Simulating the Senate

Some tentative projections of how things might look in the Senate, after electoral reform and a double dissolution have done their work.

One of the factors that was overlooked in the recent debate over Senate electoral reform was the hours of enjoyment election watchers used to have projecting Senate outcomes, with help from Antony Green’s calculators, thanks to the exquisitely predictable preferences flows that arose from the group voting ticket system. The fact that voters will now be largely making up their own minds means we are headed into unknown territory with the coming election. But that’s no reason to give up altogether, and I believe I’ve managed to come up with a reasonably stimulating projection exercise based on the following:

• Experience in the Australian Capital Territory tells us that voters overwhelmingly do what the ballot paper tells them to do, and no more than that, even if they are technically allowed to exploit savings provisions to number fewer boxes than indicated. On that basis, my model assumes every vote consists of six numbered boxes above-the-line.

• A ballot paper study from the 2010 Victorian state election tells us that how-to-vote cards are obeyed by 45% of Liberal voters, 41% of Labor voters and 25% of Greens voters, but by next to no micro-party voters as they lack the infrastructure to staff polling booths on a large scale. We don’t actually know what those how-to-vote cards will say at this stage (and I dare say controversies will emerge as some recalcitrants recommend a just-vote-one option), but the parties’ group voting tickets from 2013 are, at the very least, a useful starting point.

• For those who don’t follow the how-to-vote card, there is a very useful resource available in the form of the below-the-line voting data from the last election, which records the full preference order of every single voter who made up their own mind. I’ve used this to develop matrices indicating how many voters for each party included the various other parties in their top six. These are then mixed together with the how-to-vote card followers in the proportions indicated in the aforementioned Victorian election study (though only for Labor, Coalition, Greens, Australian Christians and Family First – for all other parties, I’m going entirely off the below-the-line data).

• The parties’ primary votes are, in the main, determined by applying the state-level swings currently recorded by BludgerTrack to the 2013 election result, with exceptions noted below. Whatever’s left over is distributed between the smaller parties in proportion to their share of the vote in 2013.

With those starting primary votes and assumptions about preference behaviour, it becomes possible to simulate an overall result. I must stress that I wouldn’t go betting the house on any of this, as a lot of the factors involved are tied to the last election result, and there is good cause to expect things will be different this time. The cast of minor parties is assumed to be the same as last time, and only in South Australia has the Nick Xenophon Team been accommodated. Nonetheless, this is a useful starting point for exploring how the Senate result might look, and will be refined over the rest of the campaign as more information becomes available. The results look a little something like this:


Now here’s the fine print, followed by a display showing key points in the projected counts:

NEW SOUTH WALES: I’ve arbitrarily clipped 6% off the Liberal Democrats and handed it over the Coalition, based on my guesstimate of how many of the party’s voters were confusing them with the Coalition when they drew first spot on the tablecloth ballot paper in 2013. That still leaves them 1.3% clear of the Christian Democrats as the fourth best polling party, which changes little through to Count 54, at which point the Australian Christians drop out and leave the last three positions to the three candidates remaining in the count: Coalition #5, Labor #5 and, taking the twelfth spot, David Leyonhjelm.

VICTORIA: The Sex Party survives to the final count, but lands around 3% short of Labor’s fifth candidate when it gets there. To cover that gap, it would need to improve on its 1.9% primary vote from 2013, and/or have Labor do a few points less well than the 35.5% projected for it. Family First makes it to the second last count, but the Coalition vote is strong enough on this accounting to ensure a monopoly on the five right-of-centre seats in Victoria.

QUEENSLAND: The projection has Coalition #5, Labor #4 and Katter’s Australian Party in a very close race for the last two seats, as indicated in Count 46, with Labor losing the game of musical chairs. However, Katter’s Australian Party ran a very high profile campaign in 2013 with a name candidate in country singer James Blundell, who was long thought a shoo-in before being elbowed aside by Clive Palmer and Glenn Lazarus. It’s hard to say if the party now stands to decline as a result of dwindling publicity, or clean up as voters desert Palmer United. Perhaps the more important lesson to be drawn from this is that the projection rates a base vote of 3.3% as a potential springboard for victory.

WESTERN AUSTRALIA: I’m dubious about the 10% swing against the Coalition in Western Australia recorded by BludgerTrack, and doubly so about Senate projections based on its results, since it reduces both major parties to the twenties. This inflates the minor and micro-party vote to cover the loss, which includes the Nationals and doesn’t include the Greens. For what it’s worth though, this results in a seat for the Liberal Democrats and also one for the Nationals, who may indeed by a show, but probably not with the level of certainty indicated by the model.

SOUTH AUSTRALIA: I haven’t made any adjustment at all to the 2013 election result, since the Nick Xenophon factor renders meaningless any comparison between the Senate result then and House swings now. The result effectively becomes a three-way tie for the past two seats between Liberal #4, NXT #4 and Family First, as indicated at count 44. The projection has Family First dropping out, but clearly it could have gone any which way.

TASMANIA: I’ve dealt with Jacqui Lambie by leaving her vote and preference structure from 2013 intact, which is possibly a little generous to her, but it will do for a start. That leaves her very close to a quota, and once she’s there, all that’s left is for the micro-parties to be weeded out until Greens #2 gains the final spot.


Newspoll: 51-49 to Labor

Newspoll tells a familiar tale, with the two-party vote unchanged over four successive polls, Malcolm Turnbull’s personal decline levelling off, and Bill Shorten continuing to rise.

The latest Newspoll for The Australian is the fourth successive result from the pollster showing Labor leading 51-49 on two-party preferred, from primary votes of Coalition 41% (steady), Labor 36% (down one) and Greens 11% (steady). Leadership ratings echo other pollsters in finding Malcolm Turnbull levelling off after a steep decline in the early months of the year, with approval steady at 38% and disapproval up one to 50%, and Bill Shorten continuing to improve, with approval up four to 37% and disapproval down three to 49%. Turnbull’s lead as preferred prime minister has narrowed from 49-27 to 46-31. Expectations of a Coalition in have diminished considerably since the question was last asked in March, with 44% now favouring a Coalition win (down 11%) and 33% favouring Labor (up eight). The poll was conducted Thursday to Sunday from a sample of 1709.

ReachTEL: 50-50; Ipsos: 51-49 to Coalition

Two new national polls do nothing to dispel perceptions of a tight race, and they’re accompanied by another result showing Labor with its nose in front in the key outer Sydney seat of Macarthur.

Two new national polls this evening, plus a local one from the electorate of Macarthur:

• The latest Ipsos result for the Fairfax papers, which we can now expect on Friday night rather than Sunday at least for the period of the campaign, has the Coalition two-party lead unchanged at 51-49, with the Coalition primary vote down a point to 43%, Labor up one to 34%, and the Greens steady on 14%. Malcolm Turnbull’s lead as preferred prime minister is down from 51-29 to 47-30, and his approval rating is steady on 48%, with disapproval down two to 38%. Bill Shorten is respectively up two to 40% and down three to 46%. The poll was conducted Tuesday to Thursday from a sample of 1497, compared with the usual Ipsos survey period of Thursday to Saturday.

• Seven News brings us a new poll from ReachTEL which is the third such poll in a row to have the result at 50-50, the most recent of which was conducted a few days after the budget on May 5. However, the primary votes look better for Labor this time, with the Coalition on 42.6%, down from 44.2%; Labor on 36.6%, up from 35.1%; the Greens on 9.9%, up from 9.5; and the Nick Xenophon Team on 2.7%, down from 4.2%. Malcolm Turnbull’s lead over Bill Shorten has narrowed from 57.7-42.3 to 55.6-44.4; Turnbull’s very good plus good rating is up from 28.1% to 28.6%, and his poor plus very poor rating is up from 34.5% to 35.1%; the corresponding results for Bill Shorten record a solid improvement, with very good plus good up from 24.6% to 27.9%, and poor plus very poor down from 44.0% to 38.4%. The automated phone poll was recorded last night from a sample of 2407 – full results can be found here.

• There is also a separate ReachTEL poll for the Macarthur electorate, which the Russell Matheson holds for the Liberals on a post-redistribution margin of 3.3%, down from 11.4% at the 2013 election. Here ReachTEL credits Labor with a 51-49 lead on two-party preferred. After distributing results from a secondary question prompting the 10.7%, the primary votes are Liberal 41.4%, Labor 41.1%, Greens 7.3% and Nick Xenophon Team 2.5%. The poll was conducted last night from a sample of 628.

UPDATE: Here’s what the BludgerTrack tables look like with the ReachTEL and Ipsos result added, including the Ipsos leadership ratings. This translates into a higher result for “others” at the expense of both major parties, with no change to the national two-party result (UPDATE: Turns out this was because I’d failed to distribute ReachTEL’s undecided results, so scratch that). The Coalition is down a seat in New South Wales and up one in Western Australia, although the remarkable swing result from the latter is only slightly modified.


UPDATE 2: Here’s another chart to brighten your weekend, this one tracking the state-level swings to or from the Coalition on Malcolm Turnbull’s watch. Broadly speaking, what emerges is a levelling off since March everywhere except Western Australia, where the momentum of the government’s early year slide has been maintained. The uptick to the Coalition in Queensland is a curiosity, and may simply reflect the dominance there at the moment of a single data point, namely the 1176-sample poll there from Galaxy last week (though it’s not nearly as dominant as last week’s 3019-sample state-level poll from ReachTEL is in Tasmania). You might also find evidence that the submarines contract was a steadier for the Liberals in South Australia if you’re looking at it, but the elephant in the room there is the 22.2% others vote, which crept up from 13.5% at the start of Turnbull’s tenure to 14.6% at the end of the year, and has since swelled to 22.2%. I’ll provide a more detailed display of state-level breakdowns soon, I promise.


BludgerTrack: 50.1-49.9 to Labor

Labor pokes its nose ahead on two-party preferred in the latest reading of the BludgerTrack poll aggregate, but a stronger showing in Queensland finds the Coalition keeping its head above water on the seat projection.

The flurry of national polling conducted after the budget for release at the onset of the official campaign has been followed this week by a lull in new results at national level, but with Galaxy and ReachTEL making sizeable entries in state-level federal polling from Queensland and Tasmania respectively. The only national results were the regularly weekly Essential Research and the first campaign poll from Roy Morgan, the latter of which was strong enough for Labor that they have moved back into the two-party lead by the barest possible margin. However, the strong showing for the Coalition in the Galaxy Queensland poll causes them to register 1.2% higher this week in that acutely sensitive state, translating into two extra seats to partly cancel out losses of one each in New South Wales, Victoria and Western Australia. Essential Research has provided new numbers for leadership ratings this week, and these seem to suggest Malcolm Turnbull’s slump is levelling off.

Some of you will no doubt be looking askance at that swing currently projected in Western Australia, and I don’t blame you. There are seven data points in the model from the past three weeks with a combined sample of 1048, which individually have the Coalition’s primary vote in the state ranging from 35% to 44%, compared with 51.2% at the 2013 election. However, I suspect that if you look back in a week or two, you will find the projection moderating somewhat. It’s also worth observing that the model is now crediting Palmer United with all of 0.1% of the national vote. The only pollsters who are still tracking the party are Ipsos and Morgan, with both ReachTEL and Essential having swapped them in their questionnaire for the Nick Xenophon Team. The last five data points for Palmer United are all 0%, and the previous ten were evenly divided between 0% and 1%.


News snippets:

The Advertiser reported yesterday that a privately conducted ReachTEL poll had produced an encouraging result for Matt Williams, Liberal member for the marginal Adelaide seat of Hindmarsh. Williams was credited with 41% of the primary vote, compared with 25% for Labor candidate Steve Georganas, whom Williams unseated in 2013, 14% for Nick Xenophon Team candidate Daniel Kirk, 8% for the Greens, and 7% undecided.

• Nick Xenophon told the ABC’s Lateline his party’s strongest lower house prospect, Mayo candidate Rebekha Sharkie, was polling in the twenties. How formidable that makes her would depend entirely on how much of it was gouged from the vote for Liberal member Jamie Briggs, who recorded 53.8% of the primary vote in 2013.

• Labor has hit trouble in a sensitive spot in the inner Melbourne seat of Batman, after it emerged that David Feeney had failed to declare a negatively geared $2.3 million property in Northcote on the register of members interests. The news media is now applying the blowtorch to other aspects of the real estate portfolio of Feeney and his wife, and bringing unwelcome attention to his once close association with controversial ex-Health Services Union identity Kathy Jackson. Feeney is under pressure in Batman from Greens candidate Alex Bhathal, who outpolled Liberal candidates in her previous runs for the seat in 2010 and 2013, respectively finishing 7.9% and 10.6% behind Labor at the final count.

• A week after Labor dumped its candidate in the seat, there have been headlines about the contentious views of Sherry Sufi, the Liberal candidate for the Western Australian seat of Fremantle. Sufi’s conservative positions on matters such as same-sex marriage and the stolen generations apology had been well known, but Malcolm Turnbull contrived to make an issue out of them when he visited the electorate on Monday to spruik a local shipyard’s contract to build naval patrol boards, and neglected to invite his candidate. There have also been questions raised about the accuracy of Sufi’s employment record as presented on his candidate nomination form. Also absent during Turnbull’s shipyard visit was Premier Colin Barnett, whose leadership is increasingly coming under pressure amid deterioriating opinion polling.

Essential Research: 51-49 to Labor

No change from Essential Research this week, which also records Malcolm Turnbull dipping into net negative territory on personal approval for the first time.

The latest result for Essential Research is largely unchanged on last week, with the Coalition steady on 42% of the primary vote, Labor steady on 38% and the Greens down one to 9%. One change is that the pollster has dumped Palmer United from its survey and replaced it with the Nick Xenophon Team, which opens it account on 3%. The poll also features Essential’s monthly leadership ratings, which find Malcolm Turnbull up one on approval to 40% and up three on disapproval to 42%, Bill Shorten up four on approval to 34% and down one on disapproval to 43%, and Turnbull’s lead as preferred prime minister narrowing from 44-22 to 43-28. There is also a suite of questions on social class, something 81% agreed existed in Australia, with only 8% saying otherwise. Only 2% of respondents identified as upper class, yet 53% thought the Liberal Party mainly served that party’s interests. Forty-eight per cent of respondents identified as middle class, which 15% thought mainly served by Liberal and 17% by Labor, while 34% identified as working class, which 39% thought mainly represented by Labor and 4% by Liberal. The poll also found 48% approval of the budget’s internships scheme for the young unemployed, and 52% rating the election campaign too long versus 5% for too short and 32% for about right.

Morgan: 52.5-47.5 to Labor

The first poll of national voting intention in nearly a week records a significant drop in the Coalition’s primary vote.

Morgan has released a poll of federal voting intention derived only from this weekend’s round of face-to-face and SMS polling, and not combining two weeks of polling as per its usual practice. Compared with the poll from the two weekends before the election was called, the poll records a solid drop in the Coalition primary vote from 40% to 36.5%, with Labor up half a point to 33%, the Greens up two to 15.5% and the Nick Xenophon team up one to 5%. On the headline respondent-allocated measure of two-party preferred, this converts into an increase in Labor’s lead from 51-49 to 52.5-47.5. Going off 2013 election preference flows, the change is from 50.5-49.5 to 52-48. Despite the shorter than usual field work period, the sample is a considerable 2318.

ReachTEL Tasmanian electorates polling

A poll of Tasmania’s electorates finds the Liberals grimly hanging on in the three seats gained from Labor in 2013, and independent Andrew Wilkie going untroubled in Denison.

Today’s Sunday Tasmanian has results from ReachTEL polling of each of the five lower house seats in Tasmania, from a combined sample of 3019. The report says the poll credits the Liberals with 51-49 leads in Bass and Lyons, independent Andrew Wilkie with an increased majority in Denison, Labor member Julie Collins with a lead of 54-46 in Franklin, and Liberal member Brett Whiteley with a primary vote lead of 42.7% to 32.6% in Braddon, suggesting little change on his 2.6% winning two-party margin in 2013. The Jacqui Lambie Network would find “solid support” in the northern electorates, particularly her home base of Braddon, but has just 2.7% support in Denison and 2.5% in Franklin (this being before exclusion of around 7.5% undecided). I will be able to go into greater depth on these results tomorrow, but will be beaten to it by Kevin Bonham, who promises to publish a comprehensive overview at 8.30am.

In other partly reported poll news, Brisbane’s Sunday Mail has a tranche of state results from that Galaxy poll that provided federal results yesterday, but none of the voting intention numbers are provided in the online report. The report does relate that Tim Nicholls’ coup against Lawrence Springborg the Friday before last had 42% approval and 27% disapproval, and that Annastacia Palaszczuk leads Nicholls as preferred premier by 44% to 29%. Much is made of the fact that this isn’t as good for Palaszczuk as the 54-26 she happened to record against Lawrence Springborg in November. There will be voting intention eventually, I promise.

UPDATE: Kevin Bonham details the full results from the ReachTEL poll. The published respondent-allocated results have the Liberals leading 51-49 in Bass (54.0-46.0 at the 2013 election), 53-47 in Braddon (52.6-47.4) and 51-49 in Lyons (51.2-48.8), with Labor ahead 54-46 in Franklin (55.1-44.9). Each of these results is better for Labor than a 2013 election allocation would have been, particularly in Franklin (where Labor’s lead would have been 52.4-47.6) and Lyons (where the Liberals would have led 54.1-45.9). In Denison, Andrew Wilkie records 33.2% of the primary vote, down from 38.1% at the election, with Labor up from 24.8% to 27.3%. However, ReachTEL has published a Wilkie-versus-Liberal two-party result rather than Wilkie-versus-Labor, of 66-34, even though it was Labor who finished second last time, and would do so again on these numbers. The Jacqui Lambie Network’s average across the five seats is 5.3%.

Galaxy: 54-46 to federal Coalition in Queensland

After a week spent splashing cash at marginal seats in South Australia and Tasmania, new polls arrive from Queensland and western Sydney to steady the Coalition’s nerves.

The Courier-Mail today brings the Coalition one of its most encouraging poll results in a while, crediting them with leads on federal voting intention in Queensland of 54-46 on two-party preferred, and 46% to 33% on the primary vote. This compares with 57.0-43.0 at the 2013 election, and primary votes of Coalition 45.7% and Labor 29.8%. The only seats a uniform swing of 3% would net for Labor would be the Rockhampton region seat of Capricornia (margin 0.8%), which Labor has only lost three times since 1961, and the northern Brisbane seat of Petrie (0.5%). The poll was conducted Tuesday and Wednesday evening from a sample of 1176.

Also from Galaxy, the Daily Telegraph has electorate-level polling showing the Liberals leading 54-46 in Lindsay and by unspecified amounts in Gilmore and Reid, with 50-50 results from Banks and Dobell and a 51-49 lead for Labor in Macarthur, the scene of last night’s leaders forum. More precise figures on that will be available at some point, hopefully soon. The polls were automated phone surveys of around 500 respondents per electorate.

I’m aware at least one other big set of regional polling that will be with us this evening, so stay tuned for that one. Other news:

• The small sample of attendees at last night’s leaders forum came down 42-29 in favour of Bill Shorten over Malcolm Turnbull.

• Family First Senator Bob Day’s constitutional challenge against Senate election reforms got short shrift from the High Court in yesterday’s judgement, which said in reference to the plaintiff’s submission: “None of the above arguments has any merit and each can be dealt with briefly.”

• The government has maintained its recently developed interest in South Australia with a visit to the state yesterday by the Prime Minister, in which he committed to funding half of an $85 million rail project connecting Flinders University to the central business district, with a scheduled completion in late 2018. This helpfully runs through the electorate of Boothby, to be vacated at the election by Liberal member Andrew Southcott.

• Labor and Tasmanian independent Andrew Wilkie have attacked the federal and Tasmanian state governments over $22 million in grants from the Tasmanian Jobs and Investment Fund that were announced this week. Most of the money had been freed up by the demise of a proposed tourism visitors centre at the Cadbury’s factory in the northern Hobart suburb of Claremont, in Wilkie’s seat of Denison, but the bulk of the new projects were in the three marginal Liberal seats in the state’s north. The Hobart Mercury reports that $6.29 million has gone to Lyons, $5.55 million to Bass and $3.59 million to Braddon, compared with $3.6 million in Denison and $2.91 million in Labor-held Franklin.

• Some anonymous public-spirited individuals have put together an outstanding interactive data visualisation site through which you can explore disclosures of political donations.