A border adjustment for the EU ETS: reconciling WTO rules and capacity to tackle carbon leakage

This article compares several configurations of a border adjustment (BA) to the EU Emissions Trading Scheme (EU ETS) that are designed to maximize their World Trade Organisation (WTO) compatibility, either with the GATT general regime or with Article XX (its environmental exception rule). The different BAs are assessed quantitatively using the partial equilibrium model CASE II, which represents four sectors included in the EU ETS (cement, aluminium, steel and electricity). The main findings indicate that the inclusion of imports and exports would reduce world emissions more than the inclusion of imports alone, that an obligation to buy EU allowances is more compatible with WTO rules than one based on a tax, and would be better at reducing world emissions. Moreover, if the BA is based on best available technologies, more precisely on the recently defined EU product-specific benchmarks, then the adjustment would only be partial and carbon leakage would nevertheless be significantly reduced. The popular view that a BA contributes to both carbon leakage limitation and to domestic production preservation is discussed, and it is argued on the contrary that although a BA would efficiently limit leakage, a decrease in European production of GHG-intensive products is to be expected. Industries that consume cement, aluminium and steel would pay more for these goods with a BA. Consequently, the price signal should be preserved and diffused in downstream sectors, an expected key result of climate policy. On the contrary, free allocation efficiently preserves domestic production, but does not preserve and diffuse the price signal and is less efficient in limiting leakage.


Introduction
Several countries are experiencing increased political pressure for some form of border adjustment (BA) to complement a stringent climate policy. This would level the playing field between domestic producers and foreign producers who experience little or no constraint on their GHG emissions. Arguments that justify such a trade measure generally refer to competitiveness concerns and/or the weakening of environmental efficiency. 1 Indeed, if a firm faces higher costs in its home country, it may struggle to compete and move offshore, or lose its business to offshore companies. The domestic industry will then lose market share in both domestic and foreign markets, resulting in job losses at home and an increase in GHG emissions abroad, in other words, carbon leakage.
However, as a trade measure, a BA may be contested by a World Trade Organisation (WTO) member under its dispute settlement mechanism. In fact, some early observers considered a BA in complement to a domestic emissions trading scheme or a tax on GHG emissions to be incompatible with WTO rules (see references in Godard, 2011). However, several analyses (de Cendra, 2006;Ismer and Neuhoff, 2007;Pauwelyn, 2007;Eichenberg, 2010), including a recent report from WTO/UNEP (2009), have questioned this position and concluded that, under some conditions, such a measure may be WTOcompatible. Not all experts agree on the conditions, but some common elements can be identified, thus allowing for the draft of a BA that would be compatible with WTO rules -or at least increasing the chances of it being so. Yet the design of the BA also impacts on carbon leakage mitigation. The analysis in this article examines this point in the context of the EU Emissions Trading Scheme (EU ETS).
The aim of the present article is not to reach conclusions on the opportunities for implementing an EU BA, but to propose a quantitative analysis of the most WTO-compatible BAs. In particular, their efficiency in limiting carbon leakage is examined, as well as their impact on the production and market shares of European firms.
The quantitative analysis is based on CASE II, a static and partial equilibrium model that represents four sectors included in the EU ETS (cement, aluminium, steel and electricity). The analysis examines the implications of different BA designs for production, prices and trade flows in each industry, as well as on the leakage-to-reduction ratio for each sector and for the whole ETS. Throughout this article, it is assumed that allowances are fully auctioned for two reasons. First, a BA is much more difficult to justify under free allocation than under auctioning. This is a common conclusion of the literature ( de Cendra, 2006;van Asselt and Biermann, 2007;Kommerskollegium, 2009). 2 Second, with a BA, European industry suffers less from competitive disadvantage, so there is little rationale for free allocation that creates economic distortions (Neuhoff and Matthes, 2008). 3 Based on the law literature, the design of a potential BA to the EU ETS is discussed in Section 2. The model and the scenarios used are presented in Section 3, with the results discussed in Section 4 and conclusions offered in Section 5.

How to maximize the WTO compatibility of a BA?
To implement a BA, several elements must be defined, particularly the form (e.g. tax-based or allowances-based), the coverage (e.g. imports and exports, or only imports; direct and indirect emissions or only direct emissions) and the adjustment base (e.g. EU or foreign average specific emissions, best available technologies or plant-specific emissions).
Based on the recent law literature, we identify points critical for WTO compatibility. Generally, two options are examined: first, compatibility with the General Agreement on Tariffs and Trade (GATT) general regime, and then (if the BA has not been judged compatible with the general regime) a possible recourse to the environmental exception rule, Article XX. Each option can refer to very different requirements. Moreover, it must be kept in mind that WTO compatibility depends not only on its design, but also on the manner in which it is implemented, especially for the 'Article XX' option, which cannot be taken into account in our modelling.

General regime
Under the GATT general regime, several articles must be taken into account to avoid being found in violation of GATT rules, because different ones are relevant to the import and export parts of the BA. GATT Articles I-III are particularly important with respect to imports, with GATT Article XVI, as well as the 1994 Agreement on Subsidies and Countervailing Measures (SCM), being important to exports (Eichenberg, 2010). Specifically, Article III requires WTO members to treat foreign goods no less favourably than comparable domestic goods, and GATT Article XVI expresses general disapproval of subsidies, and instructs signatories to avoid their use wherever possible (Kommerskollegium, 2009).
According to Pauwelyn (2009), carbon equalization measures on imports at the border (a border duty, allowance requirement or performance standard) may be modelled in compliance with WTO non-discrimination rules if the measure does not impose a heavier burden on imported products than on domestic products. Moreover, Kang (2010, p. 7) states that 'once the products at issue are found like products', taxes on imports even slightly 'in excess of like domestic products will be found in violation of GATT Article III:2'. This is then a crucial point to respect.
In the EU ETS, it would be simpler to ensure similar treatment between European and non-European firms by basing carbon equalization measures on allowances (Genasci, 2008;Monjon and Quirion, 2010). If the measure is based on an import tax, it would be difficult to determine the value of the tax that would result in similar treatment. The same arguments apply to exports.
The BA requires an estimation of the carbon content of imported products. Although European firms included in the EU ETS are obliged to provide their emission level, a similar obligation can hardly be imposed on importers. Ismer and Neuhoff (2007) propose the use of best available technology (BAT) for importers in order to ensure a similar treatment or at least a more favourable treatment of imported (rather than European) products. Yet the determination of a world BAT can be problematic (Monjon and Quirion, 2010). An easier way to define a BAT would be to use the product-specific benchmarks set in the EU ETS for industry products. 4 Although defined for a different purpose, that is, to set the amount of freely allocated allowances, they are a good candidate for defining the quantity of emissions that should be imputed to imports. Moreover, if a particular exporting firm that has unitary emissions lower than the EU benchmark can prove (through independent third-party verification) that its emissions are lower than the reference value, then it must be allowed to use this value (Pauwelyn, 2007).
According to Eichenberg (2010), the export part of a BA might be authorized by the SCM. The crucial point is that it does not advantage domestic producers with respect to Article XVI:4 (WTO/UNEP, 2009).
Calculating the amount of the BA -for both imports and exports -raises many challenging issues, but addressing indirect emissions costs would add even more complexity. Indeed, the inclusion of indirect emissions costs would require a calculation of the emissions involved in producing electricity, which can be controversial: different methods used to assess the CO 2 intensity of power generation (e.g. marginal vs. average) generate different results, and interconnections between adjacent countries further complicate the analysis. Moreover, the cost pass-through in the electricity price must also be evaluated, because overestimating pass-through might result in an excessive adjustment (Genasci, 2008). For these reasons the authors have excluded them in the designs that are examined.
Finally, the most-favoured nation principle (Article I) requires the imposition of the BA on all WTO members. It is not therefore possible to exempt a group of countries because they, for instance, are engaged in an international climate agreement, or are the least developed countries.
Given the different elements relevant to the GATT general regime, it is proposed to test an allowance-based BA on imports and exports with an adjustment coefficient for imports based on a European benchmark. Because the benchmark is lower than both the EU and foreign specific emission levels, the adjustment is only partial. For the export part, European firms are not obliged to surrender allowances when their production is exported outside the EU.

Article XX
If the core provision of world trade law prohibits a BA, it may still be allowed by one of the general exception provisions of Article XX. In particular, Article XX allows trade restrictions to 'protect human, animal or plant life or health' (Article XX (b)) or to ensure 'the conservation of exhaustible natural resources' (Article XX (g)). It cannot be invoked to offset competitive disadvantages for domestic industry. Wiers (2008) emphasizes that the measure should really contribute to its own environmental goal. Hence, the first indicator to be examined must be the impact on world emissions, rather than that on carbon leakage. In relation to this point, some analyses conclude that the BA must be limited to imports (Ruiz-Fabri and Reynier, 2010), while Ismer and Neuhoff (2007) and Godard (2011) conclude that the export part of the BA can also be compatible with Article XX.
Even if it is demonstrated that the measure achieves its environmental goal, its acceptance will be determined by the balance between its contribution to climate protection and its trade restrictiveness (Kommerskollegium, 2009). The introductory clause of Article XX (its 'chapeau') must also be respected; this states that 'the measures are not applied in a manner which would constitute a means of arbitrary or unjustifiable discrimination between countries where the same conditions prevail, or a disguised restriction on international trade'. The objective of the chapeau is to prevent the 'abuse of exceptions' in Article XX and to ensure that they are 'exercised in good faith to protect interests considered as legitimate under Article XX' (Bordoff, 2009, p. 19). According to Charnovitz (2007), the adjudication focus of appraisal is on how the measure is applied, rather than how it is designed. This aspect cannot be taken into account in our modelling, but the common interpretation of Article XX is that it requires that, before imposing a unilateral BA, a country should have made all efforts to reach an international agreement. The Appellate Body's interpretation of the Article XX chapeau also suggests that the EU should take the initiative in negotiating with countries that might be affected by the BA. The BA should take into account the efforts of trading partners to abate GHG emissions 5 , and this may result in lower (or no) BA on imports from countries having measures comparable in effectiveness.
What to retain for our modelling? First, demonstrating the environmental benefits of the BA is crucial. Second, there is no clear conclusion concerning the legality of the export BA. As already stated, the objective of this article is not to settle this issue but to evaluate the environmental performances of a BA that covers both imports and exports, and one that does not cover exports. Third, there is no clear conclusion concerning the adjustment base to adopt. It may be possible to use BAT emissions or average specific foreign emissions to calculate the BA on imports. Both options are examined, 6 that is, an adjustment based on the EU benchmark and an adjustment based on specific foreign emissions. 7 Finally, to avoid the BA from being considered a 'disguised restriction', the evolution of the market shares of European firms is examined as an indicator for demonstrating that the BA does not have trade-restricting objectives (de Cendra, 2006).

Which option has the most chance of compliance?
The objective here is not to choose an option but to provide insights into the ability of different BAs to limit carbon leakage and reduce world emissions. Nevertheless, in the law literature, it is sometimes thought that the 'Article XX' option has more chance of being accepted (Cosbey, 2008;Horn and Mavroidis, 2010). Pauwelyn (2007), however, emphasizes that Article XX may oblige a graduation, or even exclusion, of (i) countries with their own climate policies in place, and (ii) countries at a low level of development. This would considerably reduce the BA coverage, because the present world is characterized by asymmetric carbon constraints rather than being composed of one group of countries with a climate policy and another group without. 8 Nevertheless, Godard (2011) moderates this last position.

The model and the scenarios
CASE II is a static and partial equilibrium model that aims to evaluate the impact of different EU ETS Phase III (2013 -2020) designs. The sectors represented (cement, aluminium, steel and electricity) share a potentially large impact on carbon pricing, but have contrasting direct and indirect emissions, as well as different exposures to international competition (Hourcade et al., 2007). They represent approximately 75% of the emissions covered by the EU ETS (Kettner et al., 2008;MEDDTL and CDC, 2010). The model features two regions: the European Union 27 (EU 27) and the rest of the world (RoW). 9 Although general equilibrium effects have significant implications for climate policy, above all they play a role in leakage through the international energy markets channel, but are less important for comparing anti-leakage policies (Fischer and Fox, 2009). Moreover, the competitiveness channel of leakage is a microeconomic mechanism, so a partial equilibrium model is relevant for its analysis (Quirion, 2010). Admittedly, the neglect of general equilibrium effects is a limitation, because these two leakage channels are, to some extent, interrelated (cf. Kuik and Hofkes, 2010), and there are associated difficulties with analysing whether the different policy scenarios allow the carbon price signal to travel down the production chain to the consumer. However, general equilibrium models suffer from a higher sectoral aggregation than our model. In particular, they do not isolate cement or aluminium.
When a climate policy is implemented in the EU, domestic firms incur three types of additional cost: n Abatement cost: this is based on marginal abatement cost (MAC) curves taken from the POLES model for the year 2020 at the aggregated EU 27 level. In POLES, MAC curves are available for CO 2 energy emissions from, among others, non-metallic minerals, steel and electricity sectors. MAC curves give the decrease in emissions for each CO 2 price. Abatement costs are incorporated into the model as variable costs as in Fischer and Fox (2009). POLES does not include a MAC curve for the aluminium sector, so data from the Energy Modelling Forum EMF-21 project on multi-gas mitigation was used (Weyant et al., 2006). n Purchase or sale of allowances: if needed or possible. n Increase in electricity price: the marginal production cost of cement, aluminium and steel increases when the electricity price rises. A cost pass-through of 100% in the power sector is assumed, whatever the policy scenario modelled. The electricity price rises by the sum of the abatement cost and the purchase costs of allowances.
Owing to space constraints, the details of the model are presented in an Appendix available online from the website of Climate Policy.
Compared to the version of the model used in Monjon and Quirion (2011), the main difference is that economic growth assumptions have been updated to account for the economic downturn. Gross domestic product (GDP) growth projections are taken from IEA (2010). Moreover, the scenarios tested differ significantly as explained below. Finally, the EU ETS is modelled for 2020, assuming a cap at 21% below 2005 emissions, but also assuming that credits from the Clean Development Mechanism (CDM) and joint implementation (JI) will amount to 6% of the cap in 2020 10 , so the cap is set at 15% (21% minus 6%) below 2005 emissions.
Scenario 0 is a no-policy scenario that assumes no ETS. Scenario 1 (Auction) features full auctioning of allowances without rebating of the auction revenues to the firms covered by the ETS and without any anti-leakage provision. In scenario 2 (Free allocation), allocation is simulated in a way that is as close as possible to the actual allowance allocation in the EU ETS. Firms in the cement, steel and aluminium sectors that are exposed to leakage risk receive free allowances, whereas firms in the power generation sector do not. The amount of free allocation is calculated as the product-specific benchmark (in tCO 2 e/ tonne produced) multiplied by the production of the sector in 2005. Then, a reduction factor is applied if the sum of the allowances in the cement, aluminium and steel sectors is above 85% of their 2005 emissions. Moreover, new firms entering the market receive allowances, under the 'new entrant reserve' provisions, while firms exiting the market do not longer receive them because of the 'closure rules'. Unlike output-based allocation, the allowance a firm receives is not proportional to its current output level.
Scenarios 0 -2 are reference situations against which the following scenarios, with BA, can be compared. In scenarios 3 -6, the designs are defined in the ways that were identified previously.
Scenario 3 (Tax-based M + X(bch)) assumes auctioning with a BA both on imports and exports. As in Monjon and Quirion (2011), a 'tax-based' BA is simulated to compare its performance with that of an 'allowance-based' BA: importers have to pay a carbon tax and exporters receive a rebate. Consequently, the impact on the allowance market is less direct than with an 'allowance-based' BA, under which importers are obliged to buy some allowances. The import part is set at the level of the CO 2 price on the EU ETS market multiplied by the EU product-specific benchmark. The export part is based on the EU average specific emissions.
In scenarios 4-6, an allowance-based adjustment is simulated: importers have to buy allowances but exporters are partly exempted. This impacts the demand for, and hence the price of, allowances.
Scenario 4 (M + X(bch)) differs from scenario 3 in that it is, as explained above, allowance-based. Compared to scenarios 1 -3, the supply of allowances is adjusted to the new group of emissions covered, that is, the fact that importers now have to buy allowances but exporters do not (see Figure 1). More precisely, the reduction factor of 15% now applies to the 2005 emissions of EU ETS installations plus emissions from imports, minus emissions from exports. With this scheme, the cost to European firms rises only by the electricity price increase when production is exported. Moreover, it is supposed that the constraint imposed on importers does not encourage them to reduce their specific emissions.
Scenario 5 (M(bch)) is also allowance-based, but the adjustment covers only imports. Consistently, the reduction factor of 15% is now applied to the 2005 emissions of EU ETS installations plus the emissions from imports ( Figure 1).
Finally, scenario 6 (M(RoW) is similar to the previous scenario except that the adjustment on imports is set at the level of the CO 2 price on the EU ETS market, multiplied by the average specific emissions in the rest of the world (Figure 1).

CO 2 price
Because policy scenarios are compared for a given emission reduction target but not for a given CO 2 price, variation of the latter across scenarios is worth considering. As shown in Figure 2, the CO 2 price in 2020 is around EUE20/t and is consistent with current forecasts for a target of 221% in 2020. It is highest with free allocation for the following reason: free allocation is economically a subsidy that a firm receives, regardless of its production level, as long as it remains in activity; 11 this FIGURE 1 Different emission coverage and caps of the EU ETS in 2020 FIGURE 2 CO 2 price (EUE/allowance) and public revenues (EUE billion) encourages new firms to enter the market; competition 'à la Cournot' is increased, so decreasing the price on the EU market and raising quantities sold. 12 This increases the demand for allowances, hence the CO 2 price, which in turn reduces specific emissions.
With BAs, the price is higher than in Auction because the BA limits the substitution of foreign production for domestic production (which is a way of reducing CO 2 emissions in the EU). Thus, to prevent aggregate emissions exceeding the cap, lower specific emissions must be achieved, which requires a higher price signal. Among the scenarios with a BA, the price is higher in allowance-based scenarios because the model supposes that adjustment on imports is not based on the specific emissions of the plant used to produce each good, but on the EU benchmark or on the average specific emissions in the RoW. In both cases, the adjustment is exogenous for a firm that exports into the EU, so these firms have no incentive to reduce their own specific emissions. For aggregate emissions not to be in excess of the cap, this must be compensated by lower specific emissions in the EU; that is, the CO 2 price must be higher.

Public revenues
In all the scenarios, public revenues raised by selling allowances to the sectors covered by the model are between EUE20 and E30 billion in 2020. Free allocation gives the lowest amount, although this is not so different from that generated in the other scenarios. The reasons are that power generation firms, which account for the majority of emissions, still have to buy allowances, and they buy them at a higher price. Public revenues are higher with a BA, above all when the BA is allowance-based. The reason is a higher CO 2 price and, for scenarios 4 -6, a larger number of allowances sold (Figure 1). This is true even if the adjustment concerns both imports and exports, because emissions from the latter are lower than emissions from the former.
The extra public revenues under the scenarios with a BA (around +E0.2 -2 billion, most of which is due to the increase in the CO 2 price, as can be seen in Figure 2) would, of course, be welcome as most EU governments are trying to reduce public deficits. However, these revenues could also be used for climate change mitigation and/or adaptation in developing countries to facilitate an international agreement on climate change, as highlighted by Godard (2011). It would also demonstrate the 'good faith' of the EU, which is an important point relative to Article XX.
For scenarios 3 -6, Table 1 shows the increase in unitary production costs for European firms and for imports from the rest of the world to the EU under a BA. Neither the general regime nor Article XX requires a comparison of the increase in costs for home and foreign products. The general regime just requires that the regulatory treatment of imported goods is no less favourable in practice than that for domestic products. Nevertheless, it is interesting to use our model to examine how European and foreign costs evolve in the different sectors and for each scenario. The cost increase for European firms is systematically higher than for foreign firms, except when the BA import part is based on the average specific emissions of the RoW (where it is much higher for steel). In the aluminium sector, the cost increase for European firms is much higher than for foreign firms because the main part of this increase comes from indirect emissions not included in the BA. Figure 3 presents the change in world emissions from the cement, aluminium, steel and electricity sectors compared to the no-policy scenario and the leakage-to-reduction ratio, that is, the increase in RoW emissions divided by the decrease in EU emissions. 13 EU emissions are constructed to be the same in the first three scenarios. Emissions in the RoW increase in all scenarios, although by a small amount, and the leakage-to-reduction ratio is always below 11%. Compared to the Auction scenario, leakage is only slightly reduced by free allocation, but is at least halved by BAs. The most efficient adjustments for reducing leakage are those that cover both exports and imports, followed by scenario 5, which uses RoW average specific emissions rather than the EU benchmark. The same ranking is found in terms of world emissions reductions, except for the tax-based BA, which is less efficient in this respect than all allowance-based BAs.
An important conclusion is that all BA scenarios reduce world emissions more than the Auction scenario. This provides a clear environmental justification for the mechanism, crucial for WTO compatibility under Article XX.
A comparison of scenarios 3 and 4 shows that the allowance-based BA is more efficient in limiting leakage and even more so in decreasing world emissions, so an allowance-based BA is preferable to a tax-based one, both in terms of WTO compatibility and environmental objectives. When the BA is applied only to imports, its environmental performance is better if the adjustment base is the average specific emissions of the RoW, but the WTO compatibility of this option could be difficult.

FIGURE 3 Changes in emissions compared to no-policy scenario
Another interesting insight from this modelling exercise is that the decrease in world emissions is higher when exports are included in the adjustment. This result is due to the fact that EU specific emissions are lower than those of the rest of the world, mainly because the EU industry pays a CO 2 price.
For a BA to be WTO-compatible and acceptable to trade partners, it should not constitute arbitrary discrimination or disguised trade restriction (see Section 1). It may be argued that, as long as a foreign firm does not pay a higher CO 2 price than a European firm with the same specific emissions, a BA does not constitute an arbitrary discrimination. However, if foreign producers can show that they lose market shares due to EU climate policy, they may use this argument in a possible WTO case, or may be more likely to engage in trade retaliations. For this reason, Figure 4 presents the market shares of EU producers on the world market, that is, the sum of the EU and RoW markets. In all the scenarios with a climate policy, the market shares of the EU are reduced compared to the no-policy scenario. The differences among scenarios 3-6 are very small. This result suggests that, whatever the design of the BA selected, there is no disguised restriction of trade. Note that the EU market share is lower with BAs than with free allocation, although, as has been seen, the latter is less efficient at reducing leakage. The explanation is a composition effect: the size of the EU market (in which the EU market share is naturally higher) is less reduced with free allocation, because the product price increase is lower (Figure 6).
Although a BA is more efficient than free allocation in reducing carbon leakage, a greater production decrease in GHG-intensive sectors is to be expected ( Figure 5). The simulated BA scenarios do not shelter aluminium significantly, because indirect emissions, which are higher than direct emissions, are not included. For cement, the BA mitigates the substitution of imports for European production (and European exports by foreign production when the BA has an export part), but it raises the CO 2 price, reduces the European demand for cement and, therefore, reduces EU production for the domestic market. There is also a decrease in European steel production, although to a lesser extent: BAs mitigate production losses by one-third at most in Scenario 3. Steel is more open to international trade, but has a lower CO 2 /value added ratio than cement. Among the scenarios with a BA covering both imports and exports, the one based on allowances features a slightly higher production loss than the one based on a border tax. These results highlight a frequent misunderstanding. It is often thought that one instrument, for instance a BA, can contribute both to carbon leakage limitation and to domestic production preservation. The results presented in Figures 4 and 5 lead to a different conclusion: correctly defined, a BA would perform well in limiting carbon leakage but would not prevent a significant decrease in European production. Indeed, the BA increases the price of CO 2 -intensive goods in the EU, as seen in Figure 6, because European firms carry a CO 2 price and, in addition, imports now have to pay a CO 2 price. Hence, industries that consume cement, aluminium and steel, such as the construction and automobile sectors, would pay more for these CO 2 -intensive goods with a BA, leading to a decrease in European demand. 14 Consequently, the price signal is diffused in the rest of the economy, a key expected result of climate policy. Cost-efficient abatement requires GHG-intensive goods to be replaced by cleaner goods; for example, wood would replace cement and steel in buildings. 15 Free allocation in scenario 2 would limit the increase in these product prices, hence the European production FIGURE 5 Changes in EU production compared to no-policy scenario FIGURE 6 Changes in the price index in the EU compared to no-policy scenario decrease. However, it performs poorly in limiting carbon leakage, and would prevent much of the above-mentioned substitution.

Conclusions
Different design options (with the best chance of being WTO-compatible) were analysed for a BA to the EU ETS. Simulations with the CASE II model indicate that all the tested options efficiently reduce carbon leakage to at least half that in a scenario with full auctioning and no BA. Public revenues are also higher with a BA.
However, BAs lead to a production decline in EU GHG-intensive sectors, due mainly to a decrease in European consumption rather than a reduction in EU market shares. With a BA, the price signal is preserved and is diffused in the rest of the economy, a key expected result of climate policy. Nevertheless, the decrease in European production would harm some politically influential industrial sectors and constitute a political hurdle to the implementation of a BA.
Among the various options tested, some are more environmentally efficient than others. First, an allowance-based adjustment, which obliges importers to buy and surrender allowances, reduces world emissions more. Second, an adjustment on both exports and imports reduces world emissions more than an adjustment on imports alone. Finally, although an import adjustment tackles leakage more efficiently if it is based on specific emissions in the RoW rather than on EU-specific emissions, this option is less likely to be WTO-compatible.
In conclusion, the BA used in scenario 4 (i.e. allowance-based, covering both imports and exports and using EU product-specific benchmarks) is the most efficient in reducing world emissions and in limiting carbon leakage. Moreover, it is among the options most likely to be accepted by a WTO dispute settlement panel. Hence, if the EU wants to complement the ETS by a BA (a question that goes beyond the present article), this option seems to be the most attractive.
6. Whatever measure is applied, importers must be permitted to demonstrate how much carbon they emitted individually and pay for allowances on that basis (Bordoff, 2009). 7. An adjustment based on foreign specific emissions is problematic to evaluate, because most non-EU production installations have no obligation to declare -and thus do not know precisely -their CO 2 emissions. An option is to ask importers to provide certified information on the carbon content of the products they want to import in the EU, but it is difficult to oblige importers to do so as for a small importer the administrative burden could be high in proportion to its sales. Another option is to use the average emissions per tonne in the exporting country for every product covered by the BA, but this value could be difficult to compute, especially if the country is reluctant to participate. Consequently, the practical feasibility of this scenario is not certain. 8. Conversely, if Article XX is not used, it is not possible to exempt a group of countries because, for instance, they are engaged in an international climate agreement, or are the least developed countries, due to the GATT most favoured nation principle (cf. previous section). 9. Because the model aggregates all foreign countries into one RoW region, a BA based on country-of-origin specific allowance obligations cannot be assessed, which could reduce leakage further. 10. Over 2008 -2020, the limit of credits from the CDM and JI in the EU ETS is set at 1.68 Gt CO 2 e, hence 6% of the cap (Turner, 2010). It is assumed that this limit will be binding and that credits will be used homogenously over the period. 11. In the EU ETS, a firm must emit more than a minimum threshold to receive its allocation (EC, 2010). In the model, all the firms of the same sector are symmetric and then produce the same quantity, which is higher than these thresholds in all scenarios. 12. The pricing behaviour of the firms is different in the EU markets and in foreign markets. Indeed, in foreign markets, the competition increases less because EU firms have only a small market share. 13. The variations are calculated between a scenario with the climate policy and a scenario without, and include emissions from the power sector. 14. The model does not allow taking into account the climate policy implemented in the rest of the economy. For instance, building retrofitting would moderate the decrease in cement demand. 15. However, in some downstream industries, the cost increase would be weak. For example, according to ADEME (2007), the emissions related to the use of steel and aluminium in a car of 1 tonne is around 1.6 tonne of CO 2 . If a CO 2 price of around EUE20 is assumed and that the cost pass-through is complete in the electricity, steel and aluminium sectors, the cost increase to produce a car would be around E30-35.