legal challenges are being mooted.

After a fair amount of discussion of this matter yesterday my current view is that the VEC's statements are indeed completely consistent with the VCAT ruling in Sheed and their instructions in their candidate handbook, though their separate checklist could do with some rewording.  A legal challenge could raise interesting issues about whether a court can ever be persuaded by empirical evidence regarding actual rates of confusion, and whether a court would want to consider where votes are cast or who they are cast for in deciding the matter (I rather doubt the court would go there).  But above all else, this is silly and so typical of the state of Victoria's electoral laws. Candidates are being "saved" from maybe/maybe not losing a handful of their own votes through their own campaigning choices while all the time the house is burning down: Victoria has the worst lower house informal voting rate in the nation.

How to vote card registration laws

In federal elections there is no requirement to pre-register a how to vote card that is going to be distributed.  In Victoria, however, a proposed how to vote card that is to be distributed on polling day (but not prepoll) needs to be pre-registered with the VEC.  This means that the VEC must form a judgement on the compliance of cards with the relevant laws, whereas federally the AEC is not obliged to positively rule that material is OK in the absence of complaints.  

That's one difference, and another is that the Victorian laws are more extensive.  Under Section 79 3(a) the VEC must refuse to register a card that "is likely to mislead or deceive an elector in casting the vote of the elector".  This parallels provisions found elsewhere, but also under Section 79 3 (b) the VEC must refuse to register a card that "is likely to induce an elector to mark the vote of the elector otherwise than in accordance with the directions on the ballot-paper".  The VEC's interpretation, upheld in Sheed, is that Section 79 3 (a) disallows any image of a ballot paper containing blank boxes.  If a candidate wants to issue an open how to vote, then the VEC wants it to use a similar format to this card that Sheed used in 2014.  The key differences are that it clearly states the number of boxes to be filled, and that it does not depict blank boxes.  VCAT suggested (apparently just intuitively as no evidence is referenced) that a voter who was confused by this card would be more likely to ask for help rather than make a mistake.

(Some teal supporters have referred to section 79 2 (c) that they have argued allows any card that "indicates the order of voting preference for all candidates listed on the card or contains a statement that a number must be placed against the name of each candidate", but this is incorrect.  Section 79 2 (c) is only something the VEC must "have regard to", it clearly does not override Section 79 (3) (a) or (b) if a card infringes those sections.)

The VEC's position is stated in crystal clear language on page 30 of the district candidate handbook (PDF link): "If, for example, an HTVC for a district election contains empty boxes, the voter following it might complete their ballot paper the same way, leading to an informal vote. For this reason, the HTVC application would be refused."  There is a more confusing statement in the applicant checklist that in my view should be reworded:

(Interpretations have varied but I read this as applying "no boxes can be blank" to a candidate who wishes to indicate an order of voting preference for all candidates, but not necessarily to one choosing not to.)

In applying laws of this kind, the normal standard is to accept that there are a range of electors and that some of them are more vulnerable to confusion than others.  If something is potentially misleading to a voter who has low literacy skills, who does not use English as a first language or who is, as VCAT puts it, "unsophisticated", then that's enough.  It doesn't matter if the number of voters potentially misled is very small - if there's a serious risk of confusing even one elector, that can be enough to trigger these laws.

Thus in Sheed, VCAT found that "Turning to the Card, first it offers an image which can be interpreted as how the ballot paper should be filled to vote for Ms Sheed. It can be interpreted that all one has to do is place No. 1 in the box next to her name and leave the other boxes blank. It is likely that vulnerable voters would focus on this image rather than the text on the Card with that result."  VCAT also referred to evidence that, statewide, around 1% of Victorian voters were casting this kind of just-1 informal vote.

Do these cards cause informal voting?

Fortunately we have had a recent test of whether cards like this cause 1-only informal voting, via Monique Ryan's campaign at the federal election.

Teeth were gnashed when overheated concerns were raised that thousands of voters might vote informally on seeing the above card, but in this case data wasn't the plural of anecdote.  Scrutineers for multiple candidates have informed me that the rate of 1-only votes for Ryan was (i) negligible (from the comments I've seen it may have been less than 0.1% of intending Ryan votes) (ii) lower than for other candidates.  Informal votes caused by sequence errors (omitted and repeated numbers) were much commoner but not more so than for other candidates.  The overall official informal rate for Kooyong fell 0.08% to 2.89%.  There is no evidence that Ryan's card caused any significant level of informal voting.

Do these cards reduce informal voting?

Teals and teal supporters have been tweeting stuff like this, arguing that the Ryan card actually reduced the informal voting rate in Kooyong.

Firstly, Torney's tweet is incorrect (but the error is in her favour); the national Reps informal rate was actually 5.19% with 4.79% in Victoria (her figures are for the Senate).  Kooyong was way below the national and state average, but that is Kooyong; it has only gone above 3% once in the last twelve general elections.

I want to go into this in some detail (===wonk factor 4/5 sector ahead ===) because informal voting analysis is complex and no-one in the above tweets knows what they are talking about.  Different areas of the country have much higher or lower underlying informal voting rates that are correlated with socio-economic indicators such as education levels and proportion of voters who do not speak English at home, but that are also affected by what electoral system exists at state level.  The seats of Blaxland, Fowler and Watson perennially appear at the top of the informal voting rates because they are both demographically prone to informal voting and also located in NSW, which has optional preferencing at state level.  

At given elections, the rate of deliberate informal voting rises and falls, and there can also be overall trends in the rate of accidental informal voting.  Accidental informals can go up federally at state level, for instance if a federal election is held just after a NSW state election as in 2019.

Because of the wide variation in underlying informal rates, just looking at a raw informal rate in a division will never prove much by itself; it is better to look at how the rate is changing and how that compares with the rate of change elsewhere (especially in the same state.)

Federally, changes in candidate number are a strong predictor of changes in the informal rate by seat.  There are two useful predictive indicators, firstly the raw change in candidate numbers, and secondly the eight-candidate status change.  The eight-candidate status change refers to whether the number of candidates changes from fewer than eight to eight or more (in which case I code it as +1), or vice versa (-1).  If it stays on the same side of eight, I code it as zero.  The reason it matters is that when the number of candidates reaches eight, a vote that mistakenly follows the Senate instructions (1-6 minimum) on the House ballot paper becomes informal.

Beyond this, any number of things might cause changes in the informal rate in certain divisions - the nature of the campaign, demographic shifts in the electorate, the range of candidates available, the issues mix in the election and so on.  As such, while the informal vote fell in Kooyong by 0.08%, that could be caused by more or less anything, especially as the rate in Goldstein (which had similar cards for Zoe Daniel) went up by 1.22%.  Indeed the drop in Kooyong didn't even match the national average drop of 0.35%.

In fact the Kooyong informal rate result is still pretty good, just not in a way that proves anything.  This is because the number of candidates in Kooyong increased by three from eight to 11.  A regression for all Victorian divisions (excluding the new seat of Hawke) gives the following:

change in informal = 0.398*(eight-candidate status change) + 0.507*(candidate number change) - 1.019 +/- 0.754

(Interestingly this suggests that had the number of candidates not increased, informals in Victoria could have fallen by a full percentage point.  Increased postal voting may have had something to do with that, as well as the nature of the election.)

For Kooyong the expected change by this regression was an increase of 0.5% so a drop of 0.08% is pretty good - there is just no evidence it was caused by any specific thing.  And indeed in Goldstein (which rose from six to nine candidates and therefore had an eight-candidate status change) the predicted change was 0.9% and the actual increase exceeded it.  This also happened in Indi where Helen Haines had a similar card (I haven't checked yet if she had it in 2019 as well).  

So how about the very low rate of just-1s on Ryan's own votes, does that prove anything?  No.  They are the educated end of an educated division; one would not expect any more than a handful of voters for Ryan to just vote 1 by mistake and stop.  

Conclusion (===and end of wonk factor sector===): there is no evidence Ryan's card caused any decrease in informal voting either.

Why do the teals want to use these cards?

As best I can determine, the teals don't want to use the Sheed 2014 style reproduced above (with no ballot paper mockup, just an instruction to vote 1 for the candidate and an instruction to number all of the boxes) because they think that it could cause informal voting.  Here again, given that they are pitching to the educated end of educated districts, I'm very doubtful that the risk exists.  Much was made of this in the 2018 Sheed case (where Sheed argued confusion in 2014 had increased the informal vote) but even in Shepparton (a district with far higher underlying informal voting rates) the change in informal vote simply matched the state average (an increase of about 0.3%).  That was a relatively poor result given that there was one fewer candidate in Shepparton in 2014 than 2010 (meaning that the projected result would have been a fall of 0.2%), but again not enough to suspect any one thing would have caused it.  In the following graph, Shepparton 2014 is the red dot - if its change in informal voting rates is interesting then so are the changes in dozens of others:

One advantage of the teals' preferred design could be that it helps the voter locate the candidate on the ballot paper.

The elephant in the room

Generally I am all for measures that reduce unintended informal voting, and generally I welcome Victoria's stricter approach to misleading electoral material, but this situation is absurd.  Even if the cards proposed by the teals really are misleading voters and are doing so more than the approved 2014 version would, then the damage in the seats targeted by the teals will be (i) very very minor and (ii) entirely self-inflicted.  While this may very well turn out to be an entirely correct reading of Victorian electoral law by VEC and VCAT, it's paternalistic to protect candidates from their own supposed folly in this way, so if that is the law, then in this case it's silly.

Meanwhile, the Victorian Parliament does nothing legislatively about the fact that Victoria has the highest informal voting rate in the country, as a result of compulsory preferential voting without useful savings provisions in the state with the second-highest share of vulnerable voters after NSW (which has optional preferencing.)  At least the VEC is being resourced for studies to try to reduce informal voting (more than can be said for Tasmania where unintended informal voting is soaring and nobody official shows any public sign of caring.)  With a scarcely believable increase in candidate numbers of 2.6 extra candidates per seat there is potential for this to get worse, though underlying informal rates are hard to predict and the recency of the federal election (for instance) might save the day.  At the least, some voters in four divisions with 14 or 15 mostly superfluous lower house candidates will have a very challenging time ranking them all without error.