Recount: Recount of above the line and informal votes granted Oct 10. Recount ran from Oct 17 to Nov 1. Various corrections were made but 1370 votes were lost, could not be rechecked, and were excluded from the recount.
Declared Result: Ludlam (Green) and Dropulich (Sports) declared winners by 12 votes at key exclusion.
But There's A Problem: The lost votes would have advantaged Pratt (ALP) and Wang (PUP) by 13 votes, meaning that if there were no errors in the counting of those votes, Pratt (ALP) and Wang (PUP) could have won by one vote.
Where to from here? Outcome referred to Court of Disputed Returns by AEC requesting that election be ruled void. Losing parties may also petition seeking the same or that they be declared winners. Full state by-election likely.
Updates to this post appear with the newest at the bottom.
|Democracy in action! Screenshot of the critical point of the preference distribution.|
Just a post to explain what is going on with the Western Australian Senate count to those not already aware of it. The button was pressed today on the most exciting of the eight Senate post-counts.
Through the complex process of modelling the outcome undertaken by Truth Seeker, Antony Green and others, one point of the long list of exclusions of minor candidates became increasingly crucial and incredibly close. That point concerned which of the Shooters and Fishers and the Australian Christians was excluded first. Neither of these parties had the slightest chance of winning, but because of their different group voting ticket preferences, which one was excluded first would have a great impact on the outcome. If the Australian Christians were excluded, this would then lead to the exclusion of the Sports Party's Wayne Dropulich, and this would in turn result in Labor's Louise Pratt and Palmer United Party's Zhenya Wang getting the final two seats. If the Shooters and Fishers were excluded, this would then lead to Dropulich winning on a preference snowball and taking the Greens' Scott Ludlam with him.
The Shooters and Fishers had a lead on the ABC Calculator, but it assumes all votes are above the line votes, and ignores the impact of below the line votes that do not follow the script. If only reliably "locked in" votes (ticket votes and a share of votes from a party's second and subsequent candidates) were included then the Christians had a wafer-thin lead. But modelling by Truth Seeker here suggested that the Shooters and Fishers were slightly more likely than not to do just well enough on below-the-lines from other parties to beat the Christians at this point. The final Truth Seeker projected margin was 54 votes to S+F, but with a large degree of uncertainty, such that "Is this significantly different to zero? No."
When the button was pressed this key exclusion was indeed resolved in Shooters and Fishers' favour by 14 votes (23515 to 23501) meaning that the cascading series of preferences that determined two seats has been decided by 14 votes out of the 1,311,440 formal ballots cast in WA; an astonishingly close outcome. See Antony Green here for a blow by blow account.
The Greens at least are expected to request a recount. Whether the Sports Party is likely to as well is unknown. Whether such a recount would be granted in full or of just some ballots is unknown to me and I will post news on the AEC's decision when I have it. A full Senate recount would be expected to be a long process potentially taking weeks. The Senate is not bound by the same rules as the House of Assembly and a margin of <100 votes at some stage in an exclusion does not automatically trigger action. Indeed at early exclusion stages minor candidates are often tied and excluded by random draw.
I have not seen the full Senate recount rules but similar general issues apply as for House recounts apart from the lack of an automatic margin. The candidate(s) appealling for recounts must specify their reasons for requesting a recount. Where possible the recount will be confined to a subset of all counting rather than all ballot papers. If there is a recount, a result is declared and the losing candidate(s) can challenge the declared result in court. A successful court challenge could lead to a new result being declared, or could potentially (if it was found that voters were improperly unable to vote, for instance) result in a full state Senate by-election for all six seats.
Update: It has been confirmed by Senator Ludlam that the Greens will request a recount of "elements of the WA count". The Sports Party is also requesting a recount. It is only necessary for a recount to confirm that the Shooters and Fishers did in fact exclude the Australian Christians, a count that involved only a few percent of all ballot papers, but I would think that doing that properly would require checking the 1 vote on every single ballot paper to determine whether any vote belonging to either of these two parties was mistakenly counted as belonging to anyone else, and also checking the informals pile for the same reason.
The AEC expects to decide whether or not to grant the recount requests within 24 hours.
Wikileaks: There has been quite a bit of comment flying around about one of the more notorious preferencing decisions, that by the Wikileaks Party to preference the Nationals ahead of the Greens.
The fact is that this, in the end, did not cause Ludlam's apparent defeat, because the Nationals were excluded, and the Wikileaks Party's above-the-line preferences reached the Greens at full value. People are remembering that Wikileaks shafted the Greens in preferencing and assuming that this caused the result without checking their facts; it didn't.
It is, however, the case that the Wikileaks Party's preferences went to the Sports Party when they could have gone to HEMP or the Animal Justice Party, both of which would seem more likely to be agreeable to Wikileaks Party supporters. Some people have pointed out that had they gone to the AJP, the exclusion of HEMP would then have resulted in the Sports Party and Greens getting up. (In this sense, the Wikileaks Party may have caused the Sports Party's doom ... by preferencing it!)
Of course, Wikileaks might just as easily (if preferencing ideologically) have decided to preference HEMP. And these sorts of silly outcomes will be found scattered through the decisions of many parties.
I suggest keep an eye on Truth Seeker here for more of the absurdities of this count. After what we saw in Tasmania, I think there will be too many for me to keep up with!
Oct 3: No word on the recount yet beyond that a decision will be made today, possibly by around 3 pm WST (5 pm AEST). You can read Antony Green here regarding a precedent for a Senate recounts. The one he mentions was under a vastly different system.
Antony also mentions recounts based on the closeness of results of a stage prior to the final stage. A similar example happened in the 2011 Hobart City Council election for Deputy Lord Mayor, in which two candidates were running very close for second and whichever finished second would win on the other's preferences.
Oct 3 5:20 pm: Have seen widespread reports that the request for a recount has been initially denied, but the denial can be appealed to higher up in the AEC before we get to the declaration and lawyers-at-30-paces stage. Declaration has been postponed until tomorrow. Going out for a few hours now; will update much later tonight.
Oct 3 9:45 pm: The AEC confirms the requests are provisionally declined:
"The requests for a recount did not identify any specific issues, which would have warranted the conduct of a recount and as a result the requests have been denied."
Importantly, you don't get a recount by just arguing that the margin was extremely close and that if it was recounted the result might change. The following is especially important in the AEC's information provided to candidates:
"A request for a recount needs to identify specific ballot papers and associated significant counting process errors or irregularities that could change the result of an election within a division."
If those asking for a recount did not specify anything of the sort, it could be because there is nothing to identify. Or it could be that their parties' scrutineering efforts have been insufficient to gather such evidence.
That said I have not seen the text of the recount requests thus far.
There is a detailed radio interview with AEC state manager Peter Kramer here.
The refusal can be appealed to the Australian Electoral Commissioner, Ed Killesteyn. This will need to occur before the scheduled declaration of the poll at 11:30 am WST and will presumably lead to a further delay in the declaration.
Put That Fairfax Analogy Down!
There are a lot of analogies involving Fairfax being waved around at present. The Fairfax situation is fundamentally and multiply different and the analogies are just not valid. Some may think that the vastly greater number of ballot papers makes a recount much more likely to shift the result here, but that is not necessarily so.
In the Reps count any vote that changes status from formal to informal or vice versa affects the final contest by one, and any vote that changes camp between the main contenders affects the final contest by two. In the Senate count, the two parties whose votes determine the crucial exclusion have less than 3.6% of the state vote between them and a given error on a ballot paper will often affect neither. If the ballot paper is at reduced value the impact of the correction may be lost due to rounding.
Secondly the House of Reps formality guidelines require the numbering of nearly all the squares without errors, so debate about whether a voter has marked, say, a 5 and an 8 or two 8s can determine the entire fate of a vote. In the Senate these situations only determine whether or not a vote exhausts at a given point. There should therefore be a much higher rate of borderline formality rulings on House ballot papers.
Thirdly the vast majority of Senate voters just vote 1 above the line, a vote which is generally not at all ambiguous as to its formality. I suspect that voters who have trouble voting neatly - for instance, voters with very unsteady hands for whatever reason - would be most likely to vote above the line to avoid the risk of their vote being misread.
So I wouldn't assume that just because the count in Fairfax has shifted by 29 votes in the examination of a mere portion of a single division, that a recount in the WA Senate would necessarily make the same, let alone a much greater difference.
Still, fourteen votes ... tomorrow will be an interesting day ...
Oct 4: House - Senate Discrepancies: Truth Seeker here reports on numerous discrepancies, in both directions, between Senate and House of Reps vote totals for specific booths in WA. While a possible explanation for these discrepancies is available (see below) the two cases of discrepancies of 50 certainly stand out. I am looking at these.
Irrespective of the ability of the Green and Sports candidates to make a decent case for a recount under the AEC guidelines, I wonder if there is anything that prevents the Australian Electoral Commissioner from employing his power to direct a recount himself under S 278. I just cannot see how justice will be seen to have been done by supporters of the losing parties otherwise.
Oct 4: An anonymous commenter on Truth Seeker's site suggests that the "binning" rate (voter fails to lodge one or both papers, either discarding the ballot in the bin or taking it out of the polling booth entirely) may be sufficient to explain discrepancies at particular booths. This warrants investigation for outliers. Meanwhile a new discrepancy has surfaced by way of the booth of Geraldton-Waggrakine where a surprisingly small number of below-the-line votes are recorded in the AEC's data displays - just one in a booth that recorded fifty in 2010.
The Australian Electoral Commissioner is considering the Greens' appeal against the denial of a partial recount. The result will not be declared today pending that decision. It is not known at this stage how long a decision might take.
The Greens' letter of appeal to the AEC has been published by Crikey here. The appeal letter mostly concerns the closeness of the result and does not contain evidence concerning irregularities. However it indicates that some such evidence may have been provided in the Greens' initial submission.
Oct 4: A tweet by Fairfax Radio reporter Frank Keany states that a decision on the Greens' appeal will not appear today. This is now confirmed by the AEC. That makes the weekend available for modellers looking for errors.
I have investigated two outlier House-Senate booths found by Truth Seeker, Derby (Durack), Kambalda West (O'Connor). These booths had discrepancies of 50 votes in the House and Senate totals. In the case of Derby the most likely explanation appears to be that the informal vote total for the House is excessive by enough to explain it. In the case of Kambalda West I cannot find anything "wrong" with the House votes, and the Senate vote in this booth is substantially down on that in 2010. Hmmm...
Oct 7: Truth Seeker here points out that a miscount of the Labor vote by 50 in one house or the other is the most likely cause of the Kambalda West discrepancy.
Various reports have indicated October 9 (this Wednesday) as the most likely decision day, or perhaps October 8 (Tuesday).
Oct 9: Today was expected to perhaps be the day, but it turned out not to be.
Oct 10: Recount granted: a recount of above-the-line and informal votes has been announced by the AEC here. Expect this to take a long time.
Oct 17: A plea for scrutineering help circulated by one of the provisional winners (Louise Pratt) indicates a timetable of Oct 17 - Nov 8 for the recount. This is also compatible with an estimate by the AEC in a letter I have seen that they would have finalised all election results in about three weeks from now.
Oct 18: The recount is underway and details are being posted by the AEC here. Details published today and available in the CSV file are for the Division of Cowan and include most papers for that division.
It is not clear from the information published whether we are going to get any form of running update on the primary vote totals as errors are found and corrected or whether the revised totals and outcome will come out only at the end of the recount. What we can tell from the information published so far is:
* Small discrepancies between the original and the new vote totals are present in about 70% of booths. But if you were hoping these would resolve the House-Senate discrepancy issues in most booths, it seems from the first ten booths at least that they often don't, and in cases increase the gap between the two.
* The rate of challenged ballots is relatively low, suggesting a normal challenging environment rather than a repeat of what is going on in Fairfax. Only one vote is shown as challenged for authenticity so far.
Meanwhile for those who really want to get microscopic about Senate analysis, there is some very detailed commentary about BTL votes by an anonymous commenter to a Truth Seeker thread here. When a number is repeated or omitted, the vote exhausts at that point if it has been through all the candidates above that point (because they are all elected or eliminated). A small number of repetitions or omissions doesn't cause a vote to become informal but may cause it to become worthless at a certain point. Anonymous shows that certain combinations (such as two 4s and no 41s) are unusually common, and that in most such cases it's clear from the placement of surrounding numbers which number the voter meant to be which. These might be cases of poor handwriting or cases where both data entry staffers made the same mistake. (Hopefully any ballot shown as exhausting is automatically checked to guard against this.)
Anonymous then looks at the possible impact of such votes on the count if they were counted for the candidate who was most probably intended. Anonymous finds that the impact on the critical point is very small and would probably slightly widen it if these votes did not exhaust.
Oct 24: Apologies for the slow pace of updates on this article. We've seen the process above continue through various electorates with Brand, Canning, Cowan, Fremantle and Moore completed and another two started. It looks like figures in the current download aren't complete for Moore. That aside there are numerous small discrepancies with two substantial ones: Greenfields PPVC (Canning, +79 compared to expected) and Absent (Fremantle, +106). Whether these two result from miscounting or just from classification issues involving other electorates I don't know.
Assuming they are just classification issues but also that there are some booths that have errors in both directions (eg a booth with 3 extra votes might have 4 extra for one candidate but one less for another) then it's possible the total number of errors is about 1 in every 1000 votes, ie over 1000 votes worth of errors in the total for the state. Even with only a few percent of votes being cast for one of the critical parties or their feeders that might sound like something with a fair chance of upsetting the margin, but it seems unlikely either feeder party would gain or lose enough to make a difference (since errors affecting them will occur in both directions), and it's more likely large bundles would be miscounted at booth level than small. So the most likely cause of changes that might actually overturn the margin is in my view outright mis-sorting (votes for AC-vs-S+F feeder parties in the wrong pile).
Oct 25: The process continues and the electorate of O'Connor has had attention today, with the result that the Kambalda West discrepancy has been resolved (53 votes found). What is interesting is that another booth, Castletown, which so far hasn't been picked up as an issue, was 51 votes out in the other direction. Of yesterday's issues, the Absent total in Fremantle now squares (within a vote) but Greenfields PPVC remains 79 out. Double-figure differences (usually under 20) are also cropping up in a number of booths.
Differences of this size vindicate the recount as having been a good idea no matter what the outcome. But they also raise the question of why such large errors in the counting of individual booths are happening and whether more should be done by way of quality control to ensure such errors are avoided earlier in the process - not just at Senate level but also for the House of Reps.
Oct 28: The most recent update shows that the Geraldton-Waggrakine booth has also now been
(Note added Oct 29 There are still issues with this booth - see comments - the number of votes shown as accounted for is 71 fewer than expected, which is one of very few large differences in that column.)
Oct 31: An explosive (and for the AEC disastrous) development in the race today with an announcement by the AEC that 1375 votes were not just initially miscounted, but are considered actually missing. It is not yet clear what the breakdown of the missing votes is and whether they are a sum of the many small discrepancies already noted or whether they relate to booths not yet partly or fully documented (the CSV is out of date). The AEC looks set to press ahead with a new distribution of preferences and a new margin but it is hard to see that this has a snowflake's chance of surviving a court challenge from the losers, and it may well be that the AEC itself petitions for a no-result to be declared and a by-election initiated. With new Senators coming on board only from mid next year whoever is "elected" may not even get to take their seat before the election is annulled and a by-election called. I cannot see this going any other way from here.
Nov 1: Commenters at Truth Seeker have been looking at revised figures and believe that the Australian Christians' position on ATL votes has improved enough for them to now possibly surpass the Shooters and Fishers at the key point of the count, which could flip the outcome to a "win" for Ludlum and Dropulich if so. However the margins involved are microscopic, there is still the possibility of some or all of the missing votes being found, and in any case as things currently stand the outcome mainly has meaning should the CDR not determine the case before the new Senate takes its place. Unless the votes can be found mighty fast then what we have here may well be a zombie election* that continues as required by law as if it is a living recount, but is actually already dead.
There is the possibility - canvassed by William Bowe here (might be paywalled) that a court could find some way to save the whole election and accept the result - for instance by accepting the count as it stands on the understanding that the missing votes had been counted once and could only not be recounted. So the court might require a petitioner to show cause (based on scrutineering or other evidence) why any of the missing votes should be assumed to have been wrongly counted in the first place. (Based on what we have seen there are probably some counting errors in the original counts for the booths with missing votes, but they may well not affect the result.)
The missing votes come from four booths: "Bunbury East in the electorate of Forrest, and Henley Brook, Mount Helena and Wundowie in Pearce". Transportation of votes is a possible issue here - there is probably not anything corrupt going on - but we will have to wait for the full investigation to possibly determine how it happened. Thus the missing votes are a completely fresh issue, independent of the mostly minor count variations that have been seen in a very large number of booths.
Many of the issues that arose with this Senate election were already being compared to the many irregularities in Florida in 2000 that affected the outcome of the election for President. Until now none of those have been valid, because to the limited extent that they resulted from electoral authority error (Coolum Beach in Fairfax, the "missing" (actually mislabelled) votes in Indi) they were properly rectified before they caused more than passing embarrassment. Most of the issues with this election until now resulted from bad system design by politicians and not the action of counters. This unfortunately is different - this is looking like a critical failure that will invalidate an otherwise valid (albeit absurdly close) result. The possible silver lining then is that it kills an outcome that was already the product of a farce, and maybe the whole state by-election will have a more proportionate result. Then again, given the likelihood of a lousy turnout, perhaps not.
See Poll Bludger for a rundown of past Senate and multi-member PR recount-like situations.
* This expression sounds like mine but was my partner's.
Button Press Take Two: The button will be pressed on Sat Nov 2, apparently in the afternoon WA time, with declaration of the poll at 5 pm Nov 2 (WA time).
More: the anonymice have been busy at Truth Seeker again and because there are still the small number of fresh BTLs to be added, it's difficult to tell exactly which way the zombie count is going to go. (It would be too funny if the key exclusion was a tie and a candidate had to be randomly eliminated.)
The comment at 8:45, taken in conjunction with the detailed one ahead of it, is interesting. If the AEC really pushes the button on recounted data that is missing the missing votes (even if it is known who the missing votes are for) then surely the CDR will not accept that as an outcome. The question is then whether the CDR will attempt to find an outcome at all or just can it.
Nov 2: The button press is expected to occur at 2 pm WA time (5 pm for eastern states on Daylight Savings.) The latest comments on Truth Seeker indicate that if the count is based on the recounted totals without the 1375 lost votes, then Ludlam and Dropulich may win by a whisker, while if it is based on the recounted totals with the lost votes included, the original result may remain. Either way there seem to be just a handful of votes in it at the critical point.
Nov 2: Scott Ludlam has tweeted "We won" suggesting that the official outcome has flipped. Developing ...
Update: Confirmed. The declaration of Ludlam and Dropulich as "winners" will occur on Monday but is extremely likely to be thrown out by the court. The AEC's wording suggests that ATL votes not found during the recount were not included. We now await the margin at the key exclusion.
Update: The margin at the key exclusion point is Australian Christians over Shooters and Fishers by 12 votes. But we know that the original count of the lost ATL votes favoured S+F over Australian Christians by 13, meaning that the true result (had the missing votes been recounted) may have been as close as one vote, or could have even (allowing for the possibility of an error in those booths) have been a tie requiring a rather arbitrary countback system to determine the final seats. (For those interested it would go the way of Ludlum/Dropulich if that happened.)
However, the CDR does have the power to reexamine disputed votes from the recount, and on that basis it might determine several votes differently. That is assuming it even attempts to do so rather than simply deciding that a new vote is required irrespective of whether the actual margin is (say) 0, 1, 5, 10 or 20.
Nov 3: Should The Original Count Stand As Proposed By Clive Palmer?
Antony Green has useful discussion of the territory ahead of us here. Ben Raue has similar dissections to those performed by Anonymous at Truth Seeker yesterday, in this case finding a real margin of about three votes to Pratt/Wang. (Edit: Raue now agrees with one vote.)
Some may sympathise with the position of Clive Palmer that the original count was the only one that included all the original votes and therefore it should stand. I just want to make it clear why this is wrong. Firstly, the original count did not really include all the votes. Rather, we know that a bare minimum of 678 voters were disenfranchised by miscounting in the original count, most likely as a result of counting errors (including undercounts of 79 and 53 votes at two specific booths). Secondly we know that up to 774 votes were wrongly excluded from the original count because they were wrongly flagged as informal.
Both the original count and the recount are demonstrably wrong - the original count because of the errors found in it during the recount, the recount because of the missing votes. There's a view about that generally an original count and a recount are just two different counts of the same votes and each is as good as the other, but this is false. Rather, during a recount votes originally counted correctly will remain correctly counted (barring an incorrect formality ruling, which is a very rare event) while votes that are originally counted incorrectly will usually be detected and fixed.
The normal principle is that the more times an election is recounted, the more accurate the count becomes, and indeed if you repeat the recount process often enough you eventually reach a point where each successive recount gives exactly the same result. (I have seen this sort of thing in action several times, including in the 2011 Hobart Deputy Lord Mayor ballot, which was recounted six times at a critical exclusion point.)
However in this case the principle of the superiority of the recount over the original count is violated by the loss of 1375 votes and their exclusion from the recount - especially as those votes are not representative at the key exclusion point and would have probably changed the outcome.
(Some might wonder how this relates to Fairfax, where the margin was 36, then 7, then 53. Surely the second margin there is less accurate than the first? The catch there is that the second margin was not one determined by a full recount, and only the third margin is actually the outcome of a full recount of ballots. The second margin was the margin after the recheck of those votes that were not primary votes for either Clive Palmer or Ted O'Brien, during the distribution of preferences. Since the rechecked votes heavily favoured Palmer it is possible that the correction of errors in those votes would disadvantage him more than O'Brien.)
The CDR's only credible choices in my view are:
(i) to attempt to find an outcome using the recounted results where available, the original results from the lost booths and its own rulings on the 942 votes challenged by scrutineers; or
(ii) to declare the election void.
And (i) is only credible if the end point is a margin that establishes beyond reasonable doubt that one or the other pair of candidates won.
Something we have not seen yet, but may well see when the court case gets underway, is the emergence of claims about votes being wrongly included or excluded - administrative errors in processing absent votes, and double-voting being typical examples. Genuine double voting (as opposed to apparent double voting caused by administrative error) is rare, but it does occur (often because elderly voters forget they have cast a pre-poll vote, and vote on the day as well). In the context of a disputed election including over a million votes, it shouldn't be hard to find enough such irregularities to knock over an implied margin of fewer than ten.
All up I think a full Senate by-election for WA is extremely likely.
Nov 4: The declaration of the phantom Senators has now occurred and Dropulich and Ludlam are now officially Senators-elect. On the assumption that the Court will order a completely fresh election for the state, ALP jockeying for preselection has already begun. Meanwhile a commenter at Truth Seeker has a possible explanation for the Waggrakine mess - the BTL and informal bundles for the booth could have been mistakenly juxtaposed.
Nov 6: Georgatos Nearly Missed Therefore Assange Will Win?
One of the odder sideshows to the potential by-election is coming from the Wikileaks Party, which intends nominating Julian Assange as its lead candidate to remedy the "injustice" of Assange not winning in Victoria. While it is true that Assange got more votes than Ricky Muir, I am not really sure why failing to win a Senate seat with 1.24% of the primary vote is such an outrage.
Anyway today the Wikileaks WA Senate candidate, Gerry Georgatos, said:
"I only missed being elected to the Senate by 3,000 plus primary votes. Therefore Julian Assange will be elected."
The first chink in this chain of argument is that a new Senate election will be held with new candidates, new primaries, and new preference deals. It therefore does not follow that any number of votes (short of a quota) will elect any candidate, nor even that a ticket that missed election this time will not win by polling the same (or even a worse) vote next time around. All this is up in the air.
The second concerns the actual election. At the point of his exclusion, Gerry Georgatos was 3202 primary votes shy of the Animal Justice Party and therefore eliminated. I suspect this is the 3000 + that he refers to. But had Georgatos got over the AJP, that would not have been enough. To win, Georgatos needed to beat the HEMP Party who were 4324 votes ahead of him at this time. Finishing between HEMP and AJP, or even ahead of HEMP and AJP in that order, would not have been enough, since the AJP's preferences would have gone mainly to HEMP with only a small helping to Wikileaks. So an increase of a few thousand in the Wikileaks primary would not have helped without specifying where those votes came from.
For Wikileaks to win, they needed the order of the parties at this crunch point to be Wikileaks, AJP, HEMP instead of what it really was, which was HEMP, AJP, Wikileaks. This could have happened with a movement of 2510 votes (1817 from HEMP, 693 from AJP) to Wikileaks in the primary count. However, when you get to specifying not only a number of votes to move, but also that they must come from specific sources, it becomes a lot less close, and there were probably many closer misses than that nationwide. A miss by thousands when you polled less than ten thousand isn't really all that close, and provides no guarantee that having the ticket topped by a celebrity candidate who is clearly not much of a vote-puller would have made the difference.
Nov 8: The AEC has released official details of the 1370 (not 1375 as earlier indicated) missing votes and this confirms what modellers determined last week: that the current real margin is one vote in favour of Pratt and Wang.
I notice that the final count of ballot papers assessed in the recount is 483 fewer than the original expected total even with the 1370 added in, presumably as a net total of counting errors. However the final count of formal votes is up by 118 (if the missing votes are included), so it seems there are 601 more formal votes (including the missing votes) now than before. This doesn't square with the 774 ballots indicated as "for CSS" (ie for below the line entry). However some possible explanations for the difference are (i) some ballots forwarded to the CSS system may have been rejected by it as informal (eg because of voter errors) (ii) some ballots forwarded to the CSS system may have been votes originally classed as above-the-line (for instance if a voter voted above and below the line.)
It's almost certain that at least one of the unchecked informals is actually a formal vote, though it's fairly unlikely that any unchecked informal is a vote that affects the notional one-vote margin. (Some figures in this update were corrected on November 11.)
Nov 15: The AEC has done the decent thing and petitioned the Court of Disputed Returns asking that the election for the WA Senate be declared void and re-run.
Nov 18: As helpfully posted by Anonymous below, the petition itself can be viewed here.
Dec 6: Keelty Report Released: The Keelty Report into the loss of ballot papers has been released and can be viewed here. It finds that the fate of the missing ballot papers is unknown and that the parlous condition of the chain of custody for ballot papers contributed not only to their loss but also to the difficulty in establishing their fate. Here's some highlights (apologies for the formatting issues with this bit - copying and pasting is sometimes a problem with these things):
"Compared to operations elsewhere, the packing and marking of boxes containing ballots from
the Division of Pearce was well below standard. Instead of each box being properly labelled, hand written felt pen markings depicted the origin and type of contents of these boxes. This part of the process is clearly inadequate given that printed labels were available and were used by other DROs
Another factor impacting the boxes of Pearce ballots was that following fresh scrutiny the Pearce boxes were moved to the Welshpool warehouse in three separate loads. Two loads were taken by the contract courier and one load was taken by the warehouse manager himself. Dividing a load in this way is contrary to all of the controls applied elsewhere to maintain some idea of the location of goods. Added to this confusion was the fact that the three loads were not on pallets. These boxes were individually transported.
During this inquiry, it was also discovered that the attitude towards Senate ballots is different to those of the House of Representatives. There is less concern for the security and integrity of Senate
ballots because it is considered that they have less of an impact on the election outcome and in any event are warehoused for six years. This is a cultural problem within the AEC and it needs to be addressed. The fact that it had been thirty years since the last full recount of Senate ballots most likely added to the loss of care in routinely dealing with those ballots during the election" [pp. 8-9]
"The Inquiry noted that prior to the election, the transport contractor offered the WA AEC a cargo
tracking system, at no cost, to assist in the tracking of AEC material during the election. The WA
AEC and the AEC national office did not pursue this offer and the Inquiry concluded that this was
a missed opportunity to implement best practice materials management." [p. 15]
"As mentioned elsewhere in this Report, it appeared that packaging, labelling and general material management practices varied markedly across WA. These practices, especially the lack of detailed consignment notes, poor labelling and the reuse of boxes, compounded by insufficient oversight of
the process, contributed to an environment where the AEC could not subsequently account for all
votes." [p. 23]
"d) ballots were left in open, unsecured boxes at the recount centre overnight, in the custody
"d) ballots were left in open, unsecured boxes at the recount centre overnight, in the custody
of a lone security guard, who had not been vetted by the AEC for political neutrality,without CCTV coverage" [p. 33]
There is so much more of it, with photographs too. The conclusion is rather damning:
"In the 2013 Senate election in WA, the AEC failed to meet its own high standards and damaged
its reputation with the community and Parliament. The loose planning culture in WA and the complacent attitude toward ballot papers following the fresh scrutiny were manifested in: inadequate
skill development and training, a series of minor errors, poor material management practices,
flawed contract management and quality control, erroneous assumptions and lax supervision and
enforcement combined to create an environment where these high standards could not be met."
The recommendations are many, including ending a culture in which some counted ballots were regarded as no longer "live", and the AEC will begin implementing them. Hopefully following the likely WA Senate re-run there will be an audit of its success in doing so.