Update (Saturday 11pm)
Two number-crunchers out in web-land have made the effort to identify which votes have gone missing by comparing booth results from the first and second counts, and they have reached the same conclusion: but for the missing votes, the result at the decisive point in the count would have been very close to a tie. One is Ben Raue at the Tally Room, while the other is an anonymous commenter on a pseudonymous blog that of Truth Seeker, whose statistical work on the Senate count process has won great acclaim. Despite the obscurity of the latter source, he or she is clearly well on top of the situation.
It is clear that what has gone missing is bundles of votes for particular parties, the numbers of which can be determined with considerable precision despite the minor adjustments to the vote totals which legitimately resulted from the recount process. I will now deal with the four booths in turn, using the numbers determined by Anonymous. Keep in mind that the only votes with the capacity to change the result are those cast for Shooters & Fishers, Australian Christians or the smaller micro-parties which fed them preferences: Australian Independents and the Fishing & Lifestyle Party in the former case, No Carbon Tax Climate Sceptics in the latter.
Mount Helena: This booth produced the most important discrepancy with the disappearance of its complement of above-the-line votes for Shooters & Fishers, of which there were 14. Also gone are the Liberals’ tally of 370 above-the-line votes, together with nine for Animal Justice.
Wundowie: Together with its 29 informal votes, the above-the-line votes of nine different parties have disappeared from this booth. Those relevant to the result are three votes for Australian Christians and one each for No Carbon Tax and Australian Independents. Also gone: 166 Liberal, 164 Labor, eight Smokers Rights, seven Help End Marijuana Prohibition, five Wikileaks, and one Katter’s Australian Party.
Bunbury East: The 152 votes missing from this booth appear to be the above-the-line votes for nine parties, the only ones relevant to the result being three for Australian Independents and one for No Carbon Tax Climate Sceptics. The others were 112 votes for the Greens, 12 for Animal Justice, 11 for Family First, six for Wikileaks, three for Stop the Greens, two for Katter’s Australian Party and one for the Secular Party, together with 77 of the 81 informal votes cast at the booth.
Henley Brook: This one was straightforward and, ultimately, not important: 349 Liberal above-the-line votes are missing, leaving 286 still in the count.
Tallying that up suggests Shooters & Fishers have lost 18 votes and Australian Christians five. Ben Raue’s calculation is very slightly different at nineteen and four, as he counts a missing Australian Fishing & Lifestyle Party vote and has one less vote missing for Australian Christians, which Anonymous thinks likely to have been genuine recount corrections. Either way, the numbers suggest that reverting to the first-count results for these booths, as I have been advocating, would cause the result to flip back to Louise Pratt and Dio Wang based on the smallest margins imaginable one vote by Anonymous’s reckoning, and three by Ben Raue’s.
Original post
For the vast civilian majority in Western Australia that would sooner not have to vote again, today’s conclusion of the Senate election recount has delivered the worst possible outcome, with the original result overturned by an excruciatingly narrow margin. On the basis of the unquestionably flawed and incomplete recount, Greens Senator Scott Ludlam and Australian Sports Party candidate Wayne Dropulich will be declared the winners in place of Labor Senator Louise Pratt and Palmer United Party candidate Dio Wang.
The key to the result remains the point at which the lead candidates of either Shooters & Fishers or the Australian Christians are excluded. If it’s the former, Shooters & Fishers preferences sustain Dropulich at a point in the count where he would otherwise be excluded, ultimately allowing him to finish ahead of Wang off the base of a tiny primary vote after harvesting preferences left, right and centre. Wang’s preferences then flow to Ludlam, giving him victory ahead of Pratt. But if Shooters & Fishers stay in the count, Dropulich is unable to overtake them, a point arrives where he is excluded, and Shooters & Fishers themselves are ultimately unable to get ahead of Wang, who wins a seat when the various micro-party preferences are distributed. With Wang’s votes used to get himself elected, his exclusion does not provide Ludlam with the mass transfer of preferences he needs, causing him to be left holding the bag with Pratt’s election to the final seat.
In the original count, Count 139 famously delivered a 14-vote victory to the Shooters & Fishers candidate with the following numbers:
BOW (Shooters & Fishers) 23,515
VAN BURGEL (Australian Christians) 23,501
However, the recount has turned this into a 12-vote lead for Australian Christians at what’s now Count 141:
VAN BURGEL (Australian Christians) 23,526
BOW (Shooters & Fishers) 23,514
But of course, the latter result excludes the famous 1255 missing above-the-line votes, which were cast at the Bunbury East, Mount Helena, Henley Brook and Wundowie booths.
If I had anything to do with the matter, I’d be very determined to craft an outcome that didn’t require a fresh election, which will have a disastrous impact on the public’s confidence in the system. The best chance of avoiding that would have been for the recount to have reaffirmed the original result, which might have given the High Court the confidence to determine that the recount should stand, warts and all. However, that’s gone out the door now that counts have produced different results with respect to not one but two seats.
The other chance of an out involved the requirement that a discrepancy be big enough to be decisive before it can be used as a basis for overturning a result. Under happier circumstances, the availability of first count results from the four affected booths might have served as evidence that this wasn’t so. While 1255 votes sounds like a lot in the context of the margins under discussion, it must be kept in mind that the only votes which actually have the potential to affect the result are those cast either for Shooters & Fishers, Australian Christians or other micro-parties which fed them preferences (remembering we’re only dealing with predictable above-the-line votes here below-the-line votes were not included in the recount process, presumably on the basis that they were thoroughly scrutineered during the data entry process). Those parties are the Australian Fishing & Lifestyle Party and Australian Voice in the case of Shooters & Fishers, and the Rise Up Australia Party in the case of Australian Christians. Based on the published results for the first count, the four polling booths at issue delivered 61 votes to the Shooters & Fishers and 60 votes to Australian Christians (UPDATE: The previous two sentences are not quite right see the update at the top of the post for a more accurate account). However, it’s not enough simply to add those votes to the existing totals and achieve a hypothetical result, as not all of the votes from the four affected polling booths went missing. In short, the situation appears far too unclear to say that the missing votes could not have affected the result, given the narrowness of the margin involved.
My own preferred solution, which I advocated in an article for Crikey yesterday, was for the High Court to direct the Australian Electoral Commission to use the first count results from the four affected polling booths and the recount results from everywhere else. I wouldn’t presume to say that there are no legal difficulties involved in this, but it would appear to me to pass the common sense test. Given that the number of missing votes on which the result might hinge is around 120, and the likelihood that these particular votes were counted more-or-less correctly the first time around, it could have been hoped that the margin at the key point in the recount would have been big enough to allow such a result to proceed with an adequate level of confidence. But with just 12 votes in it, that becomes a lot harder to do.
Sorry if this has been covered by why would the AEC declare based on an extraordinarily flawed recount rather than relying on the original count which presumably mere has ordinary flaws of counting embedded?
[Anyone have a better theory?]
Surely it’s much more likely that the AEC screwed up.
[Surely it’s much more likely that the AEC screwed up.]
But that’s very dull. My fallback theory involves aliens.
William, thank you. I really appreciate you and the other bloggers providing this analysis. There is nothing of this level in the MSM.
WeWantPaul@51
Having chosen to do the recount they are required to declare based on the recount. The legislation is not very flexible.
I add that the total number of ballots affected by “ordinary flaws of counting” in the original count exceeds the number of votes missing in the recount. Apart from the missing votes, the recount would otherwise be much better counted and more reliable. It just happens that the margin is so tiny that the missing votes are enough to sink it, and that is only so because the missing votes lean a particular way at the key point.
Even if it was possible to declare the original count, the original count is just as discredited as the new one.
WeWant Paul
Antony Green covered this. What has happened is correct under the law. The AEC must declare the result, then others can dispute it in Court.
When you consider they lost .1percent of votes and we only know this because of the recount, it raises the question of the security of votes, the checks and balances in place and whether it has happened before. While I believe every vote to be important you have to admit the vast number of votes, the casual workforce and the push to get the results determined in a timely manner all lead to the increased probability of something falling through the cracks. It’s very easy to blame someone rather than just acknowledge that nothing and no one is perfect. New technology in provides options to limit issues such as this, and hopefully the review will look at that.
Will be interesting to see how the likely court dispute goes.
Seems to me that by the time it gets to court there will have been a vast amount of analysis of both results done and thankfully, the conspiracy theory of active tampering is looking more and more like hyperbole and fluff.
I don’t think that anyone arguing for either result to stand will be successful as they are both so close as to be arguable, and with the missing votes issue in play there seems no prospect of a recount giving a definitive, reliable result that all parties will accept.
I reckon that there is now a good argument that any court where this is heard should look strictly to the principle of integrity of the result and confidence in the system so that any legislation passed or blocked by the new Senate is seen to be done so properly. The only definitive way to do that is a new election for those 6 W.A. seats with everyone keeping their fingers crossed 🙂 that the margins from that are wide enough so that its an easy result to declare.
Psephos:
[I don’t accept Fran’s assertion that Greens are too pure at heart to do such a thing. There are individuals in all parties who would do it if they thought they could get away with it. ]
That’s either a sweeping generalisation or a factual claim based on your personal knowledge. If it’s the latter, you should be more specific, and declare that you know such individuals (which would raise the question of why you haven’t brought your concerns to the approriate authorities.
If it’s just an inference based on the general maxim that there are rotten apples in every barrel then it’s really no more impressive than my belief that nobody in my party would contemplate doing such a thing.
Of course, I know a little more of my party’s internal culture than do you, so I’m going to say that my interpretation should be preferred, at least until there is impressive evidence to the contrary.
For the record, I’m not asserting that such a thing is utterly inconceivable — just highly improbable. We don’t care enough about winning to cross the kinds of ethical boundaries that people who fancy that they might achieve government seem willing to ignore. If we ever do get close to leading a government, I suppose that might change, and we will need to be a lot more vigilant about shady characters than we are now, because we’re more likely to attract the wrong kinds of people, but at least at this stage, I’m saying we’re squeaky clean.
oops:
[(which would raise the question of why you haven’t brought your concerns to the appropriate authorities).]
Assantdj
[When you consider they lost .1percent of votes and we only know this because of the recount, it raises the question of the security of votes, the checks and balances in place and whether it has happened before.]
Absolutely. I’m seriously questioning whether this happens frequently but it’s not picked up.
Obviously it would very seldom make any difference but it’s a bit of a wake-up call for all of us who laugh at the standard on elections in other countries.
61
The votes were obviously stolen by some foreign despot wanting to make their elections look less bad.
[The votes were obviously stolen by some foreign despot wanting to make their elections look less bad.]
Obama?
I was thinking Mugabe or Putin.
I betcha it was Fran all along.
thanks for the answers re the legislation requiring them to use the dodgy recount
Serious psephological questions: Is there a well-developed theory or body of research around the phenomenon of cliffhanger election results?
Knife-edge results happen far more often than is statistically probable. The zero hypothesis of random results is disproven. Obviously political parties target the centre, which in a two-party preferred system vastly increases the likelihood of splitting the vote very closely down the middle…
…but still, what gives? Can anyone point us to a body of quant research that gives a credible account of why knife-edge results happen far more often than random chance would permit?
Al Dente @ 67: Interesting question. But are you talking about close results in individual seats, or about Parliaments which are almost evenly split? And what makes you think that close results, however, defined, are happening improbably often?
Close results in individual seats is what I’m asking about (or, as we saw in WA, Senate seats falling to only a few dozen result, hence why I’m posting the question in this thread).
How often do you hear about results that are within <100 votes either way, out of electorates that are ~100,000 in size? More often than 1 in 1,000 seats that's for sure…
…It wouldn't happen with a random roll of the die. So what gives??
This particular Senate election was extremely prone to knife-edge results because of the whole snowballing thing, and the number of parties contesting.
Beyond that the distribution of 2CP margins in HOR would be an interesting exercise. The distribution would not be random (eg 52-48 is much more likely than 80-20) but whether virtually 50-50 is the mode outcome, and how much commoner say 50.2 is than 50.8, would be interesting.
I might have a look at this someday.
Psephos
“On the question of the missing votes:
* It seems highly suspicious that bundles of votes have simultaneously gone missing at four different booths in the one state during the recount of a closely contested Senate election.
* This suggests to me that the bundles have been deliberately misplaced (ie, stolen) by someone. That someone would have to be an AEC employee, since no-one else has access to the ballot papers.
* Since it has happened in four different places, it is unlikely that the theft was done on impulse – it suggests a plan co-ordinated by someone, which could only have had the objective of influencing the outcome of the recount. And as we see that objective has been achieved, at least temporarily.
* So if we apply the test of cui bono, we must conclude that the instigator of this plan was someone who wanted Ludlam or Dropulich to win.
* Since I doubt Dropulich’s party has the capacity to do such a thing, that points the finger at the Greens.
* I don’t suggest that Ludlam or the Greens state leadership are responsible. But it’s quite possible that some local zealot who had gained employment as a temporary AEC worker for the election count decided that such an action was justified by the greater good of getting Ludlam re-elected, and that he or she then texted three mates in other booths (or wherever the recount took place) to do the same thing.
* I realise the hole in this theory is that the thieves would have had to know precisely which bundles of votes to steal, but it wouldn’t have been impossible to figure that out. Anyone have a better theory?”
Is this really how your mind works? You develop a theory, don’t test it, but present it anyway?
This is really quite interesting… Wow.
[The distribution would not be random (eg 52-48 is much more likely than 80-20) but whether virtually 50-50 is the mode outcome, and how much commoner say 50.2 is than 50.8, would be interesting.]
Would it not be a matter of testing the spread of margins for kurtosis? If the distribution is leptokurtic, then Al Dente is right and the incidence of close results is greater than “normal”.
Dear proto-Dr Bowe,
What is this kurtosis and leptokurtic of which you speak?
I’d look it up in Wikipedia, but Greg Hunt is using my copy of it at the mo’…
Perhaps a picture might be worth a few dozen words:
[The ballot papers disappeared at a time and place somewhere between the RO’s office and their retrieval from the storage centre for the re-count. When and where they disappeared between these two points is now the subject of inquiry.]
Given the stories we’ve heard on this blog about spilled ballot boxes etc in transit due to AEC contractors not maintaining agreed security procedures, and the fact that transit is inherently one of the most insecure parts of the process I am (completely without foundation) going with the provisional theory that these bundles were lost in transit between the DO and the central store.
Ah, I see.
Thank you proto-Doctor. Be a dear and find out how leptokurkic are the good burghers of Downunderistan, would you? 😉
William Bowe @ 72: I’m actually not sure that it makes sense to talk in the abstract about what number of close results is “statistically probable”. To calculate the relevant theoretical frequencies and compare them with observed ones, you need to have an underlying model of elections as a probabilistic process. There would be quite a few ways of doing that, and they wouldn’t necessarily make the same predictions. You can’t just treat constituencies as being randomly sampled (without replacement) from the 12 million voters, because if that were the case, the same party would be expected to win every seat.
Al Dente 76: If you are turned on by the leptokurtic, you should really enjoy a bit of heteroskedasticity.
[Is this really how your mind works? You develop a theory, don’t test it, but present it anyway? ]
Presenting a theory for scrutiny and refutation is the best way of testing it.
Psephos
“Presenting a theory for scrutiny and refutation is the best way of testing it.”
Only when your too lazy to do your own thinking first.
And when the theory is to claim criminal intent on people you’d think the first thing you’d do is be skeptical of your own theory first. It’s a pretty shallow and transparent political attack.
I think Psephos theory contained for him the welcome idea that the perfidious Greens were to blame all along
Perfik …as the kids would say
Psephos Posted Sunday, November 3, 2013 at 11:03 am @ 31
Maybe they were accidentally bundled with HoR votes.
[Anyone have a better theory?]
AEC stuff-up.
My solution: The HC meets for ten minutes, hears no evidence, and orders that Pratt and Ludlam be seated on the grounds that they got the most votes. The end. 🙂
William @ 18.
So that’s why you’re the poll ‘bludger’ !
Psephos Posted Monday, November 4, 2013 at 10:30 pm @ 84
It would be worth it just to see Clive Palmer’s reaction.
As if leptokurticism were not enough… Today’s new academic work brings a proposal for ‘quadratic voting’. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2343956
It’s from a Chicago theorist with a famous name in legal scholarship.
Giving “proper weight to preferences of varying intensity” hey. This is assuming that voters are able to understand the best way to vote to match the formulas used to give proper weight to their preferences.
A novel aspect of the new WA Senate election if there is one is that there will be no marginal seats sucking in most of the campaign resources as usually happens. There might be areas where there are more swinging voters than others, but generally a swinging voter in a safe seat will be worth as much as a swinging voter in a marginal seat. The parties may need to change where they direct their resources for this one.