How does enforcement of an ETS work? And why is it that firms still comply at nearly 99% when enforcement of violations seems low. Those are the questions I tackle in my new podcast episode with Frank Venmans. He, together with Raphael Calel and Antoine Dechezleprêtre wrote a paper “Policing Carbon Markets”, which studies compliance and enforcement in the EU Emissions Trading System.
And their headline result is striking: compliance is extremely high, at around 99%. But the standard economic explanations fail to explain much of the variation in compliance. So what is going on? Well, tune in to find out!
Also, what I think the paper does really well is show a less glamorous but very important lesson: carbon markets are technically complex. Registries, reporting systems, verification, financial regulation, and data management all matter enormously for whether the system works in practice!
Paper: https://www.tandfonline.com/doi/full/10.1080/14693062.2025.2464699 , and
Frank Venmans’s site: https://www.fvenmans.com/
Transcript
Arvid Viaene:
After the past two episodes on the social cost of carbon, I want to return to one of our main tools for fighting climate change: carbon markets, and more specifically, the European Union Emissions Trading System.
We have already covered the EU ETS quite a bit on this podcast — its history, how it works, and possible future developments. But one element we haven’t tackled yet is how to deal with non-compliance by firms. You can set a cap on emissions all you want, but if firms don’t follow the rules, it might all be for nothing.
Carbon markets are also relatively new, so this problem can be challenging. When I discovered a paper called Policing Carbon Markets that tackles exactly this question, I was very excited. The paper was published in Climate Policy in 2025 and presents a comprehensive analysis of compliance in the EU Emissions Trading System, combining some impressive data collection with detailed empirical analysis.
I’m delighted to have one of its authors with me today, Frank Venmans. Frank is an Associate Professor and Deputy Research Director at the Grantham Research Institute at the London School of Economics. He studies how we can design effective climate policy under uncertainty, with a particular focus on the path to net zero, technological change, inequality, and climate risk. His work combines economic modeling with empirical research on carbon markets, climate litigation, economic growth, biodiversity, and carbon removals. With a background in forestry and soil management, he brings a broad perspective connecting economics, climate science, and policy.
Frank, welcome to the podcast.
Frank Venmans:
Hello, Arvid. Thanks for inviting me.
Arvid Viaene:
I’m very excited to have you on. Sometimes I become curious about a particular question, then discover a paper tackling exactly that issue, and I thought this was a really great paper. So to start, what question are you trying to answer?
How High Is Compliance in the EU ETS?
Frank Venmans:
The first question is simply: how good is compliance? What is the compliance rate? And there we observe relatively high compliance.
The second question is what can explain why some countries perform better than others. What mechanisms do regulators use? How do those mechanisms differ between countries, and can those differences explain some of the non-compliance we observe?
Then there is the question of enforcement: what actually happens when there is non-compliance? The general finding is that the consequences are not particularly dramatic in practice. That creates something of a contradiction. In a standard economic model, we would think of a company as maximizing profits. If there is an opportunity to cheat, it weighs the expected consequences — the punishment multiplied by the likelihood of getting caught — against the gains from cheating.
But that model turns out not to be especially well supported by the data. So we are left with something of a puzzle: very high compliance alongside relatively limited enforcement. We haven’t completely solved that puzzle, but we investigate a number of mechanisms that might explain it.
Arvid Viaene:
Could you say more about the actual rate of compliance you found?
Frank Venmans:
On average, 98.8 percent of companies are compliant, excluding 2005. The year 2005 was the first year of the system, and compliance was much worse. Around 22 percent were non-compliant. But this was effectively a trial phase, and the system had been established very rapidly.
The relevant EU legislation came in 2003, and trading started in 2005. That is also why it was smart to make the first phase largely separate from the later phases. You couldn’t bank allowances left over in 2007 and use them in 2008. Whatever went wrong during the first phase was therefore largely contained within that initial period.
Once you move beyond that phase, though, compliance is extremely high.
Arvid Viaene:
So essentially 99 percent of firms comply. Were you surprised by how high that number was?
Why Complying With an ETS Can Be Easier Than With Traditional Regulation
Frank Venmans:
We didn’t necessarily have a strong prior before starting the project.
More generally, I think large industrial companies tend to have a preference for complying with environmental regulations. But there is also something specific about the EU ETS.
Imagine a command-and-control regulation where a facility is not allowed to emit particulate matter above a particular level. Something might go wrong with an industrial oven, for example, and suddenly emissions exceed the regulatory threshold. There may not be an easy immediate solution.
With the EU ETS, the situation is different. If you emit more than your free allocation, you can simply buy additional allowances. That means non-compliance is much more clearly a choice. I used the word “cheating” earlier, although when you look closely at the data, not all non-compliance really qualifies as cheating.
We found examples of companies that simply did not know they were included in the system. I spoke with the Italian regulator, for example, and they described literally calling companies and telling them: “You know you’re part of the carbon market? You need to report your emissions.”
Those firms might initially be classified as non-compliant. They would eventually produce their emissions report and surrender the allowances, but perhaps too late. I wouldn’t describe that as cheating. It is more a case of poor information or administrative sloppiness.
Arvid Viaene:
And that makes the initial trial phase especially useful. Once those informational problems are resolved, firms seem to comply extremely well. I also hadn’t really thought about the difference you mentioned. With conventional regulation, if something goes wrong at your factory and emissions exceed the standard, switching technology immediately may be impossible. Under an ETS, however, you can simply buy additional allowances.
Frank Venmans:
Exactly.
Take the example of an industrial oven. If something goes wrong and the oven is not at the right temperature, it can generate significantly more pollution. There is an optimal temperature and oxygen input, much like a household heating system: if there isn’t enough oxygen, you can get soot coming out of the chimney.
Under conventional regulation, that incident could immediately put you out of compliance. Under the EU ETS, you can buy the additional allowances necessary to cover the extra emissions. In that sense, the system makes compliance easier.
The Gap Between Theoretical and Actual Fines
Arvid Viaene:
That brings us nicely to fines. Compliance is very high, but you also looked closely at what happens when firms do not comply. What did you find?
Frank Venmans:
First, I should make the compliance figure slightly more precise. The high compliance rate refers to the official compliance code regulators assign to companies. But in the underlying data we sometimes observe a company receiving a compliant code even though it apparently did not surrender enough allowances to cover its emissions on time.
We can see that it makes up the difference the following year. So perhaps it was two weeks late or a month late, and the regulator essentially said: “That’s fine. The substance of the requirement has been met; the administration was just late.” For our analysis, if the regulator classifies the firm as compliant, we treat it as compliant.
For the officially non-compliant firms, however, we calculate what the theoretical fines should have been. The EU-wide penalty was €40 per excess ton during the first phase and €100 per ton thereafter.
And importantly, paying the fine does not remove the obligation to surrender the missing allowances. You still have to provide those allowances later. So deliberately failing to comply can be extremely expensive.
When we calculate the theoretical fines based on observed non-compliance, we get approximately €13 billion. But we can only find evidence of roughly €2.1 billion actually being collected. So less than 20 percent of the theoretical fines appear to have been paid.
Arvid Viaene:
Why is there such a large gap? Do regulators have discretion over whether to impose the fine?
Frank Venmans:
One recurring issue in the interviews was what happens when companies close.
Suppose the regulator discovers non-compliance and tries to contact the company, but the company has already disappeared. There is nobody left in the office.
The regulator can block the company’s account. In the best case, there are still allowances in that account that can cover the emissions. But the company is still technically non-compliant.
In the worst case, the company sold all its remaining allowances while trying to stay afloat and then disappeared. That is true non-compliance in the sense that its emissions were never covered by allowances.
The regulator could begin a judicial process, but it would be pursuing a company that no longer exists. That can take years and may recover very little. So this is one reason some theoretical fines are simply never collected.
Are Firms Reporting Their Emissions Truthfully?
Arvid Viaene:
Another part of the paper I found really interesting concerns whether reported emissions themselves are truthful. It is one thing to observe a 99 percent compliance rate, but that only means something if the emissions firms report are accurate. You use some interesting statistical tests to investigate that. Could you explain them?
Frank Venmans:
One test we use is based on Benford’s Law.
Benford’s Law predicts the frequency with which different leading digits should appear in naturally occurring numerical datasets. Because our emissions data include everything from relatively small companies to extremely large firms, we would expect the leading digits in truthful emissions reports to follow that distribution.
The digit one, for example, appears as the first digit more frequently than nine. The law gives us a precise prediction for how frequently each leading digit should occur.
Manipulated data often deviate from Benford’s Law because people are not very good at inventing numbers that reproduce the natural distribution of digits. So if the reported emissions were being systematically manipulated, we might expect to see deviations.
The general result is that Benford’s Law fits the data well. We therefore do not see broad evidence of fabricated or manipulated emissions reporting.
There is one suggestive result involving smaller firms. Large firms are required to meet fairly strict verification standards, with around a 1 percent maximum measurement error. Smaller companies can use methods that permit larger measurement errors, potentially up to around 4 percent, in order to avoid excessive verification costs.
After the non-compliance penalty increased to €100 per ton, the smaller companies show somewhat larger deviations from Benford’s Law. We cannot perform a decisive statistical test on that comparison, so I would not interpret it as proof of manipulation. But it is suggestive that something may be happening.
Still, the overall picture is that the law fits well and we do not find general evidence of widespread misreporting.
Does the Choice of Emissions Verifier Matter?
Frank Venmans:
We also look at differences in verification rules across countries.
Emissions reports are verified by a third party — perhaps a company such as Bureau Veritas, Deloitte, or another accredited verifier. The regulated company generally pays the verifier. That potentially creates a conflict of interest. Because the company is paying the verifier, you might worry that the verifier has an incentive to keep its client happy.
Some countries therefore allow the regulator to appoint the verifier rather than letting the company choose. We test whether compliance improves in countries where regulators can appoint verifiers. It does not seem to.
We also examine whether emissions reports appear to bunch around the level of free allocation. During much of the period we study, many firms received free allowances. You can think of that free allocation as a kind of benchmark. If firms and verifiers were collaborating to underreport emissions, you might expect reported emissions to cluster just below the number of free allowances, because then the firm would avoid having to buy additional allowances.
We test for this bunching. We look at how many reports fall just above versus just below the free-allocation threshold. We do not find strong evidence of manipulation.
We also test other institutional differences. In some countries, regulators can train and supervise verifiers. That does not explain much variation either. In some countries, regulators can conduct their own on-site inspections. Again, we do not find that this explains much of the variation in compliance.
Do Larger Fines Actually Increase Compliance?
Arvid Viaene:
You also test the classic economic model directly: if the probability of being punished is greater, firms should be more likely to comply. What do you find?
Frank Venmans:
Yes. We wanted to test the canonical model in which a company maximizes profits. Under that framework, if I cheat, I avoid paying the carbon price. But the expected cost of cheating is the probability that I get caught multiplied by the fine I would pay if caught.
So we examine whether countries where non-compliant firms are more likely to be fined have higher compliance. They do, slightly. But the effect explains very little of the overall variation — perhaps around 5 or 10 percent. There is a great deal of compliance that the standard profit-maximizing model does not explain.
The Harrington Paradox
Arvid Viaene:
And that is where the paper becomes especially interesting. You have extremely high compliance, but when you examine the classic explanation — fines multiplied by the probability of getting caught — it explains only a small part of what you observe. How did you approach the remaining puzzle?
Frank Venmans:
This puzzle is actually not new. It is known as the Harrington paradox, after a 1988 paper studying water pollution regulation in the United States.
The paradox is essentially the coexistence of high compliance and low enforcement. Fines are relatively rare, and yet most companies comply with environmental regulations.
Economists have proposed several possible explanations over the years, and we test a number of them in the context of the EU ETS.
One possibility is naming and shaming. If a regulator publicly identifies a company as a polluter or a non-compliant firm, the resulting reputational damage might give companies an additional incentive to comply.
Interestingly, naming and shaming is explicitly part of regulatory policy in some countries but not others.
Arvid Viaene:
What does it mean for naming and shaming to be an explicit policy? Is it literally written into the regulatory approach that the names of non-compliant firms will be published?
Frank Venmans:
Yes.
Every country has to submit reports to the European Commission explaining aspects of how it administers the system. One question asks whether the regulator can appoint verifiers. Another asks whether the country uses naming and shaming for non-compliance.
In practice, a national registry might have a website reporting emissions and identifying companies that are non-compliant. There is already some transparency at the EU level because installations and their compliance information are available through the EU registry. But if you are just one non-compliant installation among thousands, you might reasonably hope nobody notices.
Active naming and shaming — putting the company on a prominent website or mentioning it publicly — is different.
We therefore compare compliance in countries that actively name and shame with countries that do not. Compliance is somewhat higher where naming and shaming is used, but the effect is not statistically significant.
We also test whether the effect is stronger for publicly listed companies, because they might care more about their reputation with shareholders. It is not. We test whether domestic companies with headquarters in the country are more sensitive to naming and shaming. Again, we do not find a stronger effect.
So our conclusion is that the impact is probably quite small, perhaps even zero.
Why the Paradox Remains a Paradox
Arvid Viaene:
So we are back at the Harrington paradox. Compliance is very high, but the enforcement mechanisms themselves do not seem to explain very much of it.
Frank Venmans:
Yes. In a way, the paper treats the Harrington paradox as a genuine paradox because several of its standard explanations do not receive much empirical support in our setting.
Another potential mechanism is that the regulator can punish a non-compliant firm informally in other ways. For example, perhaps a regulator sees that a company has failed to comply and responds by reducing its free allocation in the next phase. There would be no court process or public confrontation. The company would simply know that annoying the regulator has consequences.
We test whether non-compliant firms subsequently receive less free allocation. There is a small effect, but again it is not statistically significant.
That mechanism is also no longer really available because rules governing free allocation have become harmonized across the EU, which is actually a good development.
Another explanation used in traditional pollution regulation concerns targeting firms with particularly high abatement costs. A regulator may know which company will struggle most to meet an emissions standard and therefore inspect that firm more frequently.
That mechanism does not really apply to an emissions trading system because marginal abatement costs are equalized through the market. If your abatement costs are above the allowance price, you buy allowances. If they are below the allowance price, you reduce emissions and potentially sell allowances.
Another possibility is threatening to suspend an operating license. Some regulators have the authority to suspend licenses for non-compliant companies, whereas others do not. We test whether compliance is higher when regulators have this power, and again we do not find a statistically significant difference.
But based on the interviews, I actually think this kind of relationship matters quite a lot.
The Importance of Long-Term Relationships With Regulators
Frank Venmans:
I interviewed companies in industries such as cement, bricks, and lime. These are long-lived industrial facilities. A company may have been operating at the same site for 40 years and may have another 30 years remaining in its quarry.
The regulator is also going to be there for decades. So the firm and the regulator have a long-term relationship. Damaging that relationship over a relatively minor amount of emissions simply is not very smart.
At some point, the company may want to expand production. It may need its environmental permit renewed or extended. It may want approval for additional production capacity. Whenever a firm can keep the regulator happy at relatively low cost, there is a strong incentive to do so.
I think that repeated interaction is probably part of the explanation.
Arvid Viaene:
That makes sense. The standard economic framework almost treats compliance as a one-off decision. But in reality, this is a relationship that lasts for decades. If you are missing allowances for 10 percent of your emissions, you can simply buy them. Trying to hide the problem and risk damaging your relationship with the regulator may not be worthwhile.
Frank Venmans:
Exactly. In economics, we would call this a repeated game. Repeated games make cooperative outcomes easier to sustain.
There are also additional legal penalties countries can impose on top of the standard EU fine. The EU-level penalty is €40 or €100 per ton, depending on the period. But countries can add their own legal penalties for things such as misreporting.
Those additional financial penalties range from zero to as much as €15 million. Prison sentences for misreporting range from no prison penalty at all to as much as ten years.
We examine whether countries with harsher additional penalties have higher compliance. Again, the relationship is weak.
So I have to confess: we have not really solved the paradox. We tested many standard explanations, but we do not find strong statistical evidence for most of them.
Why Repeat Offenders May Be Different
Frank Venmans:
One other possible mechanism concerns repeat offenders.
Under conventional regulation, companies that repeatedly exceed pollution limits may face escalating punishment. The EU ETS penalty is set in law, which limits the regulator’s discretion. But because we observe cases where there is technical non-compliance without a fine being imposed, the regulator may still have some room to say: “This time we will let it go, but next time you will pay.”
That could help explain why firms do not repeatedly offend. The first instance creates a credible warning.
Arvid Viaene:
That brings us to another part of the paper I really enjoyed: the interviews with regulators and enforcement officials. After doing all of the quantitative analysis, what did you learn from talking to the people actually administering the system?
What the Interviews Revealed About Running a Carbon Market
Frank Venmans:
The interviews were very useful and actually quite fun. One of my main goals was simply to understand the data.
I think economists, policymakers, and especially countries currently establishing new emissions trading systems can underestimate how difficult the technical side is. The EU ETS was a European project that needed many countries on board, and not every country was equally motivated or equally well prepared.
Initially, the registries were managed at the national level. That is not necessarily the easiest way to run a large integrated system. You need complicated protocols governing how national registries communicate with the central European system.
France, for example, used an external IT company to help manage its registry. At one point, there was a coding error involving foreign allowances. As a result, a lot of data were missing.
The regulator might still have known whether a company was compliant and been able to assign the correct compliance code, but there were gaps in the underlying centralized dataset.
I also spoke with the regulator in Malta. They spent around two years deciding whether they even needed their own registry because there were only two regulated installations in the country, and creating a registry was expensive.
Eventually they did create one. But when I looked at the EU database, data for those two companies were frequently missing. The Maltese regulator told me: “I have the data. It is in my registry. I can see the allowances, verified emissions, and surrendered quantities.”
So the problem was not necessarily missing information at the national level. It was communication between the national and central registries.
I also performed basic data-quality checks. For example, the database contained cumulative emissions as well as annual emissions, so you can check whether cumulative emissions equal the sum of previous annual observations. Sometimes they did not.
When I raised issues like this with someone managing the central system, they became rather nervous. At some point, they seemed to wonder whether I had an agenda to damage the reputation of the EU ETS.
But I was simply doing ordinary data cleaning.
The broader point is that managing a huge dataset involving many parties is difficult. Once the system moved toward a centralized registry around 2013, it became much easier. There was one manager and one system.
But again, this reflects political reality. You have to convince a large number of countries to participate, and institutional design develops gradually.
What New Carbon Markets Can Learn From Europe
Arvid Viaene:
Given everything you found — very high compliance, relatively low collection of fines, weak effects from many of the enforcement mechanisms, and the technical challenges — what are the main takeaways from the paper?
Frank Venmans:
One takeaway follows directly from the technical issues we just discussed.
Suppose you are Vietnam, Mexico, Indonesia, or another country establishing an emissions trading system. Be very mindful of the technical complexity. You need to think carefully about which institution is going to administer and enforce the system.
If possible, it may help to place these responsibilities within an existing government institution that already has experience collecting money, managing company data, and performing functions similar to those required in tax administration.
Do not underestimate the complexity of building these systems. And do not underestimate the extent to which criminals may try to exploit the fact that you have created a valuable financial asset.
The EU ETS experienced serious problems with VAT carousel fraud. At one point, very large trading volumes were essentially connected to VAT fraud, contributing to billions of euros in lost tax revenue.
Initially, identity checks for people trading allowances were weaker than those applied to conventional financial assets. That problem has largely been addressed. Emissions allowances are now treated under financial regulation much more like stocks, bonds, and other financial assets. Repeating those VAT-carousel schemes would therefore be much harder today.
But the lesson remains: technical and financial-market regulation is crucial.
Why High European Compliance Should Not Be Taken for Granted Elsewhere
Frank Venmans:
The second takeaway is that although we struggle to distinguish the effects of relatively small differences in European enforcement rules, that does not mean enforcement institutions do not matter.
There is research from India, for example, examining the appointment of environmental auditors. That work finds substantial problems when companies themselves choose and pay their auditors. When auditors are assigned independently rather than selected by the regulated company, reporting becomes much more truthful.
We do not find a comparable effect in the EU ETS, but the Indian evidence shows that institutional context matters enormously.
Europe may simply be a relatively easy environment in which to run this kind of system. Institutions are generally strong, corruption is comparatively limited, and companies are accustomed to environmental regulation.
So I would not take 99 percent compliance for granted when designing emissions trading systems elsewhere.
Arvid Viaene:
Exactly. Within the European institutional setting, some of these differences may not explain very much. But in another institutional context they could become extremely important.
Frank Venmans:
Yes, exactly.
Closing Thoughts
Arvid Viaene:
Frank, thank you so much for coming on. I think this was a really interesting deep dive into an extremely important aspect of carbon markets: how to make sure firms actually comply.
Frank Venmans:
Thanks for inviting me.
I’m very motivated to make these results and this research accessible to a wider public because this work should ultimately be useful. Emissions trading systems are expanding. Many new systems are being established in different parts of the world.
The European experience is especially valuable because it was the first very large emissions trading system. It provides an example from which other jurisdictions can learn. We should therefore try to learn as much as possible from that experience.
Arvid Viaene:
Absolutely. Thank you very much, Frank.
Frank Venmans:
Thank you.
