Shifting concerns for the EU ETS: are carbon prices becoming too high?

Carbon prices in the EU Emissions Trading Scheme (EU ETS) have risen from around 5 euro per ton of CO2 in 2017 to above 90 euro in 2021. One probable explanation is the cancellation mechanism implemented along with the Market Stability Reserve (MSR) of the EU ETS in 2018. We identify realistic conditions under which the MSR results in truly massive cancellation of emissions allowances, pointing to the steepness of the emissions pathway over time as essential. A flattening of the emissions pathway implies huge reduction in cumulative emissions, suggesting much higher ETS prices. The concerns about too low and ‘ineffective’ carbon prices may turn into concerns for too high prices. The results have important ramifications for planned revisions of the EU ETS.


Significance statement
Recent changes in the design of the EU Emissions Trading Scheme (EU ETS) will cause large amounts of allowance cancelation in years to come. Although the academic literature has widely acknowledged this basic fact, there currently does not exist an overarching picture of what drives the exact extent of cumulative cancellation in the EU ETS. In our paper, we offer a constructive analysis that reveals one crucial driver: the time structure of demand for emissions. Being both the EU's major climate policy and the world's largest (in terms of value) and most ambitious emissions trading scheme, a clear understanding of how emissions in the EU ETS are determined is key in improving existing, or developing new, policies to fight climate change.

Main text
The European Union's Emission Trading System (EU ETS) is the flagship EU climate policy, covering roughly half of the EU's CO 2 emissions (Bayer and Aklin 2020). The EU ETS has had a tumultuous past with volatile and mostly low prices, especially in the wake of the 2008 financial crisis. In the years 2012-2017, the price hovered consistently below €10 per ton of CO 2 , falling short of even the most conservative estimates for the social cost of carbon (Pindyck 2019). Then prices started rising-a trend that gained momentum despite the COVID-19 pandemic and led to record-breaking prices throughout 2021. What happened?
There probably are three main explanations. The price spike throughout 2021 is likely related to high natural gas prices (as opposed to very low gas prices in 2020), inducing substitution towards coal in power production and hence increased demand for emissions allowances. The second explanation is the EU's more ambitious emissions reduction target for 2030, strengthened from 40% to 55% (vis-a-vis 1990). The third is a key reform of the EU ETS in 2018, when a stabilization mechanism was incorporated into the Market Stability Reserve (MSR). This reform implies that cumulative emissions over time are no longer dictated by predetermined EU ETS supply schemes but depend on ETS market dynamics. The EU ETS thus changed from a system with an exogenous cumulative cap on emissions to one with an endogenous cap (Perino 2018, Rosendahl 2019, Gerlagh et al 2020b The MSR is essentially a vault, taking in allowances as long as the surplus in the market at the end of a year ('total number of allowances in circulation' or TNAC) exceeds 833 million allowancesthe supply of new allowances is reduced by the same amount. When the TNAC falls below 400 million allowances, allowances are released from the MSR. However, when the MSR holds more allowances than a given threshold, the excess is permanently canceled, thus reducing the future supply of allowances and thereby the cumulative emissions cap. This regulatory change is expected to cause massive cancellation of allowances in years to come, implying large reductions in emissions (Perino and Willner 2017, Kollenberg and Taschini 2019, Bruninx et al 2020, Gerlagh et al 2020a, Osorio et al 2021 5 . The EU ETS is not the only cap and trade scheme to use stabilization mechanisms. California's ETS (Borenstein et al 2019) and the Regional Greenhouse Gas Initiative (MacKenzie 2022) also use them. There are two critical differences between both sides of the Atlantic, though. First, US stabilization mechanisms are triggered when allowance prices pass a threshold, while the MSR relies on quantity-based triggers. The US mechanisms thus turn a cap on quantities into a hybrid price-quantity policy, while the EU has doubled down on the quantity aspect (Murray et al 2020) 6 . Second, while the intentions for the EU MSR were to raise prices, California's mechanism has both a price floor and a ceiling; the goal is to stabilize prices in both direction, not just to raise them.
Although several papers build models to simulate the EU MSR cancellation mechanism, previous work such as Bruninx et al (2020) and Osorio et al (2021) have approached this issue from a supply-side angle 4 The question arises, why did not the EU ETS allowance price respond to the reform straight away? We offer two rationales. First, the MSR was prepared for some time but its implementation and details remained conditional upon formal approval by the European Parliament. Moreover, there has been an ongoing discussion in the EU on whether and how to strengthen the MSR rules, in particular the rate of inflow into the MSR (cf supplementary information available online at stacks.iop.org/ERL/17/054018/mmedia). This ambiguity may have dampened firms' initial responses to the reform. Second, if firms maximize profits over a finite and rolling time horizon, delayed responsiveness of the EU ETS market to changing supply-side policies is likely; see Quemin and Trotignon (2019) for such a model. 5 Only allowances in the MSR are canceled. Those held by private parties can be banked without constraint and remain valid. This rule was implemented after Phase I (2005)(2006)(2007) in the EU ETS; allowances allocated in that phase lost validity after the end of Phase I, which led to zero prices when supply exceeded demand. 6 Price and quantity mechanisms are not equivalent. Heijmans (2022), for example, shows that policies aimed at banning future emissions have opposing effects under price mechanisms compared to quantity mechanisms. using market models 7 . We, in contrast, approach it from a demand-side angle in a more stylized way by focusing on a critical yet often neglected driver: the steepness of the emissions pathway over time. We summarize this dynamic in the simplest way possible by addressing a very basic question: will emissions decline more or less gradually over time?
With an exogenous cumulative cap (without a cancellation mechanism), the time-structure of emissions does not affect cumulative emissions. A flattening of the emissions pathway involves lower current emissions combined with a corresponding increase of future emissions. Graphically speaking, the two emission pathways E1 and E2 in figure 1 exhibit the same area under their curves.
Things work differently when the cumulative cap on emissions is endogenous due to the cancellation mechanism. As a flattening of the emissions pathway involves lower initial emissions, this leads to a larger surplus of allowances in early years, by construction causing an increased inflow of allowances into the MSR and thus more cancellation over time; the cumulative supply of allowances, and hence emissions, consequently drop. This is illustrated in figure 1. Both E1 and E3 are emissions pathways consistent with the updated MSR rules as recently proposed by the European Commission (2021b) (E2 is not and only serves as illustration of an exogenous cap). We notice that E3 is not only flatter than E1 (like E2)-it is also substantially lower. Emissions must start much lower-otherwise there will be a shortage in the ETS market in later years because of the cancellation mechanism. In other words, the area under the E3 curve is much smaller than the area under the E1 curve; the difference is caused by the MSR reacting strongly to a flattening of the emissions pathway. While it is not surprising that the areas differ, the size of the effect is remarkable.
To disentangle the effects summarized in figure 1, one may compare figures 2 and 3. These figures present the number of allowances (a) banked (TNAC), (b) entering the MSR, (c) released from the MSR, and (d) canceled in a given year when the emissions pathway declines by 4.5% and 3.0% (of 2021 emissions) annually, respectively. The figures also show the exogenous supply of allowances, i.e. without adjusting for the MSR 8 . The two figures look 7 Bruninx et al (2020) and Osorio et al (2021) assess the sensitivity of emissions in the EU ETS with respect to the reduction over time of the supply of allowances. We instead focus on reductions of the demand for emissions over time. Thus, while these authors consider sensitivity of the MSR to parameters of the EU ETS, we consider sensitivity to market characteristics. 8 There is an annual supply of allowances, which are either auctioned or given out for free to installations. As mentioned before, there is no constraint on banking of allowances. On the other hand, borrowing is not allowed except for a small window between allocation early in year t and surrender of allowances for year t − 1 in April year t (this only applies within each phase though). Scenario E3: emissions decline by 3.0% of 2021 emissions annually and supply follows MSR rules. The scenario E2 keeps MSR retirement the same as in scenario E1. The area between E2 and E3 measures the reduction in cumulative emissions due to extra MSR cancellation caused by the change in the slope of the emissions pathway. E1 and E3 are consistent with MSR rules and also part of figure 4; E2 serves as an illustration. More details about scenarios E1 and E3 are found in figures 2 and 3, respectively. distinctively different. As mentioned, the surplus of allowances (banking) is substantially larger in the 3.0% scenario. This is intuitive: if emissions are to decline less rapidly over time, more allowances need to be saved for future usage. An immediate implication is that the inflow of allowances into the MSR is greater, and therefore so is allowance cancellation. The total number of allowances canceled in the 3.0% scenario is hence substantially larger than in the 4.5% scenario.
To investigate this further, we calculate the cumulative emissions consistent with the EU ETS and updated MSR rules for various slopes of the emissions pathway. Figure 4 shows cumulative emissions and initial emissions (in 2021) for each of these paths. If emissions, and thus demand for allowances, are set to decrease by 4.5% annually (i.e. 4.5% of the 2021 level), cumulative emissions add up to 16 GtCO 2 . If emissions are set to decrease more slowly, i.e. by 3.5% of the 2021 level annually, emissions must start at a much lower level in 2021, and cumulative emissions drop to 9 GtCO 2 . Cumulative emissions are reduced by almost one half in response to a seemingly moderate change of the slope of the emissions pathway.
We emphasize that our results follow mechanically from the rules of the MSR and do not involve any economic assumptions on the EU ETS market (such as price-sensitivity of demand). The MSR is a purely quantity-based protocol for backloading and canceling allowances (one potential exemption being Article 29a of the EU ETS Directive, see below). The time-structure of emissions, together with the rules of the MSR, are all we need to fully determine the outcomes depicted in figure 4. Indeed, the only two assumptions entertained in the analysis are, first, that emissions decline linearly over time, and second, that firms are not allowed to hold allowances beyond 2070 (the supply of allowances drops to zero after 2040). Neither of these are important drivers of our results. We discuss the role played by a linear emissions reduction below. As to the market ending at the latest in 2070, we observe that if the EU ETS were to continue after 2070 (so allowances issued before 2041 can be used after 2070), cumulative emissions would be even lower than shown in figure 4.
The intuition for the magnitude of the effect in figure 4 is as follows. If emissions decline less rapidly over time, they must start at a lower level. Lower emissions in early years imply a larger initial surplus, so more allowances are stored in the MSR. Of those, a substantial fraction gets canceled. The cumulative supply of allowances therefore decreases and emissions adjust downwards, including in the early years. But this means we have to repeat our argument, with a larger initial surplus and more cancellation again and again. The cancellation feedback loop spirals on and  on, stabilizing only after a massive amount of allowances got canceled. Our argument thus predicts that if or when the market expects a flattening of the emissions pathway, there should be a markedly upward shift in the ETS price due to the substantial reduction in cumulative supply of allowances. Observe that the logic or intuition behind our results is general and does not rely on linear emissions pathways; linearity was assumed only to simplify the exposition.
What do we know about the time-structure of emissions? In general, it depends on a number of factors, such as technological change and the development of (marginal) abatement costs over time (Bruninx et al 2020). Some observations suggest, however, that the emissions pathway has become flatter over time than previously believed (Flachsland et al 2020). From a polluting firm's perspective, an emission allowance is a non-renewable resource that is perfectly substitutable over time. For this reason, the allowance price should follow Hotelling's rule and rise with the interest rate (Hotelling 1931), at least approximately. The interest rate has fallen consistently over the last years, implying a near-constant allowance price over time. The latter implication is borne out by the data: EU ETS prices for future allowances are close to current spot prices. All else equal, a flatter price path will tend to make the emission path decline less steeply. Combined with our findings above, the MSR thus supports a price path that reaches a fairly constant yet very high level, suffocating emissions. Due to this very low level of emissions and the implied substantial surplus of allowances in early years, a large majority of allowances will be canceled by the MSR, self-enforcing the extreme price level. This effect does not rely on the precise linear shape of the emissions pathway. The only thing that matters is that the emissions pathway is flat relative to the path of supply leading to a surplus of allowances in early years. After a decade of worrying about low allowance prices, the EU may soon find itself otherwise occupied, struggling with a carbon price that spirals out of control.
Obviously, scenario E3 in figures 1 and 3, involving a sudden drop in emissions by around two thirds from current levels, seems unrealistic. Economic activity levels and technology choices are less elastic in the short run than in the longer run. This rigidity will soften the outcomes revealed in our analysis. Future research could examine this further using high-resolution market models as in e.g. Bruninx et al (2020) and Osorio et al (2021).
Nevertheless, our results reinforce calls for an MSR revision that incorporates price triggers (Borenstein et al 2019, Murray et al 2020. A pricetriggered MSR was recently proposed in the EU's Fit for 55 legislation package as market stability mechanism for a second, completely independent EU emissions trading system to cover buildings and road transportation. That said, a simple price ceiling or floor (Abrell and Rausch 2017) is not so easily constructed in the EU ETS, largely because of the MSR. For instance, Article 29a of the EU ETS Directive stipulates that when over six consecutive months prices are more than three times higher compared to the average over the previous two years, 100 MtCO 2 of allowances are moved back from the MSR to the market. Our calculations suggest almost all of these allowances are returned back to the MSR at later stages, implying negligible effect on cancellations and cumulative emissions.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.