Tuesday, May 5, 2020

Eden-Monaro By-Election 2020: How Loseable Is It?

By-election July 4th
Cause of by-election: retirement of Mike Kelly (ALP)
Outlook: Loseable but Labor probably should retain

Welcome to what I intend to be my main pre-analysis page for the Eden-Monaro by-election, though I may split it into multiple pages if the campaign drags on for a long time, raises unexpected electoral issues, or generates too much detail.  I expect to have a live page on election night.

There are two main narratives about this by-election as the parties compete for the role of underdog and try to manage expectations in advance.  The first is that the loss of an opposition seat to an incumbent government in a federal by-election is literally a once in a century event (it happened for the only time in 1920) and that therefore a government win is unrealistic.  The second is that the seat's marginal nature combined with the high personal vote of outgoing ALP incumbent Mike Kelly makes the seat extremely difficult to defend in the current environment.  I argue here that both these narratives are wrong.   The by-election is much more loseable than the "100 years" history suggests, but [s]most[/s] all of the arguments as to why it could be lost are being overplayed.


The by-election timing was delayed slightly by the need to obtain advice from the AEC, which took a fortnight, pushing the date back to mid to late June.  The date subsequently landed at the start of the July to August window I was expecting, being earlier than it might have been perhaps because of encouraging progress in the containment of COVID-19.

There was no requirement to hold the by-election at any specific time; it was a matter of balancing risks against the negatives of having the seat vacant.  Those negatives were somewhat reduced for the time being by the limited nature of current parliamentary business.  See Antony Green for extensive discussion of management of the by-election.


Eden-Monaro surrounds the ACT in far south-eastern NSW.  It was long famous as Australia's champion bellwether seat, going with the party that won government at fifteen elections in a row from 1972 to 2013.  This streak was snapped when Mike Kelly recaptured the seat from one-term incumbent Peter Hendy in 2016 and the mantle has now passed to Robertson (on a streak of 14 elections starting from 1983).

Eden-Monaro has never been won by the National Party or its precursor the Country Party.  The Country Party came close in 1972 (lost by 503 votes with 49.5% 2PP) and even closer in 1974 (146 votes, 49.87% 2PP).  In both these cases the flow of preferences from the Liberal Party was very strong (95% and 97.4% respectively) but the fact that not absolutely every Liberal voter preferenced the Nationals saved Labor (this also happened with National to Liberal preferences in 2019, though the Liberals would have needed all bar a few dozen to win.)

The main overlapping state seats are Bega (held by the Liberals' Andrew Constance) and Monaro (held by the Nationals state leader John Barilaro).  Bega has been held by the Liberal Party solidly though at times fairly narrowly since its re-creation in 1988, while Monaro has been held by the Nationals throughout this time except for two Labor wins in 2003 and 2007.  There was a large swing to the Nationals in previously marginal Monaro at the 2019 NSW election, clearly off the back of Barilaro's profile as leader and against the backdrop of a good election for his party in the near-coastal regionals.

Opposition By-Elections Generally

In recent years governments have normally not contested Opposition seat by-elections.  The Howard government from 1996 to 2007 never contested any.  The Rudd government took a swing at Gippsland in 2008, probably hoping to translate Kevin Rudd's popularity into at least some embarrassment for the Opposition, but it backfired with a 6.1% swing to the Nationals' Darren Chester.  The Abbott government took a more realistic shot at Rudd's own seat of Griffith in 2014 and picked up a 1.25% swing off the loss of Rudd's personal vote.  (Such a swing would win Eden-Monaro, but Rudd was an ex-Prime Minister, after all.) The then Turnbull government contested two of Labor's four vacancies in the 2018 Super Saturday by-elections, but ALP incumbents Justine Keay (Braddon) and Susan Lamb (Longman) won with 2PP swings to them of 0.1% and 3.7% respectively.  Labor was bolstered by unimpressive government candidates (especially in Longman) and a government campaign blunder (picking a fight with a local independent) in Braddon.  Both ALP incumbents went on to lose in the 2019 general election.

Prior to these four cases, the last time a government contested a by-election was Groom 1988.  Labor may have had some hopes of exploiting Coalition disunity following the demise of Joh Bjelke-Petersen but ended up running third in a three-cornered contest (and would probably have had a 2PP swing against them anyway).

The long-term average two-party swing in Opposition by-elections that are contested by the government of the day has been 1.2% to the Opposition.  This is less than the average swing for all by-elections contested by both Government and Opposition, which has run at around 4% since WW2. The reasons for this difference include:

* In an Opposition seat by-election, the Opposition is usually losing the personal vote of an MP who has resigned or died (Braddon and Longman in 2018 being exceptions to this).

* Governments have increasingly cherry-picked which Opposition by-elections to contest and have typically avoided running those in which embarrassing swings against them appear likely to occur.

* When Opposition MPs throw in the towel soon after losing an election, the government may benefit from honeymoon effects, desire to stabilise the parliament following a close result, or voter contempt for the sitting MP's decision to quit.  There have been cases of Oppositions losing by-elections to Governments in this situation at state level, notably the Victorian examples of Benalla 1999 and Burwood 2000.

While the one case of a Government winning an Opposition seat at a federal by-election (Kalgoorlie 1920 following the expulsion of Hugh Mahon) provides a historical strike rate of only a few percent, that history is misleading because few of the seats contested were super-marginal.  Swings exceeding 1% to the Government of the day have happened in around 30% of Opposition seat by-elections.

Mike Kelly's Personal Vote (and its implications ...)

Warning: this section is a bit number-heavy, rising in places to 3/5 on the Wonk Factor scale.

Mike Kelly, decorated Army colonel, PhD scholar and barrister, was the member for Eden-Monaro from 2007 to 2013 and again from 2016 until his retirement, making him a 10-year incumbent.  Kelly was clearly a popular MP and has been praised on all sides of politics following his decision to step down because of health issues stemming from his military service in some of the hotter places on Earth.  But can we put any numbers on just how popular he is?

Looking at graphs comparing Eden-Monaro with NSW as a whole (see Antony Green again) it's notable that the seat was running close to the state average until Kelly's first defence of the seat in 2010.  With Kelly as a candidate who had served as the seat's MP from 2010 onwards, the seat then ran 5.4, 3.8, 3.4 and 2.7 points above the state's average for Labor, in spite of redistributions following the 2007 and 2013 elections knocking 1.1% and 2.7% off Labor's position.  With these factors considered Labor ran 4.9% to 6.5% above their previous relative standing with Kelly as an established or former incumbent, but this is coming off a baseline that includes Liberal Gary Nairn's personal vote as an 11-year incumbent.

One term 2013-6 incumbent Peter Hendy did not appear to generate any personal vote, unlike the normal pattern of new MPs gaining through double sophomore effect.  As an aside, this wasn't because the Turnbull government's campaign played badly in Eden-Monaro, if anything slightly the opposite.  (The swing to the Coalition in Eden-Monaro in the Senate in 2016 was three points higher than the NSW average despite the Reps swings being the same, though about 1.4 points of this is explained by a high rate of Liberal Democrat confusion in the Senate in the seat in 2013).

Estimating personal votes for incumbent MPs, beyond that they usually exist, is very tricky.  One method is to subtract the relevant party's Senate vote from their House vote, on the grounds that the more voters vote for the party in the House but not the Senate, the higher the incumbent's personal vote probably is.  On this measure Kelly does extremely well, polling 10.52 points above Labor's Senate vote in Eden-Monaro, putting him fifth of all the Labor MPs in NSW.  But Labor did 4.74 points better in the Reps than in the Senate in NSW as a whole, so Kelly is 5.78 points above Labor's statewide average.  A 2PP edge of nearly six points sounds enormous (enough to put Labor underwater after accounting for the existing margin and the normal anti-government factor) but it also happens that the Coalition in Eden-Monaro did 5.11 points better in the Reps than the Senate in this seat.  That compares to a statewide difference of 3.99 points.  So it is not as if Kelly is taking such a large share of the 5.78 points in performance above Labor's state average from the Coalition; rather he is taking some from the Coalition and some from minor parties.

That said, the Coalition are advantaged in this comparison by Eden-Monaro being a three-cornered contest, and once three-cornered contests are adjusted for at both state and local levels, Labor's advantage over the Coalition on Reps performance relative to Senate performance lands at 6.26 points.  In other words, a 2PP swing advantage of 3.13%.

(There are reasons why both sides would, all else being equal, have done well on the Reps minus Senate indicator in Eden-Monaro.  The seat had no One Nation Reps candidate and no prominent independents, both candidate types which drag down the minor parties in several other NSW seats.  While it may seem independents mainly drag down the Coalition vote, there is very strong evidence from comparing Reps and Senate votes that in the rural seats of Cowper and Farrer it was mainly Labor voters who went over to independents Cowper and Mack.  Eden-Monaro also has a fairly high Shooters and Fishers vote.)

Another, simpler way of calculating candidate effects between the major parties in the Reps is to subtract the Senate 2PP from the Reps 2PP.  (I've just used the #1 candidate for each major party ticket to calculate the 2PP.)  In cases where there is a bad candidate for one of the parties, this method doesn't say whether the difference was caused by one party's candidate being good, the other party's being bad, or some combination of both.  It is also a rough measure because voters for particular Senate parties might preference one major over another but be less likely to distribute their Senate preferences.  Anyway, the 2019 Senate 2PPs in Eden-Monaro and NSW statewide were virtually identical (52.78 to Coalition vs 52.8).  However the Reps 2PP in Eden-Monaro was 49.15% for Coalition, compared to a statewide 51.8%.  On this comparison, Labor's performance through Mike Kelly comes out 2.63% better as a 2PP swing (or double that on 2PP margin).

But Senate vs Reps personal vote comparisons have one obvious flaw however they are done - they don't say to what extent a popular local member lifts both the party's Senate primary vote and its Reps vote.  Maybe the lift is minor, but it probably exists.  Given all the evidence, I think Antony Green's estimate of Kelly's personal vote at three to four points is likely to be on the money.

(William Bowe has also looked at personal votes by measuring differences from a vote share projected off demographic factors that influenced a specific election, but that method works less well in regional NSW).

So Labor is losing 3-4 points of personal vote, is that fatal?  Well, no, not by itself.  The average swing in Opposition by-elections is 1.2% to the Opposition, but that includes many vacating MPs who took large personal votes with them, former PMs Rudd and Fraser among them.  The best baseline for deducting Kelly's personal vote from is the typical swing from all by-elections, and this only cancels out that swing or nearly so.

So before we consider special factors, Labor seems slightly ahead.

Special Factors

The following are some special factors that may influence the election result:

* Candidates:

Labor have aimed for authenticity and crisis experience (and initially to cover off against Constance) with their selection of Kristy McBain, the first-term Bega Valley Mayor and second-term councillor who also became suddenly prominent leading local responses to the bushfire crisis, starting immediate speculation of a bigger career.  McBain, a lawyer and lapsed-then-recently-rejoined ALP member, does not come across like a career politician either, though she is being quick to stress that she has the experience to do the job.

After a messy pre-preselection stoush involving John Barilaro, Andrew Constance and others the Liberals have gone for their 2019 candidate Fiona Kotvojs, who was the first female two-party-preferred candidate ever picked for this seat.  Kotvojs is a well-credentialled candidate with a PhD in education, experience in farming and charity (Oxfam) and as a voluntary counsellor, Army Reservist and fire service volunteer.  Kotvojs' past comments on climate change have come under some attack from the online left but these attacks are overblown, painting Kotvojs as a denialist for any remark seen as downplaying the threat (even in an area where Kotvojs has academic expertise).

With Constance out, candidate factors are likely to favour Labor unless McBain has any political skeletons in the closet.  (The closet these days is usually Facebook).  The swing in Eden-Monaro in 2019 was not greatly different to the national average, so provides little evidence either way about Kotvojs' appeal, especially since Hendy was regarded as a dud incumbent.

Comments regarding Andrew Constance's possible appeal are no longer relevant and have been removed.  Constance was initially a confirmed candidate for preselection then withdrew the day after.  Some views of his likely vote-pulling appeal were probably overblown.

* The bushfires: Eden-Monaro was very severely impacted by the bushfires, which destroyed nearly 1000 properties in Constance's Bega electorate alone and caused severe smoke pollution even in areas not directly threatened by the fires.  The electorate was ground zero for the backlash against the Prime Minister's handling of the fires, including the infamous Cobargo incident in which PM Morrison was heckled by locals, some of whom were a few months ahead of their time in their desire to avoid shaking hands.  To what extent locals who would normally vote for the Coalition might still feel angry or abandoned could be a major factor here.

* Coronavirus:  Of course economic recovery issues connected with the COVID-19 shutdowns are bound to be prominent in the campaign. A common theme has been that the COVID-19 crisis will give the government a massive lift in the by-election.  For sure, there is a reasonable historic relationship between government polling at the time of a by-election and swings at that by-election.  However, the by-election is still months away and a lot can change in terms of the government's standing at the time.  The other problem is that, as that relationship goes, there isn't much evidence even now that COVID-19 is shifting federal voting intentions.  Rather it so far seems to be mainly just boosting the Prime Minister's own ratings.  This is a little surprising, given that Australia has seen remarkably good results in containing COVID-19 thus far without needing to lock down as severely as some countries.  It is worth noting (see Tally Room) that there isn't much evidence of the disease in this electorate to this stage.

Something I don't expect to now be a factor is the three-cornered contest (if there is one).  If the Nationals do run at all without Barilaro they will probably just poll a similar vote of several percent to last time with some of it leaking to Labor (in 2019 nearly 13% of Nationals preferences went to Labor).  That is not to say that a token Nationals candidate would be damaging.  Probably the votes they would leak to Labor would be voters who personally knew the National candidate, or idiosyncratic voters who would have preferenced Labor above the Liberals anyway.  The spoiler effect might look compelling if Labor again wins off this "leakage" but that wouldn't mean it was actually the cause.

I'd also expect the Coalition chaos involving Barilaro and Constance, and the wider National Party infighting to have blown over in the minds of most voters by election day, but it's worth keeping an eye on because it could damage any candidate who was linked to it (such as Barilaro if he changed his mind and decided to run.)


There have been various internal polling rumours that don't deserve that much attention.  A Nationals poll of a matchup between John Barilaro and Jim Molan was said to show Barilaro beating Molan on primaries then winning 52-48, which in view of the accuracy issues with seat polls lately would mean little even with Molan as the candidate.  It would not even show that Barilaro would beat a generic Liberal, as Molan has a cult following but would also have significant negatives in a within Coalition contest among moderate and possibly younger Liberal voters.  The Liberals are said to have polling by Crosby-Textor showing Constance beating Barilaro and McBain but no credible numbers or further details have been published.  Mutterings about a result in excess of 60-40 to journalists need not detain us here.

In mid-May an Australia Institute uComms robopoll was reported with a last-election 2PP of 51.1 to Labor off primaries that according to William Bowe came out as "Labor 39.8%, Liberal 34.3%, Nationals 7.3%, Greens 6.7% and One Nation 6.5%" after exclusion of what this pollster calls "undecided".  Seat polling is very inaccurate for many reasons.  The poll found coronavirus and bushfire recovery to be taking a back seat as nominated issues to the economy, though concern about the economy may well be dominated by coronavirus-related impacts.  Other findings are bound to include the usual skewed waffle in TAI's issue questions.

On 25 May a further uComms robopoll commissioned by GetUp! was reported by the Guardian, supposedly showing agreement with a claim that the government was not doing enough about climate change.  Until full details of this poll have been published including all wording of all preambles and questions in the order asked I am not taking it seriously.


Betting is not reliably predictive but it is amusing to keep an eye on.  At a certain site that is in the doghouse for paying out early on the 2019 election and therefore is not being named on here at the moment, the Coalition was a narrow favourite a few days ago and has now come down to 1.55 vs 2.55.

Update 6 May: Following Constance's withdrawal odds were 1.80 vs 2.10, but I expect McBain to now become favourite.  (Further update: by the end of the night that was true, McBain 1.85 vs 2.10.)

7 May: McBain 1.65/2.20

12 May: 1.60/2.25

16 May: 1.70/2.25

24 May: 1.60/2.30

Other Candidates

Aside from the question of whether or not the Nationals run, other candidates are likely to just be making up the numbers, but there will probably be a lot of them and they could well drive up the informal rate.  The Greens are expected to run and other declared candidates include:

Dean McCrae (Liberal Democrats)

Serial candidate James Jansson of the so-called Science Party, which is considered unworthy of the name on these pages for its unscientific opposition to Senate reform (see comments here)

Andrew Thaler, a serial independent who ran for this seat in 2013 and 2016 polling just over 1% each time, has also run for NSW state parliament and has a colourful internet history

Karen Porter, an independent running on behalf of an unregistered party called the New Liberals, in fact an anti-Liberal small-l liberal/populist outfit.  It will be interesting to see whether this party succeeds in being registered under this name when the time comes as the Liberal Party are highly likely to challenge the name and might have some prospects of successfully arguing that it implies a connection with the Liberal Party,

See also

Poll Bludger guide
Tally Room guide

Saturday, May 2, 2020

Legislative Council 2020: Rosevears And Huon Not Live

In some alternative universe, the polls will close about four hours from now ...

In the normal scheme of things, today would have been the day for the Rosevears and Huon Legislative Council elections.  I think it is worth a quick post to reflect on that fact and to summarise where things are with the postponement of these elections, which I have also been covering in an article that is now well down the list.

The elections were postponed because of risks associated with the current coronavirus outbreak.  Indeed in recent weeks Tasmania has had the nation's proportionally most severe outbreak of COVID-19, but it has been almost entirely confined to the north-western health system and close contacts of individuals within it.  A very small number of cases within that outbreak have been diagnosed in the North and South rather than the North-West, but beyond that the South has had only two cases in the last month (for one of which on 6 April, no detail ever appeared to my knowledge) and the North has not had any.