Sunday, November 13, 2016

Nats Under The Gun In The Orange Postcount

ORANGE, NSW (Nat 21.7 vs ALP)
GAINED by Donato (Shooters Fishers and Farmers) by 50 votes after recount


Key questions:

1. Will Labor overtake Shooters Fishers and Farmers for second place?  (If yes, Nationals win)
Assessment: No

2. If no to 1, will the Shooters Fishers and Farmers catch the Nationals on preferences?
Assessment: Very close, currently Shooters appear to be ahead

Three state by-elections were held in New South Wales today.  Labor very easily retained Canterbury against token opposition, and held Wollongong now that their regular Independent opposition there no longer has the Noreen Hay factor to capitalise upon.  But the third by-election, the one that was always likely to be the most interesting, has lived up to its billing, and then some.

In the by-election for Orange, held by the National Party (and its precursor the Country Party) since 1947, the Nationals have suffered a primary vote swing that is currently running at 35.4%.  Their candidate Scott Barrett leads on primaries on 30.27% with 73.6% of enrolment counted, but is closely followed by the Shooters, Fishers and Farmers' Philip Donato on 24.73%, with Labor's Bernard Fitzsimon on 18.85%.  The rest of the field includes one Green, one Christian Democrat and three independents.

Even if the Nationals retain the seat, the result is still dismal.  This is best seen in the stunning booth swings.  The Nationals were down well over 20 points in all but two booths.  This is partly down to the greater number of candidates running (the SF+F and three indies have replaced only No Land Tax at the last election).  However in three booths they are down by 60 (!) points or more and in two more by over 50.   If these were tiny rural booths with small samples this might be less surprising but one has over 1200 voters!  SF+F are the main beneficiaries of the swing.

At some booths over 80% of voters who voted National last time didn't do so at the by-election.  These are staggering numbers, even by by-election standards, and will come as a very rude wake-up call to the party after a strong performance at the federal election.  The swing is thought to be driven primarily by proposed forced council amalgamations, with the now-retracted ban on greyhound racing also prominently in the mix.  Some are even seeing a Donald Trump factor at work, which seems a pretty long bow to draw, though there is little doubt the US result emboldened the Nats' opposition in the final days.

Labor and the Shooters have a preference-swap, although the impact of this will be blunted somewhat by optional preferential voting.  Of the remaining candidates, the CDP and the independents Duffy and Donald apparently did not direct preferences (though probably few voters would care much if they did) and the Greens preferenced only Labor.  I assume Munro preferenced SF+F ahead of the Nationals.  Union campaigns encouraged putting the Nationals last.

It seems unlikely that Labor can get close enough to the Shooters on primaries in the postcount to overtake them on Green preferences (if this happens the Nationals win easily).  This shouldn't be dismissed entirely though - at present the Greens and Munro, who both preferenced Labor, have 14.9% between them, from which Labor would need to close about six points.  Assuming the final two are Barrett and Donato then the current margin would require Donato to catch Barrett at a rate of 0.123 votes for every available preference.  So for instance if about 60% of preferences exhausted, Donato would need about a 65% split of those that didn't.

The last time this seat had a by-election (1996) final turnout was 89%, and that is at the high end for NSW by-elections, which usually see turnouts somewhere in the 80s (rarely just over 90 and in some cases below 70.)  So there might conceivably be about 16% to go, or it may not be that much.

To place all this in context, at the 2015 NSW state election, the Greens won Ballina with a gain rate of .412 votes per vote, and they also gained at .353 in Lismore - even given the use of optional preferencing.  That was mainly off Labor preferences though, and in this case Labor votes account for less than half of the total.  Even so, given the mood shown by the massive Shooters vote, I would think they could get at least .2 votes/vote if they had to.  If I am right, the Nationals will need to gain very rapidly on the postal votes and other non-ordinary votes to have what only might be a serious chance of retaining. To set the bar that high the Nationals might need to beat the Shooters by about 30 points on remaining primaries (they beat them by 15 points on iVotes and nine points on prepolls).   We'll see, but I am not liking their chances.

As I write there is no two-party preferred count for the seat.  The two-candidate count was scheduled to be between Nationals and Labor, but the NSWEC has realised this is probably wrong and taken it down.  Preferences will have to be thrown between Nationals and SF+F, probably early this week, and that should give us a very much more confident idea of who is winning.

I'll be posting updates on the progress of the post-count to this by-election over coming days.

Sunday: Some more prepolls have been added and they must have been very strong for the Nationals, lifting their total lead on prepolls to 12.5% and their primary lead to almost seven points (31.14 vs 24.35) with 78% counted.  I infer that the Nats beat Donato on these votes by about 28 points.  The required gain rate for Donato on preferences lifts to .153 votes per vote so I'd say the Nationals are not yet out of this.

Sunday afternoon: The Nationals have continued to do strongly on postcount primaries, but it seems it might not be enough.  A batch of 1134 postals has broken 491-147 their favour, with Labor getting 179.  Barrett now leads by 8.14% with 86.4% counted, and the gain rate required for Donato has lifted to .183 votes per vote, and will presumably go higher as more votes are added.  There is also some life in the possibility of Labor coming second as they are now only 5.3% behind, although I still think that's too much.

6:10 The NSWEC have now released preference distributions for several booths between the Nationals and SF+F.  Assuming that SF+F are indeed second, this is grim stuff for the Nationals, as SF+F are leading 53.3:46.7 on preferences from booth votes despite the booth votes being a slightly worse than average sample for them. On booth votes, the gain rate is .215 votes/vote.  On postals, the SF+F gain rate is only .111 votes/vote, below the total they need overall (showing the influence of How-To-Vote cards on booth voting).  I am currently calculating some new projections for the seat.

7:40 I have been crunching today's numbers in projections.  It seems that the strongest predictor of the Shooters, Fishers and Farmers gain on preferences in a booth is their percentage of the primary vote in that booth - they are very strong on preferences where they are strong on primaries. In the booths processed so far they are slightly stronger on preferences than in booths overall, so this suggests that their gain rate on remaining booth votes won't improve.  Based on the projected gain rate of .201 votes/vote I project Donato to win booth votes on preferences by 1469. (Note: I excluded Alectown Hall booth as the numbers are not close to adding up).

We know that SFF gain 162 votes on the Orange prepolls and lose 289 votes on the known postals and also that the Nationals are about 1500 ahead on the primaries for the iVotes and remaining prepolls.  If the count stopped now, this would leave Donato needing a gain rate of only about .06 votes/vote on the remaining prepolls and iVotes to win, which he should just do (by, say, 200 votes) despite his weakness in the prepolls.  But the question is just how many remaining postals remain and what they will do to the required getting rate. As the margin projects to be so close and getting closer, in my view the seat is much too close to call.

Monday 12:30: The NSWEC has uploaded a check count which has Barrett and Donato tied on 11,031 votes after preferences.  However it is immediately obvious that this check count includes a higher proportion of postcount votes (which are on the whole bad for Donato) than booth votes (which are generally good).  Furthermore my projection of the gain rate for Donato on ordinary votes has improved to .210, which is worth just over 100 more votes for him in the projection.  On the downside, his gain rate off the iVotes was a miserable 0.044 votes/vote, which is well below target.  As Antony notes on his coverage, breaking apart the prepoll booth totals is complicated by the way they are reported, making projections very difficult.

Monday 1:20: The tie is broken - Donato leads in the check count by nine.  He's dropped back a little on my projection for the booth votes too.  This remains all very unclear depending on how many votes might still be to come that are not in the primary count.

Monday 2:48: Donato is 76 ahead in the check count which now crucially includes all the known non-ordinary votes (prepolls, ivotes etc).  What we have left now are booths and any remaining postcount votes such as late postals.  I project that with the remaining booths counted Donato will lead by 228, mainly because of a massive blow to the Nats in the Molong booth.

Monday 3:33: Booths all done and Donato leads by 204.  It should be noted though that the NSWEC and ABC differ over the size of the roll, with the NSWEC only showing 81.3% counted.  I don't have any information on what the size of the remaining postcount might be.

Monday 5:10: Institution, Enrolment and Provisional/Silent votes have been added to the primary count but have yet to appear in the 2PP.  The primaries in these break 230-118 to Barrett (mainly off the institution votes) with a further 390 available as preferences.  I would not expect the preferences will do much either way off these so we may well be headed for a sub-100 vote margin which may trigger a recount.  In 2007 two districts with final margins of 68 and 106 votes were recounted.

Monday 5:30: The margin with these votes added is 103 votes, which Antony says is the "final update". However that's probably just final for day; there will be an unknown but probably small number of postals to come.  On the rate of postals so far there would need to be about 400 to overturn the lead, but the other chance for the Nationals would be getting close then winning on a recount.  However because many of the votes have been counted more than normal already, a recount isn't likely to find more than a few dozen votes in errors.  We still don't know for sure that SF+F aren't passed by Labor in the preference cutup, but it seems very unlikely that they are.

Tuesday 2:10: 233 postals have been added, breaking 74-45 to the Nationals on primary votes.  These votes are not yet included in the preference count, but when they are I expect they will benefit the Nationals by around 15-20 votes, pushing the margin below 100.  However if those were all that was to hand I suspect there will be very few to come.

Tuesday 5:10: The Nationals gained 25 votes on postals but the Shooters have gained a vote somewhere else.  The lead is now 79 with 88.2% counted and all primaries counted to preferences.  A formal distribution of preferences will be needed to confirm the Shooters make the final two.  This should occur on Thursday with the postal vote deadline tomorrow at 6 pm.

Wednesday 1:20: Another 62 postals have been counted but not yet included in the check count; they split 21-21 between the contenders so if anything Donato might pick up a few votes here.

Thursday 1:30: Donato won the check-count by 84 votes. The distribution of preferences is being conducted today, which should confirm whether he stayed ahead of Labor.  If he did then the Nationals require errors to win.

Friday 7 am: Late yesterday unofficial reports surfaced that the lead had dramatically flipped in the preference distribution with Barrett now 65 votes ahead! This sort of change from check-count to preference distribution is unusual in a heavily scrutinised count but far from unheard of (in less scrutinised counts I have seen changes as large as 400). The Nationals' copy of the distribution is posted here (thanks to Antony Green).  The Shooters believe there is a 100-vote error.

Friday 11 am: Antony reports that it appears the Shooters were right - a misplaced bundle has been found and they are leading again.

Friday 1:30: More detail from Antony on the error.  In terms of the margin, the error counts for 117 votes in favour of Donato, so all else being equal he is now about 52 ahead.

Friday 5:20: The margin appears to have stayed about that after all rechecks with the Shooters reporting that "final data entry" is occurring and they are up by 55.  However there will now be a recount beginning on Monday and the declaration of the seat will be postponed.

Monday 5 pm: It's over! Donato has won the recount by 50 votes and will be declared as the winner and seated.  The Nationals have lost Orange after 69 years.  There is no such thing as a safe seat anymore.

Is This A Record?

There have been various comments about this being the largest by-election swing against a government in NSW history, but also some that have pushed it further by dropping the qualifiers "by-election" or "NSW".

The 2PP swing doesn't apply because it doesn't involve the same party, but as concerns the primary vote it is surprisingly difficult to find a case of a 34% or greater swing against a sitting Government - excluding cases where somebody was elected unopposed.  Even the bloodbaths of Queensland 2012 and 2015, NSW 2011 and NT 2016 do not appear to have such things.  There was a 36.7% 2PP swing against Labor in Bathurst in 2011 but the primary vote swing against Labor was 32.5%.

Finally I have found the following highly artificial example.  Labor's Frank Scully polled a primary vote of 90.2% in Richmond (Vic) in 1952.  By 1955 Scully had been expelled from the party and was the candidate for the Australian Labor Party (Anti-Communist) which was the proto-DLP.  Scully, alone of the Anti-Communist expelled MPs who had brought down John Cain Snr's Labor government, was re-elected and Labor's primary vote slumped to 39.2%, a swing of 51.0%.

In 1957 in Queensland several seats saw swings above 34% against Labor on the primary vote, and many of these differed from the above in that the Opposition had actually contested the prior election.  However in these cases the sitting member had defected to Queensland Labor (another proto-DLP) before the election.

As yet I have found no Australian case where a government occupying a seat at an election, and with the Opposition having contested it at the previous election, has suffered a primary vote swing of 34%.   (But see comments.)


  1. Seems to me it's a fine point whether you see this as a swing against the Nationals, or just a split in the Nationals. It's not as if there's some huge shift in the general opinions of the average Orange voter - they're just voting for a more angry and emphatic version of the party they've always supported.

    1. Yes, it's hardly a philosophical realignment; it's the voters saying that they haven't changed but they think the Nationals have changed and sold out.

  2. If we're looking for artificial swings, the 1931 federal election in NSW is fertile ground thanks to Lang Labor. Considering Federal Labor (FLP) to represent the government and comparing it to Labor's 1929 vote, we have:
    -19.2 in Eden-Monaro (lost to UAP, sitting MP Cusack FLP but contested Cowper)
    -22.6 in Hume (lost to Country Party, sitting MP Fuller FLP)
    -28.8 in Gwydir (lost to Country Party, sitting MP Cunningham FLP)
    -29.4 in Darling (easily retained, sitting MP Blakeley FLP)
    -31.8 in Riverina (Country Party-held)
    -32.6 in Calare (lost to Country Party, sitting MP Gibbons FLP)
    -34.9 in South Sydney (lost to UAP, sitting MP Riley FLP)
    -37.5 in Macquarie (narrowly lost to UAP, sitting MP Chifley FLP)
    -38.5 in Parramatta (lost to UAP, sitting MP Rowe FLP)
    -38.8 in Robertson (UAP-held)
    -40.5 in Martin (lost to UAP, sitting MP Eldridge Lang but contested Barton)
    -46.1 in Lang (lost to UAP, sitting MP Long FLP)
    -46.3 in Parkes (UAP-held since by-election)
    -46.6 in Reid (lost to Lang Labor, sitting MP Coleman FLP)
    -48.4 in Barton (lost to UAP, sitting MP Tully FLP, also contested by Lang MP for Martin)
    -53.2 in Cook (narrowly retained, sitting MP Riley FLP)
    -55.1 in East Sydney (lost to UAP, sitting MP Ward Lang)
    -56.0 in Werriwa (lost to Country Party, sitting MP Lazzarini Lang)
    -58.3 in Dalley (lost to Lang Labor, sitting MP Theodore FLP)
    and the winner ...
    -71.6 in West Sydney (lost to Lang Labor sitting MP Beasley)

    (FLP also lost Hunter to Lang Labor, with the sitting Lang MP having been elected unopposed in 1929. FLP polled 8.7%. FLP retained Newcastle with the sitting MP Watkins on 30.5%; this had also been uncontested in 1929.)

    1. Thanks, those are great! The Dalley example ticks both boxes of being held by the government at the time of the election and also having been contested by the opposition in both cases.