Below the lines in Tasmania show potential for upset


The Senate data published on election night was very messy.

For a start, it didn’t make any effort to distinguish above-the-line and below-the-line votes, instead throwing all votes for each group into one pile.

Secondly, a lot of votes that will turn out to be formal were treated as informal and thrown into that pile, to be later resolved. Apparently this applied particularly to people who just voted ‘1’ above the line, but it’s not clear how consistent the AEC was with this rule – it’s possible to imagine that this happened more in busier booths and in states with bigger ballots.

In this post I want to focus on the first two problems: incorrect informality and no below-the-line data.

TLDR: There is a lot of below-the-line votes reporting for Lisa Singh and Richard Colbeck, who have the potential to be the first candidates to win seats off the back of below-the-line votes, potentially defeating their party’s preselected ticket. This data also suggests that Labor and the Greens may only get six seats, whereas the party totals suggest seven.

We are now, slowly, starting to see a few polling booths processed, so that informal votes have been correctly assessed and below-the-line votes are assigned to the individual candidate. I’ll look at Tasmania both because there are only five seats to examine, and because of the talk around major party candidates Lisa Singh and Richard Colbeck overcoming their unwinnable ticket position on the back of below-the-line votes.

I should stress that this sample is very small – we only have the divisional office pre-poll from three out of five seats, and one of the special hospital booths from each of these three seats. We’re talking about 1000 votes out of a total of 280,000. I’ll repeat this exercise in coming days to see if it stands.

Firstly – the informal rate drops from 4.7% for the rest of the state to 1.6% in these booths.

Secondly, Lisa Singh and Richard Colbeck appear to be getting enough below-the-line votes to have a chance of overtaking their party order. Singh is currently polling 12.2% of the total Labor vote, and Colbeck is polling 11.5% of the total Liberal vote. As a comparison, in 2013 no candidate on either major party ticket polled over 5% of the party’s vote in Tasmania. It’s also the case that the vast majority of below-the-line votes usually go to the party’s lead candidate, yet Singh is polling 2.4 times the vote of the Labor lead candidate, and Colbeck is polling 1.6 times the vote of the Liberal lead candidate.

If you assume this proportion of the vote stays for the rest of the count, Singh would poll 0.54 quotas in her own right, with the rest of the Labor ticket polling just under 4 quotas. If this is the case there would be no votes left for Labor #5 candidate John Short, and Singh would have a good chance of winning the seat, and would be the beneficiary of any Labor above-the-line preferences.

If we do the same exercise for Colbeck, he would poll 0.474 quotas, and the rest of the ticket would poll 3.65 quotas. This could possibly lead to both the fourth Liberal (Bushby) and Colbeck to win seats, possibly at the expense of either Singh or Greens second candidate Nick McKim.

To understand this better, this is what the race would look like if we make these assumptions. Firstly, the first three Labor candidates, first three Liberal candidates, Greens lead candidate Peter Whish-Wilson and Jacqui Lambie are elected on primary votes.

  • Bilyk (Labor) – 0.866
  • Bushby (Liberal) – 0.608
  • Singh (Labor) – 0.543
  • Colbeck (Liberal) – 0.475
  • McKim (Greens) – 0.4542
  • McCulloch (One Nation) – 0.328
  • Madden (Family First) – 0.253

And these guys would be in a race for the last four seats.

Another thing worth pointing out is that, if you just look at the group totals, it seems very likely that the Greens’ McKim will win a seat, and so Labor will win five. But the more efficient allocation of the Liberal vote between Bushby and Colbeck creates a real risk that either Labor or the Greens might lose.

Bear in mind that my latest predictions had Labor + Greens + NXT on 39 seats – a slim majority. If they lose a seat in Tasmania, they may not have that majority, although it looks like they would have enough to block legislation from the (likely) Coalition government.

As I said earlier, this is very small sample – I’ll do this same analysis again once we have more data.

PS: One more thing to note: we are still waiting for a bunch of primary votes to be added (although Tasmania is more complete than others) and on the votes counted so far we’ve seen the Greens position improve 0.3% since Monday evening. I don’t know how much more that will improve but bear that in mind.

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


  1. Well, if some odd results end up discouraging the parties from dropping popular Senate incumbents to unwinnable positions in the future, I think that’s probably a good thing.

  2. I worked at a medium sized booth in the ACT, around 2.1k votes. There, Senate votes were formal if there were either 6 valid preferences below the line or a single 1st preference above the line. There were lots of votes completed both above and below the line, and if both were formal, they were counted in the BTL total.

  3. It may ENCOURAGE parties to run popular candidates low on the ticket as sometimes this can grant them an additional senator.

  4. Keep an eye on the ACT as well. There may be more than a few “Liberal” votes that are actually below the line votes for the support candidate with Zed Seselja much further down or left off.

  5. In some case – putting the popular candidate lower down may mean you end up with both getting elected.

    In other cases I suspect you end up with neither

  6. In Victoria there was talk amongst Greens voters to preference Janet Rice ahead of Richard DiNatale (she definitely has comparable if not higher personal vote than DiNatale + general dissatisfaction with DiNatale) if the trend in Tasmania is anything to go by I imagine Janet Rice could benefit from first pref below the line votes. Another one to keep an eye on.

  7. Neither elected seems very unlikely. If one excludes ATL will flow to the down ticket candidate and I’d image the down ticket BTL votes would typically flow up.

  8. This shows a couple of interesting things about voters in Tasmania. First, they’re not prepared to put up with party room shenanigans in “punishing” candidates by putting them in so-called unwinnable spots. Secondly, because Tasmania voters are used to Robson rotation of parties/candidates on ballot papers, they are used to how-to-vote cards having less influence. Thirdly, because grouped candidates appear alphabetically within that group voters are used to searching for the candidate they want, rather than numbering from the top down if the candidates were in the party’s preferred order.

    ACT is similar – I wonder if there will be a similar effect.

  9. These BTL votes we have so far are very unrepresentative – a disproportionate influence of Special Hospital Team votes (plus also some Divisional Office prepolls). It is difficult to get a clear impression in scrutineering (other than booth samples on the night, which are few) because of the level of stratification in the votes coming through and the vast variation between electorates and parts of electorates. The votes for Colbeck and Singh are certainly large and the risk of McKim being trapped behind four major party candidates seems to exist, but it will be a while before we have any remotely accurate handle on it.

    Another trap for beginners – for some booths they are now including the ATLs and not the BTLs, which is highly annoying.

  10. Re how to vote cards – it’s not very relevant because the 1 Liberal preferences won’t be distributed unless Colbeck has quota but the spray of them is incredible. Maybe 20% follow the card (2 to CDP, 3 to SFF etc) but the rest go absolutely everywhere, with Labor probably getting the biggest slice. I saw 1 Liberal 2 Animal Justice Party four times in a few hours. Minor party preferences spray a lot too.

  11. Hi Ben. You mention that your latest predictions had Labor + Greens + NXT on 39 seats – a slim majority. Have you published this prediction analysis anywhere? Great work. Cheers.

  12. kme:

    “Well, if some odd results end up discouraging the parties from dropping popular Senate incumbents to unwinnable positions in the future, I think that’s probably a good thing.”

    Alas in terms of learning the lesson, Labor currently looks like benefiting from doing so (since there is a good chance both Singh and Bilyk will win) which would be doubly unjust given that they voted against the system which may save them from their own stupidity.

    Senator Polley today tweeted a photo of the Labor Senate team that conspicuously didn’t include Singh.

    Senator Bilyk (ALP number 4) is on zero below-the-line votes in Braddon out of 1171. The only other candidates on zero BTLs in Braddon are the two CEC candidates and 10 supporting micro-party candidates.

    The BTL rate in Tasmania is looking encouragingly high, I have it tracking for about 28%.

    The data provided is a nightmare because some booths are uploaded segregated into Ticket Votes and Unapportioned.

  13. One thing Tom Clement pointed out to me is that, if two candidates of the same party are both in the race, all of the ATL preferences from other parties will favour the higher-ranked candidate until they’re elected.

    So if Singh is on say 0.9 quotas and Bilyk is on 0.48 quotas (which is what the latest data suggests if you don’t adjust by electorate) then anyone who votes say LDP 1 ALP 2 will go to Bilyk. Which is probably what you want.

  14. Ben – while that’s true, I would imagine that a lot of BTL preferences would end up also going to Singh first – people aren’t likely to vote BTL if they’re going to go down the columns in the first place, and if Singh has run such a strong campaign, I could see Greens voters, for example, voting below the line to have their preferences flow to Singh rather than other Labor candidates.

  15. My hope would be the lesson taken is that BTL voting in the senate is the way to go!

    Unfortunately I can’t see that really going beyond Tasmania or ACT (which in the latter it really would make no difference)

  16. On current numbers I fail to see how Lambie can get two. She is only on 1.1:quotas. I don’tunderstand why journalists claim she is in the running. It looks like 5/4/2/Lambie to me.

  17. QO – it’s all down to how the preferences flow. I could see most voters who went for Hinch, LDP, PUP, and Xenophon putting Lambie as their second (or, in LDP case, 5th) preference in relatively large proportions. Hinch+LDP+PUP preferences would push Lambie well ahead of Xenophon, and then Xenophon preferences would push Lambie up into a fairly strong position.

    From there, if they stay ahead of Family First, then when Family First is knocked out, ALA preferences flow to Lambie.

    And from there, it’s mostly preference leakage and vote exhaustion, unless Labor gets particularly bad preference flow and gets knocked out (in which case Greens overflow will see many Labor votes flowing to Lambie).

    Note that I don’t think she’ll get a second seat. But it’s certainly not implausible.

  18. The spray in preferencing here is such that assumptions based on HTV cards just do not work at all. Assumptions can be made based on ideological-fellow-travelling only and even then the flows are fairly weak. I don’t think any party will be getting half of any other party’s preferences, and I don’t give Lambie any chance at all of getting two.

    Regarding BTLs there is significant “leakage” within the party tickets, eg quite a few people voting BTL 1 Abetz 2 Colbeck. Also a fair bit of BTL preferencing to Singh.

    Adjusting by electorate is important – Singh is on about 10% in Denison and Franklin combined and a few percent everywhere else, Colbeck is on 9 in Braddon but not breaking 5 outside it. I currently have .77 for Singh, .63 for Colbeck.

  19. Any update on that Vic Greens BTL boilover situation? How’s that theory working out so far?

Comments are closed.