Carbon leakage through supply chain adjustments

Wang, Hanyi Working Paper Carbon leakage through supply chain adjustments IDOS Discussion Paper, No. 10/2025 Provided in Cooperation with: German Institute of Development and Sustainability (IDOS), Bonn Suggested Citation: Wang, Hanyi (2025) : Carbon leakage through supply chain adjustments, IDOS Discussion Paper, No. 10/2025, ISBN 978-3-96021-255-3, German Institute of Development and Sustainability (IDOS), Bonn, https://doi.org/10.23661/idp10.2025 This Version is available at: https://hdl.handle.net/10419/316405 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Carbon Leakage through Supply Chain Adjustments Hanyi Wang IDOS DISCUSSION PAPER 10/2025 Carbon leakage through supply chain adjustments Hanyi Wang Bonn 2025 Hanyi Wang is an economist who studies the cost-effectiveness and distributional impacts of climate policies on firms and their trade implications. She also studies sustainable investing’s impact on firms and how regulatory pressures facilitate green transitions. Dr Wang received her PhD in Economics from the University of California San Diego in June 2024. The German Institute of Development and Sustainability (IDOS) is institutionally financed by the Federal Ministry for Economic Cooperation and Development (BMZ), based on a resolution of the German Bundestag, and the state of North Rhine-Westphalia (NRW) as a member of the Johannes-Rau-Forschungsgemeinschaft (JRF). Suggested citation: Wang, H. (2025). Carbon leakage through supply chain adjustments (IDOS Discussion Paper 10/2025). Bonn: German Institute of Development and Sustainability (IDOS). https://doi.org/10.23661/idp10.2025 Disclaimer: The analyses expressed in this paper are those of the author(s) and do not necessarily reflect the views or policies of the German Institute of Development and Sustainability (IDOS). Except otherwise noted, this publication is licensed under Creative Commons Attribution (CC BY 4.0). You are free to copy, communicate and adapt this work, as long as you attribute the German Institute of Development and Sustainability (IDOS) gGmbH and the author(s). IDOS Discussion Paper / German Institute of Development and Sustainability (IDOS) gGmbH ISSN 2751-4439 (Print) ISSN 2751-4447 (Online) ISBN 978-3-96021-255-3 (Print) DOI: https://doi.org/10.23661/idp10.2025 © German Institute of Development and Sustainability (IDOS) gGmbH Tulpenfeld 6, 53113 Bonn Email: [email protected] https://www.idos-research.de Printed on eco-friendly, certified paper. IDOS Discussion Paper 10/2025 III Abstract This paper examines carbon leakage through supply chain recalibration in response to European carbon policies. Using input-output data and a high-frequency identification approach for carbon policy shocks, this paper investigates whether stringent carbon regulations in Europe affect the imports of carbon-intensive inputs from major emerging economies lacking similar policies. The findings reveal a temporary increase in the rate of change of imports from emerging countries relative to all inputs in the carbon-intensive sectors following carbon policy shocks, with effects peaking after two years before dissipating. While not directly quantifying emissions transfer, this study suggests some evidence of short-term input substitution patterns consistent with carbon leakage through international supply chains. IDOS Discussion Paper 10/2025 IV Contents Abstract Abbreviations 1 Introduction 1 2 The European carbon market background 5 3 Data 6 3.1 Emerging economy import shares 7 3.2 Sectoral-level carbon pricing data 7 3.3 Environmental policy stringency (EPS) 8 3.4 Other variables 8 4 Summary statistics 8 5 Empirical methods 9 5.1 Identification challenges 9 5.2 Emerging economy import shares 10 5.3 Dynamic effects of carbon pricing on import patterns from emerging economies 11 5.4 Heterogeneous effects 11 6 Results 12 6.1 Main results 12 6.2 Heterogeneity analysis 14 7 Robustness 16 7.1 Responses to carbon tax 16 7.2 Alternative environmental policy measure 17 7.3 Sector composition 17 8 Conclusions and policy implications 18 References 19 Appendix 21 IDOS Discussion Paper 10/2025 V Figures Figure 1: Carbon price and carbon-intensive import ratio trends 2 Figure 2: Carbon policy shocks in the European carbon market 6 Figure 3: Impulse responses of import patterns to carbon policy shock (pct) 13 Figure 4: Heterogeneity by European (importing) countries’ GDP 14 Figure 5: Heterogeneity by European (importing) countries’ carbon tax status 15 Figure 6: Heterogeneity by sectoral carbon intensity 16 Tables Table 1: Summary statistics 9 Figures in the Appendix Figure A1: Cross-sectional and temporal variation of EPS index 21 Figure A2: Robustness: impulse responses to carbon tax rate 22 Figure A3: Robustness: restricted sample excluding energy-intensive sectors 23 Tables in the Appendix Table A1: Variable description 21 Table A2: Environmental policy stringency and import shares from emerging countries 24 Table A3: Environmental policy stringency and import shares from emerging countries 24 IDOS Discussion Paper 10/2025 VI Abbreviations CBAM Carbon Border Adjustment Mechanism CGE computational general equilibrium CO2 carbon dioxide EEA European Economic Area EE MRIO extended multi-region input-output EPS environmental policy stringency ETS Emissions Trading System EU European Union EU ETS European Union Emissions Trading System GDP gross domestic product GMM generalised method of moments ICIO inter-country input-output IEA International Energy Agency IPPC Intergovernmental Panel on Climate Change ISIC International Standard Industrial Classification NACE Statistical Classification of Economic Activities in the European Community OLS ordinary least squares PCT percentage change PPI producer price index R&D research and development VAR vector autoregression WDI World Development Indicators (World Bank) WIOD World Input-Output Tables IDOS Discussion Paper 10/2025 1 1 Introduction Climate change represents a global environmental challenge that requires coordinated international action. While governments worldwide have introduced policies to curb carbon emissions, the effectiveness of unilateral regulations remains debated due to potential carbon leakage – where stringent carbon policies in one region lead to increased emissions in less regulated areas (Copeland, Shapiro, & Taylor, 2022). This concern reflects the broader pollution haven hypothesis from trade theory, which suggests that environmental regulations in developed countries can shift pollution-intensive production towards regions with lower abatement (Levinson & Taylor, 2008). Climate policy is particularly vulnerable to such effects given its global nature and the significant variation in regulatory stringency across countries. The potential for production and emissions to shift to less regulated regions can significantly undermine the effectiveness of unilateral climate policies in reducing global emissions, posing a critical challenge for climate policy design and implementation. Carbon leakage can occur through three distinct channels (Colmer, Martin, Muûls, & Wagner, 2024). First, firms may shift their supply chains to source more intermediate products from unregulated suppliers, potentially reducing compliance costs while sacrificing some value added. Second, market forces may redistribute production to firms in unregulated sectors, either domestically or abroad, as firms in regulated sectors face higher costs. Third, companies with multiple facilities might reallocate production within their network from regulated to unregulated locations. Each channel represents a pathway through which environmental regulations could lead to unintended increases in emissions in less regulated regions. While previous studies such as Dechezleprêtre, Gennaioli, Martin, Muûls, and Stoerk (2022) have focused primarily on multinational firms or production relocation and found limited evidence of carbon leakage, this study examines the potentially more responsive supply chain channel. Specifically, this research investigates whether European carbon policies lead to increased sourcing of carbon-intensive inputs from emerging economies, where environmental regulations are typically less stringent. This focus on input sourcing, rather than complete facility relocation, is motivated by several key insights from the trade and environmental literature. Ederington, Levinson, and Minier (2005) argue that pollution-intensive industries often face significant barriers to relocation due to high transportation costs, substantial plant fixed costs, and benefits from industrial agglomeration. Instead of relocating entirely, these industries might find it more feasible to adjust their input sourcing patterns. Using detailed input-output data, this paper examines how sectors adjust their international sourcing decisions in response to carbon policy changes, focusing particularly on shifts toward suppliers in emerging economies. This approach allows us to identify a potentially important but understudied channel of carbon leakage through supply chain recalibration. Figure 1 provides suggestive evidence of a supply chain adjustment channel. The ratio of carbonintensive imports from emerging economies closely tracks the evolution of European Union Emissions Trading System (EU ETS) carbon prices, averaged across Exiobase sectors based on Intergovernmental Panel on Climate Change (IPCC) classifications for Emissions Trading System (ETS) coverage. Import ratios increased from around 4.5 per cent in 2000 to over 6 per cent by 2020, with notable acceleration after the EU ETS introduction in 2005. Particularly sharp increases in import ratios coincided with periods of high ETS prices, such as 2005 to 2006 and 2019 to 2020. This parallel movement between sectoral carbon prices and import patterns, while not establishing causality, suggests that European industries may adjust their sourcing decisions in response to carbon policy changes. While this correlation is consistent with potential carbon leakage through supply chains, establishing causality requires addressing important empirical challenges. IDOS Discussion Paper 10/2025 8 carbon policy shocks as the key independent variable while controlling for both ETS prices and carbon tax rates to capture the overall carbon pricing landscape. 3.3 Environmental policy stringency (EPS) The Environmental Policy Stringency (EPS) index is obtained from the OECD. This countryspecific measure covers 40 countries from 1990 to 2020 and quantifies the degree to which environmental policies put an explicit or implicit price on pollution behaviour. The index incorporates 13 instruments related to climate and air pollution.8 As a standardised measure, the EPS index enables cross-country and intertemporal analyses of environmental regulation effects. It uses a scale from 0 (not stringent) to 6 (highest degree of stringency), reflecting the relative stringency of a country’s environmental policy instruments in a given year.9 This EPS index is employed here in the robustness tests to validate the main findings and ensure the consistency of the results across different measures of environmental policy stringency. 3.4 Other variables This analysis incorporates several control variables to account for various economic and environmental factors. EU ETS prices and carbon tax rates are included from the World Carbon Pricing Database (Dolphin & Xiahou, 2022) to control for the existing policy environment.10 Carbon intensity data from Exiobase environmental accounts controls for differences in production efficiency across countries and sectors. Country-level GDP from the World Bank’s World Development Indicators (WDI) accounts for economic development and market size. Producer price indices (PPI) control for broader cost pressures affecting industrial production decisions. Trade openness (trade-to-GDP ratio) captures countries’ integration into global markets and their propensity to trade. Together, these variables help isolate the effects of carbon policy shocks from other factors that might influence import patterns. 4 Summary statistics The analysis covers 1999 to 2019, ending before potential confounding events: the Covid-19 pandemic, the Ukraine conflict, and the EU ETS Market Stability Reserve introduction. The final sample comprises 76 Exiobase sectors across 21 European countries. Table 1 presents comprehensive statistics for the key variables. Panel A shows the distribution of carbon-intensive import shares measured three ways. The ratio to total imported inputs (ImportShare1) averages 14.62 per cent, with considerable variation (SD=15.32 per cent). When measured against total inputs (ImportShare2), including both domestic and imported, the average share is lower at 5.97 per cent (SD=10.71 per cent), reflecting the importance of domestic inputs in production. The ratio to total domestic supply (ImportShare3) shows a higher mean of 47.92 per cent but with substantial heterogeneity (SD=253.19 per cent), indicating significant cross-sectional variation in import dependence relative to domestic production capacity. 8 The EPS index does not include water and waste management policies since the data are not available in a large cross-country panel and are also hard to turn into a quantitative cross-country indicator. 9 Scores are assigned according to the distribution of the observations with the respective policy implemented. 10 See in this connection https://github.com/g-dolphin/WorldCarbonPricingDatabase. IDOS Discussion Paper 10/2025 9 Panel B presents the carbon policy shock measure derived from surprises in carbon permit prices. These shocks, constructed as differences between actual and expected price changes, have nearzero means. This aligns with rational expectations theory, as systematic bias in expectations should be arbitraged away. However, the substantial standard deviations reveal significant unexpected policy variations that could influence sourcing decisions. The shock distribution suggests frequent policy surprises in both directions, providing variation for identifying causal effects. Panel C summarises the control variables capturing various economic and policy dimensions. Carbon intensity exhibits substantial sectoral heterogeneity, reflecting differences in production technologies and energy efficiency. Economic indicators like GDP, PPI, and trade openness show considerable variation across countries and time, highlighting the importance of controlling for different stages of economic development and market integration. Although carbon taxes vary by country due to national policy differences, ETS prices are uniform across all EU countries in any given year. In the summary statistics, both measures are averaged across Exiobase sectors to reflect sectoral coverage of both carbon policies. Table 1: Summary statistics Observations Mean SD P1 P50 P99 Panel A: Outcome variables ImportShare1(over total imported inputs) 28318 14.62 15.32 5.04 9.63 17.89 ImportShare2(over total inputs) 28319 5.97 10.71 1.02 2.51 5.72 ImportShare3(over total domestic supply) 28319 47.92 253.1 1.42 4.01 11.88 Panel B: Carbon policy shock CPShock(pct) 33516 -1.66e-12 2.32 -1.59 -.51 1.30 Panel C: Control variables Log(GDP) 20748 20.00 1.32 19.09 19.77 21.37 Log(carbon intensity) 33516 .45 1.46 .012 .06 .27 ETS price 33516 6.70 11.62 0 0 9.45 Carbon tax rate 33516 2.50 11.43 0 0 0 PPI 25384 85.51 10.14 79.80 87.34 93.14 Trade openness 33516 98.82 40.79 64.54 86.39 127.77 Notes: This table presents the summary statistics for outcome variables of import patterns, carbon policy shock variables, and control variables. The variables are defined in Appendix Table A1. 5 Empirical methods 5.1 Identification challenges Evaluating the impact of carbon pricing on potential leakage effects ideally requires a study design that isolates the changes induced by these policies. However, several challenges complicate this ideal scenario. First, identifying and isolating the impact of carbon policies is difficult given various factors simultaneously influencing firm and industry behaviour. Second, participation in carbon pricing schemes, such as the EU ETS, is not random, introducing potential selection bias. In practice, a randomised control trial for carbon pricing is politically unfeasible. Researchers use econometric methods, such as difference-in-differences, assuming that carbon policies do not affect unregulated firms or sectors. However, this assumption may be violated as policy effects IDOS Discussion Paper 10/2025 10 can be transmitted through supply chains, potentially contaminating control groups. Moreover, when studying potential leakage through import patterns, additional challenges arise. Various factors beyond carbon policies, including global economic trends, trade agreements, and technological changes may influence changes in import ratios. The gradual implementation of carbon policies also makes it difficult to identify clear “before” and “after” periods for analysis. To address these identification challenges, the high-frequency identification approach developed by Känzig (2023) was utilised to quantify carbon policy shocks. This method builds on techniques originally developed for monetary policy analysis (Gürkaynak et al., 2004; Gertler & Karadi, 2015; Nakamura & Steinsson, 2018), where researchers measure asset price movements in narrow windows around policy announcements to isolate policy impacts. The EU carbon market’s frequent policy updates and active futures trading make it ideal for high-frequency identification. By measuring price changes in tight windows around regulatory events in the carbon market, this approach can plausibly rule out reverse causality since broader economic conditions are already incorporated in pre-event prices and unlikely to change within the narrow event window. The constructed carbon policy shocks are considered parallel to exogenous shocks such as weather events or monetary policy changes. This method allows for the isolation of the impact of carbon policies from other confounding factors and overcomes the challenges of non-random policy implementation. While initially developed for monetary policy, this identification strategy has proven effective in various policy contexts, including global oil markets and emissions trading schemes. 5.2 Emerging economy import shares Three measures of import shares from emerging economies are constructed as follows: ImportShareik,t m= Carbon−intensive inputs sourced from emerging economies𝑖𝑖𝑖𝑖,𝑡𝑡 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶−𝑖𝑖𝐶𝐶𝑖𝑖𝑖𝑖𝐶𝐶𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑑𝑑𝑖𝑖𝐶𝐶𝐶𝐶𝑑𝑑𝑖𝑖𝐶𝐶𝐶𝐶𝑖𝑖𝐶𝐶𝐶𝐶𝑚𝑚, where ImportShareik,t m denotes the fraction of carbon-intensive inputs that sector i in country k sources from emerging countries without carbon policies in year t, relative to the total carbonintensive inputs or supply from all sources. Carbon-intensive sectors are defined as those in the top 40 per cent of carbon intensity across all sectors. The superscript 𝑚𝑚 ∈1,2,3 indicates three different measures, each using a different denominator: • 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑚𝑚𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑1= total carbon-intensive imported inputs • 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑚𝑚𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑2= all carbon-intensive inputs (imported and domestic) • 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑚𝑚𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑3= total carbon-intensive domestic supply. These measures allow for the assessment of the relative importance of carbon-intensive imports from emerging countries that lack carbon policies. The first measure focuses on the composition of the imports. This is useful for directly observing changes in international sourcing patterns. It shows whether carbon-intensive imports from emerging economies are increasing relative to imports from other sources. The second measure includes both imported and domestic inputs. It provides a broader perspective on how the share of carbon-intensive imports from emerging economies changes relative to the total input mix. This measure is particularly important as it indicates a potential input substitution effect, showing whether there is a general shift towards foreign sourcing or just a recomposition of existing imports. As such, it is the primary focus of this paper, offering the most direct evidence of whether firms substitute domestic inputs with imports from emerging economies in response to carbon policies. The third measure compares carbonintensive imports to the total domestic economic activity in the sector. It can indicate whether imports are growing relative to domestic production, which is particularly relevant for assessing potential domestic industry impacts. IDOS Discussion Paper 10/2025 11 5.3 Dynamic effects of carbon pricing on import patterns from emerging economies The dynamic causal effects on the three measures of carbon-intensive import shares are estimated using (panel) local projections à la Jordà (2005). Δℎy𝑖𝑖𝑖𝑖,𝑖𝑖+ℎ 𝑑𝑑=α+βℎCPShock𝑖𝑖+∑θ𝑝𝑝 ℎ𝑥𝑥𝑖𝑖𝑖𝑖,𝑖𝑖−𝑝𝑝 ′ 𝑃𝑃 𝑝𝑝=1 +σ𝑖𝑖 ℎ+ϵ𝑖𝑖𝑖𝑖,𝑖𝑖+ℎ, where h denotes the horizon at which the relative effect is estimated. The dependent variable Δℎy𝑖𝑖𝑖𝑖,𝑖𝑖+ℎ 𝑑𝑑≡y𝑖𝑖𝑖𝑖,𝑖𝑖+ℎ 𝑑𝑑−y𝑖𝑖𝑖𝑖,𝑖𝑖−1 𝑑𝑑 is defined as the cumulative difference of the import share outcome variables (m∈1,2,3) measuring the fraction of carbon-intensive inputs that industry sector k in European country i sources from emerging economies in year t+h, as defined above. For panel data local projections, one would normally project the outcome variables on the shocks and control variables, including the lag of the outcome variable. However, including the lagged outcome variable with fixed effects creates biases in the estimation that would require more complex generalised method of moments (GMM) methods to address. To avoid this issue while maintaining simple ordinary least squares (OLS) estimation, the approach follows Jordà, Schularick and Taylor (2015) and projects the cumulative difference y𝑖𝑖𝑖𝑖,𝑖𝑖+ℎ 𝑑𝑑−y𝑖𝑖𝑖𝑖,𝑖𝑖−1 𝑑𝑑$ on the righthand side variables, excluding the lagged dependent variable. The independent variable of interest, CPShock𝑖𝑖 denotes carbon policy shocks at year $t$, extracted from a proxy-VAR model from Känzig (2023). Following Känzig and Konradt (2023), lagged shock variables are not included in the specification since the shock series shows no significant serial correlation (Ljung-Box test p-value = 0.88). Percentage changes rather than baseline (euro) changes in carbon prices are employed as the shock measure, as percentage changes better reflect how sectors evaluate relative costs and make sourcing decisions. The vector 𝒙𝒙𝒊𝒊𝒊𝒊,𝒕𝒕−𝒑𝒑 ′ includes lagged control variables including carbon intensity, GDP, ETS price, and carbon tax rate, with lags up to order P, allowing for richer dynamics in economic factors that might influence import shares. Sector (destination country’s) fixed-effects 𝜎𝜎𝑖𝑖 ℎ is included to account for time-invariant characteristics. The superscript h on the coefficients indicates that separate regressions are estimated for each horizon h, allowing for dynamic effects over time. 5.4 Heterogeneous effects Heterogeneous effects of carbon policies on import patterns from emerging economies are explored across different sector types, country sizes, and existing carbon pricing regimes. First, the various effects across importing sectors with different levels of carbon intensity are examined. Importing sectors of the European countries are categorised into low, middle, and high carbon intensity groups (terciles) based on their pre-shock carbon intensity (ton CO2 per thousand-dollar gross output). By examining heterogeneity across carbon intensity levels, the assessment can determine whether carbon policies disproportionately affect high-emission sectors, potentially leading to greater carbon leakage. Next, heterogeneity across European importing countries is investigated based on their economic size. Countries are grouped into terciles according to their logged GDP levels (in constant 2015 dollars). In the analysis, these are referred to as low, middle, and high GDP countries, respectively. This approach helps one to understand if the impact of carbon pricing on import patterns varies with the economic size of the importing country, potentially revealing differences in adaptation strategies or vulnerabilities to carbon leakage across economies of different scales. Finally, heterogeneity based on the presence of national carbon taxes in European importing countries is explored. A distinction is made between countries without additional national carbon taxes and those with national carbon taxes on top of the EU-wide ETS system. This distinction IDOS Discussion Paper 10/2025 12 allows for examination of whether the presence of additional national carbon pricing mechanisms influences the effect of broader carbon policy shocks on import patterns from emerging economies. 6 Results 6.1 Main results How do carbon policies affect the import patterns of the European countries from the emerging countries? The baseline carbon policy shocks are extracted from surprises in euro-denominated carbon price changes relative to prevailing wholesale electricity prices, as constructed by Känzig (2023) The three outcome variables measuring the fraction of carbon-intensive inputs sourced from emerging economies are examined relative to i) total imported inputs; ii) all inputs (imported and domestic); and iii) total domestic supply. The results suggest modest evidence of European sectors adjusting their sourcing of carbonintensive inputs in response to carbon policy shocks. This is examined through three measures, each capturing different aspects of potential carbon leakage. Figure 3 presents impulse responses to an unexpected increase in carbon policies for all three measures. The middle panel shows the primary measure – the ratio of emerging economy imports to all inputs (both imported and domestic) – which directly captures substitution between domestic and foreign sources. This measure reveals a small but positive cumulative response of approximately 0.2 percentage points that peaks around two years after a one standard deviation increase in carbon policy shock (normalised to increase energy prices by one per cent on impact), before gradually returning to zero. While Känzig (2023), who developed the carbon policy shocks used in this study, found that these shocks lead to a 0.6% reduction in domestic GHG emissions, our findings reveal a significant leakage channel through supply chain adjustments. The increase in carbon-intensive imports indicates that carbon leakage through supply chains may offset some of the domestic emissions gains, though precise quantification of this offset would require additional analysis. This result contrasts with Colmer et al. (2024), who find no statistically significant changes in French firms’ importing behaviour under the EU ETS and conclude that supply chain leakage is not a major driver of emissions reductions. This study's findings can be considered alongside earlier work by Sato and Dechezleprêtre (2015), who examined a different time period (1996-2011) and found that a 10 per cent increase in energy price differences between countries leads to a 0.2 per cent increase in overall imports. The transitory nature of this effect suggests initial adjustment through international sourcing, followed by longer-term adaptation through technology upgrades or efficiency improvements. The top panel presents the ratio of emerging economy imports to total imported inputs, showing similar patterns in how import composition shifts. The bottom panel shows the ratio to total domestic supply, where effects become less significant, likely due to broader economic fluctuations affecting this measure of domestic production. These findings indicate that, while stricter carbon policies may lead European sectors to temporarily increase their reliance on carbon-intensive inputs from emerging economies, the effects dissipate over time rather than resulting in permanent shifts in sourcing patterns. This temporal pattern reveals important dynamics in how firms adapt to environmental regulations. Initially, firms appear to respond through the most flexible available channel – adjusting their international sourcing patterns to maintain competitiveness. However, as firms adapt over the longer term, they likely develop more sustainable solutions such as technology upgrades, efficiency improvements, or process innovations that reduce carbon intensity. This evolution from short-term trade adjustments to longer-term technological adaptation suggests that, while carbon leakage through supply chains may occur initially, firms ultimately find ways to maintain production while complying with stricter environmental regulations. The transitory nature of the effects also indicates that concerns about permanent production relocation or lasting damage to IDOS Discussion Paper 10/2025 13 domestic industry competitiveness may be overstated, though short-term adjustment costs remain important considerations for policy design. Figure 3: Impulse responses of import patterns to carbon policy shock (pct) Notes: This figure plots the impulse responses of input sourcing from emerging economies following a carbon policy shock, estimated using local projection. The top, middle, and bottom panels show the fraction of inputs that industry sector k in country i sources from emerging economies in year t+h, relative to total imported inputs, all inputs, and total domestic supply in carbon-intensive sectors, respectively. The x-axis represents years after the shock. Carbon policy shocks (pct) are extracted from the carbon policy surprises measured as euro change in carbon price, relative to the prevailing wholesale electricity price (Känzig, 2023). Solid lines represent point estimates, while darkand light-shaded areas indicate 90 per cent and 95 per cent confidence bands, respectively. Source: Author’s calculations IDOS Discussion Paper 10/2025 14 6.2 Heterogeneity analysis The heterogeneity of our results is examined across multiple dimensions, focusing on both sectorspecific and country-specific characteristics. Figure 4 illustrates the heterogeneous effects of carbon policies across European (importing) countries with different GDP levels. Lower-GDP countries (bottom tercile) experience a more pronounced increase in the ratio of carbon-intensive imports from emerging economies to total imported inputs. This cumulative effect peaks at over 0.5 per cent around year two post-shock, before gradually diminishing to zero by year four, suggesting a temporary but significant adjustment in sourcing patterns. In contrast, higher GDP countries (top tercile) show a slight decrease in this ratio, potentially indicating their greater capacity to maintain or even strengthen domestic production despite stricter environmental regulations. The contrast between these groups suggests that countries with lower GDPs are more vulnerable to carbon leakage effects, possibly due to the limited technological and financial resources required to adapt their production processes. This finding implies that economic size may play a crucial role in a country’s ability to maintain domestic production in the face of stringent carbon policies, with smaller economies potentially more susceptible to outsourcing carbon-intensive production to emerging economies. The divergent responses also highlight the importance of considering country-specific characteristics when designing carbon policies, as uniform regulations may have uneven distributional consequences across countries with different economic capacities. Figure 4: Heterogeneity by European (importing) countries’ GDP Notes: This figure plots the impulse responses of input sourcing from emerging economies following a carbon policy shock, estimated using local projection, for low GDP countries (orange), middle GDP countries (gray), and high GDP countries (blue). European countries are grouped into terciles based on their logged GDP (2015 constant) levels. The x-axis represents years after the shock. Carbon policy shocks (pct) are extracted from the carbon policy surprises measured as euro change in carbon price, relative to prevailing wholesale electricity price (Känzig, 2023). Shaded areas indicate 95 per cent confidence bands. Source: Author’s calculations IDOS Discussion Paper 10/2025 15 Figure 5 demonstrates heterogeneous effects based on European countries’ carbon tax status. Countries without national carbon taxes (in addition to the EU-wide ETS) show a larger increase in carbon-intensive imports from emerging markets compared to those with additional national carbon taxes. This divergence becomes particularly pronounced two years after the shock, suggesting a delayed but significant adjustment in sourcing patterns. This pattern mirrors the results in Figure 4, likely because carbon taxes have been primarily implemented in wealthier Western and Northern European countries, which typically have more resources to invest in cleaner production technologies. These findings indicate that the implementation of national carbon taxes, complementing the EU-wide Emissions Trading System (EU ETS), may enhance efforts to mitigate carbon leakage through multiple policy instruments working in tandem. Figure 5: Heterogeneity by European (importing) countries’ carbon tax status Notes: This figure plots the impulse responses of input sourcing from emerging economies following a carbon policy shock, estimated using local projection, for countries without national carbon taxes (purple), and countries with national carbon taxes (navy) on top of the EU-wide ETS system. The x-axis represents years after the shock. Carbon policy shocks (pct) are extracted from the carbon policy surprises measured as euro change in carbon price, relative to prevailing wholesale electricity price (Känzig, 2023). Shaded areas indicate 95 per cent confidence bands. Source: Author’s calculations Figure 6 examines whether sectors with different carbon intensities respond differently to carbon policy shocks. This analysis finds that sectors across all carbon intensity levels show similar responses, with high-carbon sectors being slightly more sensitive to carbon pricing. This pattern suggests that European carbon policies lead to broad-based shifts in sourcing patterns towards emerging economies, regardless of sectors’ emission intensities. Such homogeneous responses across sectors point to the possibility that supply chain networks transmit policy-induced adjustments throughout the industrial structure, rather than being confined to the most carbonintensive activities. IDOS Discussion Paper 10/2025 16 Figure 6: Heterogeneity by sectoral carbon intensity Notes: This figure plots the impulse responses of input sourcing from emerging economies following a carbon policy shock, estimated using local projection, for low carbon intensity (green), middle carbon intensity (gray), and high carbon intensity sectors (brown). The carbon intensity (pre-shock) is measured as CO2 emissions (in tons) per 2015 thousanddollar gross output. Sectors are grouped into terciles based on their carbon intensity. The x-axis represents years after the shock. Carbon policy shocks (pct) are extracted from the carbon policy surprises measured as euro change in carbon price, relative to prevailing wholesale electricity price (Känzig, 2023). Shaded areas indicate 95 per cent confidence bands. Source: Author’s calculations 7 Robustness 7.1 Responses to carbon tax The main identification strategy using carbon policy shocks is complemented with a “controlbased” approach following Metcalf and Stock (2020) and Känzig and Konradt (2023). This approach identifies the effects of carbon taxes by controlling for various economic and sectoral factors to isolate plausibly exogenous variations in carbon prices. This complementary analysis serves both as a robustness check and provides insights into how different types of carbon pricing policies might affect supply chain adjustments. The following (panel) local projection is estimated: Δℎy𝑖𝑖𝑖𝑖,𝑖𝑖+ℎ =α+βℎCtax𝑖𝑖𝑖𝑖,𝑖𝑖+� θ𝑝𝑝 ℎ𝑥𝑥𝑖𝑖𝑖𝑖,𝑖𝑖−𝑝𝑝 ′ 𝑃𝑃 𝑝𝑝=1 +σ𝑖𝑖 ℎ+ϵ𝑖𝑖𝑖𝑖,𝑖𝑖+ℎ, where the independent variable of interest Ctax𝑖𝑖𝑖𝑖,𝑖𝑖 is the carbon tax11 imposed on country i and sector k. The outcome variable is the fraction of carbon-intensive inputs that industry sector k in country i sources from emerging economies in year t+h, relative to all inputs (m=2). The vector 11 As in the World Carbon Pricing Database: net tax rate (accounting for exemption) in current local currency unit per tonne of CO2. IDOS Discussion Paper 10/2025 17 𝒙𝒙𝒊𝒊𝒊𝒊,𝒕𝒕−𝒑𝒑 ′ includes lagged control variables including the country’s GDP, sectoral carbon intensity, producer prices, and trade openness. These controls help account for various economic factors that might influence sourcing decisions independently of carbon taxation. Similar to the main regression, sector (destination country’s) fixed-effects 𝜎𝜎𝑖𝑖 ℎ is included to account for time-invariant characteristics that might affect the propensity to source from emerging economies. Figure A2 in the Appendix shows small but positive responses to carbon tax changes, with effects becoming significant in year 3. While these results align directionally with the main findings from the policy shock analysis, the effects are notably smaller in magnitude and take longer to materialise. The smaller and less significant effects likely reflect the gradual, anticipated nature of carbon tax changes compared to the unexpected policy shocks in our main specification. This difference in response patterns highlights the importance of the high-frequency identification strategy in capturing market responses to carbon policy changes, as it better isolates the immediate supply chain adjustments to policy innovations. The delayed response to carbon taxes also suggests that firms may have more time to plan and implement alternative adjustment strategies when facing gradual, predictable policy changes compared to sudden policy shocks. 7.2 Alternative environmental policy measure The environmental policy stringency (EPS) index is also employed as an alternative measure of environmental regulation. Developed by the OECD, this comprehensive index provides a quantitative assessment of environmental policy stringency across 40 countries from 1990 to 2020, capturing the extent to which national policies impose explicit or implicit costs on polluting activities. The index incorporates both market-based instruments (such as environmental taxes and trading schemes) and non-market regulations (including emission limits and research and development (R&D) subsidies), providing a broader perspective on environmental policy than carbon pricing alone. Figure A1 in Appendix shows the cross-sectional and temporal variations of the EPS index in European countries, highlighting substantial heterogeneity in environmental policy stringency across both countries and time. The following OLS regression is estimated: ImportShare𝑖𝑖𝑖𝑖,𝑖𝑖=α+βEPS𝑖𝑖,𝑖𝑖+θ𝑝𝑝𝑥𝑥𝑖𝑖𝑖𝑖,𝑖𝑖 ′+σ𝑖𝑖+ϵ𝑖𝑖𝑖𝑖,𝑖𝑖, where EPS𝑖𝑖,𝑖𝑖 is the environmental policy stringency index for European country i in year t, 𝒙𝒙𝒊𝒊𝒊𝒊,𝒕𝒕 ′ is a vector of control variables, 𝜎𝜎𝑖𝑖 represents sector fixed effects, and 𝜖𝜖𝑖𝑖𝑖𝑖,𝑖𝑖 is the error term. Table A2 in the Appendix reports the results from this regression. The estimated coefficients on EPS are positive and highly significant for import share variables (m=1, 2), indicating a robust relationship between environmental policy stringency and carbon-intensive imports from emerging countries. This aligns with the main regression results and provides additional evidence that stricter environmental policies may lead to supply chain adjustments through increased sourcing from countries with weaker environmental regulations. Moreover, additional tests are conducted using a 3-year moving average of the EPS index to capture more persistent policy effects. These results (reported in Appendix Table A3) show similar patterns, further supporting the robustness of the main results. 7.3 Sector composition While the main analysis compares responses across sectors of different carbon intensities, the robustness of the findings is further tested by excluding energy-intensive sectors. These sectors, including mining and quarrying (NACE B05-B08), petroleum products (C19), chemicals (C20), non-metallic minerals (C23), and basic metals (C24), typically show high sensitivity to carbon policies and often receive special regulatory treatment such as free allowance allocations in the IDOS Discussion Paper 10/2025 24 Table A2: Environmental policy stringency and import shares from emerging countries (1) (2) (3) VARIABLES ImportShare 1 ImportShare 2 ImportShare 3 EPS 2.608*** 1.299*** 9.203 (0.361) (0.251) (7.417) Log(GDP) 2.397*** 1.179*** 8.732 (-0.918) (0.190) (5.398) Log(carbon intensity) -0.918 0.190 26.08 (1.181) (0.900) (25.33) PPI -0.331*** -0.187*** -0.512 (0.0388) (0.0322) (0.626) Trade openness 0.239*** 0.157*** 1.601*** (0.0249) (0.0201) (0.433) Constant -22.38*** -5.626 -64.21 (7.167) (4.543) (67.99) Observations 13950 13951 13950 R-squared 0.290 0.315 0.160 Adjusted R-squared 0.286 0.311 0.155 Notes: ImportShare1,2,3 represent the share of imports over (1) total imported inputs, (2) total inputs, and (3) total domestic supply, respectively, in carbon-intensive sectors. Robust standard errors are clustered at the sector level. *, **, and *** indicate significance at the 10 per cent, 5 per cent, and 1 per cent levels, respectively. Source: Author’s calculations Table A3: Environmental policy stringency and import shares from emerging countries (1) (2) (3) VARIABLES ImportShare 1 ImportShare 2 ImportShare 3 EPS_3MA 2.419*** 1.284*** 14.06* (0.390) (0.276) (7.593) Log(GDP) 2.497*** 1.190*** 8.647 (0.290) (0.244) (5.639) Log(carbon intensity) -0.879 0.272 31.40 (1.201) (0.939) (26.42) PPI -0.387*** -0.215*** -0.708 (0.0438) (0.0365) (0.698) Trade openness 0.236*** 0.155*** 1.605*** (0.0248) (0.0201) (0.444) Constant -29.42*** -16.56*** -245.3* (6.106) (5.291) (143.7) Observations 13147 13148 13147 R-squared 0.295 0.320 0.163 Adjusted R-squared 0.291 0.316 0.158 Notes: ImportShare1,2,3 represent the share of imports over (1) total imported inputs, (2) total inputs, and (3) total domestic supply, respectively, in carbon-intensive sectors. Robust standard errors are clustered at the sector level. *, **, and *** indicate significance at the 10 per cent, 5 per cent, and 1 per cent levels, respectively. Source: Author’s calculations