Senate preferences – unpacking the data


One of the exciting things about the new Senate voting system is that we are on the verge of gaining a lot of new data on how people preference. The AEC has released the full dataset of Senate preferences for the last few elections, but the vast majority of preferences followed pre-registered tickets, so the data was less interesting.

This time, everyone chooses their own preferences, and it appears that the vast majority of voters have marked their own preferences. Using this massive dataset we can do a lot of things to learn more about how voters preference, how many follow how-to-vote cards, and how they preference if they don’t. We can also see how these trends vary between seats, states and booths.

I haven’t had time to do any of this at the moment, but some others have done some cool things with the data. I just wanted to shout out to these people, and then address a possible problem with the method of deciding which senators get six-year terms.

David Barry, who does a lot of cool things with election data at his website, has put together a preference explorer. For voters who voted above the line (most of whom numbered six boxes), you can see how many people followed any particular preference flow.

Grahame Bowland has made his own software which takes the AEC preference data and runs his own count, to verify the accuracy of the AEC count. So far the Tasmanian count has passed the test. His page shows how the count progressed in a more user-friendly way than the official AEC PDF file.

Finally, there was some discussion last night around the interpretation of the section of the electoral act which deals with the special count to identify the top-six senators in each state (section 282).

Under section 282, a special count is conducted between the twelve candidates who have been elected from each state. Any votes which give a first preference for anyone other than these twelve candidates is distributed to one of them, and if there is no preference for an elected candidate, the vote is treated as informal. A new quota is struck as 1/7 of the remaining votes, and a new count is conducted. The Senate is then free to use the results of this count, if they choose, to determine who gets a six-year term and who gets a three-year term.

Grahame Bowland and Dean Ashley (who has built his own version of the counting software) both attempted to do a special recount as specified under section 282, but have discovered an ambiguity which makes it unclear who gets elected.

Dean goes into much more detail at his blog, but the short explanation is that it is unclear whether a below-the-line vote which is originally formal, and has a preference for at least one of the elected candidates, but has less than six preferences for elected candidates, is treated as formal in the recount.

It seems likely to me that any vote that was formal stays formal, as long as there is someone to receive the preference – but I can’t say the definitively from my reading of the legislation.

Liked it? Take a second to support the Tally Room on Patreon!


  1. I wasn’t able to find much analysis about difference in outcomes between the order elected method and the S282 recount method, so I’ve done my own back of an envelope look at it (based on the assumption that the AEC doesn’t have to make any such silly interpretations of legislation):

    TAS: 2LIB – 2ALP -1GRN -1JLN regardless of method used.

    WA: 3LIB – 2ALP – 1 – GRN looks pretty likely with either method.

    SA: Order elected method would be 2LIB – 2ALP – 2NXT. If the S282 recount method were used then the second NXT would be well short of quota prior to the redistribution of votes from unsuccessful candidates. However, they would also start off well ahead of the 1st GRN and 3rd LIB and I don’t think either would be likely to catch up. So, most likely no differences between methods in this state either, but plausible the Greens or Liberals could take a seat from NXT.

    QLD: Order elected would be 3LNP – 2ALP – 1PHON. S282 method could have the same outcome, or could see the 3rd LNP lose out to a GRN. Prior to redistribution of votes from unsuccessful candidates those two candidates would start out pretty even. The 2nd ALP and PHON would start out below quota so would each soak up preferences before they could reach either the 3rd LNP or GRN. Could be close then.

    VIC: Order elected would be 3LNP – 2ALP – 1GRN. Again, this is a distinct possibility under a recount also. However, I think Hinch would have a reasonably good shot at depriving the LNP of their 3rd senator. Plausible he could take it from a Green instead, but doubtful.

    NSW: Order elected would be 3LNP – 3ALP. For a recount the Greens would start out well ahead of the 3rd ALP, LDP and PHON and in my opinion would likely stay there to take a seat from the ALP for a 3LNP – 2ALP – 1GRN result. However, the Green and the 3rd LNP do start out far below quota so something surprising happening cannot be entirely discounted.

    Order elected:
    16 LNP, 13 ALP, 3 GRN, 2NXT, plus Jacqui Lambie and Pauline Hanson

    13-16 LNP, 12 ALP, 3-6GRN, 1-2NXT, 0-1 Hinch, plus Lambie and Hanson

    It’s a distinct possibility then that the only difference between the two methods would be whether Lee Rhiannon or Deborah O’Neill gets a six-year term in NSW. If it came down to such a personality contest then I suspect O’Neil would win, as the ALP would back their own and the LNP are not big fans of Rhiannon. Rhiannon’s best hope then would be for differences in outcome between the two methods in other states.

    If the s282 recount method also deprived the LNP of two long-term senators, giving one each to Hinch and the Greens, then that would pose a dilemma for the ALP, who may be the key swing vote on which method is used.

    The arithmetic of the 2019-2022 Senate is likely to mean that the ALP will need the support of the Greens on any motion opposed by the LNP; The only question being how many additional crossbenchers they would need on board and the number of crossbenchers to choose from. Switching out two LNP for a Green and Hinch would reduce the number of crossbenches needed by one and increase the pool to choose from by one. The loss of O’Neil to a Green would not influence the maths of this at all. So despite loosing a Senator under the S282 recount method, the ALP would find themselves better able to achieve their legislative goals. Doing so, however, would come at quite a cost to the ALP. One less Senator for three years means less funding, staffers losing their jobs and one less Senator to do the party’s work. Personally I think the ALP would be wise to give up a Senator if it meant the LNP gave up two, but I suspect the spoils of office would be rather enticing.

    I originally wrote this in the comment section of Dr Bonham’s blog, but it seems relevant so I’ve copied and pasted it here too.

  2. Not a serious exercise, but out of interest (OK I need a life), I decided to find the SA ballot that was transferred the most times. The answer is a below-the-line ballot that was transferred ten times ( including nine times at full weight) and so sat with eleven different candidates at various points of the count.

    For those with the data it’s Paper 7 from Batch 35 at Collection Point 322, which is Seacliff, in the division of Boothby. The full journey of the ballot was:

    Count 1 – Initially with Mohammad Ali (IND) at full weight (preference 1)
    Count 60 – Transferred to Malcolm Davey (IND) at full weight (preference 2)
    Count 111 – Transferred to Christopher Cochrane (IND) at full weight (preference 3)
    Count 135 – Transferred to Ron Waters (IND) at full weight (preference 4)
    Count 152 – Transferred to Adam Richards (IND) at full eight (preference 5)
    Count 227 – Transferred to Jessica Knight (VEP) at full weight (preference 8)
    Count 336 – Transferred to Tania Noble (AJP) at full weight (preference 10)
    Count 416 – Transferred to Ryan Parker (HMP) at full weight (preference 16)
    Count 426 – Transferred to Steven Burgess (ON) at full weight (preference 42)
    Count 446 – Transferred to Sarah Hanson-Young (GRN) at full weight (preference 59)
    Count 457 – Transferred to Alex McEwen at surplus value 0.02450 (preference 61)

    Preferences 62, 63 and 64 were Alex Gallacher (ALP #3), Bob Day (FFP) and Skye Kakoschke-Moore (NXT).

    For what it’s worth, I think this voter was trying for maximum transfers. The only pattern I could see in the preferences was that they started with very small parties, worked through larger ones and the final preferences seem to have been deliberately given to candidates that might be vying for the last seat.

    The next highest highest number of transfers was eight (so was counted for nine candidates) and there were five such ballots.

Comments are closed.