Carbon offset additionality
A Socratic walk-through of carbon offset additionality — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why can paying to protect a forest that was never threatened still earn carbon credits?
A carbon credit is a claim that one tonne of carbon dioxide was kept out of the air. Sit with that for a moment. Kept out compared with what? Not compared with last year, and not compared with any measurement — compared with a world that did not happen.
That is an odd sort of merchandise. Every other commodity you can put on a scale. This one is the size of a gap between reality and an imagined alternative, and only one half of that gap can ever be observed.
Reasoning it through
REASONING #So the first question is not about forests at all. It is: who supplies the half that cannot be observed?
The imagined alternative has a name — the baseline — and it is written down before the project starts, usually by the project developer, checked by a validator the developer selects and pays. The credits issued are proportional to the distance between that baseline and what is later observed on the ground.
Now hold those two facts side by side. The seller's revenue is proportional to a number the seller proposes. What does that arrangement reward?
Not fraud, necessarily — something more ordinary. Consider a developer choosing between two forests. One sits on a logging frontier with a road already cut toward it; protecting it is expensive, contested, and if it is protected the baseline will be hard to defend because the surrounding rate of clearing is genuinely uncertain. The other is remote, steep, and nobody has any near-term intention of touching it. Protecting it costs almost nothing, because nothing has to change. If the baseline for that second forest can be written as "cleared at five percent a year", the project earns credits for a decade of doing exactly what would have happened anyway.
Which forest does a rational developer bring to market? And notice what has happened: the market has not merely tolerated the non-additional project, it has actively selected for it, because the least threatened forest is the cheapest to protect and the most profitable to overstate. The very thing that makes a project useless makes it attractive.
Does the evidence bear this out? Reasonably strongly, though it is contested. A 2016 study by the Oeko-Institut for the European Commission assessed projects under the Clean Development Mechanism and judged that only around 2 percent had a high likelihood of being additional, while roughly 85 percent had a low likelihood. In 2023 a study in Science by West and colleagues compared voluntary forest-protection sites with statistically matched control areas and found that most had avoided far less deforestation than they had been credited for. Registries have disputed the methods; the direction of the finding has been hard to dislodge.
There is a sharper case still, because it shows the incentive doing something worse than nothing. Under the Clean Development Mechanism, destroying the industrial waste gas HFC-23 earned credits at a very high rate, since the gas is thousands of times more potent than carbon dioxide as a greenhouse agent. The credits became worth more than the refrigerant the plants were nominally in business to make. Several plants ran, it appeared, partly in order to generate waste gas to destroy. The European Union barred those credits from its emissions trading system from 2013. The scheme had not merely paid for nothing; it had paid for the pollution.
And here is why a non-additional offset is not simply wasted money. A wasted subsidy costs you the subsidy. A wasted offset issues a permit — someone buys it and emits a real tonne on the strength of it. The accounting says zero; the atmosphere gets one tonne. A bad offset is worse than no offset.
The analogy
THE ANALOGY #Think of a doctor paid a bounty for every life saved, who also writes the prognosis. Nothing forces a false diagnosis. But a doctor who describes healthy patients as gravely ill, and then treats them, collects the bounty for every recovery — and the healthy patients are cheaper to treat, easier to cure, and less likely to die and spoil the record.
the doctor's claim can eventually be tested against the patient, who is standing right there and might have got a second opinion, whereas the counterfactual forest never exists to be examined at all — the baseline can only ever be argued about, never checked.
Clarifying the model
THE MODEL #A few refinements keep this from tipping into cynicism.
Additionality is a property of the payment, not of the forest. A standing forest holds real carbon and does real ecological work whether or not any credit was ever issued against it. What fails is the claim that the payment caused it.
Nor is additionality unknowable in principle. It is easiest to establish where a project has no other reason to exist — a landfill gas flare that produces nothing saleable, say — and hardest where the project is something a landowner might plausibly have done regardless. The problem is not that the counterfactual is mysterious; it is that the person who benefits is the one describing it.
That points at the repair, which is not more scrutiny of individual claims but moving the estimate away from the beneficiary: baselines set from what actually happened across a whole region or jurisdiction, updated as conditions change, so that a project earns only by beating its neighbours rather than by beating its own forecast. Whether crediting can be made reliable enough for offsetting — as opposed to funding conservation without licensing an emission somewhere else — remains genuinely disputed.
Additionality is also only the first of three tests. Permanence asks whether the carbon stays put, which a fire or a change of government can undo. Leakage asks whether the logging crew simply moved to the next valley. A project can be perfectly additional and still fail both.
A picture of it
THE PICTURE #How to readStart at the input shape on the left of the top row — the safe forest that a developer would rather bring to market. Follow it into the setup step where the baseline is written, then into the diamond, which is the only genuine question in the whole scheme. The honest branch ends almost immediately in few credits. The claimed branch runs down through issuance to the hazard node at the bottom, where a permit has been created against a saving that never occurred. The arrow returning from issuance back up to the baseline step is the point of the whole picture: the money earned is the reward for the number the seller chose.
What became clearer
WHAT CLEARED #A credit is priced off a counterfactual, and the counterfactual is written by the party being paid. Everything else follows from that single structural fact — not from bad faith, but from an incentive that quietly steers the market toward the projects where the imagined alternative is easiest to inflate and hardest to check. The forest was never in danger; the paperwork says it was; and somewhere a real tonne has just been emitted under cover of the difference.
Where to go next
ONWARD #- Jurisdictional and dynamic baselines, and whether they actually blunt the incentive or just move it.
- Permanence and buffer pools — what happens to a credit when the forest burns.
- Contribution claims versus offset claims, and why dropping the one-for-one licence changes the incentive completely.
Key terms
TERMS #| Term | What it means |
|---|---|
| Additionality | the requirement that the emissions reduction would not have happened without the crediting revenue. |
| Baseline | the counterfactual scenario against which reductions are measured. |
| Leakage | emitting activity displaced outside the project boundary rather than prevented. |
| Permanence | whether stored carbon stays stored, and for how long. |
| Clean Development Mechanism | the Kyoto Protocol crediting scheme under which the HFC-23 and additionality problems were first documented at scale. |
Every term the collection defines is gathered in the glossary.