Navigating the turbulence: Analyzing Potential Carbon Leakage in EU ETS Aviation Policy Nils Malmström & My Selg Abstract: Aviation accounts for approximately 2.5% of global CO2 emissions. However, this seemingly low number is deceiving for several reasons. For one thing, it does not account for other warming effects of aviation, and further, it does not show the increasing trend in carbon emissions from aviation during the last decades. To reach global climate targets, policy adjustments for carbon emissions from aviation are more than likely to be needed. In this study, we investigate the climate policy for aviation in the European Union Emissions Trading Scheme (EU ETS). More specifically, we examine the potential carbon leakage that can arise from removing the free allocation of emission allowances (EUA) for the aviation sector in the EU ETS by 2026. To narrow down the scope of the study, we only analyze how the policy change would affect demand changes for three charter destinations from Landvetter Airport, Gothenburg, and the potential carbon leakage that could arise because of the demand changes. For the analysis, we use a regression model, containing time series data on number of passengers and fuel costs as a proxy for ticket prices, to estimate how variations in ticket prices affect demand for each destination. Using the estimations together with predicted future costs of EUAs, we find a reduction in demand of 10 to 20% in 2026. To illustrate the potential carbon leakage, we employ four different scenarios of travel substitution to a destination that is not affected by increased costs from EUAs, namely Antalya, Turkey, and find a leakage of 1.7 to 6.9 million kg CO2. Bachelor’s thesis in Economics, 15 credits Spring 2024 Supervisor: Håkan Eggert Department of Economics School of Business, Economics and Law University of Gothenburg Acknowledgements We would like to express our sincere gratitude to the people who have helped us in the process of writing this thesis. Firstly, we would like to thank our supervisor, Håkan Eggert, for the assistance in the analysis part and for overall comments throughout the thesis. We would also loke to thank Jörgen Larsson at Chalmers University of Technology for help with the research idea and background information on the subject. 1 Table of Contents LIST OF ABBREVIATIONS ..................................................................................................................... 4 1. INTRODUCTION .......................................................................................................................... 5 2. THEORY ...................................................................................................................................... 8 2.1. REGULATION OF A MARKET FAILURE .................................................................................................... 8 2.1.1. Externalities ...................................................................................................................... 8 2.1.2. Social Cost of Carbon ........................................................................................................ 8 2.1.3. Cap and trade ................................................................................................................. 11 2.1.4. Carbon leakage and the Waterbed effect .......................................................................... 12 2.2. DEMAND THEORY ........................................................................................................................ 13 2.2.1. Pass-through ................................................................................................................... 13 2.2.2. Elasticity ......................................................................................................................... 14 2.2.3. Travel Substitution ........................................................................................................... 14 3. BACKGROUND .......................................................................................................................... 17 3.1. LITERATURE REVIEW ..................................................................................................................... 17 3.2. INTERNATIONAL AVIATION POLICY .................................................................................................... 18 3.2.1. ICAO ............................................................................................................................... 18 3.2.2. CORSIA ........................................................................................................................... 19 3.3. AVIATION POLICY IN THE EU ........................................................................................................... 20 3.3.1. EU ETS ............................................................................................................................ 20 4. METHOD ................................................................................................................................... 24 4.1. ESTIMATING DEMAND CHANGES ...................................................................................................... 24 4.2 ESTIMATING CARBON LEAKAGE........................................................................................................ 25 4.3 ASSUMPTIONS ............................................................................................................................ 26 4.4 DATA ....................................................................................................................................... 26 4.4.1 Testing for stationarity ............................................................................................................. 28 4.4.2 Testing for autocorrelation and heteroscedasticity ................................................................... 28 4.3 DESCRIPTIVE STATISTICS ............................................................................................................... 30 4.4 ECONOMETRIC MODEL ................................................................................................................. 32 5 RESULTS .................................................................................................................................. 34 5.1 ESTIMATING DEMAND CHANGE ........................................................................................................ 34 2 5.2 ESTIMATING CARBON LEAKAGE........................................................................................................ 38 6 DISCUSSION AND CONCLUSIONS ............................................................................................. 40 6.1 METHOD DISCUSSION ........................................................................................................................ 41 6. 2 POLICY IMPLICATIONS ........................................................................................................................ 42 7 REFERENCES ............................................................................................................................ 44 APPENDIX A - CALCULATIONS OF DEMAND REDUCTION .................................................................... 47 APPENDIX B - CALCULATIONS OF CARBON LEAKAGE ......................................................................... 48 APPENDIX C - SENSITIVITY ANALYSIS WITH OTHER ELASTICITIES ....................................................... 49 3 List of abbreviations EEA - European Economic Area EU ETS - European Union Emissions Trading Scheme LRF - Linear Reduction Factor CBAM - Carbon Border Adjustment Mechanism EUA - Emission Allowance SCC - Social Cost of Carbon ICAO - International Civil Aviation Organization CORSIA - Carbon Offsetting and Reduction Scheme for International Aviation 4 1. Introduction The temperature rise on earth is predicted to surpass the 1.5°C target established in the Paris agreement due to increased emissions of fossil CO2 and other greenhouse gasses, resulting in adverse impacts on ecosystems, extreme weather and rising sea levels (IPCC, 2018). In the public debate on climate change, the aviation sector is often a hot topic. According to Lee et al. (2021) CO2 emissions from aviation make up approximately 2.5% of total global CO2 emissions. However, considering the full scope of environmental impacts from aviation, it is responsible for about 3.5% of effective radiative forcing, i.e. global temperature increase (Lee et al., 2021). Furthermore, previous decades have resulted in a significant rise in demand for global aviation, and predictions are that this trend will continue as global wealth increases. This, combined with the difficulties in further abating CO2 emissions and limiting impacts from aviation's high-altitude effects, highlights the necessity of regulating the environmental impacts from the aviation sector to ensure that global warming stays below the 1.5°C level. Some regulations of the environmental impacts of the aviation sector are already implemented on international, EU and national levels. The different regulations functions in different ways, ranging from large scale Cap and Trade programs, such as the European Union Emission Trading Scheme (EU ETS), to simpler excise duties on flight tickets, such as the Swedish flight tax. An international market-based regulation under the name Carbon Offsetting and Reduction Scheme for International Aviation (CORSIA), is currently entering its first phase but is still largely voluntary. These programs are vastly different in how they aim to abate emissions from the aviation sector; however, they all face large scrutiny and are subject to much critique concerning their efficiency (Feichert et al., 2020, Scheelhaase et al., 2018, Wozny et al., 2022). Flights between EU member states have been covered by the EU ETS since 2012, where flight operators are required to submit emission allowances (EUAs) for each ton of CO2 emitted on flights covered by the regulations in EU ETS (European Commission, 2024b). Through the trading and pricing of emission allowances, the cap-and-trade scheme allows for cost efficient emission reduction (Fichert et al., 2020, p. 57). Historically, emission allowances have mainly been allocated to airlines for free, and complementary purchases have been possible via public auctions and via trading (Nilsson, 2023). Over the coming years, public auctions will increase as 5 the main mechanism of allocation, with the aim of completely removing free allocation for the aviation sector by 2026 and by 2034 for the entire EU ETS (Nilsson, 2023). This policy change is expected to infer significantly increased costs for flight operators subject to the EU ETS, since more than 50% of aviation emissions in the scheme have been covered by allowances from free allocation (Nilsson, 2023). Subsequently, most of these increased costs are expected to be transferred on to the consumers by raising ticket prices and, thus, affect the demand for aviation (Koopmans and Lieshout, 2016). However, almost 20% of European flights are inter-continental (Eurocontrol, 2024), meaning that a large share of EU aviation emissions arises from flights which are not covered by the EU ETS, and are only partially regulated under CORSIA. A possible future difference in stringency of the environmental regulation between intra- and inter-continental flights raises the concern of a substitution between inbound flights, subject to EU ETS, to intercontinental flights, only subject to CORSIA. This effect is known as carbon leakage and may increase CO2 emissions outside the EU as a result of the EU policy (Dray and Doyme, 2019). The purpose of this study is to analyze the possible environmental effects of removing the free allocation of emission allowances for intra-continental aviation in the EU. The aim is not to quantify the total environmental impact on a global scale, but rather to illustrate the mechanism of distorted environmental policies. Therefore, the analysis is done on a limited scale, namely in a Swedish context, only including three specific flight routes. More specifically, we estimate the demand changes on flight travel to three EU destinations, arising from increased costs of CO2 emissions for flight operators. We then use these predicted demand changes to assess the possible increase in CO2 emissions outside the EU ETS given possible substitution to inter-EU aviation as a consequence of the policy change. For this part, we chose to assume that substitution only occurs to one destination, namely Antalya, Turkey. On this basis, we form the following research questions: ● How will removing the free allocation of EU ETS emission allowances to the aviation sector affect demand for three of the most frequent charter flight destinations from Landvetter Airport, Gothenburg? ● How could these demand changes affect CO2 emissions emitted outside the EU ETS? 6 To be able to understand the approach in this study, it is important have some basic knowledge about the basis of our research question. Therefore, this paper begins by explaining some important concepts in environmental economic theory. After that, we provide a background containing a description of the current state of knowledge regarding demand for air travel, as well as the current climate policy framework for aviation in the EU and internationally. After this, the method used in this study is explained. Finally, we present the results of our study before we discuss and conclude how the results can be interpreted and what the policy implications of the results are. 7 2. Theory 2.1. Regulation of a market failure 2.1.1. Externalities An efficient market is characterized by the accurate representation of benefits and costs, determining consumption levels through market equilibrium, where the marginal utility of consumption is optimized. However, in certain instances, the market price does not accurately reflect the total societal cost of a good. This discrepancy typically indicates the presence of externalities, which are the unaccounted-for effects on third parties that neither the consumer nor the producer considers in the market (Perloff, 2013). Externalities can be classified as either negative or positive. Negative externalities occur when the consumption or production of a good inflicts harm on others, which is not reflected in its price. Conversely, positive externalities arise when the consumption or production of a good generates benefits for third parties that are not considered by the market. The presence of externalities often leads to a suboptimal allocation of resources, a phenomenon known as market failure (Perloff, 2013). A quintessential example of market failure due to negative externalities is global warming. This phenomenon primarily results from the excessive consumption of fossil fuels, which, while economically beneficial to users, imposes environmental costs not accounted for in its market price. 2.1.2. Social Cost of Carbon To solve the market failure, the social cost of carbon emissions needs to be internalized in the market price. This is possible through governmental policy intervention that forces agents to pay for the cost the emissions cause the society. However, to be able to determine the scale of intervention that leads to an efficient market, the true social cost of carbon needs to be known. The social cost of carbon (SCC), defined as “the monetary value of the damage done by emitting one more tonne of carbon at some point of time” (Pearce, 2003), is possible to estimate. However, since the value of the SCC depends greatly on the parameters used in its 8 estimation, as well as the choice of social discount rate, the true cost of carbon has proven difficult to determine, although it has been thoroughly studied (Pearce, 2003). According to Stern et al. (2022) there are two main methods of estimating the SCC. The first method implies estimating the damage carbon has on our environment (following the definition presented above from Pearce (2003)), while the other implies estimating the necessary cost of carbon in the market to ensure that global warming is limited in accordance with the Paris agreement. The authors argue that previous research determining SCC through estimation of damages has resulted in an underestimation of SCC, compared to the SCC necessary for compliance with the Paris agreement (Stern et al., 2022). A recent and comprehensive study by Rennert et al. (2022) provides evidence for a social cost of carbon of $185 per tonne, ranging from $44–$413 in a 90% confidence interval. By contrast, the implemented value for SCC in US government evaluations, at the time of their study, was $51. This highlights the large differences of the estimated SCC, which has a large impact on its efficiency when used as a policy instrument. Regulation of externalities can essentially be made in two ways: i) regulating the price of a good or, ii) regulating the quantity of a good. Weitzman (1974) presents a thorough analysis of the implementation of price and quantity-based regulations in the face of uncertainty about the social costs of the externality as well as the associated abatement costs of said externality. He concludes that a regulation on quantity, such as a cap on carbon emissions, is preferable when there is greater uncertainty about social costs than abatement costs and the opposite is true for a price regulation, such as a carbon tax. Ultimately, he recommends a combination of policies to swiftly adapt to new information about costs and benefits. 9 The environmental effects of aviation (Based on Lee et al. (2021)) According to aviation accounts for about 2.5% of global CO2 emissions. However, CO2 emissions are only responsible for one third of aviation’s contribution to global temperature increase. When accounting for all factors affecting net surface warming from aviation – including nitrogen oxides (NOx), water vapor, soot and sulfate aerosols, and increased cloudiness due to contrail formation – aviation is responsible for about 3.5% of effective radiative forcing, i.e. 3.5% of global temperature increase. This is, however, a severely simplified image of the impact aviation has on global temperature, which is explained comprehensively in Lee et al. (2021). The radiative forcing from aviation’s CO2 emissions is not a constant share of the sector’s total radiative forcing. A decisive factor for this is that different climate forcing agents impact the temperature increase on different time scales. While CO2 emissions are long-lived in the atmosphere and continue to impact the climate for centuries, non-CO2 forcing agents from aviation are short-lived and, thus, stop affecting the climate when the inflow stops. Therefore, the ratio between the radiative forcing from CO2 emissions and non-CO2 forcing agents will change substantially if future aviation emissions deviate from their current growth trajectory. Lee et al. estimates a multiplier of 3 for the total radiative forcing from aviation compared to the radiative forcing from the CO2 emissions from aviation alone (the estimated multipliers for 2018 range from 1.0 to 4.0 depending on the choice of time horizon and emission metric). However, in a scenario where emissions from aviation decrease in the future, this multiplier could fall below 1, since a steadily falling rate of non-CO2 emissions would equal removing CO2 from the atmosphere. Altogether, Lee et al. argues that using a multiplier in climate policy has limitations, as it is dependent on subjective choices. 10 2.1.3. Cap and trade The policy of a cap on a good with negative externalities is often implemented through a cap- and-trade program. The EU ETS is one example of a cap-and-trade program where a cap is set on the “good” CO2 emissions. In the program, the cap is defined as the maximum allowed quantity in a specific economic region. The cap is enforced through the issuing of allowances that in turn can be traded between the actors in the market (Fichert et al., 2020, p. 57). Through the trade of permits a price on the externality is created, but contrary to a tax, the price is determined in the market by supply and demand, not through policy decisions. In the context of climate change, cap-and-trade programs are designed to regulate the emissions of greenhouse gases. The cap represents the total allowed emissions for the participants in the program, and tradable permits will represent an allowance to emit a specific amount of emissions. When trade of permits is allowed, the societal abatement of emissions will be economically cost effective, since actors with abatement costs lower than the price of emission allowances can sell permits, and the actors with high abatement costs will purchase permits. Such a system achieves the equimarginal principle, i.e. that abatement costs are minimized, since all actors in the system will abate emissions until the marginal cost of abatement equals the permit cost. Further, this means that all actors will emit at the same marginal cost of abatement (or marginal savings of further emissions). In the design of an emissions trading scheme, policymakers will have to decide on how to allocate permits. One approach is that emission allowances are allocated based on past emissions, so called “grandfathering”. Another possible approach is that emission allowances are allocated through auctioning. The method of allocation is an important factor of the economic impacts for the participants in the program, however, it has no effect on the net emission reduction under the ETS, since the total emission cap is predetermined by the policymakers (Fichert et al., 2020, p. 61-62). In this essay we will specifically consider how a change from free allocation to allocation by auctions, i.e. charging for permits, might affect the pricing and demand for intra EU aviation. 11 2.1.4. Carbon leakage and the Waterbed effect When implementing a carbon emissions policy in one area, the aim is to reduce the negative externalities of carbon emissions. When policy changes cause differences in cost of emissions between the policy-affected area and non-affected areas there is a risk of carbon leakage. Carbon leakage is defined as the change in carbon emissions outside the policy affected area divided by the change in carbon emissions in the policy regulated area, due to the policy change. Leakage is therefore presented as a percentage share, where the leakage, in theory, can become infinitely high. 𝐼𝑛𝑐𝑟𝑒𝑎𝑠𝑒 𝑖𝑛 𝑒𝑚𝑖𝑠𝑠𝑖𝑜𝑛𝑠 𝑜𝑢𝑡𝑠𝑖𝑑𝑒 𝑝𝑜𝑙𝑖𝑐𝑦 𝑎𝑟𝑒𝑎 𝐿𝑒𝑎𝑘𝑎𝑔𝑒 = Equation (1) 𝐷𝑒𝑐𝑟𝑒𝑎𝑠𝑒 𝑖𝑛 𝑒𝑚𝑖𝑠𝑠𝑖𝑜𝑛𝑠 𝑖𝑛𝑠𝑖𝑑𝑒 𝑝𝑜𝑙𝑖𝑐𝑦 𝑎𝑟𝑒𝑎 Source: (Dray and Doyme, 2019). In this study, carbon leakage is a central concept. However, in our case, it is necessary to adjust the definition presented in equation 1. This adjustment is due to the waterbed effect inherent in emissions trading schemes with a fixed emissions cap, such as the EU ETS. The waterbed effect demonstrates that emission reductions in one sector, company, or activity will likely be offset by an equivalent increase in emissions from another entity within the same emissions cap. This effect is analogous to the mechanics of a waterbed: an indentation in one area causes a rise elsewhere, as the total volume remains constant. Due to the predetermined cap on emissions, CO2 reductions from a specific sector within the policy area are theoretically equivalent to zero, if there is a shortage on emission allowances. This implies that, under the original definition of carbon leakage, the denominator approaches zero. Consequently, if a policy change within the EU ETS leads to an increase in emissions outside the trading system, the leakage rate approaches infinity. Therefore, a more appropriate measure of carbon leakage for this study is the absolute increase in emissions outside the policy area. This also means that national policies regarding CO2 emissions from aviation inside the EU area, such as the Swedish aviation tax, do not affect the emission level. However, they do have an effect on emissions from inter-continental aviation and on non- CO2 emissions. According to Dray and Doyme (2019) carbon leakage arising from changes in aviation policy stems from two main components, firstly the change in behavior of passengers, and secondly 12 the change in behavior of airlines. The former includes changes in travel demand and frequency, whereas the latter can include fleet-swapping, where the most efficient aircrafts operate in the policy affected area and the less efficient aircrafts operate outside, leading to a shift in emissions but no absolute reduction. The focus in this essay lies on the estimation of passenger responses to the future change in EU ETS, airline responses will not be covered in the analysis. To limit carbon leakage, one possible measure is to use environmental import tariffs. The EU has decided to implement such a measure, called a Carbon Border Adjustment Mechanism (CBAM) for some specific sectors in the EU ETS. It works through putting a cost (in line with the ETS permit cost) on imports based on their greenhouse gas emissions from a life-cycle perspective. However, as of now, there is no plan to include aviation in CBAM (European Commission, 2024a). Even though aviation to the EU is not necessarily defined as imports per se, one could imagine that incoming flights could be covered by the same mechanism in the future. In this study, we only cover outgoing flights from EU airports, and thus, even if aviation would be covered by CBAM, it would have no direct implications in this case. 2.2. Demand theory 2.2.1. Pass-through Pass-through regards the extent to which an increase in costs for a producer is passed on to the consumer and its level is dependent on the market conditions, elasticity of demand and whether costs are firm- or sector-specific (Koopmans and Lieshout, 2016). In the context of this study, pass-through relates to the share of the airline's costs from emission allowances that is reflected in a price increase for the passengers. The market structure of aviation traffic is therefore necessary to investigate. It has been established that air traffic often follows a differentiated oligopolistic competition structure, whereas it is often assumed to be perfectly competitive (Koopmans and Lieshout, 2016). Pass-through in this market setting is dependent on the degree of competition between firms, elasticity of supply (related to the marginal cost for suppliers), elasticity of demand and the shape of the demand curve (Wozny, 2024). Following this, Koopmans and Lieshout (2016) establish a pass through of more than 50% for costs that 13 apply to the whole sector, which is the case with costs for the EU ETS. Estimating the correct pass-through of costs of EUAs is a difficult task, and it will not be attempted in this study. 2.2.2. Elasticity In neo-classical economic theory, consumers are expected to behave as rational individuals who, when presented with an option, will choose the option that maximizes her utility given her budget restriction. One concept used in economics is the price elasticity of demand which describes the ceteris paribus change in the demanded quantity of a good due to a percentage change in the price of the good (Perloff, 2013). Defined by the following formula: 𝛥𝑄/𝑄 𝜀 = Equation (2) 𝛥𝑃/𝑃 Where Q represents the quantity of the good, P represents the price of the good and 𝜀 is the price elasticity of demand. The price elasticity of demand takes on a negative value for normal goods, meaning an increase in price derives a reduction in demanded quantity of the good, following the law of demand (Perloff, 2013). This applies to air travel, for which numerous estimations of the elasticities of demand have been conducted. A meta-analysis on elasticity of demand for air travel found the mean estimated elasticity to be -1.146, corresponding to an elastic demand (Brons et al., 2002). This means that a 1 percent increase in price leads to a 1.146 percent decrease in demand. Another important concept is the income elasticity of demand, which is defined as the percentual change in demand from a one percent increase in income. For normal goods, such as air travel, an increase in income leads to an increase in demand. 2.2.3. Travel Substitution The availability of substitutions of a good is one factor that affects the level of elasticity of demand. Brons et al. (2002) illustrates the multiple levels of substitutions an individual is faced with in the context of air travel, see Figure 1. Each level of the model represents a choice of allocation of income, where choices are based on utility maximization. 14 Figure 1: Chart depicting a consumer’s evaluation of different substitutes to travel, based on model in (Brons et al., 2002). In the context of our essay, where we are evaluating the potential carbon leakage following the change of allocation of EU emission allowances to the airline sector, the potential substitutions of air travel are essential for our analysis. The four steps of Brons model can be described as follows: 1. The individual is faced with a choice of allocating some share of her income to travel expenditure. (Previous studies have shown that travel expenditure is income elastic.) 2. The consumer then decides how to distribute their travel budget among various destinations. We hypothesize that an increase in ticket prices for destinations within the EU may lead the consumer to redirect a larger portion, or even their entire travel budget, to alternative destinations. These destinations could be closer to her or outside the EU, where prices have not risen in the same manner. 3. After deciding on a destination and a budget, the individual can choose between modes of transportation. The figure presents a limited depiction of the possible substitutions at this level. The choice will, of course, depend on multiple factors, such as cost, time, preferences and available alternatives. If the individual has chosen a destination close to her, then transportation via train or car might be a viable option. If the individual has chosen a destination far away, for example Egypt or the USA, other modes of 15 transportation except for aviation is not a plausible alternative for the individual due to geographical limitations. 4. The final decision involves selecting among various airlines that offer the desired route. This aspect of substitution is excluded from our analysis as it is assumed to have negligible impact on carbon leakage. 16 3. Background 3.1. Literature review To the best of our knowledge, no study has attempted to estimate the scope of the potential carbon leakage originating from removing the free allocation of emission allowances for the aviation sector. What has been studied before, however, is how increased prices for aviation affect demand. Oesingmann (2022), for example, estimates how inclusion of the aviation sector in EU ETS has affected air passenger flows for intra-EEA (European Economic Area), and finds no significant effect. This might, however, be heavily influenced by the low ETS permit price (and free allocation) during the years the study covers. Oesingmann (2022) also estimates the effects of Austrian and German air transportation ticket taxes and finds that they have had a statistically significant negative effect on demand. Fageda and Teixidó (2022) uses a difference in difference approach to estimate how emissions from intra-EEA flight routes have changed from 2010 to 2016, i.e. since the inclusion in EU ETS, compared to a control group consisting of inter-EEA flight routes. They find that the EU ETS reduced emissions by 4.7% in the regulated routes relative to the counterfactual, and that the effect has been even larger on short-haul flights, 10.7% reduction. They also discuss the fact that airlines could shift their activity to non- regulated routes, and thereby cause carbon leakage. However, they do not attempt to analyze the extent of this effect. There are several studies that have estimated how demand for air travel changes with ticket prices. For instance, Brons et al. (2002) provides a meta-analysis of price elasticities of demand for passenger air travel. They find elasticities ranging from positive 0.21 to negative 3.2, varying depending on e.g. distance and type of ticket. However, we only found two studies that estimate demand changes in a Swedish context. Kopsch (2012) estimates short- and long-run elasticities of demand for domestic business and leisure air travel in Sweden using aggregate price index data. For leisure travel he finds a short-run elasticity of -0.79 and a long-run elasticity of -1.20, and finds it to be more elastic than business travel. Stråle (2021) uses the price of aviation fuel times the flight distance as an instrumental variable (IV) for estimating demand elasticities for air travel from Arlanda Airport in Stockholm. The use of the instrument is motivated by the fact that there is a simultaneous bias in the pricing of tickets, where prices affect demand, but demand simultaneously affects prices. Prices of aviation fuel, however, should not affect the demand for 17 air travel, other than through the effect it has on airlines pricing, this fulfills the instrument exogeneity assumption. Furthermore, Stråle (2021) finds that variations in fuel price and variations in ticket price have a strong correlation, which validates the relevance of fuel as an instrument in estimating the demand changes. The price elasticity of demand estimated by Stråle (2021) for international air travel is -0.76, corresponding to an inelastic demand. 3.2. International aviation policy 3.2.1. ICAO The United Nations agency International Civil Aviation Organization (ICAO) was founded following the Convention on Civil Aviation, also known as the Chicago Convention, in 1944, with the purpose of fostering cooperation between its member states to allow safe international air travel (European Commission, 2024b). Among other responsibilities, ICAO is the responsible party for ensuring the reduction of greenhouse gases from aviation following the adoption of the Kyoto protocol in 1997 (Feichert et al., 2020). Commonly, implementation of policies covering emission reduction in line with the Kyoto protocol falls on the participating states. However, due to the international nature of the aviation sector, it is difficult to assign responsibility to a sole country (Feichert et al., 2020). Following the Kyoto protocol, ICAO set out to examine possible measures to ensure the stabilization of net carbon emissions. The work took several years and resulted in the acceptance of three goals of environmental initiatives among ICAO members: “i) An average annual improvement in fuel efficiency of 1.5% between 2009 and 2020, ii) a freeze of the sector’s net CO2 emissions from 2020 [...], and iii) eventually a reduction to 50% of 2005 levels by 2050” (Fichert et al., 2020, p. 118). In order to reach these goals, flight operators and states will have to work with technology, operations, infrastructure and carbon offsetting (Feichert et al., 2020). Following this, a market- based offsetting scheme for international aviation emissions was developed, which is discussed in the next chapter (3.2.2). 18 3.2.2. CORSIA The implementation of a new global market-based carbon offsetting scheme under the name Carbon Offsetting and Reduction Scheme for International Aviation (CORSIA) was decided during the 39th Assembly of ICAO in 2016 (ICAO, 2016). Implementation of the program is divided into three phases, starting with a pilot phase, taking place from 2021 to 2023, followed by the first phase from 2024 to 2026, which are both subject to voluntary participation. The second phase, ranging from 2027 to 2035, is mandatory for ICAO members who fulfill certain criteria (ICAO, 2016). A total of 126 countries are participating voluntarily in CORSIA from January 1st 2024, whereas 88 countries enrolled from the start of the pilot phase in 2021 (ICAO, 2020, ICAO, 2023). Even from 2026, under the mandatory period of CORSIA, not all countries will be covered by the scheme. Small, developing nations and states with minimal contribution to global aviation emissions will be excluded, but are encouraged by ICAO to voluntarily participate in CORSIA (ICAO, 2016). The aim of CORSIA is in line with one of ICAOs goals on environmental initiatives: to freeze global emissions from international aviation at the 2020 level. The scheme has been designed to have eligible operators compensate for the sector’s growth in CO2 emissions compared to the baseline level (Swedish Transport Agency, 2024)1. In practice, a given operator's compensation obligation is decided based on their emissions from flights on routes covered by the CORSIA scheme over the entire year, multiplied with a growth factor for the total sectoral emissions for all CORSIA flights (Scheelhaase et al., 2018). This method for calculating is used in the pilot and the first phase and does not account for the individual operator's growth from the baseline. During the second phase, with compulsory participation for ICAO member countries, growth factors will be estimated differently, where the growth of specific airlines are included to different extents (Scheelhaase et al., 2018). The implication of the decided basis for estimating offsetting amounts in the respective phases is that operators are not incentivized to reduce their own emissions during the pilot and first phase. However, a reason presented in favor of the decided offsetting calculation is to avoid skewing the burden of offsetting to newer airlines undergoing 1 From 2024, the baseline is set to 85% of the CO2 emission level of 2019. During the pilot phase (2021- 2023) 100% of 2019 emissions level were used as a baseline (Swedish Transport Agency, 2024). 19 rapid growth, compared to the more established airlines with low growth rates (Fichert et al., 2020, p. 118). The necessary offsetting to stabilize the international aviation sector's emissions at the baseline level is achieved through operators' purchases of eligible offsetting credits. ICAO has established several criteria for ensuring the eligibility of CO2 offsetting projects, criteria are based on credibility of the offsetting by demanding correct measurements and ensuring offsetting is permanent. A list of eligible projects for offsetting is provided by the organization. Still, the credibility of carbon offsetting, and by extent the credibility of CORSIA, has been questioned by Wozny et al. (2022) among others, who found that the credits could not guarantee net reductions of emissions. As put forth by Scheelhaase et al. (2018), the environmental integrity of carbon offsets in CORSIA is integral for the reduction of the global aviation sector’s emissions. 3.3. Aviation policy in the EU 3.3.1. EU ETS Parallel to ICAO’s development of CORSIA, the European Union set out to implement their own regulation of CO2 emissions from international aviation. Already in 2008 the EU passed legislation to include the aviation sector in the existing European emissions trading scheme, beginning in January 2012 (European Commission, 2024b), compared to international regulation that was established first in 2016 (ICAO, 2016). The original ambition of the EU was to include “emissions from flights from, to and within the European Economic Area (EAA)” (European Commission, 2024b). However, the inclusion of flights with origin or destinations outside of the EU was faced with large opposition from several countries, including Russia, China, India and the United States, as well as some European airlines (Fichert et al., 2020, p. 87). As a result of this critique, and in combination with ICAO’s further development of international regulations, the EU decided, in April 2014, to reduce the scope of covered flights to only intra-EU flights (Fichert et al., 2020, p. 87). The details of the aviation sector’s involvement in the EU ETS are however not final, according to the European Commission new decisions on the scope of flights covered will be made based on the success of the implementation of CORSIA by 2024 (European Commission, 2024b). The scope of covered flights is an important 20 aspect for the analysis in this essay, as it has a direct effect on the policy area and thereby the possible carbon leakage. The EU ETS is a cap-and-trade program for carbon dioxide intensive industries in the EU. The program is divided into several phases, where the aviation sector joined at the end of phase 2, ranging from 2008-2012. Currently the ETS is in the fourth phase, ranging from 2020-2030 (European Commission, 2024c). Since aviation's entry in the EU ETS, emissions from aviation have generally constituted a growing share of the total greenhouse gas emissions covered by the cap. Until the covid-19 pandemic, yearly emissions from aviation were growing, while emissions from stationary installations generally had a downward trend. Overall, emissions in the scheme have decreased due to successful CO2 reduction from electricity producers (Nilsson, 2023). Between the phases, both minor and major modifications were implemented; notably, Phase 3 introduced a significant shift from national emission caps to a harmonized cap across all EU emissions (European Commission, 2024c). The cap for the EU ETS for 2021 is set to approximately 1,600 million emission allowances, each allowing one tonne of CO2 emission, and the cap will be systematically lowered yearly according to a linear reduction factor (LRF) currently set to 2.2% (European Commission, 2024d). Consequently, there will be a reduction of more than 50% in new emittance of allowances by 2030 compared to 2023, and given the current pace of reduction, no more allowances will be put on the market after 2039 (Nilsson, 2023). This highlights the enormous transition that is necessary in the carbon intensive industries of the EU. Not only will the amount of available emission allowances change drastically over the coming years, but the method of allocation is also changing. At the start of EU ETS, a majority of allowances were freely allocated. Which also applied to aviation at its introduction in the ETS (Nilsson, 2023). The EU has now decided to systematically phase out free allocation of emission allowances until 2030, substituting it completely for allocation via auctions held by EU member states. For the aviation sector, free allocation will end at the start of 2026 (Nilsson, 2023). 21 Given the historic levels of free allocation, the coming policy change is expected to result in substantially increased costs for flight operators participating in the EU ETS. Figure 2 illustrates the quantity of freely allocated allowances, compared to the sector's verified emissions. The discrepancy between these values constitutes the amount of allowances that was purchased, predominantly through purchases from other sectors via market trading, and some via auctions orchestrated by member countries. Between 2013 to 2019 approximately 50% of the sector's emissions were covered by freely allocated emission allowances (European Environmental Agency, 2023). However, the significant decline in aviation activity due to the Covid-19 pandemic led to allocated allowances in 2020 exceeding total emissions by nearly 20%. Figure 2: Graph depicting number of freely allocated EUAs and the aviation sectors’ verified CO2 emissions. Source: EU Emissions Trading System data viewer (European Environmental Agency, 2023). The economic impact of the policy change will naturally depend on the market price of emissions allowances. Historically, prices have varied immensely and during the financial crisis of 2008 prices dropped to almost zero euros and stayed low in the following years, varying around approximately €5 per allowance (Nilsson, 2023). For multiple reasons, such as the 22 design of the EU ETS and effects of other policies aimed at emission reduction in the affected sectors, a large surplus of allowances was generated which kept the prices at low levels (Nilsson, 2023). To lessen this problem, the EU in 2015 decided to implement the Market Stability Reserve, which tackles the “Total Number of Allowances in Circulation” (Nilsson, 2023). Following the implementation, the market price of European emission allowances has risen to approximately €50 between 2022-2024, with peaks of circa €100 in February 2023 (Trading Economics, 2024). The prediction of prices for European emission allowances in 2026 is necessary to answer the research questions for this thesis, this will be covered under section 5.1. 23 4. Method The method in this study consists of two parts, each corresponding to one of the two research questions. In the first part, we estimate demand changes resulting from removing the free allocation of emission allowances to the aviation sector to three popular charter destinations from Landvetter Airport: Chania, Gran Canaria, and Palma Mallorca. In the second part, we use the estimated demand change to calculate the carbon leakage, given four different leakage scenarios. 4.1. Estimating demand changes Usually, evaluating the effect price changes on one good has on the demand for another good involves estimation of a cross-price elasticity. In the context of our essay, cross-price elasticities for all destinations of outgoing EU flights would be necessary to evaluate the total carbon leakage of the policy change for aviation. Estimating such elasticities would be very difficult, and we found no previous research attempting such a task. Therefore, we chose to instead focus only on the potential leakage from one Swedish airport, namely Landvetter Airport, Gothenburg. Further, we chose not to estimate cross-price elasticities. Instead, we estimate price elasticities of demand for aviation and calculate how increased ETS costs would affect demand for three charter flight routes. As for the demand change to inter-EU destinations, we rely on assumed possible substitution scenarios instead of an empirical analysis. We were interested in leisure travel to EU destinations, currently covered by the EU ETS, since substituting destinations for leisure travel likely is easier than for example business travel. Previous studies also conclude that the price sensitivity of business travelers is much lower than that of leisure travelers (Brons et al., 2002). To be able to assume that most of the travel to the destinations is leisure travel, we only included charter destinations in our sample (Swedavia, 2024).2 2 This was done by filtering for charter destinations on the Swedavia webpage (Swedavia, 2024). All in all, 24 EU charter destinations were listed. However, only six of these had sufficient data to be included. 24 To estimate price elasticity of demand, we apply a modified version of the approach used in Stråle (2021)3. Because of time restrictions we do not have time to collect flight ticket prices online, like Stråle does, and therefore, we instead use fuel costs directly as a proxy for ticket prices. Because our method did not include the actual ticket prices, we chose to also apply a sensitivity analysis were we instead use the estimated price elasticities of demand in Stråle (2021) and Kopsch (2012) to calculate the demand changes the increased ETS costs would result in (Appendix C). In a first attempt at estimating a price elasticity of demand, we used a panel data approach. The dataset covered six destinations – Chania, Gran Canaria, and Palma Mallorca, as well as Larnaca, Rhodes and Tenerife – together with monthly time series data from January 2004 to March 2018. However, when running our regression using fuel cost multiplied with the distance to the destinations as our main independent variable of interest, our results were inconclusive, and we found no significant effect on the number of passengers carried. This was likely due to the largely unbalanced dataset with many missing values (we treated months with no passengers as missing values since there were no flights. However, the absence of departures was most likely a cause of low demand in the long run). Because of the difficulties estimating demand elasticities using panel data, we decided to instead use a time series approach, and estimate demand elasticities for each destination separately. For this method, only three destinations provided significant results, hence Larnaca, Rhodes and Tenerife are not included in the results. To calculate the extent of the change in traveling passengers to each destination, we used our estimated elasticities together with predicted permit prices in 2026, found in the literature, average fuel cost and a conversion factor for translating the cost for CO2 emissions to costs per liter of aviation fuel. A detailed description of the calculations can be found in Appendix A. 4.2 Estimating carbon leakage In the second part, our aim is to find a rough estimate of the magnitude of the carbon leakage as a consequence of the policy change. Because of the difficulties in estimating this effect using 3 Stråle’s model is explained in more detail in chapter 3.1 of this essay. In this chapter we focus only on the aspects in which our application differs from Stråle (2021). 25 quantitative methods, the analysis instead relies on multiple scenarios for the extent of consumers' substitution of intra-EU flights to possible alternatives. We chose to only include one alternative, namely Antalya, as the destination to which passengers substituted their consumption. The choice was made since Antalya is a charter destination that is popular for travel from Landvetter, is located outside the EU, but still is relatively similar in distance and travel time. Finally, carbon leakage for each scenario was calculated using CO2 emissions data retrieved from ICAO’s carbon calculator (ICAO, 2024). 4.3 Assumptions We make a few important assumptions in our analysis, which affect how our results should be interpreted. Firstly, we assume that the carbon offset system in CORSIA has no impact on flight ticket prices and therefore no effect on demand of air travel. CORSIA has been shown to have a limited effect in earlier studies, but to what extent this assumption would hold in practice is hard to anticipate. It is also important to note that our approach assume that EU ETS will not be expanded to cover inter-EU aviation. According to Nilsson (2023), the plan of the EU- commission is that inter-continental aviation in the EU will be included in EU ETS starting in 2027, unless ICAO by 2025 has made decisions that are judged to secure a climate-neutral aviation by 2050. However, how this will unfold remain uncertain. 4.4 Data Since our approach is similar to the method used in Stråle (2021), we chose to use similar data sources. Hence, we collected fuel price data from the US Energy Information Agency (EIA) (US Energy Information Agency, 2024), data on monthly passengers carried from Eurostat (Eurostat, 2024a), and Swedish temperature and sunshine data from the European Climate Assessment & Dataset project (European Climate Assessment & Dataset, 2024). The latter two variables were used as control variables, as we expect temperature and sunshine in Sweden to correlate with demand for air travel abroad. We also added Swedish GDP and unemployment rate as additional control variables. For this, we used quarterly data from Eurostat and Statistics Sweden respectively (Eurostat, 2024b, Statistics Sweden, 2024). 26 Our main independent variable, fuel cost, is constructed from a combination of fuel prices and the corresponding cost for emission allowances. Stråle (2021) does not include the ETS costs in his dataset, but since we only include EU destinations in our dataset, cost for ETS permits should affect fuel cost to all destinations. However, due to low permit prices and the large share of free allocation during our time sample, this likely only had minor implications for the results. The costs for ETS were included by taking the mean of monthly ETS permit prices, collected from Trading Economics (2024), for applicable period: January 2013 to March 2018. The permit prices were transformed to fuel costs in two steps: firstly, we calculated CO2 emissions per liter of aviation fuel, using carbon intensity (3.16 kg CO2/kg aviation fuel) and aviation fuel density (0.811 kg/liter), collected from ICAO (2018) and Embraer (2023) respectively. Secondly, this was multiplied with the permit prices for each year. Before this was added to the fuel cost, it was multiplied with the share of allowances that were not allocated for free to the aviation sector. This approach implies that the entire cost for permits is transferred to the fuel cost, and that the firms occupying the routes of interest need to buy allowances at the average rate of the sector. Because we only included EU destinations in our regressions, we also decided to transform the jet fuel prices from dollars per gallon to euros per liter using monthly average exchange rates collected from the European Central Bank (European Central Bank, 2024). It could be argued that using Swedish krona would have given even more representative results. However, to what extent airplanes are fueling up in Sweden versus euro countries is difficult to determine, therefore, we decided to use euros per liter as our unit. Our analysis includes monthly time series data from the period January 2004 to March 2018. This gives us a total of 171 observations (T=171). Preferably, we would have liked to include a longer time period since more observations would improve the accuracy of our model. However, the data for departures was limited further back in time than 2004. Furthermore, the Swedish aviation tax was implemented in April 2018, and we did not want effects from the tax to impact our results on demand changes. We contemplated whether to include observations from the years following the financial crisis in 2008. This meant a trade-off between more observations and the possibility of including effects of the crisis in our estimates. We decided to include all observations as well as constructing a dummy variable for the years around the financial crisis (covering September 2008 to December 27 2010). A few different date ranges for the dummy was tested; this had no significant effect on the results. 4.4.1 Testing for stationarity To be able to estimate a causal relationship between fuel prices and aviation demand, a necessary precondition is that we can assume that the data is stationary, i.e. that there is no inherent time trend in the data series. Therefore, we used Augmented Dickey-Fuller tests for unit roots with zero lag to detect stationarity in our time series data. Our main variables of interest were fuel cost and number of departing passengers, as well as the logarithmic version of these variables. The test results for the passenger data showed that the presence of a unit root could be rejected for both for the original and logarithmic time series for all destinations (p- values below 0.000). Thus, this data was assumed to be stationary. The fuel cost data was a bit more ambiguous. The Augmented Dickey-Fuller test for the original (i.e. non logarithmic) data gave a p-value of 0.0994, and thus, the presence of a unit root could not be rejected. Consequently, we cannot reject the possibility that the data is stationary. However, this does not necessarily mean that the data is non-stationary. The p-value is relatively low and assuming stationarity, therefore, still implies a relatively low risk of a type 1 error. The logarithmic fuel price data, on the other hand, has a p-value of 0.0394, and thus, assuming stationarity implies a relatively low risk of a type 1 error. In our estimations, we only use the logarithmic data, and the assumption of stationarity therefore seems reasonable. (Stock and Watson, 2020). One might argue that inflation should be considered, and that the fuel cost data, therefore, should have been transformed from nominal to real terms, using for instance consumer price index. However, what consumer price index to use in that case is not obvious, and further, to what extent it is inflation that affect fuel prices and not the other way around is not obvious either. To simplify, we therefore use fuel costs in nominal terms. 4.4.2 Testing for autocorrelation and heteroscedasticity Being able to assume our data is stationary is necessary for unbiased and consistent estimators. However, stationary data it is not enough for the OLS estimators to be efficient. The data can be heteroscedastic, i.e. that the variance of the error term is not constant, and/or be 28 autocorrelated, i.e. that the error terms for different observations in time are correlated. If the estimators are heteroscedastic or autocorrelated, the estimators are no longer efficient, which means that all inference becomes misleading and that we no longer can trust the standard errors, t-statistics or p-values. To test for heteroscedasticity, we perform the Breusch-Pagan- Godfrey test for all our regressions (Breusch and Pagan, 1979). The test results indicate problems with heteroscedasticity for all regressions (p-values below 0.01 for f-statistic, null hypothesis: homoskedasticity). We also performed the Breusch-Godfrey test with one lag to test for autocorrelation for all regressions. The tests indicated problems with autocorrelation for Palma Mallorca and Gran Canaria (p-values below 0.05 for f-statistic, null hypothesis: no autocorrelation). To manage the possible risk of inefficient estimators, we decided to run our regressions with robust standard errors. 29 4.3 Descriptive statistics Standard Variable Mean deviation Min Max Observations Monthly passengers to Chania 5008.524 2309.814 479 9686 84 Monthly passengers to Gran Canaria 4630.474 3551.15 471 11674 171 Monthly passengers to Palma Mallorca 4368.962 3649.957 3 16255 132 Fuel costs (EUR/liter) .4330344 .1228923 .1950225 .6724068 171 GDP per capita in Sweden (Million SEK) 427848.7 66293.03 306705 577308.7 171 Mean temperature Gothenburg (°C) 8.888471 6.699641 -6.15 21.19 170 Sunshine duration Gothenburg (0.01 hours) 14387.84 8809.31 760 35100 171 Unemployment rate in Sweden (%) .0753385 .0102581 .0525459 .0993745 171 Table 1: Descriptive statistics for mean, standard deviation, largest and smallest value, and number of observations for all variables in our dataset. From the descriptive statistics presented in Table 1, we can determine that Chania is the most popular destination, followed by Gran Canaria and lastly Palma Mallorca. There are some missing values for our dependent variable, monthly passengers, to both Chania and to Palma Mallorca. As mentioned in section 4.1, these correspond to months where there were no direct flight routes to the destinations, which is likely a cause of low demand. 30 An important aspect to note regarding our dataset, is that there is a great seasonal trend in the number of passengers to all the destinations, which Figure 3 depicts. For Chania, Mallorca and Antalya, the number of passengers increases during the summer months and is close to zero during the winter months. For Gran Canaria, the opposite pattern occurs. This indicates that travelers have different demand functions for different parts of the year. We address this in our model below by using monthly dummy variables. This pattern might indicate that Antalya is not a plausible substitute for leisure travel to Gran Canaria. However, in this study, we assumed that this is the case. Figure 3. Monthly departing passengers from Landvetter Airport to each destination by year. Figure 4 shows the variation in fuel price, in nominal terms, during the period of interest. The fluctuations are quite large, but no obvious time trend occurs. 31 Figure 4. Monthly average fuel prices for the time period in our dataset in euros per liter. 4.4 Econometric model When estimating our price elasticities of demand, we decided to use a static model, i.e. that the demand only was affected by fuel costs in the same time period. It is, of course, possible that not all flight tickets were necessarily bought in the same month as the departure. Thus, a dynamic model, where the dependent variable depends on lagged fuel prices as well (i.e. fuel prices in t-1), might have provided a better fit. However, to simplify the model, we assume that no such lagged dependence exists. We use two separate models to estimate how price affects demand: Model 1: 𝑌𝑡 = 𝛼 + 𝛽𝑥𝑡 + 𝛿𝑡 + 𝜀𝑡 Model 2: 𝑌𝑡 = 𝛼 + 𝛽𝑥𝑡 + 𝑋𝑡 + 𝛾𝑡 + 𝛿𝑡 + 𝜀𝑡 In the first model, the log of number of passengers, 𝑌𝑡, only depends on the log of jet fuel prices, 𝑥𝑡. Monthly fixed effects, 𝛿𝑡, are also included. The second model includes a vector, 𝑋𝑡, 32 containing the control variables GDP/capita and unemployment rate in Sweden, as well as mean temperature and sunshine duration in Gothenburg. The second model also includes a dummy for the months September 2008 to December 2010, 𝛾𝑡, to control for effects of the financial crisis. 33 5 Results This chapter of the essay presents the results of our analysis and seeks to address our research questions. The chapter is structured into two main sections, where each section relates to each of our research questions respectively. 5.1 Estimating demand change Table 2 presents the results from our regression models, presented in chapter 4.4. For each of the three destinations, two regression results are presented. The first column (1) presents the regression results from our first model, and the second column (2) provides results from the second model, where a vector of control variables and a dummy for the financial crisis are included. 34 Gran Canaria Palma Mallorca Chania Variable (1) (2) (1) (2) (1) (2) Logarithm of fuel - cost (Fuel price + -0.0384 -0.260*** -0.332*** -0.333*** -0.235** 0.00615 ETS costs) (0.0720) (0.0605) (0.152) (0.0930) (0.111) (0.112) Monthly fixed effects Yes Yes Yes Yes Yes Yes GDP per capita in 3.57E- 1.37E- 1.95E-05*** Sweden 05*** 05*** (1.66E-06) (5.11E-06) (3.61E-06) Mean temperature -0.0130 -0.0824*** 0.0247 Gothenburg (0.00822) (0.0401) (0.0315) Sunshine duration 5.98E-06 1.31E-05 1.02E-06 Gothenburg (0.01 hours) (6.04E-06) (1.13E-05) (8.02E-06) Unemployment rate 0.322 -14.8** -3.74 in Sweden (2.24) (3.97) (3.77) Financial crisis (Sept 2008 -0.202*** -0.200 -0.0619 - Dec 2010) (0.0343) (0.0836) (0.0915) Observations 171 170 132 132 84 84 R-squared 0.911 0.961 0.831 0.915 0.843 0.893 Table 2: Results from running regressions using model 1 and 2. Robust standard errors within parentheses. *** p<0.01, ** p<0.05, * p<0.1 The results show that there is a significant relationship between the changes in airlines fuel costs and the change in number of passengers traveling from Landvetter Airport, Gothenburg to these three charter destinations. The estimated coefficients are negative and significant for the 35 regression in column two (2) and the interpretations of these imply that a one percent increase in the fuel cost is associated with a percentage reduction in number of passengers traveling to the destination ranging from 0.235 to 0.332 percent (i.e. demand is inelastic). The sign of our coefficients is consistent with previous literature and economic theory since an increase in airlines costs should correspond to a price increase in flight tickets which would decrease the demanded quantity. Comparing the magnitude of our estimates to previous literature is however difficult, since most previous studies estimate the price elasticity of demand from variations in actual ticket prices, whereas we estimate elasticities using fuel costs as a proxy for ticket prices. The next step in our analysis is to estimate the change in costs for airlines facing the removal of free allocation of emission allowances in 2026. For this, we calculate the equivalent necessary allowances per liter of jet fuel and multiply this with the corresponding price of emission allowances. The calculation follows the same principle as applied on the data on fuel price included in our regression. Since the share of freely allocated allowances will be zero, and as prices for allowances have risen significantly since 2018, we expect to see a larger increase in airlines costs for fuel. The future price of emission allowances is an unknown parameter; therefore, we will rely on estimates and predictions from multiple sources presented in Table 3. A sensitivity analysis with the upper and lower bound of the estimates, corresponding to €70 and €100, is also provided. 36 Estimated price Period of estimation Source Enerdata, November €70 - €75 2020 - 2030 2023 (Enerdata, 2023) IETA, April 2023 (IETA, €100 2026 - 2030 2023) Reuters, January 2024 €100.13 2026 (Twidale, 2024) Table 3: Price estimates for EUAs. Using these predictions, we estimate that the increase in fuel costs for airlines in 2026 due to costs of emission allowances will be €0.218 per liter of jet fuel (ranging from 0.1792 to 0.2560 EUR/liter). The mean value of fuel costs in our dataset is €0.4330 per liter, and consequently, our estimated percentage increase in fuel costs in 2026 lie between 41 and 59%. Results are presented below in Table 4. For more detailed calculations see Appendix A. Assumed ETS price € 70 € 85 € 100 Emission price per liter of fuel (EUR/liter) 0,1792 0,2176 0,256 Percentage increase in fuel cost 41.38% 50.25% 59.12% Table 4: Estimated increase in fuel cost 2026 under three different assumed ETS prices. Using our estimated elasticities from the regressions, presented in Table 2, column two (2) above, we can calculate the yearly change in passengers. The result is presented in Table 5. The estimated decrease in number of passengers traveling to Chania, Palma Mallorca and Gran Canaria from Landvetter in 2026 would be 3 484, 6 753 and 7 248 passengers respectively. The total reduction in the number of passengers for these destinations should lie in the interval between 14 399 and 20 570 people, depending on the price of emission allowances. For a more detailed explanation of our calculations see Appendix A. 37 Changes in passengers 2026 € 70 € 85 € 100 Chania Percentage -9.72% -11.80% -13.88% Absolute -2869 -3484 -4098 Gran Canaria Percentage -10.74% -13.04% -15.35% Absolute -5969 -7248 -8527 Palma Mallorca Percentage -13.74% -16.69% -19.63% Absolute -5561 -6753 -7945 Table 5: Estimated reduction in number of passengers, given different EUA prices, to the three destinations, presented in percentage change and absolute numbers. Our results show that the greatest reduction in passengers will be for travel to Palma Mallorca, almost 17% reduction compared to the average number of yearly passengers. To answer our first research question, the demand for leisure travel from Landvetter due to changes in ETS cost in 2026 will result in significant reductions in number of passengers. 5.2 Estimating carbon leakage Previously we have estimated the expected demand change in travel to Chania, Palma Mallorca, and Gran Canaria. It is now of interest to evaluate the substitutions that these travelers can make. Following Brons et al. (2002) model of travel substitution (see chapter 2.2.3) we have estimated the change in consumption in level three of substitution, i.e. change in travel expenditure destination A by airplane. It is possible that the decrease in passengers by aviation to our three destinations (corresponding to destination A in the Brons model) will be distributed over multiple levels of travel substitution. For example, between modes of travel (level three), between destinations (level two) and to non-travel consumption (level one). Substitutions in level four, meaning between airlines, will not be analyzed since we assume that cost increases affect all airlines in the same way. 38 For our analysis of carbon leakage, we put forth four different scenarios for travel substitution. The share of the estimated passenger reduction to our three destinations who are now opting to travel to another destination - substitution in level two - is increasing for each scenario. The decrease of passengers used in this calculation is the sum of passenger decreases to each destination using the predicted EUA price of €85. The resulting emissions of CO2 from the increased travel to Antalya is derived from the ICAO emissions calculator. The results are presented below in Table 6. For a more detailed explanation of our calculations, see Appendix B. Scenario (1) (2) (3) (4) Level One 75% 50% 25% 0% Level Two 25% 50% 75% 100% Passenger increase to Antalya 4 371 8 742 13 113 17 484 1 726 3 453 5 179 6 906 Kg of CO2 emissions 575.09 150.17 725.26 300.35 Table 6: Outcomes on number of passengers and CO2 emissions for each of the four scenarios. Table 6 represents the answer to our second research question, regarding the potential carbon leakage from EU ETS caused by the aviation policy change. The results show a leakage of about 1.7 to 6.9 million kg CO2, in absolute terms. When conducting a sensitivity analysis with elasticities of demand from Kopsch (2012) and Stråle (2021) results are similar, with a range of carbon leakage from approximately 1.7 to 6.7 million kg CO2. See Appendix C for detailed calculations. 39 6 Discussion and conclusions Our regression results show that the removal of free allocation of emission allowances to airlines operating flights covered by the EU ETS would have a significant negative effect on the demand for leisure travel from Gothenburg to some key charter destinations in the EU. This answers our first research question. According to economic theory, this is due to the internalization of the social costs arising from the negative externalities of air travel, which affects the market equilibrium to a more socially optimal level of consumption. The predicted price of EUAs in 2026 is in the lower range of the estimated SCC by Rennert et al. (2022), implying that even higher prices of EUAs would improve market efficiency. However, if the risk of carbon leakage is not accounted for, the effectiveness of increased carbon costs in the EU could be in jeopardy. Our estimated price elasticities of demand, in the range of -0.23 to -0.33, are lower than what has been found in previous research by e.g. Brons et al. (2002), Kopsch (2012), and Stråle (2021). This was expected, since we used fuel cost as a proxy for ticket prices. Considering that fuel cost constitutes about 20 to 50 percent of airline costs (Koopmans and Lieshout, 2016), our estimates are well in line with what Kopsch (2012) and Stråle (2021) finds. Our most interesting finding in this regard is that removing the free allocation of EUAs could have a significant effect on demand. We find that the policy change would lead to a reduction in demand of about 10 to 20 percent, depending on the destination and future permit price. It is worth noting that we specifically have chosen charter destinations, which can be expected to have a higher rate of leisure travel for which demand has been shown to be more sensitive to price changes (Kopsch, 2012). However, even with this in mind, the policy change might cause considerable effects on aviation on both national and EU level. As for our second research question, regarding the potential carbon leakage originating from the demand change, we find a leakage of about 1.5 to 7 million kg CO2. This is only the predicted leakage arising from behavioral changes to the three destinations covered in this study. A larger scope of analysis for the policy change would provide a better picture of the extent of the issue. Nonetheless, even at this small scale, carbon leakage is present and impact the efficiency of 40 global CO2 mitigation from the EU ETS. The efficiency is even further problematized since the climate effectiveness of CORSIA, which aims to regulate the global environmental impact of the aviation industry, has been strongly questioned. 6.1 Method discussion There are some limitations to acknowledge in our calculations and econometric model. Firstly, it must be mentioned that we were not able to find a significant price elasticity on demand when running a regression with panel data including more destinations. This problem likely arises due to the limitations in available passenger data. This also causes uncertainty to our estimators presented in this study. Secondly, to simplify the study, we have only estimated elasticities for direct flights to the destinations. In reality, people travel in different ways, using for example stopovers to get a cheaper price. However, since our analysis is based on charter destinations, it might be reasonable to assume that direct travel is popular. Furthermore, if the length of travel is increased through stopovers, ticket prices for these routes should rise more compared to a direct transit. What effect this limitation has had on our results is evidently difficult to determine. Thirdly, in our model we control for income by including GDP/capita in the regression, which has a significant positive effect on demand for travel to all three destinations. This means that if income increases in the future, the studies policy change will, likely, not reduce demand to the extent that we have estimated. However, it is important to note that an income increase would affect demand for inter-EU flights as well. Finally, our regression model is, of course, not a complete model of the reasons people chose to travel by airplane, and it is possible that controlling for other variables, such as destination specific variables or exchange rates, would improve the model. The use of fuel cost as the base of our regression to estimate demand change has some implications on the assumed pass-through of ETS costs. Specifically, that variations in airlines costs from fuel and EUAs will have the same effect on demand changes. Since consumers are not made aware of exactly what they are paying for, we must assume that the pass-through of variations in fuel prices and EUA costs are the same. Regarding our estimation of carbon leakage, we based the analysis on hypothetical scenarios of travel substitution, due to the difficulties in estimating the leakage using quantitative methods. 41 Hence, this result should not be interpreted as a statistical estimation of the leakage, but rather, as a rough approximation of size of a potential leakage. Further, this means that our results are coupled with high uncertainty. The leakage could be higher than predicted if people would choose to substitute journeys to the EU destinations to destinations further away, e.g. Thailand or the US. It might also be lower than predicted if people choose not to travel or to travel using a mode of travel with lower emissions, e.g. train. However, these substitutions are to some extent less probable since they are all associated with significantly increased travel times. 6. 2 Policy implications The potential problem with the statutory future changes of the EU ETS, with removal of the free allocation of emission allowances to the aviation sector, is that this possibly distorts the market and, thus, could lead to carbon leakage. This happens if the stringency in climate policy differs between inter- and intra-continental flights, which is a real possibility because of the large differences in functionality between EU ETS and the CORSIA agreement (Scheelhaase et al., 2018). When aviation was included in the ETS agreement, the plan was for all flights from European airports to be included in the scheme. However, strong international opposition led to the so-called “Stop the Clock” decision, which meant that the inclusion of aviation in the EU ETS was limited to only cover emissions from flights within the European Economic Area (EEA) (Scheelhaase et al., 2018). Our analysis is based on 2026, i.e. the year where a gap in policy can be expected between intra- and inter-continental aviation in the EU. As previously mentioned, the plan of the EU-commission is that inter-continental aviation in the EU will be included in EU ETS starting in 2027, unless ICAO by 2025 has made decisions that are judged to secure a climate-neutral aviation by 2050. However, since EU has withdrawn the decision to include inter-continental aviation before, perhaps this could happen again? In that case, the validity of our results expands beyond 2026. To mitigate the potential carbon leakage that arises, it is necessary to reduce the difference in stringency between the different markets, in this case, intra- and inter-EU aviation. One alternative to achieve this is to include all aviation with either departure or destination in the EU in the EU ETS, which would be in accordance with the EU ambition. Another alternative is to implement national flight taxes, specifically for inter-continental flights, to counteract the relative price change on European air travel. This would be relevant from a Swedish policy perspective 42 where aviation from Swedish airports, as of now, is covered by a per-seat tax, i.e. the Swedish aviation tax. The tax has been debated, and its functionality from a climate perspective has been criticized by e.g. Swedish transport sector representatives (Nygren, 2023). However, the tax has been defended by researchers from Chalmers University of Technology, University Gothenburg and KTH Royal Institute of Technology (Larsson et al., 2019). From a CO2 emission mitigation perspective, it is true that the tax, in theory, has no effect on CO2 emissions from intra-EU aviation due to the waterbed effect (leaving out effects of the Market Stability Reserve). However, as our analysis show, in the absence of complementary policies for inter-continental aviation, increasing costs from EU ETS for intra-EU aviation might lead to significant leakage of CO2. The Swedish aviation tax has the potential to act as such a second-best complementary policy to mitigate carbon leakage. The elephant in the room regarding climate policy for the aviation sector, which is not addressed in this study, is the climate effects of non-CO2 forcings. These forcings are not included in the EU ETS. However, Lee et al. (2021) finds that 66 percent of the warming effect from aviation in 2018 came from non-CO2 climate forcings. Thus, even in the absence of carbon leakage, inclusion of aviation in the EU ETS, and removal of the free allocation of emission allowances, is not enough for a cost-effective aviation climate policy. The different climate forcings affects the climate on different time scale, and paradoxically, this means that changes in aviation can have a positive and a negative effect on temperature increase at the same time, depending on the time scale (Lee et al., 2021). For instance, re-routing flights to daytime could decrease the short-term warming from contrail formation, while at the same time increase long-term warming from increased CO2 emission (if the flight duration increases). From a policy perspective, this makes aviation an interesting case. However, to be able to show the mechanisms of carbon leakage in a simplified way, we decided to not include these effects in our analysis. In this study, we have barely scratched the surface of the climate policy for aviation in the EU. This study only provides an illustration of how carbon leakage from distorting policies in the aviation sector can arise. For future research, a more comprehensive analysis of the potential carbon leakage from the aviation sector in EU as a whole would be interesting to be able to understand the size of the issue. 43 7 References BREUSCH, T. S. & PAGAN, A. R. 1979. A simple test for heteroscedasticity and random coefficient variation. Econometrica: Journal of the econometric society, 1287-1294. BRONS, M., PELS, E., NIJKAMP, P. & RIETVELD, P. 2002. Price elasticities of demand for passenger air travel: a meta-analysis. Journal of Air Transport Management, 8, 165-175. DRAY, L. & DOYME, K. 2019. Carbon leakage in aviation policy. Climate policy, 19, 1284-1296. EMBRAER 2023. Airport Planning Manual. ENERDATA 2023. Carbon price forecast under the EU ETS Executive Brief – November 2023. EUROCONTROL 2024. European aviation overview 2023. EUROPEAN CENTRAL BANK. 2024. US dollar (USD [Online]. Available: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange _rates/html/eurofxref-graph-usd.en.html [Accessed 2024-04-20]. EUROPEAN CLIMATE ASSESSMENT & DATASET 2024. Indices data. EUROPEAN COMMISSION. 2024a. Carbon Border Adjustment Mechanism [Online]. Available: https://taxation-customs.ec.europa.eu/carbon-border-adjustment-mechanism_en [Accessed 2024-05-10]. EUROPEAN COMMISSION. 2024b. Development of EU ETS (2005-2020) [Online]. Available: https://climate.ec.europa.eu/eu-action/eu-emissions-trading-system-eu-ets/development- eu-ets-2005-2020_en [Accessed 2024-04-09]. EUROPEAN COMMISSION. 2024c. Emissions cap and allowances [Online]. Available: https://climate.ec.europa.eu/eu-action/eu-emissions-trading-system-eu-ets/emissions- cap-and-allowances_en [Accessed 2024-04-09]. EUROPEAN COMMISSION. 2024d. Reducing emissions from aviation [Online]. Available: https://climate.ec.europa.eu/eu-action/transport/reducing-emissions-aviation_en [Accessed 2024-04-03]. EUROPEAN ENVIRONMENTAL AGENCY. 2023. EU Emissions Trading System (ETS) Data Viewer [Online]. Available: https://www.eea.europa.eu/data-and- maps/dashboards/emissions-trading-viewer-1 [Accessed 2024-04-09]. EUROSTAT 2024a. Air passenger transport routes between partner airports and main airports in Sweden. EUROSTAT 2024b. Gross domestic product (GDP) at market prices - quarterly data. FAGEDA, X. & TEIXIDÓ, J. J. 2022. Pricing carbon in the aviation sector: Evidence from the European emissions trading system. Journal of Environmental Economics and Management, 111, 102591. FEICHERT, F., FORSYTH, P. & NIEMEIER, H.-M. 2020. Aviation and climate change: economic perspectives on greenhouse gas reduction policies, Milton Park, Abingdon, Oxon, New York, NY : Routledge. ICAO 2016. Resolution A39-3: Consolidated Statement of Continuing ICAO Policies and Practices Related to Environmental Protection—Global Market-Based Measure (MBM) Scheme. International Civil Aviation Organization Montreal, QC, Canada. ICAO 2018. Carbon Emissions Calculator Methodology version 11, June 2018. ICAO 2020. CORSIA States for Chapter 3 State Pairs Edition 1. ICAO 2023. CORSIA States for Chapter 3 State Pairs Edition 4/ rev. 1. ICAO. 2024. Carbon Emissions Calculator [Online]. Available: https://applications.icao.int/icec/Home/Index [Accessed 2024-05-15]. 44 IETA. 2023. Average carbon price expectations worldwide from 2022 to 2030, by trading system (in euros per metric ton of CO₂ [Online]. Statista: Statista Inc. Available: https://www- statista-com.ezproxy.ub.gu.se/statistics/1334906/average-carbon-price-projections- worldwide-by-region/ [Accessed 2024-05-14]. IPCC 2018. Summary for Policymakers. Global Warming of 1.5°C: IPCC Special Report on Impacts of Global Warming of 1.5°C above Pre-industrial Levels in Context of Strengthening Response to Climate Change, Sustainable Development, and Efforts to Eradicate Poverty. Cambridge: Cambridge University Press. KOOPMANS, C. & LIESHOUT, R. 2016. Airline cost changes: To what extent are they passed through to the passenger? Journal of Air Transport Management, 53, 1-11. KOPSCH, F. 2012. A demand model for domestic air travel in Sweden. Journal of Air Transport Management, 20, 46-48. LARSSON, J., ELOFSSON, A., STERNER, T. & ÅKERMAN, J. 2019. International and national climate policies for aviation: a review. Climate Policy, 19, 787-799. LEE, D. S., FAHEY, D. W., SKOWRON, A., ALLEN, M. R., BURKHARDT, U., CHEN, Q., DOHERTY, S. J., FREEMAN, S., FORSTER, P. M. & FUGLESTVEDT, J. 2021. The contribution of global aviation to anthropogenic climate forcing for 2000 to 2018. Atmospheric environment, 244, 117834. NILSSON, M. 2023. Temperaturhöjning i klimatpolitiken – en ESO-rapport om EU:s nya lagstiftning i svensk kontext. NYGREN, F. 2023. Expert: Så gynnas Danmark av svenska flygskatten – ”Helt orimligt”. Tidningen Näringslivet, 2023-11-23. OESINGMANN, K. 2022. The effect of the European Emissions Trading System (EU ETS) on aviation demand: An empirical comparison with the impact of ticket taxes. Energy Policy, 160, 112657. PEARCE, D. 2003. The social cost of carbon and its policy implications. Oxford review of economic policy, 19, 362-384. PERLOFF, J. 2013. Microeconomics with Calculus, Global Edition, Pearson Education UK. RENNERT, K., ERRICKSON, F., PREST, B. C., RENNELS, L., NEWELL, R. G., PIZER, W., KINGDON, C., WINGENROTH, J., COOKE, R. & PARTHUM, B. 2022. Comprehensive evidence implies a higher social cost of CO2. Nature, 610, 687-692. SCHEELHAASE, J., MAERTENS, S., GRIMME, W. & JUNG, M. 2018. EU ETS versus CORSIA – A critical assessment of two approaches to limit air transport's CO2 emissions by market-based measures. Journal of Air Transport Management, 67, 55-62. STATISTICS SWEDEN 2024. Befolkningen 15-74 år (AKU) efter kön, ålder och arbetskraftstillhörighet. Månad 2001M01 - 2024M04. STERN, N., STIGLITZ, J. & TAYLOR, C. 2022. The economics of immense risk, urgent action and radical change: towards new approaches to the economics of climate change. Journal of Economic Methodology, 29, 181-216. STOCK, J. H. & WATSON, M. W. 2020. Introduction to econometrics, Pearson. STRÅLE, J. 2021. The effects of the Swedish aviation tax on the demand and price of international air travel. SWEDAVIA. 2024. Göteborg Landvetter Airport/Destinationer [Online]. Available: https://www.swedavia.se/landvetter/destinationer/ [Accessed]. SWEDISH TRANSPORT AGENCY. 2024. ICAO:s globala klimatstyrmedel - CORSIA [Online]. Available: https://www.transportstyrelsen.se/sv/luftfart/Miljo-och- halsa/Klimat/Klimatstyrmedel/icaos-globala-klimatstyrmedel/ [Accessed 2024-05-16]. TRADING ECONOMICS. 2024. EU Carbon Permits [Online]. Available: https://tradingeconomics.com/commodity/carbon [Accessed 2024-04-20]. 45 TWIDALE, S. 2024. Analysts cut EU carbon price forecasts on weak industry, power sector demand. . Reuters, 2024-01-23. US ENERGY INFORMATION AGENCY 2024. Petroleum & Other Liquids. WEITZMAN, M. L. 1974. Prices vs. Quantities. The Review of economic studies, 477-491. WOZNY, F. 2024. Tax Incidence in Heterogeneous Markets: The Pass-through of Air Passenger Taxes on Airfares. WOZNY, F., GRIMME, W., MAERTENS, S. & SCHEELHAASE, J. 2022. CORSIA—A Feasible Second Best Solution? Applied Sciences, 12, 7054. 46 Appendix A - Calculations of demand reduction To calculate the percentage fuel cost increase in 2026 we need the following parameters: • Estimated price of EUAs in 2026 = €70, €85 and €100 • Average fuel cost for the data points in our regression = €0.433 per liter • Conversion factor for necessary EUAs per liter of kerosene defined through the density and carbon dioxide content of kerosene = 0.00256 EUAs/ liter The percentage increase can then be calculated as follows: €70 ∗ 0.00256 %∆𝑓𝑢𝑒𝑙𝑐𝑜𝑠𝑡 = = 41.38% 0.433 €85 ∗ 0.00256 %∆𝑓𝑢𝑒𝑙𝑐𝑜𝑠𝑡 = = 50.25% 0.433 €85 ∗ 0.00256 %∆𝑓𝑢𝑒𝑙𝑐𝑜𝑠𝑡 = = 59.12% 0.433 Calculating the change in number of passengers traveling to our destinations in 2026 is carried out with the following parameters: • %∆𝑓𝑢𝑒𝑙𝑐𝑜𝑠𝑡 presented above. • Our estimated elasticity from the regression presented in chapter 5.1 • Average yearly passengers to each destination The calculations are carried out as follows: ∆𝑛𝑟𝑝𝑎𝑠𝑠𝑒𝑛𝑔𝑒𝑟𝑠 = %∆𝑓𝑢𝑒𝑙𝑐𝑜𝑠𝑡 ∗ 𝑒𝑙𝑎𝑠𝑡𝑖𝑐𝑖𝑡𝑦 ∗ 𝑎𝑣𝑛𝑟𝑝𝑎𝑠𝑠 47 Appendix B - Calculations of carbon leakage Our estimation of carbon leakage is based on the following parameters: • Total passenger decrease to our three destinations of interest • Share of passengers substituting travel in level 2, i.e. to Antalya • CO2 emissions per passenger for a round trip to Antalya These parameters are multiplied to retrieve the total increase in emissions, which in this essay corresponds to the carbon leakage. 48 Appendix C - Sensitivity analysis with other elasticities To further test the validity of our results we decided to conduct a calculation of carbon leakage using price elasticities of demand for air travel. The values used in the calculation are found in Kopsch (2012) and Stråle (2021). These elasticities are based on how the demanded quantity, i.e. number of passengers, depends on the price of flight tickets, our data does not include ticket prices and our variable of interest is fuel costs, this requires some adjustments of the values in order to perform the sensitivity analysis more correctly. There are two possible ways to combine the elasticities and the costs from ETS to derive the effects on demand. One way is to consider the lever of pass-through of ETS costs to the ticket price, according to Koopmans and Lieshout (2016), a pass through of 50% is reasonable for sector specific costs in the airline industry. However, since elasticities require the use of a percentage change in price we would also need the average ticket price for each of our destinations, which is data we do not have. Another option is to consider the share that fuel costs make-up of the airline’s total costs, which then would assume to be reflected in the share of the ticket price. Fuel costs make up approximately 20 to 50 percent of the airline’s total costs (Koopmans and Lieshout, 2016), The following calculations can then be carried out to estimate the total carbon leakage with elasticities from the literature. %∆𝑛𝑟𝑝𝑎𝑠𝑠𝑒𝑛𝑔𝑒𝑟𝑠 = %∆𝑓𝑢𝑒𝑙𝑐𝑜𝑠𝑡 ∗ 𝑎𝑣𝑓𝑢𝑒𝑙𝑐𝑜𝑠𝑡𝑠ℎ𝑎𝑟𝑒 ∗ 𝑎𝑣𝑒𝑙𝑎𝑠𝑡𝑖𝑐𝑖𝑡𝑦 Where avfuelcostshare takes the value 35% an represents the average share that fuel costs make up for airlines. Avelasticity takes on the value 0.774 and is the average of the elasticities from Kopsch and Stråle. The other variabels follow the definition presented above. The results of these calculations are presented here in Table 7. There are similar reductions in number of passengers as our estimations presented in Table 5. 49 Assumed ETS price € 70 € 85 € 100 Emission price per liter of fuel (EUR/liter) 0,1792 0,2176 0,256 Percentage increase in fuel cost 41,38% 50,25% 59,12% Percentage increase in ticket price 14,48% 17,59% 20,69% Changes in passengers 2026 Chania Percentage -11,21% -13,61% -16,01% Absolute -3309 -4018 -4727 Gran Canaria Percentage -11,21% -13,61% -16,01% Absolute -6228 -7563 -8897 Palma Mallorca Percentage -11,21% -13,61% -16,01% Absolute -4536 -5508 -6480 Table 7: Estimated percentage increase in fuel cost and ticket price 2026 under three different assumed ETS prices and estimated reduction in number of passengers to the three destinations, presented in percentage change and absolute numbers. The carbon leakage is then estimated following the method presented above. The results are presented in Table 8 and yield similar estimations of carbon leakage in the range of 1.7 to 6.7 million kg CO2. 50 Scenario (1) (2) (3) (4) Level One 75% 50% 25% 0% Level Two 25% 50% 75% 100% Passenger increase to Antalya 4 272 8 544 12 817 17 089 Kg of CO2 emissions 1687522.98 3375045.96 5062568.94 6750091.93 Table 8: Outcomes on number of passengers and CO2 emissions for each of the four scenarios using elasticities from the literature. 51