© 2017 Nicolaus Copernicus University. All rights reserved. http://www.dem.umk.pl/dem D Y N A M I C E C O N O M E T R I C M O D E L S DOI: http://dx.doi.org/10.12775/DEM.2017.008 Vol. 17 (2017) 129−145 Submitted November 28, 2017 ISSN (online) 2450-7067 Accepted December 28, 2017 ISSN (print) 1234-3862 Aneta Włodarczyk * Regime-dependent Assessment of the European Union Aviation Allowances Price Risk  A b s t r a c t. In this article the European Union Aviation Allowances (EUAA) price risk, associated with the activity of aircraft operators within the European Economic Area (EEA), has been evaluated across the low and high volatility periods occurring on the carbon permits market. It is found that Markov-switching heteroscedasticity models distinguish well between two volatility regimes, as well as three volatility regimes on the EUAA futures market, and that the assessments of EUAA price risk are clearly different in the regimes. These findings may be explained by the European Union Emission Trading Scheme (EU ETS) design and the changes in both the EU climate policy rules and global regulations in the scope of CO2 emis- sions by international aviation. K e y w o r d s: European Union Aviation Allowances; EU Emission Trading Scheme; Mar- kov-switching model; risk. J E L Classification: C40; G32; L93; Q53. Introduction Worrying statistics concerning the over 3% share of CO2 emission from international aviation bunkers in the total carbon dioxide emission in the EU countries in 2004 and the forecasts concerning an increase in greenhouse gases emission in international aviation by about 70% in 2020 compared with 2005 led the international community to take interest in the issues of * Correspondence to: Aneta Włodarczyk, Czestochowa University of Technology, Faculty of Management, 19B Armii Krajowej Street, 42-200 Częstochowa, Poland, e-mail: aneta.w@interia.pl.  This work was financed by Czestochowa University of Technology. Aneta Włodarczyk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 130 limiting the negative impact of air transport on the natural environment. In 2008 the European Commission adopted the directive on including civil aviation into the EU Emission Trading Scheme (EU ETS), which refers to the EU’s long-term policy aimed at limiting the greenhouse effect, im- provement in the environment quality, increased energy efficiency and growth of renewable energy sources share in the energy consumption struc- ture (Directive 2008/101/WE). Aircraft operators that in the given year carry out aviation operations covered by the Attachment 1 to the Directive 2003/87/WE in the territory of European Economic Area have been included into the EU ETS. The period of civil aviation participation in the EU ETS has been divided into the two settlement periods: the first one of them comprising only the year 2012, the aim of which was to adjust the aircraft operators functioning to the functioning scheme of the remaining sectors in the allowances trading system, and the second one (common for all the sectors covered by the EU ETS) for the years 2013-2020. The EU ETS system operates according to the “cap and trade” principle. A permissible limit of CO2 emissions in the given period (“cap”) has been determined for all aircraft operators included into the system, which is gradually decreased over time. In order to asses this limit, data on average annual CO2 emission in aviation was used, which cov- ered the reference period of 2004-2006 and came from the European Organi- sation for the Safety of Air Navigation (Eurocontrol). In the first settlement period the total amount of European Union Aviation Allowances (EUAAs) to be allocated for aircraft operators was 97% of the average calculated from historical aviation emissions in the reference period. In the second settlement period the joint annual number of allowances granted to the aviation sector decreased to 95% of the same average emission. Within the established limit, in the first settlement period 85% of allowances allocated to cover the annual CO2 emission of civil aviation will be granted to aircraft operators free of charge, and 15% of aviation allowances will be sold in the auctioning sys- tem. In the period from 1 st January 2013 to 31 st December 2020 in turn, 82% of aviation allowances will be allocated free of charge, 15% will constitute allowances purchased at auctions, and 3% will be moved to a special reserve for new aircraft operators. Chin and Zhang (2013) presented mathematical formulas and described in detail the method of emission allowances alloca- tion consistent with the Directive 2008/101/EC and proposed an alternative method of allowances allocation (the Augmented EU ETS), which considers energy efficiency of aircraft operators (Chin and Zhang, 2013). The Euro- pean Commission Regulation No 601/2012 imposes on aircraft operators the following duties, which result from their participation in the EU ETS sys- Regime-dependent Assessment of the EUAA Price Risk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 131 tem: the necessity to monitor carbon dioxide emission in each year, submit- ting a verified report on historical carbon dioxide emission to the proper body and its settlement through redemption of a proper amount of emission allowances in the EU Redemption Registry. Aircraft operators may settle their own CO2 emissions by means of emission allowances assigned to sta- tionary installations (European Union Allowances - EUAs) or aviation al- lowances (EUAAs). Moreover, aircraft operators should possess a plan to monitor annual emissions, approved by the appointed in the given member country competent body. There are sanctions foreseen that can be applied to operators that fail to settle their CO2 emissions in the given year, including a prohibition on flights to the European Union. Aircraft operators also has a possibility to apply for allocation of free CO2 emission allowances for the years 2013–2020 from a special reserve, provided that they complied with the obligation to monitor tonne-kilometres throughout 2014. Such a system of CO2 emission allowances allocation was supposed to be an incentive to invest in environmentally-friendly technologies and modernise the air fleet against an alternative to incur additional costs of purchasing allowances. The emission allowances trading system was considered by the European Com- mission to be the most effective and least expensive instrument to limit the emission of greenhouse gases in aviation in the territory of the EU member states, yet, it encountered criticism. This concerned additional costs gener- ated by the EU ETS for the aircraft operators participating in it, which was connected with, among others, purchases of missing EUAAs, new environ- mentally-friendly investments, administrative costs. The growing fears of aircraft operators included into the EU ETS concerned the loss of their com- petitive position due to their lower market share, change of entry barriers or lower profits margin (Meleo et al., 2016). The most severe objection con- cerned imposing additional charges on aircraft operators from outside the EU. The charges resulted from participation in the EU ETS and were im- posed without any prior agreements, which was treated as a breach of the Chicago Convention on International Civil Aviation (1944). In response to this the European Commission introduced derogating mechanisms: „stop the clock” derogation (Decision 2013/337/EU), the exclusions mechanism in the scope of aviation operations covered by the EU ETS system (Regulation 2014/421). The abovementioned legislative regulations in the scope of obli- gations concerning reporting the emission from aviation operations within the EU ETS in the years 2013–2014, postponing the deadline to settle these emissions for 2013 and 2014, delayed the launch of aviation allowances Aneta Włodarczyk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 132 trading in the auctioning system. 1 In addition, they influenced the volatility of EUAAs prices and the volume of EUAAs derivatives trading, as well as the interest of aircraft operators in the use of this type instruments to manage the CO2 emission risk. The inclusion of international aviation into the EU Emission Trading Scheme causes that the management of the EUAA price risk is becoming increasingly important for companies covered by the EU ETS. The increase in the price allowances volatility in the secondary market and the reduction of free allowances, which are granted to companies, cause the increase in their exposure to price risk. The aim of this article is to evaluate the EUAA price risk connected with the activity of aircraft operators within the Europe- an Economic Area, distinguishing between low and high volatility periods occurring on the carbon permits market. Volatility and downside risk measures are used to assess this type of risk across different regimes of the EUAAs' future prices, which are identified by means of Markov-switching models. 1. Research Methodology The CO2 emission allowances price risk from the aircraft operators point of view may result in a danger of not achieving by them the expected returns due to a sale of excessive or purchase of insufficient aviation allowances on the secondary market (neutral risk concept). The aviation allowances price risk may be perceived as a danger of sustaining a loss, which can have sig- nificant impact on financial results of the aircraft operator (negative risk concept). Volatility measures are the tools which are most frequently used to measure risk according to the neutral risk concept, while risk measurement in the negative meaning are conducted according to the downside risk meas- ure (Jajuga, 2007). The most frequently determined by practitioners volatil- ity measures include: standard deviation, interquartile range and absolute median deviation. The downside risk measure, which exposes only unfa- vourable situations for the aircraft operator when the real returns on sale or purchase of aviation allowances was below the average, is semi-standard deviation. The first volatility measures, which have been estimated on the basis of settlement prices of EUAAs futures contracts, is standard deviation (Kuziak, 2011): 1 Aviation auctions have been conducted since 1st January 2015. Regime-dependent Assessment of the EUAA Price Risk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 133      T t t RR T 2 2 )( 1 1  , (1) where: %100*)/ln( 1 ttt PPR – return on sale/purchase of aviation allow- ance on the secondary market in t period; Pt – settlement price of EUAA in t period; Pt–1 – settlement price of EUAA in the period t–1(t = 2,.., T);  – standard deviation; tR – expected value of EUAA returns. Coefficient of variation is also used in order to assess how much risk accom- panies one unit of profit from the investment in the EUAA futures market. The lower the value of this measure is, the higher the expected profit on sale/purchase of aviation allowances is at the defined risk level or lower risk at the defined level of profit: %100 R Vs   (2) where sV is coefficient of variation. Asymmetric risk measure is semi-standard deviation, which can be defined in the following way (Booth et al., 2005):    T t td T SV 2 2 1 1 , (3) for       RRR R d t t t t Rfor Rfor 0 , (4) where SV – semi-standard deviation. This measure reflects an unfavourable for aircraft operators situation, where the returns on sale/purchase of EUAAs on the secondary market were below their expectations. Robust volatility estimators in turn are characterized by the fact that pre- sent in the sample outliers have little impact on the estimation results (Trzpiot, 2010). This is a very important property for EUAAs return series, where single outliers occur, which confirms extremely high change of avia- tion allowances price form period to period. The most frequently applied robust volatility estimator is absolute median deviation, which can be de- scribed by means of the following relation (Trzpiot, 2010): Aneta Włodarczyk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 134 }}{{)( i i t t t RmedianRmedianRMAD  , (5) where MAD(Rt) – absolute median deviation of EUAA returns. One of the simplest robust volatility estimators is interquartile range (IQR), which ignores up to 25% of lowest and largest returns (Trzpiot, 2010): )()( 13 tt RQRQIQR  , (6) where )(1 tRQ and )(3 tRQ – respectively the first and third quartile of EUAA returns empirical distribution. It is worth stressing that risk has been most often measured under the as- sumption of the worst- scenario in the market, what has been derived from the negative concept of risk and has contributed to the popularization of the downside risk measures. However, this approach may not reflect the change of uncertainty sets with respect to different market environments (e.g. calm and turbulent periods). Following Liu and Chen (2014), in this article Markov regime switching models are used to describe the time-varying un- certainty set of the first and second order moments, which are related to two main characteristics of investments in the EUAA futures, namely expected profits and risk. As a result, Markov-switching dynamic regression models with N- regimes (MS(N)-DR(p)) have been used to describe the different dynamism of the EUAAs returns series, generated by changing risk factors on the mar- ket of CO2 emission allowances (Hamilton, 1990; Doornik, 2013): )),(,0(~ ,)(...)()()( 2 t 22110 t tpttptttttt sN RsRsRssR     (7) where: i(st) – the parameter describing the influence of delayed by i-periods EUAA returns per the current returns (i=1,2,…,p), t – the error term, )( 2 ts - residual variance dependent on the valid at the given moment re- gime, st – non-observable variable modelled as homogeneous Markov chain of N regimes and the transition probabilities matrix   }1,...,1,0{,|   Njiji pP . The elements of stochastic matrix P, defining the process probability transi- tion from j regime at t–1 moment to i regime at the t moment, satisfy the Markov property (Hamilton, 1990): )|( ),...,,|( 1 00221| jsisP isisjsisPp tt ttttji     , (8) Regime-dependent Assessment of the EUAA Price Risk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 135 means that probability of moving to the next regime depends only on the current regime of a system and not on the previous regimes. Therefore, the transition probability matrix can be defined in the following way (Doornik, 2013):                              111 1 1 0 110 1|11|10|11 1|11|10|11 1|01|00|01       NNNNt Nt Nt ttt pppNs ppps ppps Nsss P , (9) at the conditions that guarantee the stochastic structure of this matrix:     1 0 j|i| .,for 0p ,1 N i ji N-10,1,...i,j p (10) Stawicki (2004) describing this class of econometric models introduced the concept of a dual stochastic process to emphasize the existence of an unob- servable variable that has controlled the changes of regimes in addition to the observable economic variable being the subject of modelling (Stawicki, 2004). Moreover, on the basis of estimated transition probabilities (elements of stochastic matrix P), the further expected duration of the system in a given regime can be determined (Hamitlon, 1990): )( 1 1 | N-1 0, 1,...,i p D ii i    , (11) where: Di – average duration of EUAAs returns in i-th regime. It is also worth stressing that the model (7) considers the heteroscedasticity characteristic for the EUAAs returns through introduction of Markov switch- ing also in the residual variance. The most frequently applied parameter estimation method of Markov switching model is the maximum likelihood method (ML), which makes use of FSQP algorithm (Feasible Sequential Quadratic Programming) (Law- rence and Tits, 2001; Psaradakis and Sola, 1998). A by-product of the Markov-switching model parameter estimation is a sequence of smoothed probabilities )|( Tt jsP  which make it possible to identify the moment of process switching between the particular volatility regimes. These probabil- Aneta Włodarczyk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 136 ity inferences allow to draw conclusions about the EUAA returns process being in a particular regime, although the regime variable st is unobserved. On this basis, the time series of the EUAAs returns have been divided into observations generated in different volatility regimes, and then risk measures (1)–(6) have been estimated for each sub-sample. Additionally, regime classification measure (RCM) has been used to de- termine the number of regimes in Markov switching models (Ang and Bekaert, 2002):       T t N i tt isP T NNRCM 1 1 0 1 2 ))|(( 1 100)( , (12) where: )|( 1 tt isP – filter regime probabilities series (i= 0,1,…,N–1 and t= 1,2,…,T), 1t – information set available up to time t–1. Alternatively, Ang and Bekaert (2002) suggested to replace the filter regime probabilities by the smoothed probabilities over the entire sample ( )|( Tt jsP  ) in (12). For the two-regime model the RCM statistics rang- es from 0 to 100 and smaller value of this measure means better regime clas- sification. High values of the RCM may indicate that the Markov switching model cannot successfully distinguish between regimes from the behavior of the data. It may point at misspecification of the number of regimes (Ang and Bekaert, 2002). 2. Data and Empirical Results Intercontinental Exchange Futures Europe in London (ICE Futures Europe), adjusting itself to the changing regulations in the scope of the EU climate policy, introduced into trade the EUAA futures contracts in February 2012. This product was dedicated primarily to aircraft operators as an in- strument of CO2 emission risk management, due to a significant increase in their exposition to this type of risk after civil aviation had been included into the EU ETS. Hedging operations with the use of aviation allowances futures should gain in popularity as the international law provisions are being tight- ened in the scope of carbon dioxide emission in civil aviation and the princi- ples of the EU ETS functioning towards introducing an auctioning system as a basic form of acquiring emission allowances by aircraft operators. An im- portant stage of CO2 emission risk management process in the civil aviation sector is the measuring price risk of EUAA futures contracts. In the empiri- cal research two types of risk measures have been used for this purpose: volatility risk measure and downside risk measure, which have been esti- Regime-dependent Assessment of the EUAA Price Risk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 137 mated on the basis of weekly returns of EUAAs futures prices listed on the ICE Futures in the period from 04.03.2012 to 01.10.2017. 2 The shaping of weekly settlement prices and returns for constructed benchmark series for EUAAs futures have been presented in Fig. 1. Figure 1. Weekly prices of the EUAA futures [EUR/tCO2] (upper panel) and their returns [%] (lower panel) quoted in the ICE Futures Europe in the period 4.03.2012 – 01.10.2017 Table 1. MS( N)-AR(4) models for the EUAAs returns Parameter MS(2)-DR(4) MS(3)-DR(4) Regime 0 Regime 1 Regime 0 Regime 1 Regime 2 Constant 0.696 [0.049] –1.104 [0.350] 0.726 [0.045] –20.529 [0.002] 0.358 [0.654] AR-4 –0.179 [0.019] –0.129 [0.234] –0.212 [0.013] –2.288 [0.001] –0.050 [0.546] sigma 4.113 (0.420) 11.356 (1.090) 3.820 (0.357) 11.089 (3.919) 8.393 (0.808) Matrix P R(0, t) R(1, t) R(0, t) R(1, t) R(2, t) R(0, t+1) 0.9349 0.1077 0.9314 0.0000 0.0739 R(1, t+1) 0.0651 0.8923 0.0000 0.2092 0.0470 R(2, t+1) – – 0.0686 0.7908 0.8791 Di (in weeks) 22.38 15.43 22.43 1.40 9.46 Returns assigned to regime 62.37% (179returns) 37.63% (108returns) 54.70% (157returns) 2.44% (7 returns) 42.86% (123returns) Note: P – transition probabilities matrix, standard errors of parameter estimates in parenthesis, p-value in brackets. Markov switching heteroscedastisity models (MS(N)-DR(p) for N = 2,3 and p = 1, 2, 3, 4) (7) were estimated to describe the dynamism of the EUAAs 2 The research uses the December prices of EUAA futures quoted on the ICE Futures Europe, presented on the website https://www.quandl.com (access 18.10.2017). Aneta Włodarczyk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 138 returns series in various volatility regimes. The best results have been pre- sented in Table 1. 3 Two regimes have been distinguished in the first approach: the low vola- tility regime (regime 0) that describes the period of calm on the EUAAs futures market and the high volatility regime (regime 1) characterized by turbulent changes on this market, which are generated mainly by changes to the regulations concerning CO2 emission in the aviation sector. The volatility parameter estimated for the regime 1 (11.356) is almost three times higher than the parameter describing the volatility in regime 0 (4.113). The differ- ences between the regimes can be observed also for the parameter describing the expected profit from the EUAAs futures transactions, which in the low volatility regime is positive and in the high volatility regime negative. More- over, each of the regimes is rather stable, as the estimated transition prob- abilities of indicating a chance of EUAAs returns process to remain in the given regime in the next period are relatively high (they amount respectively 0.9349 and 0.8923). Therefore, the low volatility regime on the EUAAs fu- tures market lasts for about 22 weeks, while the high volatility regime lasts on average for over15 weeks. In the second approach three regimes have been distinguished: the low volatility regime (regime 0), the high volatility regime (regime 2) and „spiky” regime (regime 1) for which the standard deviation of the error term took the highest value (11.089) compared to the low and high volatility re- gimes (respectively 3.82 and 8.393). The average profit from EUAAs futures transaction is highest in the low volatility regime. The spiky regime is a tran- sitional one, which means that there is a great chance (0.7908) that in the next period it will be replaced by the high volatility regime. Only single EUAAs returns have been assigned to this regime, which indicate extreme changes of EUAAs futures prices connected with the structural changes in the system of aviation allowances trade. Table 2 shows the results of diagnostic tests on standardised residuals from each model, which allow for positive verification of white noise prop- erties for residuals series. The Akaike information criterion demonstrates the model of three regimes to be better adjusted to the EUAAs returns series, 3 Basic descriptive statistics and diagnostic tests verifying the presence of structural break occurrence, autocorrelation, volatility clustering, leptokurtosis effects have been determined for both presented in Figure 1 variables. The results of conducted diagnostic tests justify the use of Markov switching heteroscedastisity models (7) to describe the dynamism of the EUAAs returns series in various volatility regimes. Their results are available upon request from author. Regime-dependent Assessment of the EUAA Price Risk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 139 while the Schwarz criterion reaches the lowest value for models of two re- gimes. Table 2. Diagnostic tests for scaled residuals from MS( N)-AR(4) models Test MS(2)-DR(4) MS(3)-DR(4) AIC SC 6.707 6.809 6.675 6.841 B-P(20) 21.120 [0.390] 18.492 [0.555] ARCH(1–5) 0.970 [0.437] 1.695 [0.136] J-B 5.614 [0.061] 4.009 [0.135] LR 80.132 [0.000] 99.299 [0.000] RCM 33.547 0.350 Note: B-P(k) – Box-Pierce test of serial correlation up to order k, ARCH(1-q) – Engle’s ARCH test for heteroscedasticity up to lag q, J-B – Jarque-Bera test for normality, LR – likelihood ratio test for non- linearity, RCM – regime classification measure, AIC – Akaike Information Criterion, SC – Schwarz Information Criterion, p-value in brackets. It is worth stressing that for both Markov-switching heteroscedasticity mod- els, the Davies (1987) upperbound for the p-value of the LR test of linearity strongly rejects the linear model (Doornik, 2013). The calculated values of the RCM statistics are rather low in the case of each specification of Markov switching model, what indicates the correct classification of regimes in the estimated models. Figure 2. Weekly EUAAs returns (upper panel), smoothed probabilities for regime 0 (middle panel) and smoothed probabilities for regime 1 (lower panel) in the period 4.03.2012 – 01.10.2017 Aneta Włodarczyk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 140 Figure 3. Weekly EUAAs returns and smoothed probabilities for regime 0 (upper panel), smoothed probabilities for regime 1 and regime 2 (lower panel) in the period 4.03.2012 – 01.10.2017 On the basis of obtained in the estimation procedure values of smoothed probabilities, the moments of process switching between particular volatility regimes were estimated. Then, the EUAAs returns series was divided into two sub-samples generated by the low and high volatility regimes (see Fig. 2) or three sub-samples generated by the regimes of low and high volatility and the spiky regime (see Fig. 3). One can observe that in each case the ma- jority of EUAAs returns have been assigned to the low volatility regime (respectively: 62.37% and 54.70%). However, the sub-sample connected with the high volatility regime is relatively numerous, which is a characteris- tic phenomenon on the CO2 emission allowance market (Sanin et al., 2015). In each sub-sample associated with the given volatility regime measures of profit and risk have been determined (1)–(6) (see Table 3). The regime comprising turbulent changes occurring on the market of CO2 emission al- lowances in aviation, the source of which is primarily uncertainty accompa- nying the changes of the EU climate policy and global regulations in the scope of pollutant emission by this sector, is characterized by much higher values of volatility and downside risk measures compared to regime 0. Fo- cusing on the neutral concept of risk, one can see that the ratio between stan- dard deviations estimated in regime 1 and 0 amounts to almost three. Having rejected the outliers in the process of risk measurement, the absolute median deviation in regime 1 is near three times higher than in regime 0. The same ratio is obtained when semi-standard deviations in high and low volatility regimes are compared, according to negative concept of risk. Aircraft opera- tors securing themselves against the EUAAs price risk by means of futures Regime-dependent Assessment of the EUAA Price Risk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 141 contracts in the high volatility periods could on average lose 0.995% per week, and in the low volatility periods they could on average earn 0.621%. Table 3. Risk measurement for the EUAAs futures transactions in particular regimes Statistic MS(2)-DR(4) MS(3)-DR(4) Regime 0 Regime 1 Regime 0 Regime 1 Regime 2 Minimum –13.246 –48.791 –8.049 –48.791 –15.214 Mean 0.621 –0.995 0.710 –22.166 0.421 Median 0.589 –1.017 0.692 –22.816 0.248 Maximum 10.828 23.660 8.701 20.516 23.660 Standard deviation 4.189 11.708 3.995 21.907 8.573 Absolute median deviation 2.701 7.835 2.556 5.776 6.322 IQR 5.291 15.572 4.905 13.684 12.262 Semi standard deviation 2.982 8.874 2.734 12.134 5.835 Note: IQR – Interquartile range, the coefficient of variation is not calculated because of negative values of expected return in some regimes. In the half of the weeks corresponding with the high volatility regime on the futures market, it was possible to lose at least 1.017%. In the half of the weeks assigned to the low volatility regime, EUAAs returns were not lower than 0.589%. Similar conclusions can be formulated for the 3-regime Markov switching model, when the estimated risk measures were compared for the high and low volatility regime. The risk accompanying the transac- tions concluded on the derivatives market was very high in regime 1 (21.907% according to the neutral risk concept, 12.134% according to the negative concept). In the spiky regime the volatility estimated with the means of the robust estimator which is absolute median deviation, amounts 5.776% and is significantly lower than other volatility measures. Consider- ing also the average value (–22.166%) and median (–22.816%) one can as- sume that the majority of observations assigned to regime 1 constitute ex- tremely large, negative EUAAs returns. The results presented in Table 3 indicate significant differences in the level of risk accompanying the transac- tions concluded on the EUAAs futures market, depending on the presence of the volatility regime in the given period. Therefore, Markov switching mod- els may constitute a useful tool that allows to identify the moment of switch- ing between regimes or determine the average lasting time of particular vola- tility regimes on the aviation allowances secondary market. Conclusions Decisions concerning tightening the EU climate policy in the scope of greenhouse gases emission from aviation and modifying the principles of the Aneta Włodarczyk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 142 EU ETS functioning will affect competitiveness of aircraft operators per- forming flights within the European Economic Area in the years 2012–2020. Therefore, it is important that the system of CO2 emission risk management is adjusted to individual needs of a given aircraft operator and targeted to implement eco-innovations, making it possible not only to comply with legal regulations concerning the size of CO2 emission but also to earn on the growth of fuel efficiency and transportation process optimization by the operator (Ko et al., 2017). Moreover, the constant control over the CO2 emissions in order to the maintenance it below the upper allowed threshold and sale at an attractive price the surplus of CO2 emission allowances leads to strengthening the competitive advantage of aircraft operators. By includ- ing civil aviation to the EU ETS system the legislators wanted to implement the principle “polluter pays”, yet, the economic practice has shown that simi- larly to the energy sector, aircraft operators started gradually transfer the cost of participation in the EU ETS on the customers. Thanks to including the cost of CO2 emission into the price of tickets, despite received free of charge share of aviation allowances, the operators recorded windfall profit. Making use of this option depends on the elasticity of the demand for aviation ser- vices, offers of competitive air carriers, state regulations of air transport market in the scope of access to both domestic transports as well as on inter- national markets (concessions, certificates) (Tłoczyński, 2015; Dyduch, 2013). Due to the fact that aircraft operators have to make decisions on the way they use aviation allowances, the measurement of EUAAs price risk on the secondary market is of vital importance in this decision-making process. If the marginal profit from the sale of aviation service is higher than the market value of EUAA allowances, the aviation allowances will be used to cover CO2 emissions derived from the air operations (Dyduch, 2013). The paper shows the changes in the EUAAs price risk level depending on the volatility regime in force on the aviation allowances secondary mar- ket. On the basis of estimated Markov switching models the Author has identified two or three volatility regimes, for each different risk measures have been determined. The high volatility regime, in which the risk accom- panying the sale and purchase transactions of EUAAs futures was several time higher than in the low volatility regime, has been assigned to the peri- ods when important regulations in the aviation sector were introduced. The periods assigned to the high volatility regime corresponded with such events as: the end of the first settlement period for the aviation sector in the EU ETS (2012-11-25–2013-01-13), conducting the first settlement of historical CO2 emissions in civil aviation for the year 2012 and publishing the decision of the European Parliament and Council on implementing the „stop the Regime-dependent Assessment of the EUAA Price Risk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 143 clock” mechanism (2013-04-21, 2013-04-28–2013-06-23), publishing the Regulation 421/2014 on further derogation in aviation (2014-04-06–2014- 05-25), works on concluding a global agreement regarding implementation of the Carbon Offset and Reduction Scheme for International Aviation, CORSIA) (2016-06-19–2016-12-25). Such an approach makes it possible to measure more precisely the risk that accompanies the sale/purchase transac- tions of aviation allowances on the secondary market. It also allows diversi- fication of the adopted by aircraft operators strategies of securing themselves against the risk of CO2 emission allowances price depending on the price volatility regime in force. References Ang, A., Bekaert, G. (2002), Regime Switches in Interest Rates, Journal of Business & Eco- nomic Statistics, 20(2), 163–182, DOI: http://dx.doi.org/10.1198/073500102317351930. Booth, P., Chadburn, R., Haberman, S., James, D., Khorasanee, Z., Plumb, R., Rickayzen, B. (2005), Modern Actuarial Theory and Practice, CRC Press, Boca Raton. Chin, A. T. H., Zhang, P. (2013), Carbon Emission Allocation Methods for Aviation Sector, Journal of Air Transport Management, 28, 70–76, DOI: http://dx.doi.org/10.1016/j.jairtraman.2012.12.013. Commission Regulation (EU) No 601/2012 of 21 June 2012 on the monitoring and reporting of greenhouse gas emissions pursuant to Directive 2003/87/EC of the European Par- liament and of the Council, http://eur-lex.europa.eu/eli/reg/2012/601/oj (12.09.2017). Decision No 377/2013/EU of the European Parliament and of the Council of 24 April 2013 derogating temporarily from Directive 2003/87/EC establishing a scheme for green- house gas emission allowance trading within the Community, http://eur- lex.europa.eu/eli/dec/2013/377(1)/oj (12.09.2017). Directive 2008/101/EC of the European Parliament and of the Council of 19 November 2008 amending Directive 2003/87/EC so as to include aviation activities in the scheme for greenhouse gas emission allowance trading within the Community, http://eur- lex.europa.eu/eli/dir/2008/101/oj (12.09.2017). Doornik, J. A. (2013), Econometric Analysis with Markov Switching Models. PCGive TM 14, vol. 4, Timberlake Consultants Ltd., London. Dyduch, J. (2013), Handel uprawnieniami do emisji zanieczyszczeń powietrza (Trading of Air Pollution Emission Allowances), PWE, Warszawa. Hamilton, J. D. (1990), Analysis of Time Series Subject to Changes in Regime, Journal of Econometrics, 45(1–2), 39–70, DOI: http://dx.doi.org/10.1016/0304-4076(90)90093-9. Jajuga, K. (2007), Teoretyczne podstawy zarządzania ryzykiem (Theoretical Foundations of Risk Management), in Jajuga K. (ed.), Zarządzanie ryzykiem (Risk Management), PWN, Warszawa. Ko, Y. D., Jang, Y. J., Kim, D. Y. (2017), Strategic Airline Operation Considering the Carbon Constrained Air Transport Industry, Journal of Air Transport Management, 62, 1–9, DOI: http://dx.doi.org/10.1016/j.jairtraman.2017.02.004. http://dx.doi.org/10.1198/073500102317351930 http://dx.doi.org/10.1016/j.jairtraman.2012.12.013 http://dx.doi.org/10.1016/0304-4076(90)90093-9 http://dx.doi.org/10.1016/j.jairtraman.2017.02.004 Aneta Włodarczyk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 144 Kuziak, K. (2011), Pomiar ryzyka przedsiębiorstwa: modele pomiaru i ich ryzyko (Measure- ment of Enterprise Risk – Measurement Models and their Risk), Prace Naukowe Uni- wersytetu Ekonomicznego we Wrocławiu, 160, Seria: Monografie i Opracowania, 187, Wrocław. Lawrence, C. T., Tits, A. L. (2001), A Computationally Efficient Feasible Sequential Quad- ratic Programming Algorithm, SIAM Journal of Optimization, 11(4), 1092–1118, DOI: http://dx.doi.org/10.1137/S1052623498344562. Liu, J., Chen, Z. (2014), Regime-dependent Robust Risk Measures with Application in Port- folio Selection, Procedia Computer Science, 31, 344 – 350, DOI: http://dx.doi.org/10.1016/j.procs.2014.05.277. Meleo, L., Nava, C. R., Pozzi, C. (2016), Aviation and the Costs of the European Emission Trading Scheme: The Case of Italy, Energy Policy, 88, 138–147, DOI: http://dx.doi.org/10.1016/j.enpol.2015.10.008. Psaradakis, Z., Sola, M. (1998), Finite-sample Properties of the Maximum Likelihood Estima- tor in Autoregressive Models with Markov Switching, Journal of Econometrics, 86(2), 369 – 386, DOI: http://dx.doi.org/10.1016/S0304-4076(98)00010-4. Regulation (EU) No 421/2014 of the European Parliament and of the Council of 16 April 2014 amending Directive 2003/87/EC establishing a scheme for greenhouse gas emis- sion allowance trading within the Community, in view of the implementation by 2020 of an international agreement applying a single global market-based measure to inter- national aviation emissions, http://eur-lex.europa.eu/eli/reg/2014/421/oj (12.09.2017). Sanin, M. E., Mansanet-Bataller, M., Violante, F. (2015), Understanding Volatility Dynamics in the EU-ETS Market, Energy Policy, 82, 321–331, DOI: https://doi.org/10.1016/j.enpol.2015.02.024. Stawicki, J. (2004), Wykorzystanie łańcuchów Markowa w analizie rynku kapitałowego (The Use of Markov Chains in the Analysis of the Capital Market), Wydawnictwo UMK, Toruń. Tłoczyński, D. (2015), Rola państwa w kształtowaniu konkurencji na polskim rynku transportu lotniczego (The Role of State in Shaping the Competition in the Polish Air Transport Market), Research Papers of Wrocław University of Economics, 401, 525–534. T r z p i o t , G . (ed.) (2010), Wielowymiarowe metody statystyczne w analizie ryzyka inwesty- cyjnego (Multidimensional Statistical Methods in Investment Risk Analysis), PWE, Warszawa. http://dx.doi.org/10.1137/S1052623498344562 http://dx.doi.org/10.1016/j.procs.2014.05.277 http://dx.doi.org/10.1016/j.enpol.2015.10.008 http://dx.doi.org/10.1016/S0304-4076(98)00010-4 https://doi.org/10.1016/j.enpol.2015.02.024 Regime-dependent Assessment of the EUAA Price Risk DYNAMIC ECONOMETRIC MODELS 17 (2017) 129–145 145 Reżimowo-zależna ocena ryzyka zmian cen unijnych uprawnień lotniczych do emisji CO2 Z a r y s t r e ś c i. W artykule oszacowano ryzyko zmian cen unijnych uprawnień lotniczych do emisji CO2, które towarzyszy działalności operatorów statków powietrznych wykonywanej w ramach Europejskiego Obszaru Gospodarczego, w okresach niskiej i wysokiej zmienności występujących na tym rynku. Pokazano, że reżimy zmienności na rynku kontraktów futures na uprawnienia lotnicze zostały prawidłowo zidentyfikowane, zarówno dla dwustanowego, jak i trzystanowego przełącznikowego modelu Markowa, a oszacowane miary ryzyka różnią się w reżimach. Występowanie różnych reżimów zmienności na tym rynku można wyjaśnić modyfikowaniem zasad funkcjonowania europejskiego systemu handlu emisjami oraz wpro- wadzaniem zmian zarówno w polityce klimatycznej UE, jak i w globalnych regulacjach doty- czących emisji CO2 przez lotnictwo międzynarodowe. S ł o w a k l u c z o w e: europejski system handlu emisjami; przełącznikowe modele Mar- kowa; ryzyko; unijne uprawnienia lotnicze do emisji CO2.