Carbon prices and firms' financial performance: an industry perspective

Abstract

This paper investigates how carbon prices influence the financial market value of the individual firm after Phase I of the EU's Emission Trading Scheme (ETS). The dataset covers 136 firms in the industries that are responsible for the majority of the greenhouse gas emission, namely the oil and gas, power and heat, cement and lime, and iron and steel industry. The paper basically follows the method and approach applied in Oberndorfer [27]. The results show there is a positive and significant effect of carbon price changes on stock market returns in all four industries. Furthermore, there is evidence for an asymmetric influence in all sectors apart from the oil & gas industry. Volatility of the market value of firms appears not be influenced by the volatility of the EU ETS carbon prices. The results appear to be robust against different specifications for the estimation of the variance. However, they are sensitive to different time periods, i.e., distinguishing between 2008-2009 and 2010-2011. We conclude that despite several inefficiencies, the EU ETS has a significant impact on the value of firms that are responsible for most of the carbon emissions in the EU.

JEL Codes:

• C12
• C13
• C22
• G12
• Q43
• Q48

Keywords:

• EU ETS
• stock market returns
• asymmetry
• volatility
• industries

Introduction

The European Union's Emission Trading Scheme (ETS) was put into effect in 2005 as a tool for reducing corporations' greenhouse gas-emissions and, as such, to reduce their external effects. It is the cornerstone of EU's policy to deal with climate change. Ellerman and Buchner [13,14] and Creti, Jouvet and Mignon [10] provide an institutional and financial market perspective of the ETS as well as an account of the design and background of the ETS. However, it is not clear what is the impact of the ETS on the value of individual firms and industries. The main idea behind the ETS is to account for external economic effects. Basically, the ETS can impact the value of the firms in two ways. First is via the direct cost of complying: the firm will have to adapt production processes and/or buy emission allowances in the market. As such, this impacts on the cash flows of the firm and thereby on its value. Second is via the risk channel. This relates to the price changes and to price volatility of the allowances which impacts on the potential changes in the value of the firm. Price volatility exposes the participants to carbon price risk. In this respect, the carbon price risk may be compared to corporate risk due to oil price changes or exchange rate volatility (see for example, [16,19,30]). This suggests that the price of factor inputs could be affected by the ETS and as such factor prices would more properly reflect relative scarcities. By pricing carbon emissions, the use of carbon emitting fuels would decrease and this was expected to lower the value of energy-intense companies. This was one of the reasons they were being compensated in various ways. However, pricing carbon emissions not necessarily needs to result in less profits for firms using carbon-based fuels as in the end this depends upon their overall costs and revenues. For example, they may invest in new technologies that improve energy efficiency. Then, the impact can be positive as well. Hence, in the end, it is an empirical issue whether and how the ETS affects firms' value. So far, this mainly has been studied for the electricity producers (see [27]). However, as energy is an input to many industrial processes, carbon prices and changes therein can impact the value of other industries as well. Löfgren, Wråke, Hagberg and Roth [22] find that the ETS does not influence investments and innovation in Sweden. Smale, Hartley, Hepburn, Ward and Grub [35] investigate the impact of the ETS for different British industries. They find that most participating sectors would be expected to profit, but there is a modest loss of market share in the case of steel and cement, and closure in the case of aluminium.

The key question we address in this paper is how carbon prices and carbon price volatility impact the financial market value of the individual firm that is eligible regarding the ETS after Phase I. We do so by closely following the study of Oberndorfer [27], but will not only investigate the electricity industry. To be specific, we investigate 136 European firms in the industries that are held responsible for most of the greenhouse gas emissions in the EU and that have to comply with the ETS, namely the power and heat, oil and gas, cement and line, and the iron and steel industry. According to the IEA [18], these four industries make up about four fifths of the greenhouse gas emissions that is eligible under the EU ETS. The focus of this study is the industry response in the period 2008-2011. The main hypothesis to be tested is that carbon prices significantly impact the stock market returns. Furthermore, we will analyse the asymmetry in the relationship between carbon prices and stock market returns as this is a feature which characterizes many industries regarding their response to oil price changes [32]. That is we ask whether the response differs between carbon price increases and decreases. In addition, we will want to analyse the effect of carbon price volatility on the volatility of the stocks. As such, we can provide an overall perspective about how the EU ETS impacts the value of different industries. To stay in line with the approach taken by Oberndorfer [27], we basically will estimate our models with the help of (pooled) ordinary least squares regressions and GARCH. The results will be of interest to academics, practioners, and policy makers in order to find out if and how the impact of the ETS differs between industries. In all, we are able to show that the EU ETS has in many respects a significant influence after Phase I of the EU ETS on the market value of corporations. This is in line with findings elsewhere in the literature (e.g. [26]). Compared to the Oberndorfer [27] results, it is interesting to find that the EU ETS has a significant impact outside the power industry too. In contrast to his study, we find that after 2007 there no longer seem to be country specifics involved.

The remainder of the paper is organized as follows. The next section discusses the main literature and develops the hypotheses to be tested in this paper. The data and methodology will be introduced in section 3. The results are in section 4. Section 5 concludes and discusses policy implications.

Development of the hypotheses

The relationship between carbon emissions and corporate behavior is widely studied in the academic literature. Kossoy and Guigon [21] provide an overview of global developments regarding greenhouse gas emissions and carbon exchanges. Mizrach and Otsubo [25] providing a thorough review of the microstructure of the European carbon markets. Several studies investigate the association between carbon markets and firm performance, examples are Demailly and Quirion [11,12], Busch and Hoffmann [8], Chevallier [9], Oberndorfer [27], Veith, Werner and Zimmerman [36], Smale et al., [35], Creti et al. [10], Aatola et al. [1], Koch and Bassen [20], Petrick and Wagner [26], and Martin, Muûls and Wagner [24].

Our study will basically follow the approach suggested by Oberndorfer [27] who investigates the impact of the EU ETS in Phase I on the electricity industry as this was the first and main study to look into firm specifics. Oberndorfer [27] analyzes electricity stock return reactions of twelve companies to changes in EU Emission Allowance prices. He takes into account the potential for differences in the relationship over time as well as between corporations. This is relevant to corporations operating in different countries that are marked by differences in allowance allocation (and therefore in possibly different initial long/short positions in EUAs) according to the different National Allocation Plans or in the market structure or institutional design of the electricity market. This could affect cost pass-through behavior. An issue not investigated by Oberndorfer [27] is the allowance distribution. To this extent, Veith et al. [36] and Koch and Bossen [20] explicitly investigate the impact of the (free) allocation of the emission allowances. Both studies suggest that the (potential) costs of the allowances have been overcompensated by the power companies as they did receive the allowances free of charge and did require an add-up on the customer prices, which did result in windfall profits. Oberndorfer [27] also investigates whether the relationship between emission allowance price changes and electricity companies' stock returns is asymmetric, which would be consistent with the effect of emission allowances on wholesale electricity prices. In addition, he applies a GARCH approach to test whether EUA return volatility and European electricity stock volatility are related. It shows that carbon price changes do significantly impact on the stock returns of the electricity industry in Phase I of the EU ETS.

Given the increasing importance of climate change, we want to investigate whether this relationship also holds after Phase I and, if so, whether it holds for other industries too. In this respect, it is important to note that Jones and Kaul [19] find such a relationship for oil prices and international stock returns. However, Huang et al. [16] only detect a significant relationship between oil futures and oil stock returns. Sadorsky [30] analyzes the characteristics of oil prices and oil price volatility and concludes that both play a role regarding stock prices. Chevallier [9] tests the empirical relationship between the returns on carbon futures and changes in macroeconomic conditions. These conditions are macroeconomic risk factors and Chevallier [9] finds that carbon futures are only remotely connected to macroeconomic risk factors, which means there is a weak link between the stock and bond market variables and carbon futures returns. His study underlines the view that carbon allowances are a very specific market among energy commodities. Veith et al. [36] investigate evidence from the capital market to assess the impact of the EU ETS in Phase I. They rely on a multifactor model and show that returns on common stock of power generation companies are positively correlated with rising prices for emission rights. Like Oberndorfer [27], they do not investigate other industries. Veith et al. [36] establish that firms in the power industry seem to have been able to pass on their share of the regulatory burden to customers. Moreover, they show that this industry has been able to achieve windfall profits by overcompensating for the costs. Scholtens and Yurtsever [32] investigate the industry specifics in the relationship between oil prices and stock returns for a large number of industries and show that there are substantial differences between the industries. They investigate the impact of oil price shocks at the industry level in the Euro area for 1983–2007. They rely on different oil price specifications and use dynamic VAR models and multivariate regression to investigate how 38 different industries respond to oil price shocks. They specifically pay attention to the asymmetry of the industries' responses regarding oil price increases and decreases. Overall, they conclude that the impact of oil price shocks substantially differs along different industries. The significance of this result appears to differ along the different oil price specifications. As to the asymmetric response to price increases and price decreases, Scholtens and Yurtsever [32] find that this differs between industries as well as between the various time periods analyzed. Benz and Trück [5] specify emission allowances as a production factor of the firm. As such, they assume that carbon price changes directly affect the value of the emission allowances of the firm and, thereby, the value of the firm. We hypothesize that carbon prices significantly impact the stock return in the industries:

H1: Carbon price changes have a significant effect on stock returns

On top of this, we are especially interested in how different industries might perform in the light of changes in carbon prices. Ponssard and Walker [29] and Arjaliès and Ponssard [4] argue that responses could be industry specific, but they do not provide empirical evidence. Smale et al [35] model the impact of the ETS regarding five energy-intensive sectors (cement, newsprint, steel, aluminium, petroleum) in the UK for different price scenarios. They find that most participating sectors would be expected to profit. However, they expect a modest loss of market share in the case of steel and cement, and closure in the case of aluminium. We want to find out whether this finding still holds for a much larger sample. But we do not provide specific hypotheses regarding which industry would be more sensitive to carbon price changes than another as there is no specific theory about how particular industries would respond.

Another issue is whether the value of firms responds in the same way to increases in carbon prices as it does to price reductions. This feature is established by Scholtens and Yurtsever [32] for industry responses to oil price changes. Huang et al. [16] and Sadorsky [30,31] use a vector autoregressive (VAR) approach to investigate the symmetry of the relationship between oil price changes and stock returns. They find that fluctuations of oil prices change over time. According to Sadorsky [30]: “after 1987, oil price movements explain a larger fraction of the forecast error variance in real stock returns than interest rates do.” Sadorsky [30] also finds that oil price shocks have asymmetric effects on the economy. Zachman and Von Hirschhausen [37] establish an asymmetric reaction between carbon price changes and German electricity prices. When emission allowance prices rise, they tend to be passed through to the electricity market (see [36]), but this appears to have not been the case with falling allowance prices. However, Oberndorfer [27] cannot establish such an asymmetric relationship between carbon prices and electricity stock prices for energy producers. We want to analyze the asymmetry in the relationship between carbon prices and stock returns after Phase I and investigate if there is an asymmetric reaction of the stock returns in various industries on carbon price changes. As such, our second hypothesis is:

H2: The relationship between carbon price changes on stock returns in the specific industry is asymmetric.

Sadorsky [30] found that not only oil prices themselves, but also oil price volatility affects the (volatility of) stock prices. In addition, Oberndorfer [27] argues that not only carbon prices may matter for the development of stock returns, but that carbon price volatility will impact on stock price (volatility) as well. He suggests that carbon volatility fluctuations may impact the expectations for future cash flows of companies covered by the scheme. Oberndorfer [27] also mentions that the attractiveness of a stock for potential investors highly depends on the price volatility of the respective stock. However, he does not arrive at significant empirical support in this respect for the electricity industry in Phase I of the ETS. He suggests that multivariate GARCH model could be used to further investigate this issue. Therefore, we will analyze the effect of carbon price volatility on the volatility of the stocks with help of this type of GARCH models. We test for a relationship between the volatilities of the EUA price and the stock returns of the companies in the four industries. Thus, our third hypothesis is:

H3: There is a significant relationship between carbon price volatility and stock market return volatility

Because of the fact that the EU ETS is operating in 30 countries that widely differ in wealth and economic structure, there may be country-specific effects of carbon price changes on the stock returns of companies in the different industries. Oberndorfer [27] suggests that the carbon market effect is both time- and country specific in Phase I. He argues there is a possibility that the amplitude of the EUA effect itself depends mainly on country-specific characteristics such as differences in EUA long/short positions due to country-specific National Allocation Plans and the structure of the national sector markets in which the companies operate. Oberndorfer [27] used a pooled regression framework in order to test for country-specific effects in the electricity industry. Therefore, we will also use the same methodology and focus on various industries after Phase I. To account for time specifics, we will also compare the first two years of the sample period with the last two years.

Data and Methods

We investigate the relationship between carbon prices and stock returns of 136 European companies in four different industries, namely Oil & Gas, Power & Heat, Cement & Lime, and Iron & Steel. The first two are closely connected with the electricity market investigated by Oberndorfer [27] and Veith, Werner and Zimmermann [36]. Furthermore, we also investigate the effect of the carbon price changes on the stock returns of companies in the Cement & Lime and the Iron & Steel industry. Both these industries are characterized by a substantial use of energy and are about as large as the traditional energy industries regarding carbon emissions [18]. We test the relationship between carbon price changes and companies' stock market returns in the four industries by using an equally weighted portfolio of the companies. Given the industries we investigate, we think that the standard objections against using equally weighted portfolios are not applicable. For example, there are no small companies in our sample as the industries are very capital intense. This implies that small companies will not have a disproportional impact on the results (see [17]). Furthermore, these stock market returns will be analysed in disaggregated form within the framework of a panel approach. This allows for identifying firm-specific EUA effects. In order to avoid any misspecification of the economic approach, the market return as well as price changes in oil, gas, iron, cement and electricity price changes are used as control variables.

We will first run conventional Ordinary Least Squares regressions in order to test for the relationship between the stock returns and the change of carbon prices as estimated by the Capital Asset Pricing Model (CAPM) [34]. With the CAPM, a distinction can be made between systematic risk and firm-specific risk. In general, a relationship between the expected returns of certain assets and the so-called systematic risk can be found. Since the focus is on the relationship between carbon price changes and stock returns, it is obvious that the carbon price will be included in the regression. However, the response of the stock returns of the tested companies is influenced by other variables as well. For example, Mansanet-Bataller, Pardo and Valor [23] find that the prices of oil and gas are crucial variables in determining the carbon allowance prices. Therefore, we will also include gas and oil prices as explanatory variables. Zachman and Von Hirschhausen [37] show that if statistically significant explanatory price variables are excluded from the regression, this may influence the estimates of the effect of carbon prices on electricity stock returns. These biased results can lead to wrong inferences regarding statistically significant effects of the carbon prices. Thus, we will include several prices as explanatory variables. We will perform pooled regressions to account for disaggregate stock returns of all companies forming the portfolios of the different industries (see [7]).

Thus, we investigate the impact of carbon price changes in the four largest greenhouse gas emitting industries. The first two industries, Oil & Gas and Power & Heat, are closely connected with the electricity market investigated by Oberndorfer [27]. For these two, we use the following regression equations: (1) (2)

where r_O&G,t, r_P&H,t and r_m,t are respectively the returns for the industry portfolios of the Oil & Gas and the Power & Heat industry and the market portfolio at the end of period t. Furthermore, r_eua,t is the change of the EUA price and r_o,t, r_g,t and r_e,t r_e,t are the price  changes of oil, gas and electricity respectively. Besides these industries, we also want to investigate the effect of the carbon price changes on the stock returns of companies in the Cement & Lime and the Iron & Steel industry. Below we give the regression equations for both industries, where r_C&L,t and r_I&S,t are the returns for the industry portfolios of Cement & Lime and Iron & Steel respectively, furthermore r_c,t represents the change of the cement price and r_i,t the change of the iron price. (3) (4)

We will perform a pooled regression to account for disaggregate stock returns of all companies forming the portfolios of the different industries. A pooled regression is an effective way to deal with panel data. In a pooled regression a single equation is estimated on all the data together. It allows us to examine how variables and the relationships between these variables transform dynamically over time. In order to conduct a solid test of the hypothesis, it requires a long series of data to arrive at a sufficient number of observations. The equations for this pooled regression-method vary only with respect to the dependent variable from the equations used for the portfolio regressions: (5) (6) (7) (8)

where equations 5(5) –8(8) represent the following industries: Oil & Gas, Power & Heat, Cement & Lime and Iron & Steel respectively, and where r_O&G,t,, r_P&H,t,, r_C&L,t and r_I&S,t represent the pooled returns of these industries.

The GARCH (1,1) model is a well-known method to explain systematic variation of asset price volatility. Sadorsky [30], Alberola, Chevallier and Cheze [3], Chevallier [9], and Benz and Trück [5] also use such a model. With a GARCH (1,1) model, the systematic variation of asset price volatility can sufficiently be explained. It allows the conditional variance to be dependent upon previous lags. With a GARCH (1,1) model, in most of the cases it is possible to capture the volatility clustering in the data. Ordinary Least Square (OLS) is not suitable to estimate GARCH models since this type of model cannot be seen as a linear form model. For this reason another technique has to be used to estimate the model. In line with Oberndorfer [27], we will use the maximum likelihood technique.

In order to create a GARCH (1,1) model it is necessary to include the volatility variables of the explanatory variables of the regression into the framework. The volatility variables are represented as follows in the GARCH (1,1) equations υ_eua,t, υ_o,t, υ_g,t, υ_e,t, υ_c,t and υ_i,t, for the EUA price and the sectors Oil & Gas (eq. 9(9) and 10(10) ), Power & Heat (eq. 11(11) and 12(12) ), Cement & Lime (eq. 13(13) and 14(14) ) and Iron & Steel (eq. 15(15) and 16(16) ) respectively: (9) (10) (11) (12) (13) (14) (15) (16) (16)

In line with Oberndorfer [27], we assume a t-distribution for the zero mean error term ε_t. Furthermore α_i, β₁, β₂, β₃, β₄, β₅, a, b, c, d₁, d₂, d₃, d₄ and d₅ are the unknown parameters that are estimated by using the maximum likelihood technique.

We argued it is of interest to test if the relationships between the carbon price changes and stock returns of the different companies are asymmetric. Borenstein, Cameron and Gilbert [6] suggest that the asymmetric diffusion pattern of positive and negative cost shocks may be estimated using Error Correction Models (ECM). However, to make an accurate comparison between the results of our paper and those of Oberndorfer [27] and Zachman and Von Hirschhausen [37], our dataset will be tested for any asymmetric reaction with the use of dummy variables. We include a dummy variable in the regression that takes the value of zero for the EUA price decreases as well as in the case in which there is no price change. This variable will take the value of one in the case of a EUA price increase. As such, we can distinct between the reaction of the explanatory variables on both EUA price decreases as well as EUA price increases. As such, we can distinct between the reaction of the explanatory variables on both EUA price decreases as well as EUA price increases. Thus, the interaction variables (dummies) D_t take the following values:

0  → EUA price decrease (or price change is equal to 0)

1 → EUA price increase

In other words, to account for asymmetric reactions, we simply extend GARCH with an additional term added (dummy) to account for possible asymmetries. The adjusted regressions for the four industries are: (17) (18) (19) (20) (20)

In addition, Oberndorfer [27] argues that the amplitude of the EUA effect itself might depend on country-specific characteristics such as differences in EUA long/short positions due to country-specific allocation plans and the structure of the national sector markets in which the companies operate. He uses a pooled regression framework in order to test for any country-specific results. We will use the same methodology. However, the composition of the country indicators differs from the one Oberndorfer [27] uses. Since every industry has another set of companies from different countries, we will select alternative country indicators for each industry. In a sensitivity analysis, we will look into country effects for countries with more than one firm per industry; the country indicator for the variable ‘other’ consists out of countries in which only one single company is located. We use a pooled regression analysis to test for country specific effects. This pooled regression is extended with interaction term coefficients for the carbon price changes and the country indicators.

To wrap up, for our analysis of the relationship between carbon prices and financial performance after Phase I of the EU ETS, we will first use (pooled) OLS to investigate the impact of carbon price changes on the industry returns. Then, we use (pooled) OLS to investigate the role of asymmetry in carbon price changes. Last is the analysis of the impact of carbon price volatility with the help of a GARCH (1,1) model. For the sensitivity analysis, we look into stability over time and country effects.

As to the data used, we also try to stay close to those used in the paper by Oberndorfer [27] and will explain any deviations. Carbon data is derived from Carbon Market Data (source: www.carbonmarketdata.com), which is a carbon market research company offering information, consulting and technology services to a wide range of organizations in the world. The main industries, based on the amount of companies active in these industries and the amount of emissions covered by the EU ETS, are: Oil & Gas, Power & Heat, Cement & Lime and Iron & Steel. Table B₁ in the Appendix provides information about the origin of the companies in our sample regarding sector and country. The 30 countries in which the EU ETS operates are the 27 EU Member States, Iceland, Liechtenstein, and Norway. In these countries, the carbon emissions from many different types of companies are covered. Appendix B₂ gives the 17 industries along which the companies are divided. The four main industries combined consist of 629 companies. After checking which companies are listed and for which data is fully available 49 companies in Oil & Gas, 32 companies in Power & Heat, 15 companies in Cement & Lime, and 30 companies in Iron & Steel remain. This results in a total of 136 firms in the four industries. The (lognormalized) returns of the selected companies are collected from Thomson Reuters Datastream and are used as the dependent variable in the regressions explained in the methodology section of this paper.

Oberndorfer [27] uses the EUA settlement price from the European Energy Exchange. The European Energy Exchange AG (EEX) was founded in 2002 as a result of the merger of the two German power exchanges, Leipzig and Frankfurt. The EEX can be seen as the leading market in European energy trading. Other large EUA marketplaces are the NordPool, European Climate Exchange, and Powernext. Most authors use the EUA price as traded on the European Climate Exchange (for example [9,15,23,32]). Sadorsky [31] argues it is better to use future or forward prices compared to spot prices because future and forward prices are less noisy compared to prices on the spot market. Aatolo et al. [1] study part of Phase II of the EU ETS and they rely on the yearly forward EUA price. However, according to Oberndorfer [27] a characteristic of the EUA markets is that there is much less trade at the future compared to the spot market. More recently, the market has matured and now the majority of EUA transactions regards futures (Kossoy and Guignon, [21,25]). In order to stay in line with the studies after Phase I of the ETS of Alborela et al. [3], Alberola and Chevallier [2], Oberndorfer [27], and Zachman and Von Hirschhausen [37], will use the settlement EUA prices from the spot market.

Based on the literature discussed above, we use six different explanatory variables besides the EUA settlement price. These are the stock market in general, oil, gas, electricity, cement and iron prices. Since the 30 countries that are active in the EU ETS are all located in Europe, a European market index is used for the market return. We will use the Dow Jones Euro Stoxx, which is a well-traded index and a sound indicator that can be used to determine the market return. Mansanet-Bataller (2011) and Oberndorfer [27] also use this index to calculate stock market returns.

Oil price changes may impact on the market return of the different portfolios. This is why we also include the oil price change in the regressions as explanatory variables. We will use the Brent index to calculate the oil price changes in Europe. Brent crude oil is a major trading classification and is sourced from the North Sea. Oberndorfer [27] uses one-month forward prices, but for reasons of consistency with the other data, we will use the spot prices. Our use of spot prices also is in line with the approach taken by Alborela et al. [3], Alberola and Chevallier [2] and Zachmann and von Hirschhausen [37].

The natural gas price change is included for the same reason as the oil price change. Oberndorfer [27] uses the one-month forward natural gas time series from Intercontinental Exchange (ICE). Oberdorfer argues that he chose this index for natural gas because there was no EEX gas data available for his complete time period. He mentions several disadvantages regarding the use of the ICE gas price. As trading on the EEX started in 2007, the EEX data is fully available for the sample period, 2008-2012. Considering the disadvantages of the ICE data as mentioned by Oberndorfer [27] and the fact that the EEX can be seen as a leading trading market in European energy trading, we will use EEX gas prices.

Oberndorfer [27] mentions the problem of the absence of a common market for electricity in the EU. However, in many respects, the Phelix index traded on the EEX may be regarded as a reliable benchmark for the European electricity price. The Phelix Futures is a futures contract which refers to spot market prices for power of future delivery periods and covers the German electricity prices. Zachmann (2008) mentions that “despite the diminishing price differences on the European electricity market, there has still not been a price convergence achieved.” To stay in line with the approach of Oberndorfer [27], we will use the Phelix contract to determine the European electricity price changes.

Since we not only test the effect of the EUA price changes on the stock returns for Power & Heat and Oil & Gas, but also for the stock returns in Cement & Lime, it is necessary to take the possible side effects of the cement price into account. According to the data, the largest part of this sector is covered by Cement related companies. In order to avoid any biased estimates in this regression it is important to take the cement price change into account. We will use the Barcelona SE Cement price index which covers the largest part of the European cement market. For this reason, we use the Barcelona SE Cement price index to represent the European cement price change.

Similarly, in order to avoid any biased estimates with respect to the effect of the EUA price change on stock returns of companies in Iron & Steel, the iron and steel price is included as an explanatory variable in the regression. For this explanatory variable, the DJTM Euro Iron & Steel price index is used. This price index reflects the iron and steel prices in the Euro zone.

We investigate the first four years of the second phase of the EU ETS. This second phase started on January 1^st 2008 and we retrieved data until December 30^th 2011. Most of the existing literature on the effect of EUA prices on stock returns covers only the first phase of the EU ETS, which is a period of three years (2005-2007). We especially will compare our analysis with that of Oberndorfer [27] regarding the effect of the EUA prices on the stock returns.

Tables A₁ and A₂ in the appendix give the descriptive statistics for the explanatory variables used in this paper. Furthermore, a correlation matrix of the variables applied in the different regressions is provided in table A₃. There is a strong positive correlation between the four different dependent variables and the market returns, which is in line with Oberndorfer [27]. Furthermore, we find a positive relationship between the EUA price changes and the stock returns of the four industries. The explanatory variables show, except for the electricity price change, positive correlation coefficients. Both steel and cement price changes correlate strongly with the other explanatory variables.

Results

In this section, we first report the results for the impact of carbon price changes on industry returns at the industry level in Tables 1-4 for each of the four industries. These tables also hold the estimation results for the model which accounts for asymmetry. Then we go into the estimation results regarding the role of carbon price volatility. Lastly, we report the results of the sensitivity analysis with respect to subperiods and country effects.

Table 1. Results for basic- and pooled OLS and asymmetries for the Oil & Gas sector. Where the Oil & Gas portfolio return is the dependent variable for the normal OLS regression and β₁, β₂, β₃, β₄, β₅, δ are the coefficients of the explanatory variables.

Oil and Gas	OLS	Pooled OLS	Asymmetry OLS	Asymmetry Pooled OLS
C	-.02	.00	-.03*	.00
β1 (market)	3.06***	.62***	3.08***	.62***
β2 (EUA)	1.98***	.03***	1.57	.03***
β3 (oil)	6.09***	.14***	6.37***	.14***
β4 (gas)	.55	.01	.74	.01
β5 (electricity)	.15	.00	.13	.00
δ (asymmetric EUA)	–	–	1.55	.02
observations	1043	51,091	1043	51,091
R^2	.62	.07	.61	.07
F-statistic	313.76	827.74	309.87	824.66

*, ** and *** show significant values at, respectively, 10%, 5% and 1% significance-levels, respectively

Table 1 gives the estimation results for the basic- and pooled OLS regressions for companies in Oil & Gas (column 2 and 3). It shows that the market has a high and significantly positive effect on the equally weighted portfolio (OLS model) return as well as on the pooled sample of the Oil & Gas companies. Furthermore, the EUA price changes have a strong positive significant impact on the returns of the companies in the Oil & Gas industry. As such, this is a confirmation of Hypothesis 1. The size of the coefficients does differ a lot between the OLS and the pooled OLS. This suggests that firm specific attributes are quite important. However, both the signs of the coefficients and their significance are the same. The results strongly suggest that in the oil and gas industry, an increase in the carbon prices is to be associated with an increase in the industry portfolio return. This implies that this industry passes through carbon price changes, as has been found by Oberndorfer [27] and Veith et al. [36]. These results also are in line with the more general findings about factor costs in the energy industry by Sadorsky [29,30]. In line with most previous studies, oil prices also have a positive and significant impact on the returns of the Oil & Gas companies. Both the price changes of gas and electricity have a positive impact on the returns, but these results are insignificant. In contrast to Zachman and Von Hirschhausen [37], Table 1 shows that there is no significant asymmetric effect of the EUA price changes (columns 3 and 4). This can be detected from the coefficient δ, which does not show any significance. It is interesting that in the OLS estimation results, the EUA coefficient is not significant any more. This suggests that carbon price changes would not impact financial performance in oil and gas companies. In all, within the oil and gas industry, firms do not seem to respond differently to carbon prices increases compared to price decreases. This is in line with the findings of Scholtens and Yurtsever [32]. Our findings do not confirm Hypothesis 2 for the oil and gas industry.

Table 2 provides the results for the companies in Power & Heat. The market coefficients have a positive and significant impact on the stock returns of these firms. The significance of carbon prices is only marginally significant in the OLS estimations, but highly so in the pooled estimation results. This suggests firm specifics do play a role here too. These findings are very similar to those of the Oil & Gas sector. The oil price has a positive significant effect on Power & Heat companies. This can be explained by the fact that companies in this industry are very energy-intense. As could be expected, the coefficient of the oil price is higher in Oil & Gas than in Power & Heat. As in oil and gas companies, the results in Table 2 seem to suggest that firms in Power & Heat are able to pass-through energy price changes to the demand side, apparently because in the short-term this demand is rather price inelastic.. In contrast to Table 1, and in line with Zachman and Von Hirschhausen [37], the results in Table 2 show that there is a significant positive asymmetry in the reaction of the financial performance of companies in Power & Heat. Thus, returns of firms in the power and heat industry respond in a different manner to a positive change of the EUA price than to a negative EUA price change. We conclude that the results for Power & Heat confirm both hypothesis 1 and 2.

Table 2. Results for basic- and pooled OLS and asymmetries for the Power & Heat sector. Where the Power & Heat portfolio return is the dependent variable for the normal OLS regression and β₁, β₂, β₃, β₄, β₅, δ are the coefficients of the explanatory variables.

Power and Heat	OLS	Pooled OLS	Asymmetry OLS	Asymmetry Pooled OLS
C	-.01*	-.00**	-.02*	-.00***
β1 (market)	1.68***	.51***	1.69***	.52***
β2 (EUA)	.56*	.02***	.53**	.02***
β3 (oil)	.82***	.03***	.86***	.04***
β4 (gas)	.20	.01	.23	.01*
β5 (electricity)	.05	.00	-.02	.00
δ (asymmetric EUA)	–	–	.77**	.02**
observations	1043	33,376	1043	33,376
R^2	.76	.12	.71	.12
F-statistic	631.30	926.99	629.50	926,56

*, ** and *** show significant values at, respectively, 10%, 5% and 1% significance-levels, respectively

The results for Cement & Lime are in Table 3. The market coefficient has a significant positive impact on the returns of the companies, which is in line with previous literature and with our findings for the other industries. For the oil, gas and electricity price changes, no significant impact on the returns of companies in Cement & Lime is detected. This contrasts with the findings for Oil & Gas and Power & Heat, where the price change of oil in both sectors did have a significant effect. This may be explained by the fact that the revenues of companies in the latter two industries depend to a large extent on oil (and gas and electricity) price changes. Investors do not appear to think that changes in the factor prices have a material impact on returns in this industry. It would suggest that e.g. taxing these inputs would not be very effective (in contrast to doing so for Oil & Gas and Power & Heat). For Cement & Lime, energy prices mainly impact costs. The main variable impacting financial performance is the cement price. As in the power and heat industry, the EUA price change has a positive and significant impact of the returns in Cement & Lime. Table 3 also shows a slightly positive significant asymmetric effect in the EUA price change, suggesting that investors value carbon price increases different from price decreases. This is consistent with findings elsewhere in the literature [31,36]. Table 3 suggests that hypothesis 1 and 2 can be confirmed for companies in cement and lime.

Table 3. Results for basic- and pooled OLS and asymmetries for Cement & Lime. Where the Cement & Lime portfolio return is the dependent variable for the normal OLS regression and β₁, β₂, β₃, β₄, β₅, δ are the coefficients of the explanatory variables.

Cement and Lime	OLS	Pooled OLS	Asymmetry OLS	Asymmetry Pooled OLS
C	-.01**	-.00**	-.01*	-.00***
β1 (market)	1.04***	.67***	1.04***	.67***
β2 (EUA)	.48***	.04***	.51**	.04***
β3 (oil)	-.02	.00	.04	.01
β4 (gas)	.16	.01	.20	.01*
β5 (electricity)	.02	.00	.12	.00
β6 (cement)	1.76***	.12***	1.78***	.13***
δ (asymmetric EUA)	–	–	.45*	.03***
observations	1043	15,645	1043	15,645
R^2	.76	.30	.79	.29
F-statistic	632.15	1092.19	625.48	1087.55

*, ** and *** show significant values at, respectively, 10%, 5% and 1% significance-levels, respectively

Table 4 gives the results for Iron & Steel. Here, the results are highly similar to those in Cement & Lime. But, of course, now it is steel prices that are the dominant factor regarding financial performance. Table 4 shows that there is a positive and significant relationship between the EUA price change and the stock market returns. In addition, Table 4 also shows a positive asymmetric effect of the EUA prices. So, here too, investors, perceive the impact of carbon price increases different from the decreases. Just as with the previous two industry types, we do find support for hypothesis 1 and 2 for companies operating in iron and steel.

Table 4. Results for basic- and pooled OLS and asymmetries for Iron & Steel. Where the Iron & Steel portfolio return is the dependent variable for the normal OLS regression and β₁, β₂, β₃, β₄, β₅, δ are the coefficients of the explanatory variables.

Iron and Steel	OLS	Pooled OLS	Asymmetry OLS	Asymmetry Pooled OLS
C	-.01**	-.00**	-.02***	-.00***
β1 (market)	1.28***	.38***	1.28***	.38***
β2 (EUA)	.67***	.02**	.55**	.02**
β3 (oil)	.38	-.01	.41	-.01
β4 (gas)	.01	.00	.05	.00
β5 (electricity)	-.04	.00	-.04	.00
β6 (steel)	8.78***	.41***	8.77***	.41***
δ (asymmetric EUA)	–	–	1.11*	.04***
observations	1043	31,290	1043	31,290
R^2	.85	.24	.85	.24
F-statistic	907.12	1648.90	906.91	1649.53

*, ** and *** show significant values at, respectively, 10%, 5% and 1% significance-levels, respectively

To wrap up, regarding our variables of interest, we arrive at a positive significant effect of the EUA price changes in all four industries and this confirms the first hypothesis, namely that carbon prices significantly impact stock market returns. These results are in line with the expectations based on previous literature for the power industry [27,35]. Oil & Gas does not show any evidence of an asymmetric response to EUA price changes. However, the other three industries do show significant asymmetric reactions. The latter is in line with the second hypothesis, namely that there is an asymmetry in the stock market valuation response to carbon price increases compared to decreases. This finding is only partly in line with Scholtens and Yurtsever [32] who found a lot of diversity along different industries. However, their sample included much more industries. The findings for Power and Heat are consistent with those of Zachman and Von Hirschhausen [37].

Next, we investigate the role of volatility. Table 5 shows the mean equations (upper panel) and the variance equations (lower panel) for the four industries. Table 5 (upper panel) gives the results of the GARCH estimations of the mean equation for the four industries. It shows a highly significant positive impact of the market price change on the portfolio returns. The significant positive effects also hold for these four industries when applying the GARCH approach. The same accounts for the EUA price effect on the portfolio returns. Table 5 clearly shows that when using GARCH, the positive significant effect of the EUA price does hold too. Thus, with this alternative methodology, we establish a positive relationship between EUA price changes and the returns of the industry portfolios. All other coefficients that are significant in Tables 1-4 are significant according to the GARCH-model as well. Oil prices are only significant for firms in Oil & Gas and Power & Heat; gas and electricity prices are not significant.

Table 5. Results for GARCH-estimation for the four different sectors. Dependent variables are the portfolio returns per sector and β₁, β₂, β₃, β₄, β₅, β₆, β₇, δ, b, c, d₁, d₂, d₃, d₄, d₅, d₆ are the coefficients of the explanatory variables.

GARCH	Oil & Gas	Power & Heat	Cement & Lime	Iron & Steel
Mean equation
C	-.02	-.01	-.01*	-.01
β1 (market)	2.98***	4.52***	1.0428***	1.32***
β2 (EUA)	3.33***	.60***	.50**	.73**
β3 (oil)	5.76***	.85***	-.05	-47
β4 (gas)	.88	-.02	.16	-.05
β5 (electricity)	-.19	.00	-.01	-.02
β6 (cement)	–	–	1.87***	–
β7 (steel)	–	–	–	8.38***
Variance equation
C	.11***	.00**	.01***	.00***
b (GARCH-term)	.66***	.91***	.36***	.98***
c (residual)	-.01***	.07***	.14***	.01***
d1 (EUA volatility)	-1.99***	.00	.02	-.04
d2 (oil volatility)	-.53*	-.00	-.08***	.01
d3 (gas volatility)	-.38	.02	-.00	-.04**
d4 (electricity volatility)	-.48***	.02***	.02***	-.00
d5 (cement volatility)	–	–	.02**	–
d6 (steel volatility)	–	–	–	-.01
observations	1043	1043	1043	1043
R^2	.61	.70	.80	.85
F-statistic	132.25	210.16	234.86	438.02

*, ** and *** show significant values at, respectively, 10%, 5% and 1% significance-levels, respectively

The results of the variance equation (lower panel) do not suggest a significant relationship between the volatility in EUA and stock market volatilities. There is a significant relationship between the oil- and gas price volatilities and the stock returns in Oil & Gas and Iron & Steel. The steel price volatility has a negative but insignificant relationship with the stock returns. The cement price volatility is positive correlated with the volatility of the stock returns. For none of the industries we find sufficient empirical evidence to support the volatility hypotheses. The relationship between the EUA price change volatility and the electricity stock returns in Oberndorfer [27] is positive (but not significant), where the relationships between the EUA price change volatility and the volatilities of the industry stock returns in our paper are negative (but also not significant).

Thus, the estimation results on the basis of the GARCH analysis reveal that the positive and significant effect of the EUA price on firm value does still appear to hold. The GARCH estimations of the mean equation for the four industries show a highly significant positive impact of the market price change on the portfolio returns. The same accounts for the EUA price effect on the portfolio returns. This suggests that both the general market and the carbon price level impact the industry portfolio returns. The same is the case for the oil price in Oil & Gas and Power & Heat, but not in the other two industries. There, again, it mainly is the price of the key product (cement, steel) which is to be associated with industry portfolio returns. These results are in line with the findings of Oberndorfer [27] and Veith et al. [36]. They also confirm hypothesis 1. The results of the variance equation (lower panel) do not suggest a significant relationship between the EUA and volatility in the market returns. This only is the case for Oil & Gas, not for the other industries. With Oil & Gas, more volatility of carbon prices implies lower portfolio returns of the industry. Furthermore, there is a significant relationship between the oil and gas price volatilities and the stock returns in Oil & Gas and Iron & Steel. The steel price volatility has a negative but insignificant relationship with the stock returns. The cement price volatility is positive correlated with the volatility of the stock returns. For none of the industries we find sufficient empirical evidence to confirm the third hypothesis about carbon price volatility.

For sensitivity purposes, we first split the sample in two periods: The first period ends in October 2008 with the fall of Lehman Brothers, the second period starts in March 2009. As such, we take out four months of observations to have a clear break between the two periods. For the OLS regressions and asymmetry analysis for firms in Oil & Gas, the main results from Table 1 hardly change. However, gas prices turn out to be significant in the first period only. The significance of carbon prices substantially drops in the second period compared to the first. Furthermore, there is no significant asymmetric relationship between carbon prices and stock returns in the second period in Oil & Gas. In Power & Heat, carbon prices do not have a significant relationship in the first period, but there is such a relationship in the second period. The same is the case for firms in the cement and lime industry. Interesting is that oil prices load significantly on stock returns in the first period, but not in the second. With gas prices, this is the other way round. Carbon prices are not significantly related to stock returns in the first period in the iron and steel industry, but they are so in the second. The same holds for the oil prices. Furthermore, there only is a positive asymmetric relationship between carbon prices and stock returns in the second period. These findings suggest that the relationships are not stable over time, but may be subject to change. This is in line with previous findings in the literature (Sadorsky, 2999, [10,30,31]).

In addition, as detected by Oberndorfer [27] for Phase I, we investigate whether there are country specific effects after Phase I of the EU ETS. For example, the amplitude of the EUA effect may depend on country-specific characteristics such as differences in EUA long/short positions due to country-specific allocation plans and the structure of the national sector markets in which the companies operate. To this extent, we used a pooled regression framework in order to test for any country-specific results. Since every sector has another set of companies from different countries, we have to select country indicators for each industry. We look into country effects for countries with more than one firm per industry; the country indicator for the variable ‘other’ consists of countries in which only one single company is located. Hence, the pooled regression is extended with interaction term coefficients for the carbon price changes and the country indicators. Table A4 in the Appendix provides an overview of the country-specific effects of the EUA price in the four different industries. The results show not enough evidence to reject any of the country-specific hypotheses which hold that there are country-specific effects of the EUA price changes on the stock returns of companies in the different sectors. From an economic perspective, this means that there is no difference in the reactions of the stock returns on the EUA price changes in the different countries covered by the EU ETS. These findings contrast with those of Oberndorfer [27]. The difference might result from the fact that we have a broader coverage of firms. We also study industries where energy is not an output (Iron & Steel, Cement & Lime). Furthermore, in Phase I, the different governments had a lot of leeway to allocate EUAs to particular firms and industries. This has more or less vanished after 2007 with the result that national policies will have had much less impact.

Conclusion

We investigate the impact of the EU ETS on the value of individual firms after Phase I of the EU ETS in several countries and industries. We try to detect whether the ETS has a material impact on firms and can be regarded as a policy instrument to internalize external effects of production. Due to the increasing emission regulations resulting from the growing importance of the EU ETS, every company that emits carbon dioxide needs to be aware of the carbon price risk it is exposed to. As such, the focus of this paper is on the influence of carbon price changes on stock returns of 136 companies in the four largest industries covered by the EU ETS during the second phase of the EU ETS.

Our first hypothesis is that carbon price changes impact stock returns. We establish that there is a strong positive and significant relationship between the EUA price changes and the stock returns of the companies in the four industries. This suggests that EUA price changes are to be positively associated with stock returns. These findings are in line with the results of Smale et al. [35], Oberndorfer [27] and Veith et al. [36], but seem to contrast the neutral results of Löfgren et al. (2013) for Sweden regarding Cement & Lime and Iron & Steel. These results seem to suggest that energy-intense industires still benefit from the policy design (see [35,20]) and are able to pass through costs to the demand side. We want to highlight that despite many inefficiencies and over-allocation issues, the EU ETS does have a ‘real’ impact, like regarding the market valuation. This is in line with e.g. Petrick and Wagner [26], who find that carbon prices influence market shares of firms in relation to their energy efficiency.

Secondly, we test the EUA price changes for asymmetric effects on the stock returns. The results show that for every industry, with the exception of Oil & Gas, there is a positive and significant asymmetric effect. As such, we reject the hypothesis of no asymmetric relationship between the EUA price changes and stock returns for Power & Heat, Cement & Lime and Iron & Steel. This implies that the stock returns of the companies in these three industries react differently on a positive EUA price change compared to a decrease of the carbon price. The Oil & Gas industry has energy as a cost and as a return factor among its value drivers, which helps explain why it deviates in a structural manner from the other industries.

Thirdly, we investigate whether the volatility of carbon prices is to be associated with stock market volatility. In this respect, we do not find any significant relationship between the EUA price volatility and the stock price volatilities. This may be due to the fact that the impact of this price volatility is dwarfed by the volatility of other prices in the period under review, especially that of oil. Furthermore, interest rate volatility was extremely high because of the global financial crisis.

In addition, we establish that the relationships are not stable over time. This finding is similar to that of the previous literature [29–31]. Furthermore, we found very limited country-specific effects of EUA price changes. Only the countries with a large cement and lime industry show significant values. This result conflicts with the findings of Oberndorfer [27], who finds strong evidence for country-specific effects in the electricity industry. But they are in line with those of Gronwald et al. [28], who conclude to an increase in the maturity of the EUA market in more recent years.

From these findings, we derive three policy implications. First is that even at price levels that are generally perceived as being (too) low, there is a significant impact of carbon price changes on equity returns of firms (see also [26]). This suggests that the ETS does have an impact and may be used to nudge firms. However, to increase its efficiency, we agree with Löfgren et al. (2013) that reforms are required.

Second is that it is important not only to focus on the impact on traditional energy firms, but that the ETS also significantly impacts stock market returns in other energy-intensive industries. Hence, it is important to note that the ETS is a policy instrument that has ramifications throughout the economy as a whole. However, the industries respond in a different manner, depending on energy being mainly an input or an output as well, and related to firms' potential to pass-through price increases to final users. So far, this is not reflected in policy design and we suggest policy makers explicitly pay attention.

Third is that price increases have a different impact than prices decreases. It is important to be aware of this behavioural characteristic of the market participants as it impacts policy efficiency. So far, this phenomenon has been neglected in both industrial and energy policy design. It might imply that different instruments need to be set in place to deal with the response to a price increase or to a price decrease. A drawback is that this too could reduce policy effectiveness.

Overall, we conclude that the EU ETS has a small but statistically significant positive impact on the value of companies, not only in the power industry (as witnessed by [27,35]) but in other industries eligible with respect to the ETS as well. Given this finding, we can conclude that the ETS has material economic consequences. Via the ETS, external effects are being internalized, albeit to a very limited and imperfect extent. Whether or not it is the most efficient policy instrument to account for external effects and whether all external effects are covered is a different issue we leave for further study. The fact that carbon prices do have a significant impact on the stock returns of these companies also means that they face carbon price risk. If firms start to realize this, it may help to manage their carbon price risk and to reduce their carbon footprint.

Acknowledgements

We want to thank our colleagues and discussants at the UN PRI conference in Paris, France, at the Carbon Accounting Workshop at the University of Saint Andrews, Scotland, UK, and at the Enerday in Dresden, Germany, for their comments and suggestions. Furthermore, we very much appreciate the constructive comments and suggestions of the editor and four anonymous reviewers. The usual disclaimer applies.

APPENDICES

Table A1. Descriptive statistics of prices of the explanatory variables.

	Market	EUA	Oil	Gas	Electricity	Cement	Iron
Mean	267.51	15.76	87.36	19.89	50.17	1234.58	544.60
Median	266.69	14.60	83.82	21.75	48.29	1125.62	502.42
Maximum	414.90	28.75	144.07	35.92	93.56	2643.77	1235.54
Minimum	169.39	6.50	33.58	7.80	18.33	840.23	244.83
Standard deviation	45.84	4.65	24.84	6.22	13.09	347.17	232.09
Skewness	0.66	0.91	-0.05	-0.16	1.03	1.84	1.31
Kurtosis	0.22	0.08	-0.82	-0.95	1.21	2.46	1.02

Table A2. Descriptive statistics of returns of the explanatory variables.

	r m	r eua	r o	r g	r e	r c	r i
Mean	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Median	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Maximum	0.08	0.17	0.12	0.13	0.87	0.08	0.19
Minimum	-0.10	-0.14	-0.14	-0.24	-0.64	-0.10	-0.18
Standard deviation	0.02	0.03	0.02	0.03	0.07	0.02	0.04
Skewness	-0.05	0.30	-0.03	-0.91	2.37	0.14	0.17
Kurtosis	3.92	4.55	3.59	9.52	52.01	1.86	3.81

Table A3. Correlation matrix of returns of explanatory variables.

	r O&G	r P&H	r C&L	r I&S	market	EUA	oil	gas	elec-tricity	cement	steel
r O&G	1
r P&H	.72	1
r C&L	.70	.72	1
r I&S	.76	.77	.83	1
market	.75	.84	.87	.88	1
EUA	.25	.22	.23	.25	.20	1
oil	.51	.46	.41	.50	.46	.26	1
gas	.09	.09	.09	.08	.07	.15	.1	1
elec.	.01	-.00	-.01	-.02	-.01	-.01	.02	.04	1
cement	.64	.74	.78	.77	.86	.18	.39	.04	-.01	1
steel	.76	.74	.79	.91	.84	.25	.56	.08	-.03	.72	1

Table A4. Country-specific results for the four different industries.

	Oil & Gas	Power & Heat	Cement & Lime	Iron & Steel
Constant	0.00	-0.00***	-0.00**	-0.00**
β 1 (market)	0.64***	0.51**	0.67***	0.38***
β 2 (EUA)	0.02** (UK)	0.02 (Germany)	0.04*** (France)	0.02 (France)
β 3 (oil)	0.13***	0.04***	0.00	-0.01
β 4 (gas)	0.01***	0.01	0.01	0.00
β 5 (electricity)	0.00	0.00	0.00	0.00
β 6 (cement)	–	-	0.13***	–
β 7 (iron)	–	-	–	0.41***
θ 1 (EUA UK)	–	-0.039***	–	–
θ 2 (EUA Italy)	-0,01	-0.039**	0.02	–
θ 3 (EUA France)	0.00	–	–	–
θ 4 (EUA Greece)	0.03	–	0.06***	–
θ 5 (EUA Romania)	0.12***	–	–	–
θ 6 (EUA Russia)	0.04	–	–	–
θ 7 (EUA Germany)	–	–	0.08***	–
θ 8 (EUA Spain)	–	-0.00	0.04**	0.02
θ 9 (EUA Poland)	–	–	–	0.01
θ 10 (EUA Sweden)	–	–	–	0.01
θ 11 (EUA Other)	0.02**	0.04***	0.01	0.02*
Observations	1043	1043	1043	1043
R-squared	.25	.13	.30	.24
F-statistic	2083.86***	859.57***	1092.19***	1648.90***

Note: The dependent variables are the portfolio returns per industry. β₁, β₂, β₃, β₄, β₅, β₆, and β₇ are the coefficients of the explanatory variables (β₂ represents the effect of the EUA price change in the most important country with the four industries), θ₁, θ₂, θ₃, θ₄, θ₅, θ₆, θ₇, θ₈, θ₉, θ₁₀, and θ₁₁ are the coefficients of the country specific variables (*, **, *** show significant values at, respectively, the 10%, 5%, and the 1% level of significance)

Appendix B1. Countries per sector tested for country-specific effects.

Power & Heat	Oil & Gas	Iron & Steel	Cement & Lime
Germany	UK	Spain	Germany
Italy	Italy	France	Spain
UK	France	Poland	France
Spain	Greece	Sweden	Greece
Other	Romania	Other	Italy
	Russia
Other		Other

Appendix B2. Industries covered by the EU ETS.

Aluminium

Bricks and Ceramics

Cement & Lime

Chemicals

Coke

Education facilities

Food & Drinks

Glass

Hospitals

Iron & Steel

Mining

Motor industry

Oil & Gas

Pharmaceuticals

Power & Heat

Pulp & Paper

Waste management