We cannot offset our way out of climate change

Carbon offsets have serious issues. Is it even possible to fix them?

Offsets are an increasingly popular way for companies to reach net-zero goals, but many of them just don’t work.

Thousands of companies have established net zero goals, and many use carbon offsets—for example, paying to protect forests—to bolster their claims about reducing emissions. Shell, for example, spent $26 million on things like tree-planting projects last year and plans to use nature to “mitigate” 120 million metric tons of CO2 that it emits each year by 2030.

“we cannot offset our way out of climate change”

In theory, these kinds of carbon offset programs—which includes everything from extra funding for forests to capturing methane at landfills to improving sustainability in agriculture—seems like a good thing. But the system has some obvious flaws. “We cannot offset our way out of climate change,” John Oliver said in the latest episode of Last Week Tonight, which spent 23 minutes outlining some of the problems with offsets.

Carbon offsets have been around for decades, but the voluntary market ballooned to $2 billion in 2021, quadruple the size of the previous year. They have serious limitations, beginning with the fact that some companies, like Shell, plan to use them instead of cutting their own emissions. It’s also hard to prove that a particular offset project is actually helping and difficult to show that the benefits will last—some forests that were protected by carbon offsets have already burned in wildfires, releasing the carbon that had been captured in the trees.

At the same time, there’s a surge of efforts to improve the system, including new standards for how companies use carbon credits and new approaches to designing and measuring the offset programs themselves.


It’s often hard to prove that offset programs made something happen that wouldn’t have happened anyway, a factor called “additionality.” JPMorgan Chase, for example, spent nearly $1 million to help conserve a forest in Pennsylvania that had been a bird sanctuary since the 1930s—and arguably wasn’t in much danger of large-scale logging. American Carbon Registry, the organization that verified the carbon credits, says that because there weren’t previously logging restrictions in place, the sanctuary might have turned to timber harvesting if it later needed cash, since logging is common on nearby land. The offset program added new legal protections, so logging can now only happen in a limited way. But it’s difficult to model the counterfactual: What would have happened at the sanctuary if the offset program hadn’t existed?

“Is it possible to design a system that creates really clearly additional credit? Maybe, but it’s really, really hard,” says Freya Chay, a project manager at CarbonPlan, a nonprofit that studies the impact of climate actions. It’s also difficult, she says, to design a system that can’t sometimes be gamed by less scrupulous organizations trying to sell carbon credits.

One study of some voluntary offset projects in the Amazon rainforest found “no significant evidence” that the projects had actually slowed down the loss of trees. Another analysis of California’s $2 billion forest offsets program found that 29% of the offsets were overestimating the benefit to the climate, meaning that polluters buying those credits got away with emitting more CO2.

The challenge isn’t limited to forests. A working paper of more than 1,000 wind farms in India that were funded by carbon credits found that at least 52% of the projects very likely would have been built even if the carbon credits hadn’t existed. And while the number of studies finding issues in the system keeps growing, there are few, if any, studies proving that particular offset projects have been a clear success. “I’m not seeing any examples of carbon offsetting programs that are working well,” says Raphael Calel, a public policy professor at Georgetown University who coauthored the wind farm study.

It’s also difficult to prove that the benefit from a particular project will last. In forest programs, trees can be lost to wildfires, or disease, or, if the right protections aren’t in place, might later be logged. Projects use “buffer pools” as insurance, protecting extra trees in case some are lost. But a study of California’s buffer pool found that wildfires have already destroyed nearly all of the offsets that were supposed to last for the rest of the century. For a buffer pool to work, it would have to be designed differently, better taking into account the impact that climate change is having on forests.

There’s also a bigger challenge: Protecting a forest, and other impermanent offsets, aren’t truly equivalent to emissions from fossil fuels. Some programs, like California’s, call a forest program “permanent” if it is meant to last for a century. “There’s a huge value judgment there that says we’re going to consider 100 years of storage equivalently valuable to your emissions, which impact the climate for literally millennia,” Chay says. (To make things more complicated, it’s also difficult to estimate precisely how much carbon a forest stores.)


Multiple efforts are underway to try to improve the carbon offset system, both in terms of how projects are created and measured and how companies make use of offsets. The Science Based Targets Initiative, a group that has helped more than 3,000 companies set voluntary climate goals, has a new net zero standard that requires most companies to cut their own emissions deeply—in most cases, at least 90% by no later than 2050. Buying carbon credits is additional, but can’t replace the first step.

The net zero standard makes it clear that offsets “must be on top of, not instead of, a company’s own deep emission cuts,” says Rosie Williams, a communications manager for the Science Based Targets Initiative. Under the standard, the 10% or less of a company’s emissions that can’t be eliminated by 2050 will have to be offset by permanent carbon removal—projects like sucking CO2 out of the air and storing it underground. Some companies, like Microsoft and Stripe, are already strategically investing in new carbon removal tech.

Another group, VCMI (the Voluntary Carbon Markets Integrity Initiative), is working on a new standard for how companies can talk about carbon credits. “We are trying to create a robust science-based benchmark for comparing claims by corporates and other non-state actors about the greenhouse gas mitigation actions,” says Mark Kenber, executive director of VCMI. Right now, he says, because there’s no global framework, it’s difficult for consumers or investors to tell what’s greenwashing and what isn’t.

Organizations that help create offset projects, like The Nature Conservancy, continue to work on ways to improve the quality. To better understand how much additional benefit comes from a project, the organization is working with partners on large-scale tests. “We basically set up a giant landscape-scale experiment where you have a ‘control’ and a ‘treatment,’” says Campbell Moore, who leads a global carbon markets team at The Nature Conservancy. An offset project, with financial incentives to better manage the forest, is compared to similar properties nearby without the same program in place, to see what happens in real life.  “It’s moving from that static baseline, where you’re measuring your additionality against a computer model, to measuring your additionality against real, extremely analogous forests in real time,” he says. The nonprofit also created its own internal standard of review.

Some companies are trying new approaches, like NCX, which pays private landowners to avoid logging based on a short-term forecast of the current economic conditions for selling lumber and a prediction of what that landowner might do without the incentive in place. But because the program pays landowners to avoid logging over one-year periods, critics say that it risks long-term losses. (The model is also new and proprietary, and hasn’t yet been proven to work.) The company argues that it can still help. “One of the things that that we know is that urgency matters,” says Zack Parisa, CEO and founder of NCX. “We see it in every single one of these IPCC reports that that are rolling out. So what we do this year matters. And what we do this year, and the scale at which we’re able to do it, defines what’s possible going forward.”

It’s important to find ways to fix any issues in the market rather than giving up on offsets completely, argues Frances Seymour, a senior fellow at the World Resources Institute, who also serves on the board of a global program to reduce deforestation and degradation.

“We know that tropical forests are disappearing quickly,” she says. “There isn’t an obvious alternative source of finance. We do have some preliminary evidence that the prospect of this market finance can incentivize governments to start doing something different. And if we don’t come up with a way to manage those risks, we almost certainly will lose those forests, which we know are critical to meeting the Paris targets. There’s no sort of IPCC scenario that allows us to keep the temperature at a reasonable level without not only stopping tropical deforestation, but reversing it.”

The organizations setting up offset programs need to show that they work, says Calel. “I think that if anyone wants to do this now at any scale, they should assume the burden of proving out that the way they’re doing it actually works,” he says. “Because I’ve seen so many examples of really smart people trying these things and not succeeding.”


Adele Peters is a staff writer at Fast Company who focuses on solutions to some of the world’s largest problems, from climate change to homelessness. Previously, she worked with GOOD, BioLite, and the Sustainable Products and Solutions program at UC Berkeley



Systematic over-crediting
of forest offsets

from (carbon)plan APR 29 2021

Carbon offsets are widely used by individuals, corporations, and governments to mitigate their greenhouse gas emissions. Because offsets effectively allow pollution to continue, however, they must reflect real climate benefits.

To better understand whether these climate claims hold up in practice, we performed a comprehensive evaluation of California’s forest carbon offsets program — the largest such program in existence, worth more than $2 billion. Our analysis of crediting errors demonstrates that a large fraction of the credits in the program do not reflect real climate benefits. The scale of the problem is enormous: 29% of the offsets we analyzed are over-credited, totaling 30 million tCO₂e worth approximately $410 million.


Summary of results from our analysis of crediting error, in terms of net over-crediting, percentage relative to the projects we analyzed, and value in dollars assuming a credit value of $13.67. Each credit represents 1 tCO₂e. Ranges report 5th and 95th percentiles of a bootstrapped distribution forming a 90% confidence interval.

This article provides an overview of how we identified crediting errors in California’s offsets program. For a deeper dive on our methods and analysis, you can read our preprint. To better understand its implications, you can read a story by Lisa Song (ProPublica) and James Temple (MIT Technology Review) that covers and contextualizes our findings. Finally, you can browse an interactive online map of the projects we analyzed, or download the open source data and code that underlies our analysis.


Carbon offset programs issue credits to projects that purport to avoid greenhouse gas emissions or remove carbon dioxide from the atmosphere. For example, an oil refinery that is subject to an emissions limit might purchase an offset credit issued to a forest owner who agrees to reduce or delay a timber harvest. The refinery can then pollute more, claiming the avoided forest emissions as compensation.

California’s offsets program plays a central role in the state’s prominent cap-and-trade program. While it is open to many kinds of offset projects, most credits come from forest projects, which can take place anywhere in the continental United States and southern Alaska. You might think these projects involve growing new forests, but the vast majority instead involve a practice called “improved forest management” (IFM). An IFM project claims to increase forest carbon storage through changes in existing forest management practices, such as increasing the length of timber harvest rotations.


California’s forest carbon offsets program by the numbers. Each credit is worth 1 tCO₂e. We analyzed 65 IFM projects for which sufficient public data were available, totaling 102 million upfront IFM credits. These represent about two-thirds of all forest offsets, and about one half of California’s entire offsets program.

The critical aspect of California’s forest offsets program is that it awards large volumes of credits at the start of a project when carbon stocks exceed regional averages. These “upfront” credits to IFM projects are responsible for more than half of the total carbon offsets program, and more than two-thirds of all forest credits.

How are these credits awarded? Projects provide “baseline” scenarios that are meant to represent the average amount of carbon that would remain under a typical harvest scenario. The difference between the initial carbon and this baseline determines the credits awarded. To prevent unrealistic baseline scenarios, the protocol requires that the average carbon stored in a baseline scenario stays above a value called “common practice,” which is defined by the average regional carbon stocks from putatively similar forest types.


IFM projects are awarded upfront credits based on the difference between “initial carbon” stocks and the 100-year projected “baseline average.” Under protocol rules, baseline averages cannot be lower than common practice. Thus, erroneously low estimates of common practice can lead to over-crediting.

As it happens, about 90% of projects report baseline averages that are equal to or within just 5% of the minimum common practice number. Thus, crediting is determined almost entirely by the value of common practice — and if common practice is set too low, that means projects are getting excess credits that do not reflect real climate benefits.

Analysis of crediting error

To investigate potential crediting errors, we developed a novel datasetby digitizing public offset project records, most of which currently exist only as PDFs and, to our knowledge, have never been comprehensively analyzed. These records contain each forest project’s tree species composition.

To test the integrity of California’s program, we asked how well each project’s common practice number represents typical carbon stocks for similar forests. As explained further below, the California program uses broad, regional averages that fail to distinguish between species. In contrast, our new database allowed us to estimate typical carbon outcomes across only those forests with similar species. If common practice is set too low, that implies over-crediting; and if it is too high, that implies under-crediting.

We found evidence that the vast majority of projects were over-credited: for these projects, common practice values are systematically low because they reflect averages based on dissimilar species types. As a result, projects received more credits than they would have under a more ecologically accurate and robust definition of common practice.

For example, in the “Southern Cascades” region of California, the common practice numbers used in the program average together temperate, carbon-dense forest types like Douglas Fir (Pseudotsuga menziesii, average 122.5 tCO₂ / acre) and Tanoak (Notholithocarpus densiflorus, average 192.4 tCO₂ / acre) with less-carbon-dense forest types that occupy more arid niches, like Ponderosa pine (Pinus ponderosa, average 60.4 tCO₂ / acre).11Common practice numbers use tCO₂ / acre, converting the C stored in trees’ biomass to CO₂. Offset credits are denominated in terms of tCO₂e because they function in a multi-gas policy environment, where non-CO₂ gases are converted to CO₂e using 100-year global warming potentials from the 2013 IPCC assessment reports. The applicable GWP for CO₂ is 1, so only the units change when converting between CO₂ and CO₂e in this policy environment.

Comparing project carbon against this average causes projects like ACR189, which is located in Northern California and is composed primarily of Douglas fir (26% of basal area) and Tanoak (49% of basal area), to receive substantial credits simply due to a mismatch between the species in the project and the species included in the regional average. If we instead compare ACR189 against Douglas Fir and Tanoak — a more ecologically robust comparison — we find the project was over-credited by 135,869 tCO₂e (90% CI: 85,481-185,917 tCO₂e), or 50.1% (90% CI: 31.5-68.6%) of its total credits.22For all estimates, we report the median (50th percentile) of a bootstrapped distribution alongside 5th and 95th percentiles in parentheses, forming a 90% confidence interval.

But ACR189 wasn’t an exception. We found this same pattern over and over again.

To quantify these errors systematically, we replaced projects’ common practice numbers with an independent, species-specific estimate. We then used the protocol rules to recalculate how many credits each project should have received using this more ecologically robust approach. (We also checked to make sure we could accurately reproduce the most recent common practice numbers and the number of credits projects actually received, in order to be confident in our ability to estimate any over- or under-crediting.)


We estimate crediting error by re-calculating the number of credits that would have been awarded to forest offset projects with a more ecologically robust measure of common practice. The extent of crediting error is shown as a percent of each project’s total credits (top) and in units of tCO₂e (middle). For comparison we also show the total credits awarded to the project (bottom). Use the menu to sort by the different metrics, and select individual bars to see more info for that project.

Our analysis relied on the digitized project records described above, as well as public data from the US Forest Service Forest Inventory Analysisprogram and the open source rFIA package. Our methods are described in detail in our preprint and all of the code and additional dataunderlying our analysis is open source and fully reproducible.

Across the program as a whole, we estimate net over-crediting of 30 million tCO₂e total (90% CI: 20.5 to 38.6 million tCO₂e) or 29.4% of the credits we analyzed (90% CI: 20.1 to 37.8%). At recent market prices of $13.67 per offset credit,33California Air Resources Board, Summary of Market Transfers Completed in Q4 2020 these excess credits are worth $410 million (90% CI: $280 to $528 million) — and likely more, as market prices would rise if market regulators took steps to correct for over-crediting.

A key feature of our study is that it does not depend on counterfactual analysis. In general, offsets must reflect “additional” climate benefits above and beyond what is expected under business-as-usual conditions. Claims about the additionality of entire projects are important to consider but difficult to evaluate quantitatively because counterfactual scenarios cannot be observed directly. In contrast, our analysis uses revealed program outcomes to directly estimate crediting errors.

The problem with averages

The fundamental challenge with awarding upfront offset credits lies in defining an ecologically robust point of comparison. The California program aggregates forest data across both species and geographic regions, creating ecologically inappropriate points of comparison for awarding most of the credits in the program. Our results suggest that these protocol incentives have led to widespread “adverse selection”, with projects preferentially located in forests where carbon stocks naturally exceed those coarse, regional averages.

This is best illustrated in the spatial pattern of projects in the Southern Cascades region described above. This “supersection” — the California program’s term for the large regions over which carbon averages are calculated — is actually composed of three smaller subregions. The subregion on the western edge features relatively wet, carbon-dense forests. But this subregion is combined with two others, which have drier and less-carbon-dense forests. Because the common practice averages across all three subregions, an “average” forest in the western subregion is automatically eligible for upfront credits — simply due to a statistical artifact, and without creating any real benefits to the climate. Almost all of the supersection’s projects cluster within this area, where they benefit from a statistical error and inappropriately enable regulated polluters to increase their emissions.


In the Southern Cascades supersection, three subregions (or “ecosections”) with distinct local carbon patterns were averaged together to yield a common practice number that distorts ecological reality. The number in each subregion shows the relative carbon compared to the supersection average, in units of tCO₂ per acre. The western-most carbon-rich subregion (green) contains almost all of this supersection’s offset projects (), which earn credits simply by having forests with higher carbon levels than the supersection average. Select a project () to see details including ID, developer, and our estimate of crediting error.

While the Southern Cascades is an extreme example, any form of averaging creates an opportunity for adverse selection. Biogeographers have long understood the challenge of drawing firm boundaries around ecological regions or categories of species. While boundaries help communicate with outside audiences, border regions are complex areas where the characteristics of separate regions interact. Any program based on such boundaries must be continually reviewed to ensure the kind of adverse selection described here is not taking place.

Why this matters

Rather than improve forest management to store additional carbon, ecological and statistical flaws in California’s offsets program create incentives to generate credits that do not reflect real climate benefits. Our work shows that these concerns have manifested in practice at a large and systemic scale. Looking more broadly, our results illustrate how protocols with easily exploitable rules can undermine policy objectives and highlight the need for stronger governance in carbon offset markets — based on rigorous, transparent, and independent science.

It’s important to note that while some of the problematic outcomes we document likely reflect intentional strategies to take advantage of poorly designed program rules, our results don’t assume or depend on market participants’ intentions. It doesn’t matter whether landowners or project developers intended to take advantage of these rules, or simply benefited from them without awareness.

While our analysis is critical and the results are disappointing, we believe forward progress only begins by understanding our mistakes — so that we can do better in the future.

Why open science

Our approach to both conducting and releasing this work fully embraces the growing trend toward open science, which differs from traditional academic research.

We are sharing our work now, as a preprint, rather than waiting months or years for publication in a peer-reviewed journal. We are taking this approach both to address an urgent set of policy-relevant concerns and so that we discuss these issues in the open, rather than behind closed academic doors.

In order to bring this discussion into the open right away, we are making all of our materials fully public and reproducible: the digitized project database, all additional data used throughout our analysis, code to generate figures from those data, and the complete code base used for our analysis. At any point now or in the future, the entire community is welcome to review our work on its merits. We look forward to further improving our analysis based on the criticisms and collaborations that come from open science. If you have feedback, please open an issue on our GitHub repository.

We are also committed to subjecting our work to critical review from our peers. We have incorporated review in several ways prior to public release and are taking additional steps moving forward. First, we implemented our own round of independent review with domain experts. Second, we shared early versions of our work with two journalists who, in the process of developing their own story, asked independent researchers for comment and asked us questions raised by feedback from affected parties — including the California Air Resources Board, as well as nonprofit organizations, project developers, and individual landowners that participate in the program. Although an adversarial review process involving program beneficiaries and conducted with journalists as intermediaries is different from a traditional journal publication process, we believe its quality control was just as good if not better. Finally, we have also submitted our work to an academic journal and are committed to seeing it subjected to that form of peer review.

It is precisely on issues of such critical importance to the public where we believe this open, transparent approach to science and government accountability matters most.


Grayson, Danny, Jeremy, and Joe designed the research; Grayson digitized the project report data; Grayson, Danny, Jeremy, Joe, and Barbara performed the research and analyzed the data; all authors contributed to interpreting the results and writing the paper. Jeremy developed the interactive graphics with input from Jonny Black of Ordinary Things.

A journal article on this work is published at Global Change Biology and can be cited as:

G Badgley, J Freeman, J Hamman, B Haya, A T Trugman, W R L Anderegg, D Cullenward (2021) “Systematic over-crediting in California’s forest carbon offsets program” Global Change Biology doi:10.1111/gcb.15943

Please cite this web article as:

G Badgley, J Freeman, J Hamman, B Haya, A T Trugman, W R L Anderegg, D Cullenward (2021) “Systematic over-crediting of forest offsets” CarbonPlan https://carbonplan.org/research/forest-offsets-explainer


CarbonPlan received a grant from Microsoft AI for Earth to support the portion of work that involved digitization of offset project records. All other research design and data analysis was performed using CarbonPlan’s unrestricted funding. No one except the authors of the paper exercised control over the research process or products. The authors are solely responsible for the content of this writeup, which does not reflect the views of any other individuals or organizations.

Grayson Badgley is a Postdoctoral Scientist at Black Rock Forest and Columbia University, Barbara Haya is a Research Fellow and Director of the Berkeley Carbon Trading Project at UC Berkeley, Anna Trugman is a professor at UC Santa Barbara, and William R.L. Anderegg is a professor at the University of Utah.

Article text and figures made available under a CC-BY 4.0 International license.

Pledge Your Vote Now
Change language