The Poll Bludger

Projected ALP swing	Projected 2PP	ALP win probability

Tuesday evening

Rechecking was again to the Liberals’ disadvantage today, costing them 189 votes compared with 105 for Labor. That more than cancelled out their gain from the 263-217 split on the rapidly diminishing postal vote count. The blind and low vision facility also reported its five votes today, which was technically the last outstanding polling booth. That leaves Labor’s lead up from 775 to 814; my model and projection continue to assume there are around 2000 postals still to come together with 600 to 700 declaration pre-polls and provisional votes, but the rate of decline in the numbers of postals being added with each day of counting suggests the former figure is inflated.

Monday evening

Today’s batch of 1157 postal votes behaved similarly to the previously counted 10,364 in breaking 608-549 in favour of the Liberals, amounting to in an inadequate 59-vote dent in the Labor lead. The Liberals also got the worst of today’s polling booth and pre-poll re-checking, which cost them 194 votes and Labor 97, thereby extending Labor’s lead by 97 votes. Taken together, Labor’s lead increased from 737 to 775. Rosie Lewis of The Australian reports Liberal sources saying Shooters preferences split around 50-50 and the Nationals around 80-20. The latter was not so radically different from the 87-13 split at the 2019 election, but with so close a result it was enough to be decisive. If you’re a Crikey subscriber, you can read my post-match account here.

Sunday evening

Developments earlier today required Labor to put the champagne on ice, and not just because of curious determination from parts of the Murdoch empire to call the result for the Liberals. The AEC’s efforts were spent on rechecking and formally processing the available postal votes, the latter of which was the subject of an informal advisory last night that Labor had 2394 and Liberal 2464 on the progress two-party count. This was a surprisingly strong result for Labor, notwithstanding the qualifications in the section immediately below, and appeared to put the issue beyond doubt. But for whatever reason, the result was not borne out with today’s progressive addition of postals in four batches, totalling 9207 votes altogether, with Labor receiving 46.3% rather than the initially reported 49.3%.

This is still 3.1% better than Labor did on postals last year, but the uncovering of a further error on rechecking means it’s not quite enough to definitively slam the door shut. This related to the two-party result from the Merimbula pre-poll booth, the correction of which cut a further 310 votes from the Labor margin. So whereas Labor had a 1718 vote lead on the raw count when the Narooma pre-poll finally reported its two-party result early this morning, it’s now down to 737, altbeit with a lot fewer votes left to come. The projections featured in the bottom rows of my full results display are now working on the basis that 3651 postals will be added in diminishing increments over the next fortnight, of which Fiona Kotvojs will need around 60%. There are also around 700 provisionals and declaration pre-polls to come, although they are more likely to give Labor a slight boost.

It is presumably on the basis of such calculations that Kristy McBain claimed victory this afternoon, with Labor sources quoted saying they were now “100% confident”. My own model only goes so far as 88.0%, down from its 96.5% peak late last night. All will be revealed over the coming days, and the Poll Bludger results page will be there to reveal it, providing near-instantaneous updates of the results and projections as the latest data comes through from the AEC.

Sunday morning

UPDATE AT 8:30am: The Narooma pre-poll two-party result is in, and proved less helpful to the Liberals than suggested below, breaking 1697-1632 their way.

The count ended last night with Labor leading by 1685 on the AEC’s official count, which did not include a two-party tally of postal votes it published showing 2464 going to Liberal and 2394 to Labor, reducing the lead to 1563. Only primary votes have been published from the Narooma pre-poll voting centre, which should reduce the Labor lead by another 200 to 250 votes when it reports its two-party result. Beyond that there should be another 8000 postals along with 700 votes’ worth of scraps, the latter of which should if anything favour Labor a little. That leaves the Liberals needing around 60% of the outstanding postals to win, an unforeseaable reversal on the progress of the postal count so far.

My results display has now moved into late counting phase, during which it attempts to project results for the outstanding votes, notwithstanding that the actual number of the latter is not precisely known. This shows up as three new rows at the bottom of the booth results table showing projections for provisionals, declaration pre-polls and postals. The first two are based on swing differentials from the election, while the latter assumes the outstanding postal votes will behave the same as those already counted, as they more or less did in Eden-Monaro through the 2019 election count. All of this adds up to a projected Labor margin of 1566, although as noted it doesn’t include the Narooma pre-poll which I expect will shave it back to around 1300.

Note that the aforementioned two-party postal count has been added to these results, but it won’t show up in the AEC results until the morning. This is a little messy on my end as it means there are more than two-party than primary votes in the system, which is not how things are supposed to work. Hopefully the AEC’s update will include the primary votes, and my system will be able to handle the transition back to normality when it does (since I won’t be awake for it).

Some further observations on the result:

• I said ad nauseam through the evening that projections might be thrown out by a surprise on postals, surprise in this case meaning a swing substantially different from the ordinary votes. In retrospect though, it stands to reason that a COVID-19 driven increase in postal voting should make the postal voter base more representative, and thus less conservative. AEC figures also show that Labor’s postal vote campaign also elicited 4447 applications compared with the Liberals’ 2908. In any case, the Liberals’ 50.8% two-party share of the postals counted thus far compares with 57.1% at the federal election, which compares with a picture of near-perfect stability on the two-party vote in polling booths (a 0.1% swing to Liberal) and pre-polls (what I’m projecting to be a 0.2% Liberal swing when the one straggler booth reports its two-party result).

• The outstanding fact of the result is that Labor achieved it with a stronger flow of preferences than last time, from roughly 50-50 overall to 57-43 (inclusive of preferences from the Nationals as well as minor parties and independents). This was despite the minor party preference pool being seemingly more conservative than last time: Shooters polled 5.4% after not contesting last time without costing the Nationals, while on the left, the Greens’ drop was cancelled out by gains for HEMP and the Science Party. It will be very interesting to see how much of Labor’s surprisingly strong preference flow was down to the Shooters, who directed preferences to Labor, and John Barilaro, who was accused of encouraging Nationals voters to do the same. We will get some idea when the full preference distribution is conducted in about a week, and a more exact one when the preference flow figures are published further down the track.

• The Liberal line on the result is that Labor’s 3.0% drop on the primary vote is disastrous for Anthony Albanese, since by-election swings to oppositions are par for the course in non-disastrous contexts. But as Peter Brent has calculated, there is actually very little historic tendency for oppositions to pick up swings in their own seats at by-elections, when they are suffering from the loss of the incumbent’s personal vote and, in most cases, an element of resentment that the member resigned. There is also the highly salient fact that there were eight candidates at the election and fourteen at the by-election, such that the Labor vote last year included a cohort who would have voted for Shooters or HEMP if given the chance. Nonetheless, the swing comes off an already low base and the result can’t be described as any better than mediocre for Labor.

Election night commentary

Continue reading “Eden-Monaro by-election live”

If you’re planning on following the action in tomorrow night’s Eden-Monaro by-election count, I flatter myself that there will be no better place to do so than this website’s live results feature, which you can see ready for service here. For an idea of how it will look when fleshed out, here’s a test I ran on the feed from the Bennelong by-election.

The top half of the display features progress vote totals and booth-matched swings, including a New York Times-style “election needle” recording a probability estimate of the final result. This is based on a somewhat novel projection of two-party preferred that uses estimates of the final primary vote shares and preference flows. The latter is the subject of its own table, which will show how minor party and independent preferences are flowing based on the latest two-party count, and how this compares with the corresponding results from the 2019 election.

The bottom half features the clearest and neatest display of the booth results you will find, in the form of a table in which you can toggle between vote totals, percentages and swings. Not the least of the benefits of this is that the results can be easily cut-and-pasted into a spreadsheet. The table also features separate sub-totals for ordinary election day polling booths and pre-poll voting centres, with swings to match — an increasingly important distinction on election night, when the latter come in quite a bit later than the former and don’t always behave the same way.

With the self-promotion taken care of, here’s the late news:

• The Australia Institute has produced its second uComms robo-poll for the campaign, conducted on Tuesday night from a sample of 643, which shows Labor with a lead of 52-48. This compares with 54-46 at the earlier poll on June 15, although the tightening is more pronounced on the primary votes. After allocating results from the forced-response follow-up for the 3.7% undecided, the poll had Labor up 0.3% to 39.3%, Liberal up 6.2% to 38.3%, the Greens down 1.5% to 7.5%, the Nationals down 1.8% to 5.2% and Shooters down 1.9% to 4.8%. Using preference flows from last year’s election, the new result rounds to 50-50, whereas the earlier poll had Labor leading 53-47. The poll also found 56.4% of respondents thought the ABC should receive more funding, 23.8% less and 19.8% about the same. Kooyong, Wentworth and Warringah were also polled, the latter finding independent member Zali Steggall with a 56-44 lead over a presumably hypothetical Liberal candidate.

• If the result indeed goes down to the wire, it will be a long night tomorrow – as well as processing large pre-poll voting centres whose results are unlikely to be reported well into the night, the Australian Electoral Commission will also be counting what it estimates will be between 5000 and 8000 postal votes, contrary to its usual practice.

• Phillip Coorey in the Financial Review reports sources from both parties say their internal polling is similar to that of the aforesaid Australia Institute poll, “but neither side was prepared to predict how it would pan out on Saturday”. A Labor source is quoted expressing concern that the Victorian outbreak could rekindle a sense of concern about coronavirus to the Liberals’ advantage (“it puts us back in the dynamic of April and May”). Elsewhere, Niki Savva in The Australian ($) writes of Labor concern that lack of awareness of the by-election could depress turnout.

• John Barilaro persists as a thorn in the side of the Liberal campaign, having contradicted the goverment’s line that it had not cut funding to the ABC, refused to rule out running for the seat at the next election and, reportedly, encouraging supporters to preference Labor. A blast from the past also emerged this week in the shape of a post-election Facebook comment in which he told Kotvojs he was glad she didn’t win.

• With scarcely an account of the race failing to note that 1920 was the last time a government gained a seat from an opposition at a by-election, Peter Brent at Inside Story notes the probably more salient fact that 14 of the 91 by-elections since 1918 have swung to the government by more than the 0.9% they will need to win in Eden-Monaro. He also calculates an average swing against the government of just 0.9% in opposition-held seats at by-elections, compared with 4.6% in government-held seats.

As the campaign enters its final week, I have given my Eden-Monaro by-election guide an overdue overhaul, such that it now offers a thorough account of the Coalition’s preselection tangles and full detail on the various reports of opinion polling that have emerged, along with the usual charts, results map and historical detail. Other news worth noting:

• The Australian ($) reports the Nationals are hawking internal polling showing a surge in their own support potentially great enough to push the Liberals over the line. Their candidate, Trevor Hicks, was said to have risen from 6% a fortnight ago to 11.5%, clipping Liberal candidate Fiona Kotvojs from 36% to 34.3% and Shooters Fishers and Farmers from 7% to 4.6%. Labor’s Kristy McBain was said to have slumped from 36% to 29.3%, with the Greens up slightly from 7% to 8.7%. The more recent poll was conducted on Thursday from a sample of 630; the sample for the earlier poll was said to be rather larger.

• A borderline comical spam email attacking Labor candidate Kristy McBain is the subject of an Australian Federal Police investigation. Purporting to be from McBain’s campaign manager, the email claimed she had withdrawn as a candidate after being hospitalised with COVID-19, before holding her largely responsible for fabricating the Catholic Church sex abuse scandal. Complaints under the “misleading or deceptive publications” provision under the Electoral Act are ten a penny, and invariably fall foul of the limitation that the falsehood must be “in relation to the casting of a vote”, (which the Australian Electoral Commission contentiously determined did not apply to Chinese language Liberal Party advertising in Melbourne at last year’s election). In this case though, the material clearly crosses the line in asserting that votes for McBain will be invalid.

• The Australian Electoral Commission has so far processed 16,098 postal vote applications, compared with a total of 7426 at the 2019 election. The thirteen pre-poll voting centres now in operation have already received more than 21,000 votes; last time a total of 35,114 votes were cast at eight locations.