Thursday, November 28, 2013

2013 Federal Election Results Finalised ... Finally!

The AEC has finally released the 2PP results for the various non-classic divisions at the 2013 election and hence the final 2PP result nationally.  Results can be seen here.

The final 2PP result is 53.49% to the Coalition and 46.51% to the ALP.  This is lower than projections based on expected results in the non-classic divisions (something like 53.6 to 53.7 was considered most likely).  It's probable that a reason for the difference from what was expected was that more National Party votes were distributed in three-corner contests than in 2010, especially because of the loss of O'Connor to the Liberals.  Nationals votes leak more to Labor ahead of the Liberals (24.6%) than Liberal votes leak to Labor ahead of the Nationals (9.2%).  In the case of O'Connor, this produces a swing "to" Labor in the 2PP result, caused solely by the Liberals rather than the Nationals winning the seat. 



The obligatory crowing: my humble aggregate on the sidebar (which, by the way, is unofficially nicknamed Aggedor) did fine despite its conspicuous lack of fancy mathematics. Its final pre-election reading was 53.5% (the same as the more sophisticated Bludgertrack, although at stages in the final weeks the two had quite different readings).  However this was an arguably lucky hit since the aggregate used the 2010-election preferences for all pollsters.  It turned out that, on average, the pollsters included were underestimating the final primary vote difference between Labor and the Coalition (perhaps because it was still widening between most of their final polls and election day) but that in 2PP terms this was cancelled out by a stronger preference flow to Labor than in 2010.  Given that a clear trend in the final week was a drift of votes from the ALP to PUP, it's possible that there was actually not much change in the 2PP going on in the final days and what we were really seeing was just a shift in primary votes that did not really affect the headline picture. 

The ALP actually won the state 2PP in Tasmania and Victoria.  In Tasmania, Labor scored over 51% of the 2PP but were rewarded with only one seat, because they won the 2PP in only two of the five seats and one of those was Denison, which is occupied by Andrew Wilkie.  This perverse outcome may seem surprising but shouldn't be; this was repeatedly flagged as on the cards in pre-election projections if the swing to the Coalition fell in a certain range. 

The 2PP preference flow by party shows that Green preferences went 83% to the ALP (up 4%), and that PUP and KAP preferences marginally favoured the Coalition (53.67% and 54% respectively).  The preference flow from generic "others" (excluding Greens and PUP) was 53%.  The preference flow from "others" including PUP was 53.3%.

The net 2PP swing was 3.69%.   According to the Mackerras Pendulum this would have been expected to claim thirteen ALP seats assuming a uniform swing.  In fact Labor lost a total of seventeen (including Dobell), four more than expected.  So the big argument about Labor facing a disproportionately bad fate in the marginals - especially NSW -  compared to the national 2PP had a bit of juice in it, but not that much.  Seats lost that were on margins above 3.69% were Barton (NSW, incumbent retiring), Dobell (NSW, incumbent expelled), Page and Eden-Monaro (NSW), Bass, Braddon and Lyons (all Tas), and Hindmarsh (SA). Seats saved that were below the waterline were Moreton and Lilley (Qld), Brand (WA) and Greenway (NSW, hopeless opponent).

The release of this information now makes a wide range of analysis possible and the following will all be coming to this site over the next week or so depending on other commitments (and I do have a few of them at present).  As they are done, this post will be edited.

* Detail on the preference splits between Labor and the Coalition of voters who voted for third party candidates who made the final two has been appended below.

* Assessment of the performance of all pollsters in the leadup to the election has been published already: 2013 Election: Best And Worst Pollsters

* New formal aggregate model taking the results of the above into consideration.

* I've updated the 2PPs for all polls since the election and posted results at the bottom of Early Abbott Era Polling Roundup, together with a revised aggregate tracking graph and a few comments.  This may be adjusted further once the new aggregate model is up.

My thanks to Adrian Beaumont for keeping an eye on the AEC's progress (or lack thereof) towards releasing these figures and also for saving me time by noticing some of the above details.

Where did crossbencher votes come from?

I've added this one here because it's useful information but a little bit "niche" to have its own article.  During the life of the last parliament there were frequent debates about the source of support for the various crossbenchers - Rob Oakeshott and Tony Windsor especially - along the lines that they got their votes from so-and-so therefore they should support so-and-so in the parliament.  The answer to this of course is that if the voters for Oakeshott and Windsor had so badly wanted a local representative who would be sure to vote Coalition, they could have just voted Coalition.  If you vote for an indie, that's what you get.

What I found odd about these debates was that people did not realise how readily available exact useful information on these preference flows is, and that you don't need to muck about with primary vote swings, Senate votes or booth analysis to infer it.  When there is a "non-classic" contest in a seat, so that the last two parties standing aren't just ALP vs Coalition, the AEC publishes a two-party preferred flow for that seat as if the ALP and the Coalition had been the last two standing.   From this, and from the distribution of preferences, it is possible to find the 2PP split of those voters who put the non-classic candidate ahead of both the ALP and the Coalition.

Here is a worked example using Tony Windsor (New England) from the 2010 election:

Votes with three candidates left (from this screen): Windsor (Ind) 58480 Coates (Nat) 23705 Smith (Country Labor) 8987.  

2PP vote (from this screen) Coalition 60907 Labor 30265.

The Coalition's 2PP includes all the voters who put it ahead of both Windsor and Smith, and also those who put Windsor ahead of it but behind Smith.  We know there were 23705 voters who put the Coalition ahead of both Windsor and Smith, so that leaves 37202 who ranked these candidates Windsor-Coates-Smith.

The remaining Windsor voters at the three-way cutoff point must have ranked Windsor-Smith-Coates.  There's 21278 of them.

So those who preferred Windsor to either Smith or Coates split 63.6% to the Coalition. 

It's a slightly misleading indicator because often in these sorts of seats the major party that is not in the contest runs dead.  They get a token vote that wins them as much electoral funding for as little cost as possible.  The Senate votes in New England in 2010 imply that the Windsor voters were about 7% more "left"-leaning in the Senate than in the Reps, but they were still a bit right-leaning on balance (see the discussions with poster Glen in the Tally Room thread here if forensically interested!)  But anyway, Windsor voters were no lefties.  And most of the votes Windsor had at the three-way point were his own, with a large share of the remainder being Green, so we can say for sure that most Windsor voters preferenced the Coalition in 2010.

Now here's the same thing applied for 2013:

In Fairfax (Qld), voters who preferred Clive Palmer (PUP) over the LNP or Labor split 56.4:43.6 to the LNP over Labor.  (A probable factor here is that the seat was vacant and Palmer scooped the LNP sitting member's name-recognition vote.)  84.1% of these votes were 1 Palmer.

In Indi (Vic), voters who preferred Cathy McGowan (Ind) over Liberal Sophie Mirabella or Labor split 69.4:30.6 to Labor over the LNP.  82.4% of these voters were 1 McGowan.  This showed that most of the Labor vote in Indi switched to McGowan once it was realised that McGowan could win the seat.  The Labor primary was down by 16.54%, and the Liberal primary down by 7.17%, but the 2PP swing between the major parties was virtually nothing.  So McGowan took both the Labor vote and a share of the Liberal vote - some voters who preferred Mirabella to Labor or the Greens maintained that preference but still preferred McGowan to any of them. 

In Denison (Tas), voters who preferred Andrew Wilkie (Ind) over Labor's Jane Austin split 64.1:35.9 to Labor over the LNP.  84.9% of these votes were 1 Wilkie. Here we can find the specific split from 1 Wilkie voters since he was the only Independent in Tasmania; it was 61.7%, meaning that voters who preferenced Wilkie above the majors after voting for someone else (mostly Green voters) split 77.6% to Labor.  In this seat there was an 11% primary vote swing from each of Labor and the Greens and a 7% 2PP swing to the Coalition, but almost no primary swing to the Liberals, so the Liberals only gained enough votes from Labor and the Greens to counterbalance their losses to Wilkie.  The pro-Labor split of the Wilkie vote was in fact down, as it was 69.4% in the 2010 election.  To what extent this was caused by 1 Liberal voters shifting to Wilkie, and to what extent by Wilkie-Labor voters becoming Wilkie-Liberal voters, is something that would take detailed booth analysis to possibly unravel.

What would be really nice here is to know the preference split from Labor between Wilkie and Liberal.  I've had one scrutineering report that Labor voters mostly bucked their side's deeply silly recommendation to preference Wilkie second-last (below the Libs and ahead of only Rise Up Australia) but there is no way to extract it from the data.  The Labor votes were not distributed in any official AEC process.

In Kennedy (Qld), Bob Katter (KAP) got an unexpected scare from the Nationals. Voters who preferred Katter over the majors (80.8% of whom voted 1 Katter) split 63.5:36.5 to the LNP. That's down from 69.9% in 2010.  With this electorate seeing a 5% 2PP swing from Labor to the Coalition, and a primary swing nearly that large, this mainly reflects Coalition-leaning Katter voters being more likely to have switched to the Coalition than Labor-leaning Katter voters (who had no real reason to do so.)  Also, of the massive 17.3-point in Katter's primary, 5.4 points of that went directly to the PUP candidate only to return immediately as preferences.

In Melbourne (Vic),  Adam Bandt's fan club aren't noted for their love of Liberals.  Voters preferring Bandt to either major (92.7% of them voting 1 Bandt) split 88.4:11.6 Labor over Liberal (even higher than the national Greens split of 83:17). Situation normal; in 2010 this split was 88:12 and in 2007 (when Bandt was second to Lindsay Tanner) it was 86.9:13.1.

Incidentally I warned that respondent-allocated assumptions about Liberal preferences in Melbourne couldn't be trusted. Liberal preferences split 63.44% to Labor over the Greens, much as they had in nearby seats in the state election; Bandt still won because his primary vote was so high and there were so many minor candidates to get preferences from.

The remaining non-classics for completeness:

In New England (NSW) voters preferring Rob Taber (Ind) over the majors preferred the ALP over Barnaby Joyce (Nat) 56:44 - a very different pattern to Windsor, suggesting that conservative-leaning Windsor supporters switched to Joyce.

In Batman (Vic) voters preferring the Greens to the majors preferred David Feeney (yes!) over the Liberals 90.4:9.6.  (An amazing breakdown! Up even on the 87.5:12.5 split to Martin Ferguson in 2010, so Greens voters weren't bothered by Feeney's role in the rise of Julia Gillard at all, and even rewarded him for it.) 

In Wills (Vic) voters preferring the Greens to the majors preferred the ALP over the Coalition 84.9:15.1.

The remainder are cases where the two Coalition partners provided the final contest.  In Mallee (Vic) those preferring the Liberals to the Nationals or Labor preferred the Nationals over Labor 89.2:10.8. (In this case we have the specific information that the split from those voting 1 Liberal was 90.79% since this was the sole electorate in Victoria where Liberal preferences were thrown between Labor and the Coalition.) In O'Connor (WA) those preferring the Nationals to Liberal or Labor preferred Liberal over Labor 80.7:19.3.  In Durack (WA) it was the same setup as O'Connor, and the Nats supporters preferred Liberal over Labor by only 72.8:27.2.

And since someone's otherwise bound to ask, in 2010 those preferring Rob Oakeshott (Ind, Lyne) over the majors split 54.6:45.4.  By the time the tanking effect on the Labor vote is considered, it's likely that Oakeshott's supporters were pretty even-handed or even very slightly left-leaning, but you'd not know it from the volume of noise over three years about him being a traitor to his support base.

See also on the subject of 2013 final preferences:
Antony Green
Pollbludger
Mumble

More Preferences Stuff

I've checked the greenie-Feeney flow above (sorry about that one!) to see if it was the strongest in the nation.  It was, at the least, close. Here's the top ten of known two-candidate-preferred (or two-party preferred in the cases of Mallee and Batman) flows from a given candidate's #1 votes:

1. 91.12% Hall Greenland (Green) to Anthony Albanese (ALP), Grayndler, NSW
2. 90.79% Chris Crewther (Lib) to Andrew Broad (Nat), Mallee, Vic
3. 90.45% Dianne Hiles (Green) to Tanya Plibersek (ALP), Sydney, NSW
4. 90.37% Alexandra Kaur Bhathal (Green) to David Feeney (ALP), Batman, Vic *
5. 90.26% Lachlan Slade (Green) to Lisa Chester (ALP), Bendigo, Vic
6. 90.26% Rod Swift (Green) to Tim Watts (ALP), Gellibrand, Vic
7. 90.1% Danielle Wheeler (Green) to Susan Templeman (ALP), Macquarie, NSW
8. 89.66% Adam Verwey (Green) to Andrew Leigh (ALP), Fraser,ACT
9. 88.74% Robyn Walsh (ALP) to Cathy McGowan (Ind), Indi, Vic
10. 88.63% Brendan Powell (Green) to Mike Symon (ALP), Deakin, Vic

The Batman figure differs from the others in that this is the flow of all votes that placed Bhathal ahead of one of the majors, not just Bhathal's primaries.  I haven't been able to find a way to extract Bhathal's own preferences but they were very likely even more strongly pro-ALP than even Greenland's - given that nearly 11% of Bhathal's vote at the cut-off point came from other sources including PUP, FFP, RUA and so on.

The weakest ALP performer on Green preferences in a classic division was Joel Fitzgibbon (Hunter) who managed only 57.97%; next weakest were Brendan Bryon (63.06% in the why-bother seat for Labor of Parkes) and Chis Hayes 65.83% in Fowler.   The only seat where the Greens' preferences didn't split to Labor in the final two-candidate contest (assuming Labor were in that contest) was Denison where, despite the Greens preferencing Labor, their voters bucked the card and opted for Wilkie, 57.2:42.8.  Where the Greens had an opportunity to put a National, PUP or KAP ahead of the Liberals, they also did so strongly in every case.

In all, 44 of the top 50 2CP preference flows recorded nationwide were Greens to Labor.  The exceptions were the ALP to Cathy McGowan and Bob Katter (!), the Socialist Alliance to Kevin Rudd and Kate Ellis, the Australian Christians to Don Randall (Lib), and the Citizens Electoral Council to the CLP's Tina Macfarlane in Lingiari (NT).  The last is clearly explained by Lingiari having a very high donkey vote, since it was the highest CEC vote in the nation, CEC preferences elsewhere split very evenly, and the CEC vote in the Senate there was nothing like in the Reps. (Something very similar happened in 2010.)

If anyone has any questions about preference trivia I'm happy to have a crack at them and see if they can be answered! (Many can't.)

PUP Preferences Did Not Decide The Election (2 Dec)

It's a little bit mean to refute a maiden speech but Clive Palmer's claim that PUP preferences decided the election is not correct. Palmer has been quoted as saying "at the last election the Palmer United Party won 5.5 per cent of the vote, becoming Australia’s fourth largest party. The Prime Minister became Prime Minister because the Coalition received Palmer United preferences." Now he's said it again here: "Tony Abbott became prime minister because the Coalition received Palmer United party preferences."

It's just not true.

As it happened, Palmer United preferences were very evenly split between the major parties.  The split of all PUP votes, as noted above, was 53.67% to the Coalition.   However, that includes Clive Palmer's own votes, which weren't distributed. 

In all, 364,349 PUP preferences were distributed to the Coalition in seats where the Coalition made the last two, but in three electorates (11,462 votes) they had nowhere else to go because the Liberals and Nationals came first and second.  If the remaining 352,887 votes had all gone to non-Coalition candidates, the Coalition would still have won the 2PP vote with a total of 50.76%.  And only seven seats would have changed: Labor would have retained Barton, Dobell, Eden-Monaro, Lyons, Capricornia and Petrie and the Coalition would have lost Solomon.  So the Coalition would still have won 83 seats and government without a single PUP preference. 

Of those seats where the Coalition would not have won without some PUP preferences, two of them were seats where the Coalition ended up getting less than half of the PUP preferences.  The highest share of PUP preferences a winning Coalition candidate needed to win was 44.7% in Petrie (they got 55.07% there).  In Eden-Monaro the target was 30.7% and in the other five electorates the target was less than 15%.

There were only three seats in which Labor needed any PUP preferences to win and won - Lilley, Parramatta and McEwen. Of these only McEwen raised any kind of serious question (Labor needed 37.7% and got 42.2%). 

All up the impact of PUP preferences on the House of Representatives contest between the major parties was minimal.

No comments:

Post a Comment