The Valuation Relevance of Greenhouse Gas Emissions under the European Union Carbon Emissions Trading Scheme

Abstract

This study examines the valuation relevance of greenhouse gas emissions under the European Union Carbon Emissions Trading Scheme. We posit that carbon emissions affect firm valuation only to the extent that a firm's emissions exceed its carbon allowances under a cap-and-trade system and the extent of its inability to pass on carbon-related compliance costs to consumers and end-users. We measure a firm's ability to pass on the future costs by its market power and its carbon performance relative to its industry peers. The results show that firms' carbon allowances are not associated with firm valuation but the allocation shortfalls are negatively associated. We also find that the negative association between firm values and carbon emission shortfalls is mitigated for firms with better carbon performance relative to their industry peers and for firms in less competitive industry sectors. These findings, which suggest that the valuation impact of carbon emissions is unlikely to be homogenous across firms or industrial sectors, have important implications for future research design and for the disclosure and recognition of a firm's greenhouse gas liabilities.

Acknowledgements

We thank participants of research workshops at the City University of Hong Kong, Monash University, the University of Otago, and Victoria University of Wellington, as well as participants at the 2012 European Accounting Association Annual meeting and the 2014 JCAE Symposium for their comments. We also thank Julie Desjardins and Alan Willis for their support and input into this project. Special thanks goes to Kevin Ranney, Director, Advisory Services, of Sustainalytics for his generous assistance in providing us with Sustainalytics’ data regarding general environmental performance for our sample firms. We are grateful to Hila Fogel Yaari and Na Li for their excellent research assistance. We are grateful for the financial support from an Australian Research Council linkage grant with CPA Australia. Gordon Richardson thanks KPMG for their generous support. Yue Li thanks Michael Lee-Chin Family Institute for Corporate Citizenship at Rotman School of Management. Finally, we are especially appreciative of the efforts of the Associate Editor, Florin Vasvari, and two anonymous reviewers whose input has greatly enhanced the study.

Notes

¹Goodstein (2002, p. 314) discusses the difference between a cap-and-trade system versus a carbon tax system, pointing out that a cap-and-trade system is much less costly than pollution taxes for affected firms when allowances are initiated through a free allowance allocation system and in addition, that a carbon tax system will create additional costs for monitoring and enforcements on the part of the government.

²This figure of 3.15% of market capitalisation is consistent with a recent discounted cash flow analysis undertaken by Deutsche Bank (2009) of the carbon liability of the top 25 Australian emitters, representing 94% of the emissions of the ASX 100. Based on a series of assumptions regarding the allocation of free allowances and the ability for cost pass through, Deutsche Bank concludes thatthe valuation impacts are relatively benign: −3% or less for 17 of the top 25 emitters. This is a result of the combination of the allocation of free allowances, and in many cases the ability to pass through the incremental carbon cost.

³The IASB withdrew IFRIC #3 in 2005. One reason for the withdrawal is that its application created significant income volatility (Ertimur et al., 2011).

⁴See Ertimur et al. (2011), PriceWaterhouseCoopers (2007) and Veith et al. (2009) for a summary of the existing accounting practice with respect to carbon liability reporting.

⁵The information in this subsection comes from European Union Climate Action website (http://ec.europe.eu/clima/policies/ets/index.en.htm), Wikipedia (http://en.wikipedia.org/European_Union_Trading_Scheme), and other sources as identified.

⁶Although the dates of the NAP submission to the European Commission and the dates for the final approval by the European Commissions are public information, it is not possible to identify the exact dates when the member states released their NAPs to the public for consultation before they were submitted to the European Commission for the final approval. For this reason, we did not implement our tests using an event study methodology.

⁷We did not implement an event study approach in this study because of the difficulty of identifying event dates (see footnote 6).

⁸For example, the Liberal Party in Australia developed a campaign platform to remove the ‘Carbon Tax’ and won the 2013 Federal election. However, due to the new Australian Senate not taking office until 1 July 2014, it remains unclear whether the new Government will succeed in doing so.

⁹Investors may still impose a valuation penalty for carbon emissions from non-EU ETS zone because of anticipated future carbon-related regulations. For example, the Province of British Columbia in Canada imposed a carbon tax for fuels on 1 July 2012 based on $30 per tonne of CO₂ equivalent emissions. As of today, there are active debates about whether to implement a cap-and-trade system to combat climate change in Australia, Canada, and the USA.

¹⁰For sensitivity purposes, we also consider the following regression equation, based on the valuation model in Collins, Maydew, and Weiss (1997):

where E is earnings to common equity and all remaining variables are as defined in the text. This model differs in that it does not require the estimation of abnormal earnings and thus represents a sensitivity check of the cost of capital measure used to calculate AE. Results based on this alternative specification of the valuation model are consistent with those reported for all analyses.

¹¹We measure stock price as the end of April to coincide with the date that carbon emissions are disclosed (see Section 4.1). Sensitivity analysis using year-end price and price three months after fiscal year-end reveal the results to be robust.

¹²The cost of equity capital is based on the CAPM with β estimated using 60 months of historical return data, R_F equal to 5.0%, and the market price of risk ([E(R_M) − R_F]) equal to 6.0%. Sensitivity analysis indicates that results are not sensitive to a reasonable range of figures for either the risk-free rate or the market price of risk.

¹³We conduct our primary analysis based on unscaled measures following Matsumura et al. (2014) given the attractive feature that the coefficient estimates on the various emissions measures have a direct interpretation as the euros (€) per tonne of emissions. Notwithstanding, in order to consider the sensitivity of our results to the unscaled form of our econometric models, we repeat all analyses after scaling all items alternatively by the number of common shares outstanding and by the book value of equity. Results based on these alternative specifications are qualitatively identical to those reported in the main tables of the paper and hence, our conclusions are robust to the choice of scalar. See Goncharov and Veenman (2014) for a summary of the ‘scale debate’.

¹⁴For sensitivity purposes, we also ran the regressions with clustering by country, finding results to be qualitatively identical. Such a finding is not, however, unexpected. While there are differences across the sample countries in the process for distributing allowances between sectors, and between firms within sectors, our interest is in the valuation implications of the firm's allocation shortfall after the allocation process has occurred.

¹⁵Results and conclusions are qualitatively identical when, alternatively, the top and bottom 3% of all variables are winsorised, and when the top and bottom 1% are deleted.

¹⁶To illustrate, consider two rivals in an oligopolistic industry sector where the low emitter has 1 tonne of direct emissions per €1000 of sales not covered by free allowances while the high emitter has 10 tonnes per €1000 of sales not covered by free allowances. Assume further that, given the elasticity of demand, both firms can pass the costs of purchased allowances for just 1 tonne of emissions on to customers. In this setting, the high emitter has no pass on ability for 9 of its 10 tonnes, since price is set by the low-cost producer (the low emitter).

¹⁷We obtain the actual emissions from the European Emission trading registry for each of the operators owned by the parent entities in our sample. We use the industry classification of the installations owned by the operators provided by the EU ETS registry in accordance Annex 1 of the Directive I. This industry classification is the one in which the operators within each industry are most likely to be homogenous with respect to the production process giving rise to emissions and thus arguably yields the most precise relative emission performance ranking.

¹⁸As a final step, we checked the completeness of our sample to both Datastream and COMPUSTAT Global, finding no additional observations. Such a finding is not surprising, however, since DataStream and Compustat Global only record listed firms whereas most of the installations in the EU ETS are private companies. To identify whether these firms are owned by a listed company, it is necessary to use a data set of private companies which records the parent entity. It was for this reason that we purchased the BVD Amadeus database, which to our knowledge is the best available data set of private companies in Europe.

¹⁹At the country level, using 2009 for illustrative purposes, the total carbon emissions generated by our sample firms as a percentage of their yearly allowances range from 81.1% for Austria, 86.7% for Finland, 91.3% for Germany, 70.9% for Italy, 55.8% for Spain, 67.5% for Sweden, and 73.4% for the UK on the high side to 23.7% for Belgium, 45.4% for the Czech Republic, 21.2% for Denmark, 13.1% for France, 26.5% for Greece, and 18.6% for Poland. Figures for the remaining eight countries (Ireland, Latvia, Lithuania, Luxembourg, Netherlands, Portugal, Romania, and Slovenia) are all less than 10%, reflective of their more limited presence within the final sample.

²⁰While the Financials and Technology sectors are unregulated under the EU ETS, companies from these sectors appear in the sample because they represent publicly listed parent companies with operations within regulated sectors. As noted above, the final sample comprises listed companies that controlled installations registered as participating in the EU ETS.

²¹Notwithstanding the current preference within the literature towards the use of clustered standard errors when working with panel data, Petersen (2009) cautions that the choice of the appropriate approach will depend on the nature of the correlation between residuals. Following his caution, in addition to using clustered standard errors, we also conduct our analysis with firm fixed effects, and firm and year fixed effects, finding all results to be qualitatively unaffected.

²²Results and conclusions are unaffected when we alternatively measure the emissions variables in monetary terms using the spot price of carbon at the end of April of the following year (the time at which the emissions data are released). Here, the coefficients (p-values) on the allocation shortfall measure now expressed in euros (AllocShort €) are −0.268 (p = .074) for 2006, −0.387 (p = .050) for 2007, −0.701 (p = .026) for 2008, −0.684 (p = .005) for 2009, and −0.541 (p = .027) for the pooled data set.

²³Sustainalytics is a global leader in sustainability research and analysis, serving investors and financial institutions around the world. The company can be accessed via the following web link http://www.sustainalytics.com/.