Carbon intensity and stock return of firms: evidence from the European Union Emission Trading Scheme

(1)

Carbon intensity and stock return of firms:

evidence from the European Union

Emission Trading Scheme

Master’s Thesis Finance

Author:

D.G. (Daniël) van Ours1

Supervisor:

Dr. R.O.S. Zaal MBA

Submitted: January 11, 2018

(2)

1

Abstract

The European Union implemented the EU Emission Trading Scheme as an incentive for polluters to reduce their emissions. Because of the program, the carbon emissions in the EU are capped, traded and priced through carbon allowances. In this paper, an empirical analysis of the stock return of carbon emitting firms under the EU ETS is presented. Firstly, the difference in the stock performance of high carbon intensity (carbon emissions relative to their sales) firms relative to the stock performance of low carbon intensity firms is examined. For this, the Capital Asset Pricing Model (CAPM) is used. Secondly, a firm-specific analysis of the exposure of carbon allowance price movements on the stock return of carbon emitting firms is undertaken. For this analysis, a modified version of the CAPM is used. Monthly data from 2009-2016 is used, covering the second and third phase of the EU ETS. Energy-intensive firms which are subject to the EU ETS are included in the dataset. These firms are divided into three different portfolios, based on their carbon intensity profile; the Low Carbon Intensity (LCI) portfolio, the Medium Carbon Intensity (MCI) portfolio and the High Carbon Intensity (HCI) portfolio. After evaluating the results, it can be concluded that the stock performance of HCI firms is not significantly lower than the stock performance of LCI firms. Moreover, the stock returns of the majority of carbon-emitting firms are not significantly exposed to carbon allowance price movements. These results indicate that firms under the EU ETS do not seem to be ‘punished’ for having a relatively high carbon intensity. And that the carbon allowances introduced by the EU ETS are not considered a relevant risk factor for CO2 emitting firms. The results remain intact after testing for robustness.

(3)

2

1. Introduction

Global warming is a big concern for the world nowadays. Scientists are pointing towards global warming as a mayor cause for the uprising of natural disasters, like hurricanes and floods. Governments are getting increasingly worried about the climate change and act together to restrict global warming. In 1997, the first major treaty between nations was adopted to cope with this threat of global warming. The Kyoto Protocol came into force, committing its parties to binding emission reduction targets. This is considered an important first step towards a global emission reduction, to stabilize the Green House Gasses (GHG) emissions2_{. With the successful negotiation}

of the Paris Agreement in 2015, a new course was set to strengthen the global climate effort. A goal for this century was set; keeping the global temperature well below 2 degrees Celsius above pre-industrial levels. Thereby, for the first time, all nations were brought together to discuss the mitigation of the GHG emissions3.

One option for governments to regulate and decrease carbon emissions, is through a cap and trade system. An acceptable emissions limit can be set for the industries which are covered by the cap and trade system. This emission limit is referred to as the ‘cap’. It is argued that a cap-and-trade system is the most cost-effective way to reach a countries abatement. It is less costly for the affected firm and the government in comparison to the implementation of pollution taxes (Hahn and Stavins, 2010). The European Union implemented such a cap and trade system to combat climate change and meet its Kyoto targets. The European Commission developed the biggest trading emission scheme in the world, named the European Union Emission Trading Scheme (EU ETS)4_{. Because of it, the carbon emissions in the EU are capped, traded and priced through carbon} allowances, named European Union Allowances (EUA). The system covers, besides the power and heat generating sectors, the following energy-intensive industry sectors; oil refineries, steel works and production of iron, aluminium, metals, cement, lime, glass, ceramics, pulp, paper, cardboard and bulk organic chemicals. The EU ETS is divided into three phases. The first phase covers the period of 2005 to 2008. This is also referred to as the pilot phase, with the intention of gathering experience to improve the implementation in subsequent periods. The second phase covers the period of 2009 till 2012, which coincided with the first commitment period of the Kyoto Protocol. In this phase, each member state of the EU had to develop a National Allocation Plan (NAP). The plan contained two main things; the quantity of carbon allowances which the country intended to issue and how they proposed to distribute these allowances to firms covered by the EU ETS. The NAP had to be approved by the European Commission. It is also important to consider the nature of the carbon allowances issued in the second phase of the EU ETS. The carbon allowances which are issued to the companies in the first phase, did not count in the second phase. Carbon allowances which are issued in the second phase can also be used in the third phase. In these first two phases, the majority of carbon allowances were given free of charge. From 2013 onwards, the ETS entered the third phase, with considerate changes compared to the previous phases. Instead of a national cap, an EU-wide emission cap was implemented. This EU-wide emission cap was set at two billion tons of CO2 emissions in 2013. This cap reduces each year,

(4)

3 aiming for an overall reduction of 21% below 2005 levels by 2020. The firms which are covered by the EU ETS must monitor their own emissions and report accordingly. Most importantly, in the third phase, auctioning is the default method for allocating carbon allowances. A total 40% of allowances were auctioned in 2013, progressively increasing towards auctioning 70% of all carbon allowances by 2020. Carbon allowances give firms the right to emit CO2. One allowance carries a pollution right of one ton of CO2. The companies need to submit the emission allowances that match the actual emissions during the compliance period. Non-compliance results in a fine. In phase one, the fine for non-compliance was €40 per European Union Allowance (EUA) not surrendered. While in phase two and three, companies are fined €100. This fine is substantially higher than the price of a carbon allowance, which at the date of writing is €6.905_{. These fines can} result in major costs, as big firms in the energy intensive industries emit millions of tons of CO26. If firms want to emit less than the carbon allowances they had received, they were able to sell these extra allowances on the carbon market. Firms who are likely to emit more than their allocation have two choices. The first is to take measures to reduce their emissions, which will require significant investments. The second is purchasing additional carbon allowances. It is possible to purchase carbon allowances from Member State held auctions or from companies who received allowances, but do not need them. The EU ETS resulted in the largest multinational carbon market in the world. The ETS system tries to create an incentive for polluters to reduce their emissions.

The actual influence of the EU ETS on the stock return of a firm remains unclear. There is no consensus in the literature, mainly because in the first two phases of the EU ETS, the allowances were granted free of charge. Veith, Werner and Zimmermann, (2009) analyzed the capital market responses of companies in the power sector in the first phase of the EU ETS. They found a positive correlation between rising prices of the carbon allowances and stock returns. Attributing this to the fact that firms achieve windfall profits, because they are overcompensated for the possible costs of the carbon allowances. Anger and Oberndorfer, (2008) investigate the role of the EU ETS first phase on firm stock prices in Germany. They do not find a significant impact of carbon allowance allocation on the stock prices of firms. This is supported by Abrell, Ndoye and Zachmann (2011), whom examine data in both the first and the beginning of the second phase the EU ETS. According to their findings, there were no significant losses of competitiveness for companies under the EU ETS. Sijm, Neuhoff and Chen, (2006) found that in the first period of the EU ETS, the free distribution of carbon allowances resulted in a profit increase for firms in the sectors covered by the EU ETS. Oestreich and Tsiakas (2015) analyzed data of the first two phases of the EU ETS. They also found an increase in stock return for carbon-emitting firms in the first phase of the EU ETS, due to the free distribution of carbon allowances. They found significantly higher stock returns for firms with high CO2 emissions, relative to the stock returns of firms with low CO2 emissions. This so called ‘carbon premium in stock returns’ was found for the period of 2003-2009, covering the first phase of the EU ETS. The difference in stock returns lead to an increased profit for carbon emitting firms. The higher profits could be attributed to two main reasons. Firstly, carbon emitting firms face higher marginal costs due to the opportunity cost of carbon allowances and can pass these costs through to consumers by increasing their prices. Secondly, the free carbon allowances not used in production, are sold for a profit. The abnormal

5_{According to the ICE database, retrieved from DataStream. Spot price on the 10}th_{of October, 2017.}

(5)

4 increase of profits for carbon emitting firms was the major critique of the EU ETS. Firms with higher carbon emissions had the opportunity to make higher profits in the first phase of the program, relative to firms with lower carbon emissions, due to the free distribution of carbon allowances.

When the ETS entered the third phase and the majority of the allowances were not distributed free of charge anymore, firms had to purchase relatively more carbon allowances to cover their CO2 emissions. Because carbon emitting firms had to buy carbon allowances in this phase, their cost of production increased. These additional costs can result in an additional risk for the company, a carbon risk. Carbon allowances are thereby fluctuating in price, so firms can also be subject to a carbon risk related to carbon allowance price movements.Investors have difficulty incorporating this carbon risk in their stock price valuations (Busch and Hoffmann, 2007). Investors nowadays are confronted with this uncertainty, because there is only little research done on the effect of the EU ETS on the stock returns of firms in the second and third phase, as it was introduced only quite recently. This paper tries to fill this gap by providing an empirical analysis of the stock return of carbon emitting firms under the EU ETS in this period. It is analyzed whether under the EU ETS, the stock performance of companies with a higher carbon intensity (carbon emissions relative to their sales) differ from the stock performance of firms with a lower carbon intensity. To address this, firms from energy-intensive industries are divided into three different portfolios, based on their carbon intensity profile; the Low Carbon Intensity (LCI) portfolio, the Medium Carbon Intensity (MCI) portfolio and the High Carbon Intensity (HCI) portfolio. To be more specific, two main questions are addressed in this paper. Firstly, is there a difference in the stock performance of HCI firms relative to the stock performance of LCI firms? Secondly, what is the firm-specific exposure of carbon allowance price movements on the stock returns of carbon emitting firms? To my best knowledge, this is the first paper to provide an empirical analysis of the difference in stock performance and carbon exposures which firms in the energy-intensive industries face, based on their carbon intensity, while using data from both EU ETS phase two and phase three.

The remainder of this paper is structured as follows. Chapter 2 presents the theoretical arguments for the hypotheses tested in this paper. Chapter 3 describes the data and the framework of the models used for the empirical analysis. Chapter 4 presents the empirical results and descriptive statistics of the models. The last chapter concludes, accompanied by closing remarks and recommendations for future research.

2. Theoretical Background

(6)

5 with stock prices in the United States. This negative relation is more pronounced for firms which are more carbon intensive. Furthermore, Matsumura, Prakash and Vera-Muñoz (2014) find evidence for a strong negative relation between carbon emissions and firm value. For each additional metric ton of carbon emissions, firm value decreases by $212. They argue that this value penalty materializes, because investors think carbon emitting firms are subject to increased carbon liabilities due to their carbon emissions. In Australia, research was conducted to assess the relation between the market value of a firm and their carbon intensity profile (Chapple, Clarkson and Gold, 2013). Interesting to note is that they did this in the light of an upcoming Emission Trading Scheme. The results indicate that the market assessed the higher carbon intensity firms with a market value penalty, arguing that these firms are most highly at risk. They found a market value penalty of 6.57% of market capitalization of these high carbon intensive firms relative to low carbon intensive firms. They attributed this to the negative financial impact of the upcoming ETS in Australia. A study with a sample of 843 firm-year observations in the EU ETS in the first phase of the program by Clarkson et al. (2014), reveals that there is a valuation penalty of €75 per ton of uncovered emissions. The amount of CO2 emissions in excess of the free carbon allowances received can be labelled carbon liabilities and negatively affect the value of a company. Busch and Hoffmann (2011), studied the effect of corporate environmental performance on corporate financial performance of the 2500 largest companies by market capitalization. They found that a lower carbon intensity is positively associated with the financial performance of a firm.

Prior literature thus suggests that a relatively high carbon intensity has a negative effect on the stock price of a firm. This effect can be attributed to both an increase in the carbon emission related costs of the EU ETS and because investors assign increased liabilities to carbon emitting firms. On top of this, there is a shift in consumer behavior. The latter is a result of the fact that there is a strong public opinion that measures are urgently needed to counter the climate change. Therefore, the public is actively supporting companies which have a positive impact on the environment. For example, they are considering low-carbon and energy-efficiency products as positive aspects in their choice of products (Brinkman, Hoffman and Oppenheim, 2008). As the public understands the necessity for tackling the climate change issues, it is in the interest of powerful stakeholders to proactively address the efforts of corporations to manage their carbon emissions (Wood and Jones, 1995).

(7)

6 with a low carbon intensity. Therefore, the following hypothesis is developed and analyzed in this paper;

Hypothesis (1). The stock performance of high carbon intensity firms is lower than the stock performance of low carbon intensity firms.

The introduction of carbon allowances in the EU ETS can influence the market’s expectation of the stock returns of carbon emitting firms. Prior research suggests that the price fluctuations of carbon allowances can be an additional source of risk for firms under the EU ETS (Benz and Trück, 2009). They find that the introduction of carbon allowances can lead to an increase in the production costs and therefore generates an additional risk for carbon emitting firms. Delarue, Voorspools and D’haeseleer, (2008) analyzed the potential effect of a price increase in carbon allowances, under the EU ETS they are named European Union Allowances (EUA). They provide evidence that overall, emission reductions increase with increasing EUA price. An EUA price of 20 euro per ton of CO2 emission is sufficient to trigger significant emission abatement. The increase in EUA price triggers firms to invest in switching to a lower carbon intensive fuel mix. This is attributed to the fact that firms which invest in carbon emission abatement will be less exposed to higher EUA prices. Firms whom heavily depend on carbon emissions and therefore have a higher carbon intensity, will be more exposed to fluctuations in the carbon allowance price. Koch and Basen (2013) study the carbon exposure of European electric utilities under the EU ETS. They reveal that the stocks of high carbon intensive firms are significantly exposed to EUA price movements. It is argued that investors require significant carbon risk premiums on stock returns from high carbon intensive firms, if the price of EUA increases. Thereby, they find that an active strategy towards a carbon-free fuel-mix will reduce a firm’s stock exposure to EUA price fluctuations.

But, it is important to consider a firm’s ability to pass on these carbon costs. Firms can mitigate their exposure to EUA price movements, by passing the extra carbon allowance related costs on to consumers. Sijm, Neuhoff and Chen, (2006) revealed that power companies in Germany and The Netherlands were able to pass through between 60% and 100% of their carbon costs to consumers. The finding that firms are able to pass-through their costs is supported by Grainger and Kolstad, (2010), who argue that carbon regulation related costs are ultimately born by consumers. This is an argument why EUA price movements might not be considered a relevant risk factor for carbon emitting firms. Thereby, Goulder, Hafstead and Dworsky, (2010) analyzed the impact of emission allowance allocation under a federal cap-and-trade program and found that freely allocating a relatively small fraction of carbon allowances already suffices to prevent profit losses for carbon emitting firms.

(8)

7 carbon allowances. The stock returns of high carbon intensity firms are therefore argued to be significantly exposed to EUA price movements. If EUA price movements is a risk factor for high carbon intensity firms, investors should demand higher returns for the higher risk (Fama and MacBeth, 1973). To be more specific, investors shareholding a high carbon intensive company should require carbon risk premiums on stock return for the extra exposure to EUA price movements. In this light, the following hypotheses are developed and analyzed in this paper;

Hypothesis (2a). Stock returns of high carbon intensity firms are exposed to EUA price movements and bear carbon risk premiums.

Hypothesis (2b). Stock returns of other carbon emitting firms are not exposed to EUA price movements.

3. Methodology

To test the hypotheses in this paper, various analyses are undertaken. The sections in this chapter are as follows. First, the criteria for the data sample and period are described in section 3.1. Section 3.2 describes the measure for Carbon Intensity and the pooling of firms accordingly. Furthermore, in section 3.3, the use of the CAPM is described to analyze the difference in stock performance of firms in the HCI portfolio, relative to the stock performance of firms in the LCI portfolio. Section 3.4 describes the model which analyzes the firm-specific exposure of EUA price movements on the stock return. In section 3.5, the monthly data used in the analyses is described.

3.1 Data Sample criteria

The role of the EU ETS is important in this study, as this cap and trade system regulates the carbon emissions and actively puts a price on it through the introduction of carbon allowances. From January 2009, the second phase of the EU ETS was initiated. And from January 2013, the third phase of the EU ETS was initiated. This is the date from which auction was gradually becoming the primary way of trading carbon allowances. The legislative procedures of the EU ETS are considered important in this paper. Firstly, it is important to consider the nature of the carbon allowances issued in the second and third phase of the EU ETS. The carbon allowances which are issued to the companies in the first phase, did not count in the second phase. Carbon allowances which are issued in the second phase can also be used in the third phase. Secondly, in 2008, a year before the second phase started, the long-term reduction target for carbon emissions was revealed. So, starting from the beginning of the second phase in 2009, the stock return of firms should also encompass a long-term signal. As the carbon intensity profile of 2009 can be used to forecast future compliance costs. Therefore, data from 2009-2016 is used, covering two phases of the EU ETS. This comes down to a total of eight years.

(9)

8 following sectors are picked from the ASSET4 database; firms in the power generation industry and firms which manufacture or produce metals, like iron, steel and copper. Both industries are energy-intensive industries. All firms in the dataset are under the ETS directive, so must deliver carbon allowances for their carbon emissions. Taken that the ETS Directive is EU legislation, its territorial scope is restricted to the territory of the EU. In this sense, any multinational company should offset the emissions of their installations to be found on the territory of an EU Member State7. Therefore, companies who have their main operations outside of the EU territory, are not included in the dataset.8

Firms in the power generation industry are handpicked and critically analyzed. The most important reason for including a firm in the dataset, is that their main activity comprises the generation of power. A total of 15 companies met this criteria and are therefore included in the dataset. The main criteria for firms in the metal industry to enter the dataset, is that their main activity is producing or manufacturing metals, which is generally a heat-producing business. A total of 14 companies met the criteria and are used in the analysis. The final analysis excludes one company from the power generation industry and three companies from the metal industry. For these firms, sufficient data is missing for the full sample period of 2009-2016. The companies in the final dataset are from the following EU nations; Italy, Spain, Finland, United Kingdom, Greece, Luxembourg, Austria, Germany and France. The final dataset comprises 25 companies from nine different countries.

3.2 Carbon Intensity measurement and pooling

To assess the carbon intensity of all firms in this dataset, a firm’s Carbon Emissions are retrieved from the DataStream database. This dataset is publicly available. The Carbon Emissions are measured in metric tonnes of CO2 emissions per company per year. To develop a measure for Carbon Intensity (CI), the Carbon emissions of a firm is divided by the Net Sales revenue in thousands, in line with previous studies who assess the carbon intensity of a firm (e.g. Chapple, Clarkson and Gold, 2013; Busch and Hoffmann, 2011). The Net Sales revenue of the firms are acquired from Datastream and are on a yearly basis. The Net Sales Revenue is transformed into euros, to compare accordingly. The following formula (1) is used to calculate the carbon intensity of a firm;

𝐶𝐼 =

𝑇𝑜𝑡𝑎𝑙 𝐶𝑂2 𝑒𝑚𝑖𝑠𝑠𝑖𝑜𝑛

𝑁𝑒𝑡 𝑆𝑎𝑙𝑒𝑠 𝑅𝑒𝑣𝑒𝑛𝑢𝑒/ €1,000

(1)

For the analyses in this paper, it is important to pool the firms of the data sample into different groups, according to their Carbon Intensity. They are therefore divided into three Carbon Intensity groups with the use of cut-off points. The firms with an average carbon intensity of 1.5

(10)

9 or higher for the period of 2009-2016, are classified as High Carbon Intensity firms. Firms with an average carbon intensity of 0.5 or lower for the period of 2009-2016, are classified as Low

Carbon Intensity firms. The firms in between which do not match these criteria are classified as Medium Carbon Intensity firms. This way, the sample is segmented into three distinct portfolios:

High, Low, and Medium Carbon Intensity portfolios. Table 1 presents the firm-specific Carbon Intensities and the resulting segmentation of firms into the different Carbon Intensity portfolios. The cut-off points used for the pooling of firms are based on my own evaluation of the data. Table 1

Classification of the firms under the EU ETS. The table reports the Carbon Intensities of all firms, calculated according to formula (1). Both the Carbon Intensity of 2015 and the mean of the Carbon Intensity in the entire period of 2009-2016 are shown. 2015 is the most recent year where both the carbon emissions and the Net Sales are known for all companies9_{. The sample is segmented into three groups, based on the firm’s Carbon Intensity; High, Low, and Medium} Carbon Intensity firms. Both the country where the firm is based and the sector in which the company operates are given for all companies.

Company Country Code Sector Carbon Intensity Carbon Intensity

(2015) (2009-2016)

High Carbon Intensity firms

ENEL IT POWER 1.64 1.63

ENDESA SP POWER 1.77 1.61

FORTUM FIN POWER 5.60 4.26

DRAX GROUP UK POWER 3.06 7.78

PUBLIC POWER GR POWER 6.16 7.66

ARCELORMITTAL LUX METALS 3.33 3.11

ERAMET SA FR METALS 1.84 1.51

Average Carbon Intensity firms

GAS NATURAL SDG SP POWER 0.92 1.12

IBERDROLA SP POWER 1.04 1.27

SSE UK POWER 0.30 0.60

EDP ENERGIAS DE PORTUGAL SP POWER 1.77 1.34

VERBUND AU POWER 0.69 0.94

A2A IT POWER 1.57 1.12

EDISON RSP IT POWER 0.55 1.14

THYSSENKRUPP GER METALS 0.79 0.69

VOESTALPINE AU METALS 1.17 1.19

SALZGITTER GER METALS 0.89 0.86

KLOECKNER & CO GER METALS 1.07 1.04

TENARIS LUX METALS 0.49 0.64

Low Carbon Intensity firms

EDP RENOVAVEIS SP POWER 0.02 0.02

ACCIONA SP POWER 0.09 0.12

ACERINOX 'R' SP METALS 0.08 0.08

AURUBIS GER METALS 0.15 0.12

VALLOUREC FR METALS 0.26 0.29

HUNTING UK METALS 0.04 0.05

(11)

10

3.3 CAPM: excess stock returns

The intention of this analysis is to investigate if there is a difference in stock performance between HCI and LCI firms. This is done to evaluate hypothesis 1. The analysis will be based on a standard factor model, the Capital Asset Pricing Model (CAPM). The CAPM is still argued to be the best asset pricing model available (e.g. Smith and Walsh, 2012; Da, Guo and Jagannathan, 2012). This model is used to measure the alpha (α) of the firms with differing carbon intensity. The High Carbon Intensity (HCI) firms are all put in one portfolio, the HCI portfolio. The Medium Carbon Intensity firms are put in the MCI portfolio. The Low Carbon Intensity firms are put in the LCI portfolio. The alphas of the different carbon intensity portfolios are evaluated with the use of the CAPM, which is specified as follows;

𝑟𝑗𝑡− 𝑟𝑓𝑡 = 𝛼𝑗+ 𝛽𝑗(𝑟𝑀𝑡− 𝑟𝑓𝑡) + 𝜀𝑗𝑡 (2)

This model describes the abnormal excess return of the different portfolios. 𝑟_𝑗𝑡 is defined as the monthly return of a carbon portfolio j at time t. 𝑟𝑓𝑡 is the monthly riskless return at time t, while 𝑟_𝑀𝑡 is the monthly return of the market portfolio at time t. 𝛽_𝑗 captures the market risk. 𝛼_𝑗 measures the over- or under-performance of a portfolio (Jensen, 1968). Last of all, 𝜀_𝑗𝑡 is the error term of this model. Model (2) is used to capture the stock returns of the portfolios and to assess how the portfolios of differing carbon intensity perform in comparison to each other. The main interest goes to the difference in alpha of the HCI portfolio in comparison to the LCI portfolio.

3.4 Econometric model: EUA price movements

(12)

11 𝜇_𝑖 = 𝑟_𝑓+ 𝛽_𝑖𝑀(𝜇_𝑀− 𝑟_𝑓) + 𝛽_{𝑖𝐸𝑈𝐴}(𝜇_𝐸𝑈𝐴− 𝑟_𝑓) + 𝛽_{𝑖𝐸𝑛𝑒𝑟𝑔𝑦}(𝜇_{𝐸𝑛𝑒𝑟𝑔𝑦}− 𝑟_𝑓) (3)

The expected return of a firm’s stock 𝜇𝑖 is equal to the risk-free rate 𝑟𝑓, plus the risk premiums for market risk, energy risk and carbon risk. The risk premiums are based on the stock sensitivities 𝛽𝑖𝑀, 𝛽𝑖𝐸𝑛𝑒𝑟𝑔𝑦 and 𝛽𝑖𝐸𝑈𝐴 for respectively the market, energy and carbon factors. With the expected returns of 𝜇_𝑀, 𝜇_{𝐸𝑛𝑒𝑟𝑔𝑦} and 𝜇_𝐸𝑈𝐴 for respectively the market, energy and carbon factors.

The ordinary least squares regression (OLS) is eventually used to estimate the firm-specific market, energy and carbon risk premiums of the companies in the dataset. To do this, an econometric model of the underlying basis model (3) is developed.

The dependent variable for the econometric model is the monthly excess return of the stock returns of firms in the dataset. The independent variables are the monthly excess return on the market index and the monthly excess return of EUA, for respectively the market risk and carbon risk. For the control variables, energy price changes are incorporated in the model to capture the energy price risk factor. This is done for coal, oil and gas prices. The motivation for including these variables lies in the fact that energy-intensive companies require significant amounts of energy to produce. Therefore, they amount for significant amounts of risks for firms included in the data sample. The regression equation for the underlying basis model (3) is:

𝑟𝑖𝑡 = 𝛼𝑖 + 𝛼𝑡+ 𝛽𝑖𝑀𝑟𝑀𝑡+ 𝛽𝑖𝐸𝑈𝐴𝑟𝐸𝑈𝐴𝑡 + 𝛽𝑖𝐶𝑟𝐶𝑡 + 𝛽𝑖𝑂𝑟𝑂𝑡+ 𝛽𝑖𝐺𝑟𝐺𝑡 + 𝜀𝑖𝑡 (4)

In this final two–way error component model (4), 𝑟_𝑖𝑡 is the t monthly excess return on the i stocks in the data sample and 𝑟𝑀𝑡 is the t monthly excess return on the market portfolio M. 𝑟𝐸𝑈𝐴𝑡 is the monthly excess return of the carbon allowances. 𝑟𝐶𝑡, 𝑟𝑂𝑡 and 𝑟𝐺𝑡 are the monthly excess returns of the energy prices, for carbon, oil and gas respectively. 𝑟_{𝐸𝑈𝐴𝑡}, 𝑟_𝐶𝑡, 𝑟_𝑂𝑡 and 𝑟_𝐺𝑡 are calculated by subtracting the risk-free return 𝑟_𝑓 of the first log-differenced prices of EUA, carbon, oil and gas. 𝜀_𝑖𝑡 is the error term of this model. The unknown parameters which are estimated by model (4) are 𝛼_𝑖, 𝛼_𝑡, 𝛽_𝑖𝑀, 𝛽_{𝑖𝐸𝑈𝐴}, 𝛽_𝑖𝐶, 𝛽_𝑖𝑂 and 𝛽_𝑖𝐺. 𝛽_𝑖𝑀 captures the market risk of the excess stock return, 𝛽_{𝑖𝐸𝑈𝐴} captures the carbon risk of the excess stock return, while 𝛽_𝑖𝐶, 𝛽_𝑖𝑂 and 𝛽_𝑖𝐺 capture the energy risk of coal, oil and gas respectively. In this model, both time-varying firm fixed effects and time-invariant fixed effects are included. The time-varying firm fixed effects are measured by 𝛼_𝑡, capturing the cyclical determinants that affect all firms similarly. The time-invariant firm fixed effects are measured by 𝛼_𝑖, capturing the unobservable firm characteristics.

(13)

12

3.5 Data asset pricing model

The dataset for the models are considered a balanced panel dataset, as there are no missing data points for any of the variables. The variables are observed for all entities and all time periods on a monthly basis. Monthly data required for the financial variables of models (2), (3) and (4) are obtained from Datastream. Because monthly data is used in this model and the sample period ranges from February 2009 till December 201610, there are 94 observations for each company or portfolio of companies. Monthly stock returns are calculated with the following method; 𝑟_𝑖𝑡+1 = ln(𝑃𝑖𝑡+1) − ln(𝑃𝑖𝑡), where 𝑃𝑖𝑡 is the return index of a stock i at time t. The monthly return of the risk-free rate is proxied by the 10-year German government bond. The market portfolio is proxied by the MSCI Europe. This is a value-weighted European market index11.

For the carbon allowance prices, monthly settlement prices of EUA contracts, traded on the Intercontinental Exchange (ICE), are used. Phase 2 and phase 3 EUA prices are considered to be a reliable price signal for investors, so future contracts are not needed. For the coal, oil and gas control variables, one-month futures contracts of energy prices for coal, oil and gas are considered, instead of spot prices. For all countries these prices can be considered the same, as all countries are situated in the EU. For coal, ICE’s API2 Cost, Insurance and Freight Amsterdam, Rotterdam and Antwerp month forward coal price is used. For oil, ICE-Brent Crude Oil Continuous one-month future price is used. For gas, the Intercontinental Exchange (ICE) Natural Gas one-one-month Forward price is used. Choosing these indicators for the control variables coal, oil and gas is in line with previous literature (e.g. Alberola, Chevallier and Chèze, 2008; Koch and Bassen, 2013; Oestreich and Tsiakas, 2015). For the carbon risk explanatory variable, the return is computed the following way: 𝑟𝐸𝑈𝐴𝑡 = ln(𝑃𝐸𝑈𝐴𝑡) − ln(𝑃𝐸𝑈𝐴𝑡−1), where 𝑃𝐸𝑈𝐴𝑡 is the price of the carbon allowance. The returns of the energy risk control variables coal, oil and gas are also computed in this manner.

4. Results

The regression results and descriptive statistics of all models are evaluated in this section. Firstly, section 4.1 reports the descriptive statistics of the models. Secondly, the results of regressing the CAPM are shown in section 4.2. Most importantly, the difference in stock performance of the HCI portfolio compared to the stock performance of the LCI portfolio is estimated. Thirdly, a rolling estimate of the difference in stock performance of the HCI portfolio and the LCI portfolio is illustrated and discussed in section 4.3. Then, the results of regressing the econometric asset pricing based model are interpreted in section 4.4. With the focus on measuring the firm-specific exposure of EUA price movements on the expected stock return. All the estimations in this paper use standard errors corrected with the Newey-West procedure, which are consistent in the presence of both heteroskedasticity and autocorrelation (Newey and West, 1987).

10_{January 2009 is skipped in this analysis, as the price of a carbon allowance was only €0.20, as this could have had} a major influence on the results.

(14)

13

4.1 Descriptive statistics

Table 2 presents the descriptive statistics of model (4). Interesting to note is that for all variables in the model, the mean excess return is negative. This indicates that over the entire sample period, the risk-free rate was higher than the average return on any of these variables. The correlation matrix shows that most variables are significantly correlated with each other. The return on the market portfolio and the stock return are significantly positively correlated, as expected. Just like all energy price return variables, which are significantly positively correlated with each other. Interesting to notice is that the returns of both the market portfolio and stock returns of individual firms are significantly and positively correlated with the EUA price returns. This could indicate that investors demand higher returns of firms if the price of EUA increases. Results of regressing model (4) should give more clarity on this. No correlations between dependent and independent variables are extremely high (highest is 0.54 between the excess return on coal and the excess return on gas), so there is no problem with multicollinearity.

Table 2

Descriptive Statistics of model (4) variables. Besides a Correlation Matrix, the Mean, Standard Deviation (SD), Minimum (Min) and Maximum (Max) of all the variables in the models are given. 𝑟𝑖 is the t monthly excess return on

the i stocks in the data sample and 𝑟𝑀 is the t monthly excess return on the market portfolio M. 𝑟𝐸𝑈𝐴 is the monthly

excess return of the carbon allowances. 𝑟𝐶, 𝑟𝑂 and 𝑟𝐺 are the monthly excess returns of the energy prices, for carbon,

oil and gas respectively. The sample has monthly data from February 2009 till December 2016. *, ** and *** denote the significance levels at 10%, 5% and 1% respectively.

4.2 CAPM regression results

The regression results of the CAPM (2) are displayed in Table 3. The CAPM-α measures the over- or under-performance of a portfolio, relative to the market performance (Jensen, 1968). When considering the entire period, the stock performance of the HCI portfolio is the only portfolio which significantly underperformed, at the 10% significance level. When looking at the EU ETS second phase (2009-2012), all Carbon Intensity portfolios are significantly underperforming, relative to the market. After evaluating the CAPM-α in the EU ETS third phase (2013-2016), it can be concluded that none of the Carbon Intensity portfolios are significantly over- or under-performing the market. The stock returns of all Carbon Intensity portfolios are highly significantly exposed to the market return, which is captured by the significantly positive market beta. This is true for all time periods. The firms in the High Carbon Intensity portfolio are the most exposed to market return in the EU ETS third phase, with a coefficient of 1.10. Indicating that overall, firms in this portfolio have to pay the highest market risk premiums for risk compensation to their shareholders during the EU ETS third phase.

Variable Mean SD Min Max 1 2 3 4 5

(15)

14 Table 3

Regression results of CAPM (2). CAPM-α is the alpha of the CAPM regression. Thereby, the Market-β and adjusted R2_{for each portfolio are given. Standard deviations are given in parentheses. Both the alphas and standard deviations} are annualized. The portfolios are; High Carbon Intensity (HCI) portfolio, Medium Carbon Intensity (MCI) portfolio and Low Carbon Intensity (LCI) portfolio. The High-minus-Low portfolio is added in this table. Firms in the portfolios and their respective Carbon Intensity measures are listed in Table 1. All portfolios are formed with equal weights. The sample uses monthly data from January 2009 till December 2016. *, ** and *** denote the significance levels at 10%, 5% and 1% respectively.

Portfolio CAPM-α Market-β Adjusted R2

Full sample period (2009-2016)

High Carbon Intensity portfolio -0.10* 0.87*** 0.58

(0.05) (0.18)

Medium Carbon Intensity portfolio -0.06 0.79*** 0.69

(0.04) (0.07)

Low Carbon Intensity portfolio -0.08 0.80*** 0.54

(0.06) (0.07)

High-minus-Low portfolio -0.03 0.073 0.01

(0.06) (0.07)

Second phase (2009-2012)

High Carbon Intensity portfolio -0.17** 0.81*** 0.71

(0.07) (0.09)

Medium Carbon Intensity portfolio -0.15*** 0.77*** 0.74

(0.05) (0.09)

Low Carbon Intensity portfolio -0.11* 0.77*** 0.71

(0.07) (0.07)

High-minus-Low portfolio -0.07 0.04 0.01

(0.07) (0.07)

Third phase (2013-2016)

High Carbon Intensity portfolio -0.02 1.10*** 0.43

(0.09) (0.17)

Medium Carbon Intensity portfolio 0.042 0.84*** 0.60

(0.05) (0.12)

Low Carbon Intensity portfolio -0.05 0.90*** 0.30

(0.09) (0.20)

High-minus-Low portfolio 0.02 0.20 0.02

(0.09) (0.18)

(16)

15 over the LCI portfolio. As can be seen in Table 3, it is also the difference between the alphas of both portfolios. It must be noted that the values differ due to the rounding off by two decimals. The negative value of the High-minus-Low alpha indicates that the stock performance of the HCL portfolio is lower than the stock performance of the LCI portfolio for both the full sample period (2009-2016) and the period of the EU ETS second phase (2009-2012). This is as expected, as it is hypothesized that the stock performance of HCI firms is lower than the stock performance of LCI firms. In the EU ETS third phase (2013-2016), the CAPM-α is positive, indicating that for this period, the stock performance of the HCL portfolio is higher than the stock performance of the LCI portfolio. But, the CAPM-α of the High-minus-Low portfolio is not statistically significant in any of the periods. Consequently, the null hypothesis that there is no difference between stock performance of High Carbon Intensity firms and stock performance of Low Carbon Intensity firms cannot be rejected. For the CAPM analysis on the difference in stock performance between firms in the HCI portfolio relative to firms in the LCI portfolio, no support is found for hypothesis 1.

4.3 CAPM rolling window

To evaluate if there is a time period where the High-minus-Low alpha is significantly different from zero, a rolling estimate of the High-minus-Low portfolio CAPM-α is made. This is illustrated in Figure 1. For this figure, a rolling window of two years is used. With the rolling estimate figure, the exact timing when the High-minus-Low alpha is significantly different from zero can be identified and evaluated. Both the 5% significance level and the 10% significance level are shown in the figure. If the upper or lower bound tips over the x-axis, the CAPM-α is significantly different from zero at that time.

(17)

16 Figure 1

Illustration of the High-minus-Low portfolio stock return. A rolling estimate of the High-minus-Low carbon intensity portfolio CAPM-α is plotted. A two-year rolling estimate is used. Each alpha is plotted as the center of the rolling window, using 12 months before this point and 12 months after. The average of this window is illustrated by each point in the graph. The area between the upper and the lower bound is the significance level, calculated by using Standard Deviations of the alpha. The sample uses data from January 2008 till December 2016. The upper panel displays the boundaries at the 5% significance level (Figure 1a), while the lower panel displays the boundaries at the 10% significance level (Figure 1b).

Figure 1a: Boundaries at the 5% significance level

Figure 1b: Boundaries at the 10% significance level

-0,050 -0,040 -0,030 -0,020 -0,010 0,000 0,010 0,020 0,030 0,040 2009 2010 2011 2012 2013 2014 2015 2016 CAP M -α o f th e Hi gh -m in u s-Low p o rtfo lio

Alpha High-Low lower bound (5%) upper bound (5%)

-0,040 -0,030 -0,020 -0,010 0,000 0,010 0,020 0,030 0,040 2009 2010 2011 2012 2013 2014 2015 2016 CAP M -α o f th e Hi gh -m in u s-Low p o rtfo lio

(18)

17

4.4 EUA price movements regression results

The firm-specific regression results of model (4) are displayed in Table 4. It presents the effect of the carbon risk, market risk and energy risk on the firm-specific stock returns. The joint F-test is rejected at the 1% significance level for all firms. Except for Arcelormittal, which has a F-test value of 0.47. The goodness of fit, represented by the adjusted R2_{, shows that the regression} model fits the data well. As most of the R2 values are in the range of 0.30 and 0.50. There are two firms where the regression model does not seem to fit the data. Just like with the F-test, this is the case with Arcelormittal, with an R2 of only 0.03. The other company where the regression model does not seem to fit the data is Public power, with an adjusted R2 of 0.06. All companies, except for Arcelormittal, show that their stock return is highly significantly exposed to market risk, which is captured by the positive market beta. The market risk exposure to individual stock returns vary largely. Hunting is the most exposed to market risk, with a coefficient of 1.26. Indicating that this firm has to pay the highest market risk premiums for risk compensation to their shareholders. For the variables coal, oil and gas, which capture the energy price risk, the exposure is not clear-cut, as both the significance levels and the sign direction vary largely. Gas price return only seem to significantly influence the stock return of EDP Energias de Portugal, showing a negative sign. For firms individually, oil price returns seem to have the most significant influence, but the signs vary.

The main interest of this section goes to the estimated parameter of EUA, showing the exposure of firms to carbon risk, due to carbon allowance price movements. A total of five firms in the dataset show a significant exposure to carbon risk. This shows on the one hand that for most of the energy-intensive companies, EUA price movements are not a relevant risk factor. On the other hand, that carbon risk due to EUA price movements is asymmetrically distributed to only a few companies. Interestingly enough, none of the HCI firms exhibit significant exposure to EUA price movements. Indicating that investors do not consider EUA prices a relevant risk factor for these firms. Of the five firms which exhibit exposure to carbon allowance price changes, two firms are in the LCI portfolio, while three firms are in the Medium Carbon Intensity (MCI) portfolio. Stock returns of the two firms in the LCI portfolio both exhibit a significant and positive sensitivity to the EUA price movements. Namely, Vallourec and Acerinox ‘R’. This indicates that investors demand higher returns of these firms if the price of carbon allowances increases. So, for these firms, investors demand carbon risk stock premiums. A2A, a firm in the MCI portfolio, also has a significant and positive sensitivity to EUA price movements. Two firms in the MCI portfolio, namely Edison RSP and Thyssenkrupp, have a significantly negative EUA beta coefficient. These significantly negative beta coefficient can be interpreted as evidence that investors demand lower returns of these firms for increasing carbon allowance prices. It must be noted that the signs of the EUA coefficient vary substantially between firms. Also between firms in the same carbon intensity portfolio. Further inspection of the data shows that for all firms, the significant EUA coefficients is not robust for the entire sample period. For A2A, Thyssenkrupp and Vallourec, the EUA coefficient is insignificant for the period 2009-2012 (EU ETS phase 2). While for Edison RSP and Acerinox ‘R’, the EUA coefficient is insignificant for the period 2013-2016 (EU ETS phase 3).12

(19)

18 Table 4

Regression results of econometric model (4). The estimated parameter values (betas) of the model are reported. Their standard errors are given in parentheses. The table reports the firm-specific regression results of the entire dataset. Market and EUA are the estimated parameters which capture the carbon risk and market risk. Coal, Oil and Gas are the estimated parameters, capturing the energy risk. The adjusted R2 _{and F-test values are also shown in the table. The} sample uses monthly data from February 2009 till December 2016. *, ** and *** denote the significance levels at 10%, 5% and 1% respectively.

Dependent Variable Market EUA Coal Oil Gas Adj-R2 _{F test}

ENEL 0.88*** -0.01 0.05 0.18** 0.07 0.29 0.00*** (0.16) (0.09) (0.11) (0.08) (0.06) ENDESA 0.91*** 0.04 0.06 -0.17 -0.06 0.28 0.00*** (0.17) (0.08) (0.14) (0.11) (0.08) FORTUM 0.91*** 0.04 0.09 -0.02 -0.04 0.36 0.00*** (0.14) (0.06) (0.12) (0.14) (0.07) DRAX GROUP 0.59*** -0.01 -0.02 0.19* 0.09 0.33 0.00*** (0.12) (0.04) (0.11) (0.11) (0.06) PUBLIC POWER 0.34*** -0.03 0.04 -0.08 0.07 0.06 0.02*** (0.11) (0.05) (0.09) (0.12) (0.07) ARCELORMITTAL 0.17 0.03 0.08 -0.06 -0.02 0.03 0.47 (0.15) (0.07) (0.11) (0.07) (0.07) ERAMET 0.68*** 0.01 0.00 -0.06 0.00 0.36 0.00*** (0.11) (0.04) (0.10) (0.08) (0.05)

Medium Carbon Intensity firms

GAS NATURAL SDG 0.60*** 0.01 0.01 -0.13 -0.04 0.21 0.00*** (0.14) (0.05) (0.10) (0.08) (0.07) IBERDROLA 0.80*** 0.01 0.06 -0.13* 0.04 0.38 0.00*** (0.11) (0.04) (0.12) (0.08) (0.08) SSE 0.77*** -0.01 -0.01 -0.06 0.02 0.39 0.00*** (0.10) (0.06) (0.09) (0.06) (0.06) VERBUND 1.13*** 0.15 0.40** 0.08 -0.18 0.41 0.00*** (0.19) (0.09) (0.20) (0.17) (0.12)

EDP ENERGIAS DE PORTUGAL 1.19*** -0.02 0.44*** 0.13 -0.17* 0.48 0.00*** (0.16) (0.07) (0.16) (0.13) (0.09) A2A 0.57*** 0.13*** 0.14 -0.14* 0.04 0.42 0.00*** (0.10) (0.04) (0.09) (0.07) (0.05) EDISON RSP 0.85*** -0.10** 0.04 -0.16* 0.11 0.40 0.00*** (0.14) (0.05) (0.10) (0.08) (0.09) THYSSENKRUPP 0.40** -0.16* -0.01 0.37** -0.05 0.21 0.00*** (0.18) (0.08) (0.18) (0.15) (0.09) VOESTALPINE 0.90*** 0.04 0.13 -0.22*** 0.01 0.49 0.00*** (0.11) (0.05) (0.09) (0.06) (0.06) SALZGITTER 1.28*** 0.05 0.00 -0.03 0.05 0.38 0.00*** (0.22) (0.09) (0.19) (0.12) (0.14) KLOECKNER & CO 1.37*** 0.01 -0.01 0.03 -0.01 0.24 0.01*** (0.40) (0.12) (0.23) (0.17) (0.14) TENARIS 1.00*** 0.07 0.09 0.09 -0.04 0.42 0.00*** (0.16) (0.07) (0.15) (0.11) (0.10)

(20)

19 Furthermore, the equally weighted carbon intensity portfolios are evaluated. The results are shown in Table 5. All carbon intensity portfolios are most significantly exposed to coal prices, with a positive sign direction. Closer inspection of the EUA beta coefficients of the Carbon Intensity portfolios show that all EUA beta coefficient are insignificant. Indicating that the carbon intensity of a firm has no influence on the decisions of investors, regarding a firm’s exposure to carbon risk, due to carbon allowance price movements. Thereby, the signs of the EUA coefficient are positive for the LCI portfolio, MCI portfolio and all firms. The sign of the EUA coefficient for the HCI portfolio is neither positive nor negative. The above results suggest that in the period of 2009-2016, firms in the energy-intensive metal and power-generating industries, are not significantly exposed to carbon risk through EUA price movements. Thereby, no significant results were found when analyzing only the period of EU ETS phase 2 (2009-2012) or EU ETS phase 3 (2013-2016)13_{. For evidence supporting hypothesis 2a, a positive and significant EUA coefficient} should be found for firms in the HCI portfolio. The EUA coefficient is both insignificant and neutral. Consequently, for the analysis on EUA price movements, the data does not find support for hypothesis 2a. For evidence supporting hypothesis 2b, an insignificant EUA coefficient should be found for firms in the LCI and MCI portfolio. Although 5 of the 18 firms in the MCI and LCI portfolio were significantly exposed to EUA price movements, the results were not robust for the entire sample period. Thereby, when analyzing the equally weighted portfolios of both portfolios, insignificant EUA coefficients are found. Therefore, it can be concluded that overall, stock returns of Medium Carbon Intensity firms and Low Carbon Intensity firms are not exposed to EUA price movements. The data thus finds support for hypothesis 2b.

Table 5

Regression results of econometric model (4). The estimated parameter values (betas) of the model are reported. Their standard errors are given in parentheses. The table reports regression results of both the equally weighted carbon intensity portfolios and all firms. Market and EUA are the estimated parameters which capture the carbon risk and market risk. Coal, Oil and Gas are the estimated parameters, capturing the energy risk. The adjusted R2 _{and F-test} values are also shown in the table. The sample uses monthly data from February 2009 till December 2016. *, ** and *** denote the significance levels at 10%, 5% and 1% respectively.

Dependent Variable Market EUA Coal Oil Gas Adj-R2 _{F test}

High Carbon Intensity 0.71*** 0.00 0.09* -0.03 -0.01 0.63 0.00*** (0.12) (0.01) (0.05) (0.05) (0.03)

Medium Carbon Intensity 0.88*** 0.02 0.08* -0.03 0.00 0.73 0.00*** (0.09) (0.03) (0.04) (0.05) (0.02)

Low Carbon Intensity 0.88*** 0.05 0.06* 0.08 0.03 0.54 0.00*** (0.16) (0.04) (0.03) (0.08) (0.02)

All firms 0.82*** 0.02 0.08*** 0.00 0.00 0.68 0.00***

(0.07) (0.01) (0.02) (0.03) (0.01)

4.4 Robustness tests

Finally, additional factors which could have affected the outcomes, are taken into consideration. This section is an extension of the empirical analysis to assess the robustness of the main findings in this paper.

Firstly, the difference in the stock performance between the HCI portfolio and LCI portfolio crucially depends on the pooling of firms into the different Carbon Intensity portfolios.

(21)

20 In this paper, firms with an average carbon intensity of 1.5 or higher for the entire period of 2009-2016, are classified as High Carbon Intensity firms. Firms with an average carbon intensity of 0.5 or lower for the entire period of 2009-2016, are classified as Low Carbon Intensity firms. The firms in between which do not match these criteria are classified as Medium Carbon Intensity firms. After analyzing the Carbon Intensity data, the composition of the different Carbon Intensity portfolios is adjusted for two separate robustness checks. In the first robustness check, for firms to be included in the HCI portfolio, their average Carbon Intensity over the sample period has to be higher than 3 (Robust 1). For the second robustness check, firms are pooled into two separate portfolios; the High Carbon Intensity portfolio and the Low Carbon Intensity portfolio. Firms with an average carbon intensity of 1.0 or higher for the period of 2009-2016, are classified as High Carbon Intensity firms. Firms with an average carbon intensity which is lower than 1.0 for the period of 2009-2016, are classified as Low Carbon Intensity firms (Robust 2). Appendix A reports the two new classifications of firms, according to the alternative criteria for inclusion of firms in the different carbon intensity portfolios.

Both the High-minus-Low CAPM-α regression results and the rolling estimates for these new classifications are analyzed. The results for the High-minus-Low CAPM-α regression results can be found in Appendix B. Firstly, the regression results are evaluated. The CAPM-α results of both robustness checks are similar to the results measured in the original analysis. For both robustness test, when looking at the EU ETS second phase (2009-2012), all Carbon Intensity portfolios are significantly underperforming, relative to the market. After evaluating the CAPM-α in the EU ETS third phase (2013-2016), it can be concluded that none of the Carbon Intensity portfolios are significantly over- or under-performing the market. Thereby, the CAPM-α of the High-minus-Low portfolio is not statistically significant in any of the periods for both robustness checks. It is only worth noting that the sign of the High-minus-Low CAPM-α is positive for the second robustness check. But it is highly insignificant (p>0.884). For the analysis of the timing of the difference in stock performance between the HCI portfolio and LCI portfolio, rolling estimates of the two robustness checks are displayed in Appendix C. As can be seen in the figures, the rolling estimates of the High-minus-Low CAPM-α are quite similar to the original Figure 1. This is further evidence that the results of difference in stock performance between the HCI portfolio and LCI portfolio remain qualitatively the same. Therefore, it can be concluded that the criteria in this paper for pooling the firms, does not have a substantial effect on the final results.

(22)

21

5. Conclusion

In this paper, an empirical analysis of the stock return of carbon emitting firms under the EU ETS is presented. Firms from two different energy-intensive industries are included in the dataset; the Metals industry and the Power-generating industry. All firms in the dataset operate in the European Union. To be able to investigate the importance of the intensity of CO2 a firm emits, a carbon intensity measure is created by dividing the total carbon emission of a firm by the total amount of sales. Firms are then divided into three different portfolios; the Low Carbon Intensity (LCI) portfolio, the Medium Carbon Intensity (MCI) portfolio and the High Carbon Intensity (HCI) portfolio. Firstly, the difference in stock performance between High Carbon Intensity firms and Low Carbon Intensity firms is estimated, using the CAPM alpha. The difference in the alpha of the two portfolios measures the abnormal excess return of the HCI portfolio over and above that of the LCI portfolio. Secondly, an analysis of the firm-specific exposure of EUA (European Union Allowance) price movements on the stock return is undertaken.

Regression results show that the stock performance of firms in the HCI portfolio is not significantly lower than the stock performance of firms in the LCI portfolio. These insignificant results are found for the full sample period (2009-2016), EU ETS phase two (2009-2012) and EU ETS phase three (2013-2016). Thereby, the timing of the difference in stock performance between the HCI and LCI portfolios is determined. It is interesting to note is that just before 2012, there is a steep increase in stock performance of firms in the HCI portfolio relative to the stock performance of firms in the LCI portfolio, reaching its peak one or two months into 2013. For a small period, starting in 2013 and ending just before 2014, even a weakly significant better stock performance is measured for firms in the HCI portfolio relative to firms in the LCI portfolio. This coincides with the first year of the EU ETS third phase. Indicating that for this period, there was a carbon premium in stock performance for firms with a higher carbon intensity. This is an interesting finding, as it was expected that the need to buy carbon allowances in auctions, after to the introduction of phase three, would result in a significantly lower stock performance for HCI firms. Instead, in comparison to LCI firms, HCI firms even experienced a carbon premium in stock returns for a brief period of time. This could indicate that HCI firms are still overcompensated, by receiving too many carbon allowances for free.

Further results indicate that the stock return of only a few firms in the dataset are exposed to EUA price movements. The exposure of EUA price movements on stock returns is asymmetrically distributed to a total of five out of twenty-five firms; two firms in the LCI portfolio and three firms in the MCI portfolio. But after testing for robustness, this carbon risk is insignificant over the entire sample period. None of the stock returns of firms in the HCI portfolio show significant exposure to the EUA price movements. It can thus be concluded that for a great majority of firms in this dataset, investors do not see EUA price movements as a relevant carbon risk factor for stock returns. Furthermore, after evaluating the equally weighted carbon intensity portfolios, it can be concluded that the exposure of EUA price movements on stock returns is unanimously insignificant for all carbon intensity portfolios.

(23)

22 industries. Adding firms from two different industries could thus have altered the results, as firms in the Power and Metal Industry were not evenly distributed among the different Carbon Intensity portfolios. Overall, relatively more firms of the Power Industry were added to the High Carbon Intensity portfolio and relatively more firms of the Metal Industry were added to the Low Carbon Intensity portfolio. Furthermore, there is an ongoing discussion about the use of the CAPM-model to describe the behavior of financial markets. Some academics question the validity of the CAPM-model (e.g. Dempsey, 2013) while others argue that the CAPM is still the best asset pricing CAPM-model available (e.g. Smith and Walsh, 2012; Da, Guo and Jagannathan, 2012). But, it is the most widely used model by financial officers (Graham and Harvey, 2001). In this paper, the goodness of fit of the CAPM for the entire sample period, calculated with the adjusted R2, is not particularly high. Extensions to the model might have given different results, like the multifactor asset pricing models proposed by Fama and French (1992) or Ross (1976).

Nevertheless, I am confident that the results in this paper are interesting. To conclude; under the EU ETS, the stock performance of firms with a lower carbon intensity is not significantly different from firms with a higher carbon intensity. Moreover, the stock returns of the majority of carbon-emitting firms are not significantly exposed to carbon allowance price movements. These results indicate that firms under the EU ETS do not seem to be ‘punished’ for having a relatively high carbon intensity. And that the carbon allowances introduced by the EU ETS are not considered a relevant risk factor for CO2 emitting firms. The insignificant results might be attributed to the fact that still, a lot of the carbon allowances are distributed for free (Goulder, Hafstead and Dworsky, 2010). Future research on carbon emitting cap-and-trade systems with full auctioning might shed more light on this matter. Thereby, future research which covers the whole EU ETS period should be done to verify the results of this paper and previous papers, as the EU ETS ranges till 2020. Hopefully, the main findings in this paper can contribute to future research and provide more insight about the implications of the EU ETS. It could contribute to the environmental policy debate, as it informs governments and investors about the possible future design of similar cap-and-trade systems, which can be implemented to reduce overall carbon emissions.

6. Reference List

Abrell, J., Ndoye, A., and Zachmann, G., 2011. Assessing the impact of the EU ETS using firm level data. Bruegel Working Paper 2011/08, Brussels, Belgium.

Alberola, E., Chevallier, J. and Chèze, B., 2008. Price drivers and structural breaks in European carbon prices 2005–2007. Energy Policy, 36, 787-797.

Anger, N. and Oberndorfer, U., 2008. Firm performance and employment in the EU emissions trading scheme: An empirical assessment for Germany. Energy Policy, 36, 12-22.

(24)

23 Benz, E. and Trück, S., 2009. Modeling the price dynamics of CO2 emission allowances. Energy Economics, 31, 4-15.

Brinkman, M. W., Hoffman, N., and Oppenheim, J. M., 2008. How climate change could affect corporate valuations. McKinsey Quarterly, 29, 1-7.

Busch, T. and Hoffmann, V., 2007. Emerging carbon constraints for corporate risk management. Ecological Economics, 62, 518-528.

Busch, T. and Hoffmann, V., 2011. How Hot Is Your Bottom Line? Linking Carbon and Financial Performance. Business & Society, 50, 233-265.

Chapple, L., Clarkson, P. and Gold, D., 2013. The Cost of Carbon: Capital Market Effects of the Proposed Emission Trading Scheme (ETS). Abacus, 49, 1-33.

Clarkson, P., Li, Y., Pinnuck, M. and Richardson, G., 2014. The Valuation Relevance of Greenhouse Gas Emissions under the European Union Carbon Emissions Trading Scheme. European Accounting Review, 24, 551-580.

Da, Z., Guo, R. and Jagannathan, R., 2012. CAPM for estimating the cost of equity capital: Interpreting the empirical evidence. Journal of Financial Economics, 103, 204-220.

Delarue, E., Voorspools, K. and D’haeseleer, W., 2008. Fuel Switching in the Electricity Sector under the EU ETS: Review and Prospective. Journal of Energy Engineering, 134, 40-46.

Dempsey, M., 2013. The Capital Asset Pricing Model (CAPM): The History of a Failed Revolutionary Idea in Finance? Abacus, 49, 7-23.

Ellerman, A., Convery, F., De Perthuis, C., Alberola, E., Buchner, B., Delbosc, A., Hight, C., Keppler, J. and Matthes, F., 2010. Pricing Carbon: The European Union Emissions Trading Scheme. Cambridge University Press, Cambridge.

Fama, E. and French, K., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33, 3-56.

Fama, E. and MacBeth, J., 1973. Risk, Return, and Equilibrium: Empirical Tests. Journal of Political Economy, 81, 607-636.

Goulder, L., Hafstead, M. and Dworsky, M., 2010. Impacts of alternative emissions allowance allocation methods under a federal cap-and-trade program. Journal of Environmental Economics and Management, 60, 161-181.

Graham, J. and Harvey, C., 2001. The theory and practice of corporate finance: evidence from the field. Journal of Financial Economics, 60, 187-243.

Grainger, C. and Kolstad, C., 2010. Who Pays a Price on Carbon? Environmental and Resource Economics, 46, 359-376.

(25)

24 Hahn, R.W. and Stavins, R.N., 2011. The effect of allowance allocations on cap-and-trade system performance. Journal of Law and Economics. 54, 267–294

Intergovernmental Panel on Climate Change, 2014. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. New York Cambridge University Press, New York.

Jensen, M., 1968. The Performance of Mutual Funds in the Period 1945-1964. The Journal of Finance, 23, 389-416

Koch, N. and Bassen, A., 2013. Valuing the carbon exposure of European utilities. The role of fuel mix, permit allocation and replacement investments. Energy Economics, 36, 431-443.

Konar, S. and Cohen, M., 2001. Does the Market Value Environmental Performance? Review of Economics and Statistics, 83, 281-289.

Matsumura, E., Prakash, R. and Vera-Muñoz, S., 2014. Firm-Value Effects of Carbon Emissions and Carbon Disclosures. The Accounting Review, 89, 695-724.

McGahan, A. and Porter, M., 1997. How Much Does Industry Matter, Really? Strategic Management Journal, 18, 15-30.

Newey, W. and West, K., 1987. A Simple, Positive Semi-Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix. Econometrica, 55, 703-708.

Oberndorfer, U., 2009. Energy prices, volatility, and the stock market: Evidence from the Eurozone. Energy Policy, 37, 5787-5795.

Oestreich, A. and Tsiakas, I., 2015. Carbon emissions and stock returns: Evidence from the EU Emissions Trading Scheme. Journal of Banking & Finance, 58, 294-308.

Ritter, A. and Wells, P., 2006. Identifiable intangible asset disclosures, stock prices and future earnings. Accounting and Finance, 46, 843-863.

Ross, S., 1976. The arbitrage theory of capital asset pricing. Journal of Economic Theory, 13, 341-360.

Sijm, J., Neuhoff, K. and Chen, Y., 2006. CO2 cost pass-through and windfall profits in the power sector. Climate Policy, 6, 49-72.

Smith, T. and Walsh, K., 2012. Why the CAPM is Half-Right and Everything Else is Wrong. Abacus, 49, 73-78.

(26)

25

7. Appendices

Appendix A: Pooling of firms for Robustness

Table 6

Classification of the firms under the EU ETS, according to Robust 1 and Robust 2. The table reports the Carbon Intensities of all firms, calculated according to formula (1). Both the Carbon Intensity of 2015 and the mean of the carbon intensity in the entire period of 2009-2016 are shown. 2015 is the most recent year where both the carbon emissions and the Net Sales are known for all companies. The sample is segmented into three groups, based on the firm’s Carbon Intensity; High, Low, and Medium Carbon Intensity firms. Both the country where the firm is based and the sector in which the company operates are given for all companies. For robust 1, only firms with a CI>3 are added to the HCI portfolio (upper panel). For robust 2, all firms are pooled into only two portfolios; HCI and LCI (lower panel).

Table 6a: robust 1

(2015) (2009-2016)

Average Carbon Intensity firms

(27)

26

Table 6b: robust 2

(2015) (2009-2016)