By increasing the cost of using fossil fuels, carbon taxes should incentivise substitution toward cleaner energy. In practice, firms may shift production and emissions across borders, a problem known as carbon leakage. This column studies carbon leakage in the context of the EU Emissions Trading System using firm-level import data. Following the introduction of the scheme, the share of dirty inputs sourced from non-participating countries has increased significantly, consistent with carbon leakage. A policy which eliminates the carbon leakage incentives, such as a carbon border adjustment mechanism, could reduce emissions by more but would likely lead to higher prices as well.
The implementation of a carbon tax to disincentivise the use of fossil fuels has been advocated by numerous economists and implemented in several countries. By increasing the cost of using fossil fuels, a carbon tax should in theory lead firms to substitute towards cleaner energy sources and technological innovations to cut back ‘dirty’ production. Costs may also rise for firms down the supply chain that use dirty input goods in production, so the total impact of a carbon tax can in theory be large. In practice, however, carbon taxes are unilateral policies, whose effectiveness in decreasing global emissions is questionable, given firms’ ability to adapt to an increase in carbon prices. Unilateral carbon prices may shift production and thus emissions across borders, a problem that is commonly referred to as carbon leakage (e.g. Dechezleprêtre and Sato 2017, Naegele and Zaklan 2019).
In Coster et al. (2024), we study carbon leakage in the context of European policies. The EU has been at the forefront of carbon policies, with the introduction of the EU Emissions Trading System (EU ETS) in 2005. This scheme is meant to set a common price for carbon across the EU and applies to a set of firms in high-emission industries, such as the production of steel, chemicals, cement, or ceramic goods. By increasing the cost of emission-intensive production, the system incentivises producers to invest into cleaner technologies (Colmer et al. 2024). The policy, however, creates leakage opportunities in downstream sectors, as products taxed within the EU can be sourced from outside the Union. To eliminate the remaining leakage and incentivise foreign firms to produce low-carbon intensive goods, the EU just extended the ETS to importers of high-emission products through the Carbon Border Adjustment Mechanism (CBAM), which will take effect in 2026.
We provide novel evidence of carbon leakage observed in firm-level import data during the 2010s, when the EU ETS system was becoming increasingly binding. We first construct a new dataset that classifies ‘clean’ and ‘dirty’ manufacturing goods by leveraging information about the scope of the European policies. Given our focus on the supply chains, we define a list of dirty inputs based on whether these goods are produced in a sector covered by the EU ETS. We merge the list of dirty goods to French firms’ balance sheet and import data to study where firms source clean and dirty goods, and how this behaviour has changed over time. By focusing on firms’ input usage, we capture the indirect impact of the policy on downstream customer firms. The use of a dataset with detailed firm-level information further allows us to control for common trends and other economic forces that may be driving patterns observed in aggregate French imports, and which would make it difficult to clearly identify the impact of the EU ETS on carbon leakage. Specifically, we use the firm-source country-product level dimension of our dataset to identify how firms have changed their relative sourcing of dirty versus clean inputs from non-ETS countries, over time.
Results are summarised in Figure 1, with the left panel comparing the import share of dirty versus clean inputs sourced from non-ETS countries, while the right panel focuses on firms’ propensity to source inputs from these countries. In both cases, the firm-level panel is balanced in the product times sourcing country dimension and the coefficients recovered from a Poisson-pseudo maximum likelihood (PPML) estimator.
Prior to the introduction of the EU ETS, the relative share of dirty inputs sourced from non-ETS countries is slightly decreasing, compared to clean inputs. The trend is reversed from the first phase of the EU ETS (2005-2007) and becomes significantly positive at the end of the second phase (2008-2012). By 2019, the share of dirty inputs sourced from non-ETS countries has increased by 15% relative to the share of clean inputs sourced from the same area, compared to 2004. As illustrated by the right panel in Figure 1, this shift in input sourcing strategies is largely driven by the extensive margin. The estimated probability of a firm importing dirty goods from non-ETS countries has risen dramatically relative to the probability of importing a clean good – up to almost 30% by 2019. We show that results are robust to further controlling for firm-product-country, sector-year, and country-year fixed effects, thus suggesting the diverging trends in sourcing strategies is driven by changes in the relative propensity to source regulated and non-regulated inputs from non-ETS countries. Finally, we also show that the positive trend is entirely driven by non-regulated firms, i.e. firms are not directly exposed to EU ETS regulation but use the regulated products as an input in production. We interpret these findings as indicative of carbon leakage, a tendency of French firms belonging to sectors that are downstream regulated sectors to adapt their sourcing to avoid paying the carbon tax. These are typical behaviours which the carbon border adjustment mechanism is meant to target.
Figure 1 Evolution of firm-level imports from non-ETS countries: Dirty versus clean inputs
Note: This figure shows the evolution of the relative import share (panel (a)) and the relative import probability (panel (b)) of dirty inputs sourced from non-ETS countries, compared to clean inputs sourced from there. All coefficients are normalised to zero in 2004, one year before the beginning of ETS. The regression controls for product-country and year fixed effects. The blue areas correspond to Phases 1 and 3 of ETS. The confidence intervals are defined at the 95% level.
We rationalise these results using a heterogeneous firm model of sourcing decisions, estimated on pre-ETS firm-level data. The model extends Antras et al. (2017) to a nested CES structure in which firms choose where and how much to source of each variety of two categories of inputs, clean and dirty. The distinction between clean and dirty inputs makes it possible to simulate various carbon price policies that mimic the EU ETS – a carbon tax on all inputs produced by regulated sectors in ETS member states – and the EU ETS + CBAM system, in which inputs imported from non-ETS countries that belong to the coverage of CBAM are also taxed.
Under the EU ETS only scenario and applying a €100/ton of CO2 carbon tax, global emissions fall by 1.8 million tons of CO2, but at the cost of a moderate increase in the price of French manufacturing products, of 0.05%. The small impact of the tax is driven by the ability for French firms to avoid paying the cost of the tax by switching input sourcing away from regulated countries, most notably Russia and China in our simulations. The model thus replicates 80% of the carbon leakage estimated in the data. When we simulate an extension of the EU ETS to include imported products, thus mimicking the future CBAM, the global reduction in emissions increases fourfold while the manufacturing price level rises by a factor of ten. The combined EU ETS + CBAM tax is far more powerful than the EU ETS alone as the CBAM eliminates carbon leakage incentives. Therefore, firms not only reshore input sourcing away from the most polluting EU ETS countries, like Romania and Bulgaria, but now also shift sourcing from countries like Russia and China outside of the regulated area, towards less polluting countries, such as France. The resulting cut in emissions comes with a cost, however, as prices faced by French consumers rise quite a bit given the higher cost of inputs.
The results underscore the importance of considering the indirect impacts of climate policy through supply chain linkages. Firms can adapt along multiple dimensions to minimise the cost of carbon policies. These adaptation strategies can be welfare-improving when incentivising clean technology investment, but they can also induce undesirable carbon leakage effects when firms adapt their sourcing strategy. While firms in our model reshore dirty inputs locally under the EU ETS + CBAM policy, this also reduces their international competitiveness which leads to higher prices faced by domestic households. Moreover, under a partial sectoral coverage of these carbon policies, the system generates carbon leakage incentives further down the value chain.
Source: cepr.org