A spate of recent studies are being used to claim a causal link for companies that offset their emissions between their use of carbon credits and their rate of internal decarbonisation. However, the available evidence tells a different story about whether or not companies exploit carbon markets as a licence to pollute.
In the wake of heavy criticism of the voluntary carbon market, a set of studies have assessed the relationship between carbon credit purchases and the internal decarbonisation of companies. These studies by Sylvera, Trove (now MSCI) and Ecosystem Marketplace conclude that the purchase of credits is positively correlated with internal decarbonisation.
Unfortunately, these findings are being wrongly cited and spun as proof that offsetting does not create a “licence to pollute” because the companies that do so are reducing their carbon footprint faster than companies that don’t. Even the studies themselves heavily imply this, and sometimes state it outright.
However, the publications by Sylvera, Trove and EM, while interesting, neither prove nor disprove any hypothesis that carbon credits offer companies a “licence to pollute”. Firstly, this is because the alleged positive correlation between the two is questionable, and secondly, even if such a correlation exists, it in no way demonstrates that the purchase of carbon credits leads to the reduction in a company’s emissions. As any student knows, causation and correlation are not the same thing.
Spurious correlations and confirmation bias
The most obvious flaw of these three studies is not related to their methodological approach, but to the interpretation of the results. Setting aside the methodological challenges discussed further down, a positive correlation between carbon credit purchases and internal decarbonisation does not prove a causal link.
The question we should be looking at is whether a typical company which purchased carbon credits would have increased its internal decarbonisation efforts had it not been able to buy those credits. Comparing companies that have bought credits to companies that have not bought credits, as the studies do, does not provide an answer to that question, because the two groups of companies may be fundamentally different in ways that affect both credit purchases and internal decarbonisation.
Beyond theoretical arguments, it’s useful to consider what real-world elements could explain this correlation. Here, we cite only a few, and invite further peer-reviewed and independent research into this question.
One element could be financial resources. Companies that are in better financial shape could be more likely to both buy carbon credits and reduce their internal emissions. Would they reduce their own emissions even more if they could not, or choose not to, buy credits? There is currently no good answer to that question.
Checking the climate ledgers
Another element could be sophistication, or experience with environmental markets. Companies that buy carbon credits could be companies that are more likely to buy other types of environmental certificates, such as Renewable Energy Certificates (RECs). RECs are certificates received by power plants that produce renewable electricity, and which companies purchase to claim that they are using renewable electricity in their operations, regardless of the actual source of electricity they use. You can learn more about what they are and why they are problematic here (external PDF).
The purchase of RECs could in turn appear to be a decrease in “internal” emissions – because corporate greenhouse gas accounting rules allow such certificates to be counted as equivalent to buying zero-carbon electricity (see more material on this linked above). Hence, more market savvy companies would be more likely to both buy carbon credits and report supposedly “lower” emissions due to the purchase of RECs.
In fact, the data from the Ecosystem Marketplace report is consistent with the idea that buyers of carbon credits are also buyers of RECs. When RECs are not taken into account, credit buyers in the sample have higher emissions than companies which do not buy credits. When RECs are taken into account, however, credit buyers have lower emissions than non-credit buyers. In other words, for the more technical readers, median emissions in the “credit buyer” group are higher than in the “non buyer” group for scope 1, scope 2 location-based, and scope 3 (indirect emissions), but significantly lower for scope 2 market-based (see table S3 in appendix 2 of the study).
Sectoral dynamics
A third element that could explain the correlation, is the fact that companies in specific sectors might be both more likely to buy carbon credits and decarbonise internally. For example, companies in the services and financial services sectors are likely to be more experienced with market-based approaches. In addition, these sectors tend to decarbonise faster than heavy industry or transport because cleaner alternatives are more freely available.
The Trove research is the only one of the three which provides results disaggregated by sector. Out of 11 sectors, only three show the same statistically significant correlation that is found for the sample across all sectors. This could be due to low sample sizes, as noted by Trove. However, it is noteworthy that among the three sectors for which a significant correlation is found, two of them are the services and financial services sectors, which are comparably easy to decarbonise and familiar with market-based approaches.
A fourth possible factor is that some companies which buy carbon credits do so in response to higher consumer, shareholder and regulatory pressure as well as scrutiny to act to tackle the climate crisis, and hence have higher commercial incentives to decarbonise. The higher levels of internal decarbonisation would be driven by the higher public pressure, rather than by any incentive created from the purchase of the credits.
While scientific evidence on the influence of carbon credits to incentivise actual emission cuts remains sparse, it is important that it continues to be analysed and judged independently for its real merits. The papers from Sylvera, Trove and Ecosystem Marketplace are welcome contributions to the debate, but the current spin on their findings is counter-productive when it comes to seeking rational and accurate answers to the question of what the role of carbon credits in corporate climate action should be.
Sylvera: ‘Carbon credits: Permission to pollute or pivotal for progress?‘
This study only focuses on scope 1 and 2 emissions, which for many companies represent a minor share of their total footprint. Under this limited approach, companies could appear to be decarbonising quickly, despite their full emissions (including scope 3) actually rising. For example, the Sylvera research cites Audi as one of “the fastest decarbonising companies around the world”, claiming that the German automaker reduced emissions at an annual rate of 11% between 2013 and 2021. Unfortunately, Audi does not report its scope 3 emissions, which typically represent the largest share of car manufacturers’ carbon footprint. The company has experienced significant growth in the number of cars produced over the period analysed by Sylvera (2013-2021, (see here and here for example)), suggesting that these indirect emissions probably haven’t fallen and likely have risen.
There is very little information about the sample, and how it was selected (except that the authors state that it is “balanced”). Overall, there are very few details about the analysis. Therefore, the evidence provided in the study is simply insufficient to support the bold claims that accompany it, such as “this study demonstrates that investment in carbon credits doesn’t stop companies from taking meaningful climate action”.
Trove Research: ‘Corporate emission performance and the use of carbon credits‘
This paper is arguably the most comprehensive of the three, but still leaves major questions unanswered, and requires a major leap of logic or faith to connect the scientific findings to the subjective conclusions.
Like Sylvera, the analysis does not cover scope 3 emissions. However, the paper claims to have performed a sensitivity analysis by including scope 3 emissions and to find the same correlation between credit purchases and internal decarbonisation.
The sample includes buyers of credits which have purchased more than 100 credits representing at least 5% of their scope 1 and 2 emissions. This is a very low bar to set. Given that a few buyers on the VCM are responsible for a big share of all purchases (as shown in the Ecosystem Marketplace report, see below), this means that many companies in the sample are likely to have purchased very small volumes of credits. This would inflate the sample with companies that potentially play a very marginal role in the VCM. In addition, the oil & gas sector, typically a large buyer of credits, is excluded from the study, and a note states that prior research from Trove found no correlation between credit purchases and internal decarbonisation for this sector.
Beyond these shortcomings, the study’s findings do suggest a positive correlation between credit purchases and internal decarbonisation. And, although imperfect, the sensitivity checks are helpful in controlling for some elements, such as the omission of scope 3 emissions. However, it does very little to address the bigger concerns related to the conflation between causation and correlation, as noted in the first section of this article. Overall, while the authors draw some interesting conclusions, the study makes some spurious and unsupported claims such as: “the evidence of the last five years strongly suggests that the voluntary purchase of carbon credits provides companies [with] an incentive to accelerate their emission reductions”.
Ecosystem Marketplace: ‘All in on climate: the role of carbon credits in corporate climate solutions‘
This is the only study which provides a (very limited) list of companies included in their sample. Contrary to the other two papers, this study does cover scope 3 emissions. It does not provide a sensitivity analysis, unlike the Trove paper.
Overall, the paper provides a useful analysis and suggests a positive correlation between credit purchases and such factors as emissions disclosure, target setting, and internal reductions. However, it is limited in that it provides very few disaggregated findings (e.g. by sector or region, or looking at large buyers vs. small buyers).
One interesting element of this study is that it provides a snapshot of the 25 largest buyers of credits included in the sample. There is strong representation of airlines, oil and gas majors and other large emitters in the sample. Taken together, the top 25 buyers are responsible for 72% of credits purchased for the volume considered in the study. Notably, many of these companies are far from being climate leaders. We collected emissions data for these top 25 companies for the years 2020 and 2021 (the EM data is for the year 2021). The result was that the top 25 buyers of credits, covering a majority of overall credit purchases, were on average increasing their emissions (average: 10.2%; median: 1.0%).
Of course we cannot conclude from this whether or not the study’s findings hold for the companies responsible for the overwhelming majority of purchases of carbon credits. It could be that the top 25 non-credit buyers are also performing very poorly compared to the rest of the sample. Still, even if the overall finding holds for this group of major buyers (which is doubtful), it would mean that these companies are simply a little less bad than non-buyers.
The EM report is somewhat more nuanced in its conclusions compared to the other two studies, and makes statements such as “the data do not broadly support the perception that carbon credits are being used to delay or avoid meaningful action on climate”. That may be true but neither does it support the idea that credits are associated with increased ambition..
