I had the opportunity today to visit the Bradfield recount, thanks to being appointed as a scrutineer. I did not attempt to challenge any ballots, but this gave me the best way to see how the process works. I was thinking of doing a podcast, and I may do that at some point if I get to interview some of the scrutineers, but for now I thought I’d summarise what I found in a blog post.
Firstly I should say that the process was careful and meticulous, with AEC staff and scrutineers all playing their role carefully and taking everything seriously. I do think if more people could see the process it would increase faith in democratic processes. While most of the time we just see the numbers slowly and occasionally changing, a lot of work goes on behind the scenes.
So the Bradfield recount is being held in a warehouse in Asquith that appears to have been used for a number of northern Sydney electorates, but today only Bradfield was being counted. A space is set aside as a break room for scrutineers, with large teams for the teal and the Liberal candidate. Each candidate had at least enough scrutineers to have one on every person doing counting.
The regular counting space is divided into a series of bays. Each bay is dealing with one polling place at a time. It appears that a bay deals with a polling place until it finishes, then moves on to another. Some are bigger than others, to deal with booths of different sizes.
The first part of the process is to conduct a fresh first-preference count. Ballot papers are already bundled up in first preference order, so the votes previously assigned to each candidate are dealt with one at a time. For some of the bigger bays, there might be tables just dealing with Boele votes or Kapterian votes, and another table dealing with the other candidates.
Ballot papers are usually bundled up in bundles of 50 votes, wrapped in a rubber band, to simplify counting. Each counter takes one bundle at a time and removes the rubber band, and then carefully checks each ballot to ensure it is formal and that the first preference is correct. It is then laid on a pile facing the scrutineers, who can observe it. This process is happening more slowly and carefully than you would expect for an election night count.
Once each bundle is fully checked, the rubber band is put back on, and they keep going until they’ve checked all the votes for that candidate. Once this is done, the counter will take each bundle and hand-count to verify each bundle of 50 is a bundle of 50, and then the other counter sharing that table will swap bundles and do the same check. In theory they could find a mis-counted bundle of 49 or 51 but I did not see this happen.
During this process, scrutineers are free to challenge a ballot. If so, the ballot is put in a box to be referred to the Divisional Returning Officer (DRO) for review. This happened with a decent number of votes – a few dozen for a normal booth size – but not excessively. I should also say that each candidate appeared to be entitled to have one scrutineer per counter. So at a table with two counters, there would be two Boele scrutineers and two Kapterian scrutineers.
Once the primary votes have all been checked, the informal pile is also checked. And then the box of challenged ballots is reviewed by the DRO.
After the DRO review each booth will then re-do the distribution of preferences (DoP), step by step. Presumably the final figures from this distribution is then updated as a fresh 2CP figure, and can change the margin. It seemed like smaller numbers of votes were also challenged through the steps of the DoP.
The DRO reviews were usually observed by the most senior and experienced scrutineers. The DRO considers each ballot carefully, and makes a ruling about the position of the ballot (who gets the first preference or informal), and stamps the back and fills out a little form explaining their decision. A scrutineer at this point can refer a ballot to the Australian Electoral Officer (AEO) for a final decision.
The AEO is the chief AEC staff member for the state, and is effectively the final arbiter for the recount process.
Every now and then, the AEO has a session with one scrutineer representing each candidate. The AEO considers each ballot referred up for adjudication one by one. The AEO has a magnifying glass and will carefully consider each ballot in line with the AEC’s formality principles.
I found that both the DRO and AEO are careful and cautious, but also very clear on their priorities. The formality principles require them to construe the ballot paper as a whole, and err in favour of the franchise. This means that sometimes when a number is not entirely clear, but context clues make it clear that, for example, it would make sense for that to be a 4 rather than a 7, then they may interpret it that way, but it needs to be clear enough to be sure which number is the 4 and which is the 7. I found officials who apply rules that can be very strict, but while giving a ballot the best chance of being counted.
I found that the scrutineers mostly left it to the DRO and AEO at this stage to make their decisions without much argument, although occasionally they’d politely make a case for a particular figure representing a particular number. There definitely wasn’t arguments or shouting. You wouldn’t know these people are in a fierce recount coming down to a handful of votes.
The vast majority of ballot papers are clearly written and have a clear sequence of 8 numbers, and are dealt with swiftly. But watching these contested ballots, it was frustrating to see votes that clearly attempted to express a preference ending up informal. In some cases it seemed like a voter made a mistake by losing count of where they were up to – a ballot might have unique numbers 1-5 and 7, but two sixes instead of a 6 and an 8.
In plenty of other cases, the culprit was bad handwriting, making it hard to tell which number was which. The AEC officials did their best to fairly determine the correct answer, including for some ballots that I doubted would be admitted to the count, but ultimately sometimes it’s too hard. Voters, try your best to write the numbers clearly and distinctly!
But in a lot of cases where votes were made informal, it was perfectly clear who they preferred between Boele and Kapterian – the confusion was irrelevant to the ultimate outcome of the race. Even if we don’t go to full OPV, I think we need to find a way to ensure these votes can be admitted to the count.
While we were watching votes being admitted or rejected, it was not possible to know if the margin was changing, because we didn’t know the previous status of a ballot paper. But the data suggests very few votes are changing from one category to another.
Following my return from the count, I’ve downloaded more media feed data to see how the lead has changed:
Monday and Tuesday were good days for Kapterian, extending her lead from 8 votes to 14. But today has been a good day for Boele, reining in Kapterian’s lead from 14 votes to just 5. This race remains extremely close.
The AEC appears to be making good progress, but there is still a lot of booths yet to be recounted. According to the AEC’s website, 19 out of 65 booths have updated their first preferences this week, and 19 booths (seemingly the same ones) have also updated their two-candidate-preferred count. This suggests that we are on track for the recount to finish late next week.
The experience of scrutineering the count is the kind of thing that should be mandatory for those people who carry pens in to vote.
There was one ballot where the person had filled it in with a pen, but their handwriting was so bad that their vote was informal.
I noticed at prepoll that the ballots were being individually printed.
I think the logical extension to this would be a system where the voter is marked off and sent to a booth where the ballot is on a touchscreen and you touch in your order of preference until the designated number is arrived at … obviously a clear and start over button would be there too!
Once completed, you just hit print! The device could, but, initially I’d suggest, wouldn’t be networked or used for counting – just a single function printer.
Now, how do I prevent someone printing more than one (say a code given on being marked off?) yet retain anonymity? That’s my dilemma! But it would eliminate any informal votes!
Thank you for writing so convincingly about the very careful and thoroughly trustworthy the process of counting votes in Australia is. There is misinformation being spread online – via Facebook and other places – throwing doubt on the honesty of our elections. This nonsense is coming from people who don’t value democracy, because they don’t like the result. Thank you for your excellent work in helping people to understand how lucky we are in our democracy.
Knight is that you Josh?
Now just 2 votes
The Big Crunch will come when they open the St Ives pre-poll packages
There are 5489 formals and 312 informals (5.3%) in those packages compared with the total of 1901 (formals) for “others”.
St Ives PPVC showed the biggest flip on the DOP
If there are similar errors in the primary count and in the same direction, then Boele is done for.
Thanks Ben for writing this up. I was involved in scrutineering at another electorate and my experience was very similar. the AEC staff are incredibly meticulous and careful, and I have enormous respect for them and the process they follow. I do wish that someone could create a documentary – everyone needs to see the process in action, and know that they can trust it.
GeoffL:
Wouldn’t the St Ives PPVC votes be good for Boele. Kapterian has a primary of 50% in St Ives PPVC. More chance for rulings of informal votes. Genuine question. Thanks.
My view is that people who make comments on social media should be able to be referred to the AEC for acceptance as Official Election Monitors (TM) and Democracy Participants (TM) on election night. Given that nothing really happens that night it would mean I might actually get a break from scrutineering and get to the after-party before all of the food and alcohol is gone.
@Ben Raue – You don’t have to go full OPV, you just have to extend the Savings provisions to allow a vote to be valid up to the point of dispute. So if you’re not able to determine whether it’s a 2 or a 5, it’s still formal for the 1. If it’s missing the 6 in an eight candidate ballot then it’s still able to be distributed to the first 5. If you can’t distinguish between the 1 and the 7, then it’s completely informal.
@J Knight – That was part of my submission to the JSCEM for the 2019 election. Basically using the same technology as the Powerball printers with a paper receipt. That way it removes the possibility of an accidental informal vote but still allows for a deliberate informal vote, with a process to reconcile the count in close seats and the occasional random electorate for integrity purposes. It also make scrutineering in a recount much easier. Ensuring ballot integrity isn’t that hard – sign off on your name, get given ballot paper, slide ballot paper into ballot machine to unlock it, fill out guided ballot, confirm and print, drop completed paper in ballot box. That way you still keep anonymity with a digital count and reconciliation with a paper copy.
Is there any obvious reason why some days are “good days” for a candidate, and others are “bad days”? If errors are completely random, one might expect, with a low probability of error and a large number of ballots being handled each day, that errors over a day would largely cancel each other out. Or are there more errors in both directions than might have been anticipated? (This ties in with some theorems in probability which go back to the 19th century https://en.wikipedia.org/wiki/Bertrand%27s_ballot_theorem.)
@GeoffL, part of the swing to Kapterian in St Ives PPVC during DOP was due to the identification of a prior transcription error. I’m not sure we can predict much from DOP as a result.
@Haddy I also expect St Ives PPVC to be better for Boele for the same reason. But we won’t know for sure until they get around to doing the full recount. There was one significant error with 11 preferences being incorrectly allocated to Boele when they should have gone to Kapterian, a 22-vote swing. Another error of the same or greater magnitude might exist which would render everything else irrelevant.
Moving away from hand-marked ballots would probably eliminate 75% of the errors. There are machines available where a ballot paper would be inserted into a machine. On the screen, all of the candidates are listed. You start marking your preferences 1, 2, 3, etc. The voter would see the order displayed. After all selections have been made, the machine will then ask to confirm your choice. If you click ‘Confirm’, the machine will print a tally of your selections. The ballot would then exit the machine, and you would deposit it in a tally box. The software would also detect mistakes, prevent the voter from entering a duplicate preference, and would not allow the voter to skip a preference (such as going from 3 to 5). However, if a CPV with a savings plan is used, the machine would allow you to confirm after the necessary number of selections has been made. When the count begins, the ballots will be inserted into a machine, and the votes will be recorded electronically. Once the count reaches a certain point, the preferences will begin to be allocated until the final vote is reached. If the vote falls below the required threshold for a recount or if there is a physical error in the process, then the ballots could be tallied manually, as the preferences would be printed on the ballot paper. The short and long of this is that the machine would eliminate deliberate informal voting and eliminate most accidental informal votes. There will never be a perfect system, but a 1-1.5% informal rate for most jurisdictions is probably a practical target.
I am not sure everyone would be happy to adopt a system for the House similar to the one used for the Senate. Since most government business originates in the House, there is something to be said for a political party having the ability to command a durable majority. Hung parliaments are never an ideal situation, and we do not need to repeat 1975.
Hi Ben, please feel free to contact me for more info on scrutineering, I’ve done it for many years now on election night, and more recently also at the central divisional location, including for the full preference distribution. I can probably fill in a few extra things for you, including error rate, why a skew in one booth isn’t indicative, and why the process is done in paper.
The info won’t be Bradfield specific though, as I’m based in Queensland.
It woud be unusual to find a miscounted bundle of 49 or 51 [outside of a large prepoll] because if the number of votes counted at a booth on election night doesn’t tally with the number handed out, the bundles are recounted. If there’s still one missing after a recount, the assumption is a voter walked out with their ballot.
Apart from that, the bundle error is picked up on the night, unless there is a matching error both ways.
From memory, both Mirabella in Indi and Palmer in Fairfax in 2013 had turnarounds in the order of a thousand and three thousand due to bundles being allocated wrongly.
That might be solved by not having the 2CP counted at the same table next to one another. Sure, the parties would need an extra scrutineer, shortage of scrutineers seems to be a problem for the Liberal Party.
In lots of countries around the world, majority governments are practically unheard of, and the democracy still functions smoothly. Hung parliaments aren’t the terrible, chaotic things the media paints them as. The 2010-2013 term was one of the most productive in terms of legislation passed in recent history.
Clarinet of Communists is completely wrong, hung parliaments are some of the most unproductive, undemocratic applications of the niches of the electoral system that exist. Hung parliaments results in parties on the fringes of the political spectrum who pick up tiny fractions of the vote influencing federal legislation and restricting the mandate of a party of government from governing in their own right. In the past parties like the Greens have sought to block important supply legislation which impinges on the financial initiative of executive government. They have no sense of the responsibility that comes with governing, they seek only to grandstand, posture and politicise instead of working through policy and negotiating. Australia strikes the perfect balance between an inherently unfair FPP system like in the US or UK, and an unworkable, unrepresentative system used in NZ.
They’ve never been given the chance to govern!
The closest they got was in 2010, and it was the most productive legislating relationship with the ALP and the Greens ever had.
If the Greens (or One Nation) were given some governing responsibility they would either rise to the occasion or be punished for failing to do so. But they’ve never had that chance.
@ Mark Yore:
“If you can’t distinguish between the 1 and the 7, then it’s completely informal.”
European 1s and 7s is sometimes addressed by the RO before counting starts.
If there’s only 6 candidates, then it’s probably a 1, particularly if it mirrors a HTV.
7 or more candidates, the voter’s intention would be clear if a HTV was mirrored on one interpretation and on the other it made no sense at all, imo.
Clarinet of Communists:
Sure, Gillard and the Indies got a lotta legislation through, that doesn’t mean any of it was good legislation and some may have been quite harmful.
An example might be legislation preventing Rare Earths from being processed in Australia. Since the feedstock is mullock piles from old hard rock mines, Australia would benefit from a Repeal, but what chance of that happening with the new Senate?
Okay, ignoring the first few words there (and your sad attempt to plagiarise my name), ‘hung parliaments’ are just governments where one party can’t run roughshod over our democracy. The point of democracy is to represent people, so the more people’s preferences you represent in government, the better. And as to ‘unproductive’, here’s some evidence from the ABC:
“For all the noise and controversy, the 43rd Parliament can claim to be productive. (Whether you agree with the outcomes is a separate matter.)
The big-ticket items like budgets, the carbon and mining taxes, plain packaging laws for cigarettes, paid parental leave and the Murray Darling Basin Plan all passed.
In fact, by the end of last year, 432 bills had become law.
By comparison, the Rudd Government passed just 409 bills in its entire first term, while the Howard Government passed 549 bills from 2004 to 2007 – and it controlled both chambers for much of that time.”
I think there shouldn’t be any debate that the 34.6% of the country who voted for Labor should have to satisfy to a bare minimum some of the 33.6% who voted for neither major. That’s much preferable to the 34.6% having completely free rein over the government.
If we’re talking about people not interested in working with policy and negotiating, look at what happened to the environmental laws last term. Plibersek and the Greens had reached a deal, and Albo came in over her head, at the behest of the fossil fuel industry, and called it off. And just to point out, the Greens have never brought a government down with balance of power by blocking supply.
As to the ‘perfect balance’, I think that this system is also inherently unfair, and while there’s no perfect electoral system, something proportional, like in many European countries, or something semi-proportional, like NZ, is good. As to NZ, whatever you think of the current government, the portion of voters who voted for a governing party is far higher.
Gympie I’m not saying everyone loved the legislation, only that it wasn’t as ‘unproductive’ as people say.
@Gympie,
One of the formality principles is to consider the ballot as a whole. Other principles are that they should assume the voter attempted to vote formally, and to err in favour of the franchise. I don’t have the exact words, they had them on a poster on the wall but I wasn’t allowed to take photos. You can google them.
So yes, if there’s only six boxes and there’s a number that’s either a 1 or a 7, and there’s no other 1, then that’s going to be interpreted as a 1.
Likewise, if there is eight boxes, numbers 2, 3, 4, 5, 6 and 8 are very distinct and unique, there’s a clear 1, and another number that could be a 1 or a 7, they will assume that number is a 7, and thus the ballot is formal.
I didn’t see any evidence of AEC workers taking into account whether a particular preference order made logical sense in terms of the political affinity of the candidates or what the HTVs had said.
Ben Raue they clearly did get the chance to govern. They rejected a perfectly good CPRS because it wasn’t “perfect” as as a result we’re still dealing with climate wars to this day.
What they got was the chance to agree or disagree with a Labor bill. That’s not governing. If Labor wasn’t so intent on just staring the Greens down, we might’ve got good climate legislation.
I never met a Returning Office (RO) I didn’t like
I have been tick-tacking with ROs for nearly 50 years and scrutineering for 20 years.
Peter Brent is an expert par excellence on ROs. His PhD thesis (originally to be called “The role of the Returning Officer in Australian Life”) was the first of some 50 articles he has written on matters relating to the AEC. He invited me to be his examiner for the thesis, but I was unable to for various reasons.
The AEC staff are much underappreciated, but I have always thanked them for the work they do. As a National Tallyroom Groupie (Peter Brent and Bill Bowe were NTR Groupies too), it sometimes took me half an hour to thank them as they were knocking down the NTR at 11PM. I do the same when scrutineering at a booth. They are invariably chuffed by this.
In inter-election periods they are also helpful to me at office visits, by phone and by e-mail.
Scrutineering is by far the most fun part of election day. I am pleased Ben enjoyed it but a bit taken aback that he had never seen done it before.
At many booths the scrutineers from all parties know one another from previous elections and the atmosphere between them is invariably civil and helpful. There is lots of banter. Even the ROs are entertaining. Sometimes one might say “Hey … I found one!” it was a drawing on a ballot paper – can you guess what of?).
My wife, however DID meet an RO she didn’t like. That was in 2013 at French’s Forest booth, just down the road from Tony Abbott’s electoral office. The poor RO was being hassled by Tony’s honcho scrutineers to pass invalid Abbott informals and to block other candidates’ formals. Eventually he was persuaded by my wife to phone the DRO for clarification. He returned sheepishly and said to Abbott’s scrutineers “Gentlemen we have to start again”. Sheer bliss!
The Greens were not in a position to pass the CPRS, and at no point were offered a chance to negotiate. What they were offered was a take-it-or-leave-it offer to pass the legislation Rudd had negotiated with Turnbull.
They then did get the chance after the 2010 election when they actually had the numbers and they passed a whole bunch of stuff.
Ben, so does that mean the Greens in the Senate during Kevin Rudd’s term were not given the opportunity to add amendments etc to the legislation.
I thought all MPs and Senators have a right to add amendments if they see fit. In the recent term with the housing legislation, the Greens chose not to exercise that right, instead issuing a blanket order to delay the legislation which was seen as a stunt.
https://www.elections.act.gov.au/elections/our-electoral-system/elections-in-the-act/technology-assisted-voting-and-counting/electronic-voting
No one seems to have mentioned that the ACT has an electronic voting option for its Territory elections. Used only for prepoll – so far too hard to have working for mass numbers on the final voting day. It is an option – not a requirement. I think around 10% use it and the results from that 10% are so close to the end point that most ACT elections are known at 6.05pm election day.
A point I cannot test on the material available is how to attend to vote but not, intentionally, cast a countable ballot. It may be you need to use the paper option and not complete it.
back up to 3 votes
@Gympie I was referencing a vote where there was the opportunity to put a 1 and a 7 on the paper and the written numbers were not able to be distinguished. You might be surprised how often that happens – 3s and 8s, 2s and 5s. The rules we normally rely on when scrutineering are “Is the intent of the voter clear?” and take it from there. One of the things you can’t use is “Does it follow the HTV?”. People vote in all sorts of mysterious ways and it’s not unusual to have 1 LNP 2 ALP or the reverse. The Longman byelection, for example, had preferences all over the place. Sometimes you look at the completed ballot paper and it’s very difficult to wrap your head around the type of person who would vote that way.
@GeoffL I agree 100%. I have noticed in recent elections that the amount of additions on the ballot paper has declined. I rarely see an Agro, or a laundry list of a candidates shortcomings. Even random penises no longer pop up as much. The best I ever saw was a full pencil sketch of a female in profile on the back of a Senate ballot. It was incredibly detailed and must have taken quite a bit of time. Unfortunately the ballot paper on the other side was blank.
And yes, because the people who scrutineer are generally more experienced there’s much more of a convivial atmosphere there. In the 2022 Griffith election I was scrutineering in one of the big booths and the two Greens who were there had never done it before. When it became obvious that the Labor vote was crashing, the ALP scrutineer and I did a quick training school for the Greens while the RO was waiting for instructions.
@Yoh An,
Anyone can offer amendments any time they want. But the path to a majority not involving the Coalition required the Greens, Family First and Nick Xenophon, so understandably the government only negotiated with the Coalition.
I don’t know what you’re talking about with the housing legislation, the Greens achieved a number of changes to the legislation.
Fair point Ben, the 2007-2010 Senate was unfavourable for Labor given that they needed to negotiate with Steve Fielding and Nick Xenophon in conjunction with the Greens. At least the 2010-2013 Senate and the new one starting 1 July are better for Labor as they only need to work with the Greens, not any right-wing minor parties to achieve a majority.
down to 2
My experincing of handing out HTVs and screw in earring goes back to 1966.
I have only had one RO not know what he was doing (not in Canowindra) and had to explain how the polling booth count should proceed if we were to get home before dawn.
I have often had to explain to conservative scrutineers that party HQ would be only interested in the first pref counts, and where minor party preferences were allocated. And that only the most egregiously informal ballots should be challenged, because they’d be all picked up later.
7 votes in Gisele’s favour at the moment.
Could you count a vote to the point where the selection becomes unclear? That is 1,2,3 but if 4 is unclear then the 4th preference is not counted and the becomes informal . The majority of votes would not be counted the the end of their preference selection
With neither candidate having a double-digit lead over the other this week, it might have to go to the Court of Disputed Returns.
@Michael Maley I know the answers to that!
The first is mathematical. Votes tend to fall in what’s described as a Random Walk. Say the odds of an LNP or ALP vote are equal – 50%. So the odds of six ALP votes in a row is theoretically (.5 x .5 x .5 x .5 x .5 x.5), or 3.125%. But that’s not accurate, because at each stage of the decision tree there’s a 50% chance it can go back to the centre of the count and a 50% chance it can head further out. Because there are lots and lots of opportunities for six of the same votes in a 100,000 count it can look like there’s an unsurmountable shift towards one candidate or another. If you’ve had 27 ballot papers in a row for the ALP in a marginal seat then the odds of the next ballot paper being ALP is exactly … 50%. That’s because each vote doesn’t care what the previous vote was.
The second is psychological. Human are accustomed to look for signs and portents in everything. That’s why so many people believe in astrology, or lucky numbers. They look for trends that reinforce their beliefs and ignore those that contradict them. That’s also why you see people claiming victory on election night, because they’re surrounded by people who want their candidate to win.
The third, and most important for voting, is non-homogenity. Votes come in globs, which may be completely different to the next glob that comes along. Normally the first booth to post figures is the smallest, for obvious reasons. In regional areas that is generally the most rural group. At which point every commentator attempts to divine the result from one tea leaf. As other booths come in from different areas they tend to even out the average election day vote. About the time those are done then the first of the pre-poll booths start coming in, which have a completely different type of voter. And finally you have the postal votes, declared institutions and other declaration votes. Somewhere down the track, if you’ve got a large military electorate like Herbert, those votes will get added all in one shot. As Ben’s very helpful maps show, different areas within an electorate can have very different voting patterns.
Finally you’ve got the distribution of preferences, which can be fairly unreliable. In order – National, then ALP, then Liberal – are the most likely to follow the party HTV. However their preferences are almost always the least likely to be counted. The Greens feed anywhere from 70% to 80% to the ALP ahead of the LNP, but that still means a net vote of 0.4 to 0.6 of a vote to the ALP. So when people add the Green votes to the ALP they’re always shocked when the ALP polls less than they expected.