WA Senate – how Pratt could win


Yesterday’s post predicted that the most likely outcome in the WA Senate election is that Liberal candidate Linda Reynolds will widen her lead over the ALP’s Louise Pratt, thanks to a large batch of postal votes yet to be counted.

I still think this is the most likely outcome, but since posting yesterday a number of points have been raised that I think are valid, and suggest ways that Pratt could perform better than my projection (which was very similar to William Bowe’s at Poll Bludger).

It is possible that the Liberal Party may suffer more serious leakage from party tickets, as Reynolds is relying on more preferences than Pratt. However if Pratt and Ludlam have performed strongly on below-the-line votes, more of these votes could be ruled informal in coming weeks, and this could partially cancel out any leakage benefit.

There is a scenario where a slightly higher Palmer United vote results in an earlier election, and frees up more votes to flow to the ALP. It is also possible that Friday’s story about Joe Bullock pushed down the election-day vote, which may mean our projections are too pessimistic when trying to predict Labor’s share of postal votes.

I explain these theories in more detail below the fold.

I was asked by a number of people on Twitter and in the comments section about whether Pratt could benefit from below-the-line votes. It so happens that I had originally written a section of my blog post dealing with this issue, but I then found information that suggested to me that below-the-line votes have already been included in the party totals, so my point was moot.

At the 2013 election, the Greens performed very strongly in below-the-line votes, while the Liberal Party performed poorly. The ALP also underperformed, although not as badly as the Liberal Party. With all of the focus on the first Labor candidate and his significant political differences with his running mate Louise Pratt, it is conceivable that a much larger number of voters would have chosen to vote for Pratt below-the-line, and it is reasonable to assume the Greens also continued to over-perform in below-the-line votes, and these votes would have favoured Pratt.

However, this is only relevant if these votes are not already in the count. If they are already in the count, then we have already factored these votes into the current margin, by assuming they are above-the-line votes for the party they are counted for. Any BTLs for Pratt and Ludlam will have roughly the same result for Pratt as ATLs for their respective parties.

I am reasonably convinced that the AEC has already included below-the-line votes in the party totals, for those polling places that have had results reported. If I am wrong, this will change things and help Pratt, but I don’t think this will be the case.

Even if we have already factored BTLs into the count, Pratt may still benefit from them. At the moment, we have no idea how many below-the-line votes exist, as they have been lumped in with above-the-line votes in party totals. The ABC calculator assumes all votes are above-the-line and transfers all votes according to a party’s group voting ticket.

The Liberal Party is relying on a larger number of preferences than the ALP is in that final race. It is likely that a larger proportion of the Liberal Party’s third quota will be made up of below-the-line votes that may not flow in the way that the calculator predicts. While both parties will suffer from preference leakage, the party that relies on more preferences will be more vulnerable.

Pratt and her supporters should still be worried about the expected large numbers of BTLs cast for Ludlam and Pratt. Below-the-line votes were not checked for formality on election night, and thus some of these votes are likely to be informal. If Labor and the Greens are overperforming on below-the-line votes, they will likely see more of their votes lost to informality as the count continues.

In addition, we are assuming that all BTL votes for Ludlam flow to Pratt, but some may not, which may partially outweigh Labor’s advantage in terms of preference leakage if Ludlam has performed particularly well on BTLs.

It’s also worth considering one particular key point in the count that has been identified by Kevin Bonham. In the current count, the Palmer United Party is elected on preferences from HEMP. However, if the PUP vote rises very slightly, they win their seat one count earlier, on preferences from the Liberal Democratic Party. Preferences from a number of small parties that are sitting with HEMP thus flow directly to the ALP, instead of flowing through Palmer United. It also means that the PUP primary vote mostly stays with PUP, as Palmer’s candidate would be elected with a much smaller surplus.

This would assist Pratt, but it’s difficult to say how likely it is. When I tried to achieve this outcome on the modeled results I produced yesterday, I needed to increase the PUP vote by 0.5% – rather a lot to expect. I also think that, on its own, it’s unlikely to produce enough of a shift to outweigh any benefit Reynolds will gain from postal votes.

The final point worth considering is whether the ALP may have performed better amongst voters who cast their ballots earlier. It’s a point of contention as to whether the story about Joe Bullock attacking his party on Friday produced a substantial impact on the ALP’s vote. If it did, it’s conceivable that the ALP’s vote may have been higher amongst postal votes, and thus these votes, while still providing a strong Liberal vote, may not be as bad as we are projecting.

It’s possible someone could test this by examining the difference in swing in each seat between the pre-poll voting centres (which have already been counted), and the election-day polling places. Unfortunately I haven’t had a chance to do so.

Overall, this race is very close, and is still in play. We’re unlikely to know who won until the button is pushed later this month. While Reynolds is the favourite, there are numerous factors that could shift the race in Pratt’s direction.

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


  1. “However, if the PUP vote rises very slightly, they win their seat one count earlier, on preferences from the Liberal Democratic Party”
    Actually on preferences distributed when the LDP are excluded, but not the LDP preferences themselves, which go to HEMP and then the Liberals.

  2. Good post, but I disagree with one point. You seem to believe bros will be taken out as the formality checks are done, hurting those with lots of them, including Scott.

    However my experience is that many polling booths throw anything that looks slightly suss below the line into the informal pile, and the net effect of formality checks is to see more added in than taken out. If this occurs again it will help Pratt, though possibly by a very small amount.

  3. Ben, the most important thing to remember is that the Nationals, Palmer United and Liberal Democrats preferences will be the last to be counted and with that in mind, the most likely outcome should be the Liberals Linda Reynolds.

  4. I don’t understand the HEMP comment.

    When HEMP is eliminated – HEMP’s preferences along with left wing minors (Socialist Alliance, Wikileaks, etc) all go straight to the ALP, they don’t go via Palmer.

    Palmer is then eliminated, and barring a few split tickets, all the excess from Palmer & right wing minors to go the Libs.

    Wouldn’t the ALP have a better chance of winning if Palmers surplus quota is significantly reduced?

  5. Sec, quite a lot of parties’ preferences flow to HEMP, and when they are excluded they flow either to Liberal or Palmer. Just under half of the preferences that build up with HEMP flow to Labor, the rest go elsewhere.

    This means Palmer’s vote goes from 0.9929 quotas to 1.1004 quotas. This means that all of Palmer’s votes (which heavily favour the Liberal Party) flow on at a value of 1/10th of a quota.

    If Palmer gets elected on the previous round with basically no surplus, then the HEMP preferences flow on at full value – and they favour Labor more than Palmer’s entire voting bloc does.

  6. Hang on, this is a bit confusing.

    When HEMP is excluded their preferences go to Labor. A number of parties whose preferences had been parked with HEMP then flow to PUP, but of those it’s only the half ot the Democrats and half of the Katter votes that then flow to Labor after PUP, the others all go to the Liberals.

    That 50% of Democrat and Katter votes gets reduced in value if it has to go through PUP first and then flows on as part of the PUP surplus. If PUP is already elected before HEMP is eliminated they flow on at full value to Labor, but if I’m following that correctly it’s only this 0.17% of the vote (half of Dems & KAP) which is affected. Could still be crucial though.

  7. I had the explanation of the tipping point slightly wrong and have fixed it now. The AMEP preferences actually cut the other way but because they are diluted in the best case scenario for Labor they have little impact.

    Trying to do some modelling of the BTLs but dealing with data files with 3,000,000 items takes a while for me!

  8. I’ve been looking today at whether BTL leakage could get PUP elected before the HEMP elimination. I currently have PUP picking up between 2800 and 4300 votes to that point, above what they would get if all votes were ticket votes.

    My methodology involves using the BTL preferences for each party from 2013 and some assumptions about what that projects to in terms of BTL votes for 2014. The range of outcomes reflects different assumptions about how “sticky” BTL votes are compared to overall votes for each party, and also the extent to which PUP preferences are assumed to increase along with their primary vote. There are also some crude assumptions about BTL totals for the new parties and their preference flows.

    The big sources of leakage to PUP are the Nationals (approx 500-1000 votes depending on other assumptions), AUC (350-500), WKP (275-400) and the Shooters (300-350) but they get a few from everywhere.

    Based on Ben’s projection above of PUP being 0.5% short of election before HEMP go out, I don’t think BTL leakage will quite get them there, but it may make it very interesting.

    Assuming PUP do still need the HEMP bloc to get over the line, the extra BTL preferences gathered will increase their surplus, which of course flows almost completely to the Liberals based on the tickets of PUP and the right-wing coterie that they collect earlier in the count.

    I see from Kevin Bonham’s blog (and now his comment here) that he’s attempting a possibly similar exercise to what I’ve done. He may well do it better. I only ended up going down this path as part of a refinement of my attempt to forecast BTL impacts on the final Liberal-ALP matchup. It turns out that my previous attempt there (suggesting a 2500 vote swing to the ALP) was too simplistic to the point of being on the unbaked side of half-baked. (I hadn’t modelled the Palmer leakage properly and my assumptions about the new parties were unsound).

    Looking purely at leakage of BTL preferences, I’m now getting essentially a break-even between Liberal and the ALP – within 500 votes either way. There’s leakage in both directions and its actually the ALP that gets more hurt by leakage to PUP (from tickets like WKP, HMP and ASXP) although the Liberals have their woes with about a quarter of the National BTL vote going to the ALP and another quarter to PUP.

    However this does not include the impact of BTL leakage on ABOVE-the-line flows via an increased PUP surplus. Adding in that effect, the overall impact of BTL leakage may favour the Liberals by c. 3000 votes in the final matchup IF it doesn’t lead to PUP getting elected earlier.

  9. @Stephen (comment #2),

    I originally thought you had a point with your comment, and that matched my previous scrutineering experience. However Antony Green in comments on his latest post has explained that actually a change in AEC practice means BTLs are overstated:

    “There is also a problem that changes in counting procedures means that the number of below the line votes will initially be inflated compared to last year’s count. The number will fall once data entry begins. This change in counting procedure is a result of last year’s re-count. The re-count revealed that votes were being incorrectly excluded as informal by Returning Officers, so this time any ballot paper with a mark below the line will come to the central scrutiny centre, and will so initially show up as a below the line vote. Once data entry begins, there will be a decline in the number of formal below the line votes, a pattern quite different from last year’s count.”

  10. Regarding informal below-the-line votes, I’m not aware that the AEC has ever offered a split of informal votes into those that were attempted BTL votes, versus being informal for other reasons.

    I tried to come up with an estimate based on the identification numbers assigned to ballots in the BTL data released by the AEC. Each ballot is identified by a batch number (which corresponds to a booth) and a “paper number” within that batch. In most batches the paper numbers are largely consecutive from ‘1’ but with occasional skipped numbers. My assumption was that the skipped numbers relate to ballots that were assigned a number and entered into the computer, but were badly enough numbered that they couldn’t be rescued as formal. The released data comprises only formal ballots.

    Using this methodology I came up with an estimate of 1.4% informal on the original count and 1.5% on the recount, with 94% of the ballots that were additionally assigned a number in the recount ending up as formal, resulting in an additional 701 BTL votes.

    The informal estimates exclude the Geraldton-Waggrakine booth at which there was clearly a problem with the original count, with my theory being that the BTL and informal bundles got mixed up for a period.

    Approx. 13% of formal BTL ballots were imperfectly numbered. Compared to that, an informal rate of 1.5% in the recount seems low, although I guess it’s conceivable that the saving provisions are rescuing the vast majority of genuine attempts to vote BTL.

    However it’s also possible that even in the re-count, there was a level of scrutiny being applied before a potential BTL ballot went for data entry. Given that the recount was being conducted by experienced AEC personnel with scrutineers present, they may still have been willing to reject clearly informal BTL ballots without attempting data entry, something that they’re now not willing to delegate to the less experienced personnel that do counting on election night, possibly leading to a larger initial over-count as Antony Green is suggesting.

  11. Not quite sure why this has happened (probably formality checking as discussed by Ben above) but the % counted has gone backwards slightly and Labor has lost over 1000 votes. The upshot is that on current figures, PUP do cross before the tipping point but Labor still lose, assuming BTLs don’t help them much.

    This is probably the first time since counting finished on election night that the current count as correctly interpreted would return 3-1-1-1.

  12. Some of the damage tonight was just data re-entry where votes were temporarily removed and re-entered but even so Labor is down several hundred at least in net position for the day.

  13. @ David Barry – thanks for this.

    A quick look appears to confirm that there must have been a lot of filtering applied in 2013 before assigning an ID for data entry, even in the recount.

    I just looked at NSW for the 2001 election (as the state with the biggest ballot – still much smaller than the 2014 WA one). There were 114,125 formal BTL votes in the final results there, while the survey numbers indicate there were 12,094 that failed the 90% test. So even if it was just those that were included in an initial count, you’d be looking at losing 9.6% of the original BTL count.

    If the original count also includes ballots with “Large Number of repeating numbers or missing numbers Below the Line” that increases to 12.3%. If it also includes “Ticks and Crosses below the Line” you get to 15.7% and if it includes “First preference only Below the Line”, it’s a massive 33.3%.

    I can’t see how any count could include “More than one number 1 Below the Line” or “No First Preference below the Line” because there’s no basis to assign to a candidate in the first place.

  14. Large gains to Libs on early postals. 10-13 points up on booth votes in five different electorates. Seems a very high degree of confidence is justified now that things will pan out as projected and Reynolds will win.

  15. I served as a scrutineer at the 2013 election in Canberra, and it was interesting how it played out in terms of BTL votes.

    The procedure we observed with regards to BTL votes was:

    – 1st prefs from BTL votes from Election Day booths were counted and allocated to the relevant candidate on the night (this has certainly happened in WA)

    – however, 1st prefs from BLT pre-poll votes weren’t allocated until weeks later, as they passed through central scrutiny (where BLT votes are manually entered into the preference calculator.

    This is why there is now a BLT PPVC option listed on the AEC website when results are listed by polling place.

    This is going to be a small batch (3-4% of the 10-15% of votes cast as pre-polls) but given past peformance, may provide Pratt to close the gap by 1000 votes or so.

  16. The result is clear now……….libs win their 3rd seat.
    I saw something that suggested the ALP postal vote exceeded their booth vote by 5 %
    a late swing against

  17. I’ve run some pre-poll, absent and postal projections into my BTL modelling and agree that its all very safe for the Liberals due to their own postal performance and a lower Greens result.

    Without considering BTL leakage, I have PUP 4,100 short of being elected on LDP preferences and Liberal beating Labor by 11,300 for the final seat.

    Adding in my base BTL scenario, PUP moves up to be only 1,000 short of the tipping point. Labor actually do a little better on direct BTL leakage, but PUP steal from both sides and this then gets converted into extra ATL votes for the Liberals from an increased PUP surplus, so the Liberal lead over Labor expands to 13,500.

    If I give PUP an extra 1500 ATL votes so that BTL leakage puts them over the tipping point, the Liberals end up with most of those extras to get PUP to the tipping point, then Labor gain about 3000 from the crossing of the tipping point, plus a further 1600 from a different pattern of BTL leakage (the Liberal vote is now coming from sources a little more prone to leakage) for a final Liberal margin of about 10,200.

Comments are closed.