An autopsy of the voluntary carbon market

The voluntary carbon market was built on a single promise: that one credit cancels one tonne. Most scrutiny has asked whether that promise holds on the supply side – whether the credits are real. This column asks a different question: what did the mere availability of offsets do to the firms that bought them? Using a buyer-linked dataset spanning the near-universe of offset retirements, and the 2023 integrity scandals as a natural experiment, the authors find that firms which walked away from offsetting cut their own operational emissions sharply more than firms that stayed. Even a market of flawless credits could still slow real decarbonisation.

Corporate net-zero strategies increasingly rely on the voluntary carbon market. A company that wants to reduce its net emissions can do it in two very different ways. It can abate at the source, changing how it produces or by investing in more efficient equipment. Or it can buy and retire carbon credits. As the latter strategy is often cheaper, the offset market blunts firms’ incentives to cut emissions: through voluntary offsets, firms can claim progress towards decarbonisation while avoiding costly changes to their production process. This would not be a big issue if the market worked well, if there were a large supply of offsets that are linked to activities that actually remove carbon from the atmosphere, and firms had to buy these offsets bringing their price to equal the marginal cost of abatement. Unfortunately, none of these conditions holds. When we started working on the voluntary carbon market with the idea of conducting an anatomy of this market, we soon realised that what we were doing was closer to an autopsy (Panizza et al. 2026). And as in any autopsy, the central story is not the visible wound but the condition underneath that the naked eye tends to miss.

The integrity of carbon offsets has long been questioned in academic and policy circles (e.g. Badgley et al. 2022, Calel et al. 2025, Guizar-Coutiño et al. 2022). We find that when the integrity issues were suddenly brought to the public eye in January 2023, many firms stopped buying offsets and subsequently cut their own operational emissions far more than those that kept buying carbon credits. 

Our analysis leads to an uncomfortable conclusion related to the demand side of the market and suggests that current efforts to clean up the supply side of the market (Pisani-Ferry et al. 2025) would not be enough to increase firms’ abatement effort. 

A question between two debates

Two questions dominate the recent debate on the voluntary carbon market.

The first asks whether the credits deliver what they promise. Critics point to three problems. Projects often issue more credits than the emissions they actually avoid. The carbon they store does not always stay stored, because a protected forest can burn or fall to the chainsaw ten years later. And projects measure their success against a guess about what would have happened without them, a guess they make themselves and tend to paint in dark colours.

The second question asks whether corporate climate promises mean anything in the first place. Dietz and Hastreiter (2026) find that net-zero targets are neither greenwashing nor a gamechanger. Acharya et al. (2025) argue that headline pledges follow a profit-driven strategy rather than empty rhetoric. Homroy and Rauf (2024) ask whether green ambitions survive contact with a firm’s own supply chain, and find that they often do not.

Both questions are about credibility – of the credit, or of the promise. Neither takes on the question that sits between them. Does the mere availability of a cheap offset change how hard a company works to cut its own emissions, even when the credit behind it is real? Call this the demand side of integrity. The rebound and moral-licensing literatures have long speculated about it (Gillingham et al. 2016; Burger et al. 2022), but almost nobody has studied it. That is the focus of our recent discussion paper. 

Putting the patient on the table

We stitch together three data sources. The first is a buyer-linked dataset from AlliedOffsets that covers the near-universe of the market, with more than 36,000 projects and, crucially, retirements matched to identifiable corporate buyers. The second is the disclosed emissions of those same firms. The third is their financial accounts.

We begin with four descriptive symptoms.

• Symptom 1. Avoidance projects dominate the market. These projects do not take carbon out of the air. They claim credit for emissions that never happened, measured against a hypothetical world in which the project never existed. A credit avoids a tonne only if that hypothetical holds and the project would not have gone ahead anyway. Much of the integrity evidence finds that this condition often fails (Figure 1).
• Symptom 2. Registries have issued roughly 5 Gt of credits, but buyers have retired far fewer. An overhang of 3.5 Gt sits unused. The result is a market for lemons, where cheap and doubtful credits crowd out the good ones.
• Symptom 3. A narrow set of flagship projects accounts for most retirements, and demand stopped growing after 2023.
• Symptom 4. Prices split by project type reveal not one market but two (Figure 2). The accounting treats every credit as the same good, one credit for one tonne, and a single price should therefore prevail. It does not.

Figure 1 VCM project supply by crediting-period start year

Figure 2 Average prices by project type

Note: Each index is defined as the top 20 project by retirements in each category. AlliedOffsets500 comprises the top 500 projects by retirements and serves as a benchmark.

The integrity shock of January 2023

Testing whether firms that buy more offsets cut less of their own carbon runs into an old obstacle: what causes what. A firm’s use of offsets and its emissions move together for an obvious reason. The heaviest emitters buy the most credits because they emit so much. Regress one on the other and you learn close to nothing, since the arrow runs both ways at once. Econometricians call this an identification problem. What we need is variation in offset use that has nothing to do with a firm’s own emissions path. The collapse of the market handed us exactly that.

On 18 January 2023, a joint investigation by The Guardian, Die Zeit and SourceMaterial reported that the leading rainforest-offset method was worthless, because the great majority of its credits stood for reductions that never happened (Greenfield 2023). The blow came from outside the firms. Journalists and scientists audited the methods, and nothing the buyers did brought the story on. The market felt it at once. Prices for the affected credits fell sharply, and the long climb in retirements stopped (Figure 3). A joint drop in prices and quantities is the textbook illustration of a negative demand shock. 

Figure 3 Effect on prices (top panel) and quantities (bottom panel) of the integrity scandal of January 2023

Evidence of moral licensing

After the shock, firms that used offsets in 2022 split into two clean groups. The stayers kept buying credits. The exiters walked away. We then follow the emissions of the two groups, and they diverge. Firms that quit offsetting cut Scope 1 emissions, the carbon from their own operations, by roughly 20% more than firms that kept buying. Scope 2, the carbon behind the electricity they purchase, falls only a little. Scope 3, the carbon of their suppliers and customers, does not move at all. This is the pattern you would expect if firms replaced the offsets they had given up with real cuts at home, because Scope 1 is where a firm holds the controls.

Leaving the market remains a choice, so we need something that pushed firms out for reasons unrelated to their own plans. The scandal supplies it. The news gutted forestry credits and left removals and other segments largely untouched. A firm’s exposure therefore depended on the portfolio it happened to hold years earlier, long before anyone saw the story coming. We combine those 2022 portfolio shares with the price collapse in each segment to build a shift-share, or Bartik, instrument. The instrumental variable estimates confirm the result and in fact strengthen it. Taking away access to offsets caused large and real cuts in emissions, and those cuts do not simply reflect who chose to leave.

The final diagnosis

Most work on the voluntary carbon market polices the supply side. Our finding, that the option to offset substitutes for a firm’s own abatement, suggests that the policy conversation should also address what drives demand.

The answer is not to abolish offsets. Aviation, cement and the other hard-to-abate sectors will need them. The task is to design incentives that reserve credits for the emissions a firm truly cannot remove, and stop credits from standing in for cuts the firm could make itself. Our results therefore reinforce the case for reforming carbon accounting, a case already made for mandatory disclosure (Bolton et al. 2021) and for the wider policy mix needed to decarbonise the economy (Bolton et al. 2026).

Source : VOXeu