The Price-Determinants of Carbon Futures and the Exchange-Specific Effects of Power Markets.

The Price-Determinants of Carbon Futures and

the Exchange-Specific Effects of Power Markets.

A view of Phase II of the EU ETS

Arieke Boersen

S1209043

Supervisor: Prof. Dr. L.J.R. Scholtens

June 11, 2010

University of Groningen

Faculty of Economics and Business

The Price-Determinants of Carbon Futures and

the

Exchange

-Specific Effects of Power Markets.

A view of Phase II of the EU ETS

Arieke Boersen

1

Abstract: This paper aims at characterizing the price-determinants of the EU Emission Allowances futures (EUA) of Phase II, traded since 2005. The results from a Threshold GARCH(1,1) model confirm the view that natural gas, oil prices and the switching possibilities between gas and coal for combustion generation are carbon price-determinants. Secondly this paper aims at analyzing the different impacts on carbon futures prices of the main European power markets. The Nordpool and UK power spot prices are the main significant carbon price determinants among the power spot exchanges, where Nordpool has the largest impact. In analyzing one-year power futures the German EEX has the largest impact on EUA futures prices and is the most important reference power market for the carbon price determinants.

Keywords: carbon emissions rights, EU ETS, power markets, GARCH JEL classification: L94, Q48, Q52

I. INTRODUCTION

On January 1, 2005 the European power generation sector was confronted with the carbon emission allowances system. To ensure a cleaner environment on the long run, the combustion sector has to deal with a new investment incentive, the right to emit carbon. The European Union implemented a scheme of tradable allowances, as part of its commitment to the Kyoto Protocol. To enforce carbon reduction, the EU has chosen a "cap-and-trade" mechanism. Member States receive a cap on their CO2 emissions and allocate the allowances across companies in the CO2-intensive emitting industrial sectors, with the power and heat sector as largest industry. This cap coincides with the Kyoto reduction target, and so creates a market for the carbon allowances, the European Union Emission Trading Scheme (EU ETS). The EU ETS’ design consists of two trading phases: Phase I covers a three year pilot period from 2005 to 2007 and Phase II the period from 2008 to 2012. In Phase II the actual objectives of the Kyoto Protocol are implemented by more stringent caps. One European Allowances Unit (EUA), is the right to emit a

1

metric ton of CO2. Unused EU allowances belonging to Phase I expired at the end of 2007 and could not be banked and used during Phase II. As a consequence, the Phase I Allowances (EUA), could either converge to zero, or reach the upper bound of € 40, the penalty price. Due to the banking restrictions enforced between the two Phases, futures prices for Phase II (2008-2012) proved to be much more reliable than spot prices during Phase I (2005-2007) (Alberola and Chevallier, 2009). The volatility of the prices for Phase I allowances was aggravated by the banking restriction on trading between periods. (Ellerman and Joskow, 2008). I choose to confine the empirical analysis to carbon futures of Phase II with maturities from 2008 to 2012, as it proves to be more reliable than an analysis based on the EUA prices of Phase I. On April 22, 2005 the European Climate Exchange (ECX) introduced EUA futures trading with delivery dates between December 2005 and December 20122. The ECX is the leading marketplace in Europe for trading CO2 emission allowances and located in London, UK.

Figure 1. Futures prices of EUA 2007 and EUA 2009 contracts and traded volume 0 5 10 15 20 25 30 35 40 Apr-0 5 Ju n-05 Jul-0 5 Au g-05 Oct-0 5 No v-05 De c-05 Fe b-06 Mar-0 6 May -06 Ju n-06 Jul-0 6 Se p-06 Oct-0 6 No v-06 Jan-07Fe b-07 Mar-0 7 May -07 Ju n-07 Jul-0 7 Se p-07 Oct-0 7 No v-07 Jan-08Fe b-08 Apr-0 8 May -08 Ju n-08 Au g-08 Se p-08 Oct-0 8 De c-08 Jan-09Mar-0 9 Apr-0 9 May -09 Jul-0 9 Au g-09 Se p-09 No v-09 De c-09 V O LU M E ( m il li o n t o n n e s o f C O 2 ) €0 €5 €10 €15 €20 €25 €30 €35 P ri ce p e r to n n e ( E U R ) Total Volume EUA 2009 EUA 2007

NOTE: Futures prices of EUA maturities December 2007 of Phase I (EUA 2007), December 2009 of Phase II (EUA 2009) and total traded volume in million of tonnes of CO2 for all carbon futures on the ECX.

Source: European Climate Exchange, http://www.ecx.eu/ECX-Historical-Data

2

Figure 1 shows the EUA prices for futures contracts of December 2007 and December 2009 and the volume for all traded EUA futures on the ECX. In April 2006, the European Commission released the emissions data for 2005, which revealed that in that year CO2 emissions were approximately 3% lower than the allocated allowances (Ellerman and Buchner, 2008). As a result, the prices of the EUA December 2007 dropped 54% to €14,25 in four trading days (April 25, 2006 to May 1, 2006). For EUA December 2009 futures, this drop was 39% to €19,30. Finally EUA-2007 prices fell under €1 in February 2007 and converged further to €0.03 as the EUA future went to maturity. At the same time EUA prices for Phase II (2008-2012) futures remained relatively steady and have been disconnected from the Phase I prices since November 2006 (Convery et al., 2008). In terms of traded volume, the EU ETS developed strongly. The total spot and futures EUA transactions for 2007 exceeded 2.1 billion tonnes of CO2 and for 2008 these numbers were 3.1 billion, worth approximately €63 billion3. Futures trading on the ECX has grown extensively and the ECX is now the most liquid platform for EUA futures trading4. Table A1 in the appendix shows that during Phase I (2005-2007) the number of futures transactions on the ECX multiplied by almost factor five between 2005 and 2006. Over 2007, the transactions came to 931 million and 88 % thereof referred to contracts with delivery dates in Phase II (2008-2012). In 2008 the traded volume was almost 2 billion with a value of €48.7 billion. Total numbers for 2009 show a 190% year growth to 3.8 billion transactions, with a total value of €50.9 billion only being 105% of the 2008 value. The smaller growth percentage of the total traded value compared to the traded volume can to a large extent be explained by the drop in carbon prices during and after the credit crunch.

The main objective of implementation of the Emission Trading Scheme is to provide an incentive for carbon-intensive industrials to reduce emissions and to invest in low carbon technologies and energy efficiency among CO2-emitting plants. The purpose of the scheme was to ensure not only that overall carbon emissions would be reduced, but also that the cuts are made by those companies that can achieve the most efficient abatement costs5. A prominent carbon research company6 calculated that compared to 2007 levels, European carbon emissions dropped by roughly 3% in 2008, even taking reduced economic output into account.

Figure A1 in the appendix shows the nine industries that are subjected to the EU ETS and their percentages of yearly allocated EU allowances. Over 70% of EU allowances go to combustion installations, and in this sector 71% of the allowances is allocated to power generation. Thus over 50% of

3

State and trends of the carbon market 2009. World Bank CF Research Report, Washington DC.

4 _{Mansanet-Bataller and Pardo (2008) calculated that of all traded EUA futures, 96% is traded on the ECX.} 5

European Commission, 2003

6

the total EU allowances are allocated to the combustion industry on a yearly basis. Trotignon and Delbosc (2008) used the CITL database7 and found that on an industrial level the combustion sector had a shortage of 0.9% on the allocated allowances, and within this combustion sector the power production sector was the only one in with a shortage of 7.1% on the allocated allowances. All other sectors eventually experienced a surplus of allowances. This coexistence of short and long positions explains why the allowance market grew over the years and why an allowances price for carbon emission abatement evolved. The short position on the combustion sector also explains why the relative switch price between natural gas and coal has been pointed as a main driver of the carbon emissions (Mansanet-Bataller et al. (2007), Rickels et al. (2007), Alberola et al. (2008), Chevallier (2009)). The switch relationship reflects the carbon abatement possibility of switching from coal to a less carbon intensive fuel like natural gas. Moreover, a survey of the EU8 in 2005 showed that 70 % of the combustion installations incorporated the value of carbon allowances into their daily management operations in the start-up phase of the EU ETS, sectors other than the combustion sector9 averaged only 36%. However, energy-intensive industries are generally locked into long-term fuel contracts and may have been unable to switch, or unwilling to do so until the price signal was more stable. In the short term the fuel switching opportunity is not necessarily a substitution within the same plant, it can best be seen as a substitution across plants (Hinterman, 2008).

75 % of the total EUA’s of Phase II are allocated to EU15 countries and the companies that are short on allowances are almost entirely power generators located in the EU15 (Ellerman and Joskow, 2008). The rationale for this explicit political decision taken by many EU15 governments was that power utilities had more abatement possibilities than other industries in the short run, by switching to less carbon intensive fuels. Also the power generating sector does not face international competition with companies outside of the EU. The political choice for a shortage on power generators makes it an interesting research subject. In the EU no common market for power exists. The price differences on European power trading places have diminished over the last few years, but prices have not fully integrated10 (Zachmann, 2008). For each country the natural resources differ, this difference influences the power generating mix11 and, consequently, the possibilities for carbon emission abatement will differ. The country-specific EUA allocation and the structure of the power markets influence the demand on EUA allowances (Oberndorfer,

7

The Community Independent Transaction Log is the central European registry maintained by the European Commission. The CITL records the issuance, transfer, cancellation, retirement and banking of carbon allowances that take place in the registry of each Member State.

8

Review of EU ETS 2005 Survey Highlights, http://ec.europa.eu/environment/climat/pdf/highlights_ets_en.pdf

9

Sectors other than the combustion sector are steel, pulp and paper, cement, refineries, aluminum and chemicals.

10

Zachmann (2008) analyzes wholesale power prices in 2002–2006 and rejects the assumption of full market integration of the European power sector.

11

2009). An increase in volatility in the power market could influence the demand on future carbon emission allowances and increase the EUA market’s volatility. This is of special relevance given the young and strong growing EUA market and the volatility of the EUA price. The objective of the EUA carbon futures is to provide an incentive to reduce carbon emissions and to invest into low carbon technologies. For industrials the relationship of their main carbon emission generator and the EUA price is very interesting. The EUA futures market is used as a price signal, each company wants to refer the future investment in abatement options to the carbon price. Some studies have analyzed the price fundamentals of this young spot market, especially on Phase I. For players on the EUA market it is useful to get a clear picture of the EUA futures market and the relationship to it’s main fundamental price determinants.

The aim of this paper is to provide an analysis of the EU Emission Allowances (EUA) futures of Phase II by focusing first on the empirical relationship between the EUA returns and its main fundamental price determinants. Secondly, this paper aims to analyze the exchange-specific effects of the main European spot and futures power markets and the possibility of the spot-volatility spillover. With the hypotheses that the power markets have different impacts on EUA futures, because the power generation mix and the political choice of EUA allowance allocation differs over countries. The daily EUA futures are from the ECX between the first quotation on April 22, 2005 and December 31, 2009. (1196 observations), are analyzed with a Threshold-GARCH(1,1) model. Combustion fuel prices, several European power prices and weather influences are considered as possible carbon market determinants. For the fuel prices in particular oil, natural gas, coal and the relative prices for switching from coal to natural gas are used. For power data I use six spot markets, the German EEX, the English APX-UK, the Dutch APX, the Spanish Omel, the French Powernext and the Norwegian Nordpool and two year futures markets of the EEX and Powernext. This analysis will also focus on the effect of the volatility of power spot markets on carbon allowances prices. Daily traded volume is used in the variance equation to straiten out the effects of access trading and uncertainty around the announcement of the National Allocation Plans (NAP).

II. LITERATURE REVIEW

The EU ETS is not the first large-scale environmental market, in 1995 the United States successfully established a sulfur dioxide (SO2) allowance market. This SO2 permit trading system allowed companies to bank allowances freely for use in future years, enabling them to incorporate SO2 prices into their long-term investment decisions. Relative to the EU ETS, the SO2 market is rather illiquid and exhibits a large (29%) number of zero-returns (Paolella and Taschini, 2008). Ellerman and Montero (2005) still find that banking has been surprisingly efficient and that aggregate behavior of the SO2bank indicates that most agents have made reasonably efficient abatement decisions during the first Phase and the first two years of the second Phase. In the environmental trading systems the SO2 and CO2 allowances can be considered as a new kind of commodities. They differ from other commodities such as natural gas and crude oil, as storage costs for allowances are zero. Also on commodity markets the main pricing issue refers to the size of stocks and reserves, on the allowances market the main question is the expected shortage of emissions. Moreover, in order to operate installations do not need to physically hold allowances, they only have to comply with verified CO2 emissions on a yearly basis.

The benefits of saving transaction costs by holding allowances on hand are called a convenience

yield. In theory a large marginal convenience yield tends to cause spot prices to exceed future prices12

(Pindyck, 2001). Surprisingly, by comparing ECX prices Paolella and Taschini (2008) discover a contango term structure meaning that the EUA futures prices exceed the expected spot prices and the EUA futures prices will decline to the EUA spot price as it reaches maturity. Borak et al. (2006) find that the price behavior of carbon allowances in the spot and futures market is substantially different from those of other commodities. They observe that the market has changed from backwardation to contango13, where futures prices for Phase II are clearly higher than the current spot prices. Moreover, Uhrig-Homburg and Wagner (2007) find that the futures market leads the price discovery process of CO2 emission instead of the spot market. This confirms my choice in this price-determinants study to focus on EUA futures of Phase II instead of EUA spot prices of Phase I.

12

For example, for a resource commodity like crude oil, one would expect the futures market to exhibit a predominant weak or strong backwardation as a result of a convenience yield. Indeed this is the case for crude oil (Pindyck, 2001).

13

EUA Price-Determinant Studies

Benz and Trück (2009) categorize the principal driving factors of environmental allowance markets into, firstly, policy and regulatory issues and, secondly, into market fundamentals that concern the production of CO2, the demand for allowances. Created by regulation, the market’s supply side of CO2 allowances is dependent on policymakers and their decisions concerning the National Allocation Plans (NAPs). As a result, industries operate with a long-term price signal that depends on uncertain future international political events. A change in policies or regulations could have substantial consequences on actual demand and supply and thus on short-term price behavior of emission allowances. The consequences of such regulatory changes can be sudden price jumps, spikes or phases of extreme volatility in carbon allowance prices. Benz and Trück (2009), model the EUA price and its volatility mainly for risk management and forecasting purposes. While very useful for companies that need to hedge against the risk embedded in carbon prices, they do not ultimately shed light on the fundamental carbon price determinants.

to oil prices. Unexpectedly their analysis suggests that weather is not a major contributing factor to EUA price development14.

Table 1a: Literature overview of papers on price determinants of EUA’s.

Authors EUA Subject Method Period Data Significant Results

Mansanet-Bataller et al. (2007) Phase I Spot prices (CO2-index, EEX) OLS Jan 2005 to Nov 2005

-Natural Gas and Brent Oil futures contract 2005 (IPE) -Power equivalent of month futures (EEX)

-German Weather data

Positive price effects of Oil, power, switching relationship and only extreme cold temperatures

Rickels et al. (2007) Phase I Spot prices (PointCarbon) GARCH(1,1) Jan 2005 to Dec 2006

-Euro Natural Gas (Zeebrugge) -Brent oil

-Coal (Global RB Index) -European Weather Index -No power data

Positive price effects of oil and gas, a negative effect of coal. Especially cold days have impact on carbon prices. Alberola et al. (2008) Phase I Spot prices (Powernext Carbon) OLS Jul 2005 to Apr 2007

-Natural gas Month Ahead (Zeebrugge)

-Coal, Month Ahead (CIF ARA) -Brent Oil Month Ahead (ICE) -Switch relationship

-Temperature and break dummies

-Power Month Ahead (Powernext)

-Clean dark and spark spreads

In the first sub period institutional decisions have a large influence. In the second period EUA prices react similarly but with more significant to energy fundamentals and clean dark spread shows a significant negative sign.

Sanin & Violante (2009) Phase II Futures 2008 (ECX) GARCH(1,1) with jump process Apr 2005 to Dec 2007

-Natural gas Month Ahead (Zeebrugge)

-Coal, Month Ahead (CIF ARA) -Brent Oil Month Ahead (ICE) -Power Month Ahead

(Powernext)

-Temperature dummies -Dummy for NAP announcements - Traded Volume EUA

No evidence on the EUA fundamentals, except for Brent oil. As drivers of the regimes shifts they find: the change in traded volume and changes in the regulatory environments Chevallier (2009) Phase II Futures 2008-2012 (ECX) Treshold GARCH(1,1) Apr 2005 to Oct 2008 -Macroeconomic variables -Natural gas Month Ahead (ICE) -Brent oil, spot

-Power Month Ahead (Powernext)

-Dummy for breaks on Apr’06 and Aug’07

The EUA market is remotely connected to macro-economic factors. Gas and Brent are significant and positive. Power is not significant, but positive. Alberola Chevallier and Cheze (ACC, 2009) Phase I Spot prices (Bluenext) Treshold GARCH(1,1) Jul 2005 to Apr 2007

-Natural gas Month Ahead (Zeebrugge)

-Coal, Month Ahead (CIF ARA) -Switch relationship

-Winter and break dummies -Power Month Ahead (Powernext)

-Clean dark and spark spreads -The combustion and iron sector production country-indices level.

Energy coefficient all show high significance and the excepted sign. The industrial production has impact on EUA price changes in Germany, Spain, Poland and the UK and underlines the central role of German power producers.

14

Alberola et al. (2008) find that EUA price fundamentals vary between sub-periods, before and after “the compliance break” in April 2006. They added the clean dark and clean spark spreads15 to their analysis, with the argument that power operators pay close attention to dark and spark spreads and to the difference between them. In the analyses correlation or multicollinearity between the clean spreads and the carbon prices is not discussed or even mentioned. But one could argue that clean spreads are not completely exogenous, as carbon costs are included, and should not be included in a determination of carbon fundamentals. Analyzing Phase I EUA spot prices using OLS regressions with a Newey-West procedure, Alberola et al. (2008) found that Brent oil, power, low temperatures16 and the switch variable were significant before the break, as in line with Mansanet-Bataller et al. (2007). Natural gas, coal, clean spark and clean dark are not significant before the break. After the compliance break spot prices react similarly but with more significance to energy fundamentals and clean dark spread shows a significant negative sign. They argue that institutional decisions seem to have more influence than the expected determinants in the first sub period.

Using a ARMA-GARCH model, Sanin and Violante (2008) do not find evidence to support the assumption that EUA returns can be explained by energy fundamentals, except for a weak significance of Brent oil. As drivers of the regimes shifts they find the change in traded volume and changes in the regulatory environments. Remarkable of the six studies overview presented in table 1a, Sanin and Violante (2008) are the only ones using the traded volume of EUA’s in their price fundamentals analysis. Their result show a destabilizing effect of large incoming volumes which translates into large negative returns and sudden volatility movements. Sanin and Violante (2008) argue that significant spikes in returns can be motivated by trades placed by large investors in relatively illiquid markets, even in the absence of important news about market fundamentals. Like Sanin and Violante (2008), I will also introduce the traded volume change as a carbon price determinant, unlike all other studies named in table 1a. Large incoming volumes destabilize the young market and lead to large price and volatility movements. This indeed encourages the use of change in traded volume as an explanatory variable of carbon prices.

Chevallier (2009) examines the empirical relationship between Phase II EUA futures and changes in macroeconomic conditions, using T-GARCH(1,1) model, with a sample from April 22, 2005 to October

15

The clean dark spread represents the difference between the price of a power unit and the price of coal used to generate that unit of power, diminished with the corresponding CO2 emission costs. The clean spark refers to the spread calculated similarly with natural gas and power.

16

1, 2008. Variables to capture macroeconomic influences are a common stock portfolio, the junk bond yield, the T-Bill rate, and the market portfolio excess return. Furthermore, power, Brent oil, natural gas and a Dummy for April 2006 are included. The results suggest that the carbon market is only remotely connected to macroeconomic risk factors. Chevallier (2009) describes the economic logic behind these results as a primary relation of carbon futures to the abatement opportunity of fuel switching for power producers. Adding that carbon allowances form a specific market among energy commodities, the results could therefore be relevant for portfolio diversification purposes. Alberola, Chevallier and Cheze17 (2009) find positive, very significant coefficients for natural gas, power and the clean spark. Coal, clean dark and the winter-dummies have a negative significant effect on the EUA returns. ACC (2009) test the combustion and the iron sector production indices at country-level. The industrial production has an impact on EUA price changes in Germany, Spain, Poland and the UK and underlines the central role of German power producers.

Hinterman (2008) discusses the inclusion of Brent oil in the analysis of EUA determinants, as oil is not widely used in Europe in the combustion sector to generate power. The transport sector has a large oil consumption, but transport is not included in Phase II of the Emissions Trading Scheme. Alberola et al. (2008), Oberndorfer (2009), Chevallier, le Pen and Sévi18 (2009) all quote Kanen (2006) as a main reason to use Brent oil prices. Kanen’s logic describes Brent oil prices as the main determinant of natural gas prices which, in turn, affect power prices and ultimately carbon prices. ACC (2009) do not use oil as a fundamental in their analysis of the carbon spot price during the pilot phase. They argue with Kanen (2006) that the effect of the oil price is already captured through the influence of the natural gas price. I choose to include Brent oil prices in my analysis, as Alberola et al. (2008) and CPS (2009) find large significant positive effects of oil prices on the EUA’s, mostly larger than natural gas effects.

A change in policies or regulations could have substantial consequences for the price behavior of emission allowances. The consequences of regulatory changes can be sudden price jumps, spikes or phases of extreme volatility in carbon allowance prices. Analyzing Phase I EUA prices with Unit root tests, Alberola et al. (2008) identify two structural changes: on April 2006, the disclosure of 2005 emissions, and on October 2006, following the EC announcement of stricter Phase II allocations. They test spot prices during the full period using dummies for each structural break. Also two sub-periods are tested:

before and after the compliance break, covering the period from April 25 to June 23, 2006. Analyzing

Phase II carbon futures for all maturities, Chevallier (2009) also uses a dummy variable for the

17

Alberola Chevallier and Cheze (2009) will further be named ACC (2009).

18

compliance break in April 2006. He adds a post-August 2007 dummy variable during the full sample testing and creates before and after August 2007 subsamples, in order to rule out the potential influence of the credit crunch crisis. Rickels et al. (2007) observe an increase of volatility after news on the NAP’s that dominated other variables, especially at the beginning of the EU ETS. Like Sanin and Violante (2009) I consider the relative change in the daily volume of future contracts traded on the ECX as a determinant for the occurrence of price-jumps. Sanin and Violante (2008) is the only study that uses the traded volume of EUA’s in their price fundamentals analysis. As Alberola et al. (2008) and Chevallier (2009) I will use a dummy for the structural break in April 2006. But the April 2006 will be for a shorter period, as Alberola et al. (2008) focuses on Phase I, were the declining period of the carbon allowances is much longer19. I choose not to include a credit crisis dummy, like Chevallier (2009) uses. Because this crisis dummy would interfere with the explanatory power of energy and power variables, as they are also affected by the economic slowdown20.

Review Power Studies

Since the second half of the 1990s, European power markets have experienced a wide liberalization process and have changed their structure from a regulated monopoly to a competitive open market. All European energy exchanges systems are non-mandatory, since producers and consumers may choose either to interact on the exchange or to enter bilateral contracts, (over the counter) for the delivery of power in the short and in the long term. Most European power exchanges only trade spot prices, a futures derivatives market began very recently or it still does not exist21. In a perfectly integrated power market, spot prices should be highly correlated. However, Da Silva and Soares (2008) indicate that, although spot prices seem to be more convergent within each one of the European markets, they are still far behind the aimed integrated level. The spot power markets show high volatility, this is due not only to underlying volatile energy markets, but also to the nascent nature of the liberalized power market in Europe (Rademakers et al., 2008). The liberalization process and the growing number of markets and players have increased the attention of power producers and distributors on financial performances and risk management, which is particularly needed to stabilize the competitive open market. This is essential to optimize the market risk depending on the observed volatility (Zanotti et al., 2010).

ACC (2009) study the country-specific effects of the production of the combustion and the iron sector on the EUA price during 2005-2007. Using a T-GARCH(1,1) model they analyze monthly sector

19

In figure 1 the difference in declining period between the Phase I and Phase II futures is very clear.

20

Including a credit crisis dummy in the model gave less significant results, especially in testing the last year separately.

21

production indices at country-level for Germany, Spain, France, Italy, Poland and the UK. They find that the industrial production has impact on EUA price changes in Germany, Spain Poland and the UK and underline the central role of German power producers.Bunn and Fezzi (2007) use a co-integrated VAR model to analyze the relationships between allowances, power and natural gas in the UK daily spot markets22. They show that the carbon and gas prices help determine the power price. Hintermann (2008) discusses whether power prices are exogenous and should be included as a carbon fundamental price determinant. Endogenity will lead to biased coefficient estimations of the price determinants. Power prices might be influenced by carbon prices, because carbon costs can be included and have an effect on the power prices for consumers. On the other hand, in the EU ETS the largest part of EUA’s are freely granted to power corporations23. Power prices constitute an important determinant of the EUA price given the proportion of allowances distributed to the power sector. The carbon costs that corporations actually bear in Phase II are only their shortages on yearly allocated allowances. When power demand rises above the granted allowances, carbon demand and carbon prices will as a consequence thereof rise as well.

Reinaud (2007) studies the interaction between carbon allowances and power prices. She states that in a competitive market, the pass-through of CO2 costs to the consumer in power prices is real and inevitable. Nevertheless, Reinaud admits not having focused on the demand side of power which is also an important component of power price levels. The rates at which carbon costs are passed through to consumer prices are affected by market competitiveness and elasticity of supply and demand. Profits increase for some generators when their power generation mix has relative low emissions, and so they benefit from power market price increases. In this way emissions trading results in windfall profits for power corporations. Sijm et al. (2006) estimate marginal pass-through rates using empirical data from forward energy and fuel markets. They conclude that in 2005 for the Netherlands 60–80% of CO2 costs have been passed on to the power markets and for Germany 60–120%. Sijm et al. (2006) note that in the Netherlands power prices are usually set by gas fired plants and that the share of non-carbon fuels in total power production is low. In countries where the power generation mix varies, the windfall profit can be substantially different from the one in the Netherlands. This makes it interesting to focus on the various power markets effects on the carbon allowance prices.

With 100% the carbon allowances in the EU ETS being freely allocated24, power generators can profit from the EU ETS and from the fact that the profit is positively related to the EUA prices. The

22

Bunn and Fezzi (2007) use daily temperature in London and a Dummy variable for April 2006 as exogenous variables.

23

Recall Trotignon and Delbosc (2008) calculated a net shortage for power corporations of 7.1%.

24

analysis of Chen et al. (2008) indicates that the rates at which CO2 costs are passed through to wholesale prices are affected by market competitiveness, merit order changes, and elasticities of supply and demand. Oberndorfer (2009) shows that EUA spot price changes have a positive effect on the stock returns of the most important European power corporations. German and UK power corporations stocks are positively related to the EUA prices and Spanish corporations exhibit a negative EUA-to-stock market relationship. Oberndorfer (2009) also notes that the pass through possibilities depend on differences in country-specific EUA long/short positions due to the NAP’s and the structure of the power markets. Moreover, it is unclear whether such an effect would be stable over time. Zanotti et al. (2010) estimate hedging effectiveness for three of the most liquid power European futures markets: NordPool, EEX, and Powernext. Results from NordPool and the EEX show that trading in futures allows to reduce the risk of power portfolios and to control the risk of adverse movements on power prices25. The only market where the hedging does not lead to a variance reduction is the Powernext market that also proves to be the most recent and less liquid market. Table 1b, gives an overview of the relevant literature in power studies and EU Allowances. The common denominator in the results of Chen et al. (2008) Oberndorfer (2009) and Zanotti et al. (2010) is that the relationship between EU allowances and power prices is time and country specific.

Table 1b: Literature overview of papers on the power markets and carbon emission rights

Authors Method Period Data Results

Chen et al. (2008) COMPETES Simulations of the short-run competition in the electricity market - Northwest continental European wholesale power markets. Belgium, France, Germany EEX and the Netherlands APX

The rates at which CO2 costs are passed through to wholesale power prices are affected by market competitiveness, merit order changes, and elasticities of supply and demand. Emissions trading results in large windfall profits, much but not all of which is due to free allocation of allowances

Oberndorfer (2009) OLS and GARCH Aug, 2005 to June, 2007

Stock returns of the most important European power corporations

EUA prices have a positive effect on stock returns of European power corporations.

The carbon market effect is shown to be both time- and country-specific

Zanotti et al. (2010) OLS and GARCH July 2002 to Feb 2006

Spot prices and futures contracts of Nordpool, Powernext, EEX

Futures hedging on the Nordpool and EEX power markets leads to reductions of the variance. Hedging on the Powernext does not lead to a variance reductions.

Like Alberola et al. (2008) and Chevallier (2009), I include the one month ahead power prices as a fundamental determinant in the EUA analyses. They found positive significant influences on the carbon prices. Their main reasons for choosing a month ahead contract instead of spot prices is, firstly, that companies only have to comply with their carbon emissions on a yearly basis. The second reason is for the sake of consistency with the other used energy variables. In this paper the spot and futures European

25

power prices are introduced as EUA price determinants. Spot prices reflect an important market in power trading, since power is not storable and not easily transported. This unique nature of power as a commodity makes the possibility of sudden price changes more likely. In the second part of this paper I analyze the spot prices of different European power exchanges. In this analysis the power spot price volatility will be used as measure in the variance equation of the GARCH model (Oberndorfer, 2009).

Apart from the power spot prices analysis I will conduct a base power year futures analysis with the available data from the EEX and the Powernext. These German and French power year futures are the only year futures available, that are traded on a large basis26. German year futures have the main focus in power futures trading (Mansanet-Bataller et al., 2007). With its large offer of futures derivatives, the German EEX has also developed into an important reference market for the carbon emission trading27. In the analysis conducted the carbon futures with December year maturities will be related to power futures with maturities in the same year. In this way the data are used in the most relevant way.

Hypotheses

The aim of this paper is twofold, first to provide an analysis of the EUA futures price changes of Phase II by focusing on the empirical relationship with its main fundamental price determinants. The returns analyzed are of the daily EUA futures prices traded on the European Climate Exchange (ECX) between the first quotation on April 22, 2005 and December 31, 2009 (1196 observations). Following previous literature I consider that as energy determinants Brent oil and natural gas have a positive price EUA effect. The effect of coal and the relative switch price between gas and coal are expected to be negative on the EUA futures, null hypothesis of there being no such significant effects on the EUA futures. The German month ahead power future is used to analyze this first model; this market is seen as an European reference market. I will use the concept of weather influence on carbon prices, but use the heating degree days index instead of temperature dummies to test the expected price increasing weather effect on carbon futures. An April 2006 dummy is included for the structural break in the EUA price data. I choose not to include a credit crisis dummy, as this would interfere with the explanatory power of energy and power variables, because they are also affected by the economic slowdown. The traded volume changes are used to capture and straiten out the effect of sudden large volatility movements, as a result of the access trading and uncertainty around the announcement of the National Allocation Plans (NAP), by

26

The Dutch Endex also introduced power year futures, but were not freely accessible. In most European markets, a futures derivatives market began very recently or it still does not exist.

27

arguing that at times decisions and announcement on allowance allocation have more impact than the expected determinants. (Sanin and Violante, 2009).

H1: The fundamental determinants of the EUA futures contract price changes are: gas, coal, oil, switch relationship, heating degree days, power month ahead futures contract.

The second part of this paper aims to analyze the exchange-specific effects of the main European power exchanges and the possibility of their spot price volatility spillover. With the null hypothesis of there being no difference in the influence among European power on carbon futures. All literature mentioned above introduce only one power variable in their analyses, with the assumption that all European power markets have the same magnitude and positive relationship to the carbon emission price. My hypothesis is that if this is not the case, power prices from different exchanges will have significantly different effects on the carbon prices, in size and significance. Exchange-specific EUA positions and the structure of the power markets are of influence on carbon emissions (Oberndorfer, 2009). Also the Emission Allowances are not evenly allocated to the power and heat sector. My research can be distinguished from other studies by empirically testing the effects of several European power base spot markets, namely: the German EEX, the English APX-UK, the Dutch APX, the Spanish Omel and the French Powernext and the Norpool exchange in Norway. I also test the volatility effect of these power markets on the volatility of the carbon futures, to research a possible volatility spillover. In addition, I analyze the influence of the base power year futures contracts of the EEX and the Powernext. With its large offer of futures derivatives, the German EEX has developed into an important reference market for the carbon emission trading. In the analysis conducted the carbon futures with December year maturities will be related to power futures with maturities in the same year. In this way the power price data are used in the most relevant way.

H2: The main European power exchanges have different effects on the EUA futures contract price changes. The exchange-specific effects are investigated for base power spot prices and base power futures contracts.

III. DATA

The daily carbon price data are EU Emission Allowance futures contracts of the maturities December 2008 to December 2012 from the European Climate Exchange28 (Chevallier, 2009). The data sample totals 1196 observations, running from April 22, 2005 to December 31, 2009. In Figure 2 the daily price series are given for all five futures. Daily data is the only realistic frequency as weekly or monthly data would provide too few observations in order to conduct a serious time series analysis. The EUA futures contracts are traded in € / ton of CO2. The returns for all EUA price data are calculated as the first difference of the natural logarithm of the price series29. As a determinant for the occurrence of price jumps, I consider the change in the daily traded volume of transactions of all futures contracts30 (Sanin and Violante, 2009). The daily traded volume reflects the volume of all EUA futures contracts traded on the ECX, see figure 1 in the introduction.

Figure 2: Price series for the EUA futures contracts of maturities December 2008 to December 2012. Source: EEX

€ 0 € 5 € 10 € 15 € 20 € 25 € 30 € 35 € 40 ap r-05 ju n-05 au g-05 ok t-05 de c-05 fe b-06 ap r-06 ju n-06 au g-06 ok t-06 no v-06 ja n-07 mrt -07 mei -07 jul-0 7 se p-07 no v-07 ja n-08 mrt -08 mei -08 jul-0 8 se p-08 ok t-08 de c-08 fe b-09 ap r-09 ju n-09 au g-09 € / t o n n e C O 2

dec-08 dec-09 dec-10 dec-11 dec-12

28

The ECX settlement prices reflect the weighted average of trades during the daily settlement period (16:50:00-16:59:59 hours UK local time).

29

EUA returns are defined by Rt=ln(Pt / Pt − 1) with Pt the daily EU allowance spot price at time t.

30

Energy Variables

The following price series are taken from Thomson Financial DataStream. For energy fuel prices I use month ahead futures, because installations do only need to hold allowances matching their carbon emission levels once a year. This to provide a better analysis of the EUA prices due, reflecting the fact that most energy fuel needs are met by forward contracting. (ACC, 2009) The natural gas price represents the one month ahead from the Intercontinental exchange (ICE) in British Pence / Thermal Units (Chevallier, 2009 and CPS, 2009). This source for natural gas was chosen, because gas trading at the EEX in Leipzig and APX in Zeebrugge is more recent and so there are not sufficient data available for the whole sample period. The disadvantage of using UK data is that gas prices can be driven by fundamentals of its domestic supply and demand. However, through arbitrage possibilities UK and continental gas prices are closely related. The oil price is the daily Brent crude Month Ahead Free-on-board price in US$/Barrel. Brent is the most relevant traded crude for European energy firms. (Oberndorfer, 2009) To represent coal price returns I use the Global Insight Coal Index Basis 6000 in US$/Gigajoule (Rickels et al., 2007). A gas price increase leads to switching effect towards coal, which increases emissions and therefore the demand for carbonallowances. A coal price increase leads to a switching effect in the opposite direction, lowering therefore the demand of allowances. The difference will influence the direct abatement options (Rickels et al. 2007). The opportunity to switch to gas grows when the gap between gas and coal prices becomes smaller. The switch relationship variable between gas and coal is defined as the return on the difference between gas and coal prices31(Mansanet Bataller et al., 2007) and (Delarue et al., 2008). To ensure that all price series are traded in the same currency, the price series are converted to euro using the European Central Bank exchange rate. I transform the price variables (Natural gas, oil, coal, the switching variable and power series) into returns by taking their first natural logarithm differences. An overview of the definition of all variables and the data source is given in table 3.

Power Variables

In power pricing, contracts can be traded with two profiles: base load or peak load. The Base load has a constant rate of delivery on all days from Monday to Sunday and during all 24 delivery hours32, and will be used in this research context (Chevallier, 2009). The current month futures price series for German power will be used for consistency with the other energy month forward variables in the first analysis and to compare results with other research. The Phelix Month Base €/Megawatt Hour is from the European

31

The switch return variable is defined as Switch=ln((Pgast-Pcoalt)/(Pgast-1-Pcoalt-1))

32

Energy Exchange (EEX) in Leipzig. The German EEX reflects the largest European power exchange (Oberndorfer, 2009). Mansanet-Bataller et al. (2007) pointed out that the EEX is an important reference market for power traders. With its large offer of futures derivatives, the EEX has also developed into an important reference exchange for the carbon emission trading33. In most European markets, a futures derivatives market began very recently or it still does not exist.

Table 3: Overview of all variables, the definition and the data source.

Name variable in research Definition Source

EUA - one-year EUA year futures contracts, December 2008-2012 ECX

Volume Traded volume of all one-year EUA futures of the ECX ECX

Gas ICE Natural Gas 1 Mth.Fwd. P/Therm DataStream

Coal Global Insight Coal Index Basis 6000 DataStream

Oil Crude Oil-Brent Cur. Month FOB U$/BBL DataStream

Switch Gas-Coal spread DataStream

Heat-days Monthly index on heating degree days of EU27 Eurostat

E DE-month Phelix Month Base E/MWH DataStream

E DE-spot EEX - Phelix Base Hr.01-24 E/Mwh DataStream

E UK-spot APX Power UK Spot Base Load Index DataStream

E NL-spot APX Electricity NL Avg All Hours DataStream

E FR-spot Powernext Elec. Baseload E/Mwh DataStream

E SP-spot OMEL-Elec. Spain Baseload E/MWh DataStream

E NP-spot Nordpool-Electricity Avg Reference DataStream

E FR - one-year Powernext Base one-year futures, maturities 2009-2012 EEX E DE - one-year Phelix Base one-year futures, maturities 2009-2012 EEX

To compare the power prices I will use six spot price series of major European power exchanges. I examine Base load spot price series of the main power exchanges, namely: the German EEX – Phelix Base, the English APX-UK, the Dutch APX-NL, the Spanish Omel, the French Powernext and the Nordpool exchange that covers Norway, Denmark, Sweden and Finland. Table A2 in the appendix provides an overview of all European power exchanges and the traded volumes in GWh over 2007. The main spot exchanges that are left out are in this research: the Italian IPEX, the Austian EXAA, the Polish Towarowa and the Belgian Belpex, as these are not offered by DataStream. Especially the missing of data for the Italian exchange is regrettable, as the IPEX has the largest spot volume with 329,949 GWH traded in 2007. Among the available data, Nordpool shows to be a large physical spot power exchange, with 290,000 GWh in traded volumes. Nordpool only offers spot trading and week futures. Monthly, quarterly and yearly forwards with a maximum time horizon of 4 years are only traded over the counter (OTC) at Nordpool. The EEX is the second largest spot exchange, with 117,321 GWh in traded volume over 2007.

33

The EEX offers by far the widest range in power derivatives, namely spot (day ahead and intraday) and month-, quarter- and year futures. Much smaller exchanges are Powernext, the APX and Omel. The APX is a provider of power and gas exchanges for spot trading in the United Kingdom and the Netherlands. The APX had the initiative to couple the Dutch, French and Belgian power exchanges34 and it has improved the border power flows among the markets. (Rademaekers et al., 2008).

Power generation structure differences influences the need for the amount of carbon allowances, per country. Table A4 in the appendix the power generation mix for the analyzed European countries is shown, also the import and export per country is given. The Nordic countries Norway, Sweden, Denmark, and Finland have a very diversified generation mix altogether. Ranging from high on hydro in Norway to very spread out over sources in Finland. Germany is by far the largest coal user, with 311 TWh and France shows a high dependence on nuclear energy, with 440 TWh. Da Silva and Soares (2008) find some evidence for local integration and some power spot price convergence. Compared to the other exchanges only France and Germany appear to be relatively integrated with higher correlation coefficients. Spain appears to be poorly integrated with the other locations as might be expected as a result of its peripheral position and limited cross-border transmission capacity. Figure B1 in the Appendix shows the German and French spot prices and the German current month series. The integration between French and German spot prices is clearly visible. Figure B2 shows the spot prices for Norway, the Netherlands, the UK and Spain. Nordpool in Norway, that shows the lowest price volatility of all exchanges, is also the most mature market in all of Europe. Rademakers et al., (2008) state that over time, power price volatility of the Nordpool exchange tends to decrease. The high volatility observations in the other spot power markets may be due not only to underlying volatile primarily energy markets (coal, gas and oil) but also to the nature of the liberalized power market in Europe (Rademakers et al.,2008).

In this analysis the measure of spot price volatility is the moving average of the standard deviation of the last 50 trading days of the power log returns. The influence of the volatility of power prices on the carbon futures will be tested in the variance equation of the GARCH model. (Oberndorfer, 2009). By doing so, the model allows the conditional variance of the futures returns error term to be not only determined by its own dynamics, but also by external power volatility factors. This approach relates to the literature of so-called volatility spillovers (Hamao et al., 1990) within our setting from the power spot market to the carbon market.

34

Power Futures Contracts

The year futures power contracts are only offered on three exchanges: the Powernext, EEX and Endex, where the EEX Phelix futures sums up the majority in volume, see table A2 in the appendix. The EEX offers Phelix futures derivative up to six year of maturity. In cooperation with the EEX, the French Powernext offers futures trading up to three years. Unfortunately the Endex series are not freely available on their site. EEX Phelix and French futures series with maturities 2009 to 2012 are taken from the EEX35. The E DE-2009 to -2011 year futures data series start at April 21, 200536. The French futures, the data series starts at August 29, 2005. Both the DE-2012 and FR-2012 year futures start at January 1, 2006. The DE-2009 and FR-2009 series mature at December 23, 2008. The DE-2010 and FR-2010 mature at December 24, 2009. Figure B3 shows the EEX Phelix and the French Base futures for 2010 and 2011. The futures show a very similar price path.

Heat-days variable

In previous literature temperature variables have shown to be fundamental price determinants of carbon futures prices. (Mansanet-Bataller et al., 2007, and Alberola et al., 2008) Extremely hot and cold days were converted into monthly dummy variables. Both studies showed that temperatures beyond a certain threshold can lead to increases in power demand and so influence the carbon price. I would like to introduce a variable with respect to temperatures and the power demand: the heating degree day. This is a quantitative index calculated by Eurostat37, the Statistical Office of the European Communities. Eurostat developed the index for climatic correction of final energy consumption for space heating purposes in the 27 Member States of the European Union. The heating degree days are defined on the basis of average temperatures, weighted by the population and their heating requirements of the representative regions that make up each country. The indices can be summed over time and place. Because the Heating degree day index is calculated on a monthly basis, the relative change cannot be used as a explanatory variable. The monthly values over the total sample will be converted into a daily time-series by taking the heating degree days value of a month and substituting that monthly value for every day in the corresponding month.

35

A cooperation between Powernext and the EEX. http://www.eex.com/en/Market%20Data

36

This gives a little less observations for the French year futures, see the descriptive statistics in table 3.

37

Descriptive Statistics

Descriptive statistics for all time series are presented in Table 3. The returns of the EUA futures all exhibit negative skewness and excess kurtosis38. This means that extreme return values occur with a higher frequency than implied by the normal distribution. These summary statistics therefore reveal an asymmetric and leptokurtic distribution. Such a fat-tailed distribution can be analyzed with generalized autoregressive conditional heteroskedasticity (GARCH) modeling as GARCH models attend well to excess kurtosis in the data.

Table 3: Summary Statistics for the log returns of all time series, except for the heat-days variable.

Mean Median Maximum Minimum

Std.

Dev. Skewness Kurtosis Observations

EUA-2008 -0.00011 0.00093 0.1865 -0.2882 0.0287 -1.32 17.69 931 EUA-2009 -0.00014 0.00064 0.1932 -0.2811 0.0287 -0.88 14.38 1185 EUA-2010 -0.00026 0.00000 0.1912 -0.2743 0.0285 -0.86 13.78 1196 EUA-2011 -0.00022 0.00000 0.1866 -0.2678 0.0282 -0.86 13.40 1196 EUA-2012 -0.00018 0.00000 0.1823 -0.2616 0.0283 -0.87 13.06 1196 Volume 0.00213 -0.00048 3.3207 -3.6835 0.5828 0.04 10.02 1196 Gas -0.00014 -0.00301 0.4769 -0.2606 0.0504 2.56 20.16 1196 Coal 0.00010 -0.00006 0.2106 -0.3490 0.0215 -3.55 102.63 1196 Oil 0.00024 0.00077 0.1557 -0.1253 0.0228 0.07 6.83 1196 Switch 0.00683 0.00000 3.8500 -2.4032 0.3297 2.35 35.39 1196 Heat-days 0.23809 0.22193 0.5867 0.0131 0.1801 0.30 1.64 1196 E DE Month -0.00010 0.00425 0.4935 -0.8712 0.0853 -2.89 33.17 1196 E DE-Spot -0.00038 0.00060 1.0964 -1.0764 0.2052 -0.14 7.65 1196 E UK-Spot -0.00037 -0.00606 1.4229 -0.8543 0.1959 0.39 7.38 1196 E NL-Spot -0.00055 0.00288 1.6892 -1.5402 0.1978 0.01 15.10 1196 E SP-Spot -0.00224 -0.00152 0.4495 -1.7401 0.1039 -4.33 72.57 1196 E FR-Spot -0.00046 -0.00689 2.3841 -2.2139 0.2036 0.28 33.25 1196 E NP-Spot 0.00016 -0.00260 0.4471 -0.6885 0.0718 -0.16 14.85 1196 E DE-2009 0.00037 0.00055 0.0691 -0.0633 0.0105 -0.19 11.41 938 E DE-2010 0.00008 0.00000 0.0693 -0.0634 0.0108 0.06 10.06 1193 E DE-2011 0.00019 0.00000 0.0732 -0.0643 0.0099 0.31 11.08 1196 E DE-2012 0.00020 0.00000 0.0668 -0.0576 0.0091 0.31 10.22 1022 E FR-2009 0.00037 0.00035 0.0552 -0.0558 0.0107 -0.30 8.20 850 E FR-2010 0.00006 0.00000 0.0654 -0.0561 0.0116 -0.05 7.51 1106 E FR-2011 0.00017 0.00000 0.0588 -0.0545 0.0106 0.22 7.51 1109 E FR-2012 0.00011 0.00000 0.0588 -0.0563 0.0104 0.06 7.51 1022

NOTE: EUA values refer to the daily carbon futures log returns. The EUA-08 future has only 933 observation, because it matured in December 2008. Volume refers to the change values calculated for all traded carbon futures on the ECX. Heat-days refers to the heating degree days and is the only variable not treaded as a log return. Elec-DE-Month refers to the month forward power return of the EEX. E-DE, UK, NL, SP, FR are the spot price returns of the associated markets. E DE and E FR refer to the corresponding year futures power contracts from the EEX.

38

Coal shows a negative skewness and a large excess kurtosis. Oil and gas both show positive skewness and a smaller excess kurtosis. The month ahead German power return (E DE-Month) has a large negative skewness. For the other day ahead power returns, the distributions show a smaller skewness, and also excess kurtosis.

Table 4 shows that the correlation between EUA-2010 futures price change and all introduced variable data are positive. The absolute values of the correlation coefficients are modest, especially among the fuel variables. Amongst the explanatory variables, the EUA price change correlates relatively strong with natural gas and oil returns, 14 and 26% respectively. For coal one could expect a negative correlation, but the value is close to zero. Some of the correlations between energy price variables are rather low, particularly between the gas price change and both the oil price changes. A reason for this might be that gas prices at European energy exchanges seem to be neglected by financial market agents, probably due to the widespread use of long-term gas contracts (Oberndorfer, 2009). The correlation between gas and oil is 11%. Gas and coal show a negative relation of 5%, which reflects the possibility of their substitution. Gas also has 12% correlation with the UK spot power, revealing the relationship of the relative large part of power generated by natural gas in the UK. The various power variables show the largest correlations. They will be tested separately, so that multicollinearity should not cause problems between the variables.

Table 4: Matrix of correlations N=1196

EUA-

2010 Volume Gas Coal Oil Switch Heat-days E DE Month E DE -Spot E UK -Spot E NL -Spot E SP -Spot E FR -Spot E NW -Spot E DE -2011 EUA-2010 1 Volume 0.03 1 Gas *0.14 0.02 1 Coal 0.00 0.01 -0.04 1 Oil *0.26 0.01 *0.11 0.03 1 Switch 0.01 0.01 0.03 *0.08 -0.04 1 Heat-days 0.00 0.00 *-0.09 -0.01 0.01 0.00 1 E DE Month 0.02 *0.09 0.03 0.01 -0.02 0.00 -0.02 1 E DE-Spot 0.01 *0.14 -0.04 -0.03 0.00 0.00 -0.02 *0.33 1 E UK-Spot 0.04 *0.09 *0.12 0.01 0.02 0.04 -0.01 -0.02 *0.10 1 E NL-Spot 0.02 *0.16 0.00 -0.05 -0.02 0.02 -0.02 *0.18 *0.43 *0.11 1 E SP-Spot 0.04 *0.07 0.02 0.00 0.01 0.01 -0.04 0.03 *0.13 0.05 *0.07 1 E FR-Spot 0.06 *0.13 -0.02 -0.03 0.01 -0.01 -0.02 *0.23 *0.43 *0.12 *0.52 *0.19 1 E NP-Spot 0.01 *0.13 -0.02 -0.03 0.00 0.00 -0.02 *0.17 *0.31 0.06 *0.26 *0.08 *0.25 1 E DE-2011 *0.39 0.00 *0.18 -0.03 *0.33 *-0.09 0.02 -0.01 -0.02 0.06 -0.05 -0.04 -0.02 0.01 1 E FR-2011 *0.36 0.02 *0.17 -0.04 *0.32 *-0.10 0.02 -0.02 -0.04 *0.08 -0.04 -0.04 -0.01 0.01 *0.85

Subsequently, I will use Unit root tests to test the return series on stationarity. A stationary variable can be defined as one, where parameters such as mean, variance and autocovariances for each given lag do not change over time or position. The equilibrium errors may therefore temporarily deviate from zero but should always revert back to zero. The use of non-stationary data can lead to spurious regressions (Brooks, 2002 p.367).

Table 5: Unit root tests on return series

ADF PP KPSS EUA-2008 -26.9** -27.0** 0.10 EUA-2009 -30.5** -30.6** 0.08 EUA-2010 -31.0** -31.0** 0.12 EUA-2011 -31.1** -31.1** 0.12 EUA-2012 -31.5** -31.6** 0.12 Volume -27.1** -84.1** 0.10 Gas -33.0** -33.0** 0.09 Coal -34.6** -34.6** 0.16 Oil -35.1** -35.1** 0.11 Switch -33.7** -34.5** 0.03 Heat-days -3.9** -1.8 0.07 E DE Month -30.7** -44.3** 0.0 E DE-Spot -24.1** -80.9** 0.1 E UK-Spot -19.9** -101.6** 0.1 E NL-Spot -24.7** -77.3** 0.1 E SP-Spot -27.0** -37.1** 0.3 E FR-Spot -22.8** -70.7** 0.05 E NP-Spot -29.6** -42.1** 0.03 E DE-2009 -28.2** -28.2** 0.32 E DE-2010 -31.0** -31.0** 0.50* E DE-2011 -31.1** -31.1** 0.30 E DE-2012 -28.6** -29.3** 0.22 E FR-2009 -26.1** -26.2** 0.23 E FR-2010 -29.2** -29.1** 0.34 E FR-2011 -23.7** -29.0** 0.25 E FR-2012 -27.9** -27.9** 0.15

NOTE: ** = 1%, * = 5% significance level. In the test equation only a constant is included. ADF refers to the Augmented Dickey–Fuller test and shows the t-statistic. PP to the Philips-Perron test, the adjusted t statistic is given. KPSS to the Kwiatkowski-Phillips-Schmidt-Shin test with the LM statistic. The lag structure in the ADF test is selected automatically on the basis of the Scharz Information Criterion (SIC). In the case of the PP and KPSS tests the bandwidth parameter is selected automatically according to the Newey and West (1994) approach.

series I have performed three unit root tests, to check for stationarity: the Augmented Dickey-Fuller test (ADF), the Philips-Perron test (PP) and the Kwiatkowski-Phillips-Schmidt-Shin test (KPSS). The ADF and PP test a null hypothesis of a unit root in series. A significant test statistic means rejection of the assumption of a unit root, and indicates stationarity. In contrast, the KPSS test assumes a stationary time series under the null hypothesis. A significant test statistic here indicates a unit root.

For the ADF test, the lag-length is chosen according to the Schwarz information criterion (SIC), the bandwidth parameters for the PP and KPSS tests are chosen according to the Newey-West approach.39 The results of all three tests are included in Table 5, the ADF and PP statistics show that all series are stationary on a 1% significance level except for the Heating degree days. For the Heat-days variable the log return has not been taken, as it is a monthly variable. This shows that by taking first natural logarithms differences of price series, they can be converted into stationary variables. The KPSS statistics doesn’t reject stationarity for any variable. Though in many financial time series, the variance of the residuals is not constant over time and appears to be related to the magnitude of variance of recent residuals. Ignoring such heteroskedasticity may result in loss of efficient estimations. The ARCH test is a Lagrange multiplier (LM) test for autoregressive conditional heteroskedasticity (ARCH) in the residuals. The EUA futures returns are regressed on a constant. The ARCH LM test is performed on the simple regression’s residuals, with the null hypothesis that there is no ARCH up to lag-order q, results are in table 6. The F-statistic shows an omitted variable test for the joint significance of all lagged squared residuals, with the accompanying probability. For the q=1 lags the null hypothesis is rejected for all carbon futures, except the EUA 08. For the q=5 lags the null is rejected for all carbon futures. Pointing out the ARCH effects in all the carbon futures, a Generalized Autoregressive Conditional Heteroskedasticity (GARCH) is explicitly designed to model time-varying conditional variances.

Table 6: ARCH LM test statistics

F-statistic (q=1) P-value F-statistic (q=5) P-value

EUA-2008 2.542 0.111 6.189 0.000

EUA-2009 7.268 0.007 9.775 0.000

EUA-2010 8.365 0.004 10.652 0.000

EUA-2011 9.515 0.002 11.609 0.000

EUA-2012 12.467 0.000 12.182 0.000

Note: The ARCH LM test results on the residuals of OLS regressions on a constant for each EUA future contract. The residuals are tested with q=1 and q=5 lag-orders.

39

Dummies

IV. MODEL & METHODOLOGY

The summary statistics in table 3 show that the carbon futures of all maturities exhibit non zero skewness and excess kurtosis. The EUA returns reveal a fat-tailed distribution with volatility clustering. The GARCH model (Bollerslev, 1986) provides a good description of two important characteristics of the time series: time-varying volatility and leptokurticity. Since EU allowances are traded since 2005, there is not many historical data available. By analyzing first empirical data I consider the appropriateness of several GARCH models specified by Patterson (2000). To eventually analyze the EUA returns, I choose, in the same order as Chevallier (2009) from these five models; the traditional GARCH, the Component GARCH, the Exponential GARCH, the Power ARCH and the Threshold GARCH model. The traditional formulation of the GARCH(1,1) model assumes a linear function where the moving average of past variances in included present conditional variances, where, p is the order of the autoregressive GARCH term and q is the order of the moving average ARCH term. The specification of the mean, here as a function of an exogenous variable with an error term, may be written as:

t

t

c

R

=

+

ε

ε

_t =

σ

_tZ_t,where (Zt) idd ~ N (0,1) (1)

where Rt is the EUA return at time t, and εt the error term. With the conditional variance:

∑

∑

= − = −

+

+

=

p j j t j q i i t i t 1 2 1 2 0 2

σ

β

ε

α

α

σ

(2)

where

α

₀ is the constant term,

ε

2t−1 represents the squared residuals from the mean equation of the previous period, (the ARCH term),

σ

2t−1 is last period’s fitted variance (the GARCH term). The conditional variance coefficients have to satisfy

∑

α

_i

+

∑

β

_j

<

1

and

α

_i,

β

j≥0,

α

₀ >0 to ensure stationarity and a strictly positive conditional variance (Patterson 2000, p.715). Identification and estimation of GARCH models is performed by maximum likelihood estimation. Although the GARCH(p,q) model overcomes the important problem of heteroskedasticity, it is limited by a non-negativity constraint on the conditional variance. Also the basic GARCH(p,q) model is a symmetric variance process and does not cover the leverage effects.

The Component GARCH model.