The figure quoted as $7.4 trillion in annual fossil fuel subsidies is mostly not money changing hands. About $6.7 trillion of it is damage the IMF says goes uncharged, three-quarters of that from air pollution and climate, on a definition other agencies reject.

The number most often quoted for global fossil fuel subsidies is $7.4 trillion a year. Almost none of it is a cheque. In the most recent IMF working paper on the subject, published in December 2025, explicit subsidies came to $725 billion in 2024, or about 0.6 per cent of global output. The remaining $6.7 trillion is something else: the authors’ estimate of environmental damage that fossil fuel prices do not include.

Whether that second figure belongs under the word subsidy is genuinely contested, and the dispute is not a fringe one.

What the IMF is counting

The working paper, by Simon Black, Ian Parry and colleagues, assesses fuel pricing across 170 countries. Its method compares what consumers pay with what the authors calculate an efficient price would be, one that covers supply costs plus the damage the fuel causes when it is burnt.

The mechanism behind that gap is old and uncontroversial in itself. When a power station burns coal, the particulates it emits reach lungs many kilometres away, and the carbon dioxide it releases affects a climate shared by everyone. Neither cost lands on the electricity bill. Economists have called this an externality for a century, and the observation that fossil fuel prices carry one is not what anybody is arguing about.

The gap between the two prices is what the paper calls an implicit subsidy. Three-quarters of it comes from underpriced air pollution and climate change. The remainder covers other uncharged costs such as congestion and road damage, which is worth stating plainly because the total is often described as though all of it were climate related.

This is an IMF working paper. Working papers are not peer reviewed, and they carry the standard note that they represent the authors’ views rather than IMF policy.

Why other institutions put the number ten times lower

In a 2017 paper in International Environmental Agreements, Jakob Skovgaard examined how the IMF and the OECD arrive at such different totals. His conclusion was that the difference is definitional rather than empirical.

The OECD uses what Skovgaard describes as a conferred-benefits approach, derived from World Trade Organization definitions: direct transfers, and revenue a government chooses to forgo. On that basis the OECD prefers the word support to subsidies. The IMF’s post-tax approach, which folds environmental externalities into the benchmark price, Skovgaard calls a radical break with previous definitions.

The practical consequence is unusual. Under the IMF method, almost every country is a fossil fuel subsidiser, including countries that levy carbon taxes, because a carbon tax set below the estimated damage still leaves a gap. Governments such as the United Kingdom and Japan, which maintain that they do not subsidise fossil fuels, dispute the label. Skovgaard notes the figures have been correspondingly unpopular with those governments and popular with non-governmental organisations.

The scale of the gap is large. Skovgaard sets the IMF’s $5.3 trillion estimate for 2015 against roughly $550 billion on the International Energy Agency’s approach, a difference of nearly ten times for the same year and the same fuels.

The damage estimates disagree too

Underneath the dollar figure sits a health estimate, and that estimate is contested on its own terms.

A 2021 study in Environmental Research by Karn Vohra and colleagues used the GEOS-Chem chemical transport model to estimate deaths from fine particulate pollution produced by burning fossil fuels. Running the model with and without fossil fuel emissions, they attributed 10.2 million premature deaths to that source in 2012, falling to 8.7 million once China’s documented emissions reductions through 2018 were taken into account.

That is more than double the Global Burden of Disease figure, which put deaths from all fine particulate pollution at 4.2 million.

The difference is not because air quality changed. As a data review by Max Roser at Our World in Data sets out, the estimates diverge for three reasons: researchers use different concentration-response functions, they include different sources, and newer evidence has pushed the assumed harm per unit of exposure upward. Vohra and colleagues used a steeper curve drawn from a recent meta-analysis, which accounts for much of the gap.

The authors are direct about the limits. Confidence intervals are widest in the most polluted regions, where measured concentrations sit beyond the range of the epidemiological studies the curve is built from. The model cannot adjust for differences in vulnerability by income, ethnicity or existing illness.

Roser’s review makes the one point all the estimates share. They disagree about whether the annual toll is four million or nine million. None of them suggest it is small.

What the modelling says removal would do

The IMF paper models two scenarios. Removing explicit subsidies alone would cut global carbon dioxide emissions by about 6 per cent by 2035, prevent roughly 70,000 premature deaths a year, and produce a net economic benefit of around 0.5 per cent of GDP.

Pricing fuels to cover the full estimated damage produces much larger modelled effects: about 1.1 million fewer deaths a year, and a 46 per cent reduction in carbon dioxide. The authors acknowledge the political obstacles to doing so.

One finding cuts against the usual defence of fuel subsidies. They are commonly justified as support for low-income households. The paper’s analysis finds the poorest fifth of households receive about eight cents of every dollar spent on them.

What this does not settle

The physical facts are not seriously in dispute. Burning fossil fuels produces particulate pollution that shortens lives, and carbon dioxide that warms the climate. Neither cost appears in the price.

What is disputed is whether leaving a cost out of a price is the same thing as paying someone to keep that price low. That is a question about language and accounting rather than about atmospheric chemistry, and the answer determines whether the number is seven trillion dollars or half a trillion.

It matters because the two framings point at different policies. Withdrawing a payment is an act of budget restraint. Charging for damage is a new tax. Governments that would accept the first have good reason to resist a definition that quietly commits them to the second.

The mortality estimates that anchor the larger figure have moved in one direction over the past decade, and it has been upward. Whether they settle there depends on exposure research that is still being done.

Edited by Lachlan Brown