There are two carbon markets, and almost every argument about carbon markets comes from mixing them up.
In a compliance market, a government sets a legal limit on emissions, hands out or auctions a fixed number of permits, and requires covered companies to surrender one permit for every tonne they emit. Miss the deadline and you pay a penalty. Demand exists because the law creates it.
In a voluntary market, someone develops a project that reduces or removes emissions, has it certified against a methodology, and sells the resulting credits to a buyer who wants to make a claim. Nobody is obliged to buy anything. Demand exists because a buyer chooses to show up.
That single difference — legal obligation versus discretionary purchase — explains the price gap, the liquidity gap, the integrity debate, and why the two markets behave nothing like each other in a downturn.
The two systems, side by side
Compliance (regulatory) market
- Who creates demand
- A government, by law. Covered installations must surrender units or pay a penalty.
- The unit
- An allowance — a permit to emit one tonne. It is created by the regulator, not earned by a project.
- Supply
- Fixed by the cap. It falls on a published schedule regardless of price.
- What sets the price
- The marginal cost of abatement across the covered sector, plus expectations about future cap tightening.
- Enforcement
- Statutory penalties, typically well above the market price, plus an obligation to surrender the missing units anyway.
- Failure mode
- Over-allocation. Too many free permits and the price collapses — the EU's own experience in Phases I and II.
Voluntary market
- Who creates demand
- A buyer, by choice — usually a corporate net-zero claim, occasionally a supply-chain requirement.
- The unit
- A credit — evidence that one tonne was avoided or removed somewhere, relative to a counterfactual baseline.
- Supply
- Elastic. Developers respond to price, so a demand surge is met with issuance rather than scarcity.
- What sets the price
- Perceived quality — additionality, permanence, measurement, co-benefits — and the reputational risk the buyer is taking on.
- Enforcement
- None, beyond reputation and the standards bodies. There is no regulator to fail.
- Failure mode
- Integrity. If credits are found not to represent real reductions, demand disappears rather than reprices.
There are two carbon markets and they run on opposite logic. In one, a government caps how much a factory may emit and the factory must buy a permit for every tonne — no choice about it. In the other, a company voluntarily buys a certificate saying that somebody, somewhere, planted trees or handed out clean cookstoves. One is a law. The other is a purchase decision. Everything else follows from that.
How a compliance market operates
Nearly all compliance trading systems are variations on cap-and-trade. The mechanics are the same whether you are looking at the EU, California, Korea or the UK.
The clever part is step 3. The regulator decides how much is emitted and lets the market decide who emits it. A firm with a cheap abatement option takes it and sells its spare allowances; a firm facing an expensive option buys instead. The total falls by the amount the cap says, at the lowest total cost the system can find.
Two features do most of the real work:
The declining cap. In the EU ETS the annual reduction accelerated to 4.3% a year from 2024, up from 1.74% in the previous phase. That schedule, not today’s price, is what gives an allowance long-dated value.
The stability reserve. Left alone, cap-and-trade markets accumulate surpluses in recessions and crash. The EU’s Market Stability Reserve automatically withdraws allowances when the number in circulation exceeds 833 million. The May 2026 surplus indicator triggered a withdrawal of 190 million allowances between September 2026 and August 2027 — down from 276 million the year before, which tells you the surplus is shrinking.
Free allowances exist to stop production simply relocating to a country without a carbon price — "carbon leakage". The EU's answer is to phase free allocation out and replace it with a border charge on imports.
The Carbon Border Adjustment Mechanism entered its definitive period on 1 January 2026. Importers of covered goods buy CBAM certificates priced off EU ETS auction clearing prices — €75.36 a tonne for the first quarter of 2026, moving from quarterly to weekly averaging from January 2027. Free allocation to CBAM sectors falls on a schedule: a 2.5% phase-out factor in 2026, 48.5% by 2030 and 100% by 2034. For an Indian or Chinese steel exporter, that schedule is the number to plan against.
Cap-and-trade works like a fixed number of parking permits in a crowded city. The council decides how many exist and reduces the number every year. Whoever can most easily stop driving sells their permit to whoever cannot. Total parking falls by exactly the amount the council intended, and the market sorts out who gives up first — which is cheaper than the council trying to guess.
How a voluntary credit is made
A voluntary credit follows a completely different path. Nothing is issued in advance; everything is earned after the fact and verified by a third party.
Everything contentious about the voluntary market lives in step 1. A baseline is a statement about a world that did not happen. If the project would have gone ahead anyway, the credits are not additional and the tonnes are fictional. If a protected forest burns down in ten years, the reduction was not permanent. If the same forest would have been logged and the loggers simply moved next door, the emissions leaked.
None of these are accounting quibbles. They are the reason a voluntary credit and a compliance allowance are not the same instrument, and should never be quoted on the same line without qualification.
Two bodies now try to police this. The Integrity Council for the Voluntary Carbon Market assesses methodologies against Core Carbon Principles and awards a CCP label to those that pass. The Voluntary Carbon Markets Initiative addresses the other end — what a buyer is allowed to claim once they have retired the credits. Together they have split the market rather than fixed it.
A voluntary credit is a claim that somewhere in the world a tonne of emissions did not happen because a project paid for it not to. The entire question is what would have happened otherwise — and since "otherwise" never occurred, nobody can check it directly. That is why this market lives or dies on the credibility of its auditors rather than on any physical measurement.
Why the prices are so far apart
One tonne of CO₂, seven different prices
Indicative 2026 ranges, USD per tonne of CO₂ equivalent
A hundred-fold spread for a physically identical tonne. The price is not paying for carbon; it is paying for the credibility of the claim attached to the carbon. Sources: World Bank State and Trends 2026, market assessments.
The global average direct carbon price has roughly doubled over a decade, from about USD 10 a tonne in 2016 to USD 21 today. That average conceals everything. Trading systems average around USD 22 and taxes around USD 19.50, but the EU sits near USD 87 while many jurisdictions price below USD 5.
On the voluntary side the market has visibly split. Legacy avoidance credits trade at USD 1–5 and are increasingly unsellable at any price. Durable, measurable removals — direct air capture, biochar, enhanced weathering — clear at USD 40 to well over USD 150, and are supply-constrained. The middle has thinned.
The volume data confirms the strain: issuances rose 8% from 2024 to 2025, but retirements fell about 11%. Supply is growing while demand shrinks — which is exactly what you would expect when buyers are optional and nervous.
The same tonne of carbon dioxide sells for one dollar and for a hundred and fifty dollars, depending on the paperwork attached to it. You are not buying the gas; you are buying how much anyone believes the claim. A European permit is worth about eighty-seven dollars because the law says a company must hold one. An old forest credit is worth a dollar because almost nobody wants it any more.
Carbon as a switching lever
A carbon price is not only a cost line on a compliance report. In power markets it is the dial that decides which fuel gets burned, and that makes it the most consequential number in European energy.
The comparison every generator runs is between the clean spark spread — the margin on a gas plant after fuel and carbon — and the clean dark spread, the same calculation for coal. Coal is cheaper per unit of energy but emits roughly two and a half times more carbon dioxide for each megawatt-hour of electricity, so the carbon price is what decides the contest.
Running it at August 2026 levels, with TTF at €66/MWh, API2 coal at USD 130 a tonne and EU allowances at €75:
Coal is €38 a megawatt-hour cheaper to run. Coal dispatches first, gas sits idle, and Europe burns more carbon — despite a €75 carbon price doing its level best to stop it.
The more useful way to express this is as a switching price: the gas price at which the two are equal.
TTF is at €66. Gas would have to fall by roughly a third before it displaces coal again.
Each €1 on the allowance price costs a coal plant €0.895 and a gas plant €0.367 per megawatt-hour — a net €0.53 in gas's favour. Divided by the heat rate, that moves the switching price by about €0.29 per MWh of gas for every €1 of carbon, or roughly €3 per €10 of EUA.
Which gives the striking number. To lift the switching price from €45 to today's €66 gas price would take another €21, and at €0.29 per euro of carbon that needs an EUA around €147 a tonne — close to double where it trades. At 2026 gas prices, the European carbon market is not strong enough to keep coal off the system. That is a policy result, arrived at by arithmetic.
This is also where the two halves of the energy transition collide. The gas price in that calculation is set by the LNG market described in The Atlantic–Pacific LNG Arbitrage, and the resulting spark spread is what pays for the flexible plant that covers the evening ramp in The Intermittency Problem. A carbon price, a shipping lane and a battery tender are all the same problem viewed from three angles.
The carbon price is really a dial that decides whether a power station burns gas or coal, because coal releases about two and a half times more carbon dioxide for the same electricity. Right now gas is so expensive that even a €75 carbon price is not enough to make it the cheaper option — carbon would have to be roughly double before European power stations switched back.
Where the two markets meet
They are not fully separate, and the joins are where most of the interesting policy action now sits.
| Mechanism | What it does | Status in 2026 |
|---|---|---|
| Offsets inside an ETS | Some compliance systems let covered firms surrender a limited quantity of credits instead of allowances, usually with a strict quality filter and a percentage cap | Widely used but tightening. Compliance retirements of credits fell sharply, notably in California |
| CORSIA | The aviation sector's global scheme. Airlines must offset growth in international emissions above a baseline using eligible credits — a compliance obligation met with voluntary-market units | Eligible credits traded USD 15–22 a tonne between September 2025 and April 2026. The clearest bridge between the two markets |
| Article 6.2 (ITMOs) | Country-to-country trading of internationally transferred mitigation outcomes, with a "corresponding adjustment" so the tonne is not counted twice | 108 bilateral agreements signed, but very few actual transfers. The paperwork is running well ahead of the volume |
| Article 6.4 (PACM) | The UN-supervised successor to the Clean Development Mechanism, issuing a credit with sovereign backing | First credits provisionally issued in 2026, to clean cookstove projects in Myanmar. Very early |
| CBAM | Extends a domestic carbon price to imports, so foreign producers face the same cost as EU ones | Definitive period live since 1 January 2026. Credits cannot be used against it — only a carbon price paid in the country of origin counts |
The direction of travel is consistent: compliance systems are becoming more willing to recognise credits, but only credits that look like allowances — verified, permanent, adjusted at the national level. The rest of the voluntary market is being left to reputation.
The two markets do touch in a few places. Airlines must offset their growth using approved credits. Countries can trade reductions with each other under the Paris rules. And Europe now charges imports the same carbon cost its own factories pay. In every case, the compliance side only accepts credits that behave like permits — audited, permanent, and officially counted by a government.
India’s CCTS: a third design
India’s Carbon Credit Trading Scheme is worth its own section, because it is neither a classic cap-and-trade nor a voluntary market — and because for anyone operating in India it is now a live compliance obligation.
The CCTS is baseline-and-credit rather than cap-and-trade. There is no absolute cap on tonnes. Instead each obligated entity gets a greenhouse-gas emission intensity target — emissions per unit of output — set at sub-sector level, using financial year 2023-24 as the baseline. Beat the target and you earn tradeable Carbon Credit Certificates; miss it and you must buy them.
The practical detail as of mid-2026:
- Seven of nine sectors now carry binding obligations: aluminium, cement, chlor-alkali, pulp and paper, petroleum refining, petrochemicals and textiles. Iron and steel, and fertiliser, are awaiting final targets.
- Roughly 490 obligated entities are covered.
- Targets are set for compliance years 2025-26 and 2026-27, with obligations having begun on 1 April 2025 and the first compliance date on 31 July.
- Certificates trade on India’s power exchanges, with first CCC trading expected around mid-2026.
- The Indian Carbon Market Portal went live on 21 March 2026 for registration, issuance and verification.
The design choice is deliberate. An intensity target lets output grow — which a developing economy needs — while forcing efficiency per unit. The trade-off is that total emissions can still rise, so the scheme reduces the carbon intensity of growth rather than capping it. It is the same architecture India used for its earlier Perform, Achieve and Trade energy-efficiency scheme, extended to greenhouse gases.
India built something different. Rather than capping total emissions, it gives each factory a target for emissions per tonne of product. Beat the target and you earn certificates to sell; miss it and you must buy them. It lets the economy grow while forcing every unit of output to get cleaner — a sensible compromise for a country that still has a great deal left to build.
The risk lens
For anyone carrying these instruments on a book, the exposures are different in kind.
Compliance allowances
Policy risk dominates. The cap trajectory, the stability reserve parameters and free-allocation rules are all set by legislation that can change. On 17 July 2026 the European Commission published proposals to slow the cap reduction, halve MSR intake after 2030 and extend the CBAM phase-in to 2038. None is adopted, but a serious proposal alone moves the forward curve.
Secondary exposures are conventional: price risk on a liquid futures curve, basis risk between an EUA and a UKA, and the working-capital cost of holding allowances between accrual and surrender.
Voluntary credits
Integrity risk dominates. The instrument can be devalued retrospectively — a methodology downgraded, a registry suspended, an investigation published — and there is no regulator to appeal to. Positions are illiquid and largely bilateral, so a mark is an opinion rather than a price.
Reversal risk is specific to this market. A nature-based credit sold on a hundred-year permanence assumption can be undone by one fire season. Buffer pools help, but they are pooled insurance, not a guarantee.
The two instruments fail in different ways. A government permit is threatened by politics — a new law can change its value overnight. A voluntary credit is threatened by scrutiny — one investigation can make it worthless, and there is no regulator to appeal to. Know which of those risks you are actually holding.
The plain summary
A compliance allowance is a licence to emit, created by law, in fixed supply, enforced by penalty. A voluntary credit is a claim about a tonne avoided or removed elsewhere, created by a project, in elastic supply, enforced by reputation.
They can meet — through CORSIA, through Article 6, through the limited offset provisions inside some trading systems — but they are not substitutes, and treating them as interchangeable is the single most common and most expensive error in this market.
Three things to watch into 2027: whether the EU’s July 2026 proposals to soften the cap and the CBAM timetable are adopted, because that repricing would be substantial; whether ETS2 launches on schedule for buildings and road transport, which would extend a carbon price to households for the first time at scale; and whether India’s first CCC trades establish a credible price signal, which will determine whether the CCTS becomes a real market or a compliance formality.
One sentence worth keeping: a compliance permit is a licence to emit, and a voluntary credit is a claim that somebody else emitted less. They are not the same thing, they are not interchangeable, and treating them as if they were is the most expensive mistake in this market.
Sources and further reading
- State and Trends of Carbon Pricing 2026 — World Bank
- Carbon Border Adjustment Mechanism — European Commission
- Compliance obligations under India’s CCTS enter into force — International Carbon Action Partnership
- Indian Carbon Credit Trading Scheme factsheet — International Carbon Action Partnership
- The Core Carbon Principles — Integrity Council for the Voluntary Carbon Market
- India’s carbon credit trading system scheme — IETA