جدول زمانی اصطکاک های معاملاتی در بازار کربن اروپا

During its trial phase (Phase I), the EU Greenhouse Gas Emission Trading Scheme (EU-ETS) collapsed because of an over-allocation of emission allowances. We evaluate the progress of this market from the trial phase to the next commitment period (Phase II) from a microstructure angle. We show that trading frictions, as measured by the relative spread, information-asymmetry risk, and market making profits decreased from Phase I to Phase II. Although volatility decreased, its noise-related component gained in importance at the expense of its information-related component, resulting in lower quality of the price changes.

The Kyoto Protocol entered into force on February 16th, 2005. By this agreement, 141 countries committed to reducing their global greenhouse gas emissions by at least 5% (8% for the European Union (EU)) against 1990 levels over the period 2008–2012. Seeking to control the CO2 emissions of energy intensive installations across Member States, the EU launched the European Union Emission Trading Scheme (EU-ETS) on January 1st, 2005.1 Nowadays, the EU-ETS is the world’s largest scheme of its kind in terms of volume traded (e.g., Newell, Pizer and Raimi, 2013). It covers around 11,000 companies in the energy and industrial sectors that account for 45% of Europe’s total emissions.2 As a network of linked cap-and-trade systems, total emissions under the EU-ETS are limited or “capped”. Polluting installations receive emission permits called European Union Allowances (EUAs) to cover their CO2 emissions for the upcoming year. Each EUA entitles the holder to emit one tonne of CO2 or greenhouse gas equivalent into the atmosphere. Installations must report their emissions every year to the accredited verifier, and deliver enough EUAs to cover their yearly emissions. The EU-ETS is structured as a sequence of compliance periods of Phases. Phase I (2005–2007) was the ‘pilot’ or ‘learning’ period (e.g., Ellerman and Buchner, 2008; Creti, Jouvet and Mignon, 2012). This period was marked by the market breakdown that started in April 2006 with a sudden 66% Phase I EUA price drop (from 30 EUR to 10 EUR) over a few days, and ended with prices being virtually zero by September 2007. It is generally agreed that the reason behind the collapse was a too lax cap on Phase I emissions together with the prohibition on transferring (or banking) EUAs from Phase I to Phase II (e.g., Alberola, Chevalier and Chèze, 2008; Capoor and Ambrosi, 2006; Ellerman and Buchner, 2008). The European Commission (EC) either over-estimated the level of CO2 emissions and the demand for EUAs or under-estimated the magnitude of the abatement that resulted from polluting facilities incorporating CO2 prices into their production decisions. Phase II (2008–2012) coincided with the Kyoto Protocol accomplishment period. This period was characterized by the outbreak of the international financial crisis (e.g., Chevallier, 2009 and 2011) and the deterioration of the real production expectations (e.g., Christiansen et al., 2005). Moreover, the market experienced problems with overlapping regulations (e.g., Böhringer, Koschel and Moslener, 2008) and scandals like the robbery of permits from members’ accounts, VAT fraud, the sale of recycled permits and cyber- and phishing attacks (e.g., Kossoy and Ambrosi, 2010; Linacre, Kossoy and Ambrosi, 2011).3 A few studies have compared Phases I and II of the EU-ETS from different angles. Creti et al. (2012) conclude that, in terms of the role played by fundamentals of the EUAs’ price, Phase I and II were different regimes.4 Montagnoli and de Vries (2010) show that prices were more efficient at the beginning of Phase II than during Phase I using data from BlueNext, a spot market with limited trading activity that no longer exists. Kalaitzoglou and Ibrahim (2013) model the duration (time between trades) and size of trades during Phase I and II up to 2008. They find that Phase II discretionary liquidity traders reacted faster to the information embedded in the order flow. As a response, informed traders increased the size and decreased the duration of their trades. Building on former studies, one could easily conclude that the EU-ETS progressed from Phase I to Phase II, the latter being an enhanced and more mature version of the former (e.g., Daskalakis, 2012; Ibikunle et al., 2013). However, issues of great relevance for potential traders and regulators such as trading costs, information-asymmetry risk, market-making profits, volatility drivers and quality of prices have been ignored in prior comparative analyses. In this paper, we fill this gap by testing if trading conditions and price formation in the EU-ETS during Phase II improved with respect to Phase I. Our methodological approach builds on a vast body of work on the market microstructure field that studies trading frictions and price formation in all sorts of financial markets.5 This literature has shown that, at high frequencies, observable prices may temporarily deviate from the true or efficient value because of trading frictions. Frictions arise because of imperfections and limitations in the market regulation and the organization of trading that interfere with price formation. Trading frictions matter because they introduce noise into the price discovery process, making prices less informative (e.g., Chordia, Roll and Subrahmanyam, 2008). They also increase the costs of trading (e.g., Stoll, 2000) and play an important role in asset pricing (e.g., Amihud and Mendelson, 1986, 1989). There are different views about the main sources of trading frictions. According to the real friction view (e.g., Garman, 1976; Stoll, 1978; Ho and Stoll, 1981, 1983), trading frictions arise because suppliers of immediacy demand compensation for the costs incurred in providing services. Such costs include economic resources (labor and capital) needed to process orders and execute and settle trades, but also assumed undesired inventory positions. Market power is also a source of real frictions (e.g., Laux, 1995). In this latter case, real frictions take the form of rents extracted by monopolistic dealers. Other sources of real frictions are price discreteness, temporary order imbalances, liquidity shortfalls, and price smoothing rules (e.g., Hasbrouck, 1996). Under the informational friction view, frictions adopt the form of compensation for the losses incurred when suppliers of immediacy face timelier or better-informed traders. On the one hand, posted quotes are free options offered by the suppliers of immediacy. When new information arrives to the market, speedy traders can pick off these quotes before being properly adjusted (e.g., Copeland and Galai, 1983; Handa and Schwartz, 1996). On the other hand, posted quotes may be accepted by traders endowed with superior (adverse) information (e.g., Bagehot, 1971; Kyle, 1985; Glosten and Milgrom, 1985; Easley and O’Hara, 1987; Glosten, 1994). As it naturally embodies the concession required for immediacy, the bid-ask spread is the most commonly used measure of friction (see Stoll, 2000). Empirical research has shown that both informational and real frictions matter in explaining the size of the spread (e.g., Glosten and Harris, 1988; Lin, Sanger and Booth, 1995; Huang and Stoll, 1996, 1997). Theoretical and structural models of price formation account for trading frictions by decomposing the observable price into an unobservable efficient or information-related component and a noisy or real-friction-related component.6 Shocks to the efficient component have a permanent impact on prices, while shocks to the noisy component are transitory in nature (e.g., Hasbrouck, 1991a).7 This information-noise decomposition extends to the volatility of the price change. Thus, transitory volatility is triggered by noise, whilst permanent volatility relates to the underlying efficient price (e.g., Hasbrouck, 1999; Pascual and Veredas, 2010). While transitory volatility is detrimental to financial markets, informational volatility is desirable since it reflects new information that is being incorporated into prices. The proportion of the volatility of the price change that is driven by the efficient component becomes a natural measure of the informativeness or “quality” of prices or “market quality” (e.g., Madhavan, Richardson and Roomans, 1997; Hasbrouck, 1993). To perform our microstructure-based comparative analysis of Phase I versus Phase II of the EU-ETS, we use detailed trade-by-trade data (including price, volume, and direction) from the European Climate Exchange (ECX), the most liquid trading platform for EUA futures contracts (e.g., Mansanet-Bataller and Pardo, 2008; Fan, Akimov and Roca, 2012).8 Our sample covers from April 2005 to December 2010. We apply three alternative microstructure approaches to measure trading frictions, decompose volatility into its friction-related and friction-unrelated components, and evaluate price quality: a stylized structural model derived from Roll (1984), a more sophisticated model proposed by Madhavan et al. (1997), and an econometric reduced-form approach due to Hasbrouck (1993). We provide robust evidence of improved trading conditions in Phase II with respect to Phase I. The median bid-ask spread fell from 0.1004 EUR in Phase I (excluding the 2007 market collapse) to 0.0186 EUR in Phase II; the median adverse selection costs decreased from 94.8% of the bid-ask spread (0.0825 EUR) to 76.36% (0.0137 EUR), and the median market making profits shrunk from 0.0129 EUR (7.86% of the median trade price) per lot round-trip to 0.0025 EUR (1.83% of the median trade price). Not surprisingly, the ECX reached its lowest level in price quality during the market collapse, confirming the large role that noise played in price changes during that period. Price quality progressively recovered during Phase II, but remained statistically below the levels achieved before the 2007 market collapse. The difference in the contribution of noise to volatility between 2005–2006 and 2009–2010 increased between 10.23% and 26.64% depending on the model. Overall, we provide robust evidence of a persistent decrease in the quality of the EU-ETS price changes after the trial phase. Our study therefore qualifies those studies claiming that Phase II of EU-ETS was a more mature market than Phase I in every dimension. Ours is not the first microstructure-oriented intra-day analysis that deals with liquidity and trading frictions in the EU-ETS. Benz and Hengelbrock (2008) find lower bid-ask spreads for ECX during Phase I than for Nord Pool (nowadays, NASDAQ OMX). Mizrach and Otsubo (2014) report a 57% decrease (from 0.0367 EUR to 0.0158 EUR) in the monthly spread of the December 2009 EUA futures contracts from January to December 2009, and narrower bid-ask spreads for EUAs than for other emission permits, the Certified Emission Reductions or CERs. Furthermore, they estimate that information asymmetry risk accounts for 75% of the spread during that period. Both studies use the same methodological approach (Madhavan et al., 1997), with which we provide consistent estimates. Our analysis of trading frictions, however, improves over the former studies in several ways. Firstly, we offer a more comprehensive picture of the evolution of liquidity and trading frictions in the EU-ETS by covering Phases I and II. Secondly, we provide estimates of market-making profits in the EU-ETS for the first time in the literature. Thirdly, previous research has studied the conditional heteroskedasticity of both daily (e.g., Benz and Trück, 2009 and Paolella and Taschini, 2008) and intra-daily EUA returns (e.g., Conrad et al. 2012), and regular intraday patterns (e.g., Rotfuß, 2009) and outliers in the intraday volatility of EUAs (e.g., Chevallier, 2011), but no other study has performed a microstructure-based intraday decomposition of the volatility of price changes into their noise-driven and information-driven components. In this paper, we study the nature of EUA volatility. Fourthly, we give strength and robustness to our conclusions by not relying on a single methodological approach. Finally, our findings question the generally assumed superiority of Phase II with respect to Phase I, as we provide evidence of a long-lasting impact of the 2007 market collapse on the quality of the EU-ETS. The remainder of the paper proceeds as follows. In Section 2, we offer the basic market background. In Section 3, we describe our database. In Section 4, we discuss the methodological details. In Section 5, we present our empirical findings. Finally, we conclude in Section 6.

We provide a comparative analysis of Phase I (trial phase) and Phase II (Kyoto’s Protocol compliance phase) of the world’s largest market for greenhouse gas emission allowances, the EU-ETS. We complement prior studies that focused on important issues such as efficiency of prices, price leadership, learning speed, or changes in price fundamentals by adopting a market microstructure perspective. Our main goal is to test if Phase II is an improved version of Phase I, as it is generally assumed in the literature, by looking at issues of high relevance for potential traders and regulators such as trading costs, information-asymmetry risk, market-making profits, volatility sources, and quality of prices. We pay particular attention to the market-wide breakdown that took place by the end of Phase I and its long-run implications in trading costs and market quality during Phase II. This market collapse was shortly followed by the outbreak of the world financial crisis. Using trade data from ECX, the most important trading platform within the EU Scheme, we provide evidence that at the end of 2010, the EU-ETS had not fully recovered from these two major events. Our findings contrast with the usual view of Phase II as an improved version of Phase I. Trading frictions, as measured by both the bid-ask spread and the relative spread, drastically diminished from Phase I to Phase II. The quoted spread decreased from 9 ticks at the beginning of Phase I to 1 tick by the end of 2010. Similarly, the relative spread fell from 0.67% to 0.07% during the same interval. Therefore, liquidity improved after the market collapse. We also report a significant decrease in the information asymmetry risk. During the pre-2007 Phase I, the informational friction component of the spread was 0.0825 EUR in median terms, representing about 95% of the spread and 46.9% of the average trade price. During Phase II, it fell to 0.0137 EUR, 76.36% of the spread and 9.05% of the average trade price. Our findings indicate that, historically, the main determinant of the size of the spread in the EU-ETS has been, to a large extent, the perceived risk of trading against informed traders, whilst other market making costs, such as inventory holding costs or order processing costs, played a secondary role. We use the realized spread to proxy for the expected market making revenues per lot round-trip. We find that the realized spread decreased from Phase I (0.250% EUR before 2007) to Phase II (0.154% EUR), and gradually during Phase II. Our results suggest that market makers actually translated their lower market making costs during Phase II to their quoted bid-ask spreads. We decompose the return volatility into its theoretical components: (i) the frictionunrelated volatility, due to public news, (ii) the informational-friction volatility, due to informed trades, and (iii) the real-friction volatility, due to noise. We find that friction-unrelated news was the most important contributor to the EU-ETS volatility. Yet, we provide evidence of an increase in the contribution of real-friction volatility from Phase I to Phase II. Market quality summary measures reveal that this increase in transitory volatility generated a persistent negative effect on the quality of the EUETS. All models considered coincide in showing that market quality progressively recovered during Phase II after the 1st quarter of 2009. However, quality levels observed by the end of 2010 were still below those found during the pre-collapse Phase I. Therefore, we conclude that the 2007 market breakdown had a long-lasting effect on the quality of the European carbon market, probably fueled by the international financial crisis.