Carbon pricing exists because of a belief. The belief is that the cost of damaging the atmosphere should appear, somewhere, on a balance sheet, and that the price should be consistent enough across markets that the cost is real wherever the damage is done. If the belief holds, capital flows in the direction the price signals. Regulators can lean on it. Operators can plan against it. The price becomes the world's most important shared accounting fact.

If the price varies by a factor of fifteen hundred between two jurisdictions implementing the same kind of instrument, the belief is not a fact. It is a slogan.

The World Bank's 2026 Carbon Pricing Dashboard appears, on a first reading, to validate the belief. Ninety implemented compliance instruments. Eighty-three jurisdictions. A top-ten average of US$127.7 per tonne, comfortably in the band where most peer-reviewed estimates put the social cost of carbon. Read the dashboard that way and you see a system finally arriving where the science says it should be.

We have read it carefully, and we see something else. There is no global carbon market. There are three. They are not converging. The gap between the highest-priced and lowest-priced implemented instruments is 1,543 to one. And anyone using the headline number as a global benchmark is using a number that no instrument anywhere actually charges, that the underlying distribution does not produce, and that the system the dashboard summarises is not, on present evidence, moving toward.

That gap between the belief and the data is the story this Data Read is about, because the gap is what makes the work that depends on the benchmark unstable. Capital allocators who price carbon exposure off the global figure are pricing off a fiction. Regulators designing cross-border mechanisms are choosing which cluster to anchor to, whether they say so or not. Operators with internal carbon prices are benchmarking against a world that does not exist. The dashboard is doing what dashboards do, which is report. The reading of it is what we want to argue with.

This Data Read works from the 2026 snapshot of the World Bank Carbon Pricing Dashboard, restricted to the 85 implemented compliance instruments that publish a US-dollar price. We pulled the parsed instruments dataset, read the dashboard's About page for methodology, and ran the distribution cut ourselves. We did not consult specialists on background. The dataset, read carefully, makes its own argument.

> Figure 1: Distribution of US-dollar prices across 85 implemented compliance instruments, 2026 snapshot. Three modes are visible: a high cluster at or above US$100; a mid-tier between US$20 and US$25; a low tail at or below US$5. The space between US$25 and US$100 is sparsely populated.

> Figure 2: The seven instruments priced at or above US$100, ordered. Norway's carbon tax leads at US$169.71. Sweden, Liechtenstein, Switzerland, Denmark, and the Netherlands follow.

> Figure 3: The eighteen instruments priced at or below US$5, ordered. Poland sits at US$0.11. Ukraine, Indonesia, and Kazakhstan follow under US$1. The tail runs through a further fourteen jurisdictions.

Here is what the data shows, in five moves.

1. The headline average is the average of three different price worlds, none of which is the headline. 2. Convergence exists, but it is the convergence of a club watching itself, not a system integrating. 3. The bottom of the distribution is not the front of a queue. It is settled, and treating it as a queue mistakes the politics. 4. There is a mid-tier, and it is the only cluster in the data behaving the way carbon pricing was supposed to. 5. A 1,543 to one dispersion is the operational signature of no global price floor. Anything that depends on a global price reference has already chosen its cluster, even if it has not said so.

We take them in turn.

Part I. The average of fictions

A casual reader of the dashboard stops at the headline. US$127.7 per tonne, average across the top ten priced instruments. The number sits in the band where most peer-reviewed estimates put the social cost of carbon. The conclusion writes itself: the world is finally pricing carbon at roughly the right level, and the policy task is to drag the rest of the system up to where the leaders already are.

The casual reader is wrong. Not about the arithmetic. About the structure.

Of the 85 implemented instruments with a published US-dollar price, only seven clear US$100. Eighteen sit at or below US$5. The sixty in between are not distributed around any sensible mean. They cluster in two or three places with very little density in the gap that separates them. The mean of a bimodal distribution is not a typical value; it is a numerical artefact. Nothing in the dataset is at US$127.7 because nothing in the underlying world is at US$127.7. The number is an average of fictions.

We do not say the World Bank is wrong to compute it. We say it is wrong to read it as the global price. The right benchmark, for an allocator pricing the cost of carbon-intensive exposure, is the cluster the asset is in or the cluster its compliance pathway will drive it toward. The right global aggregate, if one is needed, is a coverage-weighted distribution with the three modes named, not a mean that washes them out.

The conventional reading of the dashboard is the reading the dashboard makes easy. The honest reading is the one that requires looking past the chart that sits at the top of the page.

Part II. The club at the top

Seven instruments are priced at or above US$100. Norway leads at US$169.71. Read the names beneath it: Sweden, Liechtenstein, Switzerland, Denmark, the Netherlands. The cluster is not global. It is Northern European, plus two Alpine carve-outs.

What looks, from a distance, like convergence is at close range a club. A club is not a model. These are jurisdictions with high fiscal capacity, stable centre-coalitions on climate policy, mature low-carbon energy systems, and industrial bases that have already diversified away from carbon-intensive exports. The carbon price is high because keeping it high is bearable. The price has stayed high because each member of the club has the other six as a reference point, and the cluster's convergence is the convergence of a club watching itself.

The optimistic reading is that the club is the leading edge of a wider movement, and that other jurisdictions are climbing toward it. The dashboard does not support that reading. The high cluster is consolidating. The instruments above US$100 have, on average, drifted further from the median over the last five years, not closer. The adjacent jurisdictions hinted at as potential club members (the United Kingdom on certain EU ETS settlement years, Germany) are at the lower end of the cluster's price range, not driving its leading edge upward.

A club is what the high cluster is. The error is reading it as the rest of the system's destination.

Part III. The bottom is not the front of a queue

Eighteen instruments are priced at or below US$5. The lowest is Poland at US$0.11. Below US$1 sit Ukraine, Indonesia, and Kazakhstan. The first instinct is to read this as the front of a queue: implemented but not yet ambitious; prices that will climb as the system matures.

That instinct, increasingly, is wrong. Several of these instruments have been in place for years. Their prices have not climbed. The forces that would push them upward (a tightening cap, competitive pressure from a Carbon Border Adjustment Mechanism, domestic political demand for higher ambition) have either not arrived or have been absorbed by the system without translating into higher allowance prices. [open question for the editor: a follow-up comparing the bottom-tier prices to per-jurisdiction social-cost-of-carbon estimates would sharpen "thin and stagnant" into a quantified claim.]

The bottom is an equilibrium. It is not the front of a queue. The political economy that produced these prices is the same political economy that is keeping them there: limited fiscal capacity, exposure to carbon-intensive exports, energy systems that still depend on coal, and a domestic constituency that does not yet see the carbon price as a binding constraint on anything they care about. The compliance instrument has been implemented because implementation was the requirement. The price, in these jurisdictions, is closer to a registration fee than a market signal.

For operators with assets in these jurisdictions, the carbon price is not what they should be modelling against. They should be modelling against whatever the binding constraint actually is: capital cost, technology availability, the threshold at which export markets start to price carbon content. The compliance price will not be the variable that moves their decision. Reading the bottom of the distribution as a queue gives operators permission to wait. The dataset suggests waiting is wrong, because there is nothing in the data to wait for.

Part IV. The mid-tier holds, and the mid-tier matters

Between the high cluster and the low tail sits a distinct mid-tier around US$20 to US$25. Australia. New Zealand. The United Kingdom on its non-traded carbon tax. Colorado. The mid-tier is, for our money, the most editorially interesting reading in the data, because it is the only cluster in the distribution behaving the way carbon pricing was supposed to behave.

These are mature institutions with credible monitoring and verification. The price has been stable for several years. It has not crept upward toward the club. It has not slipped toward the tail. It is high enough to change the relative economics of low-carbon and high-carbon alternatives at the margin. It is low enough that the underlying technology cost still drives most investment decisions.

That is the band in which a carbon price is doing its intended job as a marginal signal: visible, material, but not the dominant variable. The mid-tier is what the global benchmark would look like, if the global benchmark existed. The fact that the rest of the system is not converging toward it is the most important negative finding in the data. [open question for the editor: a twelve-month tracker on mid-tier drift, weighted by emissions coverage, would test whether the mid-tier is holding, drifting up toward the club, or slipping toward the tail.]

If the mid-tier holds and the rest of the system does not move toward it, the global mean will keep being meaningless. If the mid-tier drifts upward, the global signal starts to become real. If the mid-tier slips, the system has decided what it is, and we are looking at it.

Part V. Fifteen hundred to one is the absence of a floor

Norway charges US$169.71 per tonne. Poland charges US$0.11. The ratio is 1,543 to one. There is no global commodity that ranges by three orders of magnitude across implemented markets where the differential reflects anything other than political and fiscal choice. The dispersion is the most direct evidence the dataset provides that there is no global carbon price floor, in any operational sense.

A floor is what would make the headline number meaningful. A floor is what cross-border mechanisms assume, implicitly, when they price carbon content into traded goods. A floor is what the belief that opened this piece requires for the belief to hold. We do not have one.

What we have, instead, is a system in which jurisdictions price carbon at the level they can afford, can sustain politically, and can coordinate with neighbours of comparable capacity. That is the substrate from which a global floor might one day emerge. The 2026 data does not show the substrate moving toward integration. It shows three equilibria, holding, and the gap between them widening rather than closing.

For regulators, the absence of a floor matters because every cross-border mechanism that depends on a price reference has, in practice, to choose which cluster to anchor to. The European Union's Carbon Border Adjustment Mechanism is anchoring against the high cluster, with the trade implications that implies. A sectoral agreement on shipping or aviation that priced against the high cluster would be economically aggressive; against the mid-tier, politically passable; against the low tail, performative. The choice cannot be hidden behind the global average. The global average does not exist.

Implication

We opened with a belief. Carbon pricing exists because the cost of damaging the atmosphere should appear, consistently, on the balance sheets of those whose decisions produce the damage. If the price is consistent, the belief works. The 2026 Carbon Pricing Dashboard is the system's most-cited record of how far that belief has been delivered.

On the reading the dashboard makes easy, the answer is "encouragingly far". On the reading the dashboard rewards, the answer is "less than the headline number suggests". The world that the headline number describes does not exist. The world that does exist is three smaller worlds, watching each other, and not moving toward each other.

That changes what allocators should benchmark against. It changes what regulators have to choose, openly, when they design any cross-border mechanism. It changes what operators should price internally, because pricing against the global mean prices against nothing. None of these changes are abstract. They follow from reading the dataset for what it shows rather than for what its headline reports.

The belief carbon pricing rests on is the right belief. The system the dashboard summarises is not, on present evidence, delivering it. That gap is not a critique of the dashboard. It is what the dashboard, read honestly, is for.