کشف قیمت و نوسانات سرریزها در طرح انتشار تجاری اتحادیه اروپا : تجزیه و تحلیل فرکانس بالا

This paper models the relationship of European Union Allowance spot- and futures-prices within the second commitment period of the European Union emissions trading scheme. Based on high-frequency data, we analyze the transmission of information in first and second conditional moments. To reveal long-run price discovery, we compute common factor weights of Schwarz and Szakmary (1994) and information shares of Hasbrouck (1995) based on estimated coefficients of a VECM. To analyze the short-run dynamics, we perform Granger-causality tests. We identify the futures market to be the leader of the long-run price discovery process, whereas the informational role of the futures market increases over time. In addition, we employ a version of the UECCC-GARCH model as introduced by Conrad and Karanasos (2010) to analyze the volatility transmission structure. The volatility analysis indicates a close relationship between the volatility dynamics of both markets, whereas in particular we observe spillovers from the futures to the spot market. As a whole the investigation reveals that the futures market incorporates information first and then transfers the information to the spot market.

Since the implementation of the European Union emissions trading scheme (EU-ETS) in January 2005, trading activity within the futures markets for European Union Allowances (EUAs) has steadily expanded over the first two commitment periods. However, as a consequence of the overallocation with allowances in Phase I, spot market trading activity broke down and prices converged to zero within this period. With the start of Phase II, spot market trading activity strongly rose and was even higher compared to the period prior to the spot market collapse. The main objective of this study is to analyze the price discovery process in the most liquid EUA spot and futures markets in Phase II of the EU-ETS. In addition, we investigate the joint volatility dynamics in both markets. Consequently, the paper directly attempts to assess the structure of information transmission in the EU-ETS. Contrary to previous studies such as Uhrig-Homburg and Wagner, 2009, Milunovich and Joyeux, 2010 and Chevallier, 2010, we make use of daily as well as intraday data at the frequencies of 10 and 30 min. We conduct the investigation of the price transmission between both markets on the basis of vector error correction models. Besides the analysis of common factor measures as suggested by Schwarz and Szakmary, 1994, Gonzalo and Granger, 1995 and Hasbrouck, 1995 to reveal the long-run price discovery process, we also investigate the short-run causality structure by means of Granger-causality tests. In order to assess the transmission of information in the second conditional moment, we estimate a dynamic version of the unrestricted extended CCC-GARCH model as developed by Conrad and Karanasos (2010), whereas each market’s conditional volatility is determined by lagged volatilities and lagged shocks of both markets. This model is flexible enough to capture negative volatility spillovers, leverage effects and dynamic conditional correlations. The first result is the absence of a cointegration relationship in daily spot and futures prices. Hence, at this frequency we cannot identify any market to be the price leading market. This result is in line with the findings of Milunovich and Joyeux, 2010 and Chevallier, 2010. However, extending the data from daily to intraday frequency, the analysis reveals a completely different picture. Based on high-frequency data, the results strongly support the existence of a cointegration relationship, and hence underpin the close link between both markets. Moreover, we show that drawing meaningful economic inference on each market’s contribution to the price discovery process requires to conduct the analysis on the basis of data at the highest frequency of 10 min. The reason for this is an increasing correlation between the innovations of the two markets at lower frequencies which induces an identification problem. Most importantly, we find that the futures market incorporates information first and then transfers it to the spot market. While at the early stage of Phase II the futures market attracts 70% of the price discovery process, this portion even increases over time. Consequently, our results considerably extend the findings of Uhrig-Homburg and Wagner, 2009 and Chevallier, 2010 as they show the close relationship between both markets and the futures market’s informational role. Second, concerning the short-run causality structure, we find univariate Granger-causality from the futures to the spot market in daily data. However, the investigation of high-frequency data reveals a bidirectional causality structure between both markets. This result is robust with respect to the choice of the intraday frequency. Third, in the volatility analysis we observe a similar pattern as in the price discovery analysis. In the early stage of Phase II we find unidirectional spillovers from the futures market volatility and from shocks in the futures market to the spot market’s volatility. There is no such impact into the opposite direction. Contrary, in the more mature stage only lagged spot market shocks but not lagged spot market volatility affect futures market volatility. In addition, the impact of lagged futures market volatility on current spot market volatility considerably increases over time. Consequently, the results of the volatility analysis confirm the existence of the close link between both markets, which we also find in the price discovery analysis. Further, these results contradict the findings of Milunovich and Joyeux (2010) who observe a weak link between both markets’ uncertainties in Phase I making use of daily data. Finally, the investigation of the DCC-structure indicates that the dynamic conditional correlation between spot and futures returns increases from about 0.1 at the start of Phase II to approximately 0.6 at the end of the sample period. In summary, we find strong evidence for a close relationship between the price and volatility dynamics in both markets that even intensifies over time. Further, making use of high-frequency data, we identify the futures market to be the price leading market. This result is consistent with previous findings for mature financial markets, as Tse (1999) among others reports. We organize the remainder of the paper as follows. In Section 2, we give an overview on the related literature while Section 3 summarizes the key elements of the EU-ETS. Section 4 describes the data and gives an overview on the relationship between commodity spot and futures prices in general. Section 5 outlines the methodology we use in the empirical analysis, while Section 6 summarizes the estimation results and provides an interpretation of the empirical findings. Finally, Section 7 concludes.

This paper addresses the question of information transmission in European spot and futures markets for emission allowances in the second commitment period of the EU-ETS making use of high-frequency data. Previous studies based on daily data such as Uhrig-Homburg and Wagner, 2009 and Milunovich and Joyeux, 2010 find mixed evidence for the relationship between spot and futures prices in the first commitment period. Moreover, based on low-frequency data and after controlling for structural breaks, Chevallier (2010) does not find a cointegration relationship between spot and futures prices in the early stage of the second commitment period. We show that this is due to the focus on the analysis of daily data. Based on high-frequency data, however, we find unambiguous evidence for the existence of a cointegration relationship between spot and futures prices. Moreover, we clearly identify the futures market as the price leading market. According to the price discovery measures of Schwarz and Szakmary, 1994, Gonzalo and Granger, 1995 and Hasbrouck, 1995, 70% of the price discovery process take place in the futures market. Further, the informational content of futures trading increases over time. The results of the conditional volatility analysis further confirm the findings of the price discovery analysis. Contrary to Milunovich and Joyeux (2010) who find evidence for a loose dynamic relationship between spot and futures volatility in the first commitment period, we find strong evidence for volatility spillovers from the futures to the spot market but not into the opposite direction. Moreover, the analysis reveals that both markets’ lagged shocks affect the volatilities in the other market. Finally, the results of the DCC-structure analysis indicate that the link between both markets considerably intensifies over time as the dynamic conditional correlation increases from 0.1 at the early stage of Phase II to about 0.65 at the end of the sample period. In conclusion, our results help to understand the mixed evidence previous studies report and highlight the informational role of the futures market. Moreover, they indicate that the price discovery process in the European allowance markets is similar to the one in more mature markets as Tse (1999) reports. This finding is remarkable due to the immature character of the EU-ETS.