Every carbon credit rests on a claim about a world that never happened: a forest left standing instead of logged, a cookstove nobody would have bought, a power plant that would have burned coal for another twenty years. The credit exists because someone built a model of that missing world, then measured the gap between it and the one we actually got.
For coal retirement, that modeling problem is close to unsolvable in practice. Once a plant shuts down, there is nothing left to check the model against: how much it would have generated, at what capacity factor, burning how much fuel, over how many years of its remaining permitted life. Every one of those numbers is an assumption, and every assumption is a place where a skeptical buyer, a rigorous validator, or a journalist with time on their hands can find something to dispute, often years after the credit has already been sold.
The number nobody can rerun
Transition credit frameworks have gotten more careful about this over time: baseline studies, conservative discount factors, independent validators, multi-year monitoring periods built to catch a plant that quietly runs longer than promised. All of that effort goes toward making a hypothetical more defensible.
It does not make the hypothetical disappear. The baseline for a retired coal plant is still a story about what that plant would have done if it had stayed open, and there is no version of that story that can be run twice to check the answer. However careful the assumptions, the plant that actually generates the credit is the plant that stopped generating. What it would have produced instead is permanently, structurally unobservable.
A baseline for a retired plant is a story about a plant that no longer exists. However carefully it is built, nobody can go back and check it against what the plant actually would have done.
That is not a flaw in any one methodology. It is the shape of the problem every retirement-based credit has to solve, and it is why additionality disputes in this part of the carbon market tend to resurface long after a transaction closes, when a new dataset or a new analyst reopens a question that was only ever answered by assumption in the first place.
A quieter version of the same problem
The same structure shows up, more quietly, almost everywhere else in the transition credit space. A fuel-switching project needs a counterfactual fuel mix. An efficiency retrofit needs a counterfactual consumption curve. Even a project that never touches a physical asset needs some version of what would have happened otherwise to calculate a credit at all.
Coal retirement just makes the problem loudest, because the counterfactual asset stops existing entirely, and because the volumes at stake are large enough that scrutiny is guaranteed. But the underlying design choice, crediting against a modeled scenario rather than a measured event, is common to nearly all of it. Buyers who have spent the last few years watching additionality claims get picked apart in the press are not reacting to one bad project. They are reacting to a design pattern.
A receipt, not a projection
An ARC, an Avoided Right to emit Carbon, is built to avoid that design pattern entirely, and it does so by changing what question the credit is answering.
A coal PPA already grants the plant a specific emission right for a specific contracted volume: a figure set by an approved emission factor, a regulatory benchmark, or an ETS allocation, recorded in a legal or regulatory document that exists independently of any carbon transaction. That figure is not created to generate a credit. It already exists, attached to the contract, before anyone structures a deal around it.
When a defined slice of that contracted delivery, called a SPARC stream, is instead fulfilled by a lower-emission generator, the delivery is metered like any other dispatch event. If the verified emissions of that delivered MWh come in below the emission right the coal contract had granted for it, the difference is issued as an ARC. The calculation is arithmetic: the embedded emission right minus the verified emissions of the delivered MWh. Nothing about that number depends on a scenario, a discount factor, or an assumption about what would have happened otherwise.
An ARC is issued only after a SPARC stream has been fulfilled by verified lower-emission delivery. Its quantity is the embedded emission right already recorded for that SPARC, minus the verified emissions of the delivering MWh. Only positive differences are issued, and there is no scenario in that formula for anyone to defend later.
Call it a receipt. Or, closer still, an affidavit: a documented statement, tied to a specific meter reading and a specific legal instrument, that a specific thing happened on a specific date, and did not happen a different way. The plant does not close. Nothing about the contract changes. What changes is who generated a defined, verified portion of the electricity it was already obligated to supply, and whether the emission right attached to that portion went unused.
In practice, that means an ARC carries a specific set of facts rather than a set of assumptions: the originating coal plant and PPA, the delivery band and interval it fulfilled, and the verified emissions of the MWh that fulfilled it. A skeptic can ask to see the plant's approved emission factor. They can ask for the metering data behind the delivered MWh. They cannot ask to see the plant's alternate future, because no such record exists to produce, on either side of the transaction.
This also reframes what a challenge to an ARC looks like. Nobody disputing an ARC is arguing about a hypothetical operating history for a plant that no longer exists. They are asking whether the embedded emission right, the approved factor or benchmark behind it, was set correctly in the first place, a question about a document and a methodology, not about an unwitnessed future. That is a narrower, more checkable argument than the one that follows a retirement credit around for the life of its crediting period.
Picture how each conversation actually goes a few years after the transaction. For a retirement credit, the question is some version of: are we sure this plant would really have run at 80 percent capacity through the mid-2030s if it hadn't closed. That is an argument about a future nobody observed. For an ARC, the question is: does the plant's approved emission factor match what the regulator granted, and does the metering data show what it says it shows. Both are real questions. Only one of them can be settled by pulling a document off a shelf.
What this means for a buyer
None of this means an ARC can do everything a retirement credit does. A retirement credit, however contested its baseline, can in principle represent the full remaining lifetime emissions of an asset that has genuinely stopped operating: a large, one-time volume tied to a single irreversible event. An ARC cannot claim that. It accumulates stream by stream, MWh by MWh, tied to delivery that has to keep happening for the record to keep growing. That is a real difference in what each instrument is trying to prove, not only in how it proves it, and buyers should know which one each instrument is built for. We have covered that SPARC-versus-retirement structure in more depth in our earlier comparison of coal transition results.
The table below is not a scorecard. It maps what each instrument is asking a buyer, a regulator, or a journalist to believe.
| What must be true | Modeled retirement credit | ARC |
|---|---|---|
| Reference point | An estimated operating history for an asset that no longer exists | A regulatory emission right recorded before the trade, attached to a contract still in force |
| When it is calculated | Before the outcome, against a projected baseline | After delivery, against a metered outcome |
| What a challenge attacks | The plausibility of a hypothetical future | The accuracy of a benchmark or allocation set independently of the transaction |
| What volume it can represent | Full remaining lifetime emissions of a retired asset, in one volume | The unused right on each verified MWh, accumulated stream by stream |
We think that trade-off favors ARCs more often than it costs them, at least for the specific problem of coal phase-out and early retirement, where transition credits have struggled hardest to hold up under scrutiny. An instrument whose central claim rests on a document that predates the trade, and a meter reading that follows it, is answering a narrower, harder-to-dispute question than one whose central claim rests on a scenario nobody can rerun. That does not make it a bigger instrument yet, or a more liquid one, or one with the market history that retirement credits and ETM structures have already built. It makes it a different kind of claim, one we think holds up better exactly where the current generation of transition credits has been challenged the most.
That still leaves real questions outside the mechanism's control. What a host country recognizes as the embedded emission right for a given plant is a regulatory determination, not something SPARC sets on its own. What that government later chooses to do with an unused right domestically, or whether it authorizes an international transfer, is a sovereign decision layered on top of the registry record. An ARC does not resolve either question. It narrows the argument down to ones that can actually be checked against a document, rather than argued from a model.
For buyers weighing eventual compliance-grade use, corresponding adjustments, letters of authorization, Article 6 or CORSIA pathways, a claim that rests on documents and meters is easier to line up against those frameworks than one that rests on a scenario. That alignment work is still ongoing. It is not a settled feature of the mechanism today, and we would rather say that plainly than let the reader assume otherwise.
What a buyer ends up with is a shift in the argument they have to be ready to make: not "trust our model of a world that did not happen," but "here is the document that granted the right, and here is the meter reading that shows it went unused." For how that argument can be structured into an actual purchase commitment, see our earlier piece on philanthropic capital. Whether that argument matters as much to the market as it does to us is not something we can settle in a blog post. It is something the scrutiny already heading toward this asset class will decide, one transaction at a time.