Environmental asymmetries in global value chains: The case of the European automotive sector

Journal of Cleaner Production 449 (2024) 141606 Available online 4 March 2024 0959-6526/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/bync-nd/4.0/). Environmental asymmetries in global value chains: The case of the European automotive sector Hugo Campos-Romero a , * , ´ Oscar Rodil-Marz´ abal b , Ana Laura G´ omez P´ erez c a Universidade de Santiago de Compostela, ICEDE Research Group, Department of Applied Economics, Faculty of Economics and Business Sciences, Av. Do Burgo s/n, 15782, Santiago de Compostela, Spain b Universidade de Santiago de Compostela, ICEDE Research Group, CRETUS, Department of Applied Economics, Faculty of Economics and Business Sciences, Av. Do Burgo s/n, 15782, Santiago de Compostela, Spain c Instituto Polit´ ecnico Nacional, Escuela Superior de Economía, Av. Plan de Agua Prieta 66, Plutarco Elías Calles, Miguel Hidalgo, 11350, Ciudad de M´ exico, Mexico ARTICLE INFO Handling Editor: Giovanni Baiocchi JEL classification: F14 F23 Q51 Q56 Keywords: EU28 Automotive sector Global value chains Emissions Environmental Kuznets curve Pollution haven hypothesis ABSTRACT The European automotive sector is deeply integrated in the EU regional value chain. Each EU member show differences their economic and environmental performance. While central European countries perform more advanced, higher-value and lower-emission tasks, eastern economies specialize in manufacturing and have a higher share of exports from an environmental perspective. However, the environmental effects resulting from automotive production, as well as the distinctions among European economies based on their development levels, remain unexplored. Similarly, the implications of participation in global value chains (GVCs) in this context have not been examined. Therefore, the aim of this paper is to analyze the Environmental Kuznets Curve and the Pollution Haven/Halo Hypothesis in the case of the EU28 automotive sector under a value-added approach. The multi-regional input-output methodology along with a panel data estimation was used to verify the hypotheses. The results verify an inverted-N shape curve for the Kuznets hypothesis. Nevertheless, the results show no evidence of compliance with the haven or halo hypotheses. Moreover, participation in GVCs implies higher environmental impacts. Based on these findings, some policy recommendations are proposed to enhance the sector’s circularity and optimize material use throughout the production chain. 1. Introduction The past few decades have witnessed a profound transformation in the automotive sector. The impetus towards electric motor, driven by increasing environmental pressures and the looming prospect of fossil fuel depletion, has instigated substantial alterations in the configuration of the automotive Global Value Chain (GVC) (Beuse et al., 2018; Cs´ efalvay, 2020; Guzik et al., 2020). Recent developments, such as autonomous driving, real-time feedback, and the supply chain disruptions that emerged during the COVID-19 crisis are poised to further amplify this burgeoning “green” trend within the sector (Wu et al., 2021). The transition towards electric motors finds its primary motivation in the growing concern for environmental conservation, coupled with the escalating institutional apprehension regarding the depletion and subsequent increase in the cost of fossil energy resources. The European Union (EU) has reinforced its commitment to this technology as it formulates its strategy towards a Circular Economy (CE), achieving the distinction of being the second-largest region globally in terms of electric vehicle sales by 2022, surpassed only by China (IEA, 2023). While the widespread adoption of electric vehicles undoubtedly promises a substantial reduction in emissions from vehicular usage, several critical issues necessitate attention, with the most significant ones outlined below. Firstly, there is a need to address how energy generation will be augmented to support a future fleet of electric vehicles that will replace conventional combustion engines. Secondly, it is imperative to persist in the advancement of technology aimed at producing engines and batteries suitable for road transportation of goods. Lastly, careful consideration must be given to the environmental ramifications stemming from the production processes of various vehicle components and parts. In reference to the third point, it’s worth noting that despite the automotive sector’s active involvement in GVCs, its pivotal status and the logistical challenges associated with transporting certain bulky and * Corresponding author. E-mail address: [email protected] (H. Campos-Romero). Contents lists available at ScienceDirect Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro https://doi.org/10.1016/j.jclepro.2024.141606 Received 12 November 2023; Received in revised form 8 February 2024; Accepted 29 February 2024 Journal of Cleaner Production 449 (2024) 141606 2 heavy components have necessitated the establishment of domestic factories in most EU countries. Within the context of intra-regional trade, owing to the full commercial integration among member nations, the sector’s trade interactions hold significant importance. Consequently, recent integration initiatives have spurred a rise in foreign investments and exports within the automotive sector, particularly benefiting the Eastern European Economies (EEE) (Ambroziak, 2016). Nonetheless, certain theories within the realm of international trade and its environmental consequences highlight that, under the presumption of unrestricted movement of goods and capital, processes of production relocation can emerge. These relocations might be driven by the pursuit of lower labor costs, less stringent environmental regulations, or other locational advantages, potentially resulting in a net upsurge in emission levels. Among the most noteworthy theories in this context are the Environmental Kuznets Curve (EKC) hypothesis, the Pollution Haven Hypothesis (PHH), and the Pollution Halo Hypothesis (HP) (Balsalobre-Lorente et al., 2019; Destek et al., 2018; Guzel and Okumus, 2020). In the European context, if these hypotheses were to hold true, it would suggest that the EEE could potentially maintain a relatively higher environmental impact compared to the more developed economies of the EU. This could be attributed to an increase in their economic activity, largely bolstered by integration processes, while simultaneously upholding less stringent environmental regulations and possessing lower technological capacities. Furthermore, it’s important to consider the role that each nation plays within the production chain of specific sectors. According to the theories mentioned earlier, these countries might assume responsibility for the relatively more emissionintensive production stages (Martínez-Zarzoso et al., 2017; Rodil-- Marz´ abal and Campos-Romero, 2021; Campos-Romero et al., 2023). Despite the large number of papers analyzing these hypotheses, various gaps have been identified in the literature. First, in terms of geographical scope, most of the studies are focused on developing countries, maybe under the presumption of better environmental performance in developed economies. However, a detailed examination of trade and investment dynamics reveals that it’s not inherently evident that developed countries, including European ones, have successfully decoupled economic growth from emissions. Indeed, in developed regions, very diverse economies of different levels of development coexist, so there is an emerging interest in examining their performance from this perspective. Secondly, regarding the sectoral level, most studies address these hypotheses from a broad, aggregate standpoint, overlooking potential variations across different sectors. This proposal, however, narrows its focus to the automotive sector. This choice is motivated by two main reasons: the sector’s unique integration into GVCs, which is predominantly regional, and its critical role in European industry with extensive production networks. Another significant gap in the literature relates to the scarcity of studies analyzing the implications of participation in GVCs through the lenses of the aforementioned hypotheses, particularly from a sectoral perspective. A fourth gap has to do with the limited literature exploring the intersection between regional economic integration, the role of EEE, and their specific environmental impacts under the EKC, PHH, and HP related to the automotive sector. Hence, the primary aim of this paper is to undertake an analysis of the trade dynamics within the European automotive sector, considering both economic and environmental aspects. This analysis relies on trade statistics that capture value-added components as well as trade data and related carbon dioxide emissions (CO 2 ). Beyond addressing the EKC, PHH, and HP hypotheses, this paper also aims to explore whether there is a distinct differentiation between the more recently integrated Eastern European countries and other countries within the integrated area. These countries often exhibit characteristics akin to the economic periphery within this sector (Pavlínek, 2022). This research offers several outstanding contributions. Firstly, it addresses a significant gap in the literature, as mentioned earlier, by enhancing the understanding of emissions perspectives and trends in EU countries, whether due to an increase in economic activity (EKC) or a higher volume of foreign investment (PHH/HP) in automotive production. Secondly, it considers the differences in the territorial distribution of labor and production through a center-periphery framework, particularly examining the distinct role played by EEE in the automotive value chain. Thirdly, it provides a set of economic policy recommendations aimed at strengthening the technological fabric of the European Union and promoting greater environmental responsibility from a mesoinstitutional perspective. Lastly, it integrates the perspectives of the Kuznets environmental curve and pollution haven hypotheses to provide a holistic understanding of the collective effects of foreign trade in conjunction with foreign investment flows, which is particularly pertinent to the automotive sector. To address these questions, the research employs a comprehensive approach. It commences with a descriptive analysis of the European Union’s automotive sector, encompassing its key characteristics and principal foreign trade patterns, with a focus on the value-added perspective. Additionally, the study provides insights into the emissions produced by the automotive industry within the member countries. Subsequently, an econometric panel data model is proposed to explore whether there is a discernible increase in emissions within the sector, correlated with its expanding involvement in GVCs, particularly in the EEE. For data acquisition to facilitate both analyses, the research leverages the multi-regional input-output methodology (MRIO), enabling the examination of intersectoral flows between different countries. In this context, particular attention is given to the trade flows within the automotive sector among the EU countries. The input-output approach facilitates the tracking of trade flows in GVCs related to automotive production. Additionally, it enables the analysis of its environmental effects through satellite accounts. Combined with the econometric model proposed in Section 3, this approach allows for the effective testing of the Kuznets, pollution, and halo hypotheses. The study relies primarily on two key databases. Firstly, TiVA (Trade in Value Added, OECD) serves as a fundamental reference database for analyzing trade in value-added components. It features an environmental extension that furnishes data on the environmental impact of major trade flows, quantified in terms of CO 2 emissions. Secondly, Eora is a comprehensive multi-regional input-output database encompassing sectoral information for a total of 190 countries over an extended period. Its satellite accounts enable the measurement of emissions associated with trade flows, facilitating a detailed understanding of environmental implications. In addition to these primary databases, the research draws upon supplementary sources such as Eurostat and the World Bank to gather essential information required for the study, including data on FDI flows and stock, population statistics, and others. In addition to the Introduction, the paper consists of four sections. Section 2 comprehensively examines pertinent literature concerning the characteristics of the European automotive sector, its integration into GVCs, and its ecological footprint. Section 3 delineates the data sources and methodology employed in this study. Section 4 provides an analytical commentary and discussion on the primary research findings, while Section 5 ultimately presents the core conclusions of the research. 2. Literature review The automotive industry emerges as one of the sectors with the most extensive involvement in international trade (Gorgoni et al., 2018). Dominated by a small number of prominent multinational corporations, it holds a pivotal and strategic role in the economic development of nations. In the context of the EU, it made a substantial contribution, constituting approximately 8% of GDP in 2022. The sector also maintained a favorable trade balance, primarily driven by vehicle exports to countries beyond the Union. Overall, it generated more than 7% of the total European employment and over 11% of manufacturing employment specifically (ACEA, 2023). H. Campos-Romero et al. Journal of Cleaner Production 449 (2024) 141606 3 Nonetheless, despite its expansive international reach, it can be argued that the automotive industry exhibits a more regional than global orientation, given that major companies predominantly sell their vehicles within the same region where they operate (Rugman and Collinson, 2004). In the case of the EU, the integration process has further deepened this regionalization trend within the sector (Brown et al., 2021; Pavlínek, 2020). The automotive sector serves as a notable example of the escalating regionalization within international relations. This trend is particularly prominent in the so-called factory regions of Europe, North America, and Asia, which constitute trade blocs characterized by a high degree of integration (Baldwin, 2006; Baldwin and Lopez-Gonzalez, 2014). The heightened regionalization of GVCs does not imply the absence of significant trade interactions between these blocs. However, it’s noteworthy that following the COVID-19 pandemic, regionalization processes have intensified, accompanied by a resurgence of certain protectionist measures. Governments and companies are taking such actions to mitigate the vulnerabilities exposed within GVCs during the pandemic (Enderwick and Buckley, 2020; Steinberg and Tan, 2023; Wang and Sun, 2021). A recent trend that aligns with the regional dynamics of the automotive industry is the growing significance of emerging economies in the production of high-value auto parts. Nations such as China and those situated in Eastern Europe have gained increasing prominence in the manufacturing of complex components for the sector (Gorgoni et al., 2018; Markiewicz, 2020). Furthermore, in addition to providing higher value-added components, suppliers in these regions have also developed capabilities more closely associated with R&D processes than traditional manufacturing processes (Guzik et al., 2020). However, it’s crucial to note that the major multinational corporations headquartered in more advanced economies persist in steering the industry’s developmental trajectories and trends. Several characteristics of the automotive sector exert a regional development influence, encompassing both production tasks and postsales functions. The imperative to foster trust-based relationships and proximity with primary suppliers necessitates an organizational structure in the form of business clusters. These clusters not only economize transportation costs but also facilitate effective communication and the establishment of formal and informal connections among various actors within the production chain. This optimization of organizational and manufacturing processes is well-documented (Albulescu et al., 2021; Lugo-S´ anchez and Guzm´ an-Anaya, 2021; Manzano-Ramos and Guzm´ an-Anaya, 2020; Mendoza Cota, 2021). Furthermore, geographic distance can adversely impact the quality of inputs sourced from external suppliers, especially when dealing with innovative or highly technologically advanced components (Bray et al., 2018). The research conducted by Chen et al. (2020) underscores how the cluster-based organizational structure, characteristic of the automotive industry, can play a pivotal role in promoting sustainable development patterns. This becomes particularly evident when the leading companies within these clusters initiate the adoption of environmental policies. Such actions serve as incentives for their primary suppliers, compelling them to adjust to the evolving production standards. Considering this, it is worthwhile to explore policies aimed at fostering the establishment of networks among companies. This approach would enable those with advanced technological capabilities to disseminate their knowledge and expertise to other firms within the cluster, thereby facilitating the adoption of sustainable business practices. The automotive sector serves as a lens to examine the operations of GVCs and the trends of regionalization in international relations. Consequently, the study of environmental impacts associated with this production structure (Meng et al., 2018; Wang et al., 2019; Wu et al., 2020; Campos Romero and Rodil Marz´ abal, 2021), characterized by extensive processes of productive fragmentation that inevitably entail a substantial volume of transportation movements –whose numbers escalate with greater supply chain fragmentation– has gained significance in recent decades (Rodrigue, 2006). Numerous studies highlight a correlation between emission levels and specific stages within the production chain. It’s observed that during the initial and final phases of a typical chain, encompassing activities like R&D, design, logistics, marketing, and after-sales, the associated emissions volume tends to be lower than in the central phases, which are more closely linked to manufacturing tasks (Shih, 1992; Campos Romero and Rodil Marz´ abal, 2024) (self-citation). Sectors with a substantial presence in GVCs tend to locate manufacturing activities in developing regions (Ge et al., 2020; Guzel and Okumus, 2020; Mahmood et al., 2023c; Saqib et al., 2023; Zhao et al., 2020). As companies in these regions develop new capabilities, they initiate upgrading and scaling processes. These processes not only enable them to secure a greater share of the value added within the overall value chain but also contribute to enhanced energy efficiency and reduced environmental emissions (Liu et al., 2018). From this can be derived the first hypothesis of analysis: H1. A greater insertion of the automotive industry in global value chains implies a higher volume of associated emissions. Although much of the literature focus on assessing the impact of the automotive sector from the perspective of road traffic, examining the effects stemming from its production process is equally crucial(Andr´ es and Padilla, 2018; Fan et al., 2018; Mazur et al., 2018; Moro and Lonza, 2018).In terms of the manufacturing impact, emissions within the sector would experience a notable escalation if we were to consider not only the emissions stemming from transportation but also those arising from the manufacturing processes. Consequently, emissions associated with the production of electric vehicles could surpass those of internal combustion vehicles. Additionally, we must account for the environmental repercussions of generating the necessary energy to sustain an expanding electric vehicle fleet. Hence, the utilization of renewable energy in conjunction with electric vehicles emerges as a pivotal strategy for achieving a substantial reduction in emissions within the electric vehicle sector (Kawamoto et al., 2019; Wang et al., 2013; Yan, 2009). From these considerations regarding the impact of producing transportation components with international trade, questions arise concerning the influence of manufacturing activity locations. In accordance with the EKC hypothesis, regions characterized by lower income levels are anticipated to experience an increase in their emissions as they progress in development (Copeland and Taylor, 2001, 2004; Mahmood et al., 2023a, 2023b; Yasmeen et al., 2019; Zafar et al., 2019). On the other hand, the PHH posits that within a context of unimpeded mobility of goods and capital, coupled with variations in environmental regulations on a global scale, businesses tend to position emission-intensive activities in regions with less stringent environmental regulations (Copeland and Taylor, 1994; Gill et al., 2018). Consequently, offshoring procedures could lead to an influx of FDI into destination countries and/or an augmentation of GVC participation rates within specific sectors, contingent on the chosen offshoring strategy. While extensive discourse exists in the literature regarding the validation of these hypotheses (Destek et al., 2018; Shao et al., 2019), numerous studies have put them to the test in specific regions and sectors (Balsalobre-Lorente et al., 2019; L´ opez et al., 2018; Mahmood, 2020, 2022, 2023; Pata et al., 2023; Rana and Sharma, 2018), primarily through the application of econometric models. In the case of the PHH, its authenticity has often been contingent on whether the pursuit of regions with less stringent environmental regulations genuinely serves as the primary rationale for offshoring. Nevertheless, these differences in environmental regulations, if present, constitute just one among several factors considered in decision-making processes concerning the location of economic activities. Other factors, such as the search for reduced labor costs or proximity to new markets, among various others, are also significant determinants (Barbieri et al., 2018; Buckley and Ghauri, 2004). H. Campos-Romero et al. Journal of Cleaner Production 449 (2024) 141606 4 The application of both theories to the automotive sector in the European context prompts questions about whether the EEE maintain relatively higher and increasing emission levels in comparison to more developed economies. It also raises inquiries into whether they might have evolved into potential pollution havens for emission-intensive sectors. Both of these issues are addressed in the second hypothesis: H2. The different pattern of participation in the global automotive industry chain explains the different emissions intensity between the EEE and the rest of European economies. Despite the process of European integration, which entails the harmonization of environmental regulations, initial disparities in technological capabilities could lead to elevated emission rates. Additionally, the lower labor costs in these countries, coupled with the advantages of goods and capital exchange facilitated by economic integration, may have incentivized the localization of manufacturing activities within the EEE (Pavlínek, 2023). This relocation, irrespective of specific environmental regulations, may bring about consequences in line with the predictions of the PHH. The different effects on growth and FDI when differentiating between EEE and other EU countries are included in the third hypothesis: H3. The increase in the stock of Foreign Direct Investment (FDI) and economic growth leads to an increase in the overall volume of emissions. Based on the literature reviewed, Table 1 provides a methodological detail of the cited papers, which focus on the EKC, the PHH/HP or both. While the explanatory variables vary according to each paper’s specific focus, the overall methodological framework for testing these hypotheses remains uniform. For EKC, the analysis typically involves estimating quadratic or cubic models in relation to GDP per capita and emissions, aiming to identify either an inverted ’U’ or an ’N’ shaped curve. In the context of PHH and HP, the inclusion of FDI and its squared term is generalized, aimed at analyzing an inverse ’U’ shaped relationship between some measure of environmental impact and FDI inflow. In the case of European countries, compliance with the PHH/HP may be uncertain due to the harmonization of environmental regulation linked to the integration process. The model outlined in Section 3 (see expressions (4) and (5)) is developed based on the literature analyzed, particularly drawing from studies that concurrently analyze the EKC and the PHH/HP. As an innovative element and in addition to per capita CO 2 , the emissions intensity in introduced as an explanatory proxy variable related to the environmental impact of the sector. 3. Data and methodology This research introduces several innovations in the examination of the environmental repercussions of foreign trade. Firstly, it adopts a mesoeconomic approach focused on the automotive sector, renowned for its substantial involvement in regional value chains. Secondly, it combines the analysis of the EKC with the simultaneous exploration of the PHH and HP. This investigation spans an extensive timeframe, encompassing data from the majority of EU28 countries over the period of 1995–2015, employing a value-added trade perspective. Thirdly, the study delves into the environmental consequences of the European automotive sector’s engagement in GVCs. Lastly, it places particular emphasis on Eastern European countries, aiming to ascertain whether the developmental disparity between this group and the rest of the EU has given rise to a form of pollution haven within the European Union. This study relies on an input-output analysis, following the methodology initially developed by Leontief (1951, 1937) adapted to a multiregional scale. Multiregional input-output analysis has gained prominence in recent decades, largely due to the availability of MRIO databases. However, the foundation for this methodology was laid by Leontief and Strout (1963) for analyzing trade among U.S. Data for this study have been sourced from various databases. Data pertaining to trade flows in terms of value added and environmental impact has been obtained with MRIO methodology from the Eora database (Casella et al., 2019; Lenzen et al., 2012, 2013). This method ins described below. Eora database was chosen for its extensive analytical timeframe and additional data on environmental impacts, energy consumption, and other pertinent variables in its satellite accounts. Additional variables, such as the sectoral stock of FDI, 1 population figures for each country, and average annual exchange rates have been obtained from Eurostat, the World Bank, and the International Monetary Fund, respectively. Los datos de IED a escala sectorial pueden obtenerse desde Eurostat con un elevado nivel de desglose. Para el caso del sector de la automoci´ on, no se han encontrado problemas de datos en esta variable. The analysis of trade flows in terms of value added, primarily based on the work by Koopman et al. (2010, 2014) enables the dissection of a country’s gross exports into as many as nine value-added components, which represent the various trade flows comprising GVCs. The methodology employed to derive the flows used in this study is elaborated upon below. For a set of c countries and n sectors, Tcn×cn is defined as the intermediate transactions matrix, Fc×c as the final demand matrix and Xcn×c is the total output vector, obtained as the sum of matrices T and F. Thus, the basic input-output relationships are the following: A=T X−1;L= (I−A)−1(1) Where Acn×cn is the matrix of technical coefficients of production,  X−1 cn×cn is a diagonal matrix containing the elements of the total output matrix on the main diagonal, the other elements being null; Lcn×cn is the Leontief inverse matrix and Icn×cn is the identity matrix. The value-added coefficients, v1×cn, are needed to subsequently obtain the trade flows relative to the GVCs. The vector of value-added coefficients is obtained by means of the matrix of technical coefficients: v=u−uA (2) Where u1×cn is a vector made up of "ones" that can also be used to perform sums by columns or by rows, in which case a vector ucn×1 would be preferred. Following Koopman et al. (2010, 2014), it is possible to decompose each element of the total output matrix according to where each value-added flow is absorbed. Note that in  Xcn×cn the elements lying on the n×n dimension diagonal represent each country’s domestic interand intra-industry exchanges, while the non-diagonal elements represent foreign trade flows, thus omitting the diagonal of  X yields a gross export matrix. To define export flows in terms of value added consider any three countries, s, d and z; and any three sectors j, k and r; such that s∕= d∕= z but intra-sector trade is possible, so j=k=r. Hence, we define the following three flows of exports in terms of value added considering a total of n columns: EXPFs=∑ n, s∕=dvs jLss jk Fsd kj ;EXPIs=∑ n, s∕=dvs jLsd jk Fdd kj +∑ n, s∕=d∕=zvs jLsd jk Fdz kr (3) The element EXPFs in equation (3) represents the so-called traditional trade, i.e. value added generated in sector j of country s, processed as an intermediate good in the domestic economy and then exported to sector j of country d as a final good. The element EXPIs represents simple and complex GVC trade. Simple GVC involves value added generated in sector j of country s, exported as an intermediate good to sector k of country d, where it is further processed and consumed as a final good. 1 Since sectoral FDI data is only available for the period 2008–2012, the sectoral FDI stock for the remaining years has been estimated from the 5-year sectoral average share and the total stock of FDI per country. H. Campos-Romero et al. Journal of Cleaner Production 449 (2024) 141606 5 Table 1 Literature review summary. Authorship Variables Sample Countries Period Method Hypothesis verified? Balsalobre-Lorente et al. (2019) Ecological footprint, FDI, GDP per capita, renewable energy consumption, and share of urban population to total Mexico, Indonesia, Nigeria, and Turkey (MINT) 1990–2013 FMOLS, DOLS Pollution halo Campos Romero and Rodil Marz´ abal (2024) Consumption and production-based emissions per capita, GDP per capita, global value chain position and participation, and renewable energy consumption East and Southeast Asian countries 1995–2018 Fixed effects panel data model Verifies EKC with differences on the emissions perspective considered Campos-Romero et al. (2023) FDI, CO 2 intensity, GDP per capita, exports, energy intensity, and spatial dummy European Union 1995–2015 Multi-level random effects model Pollution haven De Beule et al. (2022) Locational variable (dummy indicating insideoutside EU), carbon inefficiency (emissions over sales by carbon price), emission-to-cap ratio, size of the investment, market-to-book ratio, return on assets, and a dummy indicating exposure to carbon leakage 358 multinationals under the European Union Emission Trading Scheme 2013–2020 Random-effects LOGIT Pollution haven Destek et al. (2018) Ecological footprint per capita, GDP per capita, renewable energy consumption, and trade openess European Union 1980–2013 Panel data Verifies EKC Ge et al. (2020) FDI, 11th and 12th Five Year Plan SO2 reduction policy (interactive term between emission targets, actual emissions, and a period dummy), environmental enforcement, industry technology level, and export transactions China, province level 2001–2015 Difference-indifference-indifferences analysis (DDD) with panel data Pollution haven Guzel et al. (2020) CO 2 per capita, FDI share of GDP, and energy use per capita Indonesia, Malaysia, Philippines, Singapore, and Thailand 1981–2014 CCEMG and AMG models Pollution haven L´ opez et al. (2018) CO 2 , emissions embodied in exports, emissions avoided by imports, balance of avoided emissions NAFTA, BRIIAT, Eurozone, Rest of EU, East Asia, and China 1995–2009 Multi regional Input Output Pollution haven Mahmood (2020) FDI, CO 2 per capita, GDP per capita, financial market development, trade openess 21 North American countries (North America and the Caribbean) 1990–2014 Pooled OLS, panel data, and spatial model No effect found on FDI (monotonic), verify Environmental Kuznets Curve Mahmood (2022) FDI, CO 2 per capita (territory and consumption based), GDP per capita, exports and imports (% of GDP), financial market development Gulf Cooperation Council countries 1990–2019 Spatial Durbin model Pollution halo, also verify EKC Mahmood (2023) FDI, CO 2 per capita, GDP per capita, exports and imports (% of GDP), and financial development 18 Latin American countries 1970–2019 Pooled OLS and panel data (fixed effects) Pollution haven, also verify EKC Mahmood et al. (2023a) Literature review analysis China – – Most studies verify EKC Mahmood et al. (2023b) Literature review analysis. Relevance of including renewable energy consumption into the models No specific country – – Most studies verify EKC, especially when panel data is used compared to country cases Mahmood et al. (2023c) FDI, Consumption-based CO 2 per capita, GDP per capita, and exports and imports (% of GDP) 12 MENA countries 1995–2020 Panel data (fixed effects), and Spatial Durbin model Pollution haven, also verify EKC Martínez-Zarzoso et al. (2017) Dirty exports, clean exports, total environmental tax revenues, and GDP European Union 1999–2013 Panel data analysis Porter hypothesis Pablo-Romero et al. (2017) Gross value-added per capita and per hour, total transport energy use, household transport energy use, and productive transport energy use European Union 1995–2009 Panel data Verifies EKC Pata et al. (2017) CO 2 emissions for air transport, maritime transport, rail transport, and road transport; GDP per capita, air transport energy consumption per capita, maritime transport energy consumption, rail transport energy consumption, and road transport energy consumption 13 European Union countries 1995–2019 Two-stage panel data Verifies EKC (depending on transport mode) Rana and Sharma (2017) FDI, GDP, CO 2 , and trade (exports and imports) India 1982–2013 Dynamic multivariate Toda-Yamamoto Pollution haven, also verify EKC Saqib et al. (2023) FDI, ecological footprint, GDP per capita, Energy Structure, Renewable energy consumption, human capital 16 European countries 1990–2016 Cross-Sectional AGRL model Pollution halo, also verify EKC Shao et al. (2019) FDI, CO 2 per capita, GDP per capita, energy consumption per capita, trade openness, and urbanization BRICS, and MINT countries 1982–2014 Panel vector error correction, and panel co-integration Pollution halo Zafar et al. (2019) CO 2 per capita, renewable energy consumption, nonrenewable energy consumption, GDP, and trade openness 18 emerging economies 1990–2015 Panel data Verifies EKC Zhao et al. (2020) CO 2 per capita, CO 2 intensity, industry CO 2 emissions (%), employment in carbon intensive industries (%), regional carbon emissions (%), China, province level 2000–2014 Mediating effect model Pollution haven (continued on next page) H. Campos-Romero et al. Journal of Cleaner Production 449 (2024) 141606 6 Finally, Complex GVC involves those value-added flows that pass through at least three countries. In this case, it is value added generated in sector j of country s, exported as an intermediate good to sector k of country d, where it is processed and subsequently exported to sector r of country z, where it is consumed as a final good. To be precise, it’s important to note that value-added re-exports, referring to flows that circle back to the country of origin, whether as a final or intermediate good, are also categorized as GVC trade flows. Nonetheless, they constitute only a negligible fraction of an economy’s overall value-added exports and have consequently been omitted from the analysis for the sake of simplicity. In terms of econometric estimation, the study proposes two models derived from the literature, as detailed in Table 1. Expression (4) introduces a novel aspect by utilizing emission intensity as a proxy to measure the environmental pollution resulting from car production in the EU. Meanwhile, Expression (5) presents the conventional model for estimating the Kuznets hypotheses as well as PHH and HP. Both models exhibit similarities, differing primarily in the specification of the dependent variable. The first model employs the natural logarithm of emissions intensity, quantified as the ratio of carbon emissions to GDP. Conversely, the second model utilizes carbon emissions per capita as the dependent variable. Furthermore, ln (GDPpc)ct represents the natural logarithm of GDP per capita, while ln (GDPpc)2 ct and ln (GDPpc)3 ct denote its squared and cubed values, respectively. These additional elements are included to assess the presence of the EKC. Specifically, a positive sign for the first element, coupled with a negative sign for its square, is indicative of an EKC relationship. If the cubic element exhibits a positive sign, it implies the presence of a waveform relationship. Table 2 presents the descriptive statistics for the selected variables. Similarly, in line with the approach for the emissions-to-GDP per capita ratio, this analysis also integrates the natural logarithm of the ratio of FDI to GDP along with its squared term. This inclusion aims to capture the possible existence of an inverted U-shaped curve in the relationship between these variables. Depending on the positioning of each country on this curve, it could potentially verify the presence of the PHH or the HP. Furthermore, the analysis encompasses a set of variables designed to capture the influence of different trade flows in terms of value added. Specifically, ln (EXPF)ct denotes the natural logarithm of the first expression as defined in equation (3), which pertains to the traditional trade Flow. ln (EXPI)ct represents the sum of the other two flows in Table 1 (continued) Authorship Variables Sample Countries Period Method Hypothesis verified? intensity of environmental regulation, GDP per capita, FDI, secondary industry output in regional GDP, ration of net fixed assets and number of employees, investment in net fixed assets, and investment on carbon intensive fixed assets to total investment (%) Source: Authors Table 2 Data sources and descriptive statistics, 504 observations per variable. Average Standard deviation Min Max Definition Source GDPpc 578.1559 1636.801 11.9395 29960.87 GDP per capita (deflated) Eora GDPpc 2 3008067 4.28e+07 142.5519 8.98e+8 Square of GDPpc Eora GDPpc 3 6.73e+10 1.23e+12 1702 2.69e+13 Cubic of GDPpc Eora FDIGDP 0.7998 1.1027 0.005 7.834 Stock of foreign direct investment to GDP Eurostat (FDI) and Eora (GDP) FDIGDP 2 1.8534 6.2278 0.0000301 61.3729 Square of FDIGDP Eurostat (FDI) and Eora (GDP) EXPF 16.4109 7.9343 3.208 58.9648 Direct exports of final goods Eora EXPI 14.666 7.4003 3.5895 59.2594 Direct and indirect exports of intermediate goods Eora ENERGDP 0.003 0.005 0.0002 0.0438 Energy intensity (ratio of energy consumption to GDP) Eora URB 70.706 11.649 50.622 97.876 % or urban population The World Bank R&D 1.463 0.878 0.352 3.874 % of R&D expenditure over GDP The World Bank Note 1: The statistics have been obtained from the original data before logarithmic conversion. Note 2: The conversion from euros (Eurostat data) to dollars (Eora data) has been made using the annual average exchange rate of the International Monetary Fund (IMF). Note 3: Cyprus, Luxembourg, Malta, and Portugal have been excluded from the econometric analysis due to large data gaps in some of the variables. Source: Authors from Eora and Eurostat Model I :ln (CO2GDP)ct =β0+β1ln (GDPpc)ct +β2ln (GDPpc)2 ct +β3ln (GDPpc)3 ct +β4ln (FDIGDP)ct +β5ln(FDIGDP)2 ct +β6ln (EXPF)ct +β7ln (EXPI)ct +β8ln (ENERGDP)ct +β9ln(URB)ct +β10ln(R&D)ct + ε ct (4) Model II :ln (CO2pc)ct =β0+β1ln (GDPpc)ct +β2ln (GDPpc)2 ct +β3ln (GDPpc)3 ct +β4ln (FDIGDP)ct +β5ln(FDIGDP)2 ct +β6ln (EXPF)ct +β7ln (EXPI)ct +β8ln (ENERGDP)ct +β9ln(URB)ct +β10ln(R&D)ct + ε ct (5) H. Campos-Romero et al. Journal of Cleaner Production 449 (2024) 141606 7 Expression (3), which correspond to trade flows within both simple and complex GVC. Additionally, the natural logarithm of the ratio of energy consumption to GDP is introduced as a control variable. This variable is anticipated to exert a significant and positive influence on carbon emissions, whether measured as emissions intensity or emissions per capita. 4. Results and discussion The EU28 holds a significant position in automotive production, commanding a 16.2% share of global production in 2022, as reported by the International Organization of Motor Vehicle Manufacturers (OICA). Nevertheless, its influence in the global market has gradually waned over time, mirroring a similar trend observed in the union formed by the United States, Mexico, and Canada through the T-MEC treaty, previously NAFTA. In the early 2000s, both regions contributed slightly over 30% each to global automotive production. However, they subsequently ceded ground to the Asian market and Oceania, which accounted for more than 58.8% of total automotive production (in terms of manufactured units) by 2022. It’s worth noting that China accounts for nearly 29% of total global automotive production by itself. When examining the value-added generated (see Fig. 1), it becomes evident that there was a decline in the contribution of both the EU and North America to global automotive production during the 2004–2008 Fig. 1. Share of automotive value added contributed by EU28, North America and Asia and Oceania in world production, 1995–2018. Source: Authors from TiVA (Trade in Value Added, OECD, 2021 edition) Fig. 2. Percentage of exported domestic content over value added generated in the automotive sector (top) and percentage of exported domestic content over gross exports of the automotive sector (bottom), EU28, North America and Asia and Oceania, 1995-2018 Note: When calculating the exported domestic content, the domestic value added that returns to the country of origin for final consumption has been discounted. Intraregional trade is not included. Source: Authors from TiVA (Trade in Value Added, OECD, 2021 edition) H. Campos-Romero et al. Journal of Cleaner Production 449 (2024) 141606 8 period. Furthermore, since the years preceding the 2008 financial crisis, the major Asian economies have significantly increased their contribution to the sector, accounting for approximately 41% of global production by 2018. It’s important to note that the disparity in contribution in terms of value-added is somewhat smaller than the disparity in units produced. This discrepancy may arise from variations in the technological sophistication of vehicles manufactured in each region. Vehicles produced in the European and North American markets tend to possess a higher intrinsic value compared to those produced in the Asian market. This difference in technological sophistication may explain the somewhat narrower gap observed in terms of value-added contribution. In terms of trade insertion patterns, Fig. 2 illustrates the percentage of domestically produced content exported relative to the value added generated by the automotive sector and gross exports. This initial metric exhibits variations depending on the region under consideration (see Fig. 2 top). Following the 2008 financial crisis, North America’s contribution remained relatively stable, while Asia and Oceania experienced a decline in their share of the region’s economy. Conversely, the percentage of domestic content exported in relation to the EU28’s value added saw an increase following the crisis. Turning to the second metric, as shown in Fig. 2 bottom, it underscores a distinguishing characteristic of the automotive sector in comparison to other industries. The proportion of domestically generated value added that is exported within gross exports exceeded 70% in Asia and Oceania and exceeded 80% in both the EU28 and North America. This lower reliance on foreign value added reflects the inherently regional nature of the sector’s GVCs, as previously discussed. Nonetheless, a prevailing long-term trend becomes apparent, whereby the exported domestic value added constitutes a diminishing proportion of the overall gross exports. This signifies an expanding reliance on foreign value added within the automotive exports of each region, with Asia and Oceania contributing to 28% of this foreign value added. The mounting significance of this share in Asia can be attributed to the necessity to import high-tech components due to a shortage in internal technical capabilities for their development (Kobayashi et al., 2015; Nag, 2009). Within the EU28 countries (Table 3), a notable disparity becomes evident in the sector’s significance within each member country. Countries such as Germany, Hungary, the Czech Republic, and the Slovak Republic make substantial contributions, exceeding 5% of the total value added in each respective country. In contrast, the remaining nations contribute less than 1%, with a particularly pronounced difference observed in countries like Ireland and the smaller regions (Luxembourg, Malta, and Cyprus), which exhibit a stronger presence in specific service activities. These discrepancies extend to trade indicators, especially in those countries where the automotive sector constitutes a larger portion of national value added and plays a significant role in their exports. Nonetheless, even in countries with a relatively modest contribution to GDP (considering only direct activities in this context), the automotive sector represents a substantial fraction of exports. This holds true whether analyzed in gross terms or when variables are selected based on value added, addressing the issue of double counting often found in traditional statistics (Koopman et al., 2010, 2014). Among exports measured in terms of value-added, those categorized as DVA exported as final goods stand out in comparison to those exported as intermediate goods. However, when evaluating the foreign trade activities of the entire EU28 region in terms of value-added, extraregional exchanges take precedence over intra-regional ones –a significant shift compared to raw trade statistics. Drawing from data sourced in the TiVA database, intra-regional trade, when assessed within the context of traditional export statistics, constitutes 37% of the total (which exceeds extra-regional trade). However, when using value-added statistics, the situation is reversed, especially in the case of exports of final goods, where extra-regional trade surpasses intra-regional trade by 65%. In contrast, the difference between extra-regional and intra-regional exports of intermediate goods is only 10%, with extra-regional exports holding a slight advantage. These data highlight a distinctive aspect of the European automotive sector’s international market engagement. Given its inherent nature, the Table 3 The automotive sector in the EU28. Value added, gross exports and domestic value added (DVA) exported, 2018 VA automotive share of total VA (%) Gross automotive exports as a share of total gross exports (%) Automotive DVA exported over total DVA exported (%) Automotive DVA exported as intermediate goods over total intermediate DVA exported (%) Austria 1.72 8.61 6.08 4.68 Belgium 0.92 5 3.34 2.01 Bulgaria 0.82 2.38 2.11 1.81 Croatia 0.46 1.95 1.45 1.56 Cyprus 0.03 0.09 0.09 0.03 Czech Republic 5.8 26.73 19.9 15.47 Denmark 0.27 0.96 0.96 0.64 Estonia 0.74 2.8 2.12 0.99 Finland 0.65 3 2.68 1.52 France 1.49 14.87 11.07 7.96 Germany 5.21 18.64 17.4 12.31 Greece 0.22 0.35 0.4 0.27 Holland 0.87 3.74 2.92 1.82 Hungary 5.25 23.11 15.16 13.24 Ireland 0.15 0.52 0.19 0.13 Italy 1.62 9.01 8.09 6.53 Latvia 0.44 1.9 1.49 0.89 Lithuania 0.66 1.53 1.3 0.91 Luxembourg 0.08 0.02 0.03 0.01 Malta 0.56 0.44 0.48 0.33 Poland 2.32 11.75 8.89 7.13 Portugal 1.6 10.57 6.86 5.73 Romania 3.57 15.93 13.73 10.77 Slovak Republic 5.42 30.83 19.72 14.61 Slovenia 2.3 10.51 6.69 5.11 Spain 1.79 13.01 10.02 8.1 Sweden 3.4 14.44 12.4 8.14 United Kingdom 1.69 10.19 8.49 5.44 EU28 2.43 – – – EU28 (extra) – 11.40 11.12 7.58 EU28 (intra) – 11.78 8.52 6.63 Note: Foreign trade data for each member country include intraand extraregional trade. Source: Authors from TiVA (Trade in Value Added, OECD, 2021 edition) Table 4 Emissions intensity of the automotive sector (tons of CO 2 per thousand dollars of GDP), 1995–2015. 1995 2015 1995 2015 Slovak Republic 7.45 1.46 Portugal 1.72 0.45 Poland 10.20 1.25 Belgium 1.57 0.43 Hungary 10.50 1.24 Slovenia 2.29 0.42 Luxembourg 2.15 1.10 Finland 1.27 0.38 Bulgaria 4.24 1.06 Ireland 1.39 0.34 Lithuania 6.58 0.88 United Kingdom 1.53 0.30 Romania 12.97 0.86 Spain 1.14 0.29 Croatia 1.61 0.67 France 1.03 0.27 Latvia 7.88 0.57 Italy 1.21 0.25 Czech Republic 4.02 0.56 Sweden 0.89 0.23 Greece 1.50 0.55 Holland 0.66 0.22 Estonia 9.05 0.51 Germany 0.39 0.17 Cyprus 1.72 0.49 Denmark 0.27 0.09 Austria 1.20 0.46 Malta 0.01 0.05 Source: Authors from Eora H. Campos-Romero et al. Journal of Cleaner Production 449 (2024) 141606 9 automotive industry exhibits a more regional orientation rather than a global one, as depicted in Fig. 2. Interestingly, despite this regional focus, most of the value-added exports generated in the EU find their way to other countries. Turning to the environmental dimension of the automotive sector, Table 4 illustrates the emissions intensity of the sector in EU28 countries for the years 1995 and 2015. The trend in the emissions intensity of the automotive sector mirrors the broader pattern observed in the European economy, declining over time due to ongoing technological advancements in energy efficiency and continual reductions in emissions. In terms of each country’s impact, both in 1995 and 2015, the EEE emerged as the nations with the highest levels of emissions per dollar of production, partially verifying H3. Of particular significance are the results for Slovak Republic, Poland, and Hungary, which also show a substantial participation of the automotive sector in their economies. It has been previously pointed out that H3 has been partially verified. In one hand, Eastern European countries continue to maintain higher emission intensity levels. On the other hand, over the 20 years analyzed, the average emission intensity of the EEE in the automotive sector has been drastically reduced, narrowing its differential with the average emission intensity for the remaining EU28 countries (see Appendix A). Table 5 highlights the disparities between the sector’s economic and environmental impacts, unveiling significant variations in the member countries’ contribution to intraand extra-regional exports of automotive intermediate goods when viewed from the perspectives of valueadded or emissions. Significant disparities in the sector’s export share become evident when examining it from both economic and environmental perspectives. These discrepancies, particularly pronounced in the instances of Poland, the Czech Republic, the Slovak Republic, and Hungary, could be attributed to their roles in the European regional value chain. As per the EKC and PHH, the lower level of economic development and the initial disparities (prior to economic integration) may have led these countries, including those in EEE, to undertake the more emission-intensive tasks within the automotive sector. The adoption of such tasks, coupled with slower technical progress, could account for the variations in emissions intensity and the proportion of intermediate goods exports within the automotive sector, depending on whether these aspects are analyzed from an economic or environmental standpoint. To examine the presence of EKC, PHH, and HP within the automotive sector, we present the following econometric models. Table 6 outlines the results of Model I in Expression (4) for all 24 countries included in the estimation, as well as two additional estimations for EEE and non-EEE countries. Table 7 outlines the results of Model II in Expression (5), also differentiated the previous 3 cases. Table 6 shows the estimation outcomes of Model I across three scenarios: all countries, EEE, and non-EEE. Table 7 shows the results for the same models with CO 2 per capita as the dependent variable. In neither model across any scenario is evidence found to support the PHH or HP. However, an inverse linear relationship is observed between the ratio of sectoral FDI stock inflows to GDP and both dependent variables. These findings imply that increased investments in the European automotive Table 5 Share of each member country on intraand extra-regional exports of automotive intermediate goods in terms of ADV and CO 2 exported, 2015 Domestic DVA exported as intermediate goods CO 2 exported as intermediate goods Germany 28.30 Germany 15.31 France 14.90 France 13.10 Poland 2.01 Poland 8.18* Italy 9.27 Italy 7.60 Spain 7.95 Spain 7.59 United Kingdom 6.53 United Kingdom 6.42 Czech Republic 3.30 Czech Republic 5.98* Belgium 4.18 Belgium 5.83* Holland 8.07 Holland 5.67 Austria 3.39 Austria 5.09* Slovak Republic 0.84 Slovak Republic 4.01* Sweden 4.88 Sweden 3.59 Hungary 0.81 Hungary 3.26* Romania 0.87 Romania 2.44* Portugal 1.09 Portugal 1.58* Finland 1.21 Finland 1.50* Slovenia 0.40 Slovenia 0.55* Ireland 0.39 Ireland 0.44* Lithuania 0.13 Lithuania 0.38* Bulgaria 0.10 Bulgaria 0.33* Denmark 0.94 Denmark 0.27 Estonia 0.14 Estonia 0.23* Luxembourg 0.06 Luxembourg 0.22* Greece 0.09 Greece 0.16* Croatia 0.06 Croatia 0.14* Latvia 0.03 Latvia 0.06* Cyprus 0.03 Cyprus 0.05* Malta 0.04 Malta 0.01 Note: The symbol "*" identifies those countries whose share in intraand extraEuropean intermediate goods exports is higher in terms of emissions than in DVA. Source: Authors from Eora Table 6 Model I estimation results. Dependent variable: emissions intensity (CO 2 to GDP ratio). All countries (24) EEE countries (11) Non-EEE countries (13) Coefficient t-value Coefficient t-value Coefficient t-value ln(GDPpc) 0.316 1.54 0.907 2.09** −1.256 −2** ln(GDPpc) 2 −0.197 −5.73*** −0.345 −3.94*** 0.064 0.72 ln(GDPpc) 3 0.010 5.52*** 0.020 3.56*** −0.003 −0.61 ln(FDIGDP) −0.163 −9.21*** −0.160 −7.25*** −0.081 −2.34** ln(FDIGDP) 2 −0.036 −8.05*** −0.043 −6.38*** −0.015 −2.03** ln(EXPF) −0.275 −3.98*** −0.340 −4.03*** −0.013 −0.1 ln(EXPI) 0.193 2.83*** 0.264 3.16*** −0.061 −0.47 ln(ENERGDP) 0.100 3.91*** 0.020 0.57 0.189 5.21*** ln(URB) −1.952 −6.75*** −1.006 −1.62 −2.650 −7.75*** ln(R&D) −0.137 −3.66*** −0.093 −2.09** −0.320 −4.54*** constant 11.589 8.72 6.257 2.32 18.274 9.53 R 2 within 0.94 0.96 0.91 R 2 between 0.58 0.00 0.29 R 2 overall 0.61 0.28 0.36 rho 0.98 0.98 0.98 N. observations 504 231 273 Period 1995–2015 1995–2015 1995–2015 Note: “***”, “**” and “*” indicates significance at 1%, 5% and 10% respectively. Source: Authors from Eora, the World Bank, Eurostat, and IMF H. Campos-Romero et al.