ASSESSING THEIMPACT OF THE EUETS USING FIRMLEVEL DATAJAN ABRELL*, ANTA NDOYE FAYE**AND GEORG ZACHMANN†Highlights This paper investigates the impact of the European Union’sEmission Trading System EU ETS at a firm level. Using panel dataon the emissions and perance of more than 2000 Europeanfirms from 2005 to 2008, we are able to analyse the effectivenessof the scheme. The results suggest that the shift from the first phase 2005-2007 to the second phase 2008-2012 had an impact on theemission reductions carried out by firms. The initial allocation alsohad a significant impact on emission reduction. This challengesthe relevance for the ETS of Coase’s theorem Coase, 1960,according to which the initial allocation of permits is irrelevant forthe post-trading allocation of marketable pollution permits. Finally, we found that the EU ETS had a modest impact on theparticipating companies’ perance. We conclude that a fullauctioning system could help to reduce emissions but could alsohave a negative impact on the profits of participating companies.Keywords panel data, energy, climate change, uationeconometrics, firm behaviourJEL Classifications D21, C23 and Q49* TU Dresden, jan.abrelltu-dresden.de ** Bruegel/University of Strasbourg/University of Konstanz,anfbruegel.org†Bruegel, georg.zachmannbruegel.orgThe authors wish to thank Guntram Wolff, Philippe Quirion, RalfMartin, Tommaso Aquilante and Mouhamadou Sy for usefulcomments and ination. A special thanks to Hendrik Worschechfor excellent research assistance.BRUEGEL WORKING PAPER 2011/08JULY 20111 1. Introduction The European Unions Emission Trading System EU ETS is the biggest emissions trading scheme in the world. It is designed as a classical cap-and-trade system specifying a maximum amount of cumulated greenhouse gas emissions, and allocating tradable allowances to firms covered by the scheme. Allowing trade in these permits results in a market price for allowances. The price provides an economic signal of which mitigation measures are worthwhile1. A cap-and-trade system is by design effective in keeping the emissions of the participating installations below the cap. Thus, the relevant question is if this cap and thus emissions were below the emissions that one would expect in the absence of the system. There are two reasons why the cap might be too high and thus ineffective first, setting the cap ex ante is difficult. Emissions depend on numerous hard-to-predict factors most notably economic development. Therefore, setting a cap that is both ambitious and attainable is a difficult political rcise. Second, there are several flexibility mechanisms embedded in the design of the EU ETS. Most notably the transferability of Clean Development Mechanism CDM and Joint Implementation JI credits into EU emission allowances EUA as well as the bankability of allowances across phases2. Those instruments – that partly serve as security valves against too-high allowance prices – inflate the cap to an unpredictable degree. Consequently, it is ex ante unclear if companies will have to reduce their emissions due to the EU ETS. In this paper we address the following questions first, do the observed emissions reductions between 2005 and 2009 see section 3 indicate that the EU ETS resulted in emissions reductions, or are those reductions explained by changes in the economic environment Second, did the structural break between the first and second EU ETS phases led to a change in abatement behaviour Third, what are the influences of the initial allocation on the reduction effort of regulated firms Fourth, what is the treatment effect of the EU ETS on companies perances The EU ETS is divided into phases the trial phase 2005-2007; and the second phase 2008-2012 which coincides with the first commitment period of the Kyoto Protocol3. The rules of trading as well as the initial allocation of pollution permits have differed substantially between the two phases. The most notable changes are first, the cap, ie the total amount of permits allocated, was much lower in the second phase. Second, the regulation of the transfer of pollution permits between phases changed. In the trial phase the transfer of permits to future phases banking and vice versa borrowing was precluded. Thus, the trial phase was completely isolated from subsequent phases. In contrast, banking from the second to future phases is allowed. Third, uncertainty about the future availability of pollution permits decreased in the second phase as the long-term reduction target for 2020 was revealed in 20084. This motivates the question of how the structural break between phases affects the abatement decisions of firms. Studying the link between the carbon spot price and emissions is a way to answer this question. However, this carbon spot price was a short-term signal in the first phase because allowances were only to be used within the three years. By contrast the carbon spot price in the second phase should also encompass a long-term signal, as allowances are bankable at least until 2020 bankability beyond is not ruled out by the current directives and future rules of trading are subject to less uncertainty. Consequently, spot prices in the first and second phase are not comparable. Moreover, emission-reduction strategies are not entirely based on the marginal abatement cost of companies if the strategic motives of the regulated firms are taken into account. Given that initial allocation with valuable emission rights is based on a base year, firms try to manipulate emissions in that year in order to inflate their 1A comprehensive description of the rules and economics of the EU ETS can be found in Ellerman et al. 2010. 2 Under the Kyoto Protocol, Joint Implementation JI and the Clean Development Mechanism CDM reward projects that reduce GHG emissions with credits that can be used toward meeting Kyoto reduction targets. The EU Linking Directive allows JI or CDM credits to be converted by member countries into allowances usable for EU ETS compliance. 3The EUETS is thus one of the European tools to fulfill the Kyoto commitments of the EU member states. 4Given the on-going discussion about a 30 reduction until 2020, there still is some uncertainty about the future supply of pollution permits. 2 initial allocation5. Consequently, we choose to study the changes of abatement behaviour between phases instead of using the carbon price to investigate the effectiveness of the scheme. Another question arising in the context of the ETS is the impact of the rules of initial allocation on actual emissions. The invariant thesis of the Coase Theorem Coase, 1960 suggests that the initial allocation of permits is irrelevant for the post-trading allocation of marketable pollution permits. Put differently, the initial allocation does not affect the reduction behaviour of regulated firms; but, it certainly matters under distributional aspects, ie who receives the income of carbon regulation. However, the Coase theorem was derived under idealised conditions Coase, 1992. One line of theoretical reasoning against the neutrality of initial allocation originates in the theory of second best if the trading system is imposed on an economy in which taxes exists, the initial allocation matters for the efficiency of the system eg Goulder et al, 1999. Furthermore, initial allocation matters if regulated firms possess market power eg Burtraw et al, 2001. If we find that the initial allocation matters for reduction behaviour, this would have significant implications for the design of emissions trading schemes, as compensation through initial allocation would no longer be emissions neutral. Several authors have studied the effect of the EU ETS empirically. A concise overview is given in Anderson and Di Maria 2011. Our contribution is threefold. First, in contrast to other studies using country-specific firm level data Anger and Oberndorfer, 2008 we cover the entire European Union. Second, we explicitly take into account the structural break between the EU ETS phases. This allows us inter alia to study the effect of changing allocation on emissions. Third, previous literature on the effect of initial allocations on reduction behaviour has been either of theoretical nature or based on numerical simulations. With our unique data we are able to estimate the effect of initial allocation empirically. This firm-level data offers several more advantages. It allows us to eliminate the impact of aggregation over firms or installations when pering estimations. Furthermore, it allows exploiting a wide heterogeneity of firms with respect to their host country, turnover, employment, profit margin, sector and initial allocation. We find that the EU ETS induced emissions reductions in the second phase and that there were substantial differences in abatement behaviour across phases. Moreover, the initial allocation of permits and ex-post verified emissions are correlated. However, according to our findings, the EU ETS at most modestly affected profits, employment, and the added value of regulated firms. This paper is structured as follows. In the next section we describe and qualitatively analyse the dataset. Sections 4 and 5 describe the ological procedure and analyse the results of the estimation process. Section 6 concludes the paper. 2. Data Our dataset consists of a panel of European firms under EU ETS. We match the emissions data obtained from the European Commission Community Independent Transaction Log, CITL to firm level perance data from the AMADEUS database. From the CITL emission data, we extract ination on free allocation of emissions allowances and verified emissions 2005-2008 at the installation level. The availability of the data until 2008 is important since it allows us to include the second phase of the EU ETS. Some data issues with respect to the CITL data have been reported Trotignon and Delbosc, 2008. In particular, during the first phase of the EU ETS, the use of New Entrants Reserves was not available in the CITL’s public area, leading to some bias in the assessment of installation positions. We avoid these issues by selecting a balanced panel over the three years, ie we include installations that were present in the CITL’s public area already in 2005. From AMADEUS, we extracted ination on employment, turnover, profit margin, added value, labour and fixed capital costs 2003-2008. Both sets of data were matched via the addresses of the installations and we 5Another of strategic behaviour is associated with market power in either the permit or the output market or both eg Hahn, 1984; Matti and Montero, 2005. 3 end up with a set of 2101 firms 3608 installations, representing on average 59 percent of the total verified emissions6. We compute an allocation factor AF defined as the quotient of free allocation of emissions allocated to the verified emissions Anger and Oberndorfer, 2008. An AF 1 suggests that an installation has received allowances that exceed its emissions whereas the opposite suggests that this installation should either buy additional emission allowances or abate some of its emissions in order to comply with EU ETS. Table 1 in the Appendix 1 compares emissions and allowances in our sample of matched installations to the original CITL data. Our matched sample is representative of the biggest installations of the original CITL data in terms of emissions and allowances. There is also more heterogeneity in our installations than in the original CITL data. We classify firms into five sectors based on the two digit NACE Rev.2 code. Groups of countries were created with the geographic proximity as the main criteria. Firms are therefore classified in 18 regions or countries. Table 2 and Table 3 show the sectoral and regional distribution of our regulated firms. We notice that other non-metallic mineral products as well as electricity and heat sectors represent more than two-third of our sample. The two most represented countries in the sample are Spain and Germany with an aggregate frequency of 35 percent. Whereby, we retrieved 1/3 of the installations for the biggest emitting country Germany. Section 3 gives more ination on the aggregate emissions by country. Table 1 Sectoral distribution of the sample companies Sectors Number of firms Frequency Other non-metallic mineral products 806 38.36 Electricity and heat 660 31.41 Paper and paper products 416 19.8 Basic metals 159 7.57 Coke and refined petroleum products 60 2.86 6The matching procedure contains three steps. First, an automatized pre-matching identifies potential matches based on the similarity of company name, addresses and zip-code. In a second step this generous matching is narrowed down by selecting the actual matches from the computer-generated proposed matches. Finally, matches for the biggest unmatched installations are searched “by hand”. In the last two steps, in case of ambiguity additional sources of ination are drawn upon. 4 Table 2 Regional distribution of sample companies and CITL installations Total CITL installations Sample of matched firmsCountries of installations of firms of installations Country share in total sample firms Spain 1106 420 56719.99 Germany 1971 31464414.95 Portugal 277 23618311.23 France 1118 199 2919.47 Czech Rep. 421 120 2195.71 Poland 930 1142055.43 Italy 1124 1131675.38 Finland 649 103 4124.9 UK-Ireland 1247 851634.05 Bulgaria- Romania 399 73 1143.47 Sweden 798 711163.47 Austria 222 681183.24 Belgium-Lux 372 67 433.19 Slovakia 193 62942.95 Netherlands 437 47922.24 Denmark 403 39 621.86 SI-HU 365 33421.57 EE-LV-LT 280 27661.29 Greece 157 NANANA Cyprus 13 NA NANA Malta 2 NANA LI 2 NA NANA Norway 115 NANA Descriptive statistics of the main variables of interest are presented in Table 4. The relatively large difference between the value of the mean and the value of the median for these variables could indicate the presence of outliers in our sample. In the analysis it should be kept in mind, that the identified companies/installations are significantly larger than the average AMADEUS company / average CITL installation. Larger firms are overrepresented because retrieving the matching AMADEUS entry is more likely for larger firms. Table 3 Characteristics of the sample companies Added Value Employees Fixed Capital Profit Margin Allocation Factor 1 -1048 2 0 -46.69 0.50 5 470 10 309 -17.18 0.7525 2343 43 2968 0.05 1.00 Median 8673 150 12125 4.2 1.15 Mean 88541 663 159216 4.5 6.61 75 35014 447 49279 10.62 1.43 95 288316 2170 443055 25.37 3.31 Std 389039 2580 909914 14.32 178 3. The general perance of the EU ETS The EU ETS is divided into so-called phases. The first three years 2005-2007 were intended as a trial phase so that participants could become familiar with the new instrument. The current second ph