Senate fact-checking and myth-busting

The Western Australian and South Australian Senate results under the microscope: which preference decisions mattered, and which didn’t.

In a Senate election that has excited more interest than most, there has been a lot of understandably confused talk over the past few days about the impact of contentious preference decisions. What follows is an attempt to sort through some of them claims which have gained currency around the place.

Claim: Wikileaks preferences cost Scott Ludlam his seat.

Julian Assange has had no more stout defender over the past few years than WA Greens Senator Scott Ludlam, so his party’s failure to fast-track preferences to Ludlam provoked a furious response from many, not least his own mother. Most offensive to left sentiment was the decision to place Ludlam behind David Wirrpunda of the Nationals, which the party justified as supporting Aboriginal representation. When news came through that Ludlam had indeed fallen short (pending of course a possible recount), many on social media jumped to the conclusion that Wikileaks was to blame.

However, the potential of Wikileaks preferences to harm Ludlam actually went unrealised. Wikileaks preferences did flow to the Australian Sports Party and the Nationals in turn, but both were eliminated fairly early in the count. Since Ludlam had a higher placing from Wikileaks than the other late count survivors, namely the two major parties, the Palmer United Party and Shooters and Fishers, he did in fact get their votes – 8150 of them, which in the final analysis still left him 3372 short of Louise Pratt.

Claim: Wikileaks vote-splitting cost Scott Ludlam his seat.

Absent the opportunity to blame Wikileaks for their preferences, some of the party’s critics fell back on the argument that it damaged Ludlam by fielding a candidate at all. One such was Mary Kostakidis, who like many expressed her displeasure on Twitter. This would have been a telling point under first-past-the-post, but it doesn’t work so well under compulsory preferential voting, where every vote of an excluded candidate ends up with somebody else – in this case, as noted, the Greens. If anything, Wikileaks’ entry might have helped deliver the Greens preferences from libertarians who would never have voted for them directly, although I doubt there would have been much in this. It could be argued that the absence of a Wikileaks candidate might have given Ludlam’s campaign clearer air, but I doubt there’d be much in that either.

Claim: Greens preferences elected Family First in South Australia.

Confronted with the allegation that her party’s preferences were responsible for electing Bob Day of their ideological enemy Family First yesterday, Sarah Hanson-Young tweeted: “Actually SA Greens preferences went to Xenophon & Labor well ahead of both FF & Liberals”. Either the Senator was not entirely on top of her own preference ticket, or she was being exquisitely disingenuous – perhaps the former, given that her tweet was deleted shortly thereafter. It was indeed true that Xenophon himself was rated much higher than Family First on the Greens’ preference ticket, but as Xenophon had no trouble getting elected off his own bat, that was beside the point. The issue was Xenophon’s running mate Stirling Griff, who did spectacularly badly on preferences from all fronts. Whereas Xenophon was given a solid but ultimately irrelevant position a third of the way down the Greens’ preference order, Griff was buried deep below. As it happened though, the 11,065 Greens ticket votes delivered to Day after Hanson-Young’s election were surplus to requirements: he would still have been 21,257 clear of a quota without them. So the Greens did not of themselves elect Family First, though it wasn’t for want of trying.

Claim: Labor preferences elected Family First in South Australia.

By contrast, Bob Day would never have made it had not Labor also opted to put him ahead of Stirling Griff. This might inspire comparison with Labor’s preference decision in Victoria in 2004 which gave Steve Fielding his seat at the expense of the Greens, but on this occasion Labor had the Greens second. The decisive factor was actually the transfer of Sarah Hanson-Young’s surplus, which was bloated with the Labor preferences received when Don Farrell was eliminated. This amounted to 41,501 Labor ticket votes which would have elected Griff had they been directed to him instead.

The wonks among you might like to observe that this is an uncommonly severe case of the distortion caused by the “inclusive Gregory” method of calculating transfer values when distributing surpluses. In theory, those Labor votes shouldn’t have been worth so much since they have already been used to help elect Penny Wong. However, the existing system only applies one transfer value to all votes in a given surplus distribution. If votes were reduced in value in due proportion each time they formed part of a surplus transfer (the “weighted inclusive Gregory” method), the Hanson-Young transfer would have contained 23,390 Labor and 18,214 Greens ticket votes, rather than 41,501 and 11,065. In that case, Labor preferences of themselves would not have been decisive.

The value of Labor votes was again inflated when Bob Day’s surplus was distributed, at which point Labor’s remarkable decision to place Griff behind even the Liberals kicked in. This wasn’t decisive, but it might have been if the gap between Simon Birmingham and Griff had been a little narrower. Another factor to be kept in mind here is that Labor got a bigger bang for its buck out of its preference ticket decisions, since 90.8% of its votes were above-the-line against only 81.4% for the Greens.

Claim: Nick Xenophon started it by shafting the Greens.

The Greens’ motivation to punish Nick Xenophon even to the point of aiding Family First was his own decision to preference the major parties ahead of the Greens. Hanson-Young was heard to complain that Xenophon was “making a political decision, just like any other politician”. However, it seems to me that Xenophon is the only party in all this who was doing just the opposite, his preference tickets appearing at least outwardly to be all principle and no expediency.

The Senate ticket voting system allows groups to have up to three preference tickets for their above-the-line votes to divide evenly between. Xenophon used this to play the straightest possible ideological bat by having one ticket that looked how a moderate left supporter might want it to look, and another pleasing from the perspective of a voter of the moderate right. This meant one of his tickets did in fact have the Greens ahead of the Coalition parties, while the other had them behind both the Coalition and Labor. So the Greens might in fact have received half of Xenophon’s preferences, had those preferences been distributed at a point where there was no Labor candidate in the count. In the event, Xenophon preferences were not distributed at all, as Stirling Griff was left holding the bag when Simon Birmingham was elected at the final count.

The Greens’ sense of grievance that Xenophon’s idea of moderate tickets should place them behind Labor is no doubt informed by the fact that he played it differently at his first Senate election in 2007. Xenophon’s idea of ideological balance at that time was to have a single preference tickets which placed the competitive minor parties of both right (Family First) and Left (the Greens) ahead of the majors. I wouldn’t care to venture if Bob Brown spoke truly when he claimed Xenophon’s objective this time around was to loom larger on a smaller cross-bench (a goal he has spectacularly failed to achieve if so). But if he was wrong and Xenophon was indeed motivated by principle, he offers a useful case study for why so few others followed suit. Despite scoring something between six and seven times as many votes, Xenophon landed the same number of seats as Family First. Stirling Griff started the preference distribution process with 107,110 votes, seemingly well on his way to the quota of 148,348. But as preferences swelled the Family First base vote from 38,909 all the way to 180,670, Griff was only able to limp to 121,743.

Author: William Bowe

William Bowe is a Perth-based election analyst and occasional teacher of political science. His blog, The Poll Bludger, has existed in one form or another since 2004, and is one of the most heavily trafficked websites on Australian politics.

142 comments on “Senate fact-checking and myth-busting”

Comments Page 3 of 3
1 2 3
  1. 99

    The votes should be distributed equally between the candidates of a group. This is to give the voters, whose vote is electing a candidates or candidates from a group, the choice of candidates from that group get elected in a meaningful way.

  2. 100

    There has also been a significant increase in the number of votes in a WA Senate race since 1980. This has an effect on the number of errors that can be expected to be in a count.

  3. 99

    Victoria does not allow different tickets of the same group to have different orders for the candidates in that group.

    Section 69B (1) (b) (ii) Electoral Act 2002.

  4. Tom the first and best@102


    There has also been a significant increase in the number of votes in a WA Senate race since 1980. This has an effect on the number of errors that can be expected to be in a count.

    It does but I would also argue that the introduction of ATL voting must have very greatly reduced the total number of relevant errors compared to the old system despite the increased number of votes.

    Worth keeping an eye on the tactics displayed by PUP in the Fairfax recount (I am covering these here: because if there is anything like a full WA Senate recount we could well see the same thing on a vast scale.

  5. (I’m thinking there of errors like votes in the wrong pile, formality rulings etc. The number of errors caused by people failing to count would increase with more votes.)

  6. Tom @ 101
    Let’s have Robson rotation of the groups and Robson rotation within each group. Might have stopped the LD in NSW. If above the line voting is retained, deem the order of preference for candidates in the order on each ballot paper. That will stop both the micro parties and the majors from trying to game the system and let below the line voters choose their senators. Might save the majors from their mistakes.

    Abolishing above the line voting would be a better solution still. See Electoral Reform Australia ‘The case for fully optional preferential voting for the Senate’

  7. 107

    Robson Rotation would, almost certainly have stopped the LDs in NSW. It is a good idea. I am neutral on abolition of ATLs if they are distributed as near as practicable to equally as that eliminates the main objection to them (giving the parties control of which of their candidates are elected).

  8. William @100

    The margin increased from 214 to 346 also in 1980 The Senate was determined by a random sample of preferences.

    Ludlum has to demonstrate a) what votes are missing or had not been included in the count. He would have to also outline which votes he wants reviewed. He could claim scruniteers were denied crucial information during the count.

    His problem is soime what different then in 1980 in that it is not about a contest at the end of the count but a specific contest betweeen two candidates earlier in the count. if you delete all candidates except the last seven standing Ludlum does not win. he would have to look for a mismatch bundle in the above the line distribution. He would need to demonstrate that a particular batch showed an abnormal allocation. A print out of the Polling place returns may come in handy. If the data recorded on the night is different to the tally on the central collation

  9. Tom..City of Melbourne Act a group can change the order of preference within their own group

    Geoff… You can achieve the same result by mathematically distributing the votes. Not a great fan of Robson Rotation it just masks the problem with the count itself. I doubt very much that Robson Rotation will be introduced

    We need to first address issues with the method of counting the vote.

    If we counted money or a companies dividends the way we count the Senate our financial system would collapse overnight

  10. I’ve looked at some of the discrepancies mentioned by Truth Seeker, picking out the two 50-vote House/Senate discrepancies as they stick out like sore thumbs and seem too large to be explained by discarded/missing votes. (Would 50 votes ever go walkies through voters removing or binning them in a booth with <600 total? Doubt it!)

    For the booth of Derby (Durack) I point the finger at the House of Reps informal pile.

    For the booth of Kambalda West (O’Connor) I’m stumped so far. It seems consistent with missing Senate votes but the AEC would know exact figures on what they had issued there.

  11. [His problem is soime what different then in 1980 in that it is not about a contest at the end of the count but a specific contest betweeen two candidates earlier in the count.]

    I don’t see why that should matter, in either a House or a Senate count. The wrong result is the wrong result.

  12. WA

    Just counted the vote scraping the Droop quota and applying a distributed party vote and a weighted transfer no segmentation count.

    2 Liberal
    2 Labor
    1 PUP
    1 Grn

  13. William it is the odds are that the spread will increase not diminish. If you distributed the vote without segmentation distribution IE you do not stop half way through exclusions. The gap gets wider.

  14. Below is a count sheet using a weighted Surplus Transfer, Droop Quota and a reiterative count (Last iteration 3Lib, 2ALP, 1 PIP 1 Green)

    To view select the text and copy to a wordprocessor and change font to Courier 8

    Election title: Western Australia Senate 2103

    Method: Scottish STV

    Number of total ballots: 1311340

    Number of invalid or empty ballots: 0

    Number of ballots used in the count: 1311340

    7 candidate running for 6 seats.

    R|11. WANG Zhe|37. LUDLAM S|52. BULLOCK |53. PRATT Lo|56. JOHNSTON|57. CASH Mic|58. REYNOLDS|Exhausted |Surplus

    |nya (PUP) |cott (GRN) |Joe (ALP) |uise (ALP) | David (LP) |haelia (LP) | Linda (LP) | |


    1|191858.00000|181795.00000|348383.00000| 3303.00000|584028.00000| 714.00000| 1259.00000| 0.00000|562264.00000



    | Count of first choices. Candidates 11. WANG Zhenya (PUP), 52. BULLOCK Joe (ALP), and 56. JOHNSTON David (LP) have

    | reached the threshold and are elected. Candidates have surplus votes so surplus votes will be transferred for the

    | next round.


    2|191858.00000|181891.45066|348383.00000| 3358.69686|187335.00000|396956.40510| 1553.78582| 3.66156|375192.40510



    | Count after transferring surplus votes from 56. JOHNSTON David (LP) with a transfer value of

    | 396693.00000/584028.00000. Candidate 57. CASH Michaelia (LP) has reached the threshold and is elected. Candidates

    | have surplus votes so surplus votes will be transferred for the next round.


    3|191858.00000|181923.98133|348383.00000| 3418.21832|187335.00000|187335.00000|211081.92727| 4.87308|189317.92727



    | Count after transferring surplus votes from 57. CASH Michaelia (LP) with a transfer value of

    | 209621.40510/396956.40510. Candidate 58. REYNOLDS Linda (LP) has reached the threshold and is elected. Candidates

    | have surplus votes so surplus votes will be transferred for the next round.


    4|191858.00000|182053.41693|187335.00000|164335.79213|187335.00000|187335.00000|211081.92727| 5.86367| 28269.92727



    | Count after transferring surplus votes from 52. BULLOCK Joe (ALP) with a transfer value of

    | 161048.00000/348383.00000. Candidates have surplus votes so surplus votes will be transferred for the next round.


    5|191858.00000|182310.87684|187335.00000|187824.16766|187335.00000|187335.00000|187335.00000| 6.95550| 5012.16766



    | Count after transferring surplus votes from 58. REYNOLDS Linda (LP) with a transfer value of

    | 23746.92727/211081.92727. Candidate 53. PRATT Louise (ALP) has reached the threshold and is elected. The election

    | is over since all seats are filled.

    Winners are 11. WANG Zhenya (PUP), 52. BULLOCK Joe (ALP), 53. PRATT Louise (ALP), 56. JOHNSTON David (LP), 57. CASH

    Michaelia (LP), and 58. REYNOLDS Linda (LP).

  15. I have updated my blog – Waggrakine, Kambalda, etc.

    I also post links to a facebook site with >=3 people claiming to have voted BTL in Waggrakine.

    Yet only one vote was recorded as such.

  16. I take it that there’s no word yet on when we will get word on the recount decision. A day? A week? The AEC should say something.

  17. I’ve not seen anything about the timetable for a decision on the appeal. I don’t know if there is any reason why it must be decided quickly.

  18. Kevin @124 – where’s your source for this? (My work has had my head in the sand today, so I wouldn’t know if it’s been in the meeja)

    Corio @125 – Don’t know why Muir even cares. He could tell they to go stick a dog up their exhaust pipe.

  19. Victoria Senate

    Assange BTL Vote *2 preference Analysis)

    30% of Assange’s BTL Vote (8016) did not remain within Wikileaks

    2331 voters cast their second preference for another party

    5685 stayed within the party with 5340 preferencing Cannold 2nd.

    I have noticed a lot of “dirty data” 020 050 and the like. Not sure how they where interpreted by the AEC computer? Looking more closely they should have been 020 or 050 as they are the numbers missing in sequence. This error could be as a result of data entry error with the operator getting ahead of themselves as they punch in numbers. They should have been picked up in a double data entry check. a 020 could be interpreted as either 02 or 20. Had scrutineers been given access to this data file during the count they could have analyzed the data stream and verified the correctness of entry. Publishing after the event is of little use. Still not sure what a ?? or ? or * means… there are also blank entries so there is many inconsistencies in the data entry process. What is of concern is that the source code for counting the vote is not open sourced. So it is difficult to know just how good and to what extent the code has been properly tested and or certified. More reasons why the data file must be readily available during the count. Ludlum and the Greens should have addressed these issues well before now.

    Assange BTL Vote

    Group 2nd Pref

    AFLP 5
    AIN 13
    AJP 60
    ALP 359
    AMEP 14
    ASP 13
    ASXP 292
    AUC 5
    BRP 3
    BTA 39
    CEC 13
    CYA 3
    DEM 43
    DLP 19
    DRF 22
    FFP 15
    GRN 752
    HMP 40
    Ind 13
    KAP 35
    LDP 5
    LPNP 185
    ODR 2
    ON 2
    PIR 123
    PUP 99
    RPA 5
    RUA 14
    SCSG 21
    SEP 26
    SMK 2
    SOL 20
    SPA 27
    SPP 12
    TCS 5
    UN1 3
    UN2 22
    WKP 5685

  20. “Having considered candidates’ submissions and having further regard to the criticality of particular Senate candidate exclusion counts and the small margins involved in determining Senators’ elect, I have decided that there should be a recount of all formal Western Australian above-the-line ballots and informal ballots – as determined by Divisional Returning Officers for the fifteen WA electoral divisions in accordance with section 273A of the Electoral Act.” AEC Commissioner

  21. d@w
    [Where are the submissions? Should they not be published. ]

    Maybe the AEC will get around to it, or the candidates can publish them if they like. I thought Ludlam already had. You are very quick to find reasons to point the finger at the AEC.

  22. I am use to the AEC’s lack of transparency They should have published the BTL Data file during the scrutiny of the ballot. My prediction is the gap will widen (Unless they find that illusive bundle of misplaced above the line votes)

  23. WA Senate Vote distribution table allocated to van Burgel
    (Coma delimited, copy into Excel)

    2,DI RADO,3,20,6,14,12,16,6,5,6,9,2,12,8,13,7,8,9,13,10,17,12,12,15,19,12,15,14,10,28,22,18,17,25,28,1437,29,29,32,28,37,40,42,66,47,19789,43,78,48,82,78,91,65,100,86,92,104,70,150,72,128,62,167,,,,,,1,,,,
    5,van BURGEL,21285,221,1770,49,108,19,29,10,15,5,8,6,4,,2,4,2,,1,,,,,,,,1,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,

  24. Hello all,

    I have sent Phil Diak at the AEC an email outlining 7 issues of concern I have with the count.

    1. Waggrakine Discrepancy
    2. Fernandez additional vote
    3. Senate vs House
    4. Recount mechanics
    5. Below the Line non-numerical data
    6. Below the Line Zeros
    7. Software availability

    and I am very eagerly awaiting their response.

    The whole email text is a cheery little read, so I suggest you check it out on my blog:

    I will post any response I get to my website and I will also accordingly tweet it.

  25. democracy@work
    [ (Coma delimited, copy into Excel) ]

    Do you think any readers will actually do that? What will I see if I do that? Surely the only people who find such minutiae useful are doing their own comprehensive analysis from AEC data, not waiting around for selections of it to be posted here.

  26. That is subjective. These are the votes that need to be checked. I doubt that the gap will narrow.

    Truth where is your Monto Carlo analysis on teh chances of effecting change? Taking odds?

  27. WA BTL Votes worthy of Interest.

    Unfortunately I do not have a source location list for each batch allocation as this would assist in narrowing down the list even more

    The odds are that the gap between them will widen not swing.

    It is estimated that a recount of the WA Senate will costs taxpayers around $100,000

    Had the AEC provided Scrutineers access to the preference data-file then these votes would have been the subject of more detailed scrutiny during the initial count and possible avoid the need for a recount

  28. Should the WA recount produce a change of outcome or a loser margin in the count should not the Australian Electoral commissioner also invite candidates to nominate any Below-the-line recorded votes to further scrutiny?

    Scrutineers where denied access to the below=the=line preference data files and as such this prevented scrutineers from properly and independently scrutinising the BTL ballot count.

    I guess this question will be raised if and when the issue arises., But it does highlight some of the ongoing concerns at the lack of openness and transparency of the computerised data-entry counting procedures

    Clearly the JSCEM needs to review the regulations and legislation governing the conduct of the ballot to ensure that future elections are open and transparent and subject to full scrutiny and that copies of the BTL data-entry files is readily available and progressively published during the count

Comments Page 3 of 3
1 2 3

Leave a Reply

Your email address will not be published. Required fields are marked *