Reflections on the upcoming Victorian election

It is a bit over 48 hours until polls close for the Victorian state election. After eight years in office, Labor is running for a third four-year term under long-serving leader Daniel Andrews, while the Coalition is attempting to return to government with recycled leader Matthew Guy.

A Labor victory would mean that by the time of the next election in 2026, Labor will have been in power 23 out of the last 27 years, an extraordinary era of electoral dominance in Victoria.

Media commentary in the lead-up to elections in Australia is always dominated by “the vibes” rather than empirical analysis, a source of great frustration to yours truly. It is no different this election – except that there is even less empirical data than normal, so we are flying blind even more than usual! Given the limited data, in this post I will consider some of the uncertainties and conclude with my own read of “the vibes” (which you should take with a grain of salt).

The uncertainties

There are some key questions going into this election:

  • The 2018 election was a disaster for the Liberals and a thumping success for Labor. The Liberals can only go up and Labor only down from here. But by how much? A uniform swing of around 7% would be needed to endanger Labor’s majority (assuming no seats are lost to minor parties or independents, more on this below) but a swing of closer to 10% would be needed for a Coalition majority.
  • Labor has a lot of long-serving lower house MPs retiring at this election (13 MPs compared to 4 Liberals and 1 National). Will the loss of this personal vote endanger Labor seats on larger margins that normally would not be competitive?
  • The Liberals have sensationally announced that they will be recommending preferences to the Greens ahead of Labor in the Lower House for the first time since 2006. This would seem to doom Labor’s hopes of retaining Northcote and Richmond, which were always going to be close. But if the Greens have a really strong result, could they also gain other seats like Albert Park and Pascoe Vale, seriously damaging Labor’s hopes of winning a majority?
  • There has been a substantial redistribution of electoral boundaries in Victoria in response to population growth over the last decade – two new seats have been created in Melbourne’s west and north, and another seat created on the outer south-eastern fringe. In contrast, two seats have been abolished in the stagnant eastern suburbs and a third seat abolished in the south-east. The redistribution has clearly been beneficial to Labor. (You can compare the old and new boundaries here.) But with many MPs representing new areas, will this lead to greater unpredictability?
  • In the Federal election, there were substantial swings to the Liberals in some safe Labor seats in the outer suburbs, while marginal inner suburban seats swung to Labor. Will this phenomenon repeat at the state election, resulting in wildly different swings in different parts of the state?
  • Will the widely covered independent challengers to Labor in the western suburbs end up having any impact? Labor only holds two seats in the western suburbs with margins under 12% (Melton 5% and Werribee 9.1%). It would require gargantuan swings for Labor to lose any seats above 12%.
  • Prior to the 2022 Federal election, independents had been most successful in regional areas. Will prominent independents in Benambra and South West Coast defeat incumbent Liberal MPs? And can independents retain Mildura and Shepparton?
  • Finally, there’s the topic on everyone’s lips: the Teal independents. After their extraordinary success in the federal electorates of Kooyong and Goldstein, will they manage to defeat any inner suburban Liberal MPs (or the Labor MP in Hawthorn) at the state level?


Take this with a grain of salt but here are my thoughts on likely answers to the questions above and possible results:

  • There hasn’t been a tonne of polling this election, but the polling that does exist shows Labor more likely than not to be returned with a majority, all else being equal. Unless the Greens and independents do very well, I see no reason to doubt this conclusion.
  • There may be a lot of volatility across the state. Labor was so dominant in 2018, their vote will likely fall everywhere – but will the biggest falls be in their most marginal seats or in very safe seats? I expect bigger swings against Labor in the outer suburbs but this will probably only cost them seats to the Liberals in the outer south-east that are on smaller margins (Pakenham, the new seat of Berwick, Hastings, Nepean, Bass, Cranbourne). I also expect Labor to lose some/most of the eastern suburbs seats they gained in 2018 (Ringwood, Ashwood, Box Hill, Hawthorn). In contrast, I think Labor will probably hold their seats in regional areas (and maybe they will win Morwell too).
  • It is very hard to get an objective read on independent campaigns. Pollster Kos Samaras, who predicted the Teal independent surge at the Federal election, reckons that Benambra, South-West Coast, Kew and Mornington are all looking strong for independents, with Hawthorn very close. Roy Morgan Research, another pollster predicts independents will win Benambra, Kew, Hawthorn, Brighton and Melton. My thoughts on each of these seats:
    • I agree regional independents have an excellent chance against the Liberals in Benambra and South West Coast.
    • Kew, Hawthorn and Brighton, all seem the most logical wins for Teal independents, along with Caulfield (which strangely, neither pollster mentions). I’ll predict that the independent wins Kew – retiring incumbent Liberal Tim Smith has done his best to alienate his constituents there – but I’m not sure they’ll break through in the other three seats. The Hawthorn independent seems to be relatively high profile, but the local Liberal candidate (an ex-MP) also has a high profile and by all accounts has run a very strong ground campaign. In Brighton and Caulfield, I don’t even know the independents’ names – I doubt they have enough name recognition to break through. (At this point, I should disclose that I was a hard Teal sceptic at the federal election – I thought they’d retain Warringah, gain Wentworth and maybe North Sydney, but I gave them no chance in Mackellar, Kooyong, Goldstein and Curtin. So I could very well be underestimating them again.)
    • I agree Melton is the best chance for a conservative independent victory in the western suburbs. If they cannot win Melton, they are not winning anywhere else.
    • I have no clue why Mornington keeps being talked about as a possible independent victory. There is absolutely nothing about the demographics of this seat (one of the oldest and whitest electorates in the state) that would indicate it is fertile ground for an independent, not to mention the abject failures of independents in the local area at the last two federal elections.
    • So to summarise: regional independents probably win Benambra and South West Coast, a Teal wins Kew and maybe one other, and a conservative independent wins Melton, resulting in 1-2 Labor seat losses and 3-4 Liberal seat losses. This assumes I am not systematically underestimating inner city independents.
  • Prior to the Liberal announcement that they would recommend preferences to the Greens ahead of Labor, I thought the Greens were a good chance to gain Richmond and perhaps Northcote as well. Since the Liberal announcement, it seems virtually certain that the Greens will win these two seats. The question then is what about seats like Albert Park or Pascoe Vale? If I had to guess, I would say Labor hold these seats. Labor and Daniel Andrews still seem pretty popular in the inner suburbs so I don’t expect Labor’s vote to fall enough here to put these seats at risk.

If you’ve been adding up my vague seat predictions above, you would get to about 43-46 seats for Labor. Given Labor needs 45 seats for a majority, this puts them right on the verge of minority or majority government. Having said that, I expect it is more likely than not that Labor will overperform than underperform on a seat-by-seat basis – they are the incumbent government after all.

My bet on the scenarios from most likely to least likely:

  1. A narrow Labor majority of 1-4 seats.
  2. Labor falling 1-3 seats short of majority, ultimately forming government with regional independents or Teals.
  3. A comfortable Labor majority of 5 seats or more.
  4. A hung parliament with Labor falling 4 or more seats short of a majority.
  5. A Coalition majority

We’ll find out after 6:00pm on Saturday.

Other great election previews:

Best sources for election night coverage:

POSTSCRIPT SAT MORNING: We finally have a Newspoll released late Friday night, showing Labor up 54.5% on two party preferred. If that result is replicated tonight, scenario 1 (a narrow Labor majority of 1-4 seats) and scenario 3 (a comfortable Labor majority of 5 seats or more) are much more likely than a minority government.


Who will control the next Senate?

Sometime before May 2022, Australia will have a federal election. As well as voting for the House of Representatives, around half of the Senate will be up for election. This post will provide an overview of the current state of the Senate and consider possible scenarios for the make-up of the next Senate.

This post shouldn’t be red as a forecast – merely an assessment of different scenarios at this moment in time.

Senate basics

First, the basics. There are 76 seats in the Senate. 72 senators serve fixed six-year terms, with half of these seats up for election every three years. Each state gets 12 senators. 4 senators represent the ACT and the NT, who each get two senators and face voters every three years.

This means that the term of half the state senators will expire on June 30 2022, while the other half of the state senators (who were elected at the 2019 election) will continue serving until June 30 2025.

39 seats are required to pass legislation. It is extremely rare for one party to win a majority in the Senate, so governments are required to negotiate with minor parties and independents (crossbenchers) to pass legislation. If these crossbenchers are ideologically aligned with the government or centrist, passing legislation is usually manageable with a bit of horse-trading. If not, passing laws can be more challenging.

To summarise: there are 36 state senators serving until 2022, 36 state senators serving until 2025 and 4 territory senators serving until 2022. Confused? Okay!

A note on the territories

Every Senate election in the ACT and NT has always resulted in the Coalition and Labor winning one seat each in each territory. Although this is not guaranteed into the future, the result has never been particularly close in the past. In the absence of polling showing a collapse in either party’s support in the ACT and NT, I am going to assume that this result will be repeated at the next election. As such, to simplify the following analysis, I will not mention the territories again.

So let’s start this analysis by looking at the 36 Senators who we know will be in office until 2025 and will serve in the next Senate regardless of the next election result.*

The 2019 result: Senators serving until 2025

These 36 senators were all elected at the 2019 election. This election was relatively close, with the Coalition winning the two-party preferred vote by a couple of points. The seat results by party were:

  • Coalition: 17
  • Labor: 11
  • Greens: 6
  • Jacqui Lambie: 1
  • One Nation: 1

The sum ideological split is 18 right (Coalition plus One Nation), 17 left (Labor plus Greens) and the hard-to-classify Lambie.

Let’s consider the results by state. In NSW, VIC, WA and SA, the seat breakdown was very straightforward, with three seats going to the right (all to the Coalition) and three seats going to the left (2 Labor, 1 Greens). I will call this the ‘default’ result.

In Tasmania, the left won three seats but the right only won two seats, with Lambie taking away enough of the right’s vote to win a seat herself.

In Queensland, the Labor vote collapsed to such an extent that the left only won two seats (one Labor and one Greens). The right won four seats, with the Coalition winning three and One Nation winning one. I have summarised all this in the table below.

Table 1. Senate results by state in 2019.

StateLNPALPGRNONLambieRight-left-other split

What does this tell us? Unless a minor party or independent has a relatively high primary vote (eg. Lambie in TAS) or a major party’s primary vote collapses (eg. Labor in QLD), the most likely result is each state electing three senators from the right and three senators from the left.

Scenario 1: A three-three ideological split

With a three-three split being the ‘default’ result, let’s consider the scenario where this result eventuates in every state at the next election.

If each state does in fact elect three senators from the right and three from the left, this means that:

  • The right will hold exactly half the seats in the Senate.
  • The left will be one seat short of holding half the seats.
  • Lambie will be the sole non-ideologically aligned senator.

In this scenario, Lambie will be extremely significant. If she votes with the right, she can pass bills. If she votes with the left, she can deadlock the Senate.

If this scenario eventuates, it will have major implications depending on whether the Coalition or Labor form the next government (more about this in ‘Summing Up’).

Of course, this is only one possible Senate scenario. There are at least two other plausible scenarios:

  • The left or right manage to win four seats in one or more states.
  • A centrist or non-aligned candidate is elected in one or more states.

Let’s consider these alternative scenarios.

Scenario 2: A 4-2 ideological split

Which state is most likely to deliver a 4-2 ideological split? The Coalition and One Nation achieved this in Queensland in 2019 and could do so again. And after WA Premier Mark McGowan’s truly crushing election victory in March, it is tempting to think Labor and the Greens could win 4-2 in WA.

But federal polling shows things are much closer. Newspoll’s latest quarterly breakdowns for each state (covering the period January to March) do not show either Labor or the Coalition with substantial leads in any states.** In two-party preferred terms, the polling in NSW, VIC, QLD and WA shows a result of between 53-47% in the Coalition’s favour to 53-47% in Labor’s favour. In SA, Labor is leading 55-45%.

What does this mean? At this point in time, it doesn’t appear that the left or right has anything like the dominance needed in any state to achieve a 4-2 senate result. SA is the only state that comes close – but even a 55-45% split is unlikely to be enough.

Of course, this situation may change over the next 12 months – the Coalition or Labor may experience a surge in support or the polling could be wrong – but I think at this point it is fair to consider a 4-2 split in any state unlikely.

Scenario 3: Minor parties and independents

But what about the prospects of minor parties and independents?

Since the removal of undemocratic group voting tickets in the lead-up to the 2016 election, which enabled minor parties to win seats despite getting hardly any votes, getting elected requires actually winning a reasonably large chunk of votes. In 2019, the only minor parties elected in any state with a primary vote of below 10% were the Greens in NSW (8.7%), Lambie in TAS (8.9%) and the Greens in QLD (9.9%). The largest vote for a minor party that did not win a seat was One Nation in WA (5.9%).

Considering there has only been one half-Senate election under the new voting system, it is risky to make sweeping generalisations. But I think we can tentatively say that a minor party has a good chance of getting elected with a primary vote of over 8% while it is very hard to get elected with a primary vote under 6%. The 6-8% range is a bit of a grey zone.

So how are the minor parties polling?

Newspoll shows the Greens polling above 10% in every state making them highly likely to win a seat in every state (if an election were held today). Conversely, One Nation is polling 4% or below in every state except QLD, where they poll 8%. So One Nation have a good chance of holding their only seat in QLD but do not look like they have much chance of winning any others.

As for other minor parties or independents, none even come close. There are currently two incumbent crossbench senators from South Australia who are both up for election. Both are centrists: Stirling Griff from the Centre Alliance and Rex Patrick, an independent formerly of the Centre Alliance. Neither of these senators are showing up in the polling and I struggle to see either of them getting elected. Neither of them seem to have much of a public profile in South Australia (at least from what I can tell). And at the 2019 election the Centre Alliance completely bombed, winning just 2.6% of the vote. Is there a bastion of support for Griff and Patrick that is hard for outsiders to see? Perhaps. Could one or both senators significantly increase their profile in the next 12 months? Perhaps. But I think the safe assumption at this point is that neither senator will be re-elected.

Which other minor parties or independents could get elected? Honestly I can’t think of any with a realistic chance. Maybe Lambie will endorse another Tasmanian candidate to run with her backing. Maybe some new independent will throw their hat into the ring and gain a significant following. But it’s anyone’s guess.

Summing up

None of this post is predictive. A LOT can happen over the next 12 months. But right now, the most likely Senate make-up after the next election appears to be 38 senators on the right (Coalition and One Nation), 37 senators on the left (Labor and Greens) and 1 Jacqui Lambie.

Both the Coalition and Labor would find this Senate a tough proposition. A Coalition Government will always need the support of either:

  • One Nation AND Lambie, or
  • Labor, or
  • The Greens

But the situation for a Labor Government would be even worse. Passing bills will require the support of either:

  • Coalition, or
  • Greens AND Lambie AND One Nation***

This could have a significant impact on a future Labor Government’s legislative program. They are not going to be proposing progressive legislation if they need to rely on the support of One Nation or the Coalition to pass it.

Perhaps Labor could tempt a Coalition MP to defect and become speaker? But now I’m really speculating.

*Unless the government calls a double dissolution election. But let’s not go there.

** Newspoll does not publish a state breakdown for Tasmania.

***If One Nation hold two seats, Labor wouldn’t need the support of Lambie in this scenario. But let’s face it, if Labor aren’t getting the support of Lambie, they aren’t going to get the support of One Nation.


Newspoll’s latest quarterly breakdowns, used in this post, show the Coalition leading 53-47 in Queensland and Western Australia, a 50-50 split in New South Wales and Labor leading 53-47 in Victoria. If this were to eventuate at the next election, this would represent a 2% swing to Labor in WA and NSW, a 5% swing to Labor in QLD and no change in VIC. This could very well give Labor a narrow majority in the House of Representatives. So a scenario where Labor forms government while the Coalition and One Nation hold a blocking majority in the Senate is not far-fetched.


The 681 forgotten people

Major policy issues generally follow one of these three paths:

  • Resolution: A government can respond to a major policy issue in the face of public and elite pressure by developing and implementing an adequate policy response. Remember gay marriage? This was a major political issue in the mid-2010s and a key topic of discussion during the 2016 election. But the government (eventually and reluctantly) legalised gay marriage and resolved the issue. It has long since ceased to be a topic of mainstream discussion.
  • Stalemate: A government can respond to a policy issue by denying that the issue exists and therefore refusing to develop an adequate policy response. The issue may then remain in the mainstream of political and media discussion indefinitely. The iconic Australian example of this is climate and energy policy – the Federal Government does not take this issue seriously and does its utmost to make the issue go away. But strong interest from the public and political and media elites means that, even though interest peaks and troughs, the issue fails to totally go away and remains a live mainstream issue.
  • Apathy: This is the most depressing trajectory: a government refuses to develop an adequate policy response to a major issue but manages to persuade a critical mass of public and elite opinion that it has developed and implemented a policy response, thereby “solving” the issue. Without sufficient attention from political and media elites, the public loses interest, creating a vicious cycle whereby the issue leaves mainstream discussion and cannot break back in.

This latter route, sadly, has been the path of asylum seeker policy in Australia. Asylum seeker policy was a huge political issue at the 2010 and 2013 elections, and arguably the most hot-button political issue in the first half of the 2010s (alongside climate change). But since then, it has mostly faded into the background, only coming to the fore in relation to other political issues (eg. the medivac legislation, which was intertwined with the Federal Government’s loss of control of the House of Representatives in late 2018 and early 2019).

Despite the mainstream view, encouraged by the Federal Government, that this issue has been “solved”, the fact is that the Nauru and Manus Island detention centres remain open, as do a number of domestic detention centres holding asylum seekers. In recent years, even hotels have been turned into prison cells for asylum seekers.

In total, 239 asylum seekers remain in offshore detention, with 130 on Manus Island and 109 on Nauru. A further 442 asylum seekers are in onshore detention. That’s a total of 681 asylum seekers locked up by the Federal Government. (This does not include a further 1,053 people currently locked in onshore detention centres, primarily for visa violations).

The United States resettlement deal, negotiated by Malcolm Turnbull when he was Prime Minister, whereby asylum seekers are given the option of permanently resettling in the United States, has been the only bright spot (as deeply problematic as it is) in this bleak era of policy failure. 395 people were resettled in the US in 2018, 214 in 2019 and 243 in 2020. One can hope that most of the remaining people in offshore detention will experience freedom in the US in the not too distant future.

But that will not be the end of the matter. A further 12,255 people, the vast majority of asylum seekers, are no longer in detention and have been released into the Australian community on a bridging E visa. Although certainly preferable to detention, these people have no path to permanent residency. That means there are 12,255 people living in Australia, many for a number of years, who have not been provided with any assurance that they will be able to remain here in the long-term. I cannot imagine how difficult it must be to live a life with such uncertainty.

Asylum seeker policy is not an issue I normally write about in this blog. I draw attention to it not only to remind people of the immense human tragedy of Australia’s asylum seeker policy failure and the intense, unspeakable suffering it has caused. It is also a cautionary tale.

Not all policy issues are resolved or end in stalemate – some are just forgotten.

*Data on offshore detention is correct as of 31 March 2021. Data on onshore detention and bridging visas is correct as of 28 February 2021.

If you are trying to read the data tables on the Department’s website, note that the Federal Government makes heavy use of Orwellian Newspeak. For example, asylum seekers are called “illegal maritime arrivals” or IMAs. Hotels and other facilities not designed to imprison people are called “alternative places of detention” or APODs.


The five factors that determine when coal power stations close

With the closure of Australia’s dirtiest coal power station recently brought forward to 2028, now seems a good time to publish a post I began writing late last year but never finished, which analyses five of the most important factors that influence when coal power stations close.

A quick overview of Australia’s coal fleet

There are 19 operational coal power stations in Australia located across four states: eight in Queensland (Gladstone, Tarong, Tarong North, Millmerran, Kogan Creek, Stanwell, Callide B, Callide C), five in New South Wales (Liddell, Eraring, Vales Point, Bayswater, Mt Piper) and three each in Victoria (Yallourn, Loy Yang A, Loy Yang B) and Western Australia (Muja CD, Collie, Bluewaters). In the three eastern states, they supply between 70 and 80% of each state’s electricity. In Western Australia they supply about 40% of the power in the state’s main electricity grid.

They range in size from the 340MW Collie power station to the gigantic 2,880MW Eraring power station. In age they range from 12-year-old Bluewaters to 50-year-old Liddell.

Don’t we already know when coal power stations will close?

In theory, yes. The Australian Energy Market Operator (AEMO) requires all power station owners on the east coast to publish their planned closure dates here. The owners must provide at least five years notice of any closure, and risk penalties if they don’t. If we take this at face value, just four coal power stations are expected to close by 2030: Liddell and Vales Point in NSW, Yallourn in Victoria and Callide B in Queensland.

But it would be unwise to take this at face value. If a company is bleeding money on a coal power station and they see no hope of things changing, they are not going to let themselves go bankrupt – they’ll close it down. Even substantial fines are unlikely to deter this. Coal power stations can cost hundreds of millions of dollars every year to maintain and keep safe as they get older. I expect most companies would rather pay a substantial fine to shut a power station early than pump more money into assets that are unprofitable.

So take the official closure dates with a grain of salt. Instead, consider the following five factors.


Age is the most obvious factor influencing coal closure. As coal power stations age, they become increasingly expensive to maintain and increasingly prone to faults as machinery reaches the end of its technical life. Australia’s coal power stations are generally pretty old. Just five have been built since the year 2000 and the four oldest were built back in the 1970s (Liddell, Yallourn, Vales Point, Gladstone). Age also has an impact on the following factor: flexibility.


Coal power stations are often described as providing ‘baseload’ power. This means they are designed to operate at close to full capacity 24 hours a day, 365 days a year. That makes them poorly fitted to the modern electricity grid, with wind and solar providing a varying supply of electricity at different times of the day, forcing other power stations to also vary their supply in response. Electricity demand is also becoming increasingly varied over the course of the day, with demand falling significantly in the middle of the day and then quickly rising in the evenings.

It is true to say coal power stations are not designed to ramp their supply up and down but this doesn’t mean they cannot. All coal power stations have to ramp to some extent these days. But it takes its toll on the machinery. Generally speaking, the older the power station, the more difficult it is to ramp up and down, the higher the maintenance costs and the more likely the faults.

For example, the near-50 year old Yallourn power station in Victoria has real difficulties ramping up and down, so it avoids doing this as much as possible. In contrast, the Mt Piper power station in NSW, which was built in the 1990s and is a lot younger than Yallourn (although it is still almost 30 years old) is much more versatile and often ramps and down.

The more flexible a power station, the better its chances of hanging on into the future.

Cost competitiveness and electricity prices

Part of the reason coal power stations provide so much of Australia’s electricity is that they are relatively cheap to run – before wind and solar came along, coal power stations were the cheapest source of electricity in Australia, providing power at a far lower cost than gas and hydro.

But that is no longer the case. Wind and solar can provide power with virtually no running costs. As more and more wind and solar farms have been built across the country, the wholesale price of electricity has fallen and the most expensive coal power stations have been put under a lot of financial pressure.

According to an excellent recent analysis from IEEFA, the cost of running Australia’s coal power stations (know as the short run marginal cost) varies from $12/MWh to $42/MWh (unfortunately this analysis didn’t include WA). The three cheapest coal power stations are all in Victoria (Loy Yang A, Loy Yang B, Yallourn). The three most expensive power stations are all in NSW (Eraring, Mt Piper, Vales Point).

Cost competitiveness is not just important on a national level; within-state competitiveness is crucially important. For example, the Yallourn power station is the third cheapest coal power station on the east coast, but it is the most expensive in Victoria, so it loses out to its most direct competitors Loy Yang A and Loy Yang B.  Conversely, the Vales Point power station is the third most expensive on the east coast but two other power stations in NSW, Eraring and Mt Piper, are more expensive than it is. So Vales Point is less at risk than they are.

Why is NSW home to the most expensive coal power stations and Victoria to the cheapest? The biggest factor determining running costs is the cost of coal supply. All of Victoria’s coal power stations (and about half of Queensland’s) get coal from mines co-located with the power station and owned by the same company. All of NSW’s coal power stations (and half of Queensland’s) must get their coal from other companies who own coal mines.

The price of coal from these coal mines is heavily influenced by the international market price. If the seaborne coal price goes up, then so does the price of coal for domestic coal power stations. (This does not apply to Victoria because we are not connected to the international market – our coal is so dirty and has such a low energy content, no one would pay to import it). 

Owner incentives

Coal power stations have a range of different owners with varying incentives when it comes to closures.

Most of Queensland’s and WA’s coal power stations are owned by their respective state government. This insulates them to an extent from profitability concerns. WA’s coal power stations in particular are very unprofitable – if they were to be privatised, I would expect a significant amount of capacity to close very quickly.

Private companies also have different incentives. The Vales Point power station is the primary income stream for Delta Electricity – closing it would kill off much of their profit. But equally, if Vales Point becomes unprofitable, Delta is far more likely to suddenly close it than one of the big energy companies because they cannot absorb the losses. For the big gentailers, AGL, EnergyAustralia and Origin, who between them own six of the eight coal power stations in Victoria and NSW, their coal power stations are just one income stream of many (albeit a large one). 

There are also a couple of niche cases. The Gladstone coal power station, the oldest and largest in Queensland, is owned by Rio Tinto and primarily exists to supply power to Rio’s Boyne Island aluminium smelter.  The wholesale price of electricity is therefore not very important. As long as the smelter remains open, Gladstone is also likely to remain open as long as it physically can. And the reverse is true: the closure of Boyne Island will almost certainly lead to Gladstone’s closure.

Supply-demand dynamics

There are also varying supply-demand dynamics in the different states. Over the last decade, there has been a rapid increase in renewable energy generation in most states. This has been accompanied by a fall in electricity demand in NSW and Victoria. This has placed a lot of pressure on the coal power stations in both states and lead to a number of closures, with renewables outcompeting coal and eating into its market share.

In contrast, Queensland and Western Australia have seen growing demand for electricity over the last decade. This has meant that even though renewables have been growing in both states, coal power stations have not lost much of their market share. This is particularly the case in Queensland, which has seen only a very small reduction in coal capacity to date.

So, which power stations are in the firing line?

Weighing up all these factors, a few power stations stand out as being particularly vulnerable.

Victoria: Yallourn is clearly the most vulnerable. It is the most expensive in the state, the oldest and most polluting and will increasingly be unable to compete with the flood of new renewables. Owners EnergyAustralia recently announced that they would bring forward the closure date from 2032 to 2028 but that still seems wildly optimistic. I expect units at Yallourn to begin dropping off within the next four to five years, and I definitely wouldn’t rule out a full closure by 2025/26.

New South Wales: We already know that the Liddell coal power station will begin closing in April next year and fully close by April 2023. This will leave four coal power stations left standing in the state. The owners of these four power stations will likely wait to see if electricity prices go up again in the wake of Liddell’s closure before announcing any closures. If prices don’t rise or if any rise is short-lived, expect another coal closure announcement in 2024.

Eraring is clearly the most vulnerable – it is the most expensive coal power station on the east coast and Origin already seems to be managing expectations for an early closure. Eraring is so gigantic that you shouldn’t expect the whole thing to close at once – Origin will likely close one unit at a time. But depending on exactly how much pressure Eraring is under, I wouldn’t be surprised if an announcement for a partial closure is made even before Liddell closes – perhaps even within the next 12 months.

Vales Point is a bit of wild card – it is marginally cheaper to run than Eraring but a bit older. And its owners (Delta) are less able to absorb big losses. If things get bad at Vales Point and Origin hold out on an early closure announcement for Eraring, don’t be surprised if Vales Point goes very suddenly.

Mt Piper is the country’s second most expensive coal power station, but it is the youngest and most flexible in NSW. Expect the owners (EnergyAustralia) to endure the current low prices in the hope that Eraring or Vales Point close first, especially as they are likely to have their hands full with an early Yallourn closure in Victoria.

Queensland: I continue to think an early coal closure in Queensland is underrated. The state has seen a surge in solar generation in the middle of the day that is continuing to grow and the state’s coal fleet simply cannot continue to absorb it. With six coal power stations owned by the Queensland Government, expect them to announce the mothballing or closure of at least one in the next couple of years. Perhaps Callide B or a unit at Tarong/Tarong North? Gladstone also looms as a wild card: if Rio Tinto shuts the Boyne Island smelter or enacts a plan to transition it to a clean supply of electricity, Gladstone will likely close.

Western Australia: The WA Government has announced the closure of the two ‘C’ units of the Muja CD power station by 2024. It would not be surprising if this were to be brought forward and a closure date also announced for the two ‘D’ units in the next couple of years. Muja CD is the state’s oldest coal power station by a long distance and, like in Queensland, it is struggling to absorb the rapid growth in rooftop solar in the middle of the day. 


Coal in Australia, Part 3: New South Wales

This post concludes a series I began back in this blog’s early days looking at coal production in Australia. The first two posts looked at Western Australia and Victoria and Queensland. Today we turn to New South Wales.

Firstly, an explanation for the delay in profiling NSW. There is a frustrating lack of official, publicly available information on coal production in New South Wales. The state government does provide quarterly reports on coal industry statistics but you must pay $506 to read them! This is a disgrace, clearly intended to limit access, avoid scrutiny and make money. It stands in stark contrast to the Queensland Government, which provides a wealth of data on the state’s coal industry for free.

I have spent much time trying to find another comprehensive and publicly available data source but to no avail. That means that this post will be less comprehensive than my posts for other states and rely on a wider range of sources, some of which are quite out of date. 

Coal in New South Wales

New South Wales has 38 operating coal mines. Unlike Queensland, which has two distinct coal regions in the south and centre, New South Wales’ coal mines are geographically concentrated in eastern New South Wales. Many of these coal mines are located relatively close to Sydney, in contrast to Queensland, Victoria and Western Australia, where coal mines are located fairly long distances from the capitals.

New South Wales’ coal mines can be divided into roughly two regions, one to the north of Sydney and one to the south. (I have been unable to find an especially good map of NSW coal mines but the least bad one I have found is from the New South Wales Minerals Council).

The large Northern region, which includes the Hunter Valley, the Western Coalfield and the Gunnedah Basin, begins in Narrabri in the north-western corner, running directly south down to Lithgow then east to Lake Macquarie before heading north along the coast to Newcastle. From there it continues up the coast towards Port Macquarie before heading back inland towards Narrabri. This region includes 34 of the state’s operating coal mines and produces around 95% of the state’s coal. Most of the coal in this region is exported via the Port of Newcastle.

All five of the state’s coal power stations are located in this region. The Bayswater* and Liddell power stations (owned by AGL) are located right next to each other near Muswellbrook in the Upper Hunter. The Eraring power station (owned by Origin) and Vales Point power station (owned by Delta) are located near Lake Macquarie south of Newcastle. The Mt Piper power station (owned by EnergyAustralia) is located on the western edge of this region near Lithgow.

The Southern region, which includes the Southern Coalfield, begins south of Sydney running along the coast to Wollongong before heading inland to Moss Vale then north-east to Campbelltown. This region is far smaller than the Northern region, including just four coal mines and producing 5% of the state’s coal. Most of the coal in this region is exported via Port Kembla near Wollongong.

New South Wales is the most coal-reliant state in Australia, with the most coal power station capacity generating 79% of the state’s electricity in 2020. The ageing Liddell power station is scheduled to close in 2023, which should lead to a drop in coal generation.

There are a handful of new proposed coal mines in the state, although a number of these projects are having difficulty getting developed and they are located in existing coal basins. Unlike Queensland, where there are still plans for massive coal expansion, New South Wales coal production has probably peaked. 

Like Queensland, New South Wales is a massive coal exporter and exports about 85% of its coal. But its export market is significantly more concentrated than Queensland’s. 87% of New South Wales’ coal exports go to just four countries: Japan (39%), China (22%), South Korea (13%) and Taiwan (13%). This highlights just how much the future of the state’s coal industry is dependent on the energy policies of just four countries.

Another crucial difference between New South Wales and Queensland is the type of coal exported. In Queensland, 72% of exported coal is metallurgical coal, which is used in steel-making. Metallurgical coal will likely take longer to phase out than thermal coal (used to make electricity). 

The opposite is true for New South Wales: 85% of the state’s coal exports are thermal coal and just 15% metallurgical coal (as of 2015-16). This means New South Wales is going to experience the brunt of the energy transition much earlier than Queensland, with the state’s major coal markets currently drawing up plans to phase out coal power stations and by extension, thermal coal imports. New South Wales needs a plan to manage the decline in thermal coal exports or the state’s economy and mining regions will be hit very hard over the next decade.

There is no one-stop shop for information on the New South Wales coal industry but the best places to start are Coal Services Statistics and Coal in NSW Industry Sectors (note that the data in this source is from 2015-16).

*I took a photo of the Bayswater power station from the top of the Liddell power station, which you can see on the homepage of this website.


US election recap: what happened and what it means for 2024

After an extended post-lockdown hiatus, the ICENERVESHATTER blog is back for 2021 with a new post every few weeks.

With Joe Biden’s inauguration a few days away and the Georgia Senate elections wrapped up, it is time to look back at the results of the United States presidential election and identify some key trends to help understand the future of the American electoral map. This long read will (probably) mark the end of this blog’s focus on United States politics for the foreseeable future, as I move on to other topics.

A quick recap

Joe Biden won the electoral college, with 306 votes to Donald Trump’s 232 – the exact reverse of the 2016 result. Biden won five states that Hillary Clinton couldn’t (Michigan, Pennsylvania, Wisconsin, Arizona and Georgia), as well as one of Nebraska’s electoral votes (see this post for an explanation of the quirks of the US electoral system).

Biden won the popular vote by 4.4%, with 51.3% of total votes cast. That is the second largest winning margin since 1996 and the second largest primary vote percentage since 1988 (both beaten only by Obama’s 2008 victory).

Turnout was very high by American standards: 159 million people voted, equivalent to 67% of the voting-eligible population. That’s the highest percentage in over 100 years.

Democrats also kept control of the House of Representatives (albeit with a reduced majority) and won control of the Senate by the barest of margins.

What does this result reveal about the changing electoral map? Let’s go through some of the key lessons and lingering questions.

Virginia, Colorado and New Mexico are out of reach for Republicans

All three of these states used to be swing states and were won by George W Bush as recently as 2004. But since then, all three have drifted to the Democrats. It was an open question after 2016 as to whether these states could still be called swing states. The 2020 results seem to confirm that they are not: Biden won Virginia by 10.1%, New Mexico by 10.8% and Colorado by 13.5%. Maine is also increasingly uncompetitive with Biden winning statewide by 9.1% (although Trump won one of the two district electoral votes).

Ohio and Iowa look increasingly out of reach for Democrats

Despite the polls indicating a close result, the margins in Ohio and Iowa hardly budged in 2020. Trump won Iowa by 8.2% (a 1.2% swing to Biden) and Ohio by 8% (a 0.1% swing to Biden). Both states voted for Obama in 2008 and 2012 – by a large margin in the case of Iowa – but both swung strongly to Trump in 2016, who won them both. The question after 2016: was this a one-off or the new normal? With the 2020 results in, the latter now seems likely. Expect Democrats to devote far fewer resources to these states in the future.

Nevada is very winnable for Republicans

Biden won all the states that Clinton won in 2016 and he won by margins greater than 5% in all these states, except one. That exception is Nevada. Biden won Nevada by 2.4%, exactly the same margin as Clinton’s in 2016. That makes it one of just eight states that did not swing to Biden and one of just two swing states (the second state I’ll come to shortly). 

This will place Nevada firmly on the Republicans radar in 2024. 

Democrats can still win mid-western states

After Trump swept all but two mid-western states in 2016, there was much discussion as to whether Democrats could ever win these states back. While Ohio and Iowa now seem out of reach, Biden successfully regained Michigan, Pennsylvania and Wisconsin, where Trump had his three narrowest wins in 2016. For the second election in a row, Wisconsin was the tipping point state, delivering the decisive electoral votes for Biden to win. Clearly Democrats need not give up on the mid-west.

Democrats can win traditionally Republican states in the south and south-west

After much trying, Democrats finally won two Republican-leaning swing states in the south: Georgia and Arizona. These states had been trending Democratic for the last decade but in 2020 the long-awaited breakthrough finally happened.

But does this represent the start of a new era for these states or a one-off result? Biden’s winning margins in these states were extremely close – 0.3% in Arizona and 0.2% in Georgia. As Perry Bacon Jr has pointed out, there are some similarities to Obama’s breakthrough in North Carolina in 2008, which was heralded as the start of a new era for that state. Yet since then, Republicans have won North Carolina at every election – albeit by narrow margins. Time will tell if Georgia and Arizona remain competitive but slightly Republican-leaning states like North Carolina or whether they follow in the footsteps of now-strongly Democratic Virginia and Colorado.

Ohio and Florida are no longer bellwethers

Ohio had been won by the winning candidate at every presidential election since 1964. Florida’s record was pretty good too. But in 2020, both states backed the losing candidate. Both states, it seems, are no longer vital to becoming president.

Which state will become the new bellwether? Maybe it will be Nevada. With the exception of 2016, Nevada has been won by the winning candidate at every other election since 1980.

Democrats best chances of further gains are in the south……

In 2020, Trump won just three states by margins under 8%: North Carolina, Florida and Texas. If you know your American geography, you will have noticed that all three of these states are in the South. If Democrats are to further improve their position in the electoral college, it is these three large states that will help them.

…..And Texas looms as a massive prize

There has been much hype about Texas going Democratic over the last couple of years. But Texas seems to have been perceived as something of disappointment for Democrats in 2020. The polls indicated that the result would be a bit closer than it actually was. 

But perhaps these expectations were unreasonable. Texas did swing solidly to Democrats in 2020 – the Republican margin was cut by 3.4% to 5.6%. That doesn’t quite make Texas a close state. But recall that in 2012 Obama lost the state by 15.8%. The transformation of Texas over the last decade has been extraordinary and the transformation continued in 2020. On current trends, Democrats will get very close to winning in 2024 and 2028.

And Texas is a huge prize. Flipping Texas would go some way towards eliminating the structural disadvantage Democrats have in the electoral college. Texas currently has 38 electoral votes. For comparison, the three crucial mid-western states of Pennsylvania, Michigan and Wisconsin have 46 votes combined. And the upcoming reallocation of electoral votes based on the latest census could make the comparison even better: Texas could be worth 40 votes while the mid-western trio shrinks to 44 votes at the 2024 election (if current predictions are correct).

For how much longer will Florida remain a swing state?

More than anything else, I think the results in Florida significantly affected perceptions of Biden’s victory on election day (mine included). Florida was the fastest state to count all its votes in 2020 and it has a long history as a bellwether: if you win Florida, you become president. Once Florida was called for Trump, that led to media narratives developing that the election was very close, despite the fact that many other swing states would not count all their votes for several days, and Biden ended up winning almost all the rest of the swing states.

Florida is still a swing state. Trump only won it by 3.4%, continuing its uninterrupted run of not being won by any party by more than 5 points since 2000. But for how much longer will Florida remain a swing state? The trend is unmistakable. Florida was one of just three states to swing to Trump by more than 1% in 2020. This built upon successive swings to Republicans over the last three presidential elections.

As mentioned above, because Florida may no longer be a bellwether state, it may be harder to draw conclusions about what results in Florida mean for the rest of the country, both before and on election night.

If Florida is no longer vital to the presidency, a lot less weight may be placed on it in the future.

What about Minnesota?

Clinton won Minnesota narrowly in 2016. Along with strongly Democratic Illinois, it was one of just two mid-western states that she won. In 2020, Biden regained the mid-western states of Michigan, Pennsylvania and Wisconsin, with modest swings towards him of between 1.4% and 3%.

But Minnesota swung towards Democrats by 5.6%, with Biden’s margin increasing to 7.1%. Why has Minnesota diverged from all the other mid-western states and swung so strongly towards Biden? I have no idea. But I hope someone is looking into it.

What will the Republicans do next?

Usually after losing elections, political parties re-evaluate their strategy, messaging, policies and leadership to put themselves in a better position to win the next election. Will the Republican party go through this process? Losing both houses of parliament and the presidency over a two year period is clear proof that your electoral strategy is not working. But are the Republicans capable of this sort of self-reflection? Or will they maintain that the election was stolen and continue on their current path?

And how good was the polling?

Nate Silver has written a good overview of the 2020 election polling here, so I’ll refer you there for detailed analysis.

But briefly, there were no big polling surprises in 2020. 538’s presidential forecast on the day before the election predicted the correct winner in all but two states. One exception was North Carolina, which was expected to be very close but Biden was slightly favoured (by 1.7 points); Trump ended up winning by 1.4%. Florida was the only state that could be considered an upset: Biden was forecast to win by 2.5% and Trump ended up winning by 3.4%.

A better way to compare the accuracy of polling is not to see if it predicts the correct winner (although that is generally how the media and public judge polling) but to see how close it came to the actual margin. 

(For example, a poll that shows Biden winning by 0.5%, where Trump ends up winning by 0.5% would only be 1% off but predict the wrong winner. Whereas a poll that shows Biden winning by 0.5% when he actually wins by 10% would by a massive 9.5% off, even though it got the correct winner.)

So looking at the polling (using 538’s final presidential forecast) in the states that finished with margins of under 10%:

  • Two states had polling errors within two points of the actual margin (Georgia, Minnesota). Polling in these states was excellent.
  • Seven states had polling errors between two and five points of the actual margin (Arizona, Maine, Nevada, New Hampshire, North Carolina, Pennsylvania, Texas). Polling in these states could be improved but was within the margin of error. 
  • Five states had polling errors more than five points off the actual margin (Florida, Iowa, Michigan, Ohio, Wisconsin). Polling in these states was pretty poor. One of these states was Florida, which has its own idiosyncrasies. But you may notice that the other four states are all in the mid-west. I suspect that’s no coincidence.

Is there a certain type of Trump supporter in the mid-west who is particularly hard to poll? One theory is that voters in these states, with a large cohort of non-university educated white voters, are getting harder to reach; another is that there may have been an oversampling of university-educated white voters in 2020 polls, with so many office workers working from home, and presumably more likely to pick up the phone.

Was the polling error for this election unique to the unusual circumstances of 2020? Or is it related to the similar but smaller polling error of 2016? It is concerning that the polling errors in 2016 and 2020 both overestimated the Democratic vote and underestimated the Republican vote. That could be a coincidence. If not, then pollsters will need to change their methodologies to prevent it happening again.

A final thought before I bring this long post to an end: It has been pointed out by others that polling for special elections over the last three years, the 2018 mid-term elections and the Georgia Senate runoffs last week was all pretty accurate. The polling error seems unique to 2016 and 2020. Is there something about Trump that completely changes voter behaviour in ways that we still can’t account for? And if so, does that mean that when he is not on the ballot the problem just goes away?

Other good US election analysis:

How the 2020 Election Changed the Electoral Map

Democrats Hope Georgia Will Become the Next Virginia, But It Could End Up Becoming the Next North Carolina

The Polls Weren’t Great. But That’s Pretty Normal

The Battle for the Battleground States


10 days to go: A final preview of the United States election

6 weeks ago, I previewed the United States presidential election (part 1 here and part 2 here). Now with 10 days to go (polls begin closing on the morning of Wednesday 5th November Australian time), I will provide a final overview of where the race for the presidency stands. I will also provide a brief update on the Senate election, which I previewed four weeks ago.

As per my previous posts, I will be using FiveThirtyEight’s weighted polling averages for the presidential election and FiveThirtyEight’s Senate forecast for the Senate, as of Saturday 24th October.


In my last post, I noted that there are three ways Trump could still win despite being behind in the polls: a narrowing in the polls, a large polling error or undemocratic interference in the electoral process.

Now there are only 10 days to go. Has Trump narrowed Biden’s lead in the polls enough to put him within striking distance?

The answer is a resounding no. Far from narrowing, Biden’s lead has actually increased. 6 weeks ago, Biden led the national polls by 7.6 points. Today he leads by 9.7 points. If Biden wins by this margin, it would be the biggest popular vote victory of any presidential candidate since Ronald Reagan’s re-election in 1984. 

There is precious little time left for Trump to turn this lead around and only about 5% of voters are still unsure who to vote for. And with around 48 million people having already voted (that’s about one third of all votes cast in 2016), the pool of potential voters for Trump is shrinking fast.

Of course, the national popular vote does not actually determine the winner of the election – the electoral college does (read my previous post on this subject for an explanation of the process). So let’s dive into the latest state polling to get a better idea of the likely result in the electoral college.

Note: in my previous post on this subject, out of an abundance of caution, I included a section on swing states where Biden led by 10-15 points, including states like Virginia and Colorado. With just 10 days to go, it is clear that these states are not going to be competitive this year, so I will not be including them in this post.

Swing states where Biden leads by 5-10 points:

  • Minnesota: 7.9 points to Biden
  • Michigan: 7.6 points to Biden
  • Nevada: 6.5 points to Biden
  • Wisconsin: 6.5 points to Biden
  • Pennsylvania: 6.1 points to Biden

The first thing you will notice about these states is how similar the margins are compared to 6 weeks ago: the polling has been remarkably stable. Biden’s lead in the Clinton-won states of Minnesota and Nevada is slightly up, while Biden’s leads in the Trump-won states of Michigan and Wisconsin are exactly the same as they were six weeks ago, to the decimal point. The crucial state of Pennsylvania has improved for Biden by 1 point. Biden leads in these five states by between 6 and 8 points, a very solid lead (a polling error may endanger these leads but it would need to be a pretty large polling error). Simply winning these states is enough for Biden to be elected president. 

There are two other states where Biden was leading by 5-10 points six weeks ago: New Hampshire and Arizona. Biden’s lead in New Hampshire (which Clinton won in 2016) has increased by 3.5 points to 11.4 points, putting the state beyond Trump’s reach. Arizona we will come to momentarily.

Swing states where Biden or Trump lead by less than 5 points:

  • Florida: 3.3 points to Biden
  • Arizona: 3.2 points to Biden 
  • North Carolina: 2.9 points to Biden
  • Iowa: 1.3 points to Biden
  • Georgia: 1 point to Biden
  • Texas: 0.5 points to Trump
  • Ohio: 1 point to Trump

This group of close states has also been fairly stable but they have generally moved slightly in the direction of Biden over the last six weeks (reminder: Trump won all these states in 2016). Georgia and Iowa have moved towards Biden by 2.5 points and 2.9 points respectively, giving Biden the lead in both states (but they are still very close).

Biden’s lead in North Carolina has increased by 1.5 points to 2.9 points, while Florida has also shifted slightly in Biden’s direction. It is a good sign for Biden that his leads in both these states are up to around 3 points.

Trump still leads by the narrowest of margins in Ohio (by 1 point) and Texas (0.5 points), with both states hardly budging since six weeks ago. 

Arizona is the outlier here: it is the only one of these states that has swung towards Trump. Biden had lead by 5 points in Arizona six weeks ago but that lead has shrunk by 1.8 points to 3.2 points. I cannot offer an explanation for this, although it is notable that the southern states of Florida, Arizona and North Carolina are all now polling within half a point of each other.

So let’s briefly summarise:

  • Biden is leading in the mid-western states of Minnesota, Michigan, Wisconsin and Pennsylvania by 6-8 points. He is also leading in south-western Nevada by 6.5 points.
  • He is leading in the southern states of Florida, Arizona and North Carolina by around 3 points.
  • Iowa and Ohio in the mid-west, and Georgia and Texas in the south could go either way.

If Biden wins all of these states, this would result in Biden winning the electoral college 413 to Trump’s 125*, a margin of 288 and the biggest win in the electoral college since George HW Bush’s election in 1988 (he won by 315).

States where Trump leads by 5-10 points:

  • Alaska: 5.7 points to Trump
  • Missouri: 6.5 points to Trump
  • Nebraska*: 7 points to Trump
  • South Carolina: 7.1 points to Trump
  • Montana: 7.7 points to Trump
  • Kansas: 8.9 points to Trump

There has been hardly any movement in these states over the last six weeks (excluding Alaska and Nebraska, which did not have polling averages when I wrote my previous post). Kansas and Montana have weakened slightly for Trump, whereas South Carolina has slightly strengthened. Missouri has not budged.

Trump is favoured to win these states but I include them because just as a significant polling error could cost Biden his lead in states like Pennsylvania and Wisconsin, a significant polling error could cost Trump his leads in these states. 

Summarising the presidential race

Trump is very far behind in the polls. Before turning to the Senate, let’s return to the three ways Trump could win.

A narrowing of the polls? As discussed earlier in this post, Trump has been losing ground in the polls over the last six weeks. As every day passes, and millions more Americans vote, the probability of the polls tightening shrinks further. Is it possible Trump makes up lots of ground in the last 10 days of the campaign? Yes. Is it likely? Nope.

A major polling error? Biden’s national vote lead of 9.7 points is so large, Trump needs a polling error to be more than a large error; he needs it to be a gigantic, once-in-a-generation polling screw-up. The state polls are a bit closer, with Biden leading in the crucial tipping point states by 6-8 points. A large polling error could cost Biden his leads in these states but the polling error would have to be very widespread, costing Biden his leads in not just Michigan, Wisconsin and Pennsylvania but also Florida, Arizona, North Carolina, Iowa and Georgia. Possible? Yes. Likely? Probably not. And remember, a polling error is as likely to benefit Biden as it is to benefit Trump.

Election shenanigans? This continues to be a bit of an unpredictable wildcard but again, if the polls are right, Biden’s polling lead is so large that literally millions of votes would have to be tossed out to change the result. That would require a very extensive and sophisticated rigging operation. Possible? Yes. Likely? Who knows, but my gut says that it is very unlikely.

The Senate race is still close but Democrats are edging ahead

Now for a brief Senate update. The race for the Senate has remained tight and fairly stable since my last post on this subject four weeks ago but a number of key races have shifted slightly in the direction of Democrats, improving their odds of gaining a Senate majority.

(Reminder: for the Senate, I use FiveThirtyEight’s probabilistic forecast of the Senate, not polling averages. Eg. Republicans have a 78% chance of winning Alabama; 78% is the probability that a party will win, not the size of the winning margin).

Republicans are still favoured to win Alabama off Democrats, with a 78% chance (up 1%). Republicans also have a 21% chance of winning Michigan from the Democrats (down 1%). These races have hardly changed since my last post.

Democrats are still favourites to win four Republican Senate states, with their lead in Colorado in particular increasing over the last few weeks. Democrats have an 83% of winning Colorado (up 12%), a 78% of winning Arizona (up 1%), a 64% chance of winning North Carolina (up 1%) and a 62% chance of winning Maine (up 3%).

Democrats have also pulled ahead in two Republican-held seats that may be crucial to delivering the Democrats a Senate majority: Iowa and Georgia. Iowa continues to be a tossup race but Democrats have their noses in front with a 56% chance (up 9%). The most substantial change in any Senate race has been in the Georgia special election, with Democrats chances improving significantly from 19% to 54%. I recommend reading this article if you want to understand why. In short, the Georgia special election is a case study of the dangers of pandering to your base when the vast majority of voters are not in your base.

Democrats are underdogs in another six races currently held by Republicans but they have improved their chances in five of these races: they have a 32% chance in Montana (down 2%), a 29% chance in the regular Georgia election (up 3%), a 25% chance in Kansas (up 5%), a 23% chance in South Carolina (up 5%) and a 20% chance in Alaska (up 6%).

Democrats are a more remote chance in Texas (13%) and Mississippi (11%).

To sum up the Senate: Democrats will probably lose Alabama and gain Colorado, Arizona, North Carolina and Maine. That results in a Senate split 50-50. If Democrats also gain Iowa and the Georgia special seat, where they are narrowly ahead, that gets them to 52-48, a solid enough majority to enable Biden to govern effectively.

So Democrats would be feeling good about the Senate – but it is not a done deal and there is little room for error. 

*Assuming Biden wins all electoral votes in Maine and one electoral vote in Nebraska, which do not award all their electoral votes to the statewide winner.


The race for the United States Senate

The untimely death of progressive Supreme Court justice Ruth Bader Ginsburg has brought into focus the power of the United States Senate and how critical it has been for Republicans to maintain conservative control of the powerful court.

In my last two posts, I previewed the United States presidential election where Joe Biden is the favourite to beat Donald Trump. But for Biden to achieve any of the reforms he has committed to implementing, he will also need Democrats to gain control of Congress. Congress is made up of the House of Representatives and the Senate. 

Democrats took control of the House of Representatives in 2018 and they should have comparatively little difficulty holding it in 2020. The Senate, however, is a different matter. This blog post will preview the 2020 Senate election.


Compared to the electoral college (the system used to elect a president), the Senate is simple. There are 100 seats, with each of the 50 states electing two senators. Around one third of senators are up for election every two years with each senator serving a six-year term. Winning a Senate seat requires winning a statewide vote. Usually just one senator in each state faces election at the same time (eg. one senator from Michigan faced an election in 2018 and the second senator faces election this year). With 33-34 senators up for election at any one time, this means not every state will have a senate election every two years.

This means that a strong performance nationally does not always translate into gains in the Senate. This can cause some seemingly counter-intuitive impacts on the make-up of the Senate.

Here are some examples. In 2018, Democrats had a massive win in the popular vote, beating Republicans in the House of Representatives by 9 points. Yet Democrats actually lost Senate seats that year (going from 49 to 47). 

And in 2016, Republicans narrowly won a swath of Senate seats in Pennsylvania, Wisconsin and Florida. Democrats trounced Republicans in Pennsylvania and Wisconsin in 2018 and may do so again in 2020 – but this will not win them any Senate seats because 2016 senators will not face election until 2022!

Because of these idiosyncrasies, Republicans are on the defensive in many Senate seats this year but it will take a very strong performance from Democrats to gain control. Biden may win the popular vote for president by 5-8 points but Democrats may only be able to eke out a Senate majority of just one or two seats. A smaller Biden win could leave the Senate in the hands of Republicans. 

Going in to 2020, the Republicans have a narrow majority: 53 seats to Democrats’ 47.

There are a couple of seats this year that Republicans could gain and a much larger number of seats that Democrats are targeting. Let’s look at each of these groups in turn.

Rather than using a polling average like I did for my previous post on the presidential election, today I will use FiveThirtyEight’s Senate forecast, which incorporates polling averages and a range of other inputs to forecast likely results. I will include all states where either party has a greater than 10% chance of winning a Senate seat off the other party.

Please note that the numbers below are probabilities that a party will win a seat, not a projection of the winning margin (eg. Republicans have a 77% of winning Alabama. They will not win the vote in Alabama by 77%).

Potential Republican gains

  • Alabama – Republicans have a 77% chance of winning
  • Michigan – Republicans have a 22% chance of winning

There are very few pick-up opportunities for Republicans this year: senators facing election in 2020 were first elected in 2014, the year Republicans swept many competitive Senate seats and took control of the Senate from Democrats, who had held a Senate majority from 2006 to 2014.

Republicans are favourites to win Alabama. Alabama is one of the most strongly Republican and pro-Trump states in the country so it will be difficult for Democratic Senator Doug Jones to hold on to this seat. 

You may be thinking, how did Democrats win this seat in the first place? Well, a special election (similar to a by-election in Australian terminology) was held in 2017 to fill this vacant seat and Republicans selected a candidate who faced accusations of child sexual abuse. Not great for winning elections, huh? Alabama may be one of the most pro-Republican states in the country but when faced with the choice of a Republican accused of abusing children and a Democrat not accused of abusing children, it is hardly surprising that voters chose the latter. 

This year, Doug Jones will face a far less problematic Republican candidate and Republicans are favoured to win back the seat.

The only other serious pick-up opportunity for Republicans is in Michigan but with Joe Biden holding a large lead there in the presidential race, this seems unlikely. But if Trump does much better than expected, it is certainly possible for Republicans to win this seat.

Potential Democratic gains

  • Arizona – Democrats have a 77% chance of winning
  • Colorado – Democrats have a 71% chance of winning
  • North Carolina – Democrats have a 63% chance of winning
  • Maine – Democrats have a 59% chance of winning
  • Iowa – Democrats have a 47% chance of winning
  • Montana – Democrats have a 34% chance of winning
  • Georgia – Democrats have a 26% chance of winning
  • Kansas – Democrats have a 20% chance of winning
  • Georgia special election – Democrats have a 19% chance of winning
  • South Carolina – Democrats have a 18% chance of winning
  • Alaska – Democrats have a 14% chance of winning
  • Texas – Democrats have a 13% chance of winning

As you can see, there are way more potential gains for Democrats this year than Republicans. Democrats are currently favoured to win three Senate seats off Republicans, with another two seats too close to call. Democrats are also competitive in another seven seats.

The most likely pick-ups for Democrats are in Arizona and Colorado. Colorado is polling very strongly for Biden and he has consistently led in Arizona (albeit by a much smaller margin). Despite their clear vulnerabilities, the incumbent Republican senators in these states have strongly backed Trump at every turn, making their re-election that much harder. Democrats would be feeling very confident of picking up these two seats.

Democrats would also be feeling confident about North Carolina. Biden is only narrowly ahead there in the presidential race but the incumbent Republican senator has been polling pretty poorly. It’s still a bit early, but things are looking good for Democrats here.

Maine and Iowa loom as crucial states that will go a long way to determining which party controls the Senate. 

Democrats are currently polling well in Maine but there a number of factors that make this race more uncertain. A lot of this comes down to the long-serving incumbent Republican Senator Susan Collins. She has been a bit of a maverick in Maine and historically has had no difficulty winning by huge margins despite Maine being a generally Democratic leaning-state. She has also voted against Trump’s policies more than any other Republican senator. Will Collins be able to maintain support from moderate Democratic voters who have crossed party lines to vote for her in previous elections? The polling suggests not but the race may be close.

Iowa is currently the most competitive Senate race. The FiveThirtyEight forecast shows the race as a virtual tie, although the Democratic candidate has been pulling ahead in recent polls. This race could go either way but this seat will be critical for Democrats to reach 51 seats if they lose Alabama to Republicans. 

Beyond Iowa, there are seven other states where FiveThirtyEight currently gives Democrats a greater than 10% chance of winning. The most likely of these are Montana, Kansas, the two Georgia seats and South Carolina, where Republicans lead by very narrow margins in polls. Alaska and Texas are also possible gains but probably out of reach unless the Democratic candidates do better than the polls currently indicate.

Summing up

If we assume Republicans win Alabama and Democrats win Arizona, Colorado, North Carolina and Maine, this would result in a Senate split between each party of exactly 50-50.

If the Senate is deadlocked with each party voting 50-50 on a law, the Vice-President can be the tie-breaker. So if Biden is elected President, a 50-50 Senate is a de facto Democratic majority and vice versa.

But such a narrow majority would be a disaster for progressives because there is a very wide range of ideological views amongst Senate Democrats and one senator in particular looms as being especially problematic: Democratic Senator Joe Manchin of West Virginia. Despite being a member of the Democratic Party, Manchin is a conservative and if his vote is needed to pass laws, very little progressive legislation will pass the Senate and it is much less likely that there will be reform of the Supreme Court.

To bypass Manchin, Democrats actually need to win a minimum of 51 seats. That could mean Iowa’s Senate seat will be critical.

So while everyone will rightly be watching the presidential race, keep your eye on some Senate seats as well. Particularly Maine and Iowa.

The data used in this post was sourced from FiveThirtyEight’s Senate forecast on Saturday 26 September. You can see the forecast here.

For a list of Senate polls by state, see here


With seven weeks to go, the US election is Biden’s to lose: Part 2

This post is a continuation of yesterday’s post, providing an overview of the state of play in the 2020 presidential election. 

Swing states where Biden or Trump lead by less than 5 points:

  • Florida: 2.7 points to Biden
  • North Carolina: 1.4 points to Biden
  • Texas: 0.8 points to Trump
  • Ohio: 0.9 points to Trump
  • Georgia: 1.5 points to Trump
  • Iowa: 1.6 points to Trump

These are six states that could go either way in 2020. All were won by Trump in 2016. 

Ohio and Florida are swing states par excellence. Results in Florida are reliably close and this year looks like no exception. Since 2000, Florida has not been won by more than 5 points by any party. Biden has consistently led in state polling since April but the gap has narrowed in recent weeks.

Ohio is also a consistently competitive state, although it swung strongly behind Trump in 2016. It has a remarkable history as a bellwether: it has been won by the winning candidate at every presidential election since 1964, and has sided with the winning candidate all but once since 1944! Ohio’s status as a bellwether may come to end this year: despite Biden’s national lead of over 7 points, Trump leads in Ohio by just under 1 point.

Iowa is another of those mid-western states (like Pennsylvania, Michigan, Wisconsin and Ohio) that has been shifting to the right over the last decade. Obama won Iowa by 10 points in 2008; Trump won it by 9 points in 2016. 

In terms of its political lean, Iowa has often sat in the middle of the Midwestern states over the last couple of decades: it has not been quite as strong for Democrats as Pennsylvania, Michigan and Wisconsin but it has been consistently more Democratic than Ohio. However, Iowa’s shift to the right in 2016 was large and even now Trump leads in the state by 1.6 points. 

North Carolina and Georgia are Republican leaning swing states. North Carolina has only been won by Democrats once since 1980 (by Obama in 2008) but results are often close. Biden is currently leading by a bit over 1 point.

Georgia has not been won by Democrats since Bill Clinton’s victory in 1992 but demographic changes have seen Georgia slowly shuffle to the left – the 2018 election for state governor was very close. Part of the reason Democrats have not yet had a breakthrough in the state is because Georgia is one of the least elastic states in the country – that means that there are very few swing voters. The state’s troubled history of racialised politics means that many voters are either rusted on Democrats or Republicans. In 2020, it looks like Biden will have a decent chance of picking it up – Trump is leading in the polls by just one and a half points.   

Finally, Texas looms as one of the most interesting contests this election. Trump currently leads in the state polls by less than one point as demographic changes shift this formerly staunch Republican state to the left. Although Texas is unlikely to be decisive in determining the election winner (Biden will be elected if he wins just the four states where he is already leading by more than 5 points), a Biden win in Texas would have a seismic impact on the electoral map. Texas has not been won by any Democratic nominee since 1980, and it is the only large state that Republicans can consistently rely on winning. Without Texas, there are far fewer paths to a majority for a Republican nominee.

All in all, these six states are unlikely to be the tipping point that determines whether Biden or Trump wins the election. But they will determine whether Biden wins narrowly or by a landslide. 

If Biden only gains the four states where he leads in the polls by more than five points (Michigan, Wisconsin, Pennsylvania and Arizona), he will win the electoral college 289 to Trump’s 247. But if he also wins these six states, Biden’s win becomes a massive landslide: 409 to 125. That’s because these states are big and have a lot of electoral votes – with the exception of Iowa, the other five states in this category are all in the top ten most populous states in the country. 

States where Trump leads by 5-10 points:

  • Missouri: 6.5 points to Trump
  • South Carolina: 6.7 points to Trump
  • Montana: 8.4 points to Trump
  • Kansas: 9.2 points to Trump
  • Alaska: N/A

I will briefly mention the states where Trump is leading by five to ten points: Missouri, South Carolina, Montana, Kansas and Alaska. I have intentionally not called these states ‘swing states’ because all traditionally vote Republican.

In all likelihood, Biden will probably not win these states. But the fact that Trump’s leads in these states are so weak is a clear sign of the trouble he is in. Trump won South Carolina and Alaska by over 14 points in 2016, Missouri by over 18 points and Montana and Kansas by over 20 points.

Not even in Obama’s landslide victory in 2008 did Democrats win any of these states (although he came extremely close to winning Missouri). South Carolina, Kansas and Alaska have been won by Republicans at every election for the last forty years, and Montana has been won by Democrats just once (in 1992). These are most definitely not swing states. 

A Biden victory in these states may be unlikely, but just as a major polling error could endanger Biden’s leads in Pennsylvania, Michigan, Wisconsin and Arizona, so too a polling error could endanger Trump’s leads in any of these five states.

There have not been enough polls of Alaska for FiveThirtyEight to produce a polling average but the few polls that do exist indicate a competitive Republican-leaning race. 

Why Trump could still win

The polls paint a good picture for Biden’s chances. But he definitely doesn’t have it locked up. There are still at least three reasons that Trump could win:

  • A narrowing of the polls. They say a week is a long time in politics. Well, we are still seven weeks from the election, so there is still time for Trump to reduce Biden’s leads in key swing states. Nonetheless, the stability of Biden’s polling lead, the relative lack of undecided voters this year and Trump’s apparent inability to course correct for a sustained period of time means I am pretty skeptical this will eventuate. Oh, and Americans have already started voting.
  • A sizeable polling error. Polls are not perfect – they are a snapshot of a group of voters at a particular point in time. Polling is a very useful tool but it is perfectly normal for polls to be a few points off (and even more on occasion). This means that it is possible Trump is doing better than the polls suggest. But this cuts both ways – a polling error is just as likely to benefit Biden as it is to benefit Trump. The polls may have had a 3.2 point bias to Democrats in 2016 but they had a 2.5 point bias to Republicans in 2012 (for a more detailed explanation of polling error, see here).
  • Election shenanigans. US elections are often accompanied by minor and occasionally major scandals, with issues around ballot admissibility and election administration par for the course in a country where one major party knows it can only win elections by excluding some voters. But this year, with so many voters voting by mail, with the possibility that the election winner will not be known on the night of the election and with President Trump explicitly suggesting that he may not respect the election result, there is more uncertainty than usual this year. This is the real wildcard.

So while Trump’s re-election remains a distinct possibility, Biden is in the box seat. Biden’s polling lead has been large and stable for so long, I find it hard to see Trump winning unless something changes soon. Time is fast running out for Trump to bridge the gap.

All poll numbers used in this post were sourced from FiveThirtyEight’s presidential polling averages on Sunday 13thSeptember.

For an overview of each state’s voting history in presidential elections, see here.

To see an interactive map of which states Biden and Trump need to win the presidency, check out

For a cool visualisation of how the partisanship of swing states has changed over the last 20 years, see this 538 article.


With seven weeks to go, the US election is Biden’s to lose: Part 1

In just over seven weeks on November 3, the United States will hold its presidential election, alongside an election for the House of Representatives and one third of the Senate. This post will analyse the presidential polls and what they can tell us about Donald Trump and Joe Biden’s prospects for victory.

Unlike in Australia over the last year, there are new polls almost every day in the United States and there are a number of poll aggregators, including FiveThirtyEight’s weighted polling average. FiveThirtyEights’s polling average will be the main source of polling information for this post.

The national picture

As of Sunday, Biden leads FiveThirtyEight’s polling average by 7.6 points. This lead has been extremely steady all year. In fact, since Biden became the de facto Democratic nominee for president after Bernie Sanders suspended his campaign in April, Biden’s lead in the polls over Trump has never dropped below five points. 

Since June 5th, Biden’s lead over Trump has not varied by more than a few points, staying above 7 points and below 10 points. That is a very resilient lead and is particularly impressive considering all that has happened in the US over the last five months (read here in case you’ve forgotten).

If Biden wins the popular vote count on election day by his current 7.6 point margin, he will almost certainly be elected president in November.

But, as anyone who followed the 2016 election will remember, winning the popular vote does not guarantee an election victory. Hillary Clinton won the popular vote by 2.1% in 2016, yet failed to win the presidency.

US presidential elections are not determined by winning the national popular vote. Victory requires winning a majority of delegates in the electoral college. Each state is accorded electoral college delegates (from here on called ‘electoral votes’) roughly based on their population, with each state awarded a minimum of three votes (it’s slightly more complicated than that but that’s the short explanation). The most populous state, California, has 55 electoral votes. The District of Columbia, which is not a state, is also awarded three electoral votes. 

There are 538 electoral votes in total, so getting a majority requires winning 270 electoral votes.

In 48 states and DC, electoral votes are awarded to the candidate who wins the most votes statewide. In two states, Maine and Nebraska, delegates are awarded a bit differently. Maine awards two electoral votes to the statewide winner and two votes to the winner in each congressional district. Nebraska awards two votes to the statewide winner and three votes to the winner in each district. In 2020, only one electoral vote is likely to be competitive in each of these two states, so I won’t say anything more about them in this post.

The majority of states (and therefore electoral votes) are not competitive at most elections. For example, there are 14 states that have voted Democratic at every presidential election since 1992 and 13 states that have voted Republican at every election since 1980. A further eight states have voted Republican at every election since 2000. The vast majority of these 35 states are not expected to be competitive this year (with a few exceptions that will be discussed below).

So to summarise, becoming president requires a candidate to gain a majority of electoral votes by winning in competitive states.

With the explanation done, let’s turn to the state polls.

Biden is currently leading in the polls in every state that Clinton won in 2016, plus six states that Trump won in 2016.

At the moment, the traditionally competitive states (called ‘swing states’) can be divided into three general groups: 

  • Swing states where Biden or Trump lead by less than five points. These states are very competitive.
  • Swing states where Biden leads by between five and ten points. Biden is the favourite in these states, but they remain competitive.
  • Swing states where Biden leads by more than ten points. These states are unlikely to be competitive this year.

Let’s tackle each of these groups in reverse order.

I will also briefly look at the states where Trump is leading by between five and ten points.

Swing states where Biden leads by more than 10 points:

  • New Mexico: 13 points
  • Maine*: 11 points
  • Virginia: 10.8 points
  • Colorado: 10.6 points

New Mexico, Virginia and Colorado all voted for Clinton in 2016. These are states that have traditionally been competitive swing states but have been trending more Democratic in recent years. Neither Colorado nor Virginia has been won by a Republican since 2004 and New Mexico has only gone Republican once since 1992. Nonetheless, when Republicans are doing well, they can expect to be competitive in these states. 

Clearly that is not the case this year. With Trump underwater nationally by more than 7 points, these states are highly likely to be won by Biden.

Swing states where Biden leads by 5 to 10 points:

  • Michigan: 7.6 points
  • Minnesota: 7.4 points
  • New Hampshire: 6.9 points
  • Wisconsin: 6.5 points
  • Nevada: 5.9 points
  • Pennsylvania: 5.1 points
  • Arizona: 5 points

These are seven states where Biden has a decent but not overwhelming lead. Three of these are Democratic leaning swing states that Clinton won in 2016: Minnesota, New Hampshire and Nevada.

Since 1992, New Hampshire has only backed a Republican once and Nevada twice. Minnesota has historically been a safe Democratic state – in fact, it is the only state to have backed a Democrat at every election for the last forty years – but it has been edging to the right over the last decade. Biden’s leads in these states are pretty strong but not overwhelming. If Trump is going to gain any states that Clinton won in 2016, it will be one of these three states.

Biden is also leading in four states that Trump won in 2016: Arizona, and the infamous trio of mid-western states that unexpectedly delivered the presidency to Trump: Michigan, Wisconsin and Pennsylvania.

Until 2016, Michigan and Pennsylvania hadn’t voted for a Republican since 1988 and Wisconsin hadn’t since 1984. But they have been slowing trending to the right and Trump ended up winning all three states in 2016 – each by less than 1%. Had Clinton won these three states, she would have won enough electoral votes to become president. All three states loom as crucial pick-ups for Biden and crucial holds for Trump in 2020.

Arizona is a traditionally Republican state – it has been won by Democrats just once in the last forty years. But it has been shifting to the left over the last decade and this year it may finally cross the threshold.

If Biden can hold New Hampshire, Nevada and Minnesota, while gaining three out of four of Michigan, Wisconsin, Pennsylvania and Arizona, Biden will win the presidency. There is still enough time that Biden’s lead could narrow in all these states by election day and a sizeable polling error may mean that these states are closer than they appear. But you’d much rather be in Biden’s position in these states than Trump’s.

Part two of this post will be posted tomorrow.

All poll numbers used in this post were sourced from 538’s presidential polling averages on Sunday 13thSeptember.

For an overview of each state’s voting history in presidential elections, see here.

To see an interactive map of which states Biden and Trump need to win the presidency, check out

For a cool visualization of how the partisanship of swing states has changed over the last 20 years, see this 538 article.