Analyzing the EU ETS, Challenges and Opportunities for Reducing Greenhouse Gas Emissions from the Aviation Industry in Europe

Abstract

The aviation industry’s contribution to global greenhouse gas (GHG) emissions has been on an unsteady rise for the past few decades. This paper aims to identify the determinants of increasing GHG emissions in Europe in a dynamic panel setting, paying specific attention to the role of the European Union Emissions Trading System (EU ETS). Unlike previous studies, this paper proposes business tourism spending and capital investment in the tourism and travel industry as explanatory factors together with GDP per capita and jet fuel consumption. Unexpectedly, the EU ETS coverage is found to have an increasing role for GHG emissions from international aviation in countries where the system is put into effect. The results suggest that a more targeted emissions reduction policy needs to be implemented in order to mitigate aviation emissions in the region.

1. Introduction

In spite of various headwinds, such as the COVID-19 pandemic, the global aviation industry has, overall, been growing, accompanied by an increased contribution to global GHG emissions, accounting for 2% of global energy-related carbon dioxide (CO₂) emissions in 2022 [1]. In recent years, the growth rate has been higher than that associated with rail or road transportation or the shipping industry. While efforts are made to mitigate these emissions, the aviation industry arguably remains a source of threat in the context of the climate crisis. Except for the COVID-19 pandemic period, aviation emissions under the EU ETS were generally on a climb on an annual basis, totaling 49.2 × 10⁶ metric tons of carbon dioxide equivalents (MtCO₂e) in 2022 [2]. Nevertheless, a global political consensus to reduce the sector’s climate impact exists. In 2022, the 184 member states of the International Civil Aviation Organization (ICAO) adopted a long-term global aspirational target of net zero carbon emissions from international aviation by 2050. Furthermore, the EU aims to reduce emissions from transport, including aviation, by 90% by the year 2050 (compared to 1990 levels) in order to achieve its net zero emissions targets.

The aim of this study is to identify the main factors that affected the level of GHG emissions in the 1990–2021 period in Europe and to examine and compare practices, policies, and actions introduced by countries in order to reduce their aviation emissions. As the EU ETS is the main tool to combat GHG emissions in the EU, it is expected to lead to reductions in aviation emissions. However, this paper finds contradictory results for the role of the EU ETS in terms of its effects on GHG emissions from aviation.

Previous studies have analyzed the factors affecting GHG emissions in the aviation industry, focusing on specific airlines, airports, countries, or the whole globe. For instance, Olsthoorn [3] analyzed past data and forecast future emissions (1950–2050) from international aviation worldwide, finding that global GDP and jet fuel consumption increased emissions, whereas oil prices decreased them. The author also found that CO₂ emissions might increase by a factor of 3 to 6 from 1995 to 2050 depending on the extent of trade liberalization, capital market integration, and policies such as energy taxation that will be in effect globally [3]. Williams and Noland [4] demonstrated that long-distance flights offered lower CO₂ per km rates than shorter ones. Yue and Byrne [5] analyzed various factors affecting airline carbon emissions in European aviation. They indicated that CO₂ emissions were positively correlated with the number of passengers and flights. The route distance was found to increase overall CO₂ emissions; however, it reduced CO₂ emissions per available seat kilometer. Moreover, carbon emissions decreased when fuel prices rose. Lo et al. [6] used similar variables to those employed by Yue and Byrne [5], but they specifically focused on Lombardy, Italy. They pointed out that distance and aircraft size had a positive effect on CO₂ emissions from aviation. However, contrary to Yue and Byrne [5], and somewhat counterintuitively, Lo et al. [6] found that fuel price and CO₂ emissions were positively correlated.

Several studies have examined the effectiveness of the EU ETS. Using data at the airline-route level, Fageda and Teixidó [7] controlled for the effects of the total number of seats, flight frequencies, aircraft size, distance flown, the operating airline, and the aircraft type, and found that, when compared to non-ETS airline emissions, the EU ETS lowered emissions by 4.7% from 2010 to 2016. The authors chose a very short time range in order to implement the difference-in-differences method to compare two time periods more-or-less equal in length. Fageda and Teixidó [7] also found that low-cost airlines had the greatest impact on the estimated effects of the EU ETS on emissions. On routes where alternative modes of transport were a competitive choice, the EU ETS had a significantly greater impact. In contrast, Heiaas [8] deemed the EU ETS to be not successful in bringing the emissions down. In comparison to a scenario where the EU ETS was not implemented, a 10% increase in emissions was found to be associated with being regulated by the EU ETS [8].

Inspired by the operationalization of the EU ETS, Pang and Chen [9] studied three emissions pooling scenarios for the airline industry, namely a pay premium strategy, a market-based strategy, and a cooperative strategy, each corresponding to different market structures. They demonstrated that the EU ETS was always detrimental to airlines using older aircraft and that a high premium could effectively encourage airlines to pursue a low-carbon transition plan. Their results confirm that the EU ETS is not a panacea to the emissions problem and that it only works when the premium mechanism is properly designed and a fair market for emissions allowances exists.

In the sustainability literature, there are also papers on the impacts of other climate and energy policies on emissions from aviation. One example is the work of Cui et al. [10], who investigated the impact of the “13th Five-Year Plan” of civil aviation in China on energy conservation and emissions reduction in domestic routes between 2014 and 2019. The plan mainly intended that, by 2020, the average energy consumption of unit transport turnover and carbon dioxide emissions of civil aviation would drop considerably. The authors found that the plan had a negligible effect in mitigating GHG emissions. In their study of China’s carbon emissions from aviation, Han et al. [11] evidenced the positive effects of air transportation revenue and the aviation route structure on emissions, whereas the air transportation intensity and aviation energy intensity were found to inhibit aviation carbon emissions. Similarly, He et al. [12] evidenced that energy consumption per unit of turnover restrained global carbon emissions growth by about 8%. On the other hand, they evidenced that population size, GDP per capita, and turnover per unit of GDP contributed to the growth in carbon emissions from global civil aviation, by 5%, 82%, and 21%, respectively.

In the current study, the sample is comprised of the countries in the European region in order to emphasize the regional effects of several factors as well as to see whether the EU ETS coverage makes a difference in triggering or slowing down GHG emissions. In this study, two variables that have not been previously utilized are used as explanatory factors. These are business tourism spending (BTS) and capital investment in travel and tourism (CAPITAL). Focusing on the 1990–2021 period, a considerably long time span allows for the analysis of pre- and post-ETS effects on aviation emissions, as CO₂ emissions from aviation have been included in the EU ETS since 2012. By including Switzerland and Türkiye in the sample, which are not members of the EU and not covered by the EU ETS, and the United Kingdom (UK), which is no more an EU member but whose emissions are covered by the UK ETS as of 1 January 2021, heterogeneity in the region is accounted for in terms of climate policy.

The paper is comprised of five parts. The present introductory section is followed by Section 2 which presents the materials and methods that are utilized in the paper. Section 3 provides the results of the econometric analyses. Section 4 discusses the findings, and, finally, Section 5 serves as a conclusion.

2. Materials and Methods

2.1. Materials

According to data based on the Climate Analysis Indicators Tool (CAIT) [13], as of 2019, the three main sources of GHG emissions are electricity and heating, transport, and the manufacturing and construction sectors. Transportation accounts for about 16% of total greenhouse gas emissions and around 2% of total emissions come from aviation [14]. From 1990 to 2019, GHG emissions from aviation and shipping increased by more than 100% relative to other sectors [14]. Figure 1 shows the change in each industry’s GHG emissions from 1990 to 2019.

Figure 2 demonstrates the percentage change in emissions in 2018 compared to 1990 levels. The emissions figures for aviation and shipping have increased by 107% since 1990. This implies that although the aviation and shipping sector has a small share of the overall emissions, in terms of the rate of emissions growth, it is second only to industry.

Since the study focuses on Europe, sectoral emissions data for the EU27 are presented in Figure 3. In Europe, aviation and shipping appears to be the sector with the highest emissions growth. If no measures are taken, the aviation industry in Europe is on track to become one of the biggest GHG emitters in the future.

Our analysis covers the European continent. The sample includes all 27 European Union countries (EU27) (Austria, Belgium, Bulgaria, Croatia, Republic of Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden) as well as five other European countries, which are Iceland, Norway, Switzerland, Turkey, and the UK. The aviation emissions of all these countries are illustrated in Figure 4. Until the severe COVID-19 pandemic years of 2019 and 2020, the countries in the continent displayed an increasing trend in their aviation emissions. The UK and Germany had, by far, the highest levels of emissions among the others. Slovenia and Estonia had the lowest levels of aviation emissions as of 2018.

Since the sustainability policies in the European Union apply to all EU countries, the EU27 countries are considered to be a more suitable sample group for comparing GHG emissions through time, and the most convenient dataset in terms of data availability. In addition, as the United Kingdom was also an EU country before Brexit, it is also included in this dataset. Iceland and Norway are on this list because they are in the European Economic Area (EEA). Switzerland, on the other hand, is part of the single market, even if it is not an EU or EEA member. Finally, the sample includes Turkey, which started its preparations to implement an ETS long before the announcement of the European Green Deal, and, with an EU membership perspective, which has established the legal infrastructure of a monitoring, reporting, and verification (MRV) system in line with the EU ETS.

Annual data covering the period from 1990 to 2021 are used. Details regarding the dependent and independent variables utilized in the empirical analysis are shown in

Table 1. Definitions, units, and sources of the variables used in the analysis.

Variable	Abbreviation	Definition	Unit	Source
GHG Emissions from Aviation Industry	GHG	Greenhouse gas emissions from international aviation	Thousand Tonnes	Eurostat and OECD
GDP Per Capita	GDPPC	Gross domestic product per capita in constant USD	USD (Real prices)	World Bank
Jet Fuel Consumption	JETFUEL	Fuel used by airplanes to cover a certain distance	1000 Metric Tonnes	U.S. Energy Information Administration
Business Tourism Spending	BTS	Total consumption expenditure made by business travelers	bn USD (Real prices)	WTTC
Capital Investment in Travel and Tourism	CAPITAL	Governmental investment on physical assets in the travel and tourism Industry	bn USD (Real prices)	WTTC
EU ETS Coverage of Aviation	ETS	EU policy to reduce GHG emissions cost-effectively	1 if covered by ETS, 0 if not	Eurostat

.

The dependent variable is the logarithm of GHG emissions from the aviation industry (LNGHG). The independent variables used to explain emissions are the logarithms of GDP per capita (LNGDPPC), jet fuel consumption (LNJETFUEL), business tourism spending (LNBTS), and capital investment in travel and tourism (LNCAPITAL). Coverage by the EU ETS is represented by a dummy variable, ETS, which takes the value “1” if the country is covered by the EU ETS in a specific year, and “0” otherwise. Although the EU ETS came into force in 2005, it only started to cover the aviation industry in 2012.

Summary statistics for these variables are presented in

Table 2. Descriptive statistics for the variables used in the analysis.

Variable	Mean	Median	S.D.	Min	Max
LNGHG	7.13	7.12	1.62	3.06	10.5
LNGDPPC	10.1	10.2	0.781	8.17	11.6
LNJETFUEL	6.18	6.27	1.64	2.48	9.44
LNBTS	0.967	0.997	1.64	−2.22	4.52
LNCAPITAL	0.730	0.831	1.45	−3.51	3.82
ETS				0.00	1.00

. To eliminate possible heteroskedasticity, all variables were used in their logarithmic forms as abbreviated above and in Table 2.

Our first hypothesis is that as GDP per capita grows, so does aviation infrastructure and GHG emissions from using the aviation-related facilities. In addition, higher GDP per capita implies higher disposable income for the citizens of a specific country and, hence, a higher number of flights. Moreover, the development of the aviation industry is dependent on very large amounts of capital investments, which requires a high GDP level, and most airports are big enough to be considered mega-projects.

Second, it is expected that higher jet fuel consumption will increase emissions from aviation. Jet fuel consumption is the main source of emissions from aviation, which has to be controlled for in an econometric analysis of aviation emissions. Traditional fossil jet fuels produce CO₂, nitrous gases, sulfate, and water vapor when combusted. These chemicals are primary greenhouse gases in the Earth’s atmosphere [15]. Although the popularity of sustainable aviation fuels has increased in the last decades, fossil fuels still dominate the sector.

Third, business tourism spending represents the total consumption expenditure for a business trip, including expenditure on goods, services, accommodation, and transportation. To the best of our knowledge, there is no study that has evaluated the relationship between business tourism spending and GHG emissions from aviation. Business tourism spending refers to the total amount spent on goods and services in the destination country by a visitor on a business trip. For trips between 750 and 1500 miles, 85% of such travel is air travel, with the rate increasing to 90% for trips longer than 1500 miles. Therefore, as mobility increases, more business-related air travel will occur. For this reason, a positive relationship between business tourism spending and GHG emissions is expected.

Fourth, it is expected that capital investment in travel and tourism will be a mitigating factor for emissions from aviation. This variable refers to government investment in physical assets in the travel and tourism industry. No study has so far examined the relationship between capital investment in travel and tourism and GHG emissions in the aviation industry. Capital investment can help countries update their capital to the latest technology and, thus, curb GHG emissions.

Finally, it is hypothesized that the implementation of a climate policy tool should lead to declining emissions. The EU’s main climate change initiative is known as the EU ETS. All EU member states, as well as Iceland, Norway, and Liechtenstein, are covered by the system, which entered into force in 2005. The EU ETS imposes a yearly cap on how many tons of GHGs a company can emit, which is gradually reduced. All participants in the system must report, monitor, and verify their emissions. Companies must maintain a certain number of allowances on hand each year to cover their emissions, or they will be punished severely. If a company exceeds its allowance numbers, the company can either purchase additional allowances from other emitters or seek to cut their emissions. On the contrary, if a company has surplus allowances, they can keep them for the next year or sell them to other emitters. In the aviation industry, the EU ETS covers all aircraft operators within the EU and the European Economic Area, regardless of origin. The system came into effect in 2012, and Croatia joined a year later. In 2012, the EU established a cap of 97% of the average emissions from 2004 to 2006. For the years 2013 through 2020, the cap was reduced by 95%. A total of 85% of the allowances are given out for free and are based on historical emissions, known as grandfathering, while 15% of the allowances are auctioned. Additional permits can be acquired through purchases from other industries [16].

2.2. Methodology

The method of analysis consists of two sets of estimations. First, the pooled version of the data for 32 countries from 1990 to 2021 is used. A list of the countries included in the sample is provided in the Section 2.1. This first set of pooled data are analyzed by least absolute deviation (LAD) estimation, which calculates a regression that minimizes the sum of the absolute deviations of the observed from the fitted values of the dependent variable. The coefficient estimates are derived using the Barrodale–Roberts simplex algorithm in Gretl. In ordinary least squares (OLS) regression, the fitted values, ${\widehat{y}}_i$ = $X_i\widehat{\beta }$, represent the conditional mean of the dependent variable—conditional, that is, on the regression function and the values of the independent variables. In LAD estimation, on the other hand, the fitted values represent the conditional median of the dependent variable. It implies that the principle of estimation for median regression is to choose $\widehat{\beta }$ which minimizes the sum of the absolute residuals. While the OLS regression has a straightforward analytical solution, LAD is a linear programming problem [17].

The second set of estimations are derived from a dynamic panel data model, i.e., the system GMM method. Our assumption is that the basic source of GHG emissions from most aviation companies in European countries is the jet fuel consumption, given the available technological constraints, past climate policies, and companies’ investment decisions [18], which renders them locked into fossil fuels and previous levels of GHG emissions. Given the documented persistence of GHG emissions from aviation, one lag of the dependent variable, LNGHG, is included in the regressions. This procedure simply assumes that today’s GHG emissions are driven by past emissions. Dynamic panel data models are better able to address the omitted variable problems and endogeneity issues than static models [19]. The results of two-step estimation are reported (with robust standard errors). Two-step system GMM estimators are efficient and more robust to heteroskedasticity and autocorrelation than one-step system GMM estimators [20]. Tests for autocorrelation of orders 1 and 2 are provided, as well as Sargan and overidentification tests and a Wald test for the joint significance of the regressors. System GMM is useful in that it helps solve the endogeneity problems that may arise from the potential correlation between the error term and the independent variables. One drawback of employing panel data estimation in the current study is the use of non-random sampling, as a panel of the countries in the EU region, whose data have been available to a large extent, is selected for the purposes of the research. While interpreting the GMM coefficients, one should remember that they are short-run estimates and provide a ceteris paribus interpretation [17].

The dynamic relationship is characterized by including a lagged dependent variable among the regressors, as in Equations (1) and (2) below:

LNGHG_it = γLNGHG_i,t−1 + β₁LNGDPPC_it + β₂LNJETFUEL_it + β₃LNBTS_it + β₄ETS_it + u_it

(1)

LNGHG_it = γLNGHG_i,t−1 + β₁LNGDPPC_it + β₂LNJETFUEL_it + β₃LNCAPITAL_it + β₄ETS_it + u_it

(2)

As business tourism spending, LNBTS (in Equation (1)), and capital investment in travel and tourism, LNCAPITAL (in Equation (2)), were found to be highly correlated variables, one of them is used at each regression equation in order to avoid multicollinearity (see

Table A1. Correlation coefficients, using the observations (missing values were skipped) (5% critical value (two-tailed) = 0.0614 for n = 1019).

LNGHG	LNGDPPC	LNJETFUEL	LNBTS	LNCAPITAL
1.0000	0.4793	0.9889	0.8330	0.8470	LNGHG
	1.0000	0.4629	0.4146	0.3129	LNGDPPC
		1.0000	0.8424	0.8599	LNJETFUEL
			1.0000	0.8822	LNBTS
				1.0000	LNCAPITAL
				ETS
				0.0575	LNGHG
				0.1501	LNGDPPC
				0.0247	LNJETFUEL
				0.0643	LNBTS
				0.0717	LNCAPITAL
				1.0000	ETS

in the Appendix A for the correlation coefficients of the variables used in the models). To eliminate possible heteroskedasticity, all variables are used in their logarithmic forms. Panel unit root (augmented Dickey–Fuller) tests and Pesaran cross-sectional dependence (CD) test results for each set of regressions are reported as well (see Appendix A,

Table A2. Augmented Dickey–Fuller test results for the variables in the analysis.

H0: all groups have unit root Augmented Dickey–Fuller tests with constant model: (1-L)y = b0 + (a−1) × y(−1) + … + e

Augmented Dickey–Fuller test for LNGHG including 3 lags of (1-L)LNGHG N = 32, Tmin = 28, Tmax = 28 Im-Pesaran-Shin W_tbar = −3.66442 [0.0001]

Augmented Dickey–Fuller test for LNGDPPC including 3 lags of (1-L)LNGDPPC N = 32, Tmin = 23, Tmax = 28 Im-Pesaran-Shin W_tbar = −1.502 [0.0665]

Augmented Dickey–Fuller test for LNJETFUEL including 3 lags of (1-L)LNJETFUEL N = 32, Tmin = 24, Tmax = 28 Im-Pesaran-Shin W_tbar = −3.17198 [0.0008]

Augmented Dickey–Fuller test for LNBTS including 3 lags of (1-L)LNBTS N = 32, Tmin = 23, Tmax = 23 Im-Pesaran-Shin W_tbar = −2.02399 [0.0215]

Augmented Dickey–Fuller test for LNCAPITAL including 3 lags of (1-L)LNCAPITAL N = 32, Tmin = 23, Tmax = 23 Im-Pesaran-Shin W_tbar = −1.62979 [0.0516]

and

Table A3. Pesaran CD tests for cross-sectional dependence.

Model 1a Test statistic: z = 80.621118, with p-value = P(|z| > 80.6211) = 0 Average absolute correlation = 0.698

Model 1b Test statistic: z = 80.862626, with p-value = P(|z| > 80.8626) = 0 Average absolute correlation = 0.699

Model 2a Pesaran CD test for cross-sectional dependence Test statistic: z = 62.115487, with p-value = P(|z| > 62.1155) = 0 Average absolute correlation = 0.549

Model 2b Pesaran CD test for cross-sectional dependence Test statistic: z = 60.911877, with p-value = P(|z| > 60.9119) = 0 Average absolute correlation = 0.537

). The CD test is applicable to heterogeneous dynamic panels and is robust to structural breaks.

3. Results

Below, Models 1a and 1b display the findings of the LAD estimation (details available in the Appendix A,

Table A4. LAD estimation results with LNGHG as the dependent variable.

Model 1a: LAD, using 864 observations Dependent variable: LNGHG
	Coefficient	Std. Error	t-ratio	p-value
Const	0.340809	0.0727548	4.684	<0.0001	***
LNGHG(−1)	0.593586	0.0598860	9.912	<0.0001	***
LNGDPPC	0.0124326	0.00638154	1.948	0.0517	*
LNJETFUEL	0.403893	0.0604653	6.680	<0.0001	***
LNBTS	−0.0129407	0.00494968	−2.614	0.0091	***
ETS	0.0213874	0.00661634	3.233	0.0013	***
Median depend. var	7.229313		S.D. dependent var	1.623945
Sum absolute resid	87.85923		Sum squared resid	27.77170
Log-likelihood	512.0839		Akaike criterion	−1012.168
Schwarz criterion	−983.5984		Hannan–Quinn	−1001.233
Model 1b: LAD, using 864 observations Dependent variable: LNGHG
\	Coefficient	Std. Error	t-ratio	p-value
const	0.348946	0.0707652	4.931	<0.0001	***
LNGHG(-1)	0.602387	0.0564831	10.66	<0.0001	***
LNGDPPC	0.0108811	0.00751067	1.449	0.1478
LNJETFUEL	0.394637	0.0573136	6.886	<0.0001	***
LNCAPITAL	−0.0151824	0.00621690	−2.442	0.0148	**
ETS	0.0188537	0.00639177	2.950	0.0033	***
Median depend. var	7.229313		S.D. dependent var	1.623945
Sum absolute resid	87.96277		Sum squared resid	27.82255
Log-likelihood	511.0663		Akaike criterion	−1010.133
Schwarz criterion	−981.5632		Hannan-Quinn	−999.1975

Legend: *, **, *** indicate significance at the 0.10, 0.05, and 0.01 levels, respectively.

), whereas Models 2a and 2b present the dynamic panel data estimation results (details available in the Appendix A,

Table A5. Dynamic panel estimation results with LNGHG as the dependent variable.

Model 2a: 2-step dynamic panel, using 864 observations Included 32 cross-sectional units Including equations in levels Dependent variable: LNGHG
	Coefficient	Std. Error	z	p-value
LNGHG(−1)	0.352611	0.0172018	20.50	<0.0001	***
LNGDPPC	0.0698493	0.00397410	17.58	<0.0001	***
LNJETFUEL	0.643044	0.0171065	37.59	<0.0001	***
LNBTS	−0.0239465	0.00535430	−4.472	<0.0001	***
ETS	0.00403667	0.00242435	1.665	0.0959	*
Sum squared resid	26.23752		S.E. of regression	0.123222
Test for AR(1) errors: z = −2.16065 [0.0307] Test for AR(2) errors: z = 0.734376 [0.4627] Sargan over-identification test: Chi-square(481) = 1200.59 [0.0000] Wald (joint) test: Chi-square(5) = 4.30412 × 106 [0.0000]
Model 2b: 2-step dynamic panel, using 864 observations Included 32 cross-sectional units Including equations in levels Dependent variable: LNGHG
	Coefficient	Std. Error	z	p-value
LNGHG(−1)	0.357760	0.0175168	20.42	<0.0001	***
LNGDPPC	0.0639105	0.00525269	12.17	<0.0001	***
LNJETFUEL	0.646603	0.0168543	38.36	<0.0001	***
LNCAPITAL	−0.0364608	0.00788863	−4.622	<0.0001	***
ETS	0.00741715	0.00311928	2.378	0.0174	**
Sum squared resid	26.53343		S.E. of regression	0.123915
Test for AR(1) errors: z = −2.14943 [0.0316] Test for AR(2) errors: z = 0.731628 [0.4644] Sargan over-identification test: Chi-square(481) = 1193.71 [0.0000] Wald (joint) test: Chi-square(5) = 5.05242 × 106 [0.0000]

Legend: *, **, *** indicate significance at the 0.10, 0.05, and 0.01 levels, respectively.

). Models 1a and 2a include LNBTS, whereas Models 1b and 2b include LNCAPITAL, respectively.

The first two columns of

Table 3. Estimation results for the determinants of GHG emissions from aviation.

Dependent Variable: LNGHG	(1a) LAD	(1b) LAD	(2a) GMM	(2b) GMM
LNGDPPC	0.01 * (0.006)	0.01 (0.008)	0.07 *** (0.004)	0.06 *** (0.005)
LNJETFUEL	0.40 *** (0.06)	0.40 *** (0.06)	0.64 *** (0.017)	0.65 *** (0.02)
LNBTS	−0.01 *** (0.005)		−0.02 *** (0.005)
LNCAPITAL		−0.02 ** (0.006)		−0.04 *** (0.008)
ETS	0.02 *** (0.007)	0.02 *** (0.006)	0.004 * (0.002)	0.007 ** (0.003)
N	864	864	864	864
Sum squared resid	27.8	27.8	26.2	26.5
Wald (joint) test: Chi-square (5)			4.30412 × 106 [0.0000]	5.05242 × 106 [0.0000]

Notes: Standard errors are in parenthesis. Legend: *, **, *** indicate significance at the 0.10, 0.05, and 0.01 levels, respectively.

present the results of Models 1a and 1b derived from least absolute deviation (LAD) estimation. In Model 1a, a 1% increase in GDP per capita is associated with a 0.01% increase in GHG emissions from international aviation in the short term, at the 10% significance level, on average, ceteris paribus. Similarly, a 1% increase in jet fuel consumption leads to a 0.4% increase in GHG emissions at the 1% significance level. On the contrary, a 1% increase in BTS decreases emissions by 0.01%. To our surprise, being covered by the EU ETS increases emissions by 0.02% at the 1% significance level, compared to not being covered by the EU ETS. Our findings with regards to the effect of the EU ETS are in line with those of Heiaas [8], who found a 10% increase in emissions associated with being covered by the EU ETS.

In Model 1b, the coefficient of LNGDPPC turns out to be insignificant for GHG emissions. A 1% increase in jet fuel consumption leads to a 0.4% increase in GHG emissions, as in Model 1a, at the 1% significance level. On the contrary, a 1% increase in CAPITAL decreases emissions by 0.02%. Also, as in Model 1a, being covered by the EU ETS increases emissions by 0.02% at the 1% significance level, compared to not being covered by the EU ETS.

The above-mentioned results are robust when a dynamic panel data model specification is adopted. The third and fourth columns of Table 3 show the results of Models 2a and 2b, respectively. In Model 2a, a 1% increase in GDP per capita is associated with a 0.07% increase in GHG emissions from international aviation in the short term at the 1% significance level, on average, ceteris paribus. Similarly, a 1% increase in jet fuel consumption leads to a 0.6% increase in GHG emissions. On the contrary, a 1% increase in BTS decreases emissions by 0.02%. Again, to our surprise, being covered by the EU ETS increases emissions by 0.004% at the 10% significance level, compared to not being covered by the EU ETS.

The results of Model 2b are quite similar to those of Model 2a. A 1% increase in GDP per capita is associated with a 0.06% increase in GHG emissions from international aviation in the short term at the 1% significance level, on average, ceteris paribus. Similarly, a 1% increase in jet fuel consumption leads to a 0.65% increase in GHG emissions. Being covered by the EU ETS increases emissions by 0.007% at the 5% significance level, compared to being not covered by the EU ETS. On the contrary, a 1% increase in CAPITAL decreases emissions by 0.04%.

With the purpose of robustness checks, the same sets of regressions were run by adding dummies for the COVID-19 pandemic years. While the coefficients of these dummies were negative, showing the pandemic’s negative impact on aviation emissions, the direction of the effect of the other independent variables did not change. The same independent variables were used in an alternative set of regressions, this time specifying the growth of GHG emissions from aviation, G_GHG, as the dependent variable. The results of these regressions are also robust in the face of the change in the dependent variable, confirming the positive effects of GDP per capita and ETS coverage on emissions, and the mitigating effects of business tourism spending and capital investments on emissions (see Appendix A,

Table A6. LAD estimation results with G_GHG as the dependent variable.

Model 1a: LAD, using 864 observations Dependent variable: G_GHG
	Coefficient	Std. Error	t-ratio	p-value
const	1.04805	0.0143425	73.07	<0.0001	***
LNGHG_1	−0.0568391	0.0126312	−4.500	<0.0001	***
LNGDPPC	0.00214693	0.000922042	2.328	0.0201	**
LNJETFUEL	0.0557476	0.0125159	4.454	<0.0001	***
LNBTS	−0.00174479	0.000813562	−2.145	0.0323	**
ETS	0.00302906	0.00110496	2.741	0.0062	***
Median depend. var	1.005311		S.D. dependent var	0.036570
Sum absolute resid	13.82339		Sum squared resid	0.688157
Log-likelihood	2109.943		Akaike criterion	−4207.885
Schwarz criterion	−4179.316		Hannan–Quinn	−4196.950
Model 1b: LAD, using 864 observations Dependent variable: G_GHG
	Coefficient	Std. Error	t-ratio	p-value
const	1.04825	0.0131745	79.57	<0.0001	***
LNGHG_1	−0.0533098	0.0113700	−4.689	<0.0001	***
LNGDPPC	0.00143385	0.00112620	1.273	0.2033
LNJETFUEL	0.0527665	0.0111958	4.713	<0.0001	***
LNCAPITAL	−0.00228012	0.00107866	−2.114	0.0348	**
ETS	0.00287898	0.00106488	2.704	0.0070	***
Median depend. var	1.005311		S.D. dependent var	0.036570
Sum absolute resid	13.82175		Sum squared resid	0.698671
Log-likelihood	2110.045		Akaike criterion	−4208.090
Schwarz criterion	−4179.521		Hannan–Quinn	−4197.155

Legend: **, *** indicate significance at the 0.05, and 0.01 levels, respectively.

and

Table A7. Dynamic panel estimation results with G_GHG as the dependent variable.

Model 2a: 2-step dynamic panel, using 864 observations Included 32 cross-sectional units Including equations in levels Dependent variable: G_GHG Asymptotic standard errors
	Coefficient	Std. Error	z	p-value
G_GHG(−1)	−0.0117546	0.00668640	−1.758	0.0788	*
const	1.06774	0.0182784	58.42	<0.0001	***
LNGHG_1	−0.0894519	0.00233973	−38.23	<0.0001	***
LNGDPPC	0.00438225	0.00193823	2.261	0.0238	**
LNJETFUEL	0.0885422	0.00202637	43.70	<0.0001	***
LNBTS	−0.00220859	0.000458270	−4.819	<0.0001	***
ETS	0.000996171	0.000296925	3.355	0.0008	***
Sum squared resid	0.756913		S.E. of regression	0.020929
Test for AR(1) errors: z = −2.25645 [0.0240] Test for AR(2) errors: z = 0.545225 [0.5856] Sargan over-identification test: Chi-square(455) = 1059.28 [0.0000] Hansen over-identification test: Chi-square(455) = 27.8823 [1.0000] Wald (joint) test: Chi-square(6) = 6965.83 [0.0000]
Model 2b: 2-step dynamic panel, using 864 observations Included 32 cross-sectional units Including equations in levels Dependent variable: G_GHG Asymptotic standard errors
	Coefficient	Std. Error	z	p-value
G_GHG(−1)	−0.0281560	0.00725272	−3.882	0.0001	***
const	1.09662	0.0136859	80.13	<0.0001	***
LNGHG_1	−0.0883785	0.00196989	−44.86	<0.0001	***
LNGDPPC	0.00153321	0.00124354	1.233	0.2176
LNJETFUEL	0.0901358	0.00192287	46.88	<0.0001	***
LNCAPITAL	−0.00492498	0.000782591	−6.293	<0.0001	***
ETS	0.00163090	0.000304338	5.359	<0.0001	***
Sum squared resid	0.747826		S.E. of regression	0.020803
Test for AR(1) errors: z = −2.2123 [0.0269] Test for AR(2) errors: z = 0.51133 [0.6091] Sargan over-identification test: Chi-square(455) = 1057.04 [0.0000] Hansen over-identification test: Chi-square(455) = 26.3702 [1.0000] Wald (joint) test: Chi-square(6) = 5634 [0.0000]

Legend: *, **, *** indicate significance at the 0.10, 0.05, and 0.01 levels, respectively.

).

4. Discussion

In line with the observations noted in the existing literature, jet fuel use stands out as the most prominent indicator that drives the increase in GHG emissions and emissions growth in international aviation in Europe. Although the International Air Transport Association [21] reports that global sustainable aviation fuel (SAF) production exceeded 3 × 10⁸ L in 2022, implying a 200% increase with respect to 2021 figures, GHG emissions due to jet fuels are still likely to continue their upward trend. Despite being a potential competitor for SAFs and a non-carbon fuel, the potential use of hydrogen in the sector also runs into issues such as “design challenges of aircraft systems, fuel storage, higher costs associated with fuel production, and new systems development” [22]. Furthermore, hydrogen combustion generates nitrogen oxides.

Unexpectedly, a negative relationship between business tourism spending and emissions from aviation is detected. As more and more organizations and events, such as conferences, congress, fairs, meetings, etc., set more ambitious emissions reduction targets for themselves, they keep revising their travel habits. This opens up opportunities for businesses and travel companies to cooperate for decarbonization purposes [23]. As expected, there is a negative relationship between capital investments in travel and tourism and GHG emissions. This is most probably due to the airline companies’ investments in recent years focusing more and more on sustainability and green technologies. Thanks to these investments, airline companies that start using SAF, or replace their old aircraft with new ones, may play a major role in reducing GHG emissions. Furthermore, there is a positive relationship between GDP per capita and GHG emissions, suggesting that as income levels increase, so do the use of air travel and the GHG emissions it causes.

Our findings imply that relying on the coverage by the EU ETS to mitigate GHG emissions from international aviation is not a panacea for the countries in Europe due to the fact that free allowances are provided to the sector, which does not urge radical emissions cuts. In the literature, possible impacts of various levels of carbon pricing on aviation emissions have been analyzed. For instance, using data from 2019, Scheelhaase et al. [24] forecast the potential impacts of CO₂ pricing on airfares and growth in aviation at a 180 €/t CO₂ price level. Running two scenarios with different levels of CO₂ cost pass-through to passengers, they estimate the reduction in demand for airfare and, hence, the reduction in emissions. The authors find that long-haul operations cause relatively larger amounts of GHG emissions than short- and medium-haul flights, which implies that flight distance is a crucial factor to be considered when introducing regulatory measures. During the implementation of the Carbon Offset and Reduction Scheme for International Aviation (CORSIA) under the International Civil Aviation Organization, the European Commission maintains the route-based approach, “ensuring that airlines operating flights on the same routes are treated equally, regardless of nationality”, in order to avoid market distortions (Source: https://climate.ec.europa.eu/eu-action/european-green-deal/delivering-european-green-deal/aviation-and-eu-ets_en (accessed on 5 September 2023))).

The European Court of Auditors [25] reports that between 2013 and 2019, aviation received over 2 × 10⁸ free allowances in the EU. Free emissions allowances under the EU ETS have been criticized as it is observed that emissions from the sectors receiving them either have remained stagnant or have declined very slowly [25]. The EU ETS continues to try to eliminate the risk of carbon leakage by allocating some emissions allowances free of charge to facilities producing goods that are at risk of carbon leakage, that is, producers that are likely to flee the European Union as their costs increase under the ETS. When not pressed to pay for most of its emissions, the businesses do not feel the motivation to invest in clean technologies for emissions reduction. According to a study by E3G [26], the only sector that has significantly reduced its emissions under the ETS is the electricity sector, which is not entitled to free allowances. Compared to the level in 2013, when the third phase of the ETS started, the electricity sector achieved a 27.7% emissions reduction in 2019, while this rate was limited to just a 2.1% reduction in other ETS sectors. One reason for this is the existence of free allowances that have accumulated over the process under the ETS. It is seen that there has been a surplus in the system as a result of the free allowances given to the plants above their verified emissions since 2005, when the EU ETS was first introduced. If the facilities do not use the rights allocated to them that year, they can keep them for later years or sell them in the free market and earn income. This is reflected in EU ETS carbon permit prices. In 2022, the European Council and Parliament agreed to gradually remove free allowances between the years 2026 and 2034 (https://ec.europa.eu/commission/presscorner/detail/en/ip_22_7609 (accessed on 5 September 2023)). The allocation of free allowances to aviation was scaled down anyway from 2012 to 2023. Yet, to have more free allowances in the future, aircraft operators and airlines tend not to renew their flying units [24]. Moreover, European airlines are favored by fuel tax exemptions and VAT exemptions. When compared to other sectors, the aviation industry has the advantage of compensating their EU ETS expenditure with the money they save from tax exemptions [27].

According to the ICAO, airlines worldwide carry more than 4 × 10⁹ passengers annually. Aviation represents 3.5% of global GDP and creates around 65 × 10⁶ jobs worldwide. Air transport plays a significant role as a supporter of, and by providing linkages to, several service industries. It also significantly contributes to economic growth, which has been evidenced by several scholars [28]. However, despite its significant contribution to the global economy, the aviation industry also has negative effects on climate change. For instance, GHG emissions from the aviation and maritime sector have increased by 107% since 1990. This means that although the aviation and shipping sector accounts for a small share of total emissions, it is one of the fastest emitting sectors in terms of emissions growth. Despite several improvements observed in the sector (such as technological advancements, SAF, hydrogen, etc.) to reduce these negative impacts, the pace of improvement is quite slow compared to the incline of demand for air transportation.

Considering the complicated and interrelated nature of the factors influencing aviation emissions, several policy implications could arise from these results in relation to the measures applied to curb emissions in the aviation sector. First, specific attention should be paid to the characteristics of the aviation sector if climate neutrality targets are to be taken seriously. Second, as the emissions growth rates are mostly attributable to the fossil fuel intensity in the sector, it appears necessary to reconsider a fast transition to sustainable fuels. Third, as capital investment makes a difference in the emissions trends, as pointed out by our findings, the countries in the region could try to transform their technological development towards cleaner options. Fourth, demotivating or disincentivizing long-distance flights that trigger the release of harmful emissions into the air might be a desirable policy option.

5. Conclusions

It appears that there is a clear need for strengthening climate policy applied to the aviation sector as the EU ETS has not proven to be successful in encouraging leading European airlines to reduce their GHG emissions. Due to the free allowances provided under the EU ETS, airlines in Europe are not subject to any serious restrictions. However, reducing or completely removing the free allowances provided to the aviation industry under the EU ETS will adversely affect competition between airlines in Europe and those not covered by the EU ETS. In order to prevent such unfair competition, an international GHG emissions reduction policy could be adopted. Although the implementation of the Carbon Offsetting and Reduction Scheme for International Aviation (CORSIA) is such an international attempt, it might not provide an effective solution either, since it does not price carbon, but instead introduces carbon offsetting measures to global civil aviation. Under CORSIA, airlines can purchase certificates for CO₂ emissions reductions in other sectors or carbon sequestration measures, rather than directly reducing their own emissions. Such offset programs do not fully guarantee that emissions reductions in the sector will be permanent [29]. The most effective steps to be taken to reduce GHG emissions in the aviation sector could be to create a cap-and-trade system, as in the EU ETS, to cover all airlines and effectively price all the GHG emissions from the sector, or equivalently, to tax all the GHG emissions from the sector. Only then would the industry feel the need and the responsibility to undertake the burden by paying for its carbon bill and facing the economic incentive to reduce its emissions thanks to a certain price signal. According to a study by Dixit et al. [30], which modeled the combined impact of carbon-tax policy, congestion cost, and greening investment decisions of airlines on greening efforts, carbon tax was shown to be an effective tool to drive airlines towards sustainable operations. The authors evidenced that airlines do not volunteer for investing in green technologies without carbon taxes in place. Their numerical solutions prove that a carbon tax implemented together with a congestion cost policy can be an effective panacea for emissions mitigation for airlines. Furthermore, alternative tools of climate policy, including the application of CO₂ standards in the sector, could be considered.

Regarding further analysis, the number of passengers and the number of flights could be integrated into the analysis as further independent variables. In the present study, they were not used due to data availability concerns and potential multicollinearity problems. The impacts of national carbon pricing mechanisms, sector-specific carbon mitigation tools, or other types of climate policies, where available, could also be investigated.