UPDATE: Results in: Greens 20, Coalition (Nationals) 21, Hanson 22. Final upper house result: Coalition 11, Labor 5, Greens 3, Christian Democratic 1, Shooters and Fishers 1.
Tuesday, April 12. I’ve allowed this post to go dormant since a week after the election, since when the last lower house seat of Balmain was decided in the Greens’ favour. The big news now is that the button will be pushed on the Legislative Council count this morning, and that Pauline Hanson’s chances have firmed considerably after she moved ahead of not only the third Green, but also the eleventh Coalition candidate. This leaves these three candidates battling for the last two places. The general expectation was that Labor preferences would put the Greens ahead of Hanson, but there now seems an even money chance that she will win a seat all the same. Antony Green explains all. I’ve changed the time stamp on this post to move it to the top of the page, to allow easier access for anyone who wants to comment on the events as they unfold.
Thursday. The NSWEC hasn’t updated the figures, but the ABC results and various news reports tell us absent votes have boosted the Greens to a 218 vote lead over Labor, which Verity Firth will now have to rein in on independent and minor party preferences. Another coat of paint has been removed from Nathan Rees’ lead in Toongabbie, but his lead may be enough.
Wednesday. Labor’s lead has narrowed in Toongabbie, East Hills continues to drift away from them and Balmain remains as much of a wild card as it always was. However, Noreen Hay now looks safe in Wollongong. Nothing today from Oatley.
Tuesday. No further progress in East Hills or Balmain, but Oatley has slipped from Labor’s grasp in today’s counting and the margin in Wollongong has been cut still finer. The latter will come down to absent votes, none of which have been added yet a clear trend one way or the other would decide the result.
Monday. Late counting has seen any hope for Labor go in Monaro and almost certainly Swansea as well, and things are souring for them in Wollongong as well. East Hills and Oatley are still too close to call, and Balmain remains a wild card. The numbers are thus Coalition 67, Labor 19 and independents three with four in doubt, one of which could go to the Greens. The other turn-up today is that Legislative Council counting has put Pauline Hanson on to the ABC computer’s projection to win the final seat. Antony Green has written a post on why he thinks this unlikely but not impossible more on this at the bottom of the page.
Sunday. Excluding seats where the ABC computer has the margin at less than 2 per cent, the numbers currently stand at Coalition 64 (Liberal 47 and Nationals 17), Labor 18 and three independents. That leaves eight seats in doubt, although in some cases not really. These will be dealt with in turn below. The tables show the two-candidate preferred counts using the most complete figures available, swings for each type of vote matched against the equivalent result from 2007, the number of exhausted votes, the total number of formal votes counted and to give some sense of how many votes there might be outstanding for a given vote type the total number of such votes from 2007.
The NSWEC publishes election night and post-election night figures of the polling booth results, with the latter being the re-checks. In some cases the latter are not fully completed, and it is these partly complete figures which show on the electorate summary pages on the NSWEC site (although the results table on the index page uses the election night figures). Where this is the case, I have used the complete election night figures rather than the incomplete post-election night ones.
EAST HILLS (Margin: 14.1%)
Wednesday. Continues to drift away from Labor, with 3742 absent votes increasing the Liberal lead from 207 to 303.
Sunday. The Liberals led by two whole votes on polling booth figures, but they have gained ground today with 1860 pre-poll votes breaking 954-741 their way.
OATLEY (Margin: 14.4%)
Tuesday. Labor’s gain on pre-polls has been pretty much reversed by the addition of 3000 postals which have added 232 to the Liberal margin, now 321.
Sunday. The Liberal candidate had a 332 vote lead on polling booth votes, but Labor member Kevin Greene has chased down 243 with the addition of 3055 pre-polls.
SWANSEA (Margin: 10.8%)
Wednesday. Another 400 postal votes added, breaking 194 to 160 and increasing the very secure Liberal lead to 825.
Tuesday. Labor has picked up 43 votes from 3462 postals, which have gone 1544-1501, but it’s too little too late.
Sunday. Labor’s Robert Coombs trailed by 491 votes on the polling booths, and has gone a further 318 votes backwards with the addition of 1883 pre-polls and 43 institution votes.
WOLLONGONG (Labor vs Independent)
Wednesday. Absent votes have indeed behaved different to pre-polls and postals, favouring Labor 615-445. This has increased Noreen Hay’s lead to 442, enough for her to claim victory.
Tuesday. Another 1406 postals have maintained the trend of the first 1783 in shaving 111 off the Labor lead, which is now down to 263. However, with pre-polls presumably done with and the addition of postal votes down to a trickle, most outstanding votes are absents, and these may well behave very differently.
Monday. The two-candidate count between Labor’s Noreen Hay and independent challenger Gordon Bradbery made Hay appear home and hosed, with a margin of 2.5 per cent off the polling booth votes. However, subsequent counting has gone disastrously for her: pre-polls have favoured Bradbery by a remarkable 2173-1300, and he has further gained 766-680 on postals. This has whittled Hay’s lead down to 389, with the trend running heavily against her.
MONARO (Margin: 6.3%)
Monday. With 674 pre-polls breaking 3578-2890 the Nationals’ way, John Barilaro now holds an unassailable of 1275.
Sunday. The Nationals have a 1 per cent lead which it would take something remarkable to undo. The addition of 4300 pre-polls haven’t provided it, going 2108 to 1957 the way of Nationals candidate John Barilaro, who now leads Labor member Steve Whan by 754 votes.
BALMAIN (Margin: 3.8% versus Greens)
Thursday. The Greens have reportedly moved to a 203 vote lead over Labor on the primary vote, but the NSWEC figures haven’t been updated. The ABC figure has the lead at 218. Their challenge now is to keep that lead with the distribution of independent and minor party preferences, including those of Maire Sheehan, a council rival of Greens candidate Jamie Parker who polled 1373 votes.
Wednesday. About 4300 more votes have been added, mostly postals, and they have very much reflected the overall trend in slightly favouring the Liberal candidate (1468 votes) with Labor (1303) just shading the Greens (1274) for second place. However, this does not reflect the trend of 2007 when Labor did much better on postals than on ordinary votes (44.3% compared with 39.6%), and the Greens much worse (24.1% compared with 29.5%). The two main types of vote yet to be added, pre-polls and absents, were much stronger for the Greens. However, any lead the Greens open with the addition of these votes will have to be defended against a probable flow of independent preferences to Labor. In any event, Labor are currently ahead of the Greens by 139 votes, up from 111.
Sunday. The Liberals hold a narrow lead on the primary vote, with Labor and the Greens mixing it on 30.4 per cent and 30.0 per cent respectively. Given the likelihood the Liberals will stay in front, the NSWEC’s Labor-versus-Greens count is of little use. What matters is who out of Labor and the Greens finishes second, as I would assume that whichever of the two makes it to second will then overtake the Liberals on the other’s preferences. The precedent of 2007, when post-election night counting saw Labor’s vote fall 0.3 per cent and the Greens hold steady, suggests there won’t be much in it.
Thursday. Pre-polls and enrolment new votes have gone 546-521 in favour of the Liberals, and Nathan Rees’s lead is now down to 194.
Wednesday. Absents and pre-polls have strongly favoured the Liberals with Nathan Rees holding his ground on postals; taken together, the Labor lead is down to 285.
Sunday Nathan Rees led by 409 with the counting of polling booth votes, but he’s down 16 with the addition of 945 pre-polls and institution votes.
Newcastle. With the Liberals 1.8 per cent in front, I won’t be making the effort to follow this one.
Monday. It is clear enough that the Coalition will win 11 of the 21 new seats, Labor five, the Greens two, and the Christian Democratic Party and Shooters and Fishers one apiece. The final seat is a tussle between Labor, the Greens and, improbably, Pauline Hanson. As of today the ABC computer projection has Hanson in front, but this projection assumes no preferences, which is a very unsafe assumption where Labor and Greens candidates are involved. The most likely result is that whoever out of Labor and the Greens is excluded will deliver the seats to the other on preferences especially if it’s Labor which is excluded, given their how-to-vote card directed preferences to the Greens. However, as Antony Green notes, Pauline Hanson does uniquely well among minor candidates in polling strongly on the below-the-line votes that remain to be counted, so there is some chance she could get up thanks to exhausting Greens votes if Labor stays ahead of them.i>
530 comments on “NSW election: photo finishes”
So she lost because her supporters were too stupid to even fill out a ballot paper correctly?
Now that’s what I call divine justice!
In the history of Senate and NSW Legislative Council elections, no candidate has ever attracted as many below the line votes as Pauline Hanson. She attracted twice as many formal below the line votes as the Greens at the recent election despite getting only a fifth of the overall vote.
At the 2004 Senate election, she attractred 50,000 below the line votes in Queensland, and they had to number from 1 to 50 for a valid vote. They seemed to know what they were doing, and put considerably more effort into the process than those who just picked a box above the line.
Pauline Hanson raises a perfectly valid question that anyone interested in democracy should take seriously, which is whether the current structure of the ballot paper can cause confusion and induce informal votes. Why blame voters for being ignorant, criticise legislators and adminstrators who produce impractical forms of voting that confuse people.
Criticise Hanson supporters for being mis-guided or whatever, but it is the height of arrogance to simply call them stupid. We all get a vote, whatever our education levels, and we should cherish that right, and demand that the votes are cast and counted sensibly, whatever we think of the views of those who cast the votes.
Good point Antony. Its important for the voters to have their say about who gets elected but the current ballot, both federal and state, is ridiculous! And, if it was the intent of a sufficient number of voters for Pauline Hanson to be elected then so be it, I don’t like the idea of voter where the intent was clear being classified as invalid.
In other words I think the rules should be easy to follow as possible and the number of candidates limited to a pre-determined small number* and those with the most members/signed supporters get on the ballot and everyone else goes back and gets more members/signed supporters for next time!
*Why do we need 42 in the NSW State Upper House anyway???
Hanson asking for a recount.
Antony Green @ 502
Can you recall a previous STV election in Australia in which four candidates were elected on the final count with less than a quota?
James J @ 505
Mrs Hanson will be wasting her time and money seeking a recount on the grounds cited in the story. Regarding her concern that some candidates or groups may not have access to above-the-line voting, that was raised unsuccessfully before the High Court in 1984 by Cyril John McKenzie, the candidate of the “Party to Expose the Petrov Conspiracy”. The use of random sampling also came indirectly before the High Court in 1965, in relation to a recount from the Victorian Senate election of 1964.
Most who give a hoot about these things would agree that a fractional transfer system is preferable to a random sampling system, especially when the system is computerised anyway. But if you must for constitutional reasons have a random sampling system, it is much better to do it using computer generated random numbers than to rely on officials to just physically grab papers “at random”. It would, however, be interesting to know exactly what sort of random number generator is used in the NSWEC computer program.
Pauline is arguing, among other things, that random sampling is inherently unfair because a computer does it. She wants a manual count. She does not seem to realise that a manual count would also have to be done using random sampling of surpluses.
It is arguable that a computer count would be fairer because it would be testably random, whereas a manual sample might be biased. I am not sure how you actually randomly sample very large numbers of ballot papers scattered around a warehouse. The Senate Scrutiny Procedures handbook of 1984 contains a wealth of detail of how to carry out a manual count with 3 million ballot papers piled into huge stacks, but by this stage, the method was already Hare-Clarke.
It’s easy enough to “randomly” sample SATLS, because they are all the same. But BTLs and RATLs are another matter. In the LC, there were 17 surpluses from 3 parties to distribute via random sampling. In these, there were 12 to 20 thousand BTLs and a totally unknown number of RATLs.
On the question of variability of result as a consequence of random sampling, it would certainly occur and affect the numbers at any one count. It is hardly likely to affect who was elected, but it nearly did in Manly LG in 2008. The initial gap was 24 votes for the last seat. A recount produced a 12 vote gap (but there were also errors in the primaries). Simulation showed that the gap could have been reversed in about 1 out of 18 re-runs.
In the LC election of 2011, the smallest gap was Hanson’s lead of 207 over Buckingham at the 2nd-last count. It is arguable that this could have been a gap the other way on random resampling… Pauline would not like that. However, the second-last count was some 290 counts downstream from the last random sampling event- of which there were 17 and the chances of all 17 being favourably resampled towards her is 1 in 2^17, which 1:130,000. Now that is a BIG ask .
At any one count, the variability of the sampled number is binomially (actually hypergeometrically) distributed and a function of the square root of the number and the “chance” you want to take (usually 5%). I haven’t got my head around the variability associated with running 17 such tests in sequence but it would be akin to the relation of the standard error to the standard deviation in parametric statistics….. that is to say, if the 95% variability on any distribution is 100 (which is about right here), the variability over 17 successive distributions is probably 100 / sqrt(17), or a bit less than 25. That’s way to small to be worth recounting.
There is another mystery about this count though. This relates to the apparent materialization of 127,000 formal ATLs between the post-election period and the Tuesday morning count at Riverwood. I can think of a number of reasons for this, mainly associated with errors made at the booths, but have not seen it explained. As I remarked on Monday, Hanson benefited greatly from this, so it would not be in her interest to question it.
Pedant, yes, the last three NSW Legislative Council elections. It is how the system works because of the high rate of exhausted votes, plus the fact NSW elects 21 MLCs at once.
NSW is truly using the wrong electoral system for the LC. If you are going to continue using a state electorate for 21 members, they should use a divisor based system rather than a quota, which allocates seats based on the best fit of average votes per MP per party, rather than the current system which has basically become list PR with a largest remainder algorithm to determine the final seats.
You can maintain candidate choice within party by an open list system, maybe even preferences within the party list, but you ditch between-party preferences and quota.
STV is a terrible system of PR to elect 21 members because the preference distort the proportionality. STV is fine for small electorates of up to 7 members, but lousy for 21 members in an electorate of 4.5 million.
Any change requires a referendum, so expect debate.
Hanson can only request a re-count to check the vailidity of first preferences. Random sampling is specified as the method of distributing surplus to quota preferences, and as long as the sample can be verified as randomly chosen, it cannot be challenged on the grounds she states. If the court were to back her that random sampling is wrong, then the court would be exceeding its powers, as only legislation and a referendum can change that provision of the NSW Constitution.
I’m on Late Night Live with Phillip Adams at 10pm talking about the UK referendum.
Hear Hear, Antony.
Whatever one thinks of Hanson, there’s loads of ways the NSW system isn’t right.
I personally think it is outrageous that to have a NAMED box above the line you need to have a “party” of 750 members a year out before the election. Yes, you can spend a few grand and get a box above the line but the fact is, most voters don’t understand the nuances of the voting system, or they get it confused with the Senate one, and if you don’t see the name above the line, then that obviously hurts a candidate as they are that much less visible
The system of state-registered political parties seems needlessly draconian to me.
Okay, I withdraw the “stupid” bit – that was perhaps a little unfair. She lost purely and simply because she couldn’t persuade enough electors to cast a valid vote in her favor (actually, I like this interpretation just as much – it restores my faith in the validity of our democratic processes).
She is certainly entitled to complain about the voting system. Like all systems it could probably do with improvements. But would it have changed the result by enough to get her elected? When a simpler system would also have advantaged other candidates? How many squares did voters have to fill out to make their vote count – was it 15? Hardly rocket science!
I realize that I’ll probably get lambasted for this as well, but I’m not actually averse to a system that slightly disadvantages “narrow” or “single issue” candidates.
Yes, but it also disadvantages credible independents – perhaps John Hatton or similar?
edward o @512,
Exactly – I wasn’t attempting to single out Pauline Hanson. If you are an independent with a very narrow range of policies, those policies had better be darn well good enough to appeal to a significant number of voters – or you don’t deserve to be occupying an elected position where you will be expected to vote on everything.
I personally think it is outrageous that to have a NAMED box above the line you need to have a “party” of 750 members a year out before the election.
Yes, but then we get Glen Druery with his 24 incestuous parties on a tablecloth ballot
Geoff Lambert @ 507 – just out of interest, where did you manage to get a copy of a Senate Scrutiny Procedures Handbook from 1984?
The properties of random sampling in the Senate were thoroughly analysed in the late 70s and early 80s in a series of monographs by Alistair Fischer of the Economics Department at the University of Adelaide (Pru Goward’s ex), and also in an article he wrote, I think in the Australian Journal of Statistics, the citation for which I haven’t been able to find on google. The formulae worked out quite a bit more complex than a nice simple hypergeometric distribution.
Re the 127,000 votes you mentioned: I could hazard a guess that they might have been ballot papers marked both above and below the line. They might have originally been treated as below the line votes, but if the at the computerised scrutiny it became apparent that they were informal below the line but formal above the line, the above the line vote would count, under s.129F(2A) of the Parliamentary Electorates and Elections Act. But there could also be some other explanation.
Geoff: agree that that was a problem. But the solution shouldn’t punish everyone.
An independent in the lower house does not have significantly less VISIBILITY than a party member if they have a name. In the upper house, if you plan to vote above the line, the candidate you want has a lower visibility because they are an independent.
Though tbh the worst bit is having to register more than a year before the election, practically.
Actually Hanson was defeated by the donkey vote to the LNP getting first spot – worth about 0.5% and clearly worth thousands of votes. And why the ABC still thinks the Greens defeated hanson for the last spot is beyond comprehension.
vignette from corlette
the brethen own a few cottages up that way
say no more, gvnr
hawks nest is about 10mins away from corlette
Senate Scrutiny Handbook? They gave it to me or sold it to me for a pittance in 1984. The people in the AEC are really chuffed if you take an interest in what they do.
Antony correctly pointed out that electoral weirdness makes psephologists salivate.
While salivating last night, I realised that my estimates of how much Hanson could benefit from another random sampling were far too optimistic. Indeed they are very close to zero because it turns out that random sampling is not the lottery I thought it was.
First, let it be said that Pauline appears to me to be labouring under the delusion that HER votes were randomly sampled. They weren’t. Her votes never left her.
Second, random sampling only occurs when excess votes are distributed after a candidate is elected with a quota or more. There were 17 these 10 LNP, 5 ALP, 2 GRN.
In these 17 batches of excess votes to be sampled there will be 3 types of papers SATLs, RATLs and RBTLs. The SATLs are all identical to one another, so random sampling cannot affect the pattern of papers sampled, furthermore they cannot, by definition, stray to Hanson anyway. The RATLs stay with a party until that party is cut-up. The only party which had undergone random sampling and was subsequently cut up was the ALP. Hanson got 158 of these.
This leaves the RBTLs still to be considered. The RBTLs in play here as far as Hanson are concerned, are those that were BTLs for the 17 candidates elected on primaries and which somewhere during those 17 counts left the party and went to Hanson. These would, I expect, go directly to Hanson herself, though a very small number might go via other parties or via other Hanson candidates. [There are people who vote across the top (or across the bottom) of the Below the Line columns, just to be perverse-a few hundred votes in each party I estimate. There seem to be 30 LNP1 (Gallacher) voters who did something like this.]
We have the numbers involved- the SEC gives the numbers of votes that have BTL pattern with a later preference for Hanson and the number of these that actually ended up in the transfer to her. In the case of LNP 1, where there were 12,581 BTLs for Gallacher, 304 expressed a second preference for Hanson and 275 of these constituted the random sample which she received. Taylor (Hanson5) got 1 out of 1. Now, 275 represents [304 * Transfer Value (Google on it) of .904339] .This tells us one of two things:
(1) that the random sampling was perfect (unlikely) or that;
(2) the SEC is randomly sampling each “Parcel” of votes separately rather than the entire 1.9 million papers involved as a block.
A Parcel is a collection of ballot papers bearing the same pattern of votes- in this case Gallacher 1 Hanson 2. Now if that IS true- and I think it is (it happens with Roozendaal too)- then this puts the random sampling of RBTLs into the same category as random sampling of SATLs- impossible to go wrong except by 1 vote (which is why Taylor got 1 instead of 0.903 votes).
If it be true, I think this eliminates any unfairness due to random sampling completely. This sort of random sampling thus becomes just as fair as the Hare Clarke method of reduced value papers. Note that papers that DON’T get transferred at a surplus distribution go NOWHERE… they remain with the candidate elected. This differs from Hare Clarke- see final comment.
In the highly unlikely event that it NOT be true, we would still need to work out how many potential Hanson preferences didn’t reach her because the random sampling threw up aberrant figures. For Gallacher, it would be 304-275 = 29. I haven’t done the numbers on all of them yet, but it will pan out to be quite small. From Roozendaal her potential gain through cherry-picking is 8. These are not going to total up to much, I doubt more than 200 votes from the 17 elected candidates, so even cherry-picking won’t help her.
One potential unfairness remains in the LC method and this is the problem of randomly sampling only the excess votes at a surplus distribution. This only matters well down the counting process where candidates exceed a quota on some other candidate’s cut-up. This didn’t happen in 2011.
Antony, I agree strongly that open-list PR with a divisor-based system would make more sense than STV. I can’t think of any good argument for STV actually being superior to open-list PR in any situation–at most there might be some situations where STV is no worse than open-list PR.
The envelope please….
Hanson picked up 455 votes on the distribution of surpluses which is EXACTLY what one would predict. A re-run of the cut-up could not alter this.
If Hanson preferences were deliberately cherry-picked, she would have gained another 77 votes only. She needs a gain of 1306.
The case is closed.
RE.. Antony Greens post…….. what would be the result using the divisor system as you suggested? can you explain in very simple terms how it works?
why not elect a odd number in the upper house & elect in one bloc of say 45?
The results under a divisor system depend on which divisor system you use.
You can find some details of different divisor systems here:
If you elect the upper house in one bloc of 45 you’ll have all 45 members from one party. What’s the point of that?
my intention was to elected 45 via pr……. quota 1/46 ie about 2%
Divisors work on the basis of allocating seats so that elected members for each party on average represent the same number of voters. There is no quota and no preferences. Systems can be closed list, giving voters no choice of candidate within party, or open list, where voters can choose candidates from within a party list. None of these system allow candidate choice across different parties.
Divisor systems are iterative. The simplest is D’Hondt where you list all party totals and divide each total by one more than the number of members already elected for the party, and declare the highest value from this as receiving the next elected member. At the first round the divisors are (0+1), so the first seat goes to the highest polling party. In the next round, the highest polling party is divided by (1+1), and you proceed on in this fashion until all 21 LC members are elected.
In these days of spreadsheets, you just list the party totals down the left, put the divisors 1,2,3, etc across the top, create a table dividing votes by divisors, and allocating the 21 seats to the 21 biggest numbers in the table.
If you apply the D’Hondt method to the LC, it would have elected 12 Coalition members, 6 Labor and 3 Greens. Each Coalition MLC would have on average represented 161,937 votes, each Labor MLC 161,207, and each Green MP 151,041.
Note the parties that missed out had fewer votes than these numbers, Shooters and Fishers 150,741 and Christian Democrats 127,233.
As the above example shows, D’Hondt is viewed as being better for larger parties. As a result, many countries use the Sainte-Lague method, which divides by twice the number of allocated seats plus 1, giving divisors 1,3,5,7,9 etc.
By Sainte-Lague, the LC result would have been 11 Coalition, 5 Labor, 2 Green, 1 Shooter, 1 Christian Democrat plus Hanson. The average voters per seat were Coalition 176,658, ALP 193,448, GRN 226,562, Shooters 150,742, CDP 127,233 and Hanson 98043.
There is a modified version of Sainte-Lague which makes the first divisor 1.4 rather than 1, making it slightly harder for parties that would only elect a single MP. Under this modified method, The Greens would have gained the seat Sainte-Lague delivered to Hanson, reducing the Green average to 151,041.
New Zealand uses Sainte-Lague to allocate seats, but has a threshold requiring parties to reach 5% or win a single member electorate to be admitted to the divisor table. A threshold is justified under divisor systems as there is no minimum quota for election.
@ 526 thank you…. if say 45 seas were up for election would the totals be just doubled?
No. They would be even more proportional.
NSWEC Now has the statewide 2PP figure.
16.5% Swing to LNP
Has NSWEC posted actual two-candidate preference distributions for seats like Ballina and Lane Cove where the Greens came second? I can’t find them if they have.