Wednesday, February 26, 2020

Unintended Informal Voting In Tasmanian State Elections

Advance Summary

1. In Tasmanian state elections for the House of Assembly, any vote that fails to have the numbers 1-5 each once and once only is ruled informal and does not count.

2. A Bill to expand the Tasmanian House of Assembly would result in this being changed to 1-7. This would be likely to increase the rate at which voters voted informally by mistake.

3. The current rules very slightly advantage the Greens over other parties, especially Labor and most fourth parties and independents.  However, this hasn't decided any contest for a seat between parties in the last 30 years.

4. It is plausible that requiring voters to fill seven boxes without error would further increase the advantage for some parties over others, however the evidence on this is insufficient.

5. Unintended informal votes where a voter mistakenly omits or doubles numbers could be included in the count using a savings provision system already used in the ACT.

6. The view that the ACT system causes massive exhaust rates compared to Tasmania is based on a misunderstanding of the ACT computer counting system, which continues to distribute votes, creating spurious exhaust, after contests are actually over.

7. The ACT system may even help address exhaust issues partly by discouraging minor parties from needlessly running full slates of candidates.

8. Including more votes makes elections more inclusive!  We should do it. 


-------------------------------------------------------------------------------------------------------

(Warning: as might be guessed from the above, this piece is especially numbery and has been rated 4/5 on the Wonk Factor scale.)

Yesterday the House of Assembly Select Committee released its report into the Greens' Bill to restore the House to its pre-1998 size of 35 seats (five divisions of seven).  The select committee (3 Liberal, 2 Labor, 1 Green), apparently unanimously, recommended that the Bill be passed.  Both major parties have already poured cold water on this with Deputy Premier Jeremy Rockliff repeating the mantra that it is not a priority at this time, while Labor have played their usual game of being in principle in favour of things but it never being the right time.  It remains to be seen if we even see the Bill again in this term.

The committee also recommended "That a Joint Parliamentary Inquiry be established in this term of Parliament to develop a preferred model that provides for dedicated seats for Tasmanian Aboriginal people in the Parliament."  It will be interesting to see where this goes as Tasmania's Hare-Clark system is not as well placed to address any disproportionalities created by special seats as, say, New Zealand's MMP (a system which I'm not a fan of otherwise).

An issue I've been strongly concerned about with the current Bill is the likelihood of an increase in a certain form of unintentional informal voting in which a voter omits or repeats numbers by mistake.  Currently, a voter must at least include the numbers 1 through 5 once and once each without error to that point, or their vote is invalid.  (Any errors after 5 can cause the vote to exhaust prematurely, but don't make it informal.) In the days of the 35-seat system, that requirement was 1 through 7.  I don't think we should throw away a voter's first six preferences just because they are not very good with numbering and mistakenly double or omit the number 7. To do so is to discriminate against voters who struggle with number skills.  It is also generally bad practice when changing an electoral system to change a previously formal vote to an informal one.  I quote from my submission to the committee:

"Currently a voter must number at least 1-5 without omission or repetition. The Bill would amend
this to 1-7. This would be very likely to increase the rate of unintended informal voting. Historical
evidence shows that there was a sudden jump in informal voting under the old 35-seat system in
1982, and that at the last five elections under that system the informal vote averaged 5.38% and was
only below 5% in one year. Since the reduction of the House to 25 members, with the resultant
change from being required to vote 1-7 without error to being required to vote 1-5 without error,
the informal voting rate has averaged 4.55%, and has not been above 5% in any year. 

[..]

On this basis, a change in the formality rules to require seven boxes to be numbered is likely to increase the informal vote rate by between 0.5% and 1.0% of the total vote. This is a significant increase and should be avoided. If the Bill is to proceed at any time, it should be accompanied by ACT-style savings provisions such that, whatever the ballot instructions, any vote marked with a

unique 1 is “saved” as a formal vote up to the point of the first error.  (I would also support this
change being made within the present system.)"

I also argued in favour of adopting ACT savings provisions in a submission to the government's ongoing review of the state electoral act.

It's extremely "niche", but in this article I thought I'd have a detailed look at the history of unintentional informal voting in state elections back to 1992, when the Tasmanian Electoral Commission started producing very helpful and detailed informal ballot paper surveys in its annual reports.

In the table that follows, the percentage figures are the votes declared informal for repeats and omissions after position 1 as a percentage of the formal vote.  From this point on I will call such votes saveable informals.  So for instance the total of 1.90 for 1992 means that in addition to the formal votes that were counted, an extra 1.9% could have been added if all votes with a unique 1 were included.  Breakdowns are given by the main parties, and also for two groups of others.  OTH (F) is other candidates running as part of a full slate of candidates, so that if a voter voted only for the party of one of those candidates, their vote could count.  OTH (P) is other candidates who only belong to a partial slate.  Voters for such candidates would need to include multiple columns (or in rare cases all the ungrouped column) for their vote to count.

The Notes column shows cases where parties ran slates with more candidates than needed for formality (eg ALP 4x8 means the ALP ran 8-candidate slates in 4 divisions).  It also shows the identity of all non-Liberal/Labor/Green parties that ran full slates, and notes some notable independent candidates. (Click for slightly larger clearer version).


Here are several observations on these results:

* As noted above, in 1992 and 1996 under the 35-seat system, saveable informal voting was higher than in the later years with 25 seats.  This is consistent with, and the probable cause of, the informal voting rate being higher in general under the 35-seat system.

* The saveable informal rate in 1996 was much higher than in 1992 despite both elections being for 35 seats.  The only thing I can think of that might have caused this was closeness to the federal election held the week after, but it's not clear why that would have done so.

* In general, the saveable informal rate for Labor is markedly higher than for the Liberals.  However, there are three exceptions to this: 1996, 2006 and 2018.  For 1996 see below.  For 2018 a likely explanation is that Labor gained swings in inner-city areas from Greens voters while losing some working-class voters to the Liberals. 

* In 1996 Labor's rate did not increase as much as the Liberals and Greens.  It's possible that a reason for this is that by running eight candidates in most divisions, they managed to reduce the rate of voters voting 1-7 Labor and then mistakenly putting a 7 in another column (something that could happen if the voter started numbering from the bottom up, perhaps.)  However running more candidates than needed also creates a leakage risk for a party - that some voters will, for instance, vote 1-7 and leave the party's eighth candidate blank.  Notably in 1996 only 32% of Labor's saveable informals came to grief after position 5, whereas 42% of the Liberals' did.

* The rate is usually much higher for candidates running on partial slates than candidates running on full slates.  There are several examples of this where the same party ran a mix of full slates and partial slates.  The worst thing a party can do is run one candidate too few.

* A striking exception to the very high rate for candidates on partial slates came in 2010 when Andrew Wilkie polled a significant share of the "others" vote running as an independent.  Wilkie's 2010 state campaign mainly attracted splitters from the Greens and moderate "Turnbull Liberals" and his vote was concentrated in southern Clark booths that are very electorally literate.  Wilkie's saveable informal vote rate was just 0.43% of his formal vote.  In contrast, for the Bruce Goodluck/Catherine Goodluck ticket in Franklin in 1996 (Bruce Goodluck was elected), the rate was 3.3% under the old system.

* There is some pattern for major parties to be more affected by saveable informal voting in years when they poll badly.  Probably in those years voters for some candidates of the party are more likely to cut across party lines and make errors in the process, or to stop early.

These variations only make a small difference to the results and do not seem to have changed the result of any between-party contest from 1992 onwards.  In 2006 Kim Booth won the last seat in Bass by 136 votes after his Labor rival Steve Reissig was crushed by leakage in the final stages of the count.  Had Greens voters been as prone to saveable informals as Labor voters in that contest, the Greens would have started with 33 fewer primaries.  Mostly when I have looked at the disparities created between parties by saveable informal votes, it comes out at only a few dozen votes (and savings provisions wouldn't pass on all of those.)  However if there was a contest between a fourth-party right-wing candidate not running on a full slate and a Green, the disadvantage for the former could be over 100 votes.

What could be the impact of requiring seven places?

The informal ballot survey includes just the two elections under the old system, and their results were rather different from each other.  That the saveable informal rate would probably go up (at least in the absence of an extensive and unusually effective voter education campaign or improvements in ballot paper design) is not something you need an informal ballot paper survey for - more boxes required equals more informals is a standard result from federal elections with similarly strict formality rules.  What is more interesting is whether the impact would be unevenly felt by party.  Taking the averages of 1992 and 1996 would suggest yes (with Labor most disadvantaged and the Greens least) but this is not a reliable conclusion given that the 1992 and 1996 patterns are so different from each other.

There is the possibility that returning to seven places would create higher informal rates than normally seen in the 35-seat system.  Some voters might vote 1-5 and stop because they were used to doing so under the current system.  Secondly, some voters might confuse it with the Senate and vote 1-6 and stop (this appears to be a problem in federal lower house seats with eight or more candidates.)

Minor parties generally would be very happy to accept a requirement to number seven boxes even if they were disadvantaged a little on informal votes, because the reduced quota would so much increase their chances.  However the reduced quota would not always help independents (eg Andrew Wilkie in Denison 2010 almost won under the 25-seat system but would not have been close under the 35-seat system.)

ACT savings provisions and exhaust

In my submission I suggested that Tasmania should reduce the impact of saveable informal voting by adopting the ACT savings provisions and actually saving those votes.  Formality in ACT elections works like this:

* The voter is instructed on the ballot paper to "Number five boxes from 1-5 in the order of your choice.  You may then show as many further preferences as you wish by writing numbers from 6 onwards in other boxes"

* If the voter makes an error inside the first five but has a unique 1, their vote is still counted, but it exhausts when it reaches the point of the error.  So if a voter votes 1 for an obscure candidate and two twos, their vote is initially counted for their first candidate, but then exhausts when that candidate is excluded.

* There is a special provision for surplus transfers to reduce the level of exhaust.  When a candidate goes over quota, ballot papers that would have next gone to exhaust are fully retained with the candidate as part of their kept quota, while ballot papers showing further preferences are passed on (provided that their value doesn't increase).

In 2016 the savings provisions increased the number of formal votes in the ACT by 1.23%, compared to what would have been a 0.77% increase in Tasmania in 2018 assuming no-one voted differently on account of the savings provisions.  The savings provisions themselves might well contribute to the higher rate in the ACT (eg if some voters know you can get away with not numbering all five boxes) but other possible causes are the ACT's lesser experience of Hare-Clark (adopted in 1992, and the ACT does not have Council elections using it) and also population exchange with NSW which has optional preferential voting in state elections.  The savings provisions clearly don't have the effect of flooding the system with huge numbers of prematurely exhausting votes.

The TEC's understandable concern about introducing greater savings provisions is increased exhaust, and the Tasmanian Electoral Officer (see report page 117) cited the very high exhaust rate in ACT elections in comparison.  But the apparently very high exhaust rate in ACT elections is an artefact of the ACT's computer counting system.  The Tasmanian counts stop once there are only six candidates standing and all the surpluses of candidates who are over quota at that point have been thrown (Bass 2018 example here).  The ACT computer continues to exclude the sixth candidate, distribute their preferences and then distribute any surpluses beyond that point although all this has absolutely nothing to do with who has won the election!   In 2012 in the old five-seat Brindabella division, the exhaust rate is shown as a massive 15.1% (9630 votes, 0.91 quotas).  However the actual contest was decided when, with four candidates already elected with quota, Andrew Wall (Liberal, 8893) defeated Amanda Bresnan (Greens, 7761) for the final seat.   At this point only 3512 votes (5.5%, 0.33 quotas) had reached the exhaust pile.  A Tasmanian count would stop here.  The ACT count continued by throwing Bresnan's votes with the unsurprising result that only 1648 votes flowed onto Wall and the rest went to exhaust.

In the 2016 ACT election this also created substantial differences, albeit less dramatic:


The effective exhaust rate in the 2016 ACT election (.41 quota average) was basically the same as Tasmania's for the last two elections (.42 quotas in 2014 and .40 in 2018).   However, Tasmania's rate had been much lower in 2010 (.24 quotas) partly because there were no significant fourth party candidates except for Andrew Wilkie, who wasn't excluded.  A range of factors influence exhaust rates heavily, including the structure of particular contests and especially whether there are significant fourth parties running full slates of candidates.  (When there are, some voters vote just for one fourth party ticket then exhaust their vote.)  The ACT system actually has a possible advantage in managing exhaust when there is a high fourth-party vote, as there was in 2016 (way above the last two Tasmanian elections).  The ACT system doesn't create any reason for such parties to run full slates, so they often don't, which makes it more likely that voters for minor parties will distribute preferences outside the ticket.  In the Tasmanian system we have seen high exhaust when full slates for parties like Tasmania First, Palmer United, Lambie Network and Shooters Fishers and Farmers are completely excluded.

But even if allowing ACT-style savings provisions would produce a modest bump in the Tasmanian exhaust rate, I'm not inclined to be concerned.  The example given by Andrew Hawkey was the very close Woodruff-Street contest for the final seat in Franklin, where the Greens' Rosalie Woodruff was three votes behind the Liberals' Nic Street before getting 229 more surplus votes at the final preference throw (the surplus of Labor's Alison Standen).  Obviously if virtually all Standen's surplus exhausted, the outcome would have been different.  But in fact just over half of the surplus did exhaust, as is common when the last candidate for a major party is elected or excluded.  I would gladly bet that the votes that did continue on at this point were mostly doing so because the voter had marked more than five preferences, and that many of these would have started 1-5 Labor.  The sort of voter who numbers all boxes for one party then numbers another party and so on would be very unlikely to stop voting at 3 just because they find out that they can.

Does it all matter?

To results, probably not.  Whether an expansion to 35 seats resulted in stricter formality rules (as it appears the TEC would so far prefer) or more relaxed ones (as I would prefer) than at present isn't likely to change a seat result any time soon.  Nor would relaxing the rules within the existing system be likely to change any outcomes.  If it does change a seat result in some rare case then there will probably be much bigger regulation debates affecting that - advertising limits, spending caps, rules on misleading material and so on.

I think it's about the message that high informal voting sends, and the less inclusive voter base resulting from disallowing votes unnecessarily.  If someone has views but struggles with numbers or concentration, do we say, yes, we will still listen to their views to the extent that we can make sense of what they have said?  Or do we say that because we are determined to make people keep their votes in play longer, this person is an undesirable voter who must be fenced off until they get it right?

I think that the system is already skewed enough in favour of highly numerate voters, who can already punch above their weight by expressing more preferences or even through strategic voting.  I'd like to see us make more effort to hear the full range of voices.  For that reason, I suggest that we not even wait for the politicised 25-vs-35 seat debate to play out, and introduce ACT style savings provisions now.

No comments:

Post a Comment