Closer and greener? Firms' global value chain participation and sustainability

Closer and greener? Firms’ global value chain participation and sustainability Inaugural-Dissertation zur Erlangung des akademischen Grades eines Doktors der Wirtschaftsund Sozialwissenschaften der Wirtschaftsund Sozialwissenschaftlichen Fakult¨at der Christian-Albrechts-Universit¨at zu Kiel Vorgelegt von: Finn Ole Semrau, M.Sc. aus Henstedt-Ulzburg Kiel, 2024 Gedruckt mit Genehmigung der Wirtschaftsund Sozialwissenschaftlichen Fakult¨at der Christian-Albrechts-Universit¨at zu Kiel Erstbegutachtung: Prof. Aoife Hanley, Ph.D. Zweitbegutachtung: Prof. Dr. Katrin Rehdanz Drittbegutachtung: Prof. Dr. Eric Strobl Tag der m¨undlichen Pr¨ufung: 28.03.2023 Contents 1 Introduction 1 2 Firms’ global value chain participation and its environmental performance - A review of the empirical literature 11 2.1 Introduction....................................... 12 2.2 Firms’ global value chain participation and environmental efficiency in production 13 2.3 Firms’ global value chain participation and environmental innovation adoption . 17 2.4 Main results and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 3 Stepping up to the mark? Firms’ export activity and environmental innovation in 14 European countries 28 3.1 Additional appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 4 On the Drivers of Clean Production: Firms’ Global Value Chain Positioning 42 4.1 Introduction....................................... 44 4.2 Firm’s GVC Positioning and Clean Production . . . . . . . . . . . . . . . . . . . 47 4.3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 4.3.1 Data Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 4.3.2 Empirical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 4.3.2.1 Measuring a Firm’s Clean Production . . . . . . . . . . . . . . . 53 4.3.2.2 Measuring a Firm’s GVC Positioning . . . . . . . . . . . . . . . 54 4.3.2.3 Measuring Exposure to Environmental Policy Stringency in Export Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 4.3.3 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 4.4 Results.......................................... 57 4.4.1 A Firm’s GVC Positioning and Clean Production . . . . . . . . . . . . . 57 4.4.2 Instrumental Variable Approach . . . . . . . . . . . . . . . . . . . . . . . 60 4.4.3 The Moderating Roles of Export Status and Exposure to EPS . . . . . . 62 4.4.4 Robustness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 4.5 Conclusion ....................................... 65 4.6 Appendix ........................................ 68 4.6.1 Calculation of CO2emissions . . . . . . . . . . . . . . . . . . . . . . . . . 68 4.6.2 Additional statistical outputs . . . . . . . . . . . . . . . . . . . . . . . . . 70 i 5 Corporate Social Responsibility along the Global Value Chain 76 5.1 Introduction....................................... 78 5.2 A Model of CSR along the Global Value Chain . . . . . . . . . . . . . . . . . . . 82 5.2.1 Incomplete Contractibility of CSR . . . . . . . . . . . . . . . . . . . . . . 82 5.2.2 Setup: Preferences and Technology . . . . . . . . . . . . . . . . . . . . . . 84 5.2.2.1 Preferences, CSR, and Perceived Ethical Quality . . . . . . . . . 84 5.2.2.2 Production of Physical Output . . . . . . . . . . . . . . . . . . . 85 5.2.3 Sourcing with Sequential Production . . . . . . . . . . . . . . . . . . . . . 86 5.2.3.1 Relational vs. Anonymous Sourcing . . . . . . . . . . . . . . . . 86 5.2.3.2 Demand and Firm Revenue . . . . . . . . . . . . . . . . . . . . . 87 5.2.3.3 Hold-Up and Bargaining . . . . . . . . . . . . . . . . . . . . . . 88 5.2.3.4 Incremental Contribution at Stage m............... 88 5.2.3.5 Bargaining along the Value Chain . . . . . . . . . . . . . . . . . 89 5.2.4 CSR Investments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 5.2.5 Equilibrium CSR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 5.2.5.1 Industry Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . 91 5.2.5.2 Equilibrium CSR along the Global Value Chain . . . . . . . . . 92 5.2.6 Discussion and Link to the Data . . . . . . . . . . . . . . . . . . . . . . . 93 5.2.6.1 Testable Implications: Partial vs. Industry Equilibrium . . . . . 93 5.2.6.2 A Direct Measure of the Non-Contractible Variable . . . . . . . 93 5.3 GVC Position and CSR in India . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 5.3.1 Data Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 5.3.2 Measuring CSR Spending along GVCs . . . . . . . . . . . . . . . . . . . . 95 5.3.2.1 Measuring a Firm’s CSR Spending . . . . . . . . . . . . . . . . . 95 5.3.2.2 Measuring a Firm’s Value Chain Position . . . . . . . . . . . . . 96 5.3.3 Empirical Specification............................. 98 5.3.3.1 Empirical Set-Up . . . . . . . . . . . . . . . . . . . . . . . . . . 99 5.3.3.2 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . 100 5.3.3.3 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . 101 5.3.4 Robustness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 5.3.4.1 Smoothed CSR Spending . . . . . . . . . . . . . . . . . . . . . . 103 5.3.4.2 Controlling for Visibility . . . . . . . . . . . . . . . . . . . . . . 104 5.4 Conclusion .......................................104 5.5 Theory Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 5.5.1 Industry Equilibrium with Incomplete Contracts . . . . . . . . . . . . . . 107 5.5.2 Complete Contracts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 5.5.2.1 Level of CSR with Complete Contracts and Comparison to Incomplete Contracts . . . . . . . . . . . . . . . . . . . . . . . . . 109 5.5.2.2 Remaining Industry Equilibrium Results . . . . . . . . . . . . . 109 5.5.3 The Extended Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 5.5.3.1 Incomplete Contracts . . . . . . . . . . . . . . . . . . . . . . . . 111 5.5.3.2 Complete Contracts . . . . . . . . . . . . . . . . . . . . . . . . . 112 ii 5.6 Empirical Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 5.6.1 Wages:......................................113 5.6.2 OECD export exposure: . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 Erkl¨arung zum selbst¨andigen Verfassen der Arbeit und Einhaltung der DFGRichtlinien 121 Acknowledgement 124 iii Index of abbreviations CIS Community Innovation Survey CMIE Centre for Monitoring Indian Economy CSR Corporate Social Responsibility EI Environmental Innovation EPS Environmental Policy Stringency GHG Greenhouse gas GVC Global value chain IEA International Energy Agency IfW Kiel Institute for the World Economy IPCC Intergovernmental Panel on Climate Change KCG Kiel Centre for Globalization NAFTA North American Free Trade Agreement PPML Poisson pseudo maximum likelihood SDG Sustainable Development goal WIOD World Input-Output Database iv Chapter 1 Introduction The year 2015 was a pivotal year for global sustainable development. By agreeing on the 2030 Agenda for Sustainable Development and the Paris Agreement, the global community committed to sustainable development at all levels of governance. Unquestionably, both agreements are an urgent call for sustainable development, including actions related to social and environmental domains. It is now the responsibility of all countries to ensure that the stated goals are reached. At the heart of the Paris Agreement is the objective to hold “the increase in the global average temperature to well below 2°C above pre-industrial levels and pursuing efforts to limit the temperature increase to 1.5°C above pre-industrial levels” (Paris Agreement, 2015, p.3). A failure of this goal would cause tremendous climate-related hazards to ecosystems and humans. Among many others, these include losses in biodiversity, physical water scarcity, an increasing number of extreme weather events, more widespread diseases such as dengue, a rise in sea levels threatening coastal areas and even entire countries and an increased pressure on food production and access (IPCC, 2022). Even facing these tremendous consequences, it seems as if the world is far off-track reaching the goal to limit global warming to well below 2°C. Nevertheless, any effort to limit global warming is desirable, as many climate-related risks increase disproportionately with rising temperature (IPCC, 2022). The fact that the global community still has a long way to go is shown, for example, by the development of CO2emissions over time. Figure 1.1 illustrates the development of CO2emissions from fossil fuels over the last two centuries split by world region. Between 2019 and 2020, global CO2emission decreased slightly from 36.7 billion tonnes to 34.81 billion tonnes. However, this can be explained by the economic slowdown related to the global COVID-19 pandemic. In general, the figure shows that global CO2emissions continue to rise and have not yet peaked. Between 2000 and 2020, CO2emissions increased by 38% from 25.23 billion tonnes to 34.81 billion tonnes. Copeland et al. (2022) emphasise that most emission growth originates from developing and emerging countries. This pattern is also visible in the figure. Undoubtedly, China is the elephant in the room. China’s CO2emissions increased by an astonishing 210% from 3.44 billion tonnes in 2000 to 10.67 billion tonnes in 2020. In 2020, China announced its target to peak CO2emissions by 2030 and achieve carbon neutrality before 2060 (International Energy Agency, 2021a). These targets require China to decouple its development from emissions at an earlier stage of development than other advanced economies in the past, including the US, which has historically been the largest contributor to global CO2emissions. Even as India walks in China’s footsteps, its successes in green transformation set the pace for 1 Figure 1.1: Graphical illustration by Ritchie et al. (2020) of the global CO2emissions from fossil fuels split by region based on data of the Global Carbon Budget (2021) the world to tackle climate change. India’s CO2emissions increased by 149% from 0.98 billion tonnes to 2.44 billion tonnes between 2000 and 2020. As a result, India is now the third largest emitter of CO2emissions, although its per capita emissions are significantly lower compared to the historical major emitters in Europe and North America. All in all, the increase in emissions in India and China illustrates that global success in combating climate change is urgently needed not only in industrialised countries but also in emerging and developing countries. At the same time, emerging and developing countries are especially vulnerable to the risks of climate change (IPCC, 2022). Addressing climate-related risks poses additional challenges to developing countries than they already face in addressing social-related sustainable development goals, such as eradicating poverty and hunger, improving education and achieving health and well-being. In this doctoral thesis, I focus on the role of firms to reach sustainable development. More precisely, I contribute to the understanding of how firms’ global value chain (GVC) participation relates to their sustainability, covering environmental and social performance. GVCs are a mechanism, where firms from different countries are connected by the transfer of goods, capital and workers. These countries, in turn, are characterised by different stages of development and patterns of technology, knowledge and legislation relating to social and environmental aspects. Several studies show that GVC participation, such as by exporting, can be accompanied by improvements in the environmental performance of firms (e.g., Batrakova and Davies (2012); 2 Forslid et al. (2018); Girma and Hanley (2015); Richter and Schiersch (2017)). Among other things, I contribute to the literature by analysing how the characteristics of firms’ country of origin and the country to which a firm is linked through participation in GVCs, moderate the respective relationship between GVC participation and sustainable performance. Certainly, understanding the relationship between firms’ GVC participation and sustainability is only one part of a complex network of how economic activity relates to sustainable development. In their flagship report, the World Bank (2020) outlined the rise of GVCs after the 1990s and how it contributed to growth, jobs, and the reduction of poverty. On the one hand, the report points to environmental degradation, such as outsourcing dirty production steps to countries with weak environmental regulation stringency. On the other hand, they emphasise that GVC participation might lead to technology transfer, the upgrading of standards and the transmission of intangible assets, such as good managerial practices. Overall, I follow Copeland et al. (2022) who emphasise the importance for decision makers of understanding the complex relationship between globalisation and the environment. Chapter 2, 3 and 4 analyse how firms’ GVC participation relate to improvements in environmental performance.1The diffusion of clean technologies and the related green transformation of industries is directly linked to the Paris Agreement and SDGs 9 and 12. The International Energy Agency (2021b) emphasises that the technologies needed to achieve the necessary deep cuts in global emissions by 2030, already exist. What is missing is an immediate and massive deployment of clean and energy-efficient technologies, already deployed or available in industry. The transmission of these technologies is especially important for firms based in developing and emerging countries, characterised by being - on average - further away from the technology frontier (Guti´errez and Teshima, 2018). As outlined before, developing and emerging countries represent the main source of emission growth, are more vulnerable to climate risks and need to decouple economic growth from environmental degradation at an earlier stage of development than advanced economies. To begin with, in Chapter 2, I summarise the relevant empirical literature on the relationship between firms’ GVC participation and environmental performance.2By discussing the related literature, the literature review also presents the main contributions of Chapter 3 and Chapter 4. At the core of the literature review is a discussion of two relatively separate literature strands. Applying insights from both literature strands, I identify nine key findings and possible research gaps. The first literature strand, starts with the seminal contribution by Grossman and Krueger (1991), focusing on the impact of trade on the environment. It is followed by many other contributions, using aggregated-data (Grossman and Krueger, 1995; Chichilnisky, 1994; Copeland and Taylor, 1994; Antweiler et al., 2001). More recently, studies using firm level data have been gaining momentum (Cherniwchan, 2017). By using firm level data, re1Please note that the writing style of each chapter differs due to the different requirements of the journals to which the manuscripts have been submitted. 2The chapter is an earlier version of “Semrau, F. O. (2023). “Chapter 8: Firms’ global value chain participation and its environmental performance: a review of the empirical literature.” Handbook on Innovation, Society and the Environment edited by Edited by Fernando J. D´ıaz L´opez, Massimiliano Mazzanti, and Roberto Zoboli, 125-139.” The published manuscript is accessible via https://doi.org/10.4337/9781802200065.00016. Since the chapter includes a detailed discussion of the literature relevant for the Chapters 3, 5 and 4, it substitutes to some extend the literature review typically part of an introduction. 3 Timmer, M. P., E. Dietzenbacher, B. Los, R. Stehrer, and G. J. Vries (2015): “An illustrated user guide to the World Input–Output Database: The case of global automotive production,” Review of International Economics, 23, 575–605. World Bank (2020): “World Development Report 2020: Trading for development in the age of global value chains,” A world bank group flagship report, World Bank Group, Washington. 10 Chapter 2 Firms’ global value chain participation and its environmental performance - A review of the empirical literature∗† Finn Ole Semrau Kiel Institute for the World Economy (IfW) Kiel Centre for Globalization (KCG) finn-ole.semr[email protected] Abstract Global diffusion of previously developed clean and energy efficient technologies is continuing to define the pace of the international transition to a green economy over the next few years. Empirical firm-level analyses offer a powerful approach to investigate how a firm’s global value chain participation relates to the adoption of green technologies. In this chapter, I review insights referring to two relatively separated strands of literature. First, scholars from the international economics literature provide strong evidence for an environmental efficiency premium from exports, foreign direct investments and proximity to final consumption. Second, scholars analysing the determinants of environmental innovation adoption go beyond efficiency arguments and cover a diverse set of environmental domains. Results from the latter are ambiguous and crucially depend on the sample under investigation. I conclude the review with the identification of nine key findings and a discussion of possible research gaps. JEL Classification Numbers: F18, F64, O33, Q55, Q56 Keywords: Trade and environment; Environmental/ eco/ green innovation; technological change; global value chains; impacts of globalization; literature review ∗The chapter is an earlier version of: “Semrau, F. O. (2023). “Chapter 8: Firms’ global value chain participation and its environmental performance: a review of the empirical literature.” Handbook on Innovation, Society and the Environment edited by Fernando J. D´ıaz L´opez, Massimiliano Mazzanti, and Roberto Zoboli, 125-139.” The published article is accessible via https://doi.org/10.4337/9781802200065.00016. For citation please refer to the published article. Further, the chapter should not be reused or posted independently of the thesis without further permission. †I would like to thank Aoife Hanley, Peter Kannen, Jakob Lehr, Massimiliano Mazzanti, Frauke Steglich and Joschka Wanner for valuable comments during the writing of this chapter. 11 2.1 Introduction Immediate and massive deployment of already developed clean and energy efficient technologies crucially define the pace for deep cuts in global emissions by 2030 (International Energy Agency, 2021). Veugelers (2012) emphasises the urgency to run a clean innovation machine at full speed in the light of the sizeable climate challenge we face. Global diffusion of the available green technologies is necessary to reach full steam ahead in the green economic transition, e.g. reducing greenhouse gas emissions and other environmental negative impacts. In this chapter, I review empirical research on how firms’ global value chain participation, including exports, foreign ownership, positioning along a global value chain and indirect exposure to international trade via import competition, is associated with firms’ environmental performance. In so doing, I identify nine key findings and discuss possible remaining research gaps. Overall, the impact of globalisation on the environment is ambiguous (Cherniwchan et al., 2017; Copeland et al., 2022). The empirical use of firm-level data offers a powerful approach to better understand the interface between globalisation and the environment, since it enables researchers to account for the heterogeneity of firms, even in narrowly defined industries (Cherniwchan et al., 2017; Melitz, 2003).1 Summarising two main strands of empirical literature forms the core of this chapter. Scholars part of the first literature strand, I characterise as belonging to the international economics literature, frequently adopt the perspective of the Melitz (2003) model to analyse induced effects caused by global value chain participation. These effects include energy consumption, CO2 emissions and emissions from other pollutants. Scholars from the second literature strand emanate from the innovation literature. Here, the focus is on external and internal factors driving a firm’s decision to adopt environmental innovation and green management tools.2Surprisingly, both strands of literature are relatively unconnected. Accordingly, matching theoretical and empirical frameworks represents a promising avenue for future research. The international economics literature widely agrees on firm-level environmental efficiency gains induced by different forms of global value chain participation and exposure to trade such as exports, import competition, foreign acquisition and proximity to final consumption (e.g. Batrakova and Davies (2012); Brucal et al. (2019); Cherniwchan et al. (2017); Guti´errez and Teshima (2018); Semrau (2022)). Conversely, the results on how a firm’s global value chain participation links to the diffusion of environmental innovation is less clear-cut and results depend on factors such as local embeddedness, the idiosyncrasies of a firm’s home country, the type of environmental innovation and whether non-innovative active firms are excluded from the empirical analysis (e.g. Cainelli et al. (2012); Del R´ıo et al. (2016); Ghisetti et al. (2015); Hanley and Semrau (2022)). Some aspects, like positioning in a value chain or import competition are 1Analysing the whole effect of trade on the environment goes beyond the scope of this literature review. To this end, I refer to a very recent contribution on globalisation and the environment by Cherniwchan et al. (2017) and Copeland et al. (2022). 2I follow Ghisetti and Pontoni (2015) and use the term environmental innovation rather than eco-innovations. Even though many scholars use the terms synonymously, eco-innovations were initially conceived as a subclass of environmental innovations, affecting a firm’s economic and environmental performance. Environmental innovation, per se, limits the focus to a firm’s environmental performance. 12 widely neglected so far. The next section discusses the stated research strands in more detail. 2.2 Firms’ global value chain participation and environmental efficiency in production The discussion on how trade affects the environment goes back to the seminal contributions of Grossman and Krueger (1991), Grossman and Krueger (1995)), Chichilnisky (1994), Copeland and Taylor (1994), Antweiler et al. (2001) and Frankel and Rose (2005). Generally, the aggregated effects of trade on the environment can be split into a scale-, compositionand technique effect. First, trade increases the scale of economic activity that translates into a higher environmental footprint. Second, trade liberalisation can shift the place of production in line with a country’s comparative advantage that has an ambiguous effect on the environment. Third, the technique effect refers to trade-related changes in the environmental costs of production, e.g. driven by the adoption of more efficient machinery and equipment. Theoretical predictions on the total effect of trade on the environment depend on the determinants of a country’s comparative advantage (Cole and Elliott, 2003). Accordingly, empirical investigation needs to be guided by our current understanding regarding the environmental impacts of globalisation. In this chapter, I widely limit the discussion to empirical analyses based on firm-level data. Global value chains involve splitting production processes across different countries and allows firms to specialise in tasks. The Melitz model predicts that more productive firms self-select into exporting (Melitz, 2003). Bustos (2011) extends the theoretical model of Melitz to technology choice and predicts that export-related increases in revenues enables exporters to upgrade technology, e.g. increase spending on computers, software, technology transfers, patents, and innovation activities. The proposed mechanism is empirically shown for Argentinean firms for the period 1992 to 1996, a period characterised by a sharp drop of tariffs between Argentina and Brazil due to the trade liberalisation of MERCOSUR. Batrakova and Davies (2012) apply the theoretical predictions to environmental-related outcomes, more precisely to energy consumption of Irish manufacturing firms between 1991 and 2007. They find that export entry is positively connected with energy use, while it is negatively associated with energy intensity. Interestingly, they distinguish between high and low energy consumers. In particular high energy consumers can offset higher energy demand by efficiency gains. Cui et al. (2016) analyse the relation between export and environmental performance for U.S. firms in 2002, 2005 and 2008. They find that more productive firms report lower emission intensities (sulphur dioxide (SO2), carbon monoxide (CO), ozone (O3) and totals suspended particulates (TSPs) per value of sales). Conditioned on firms’ productivity, exporting firms environmentally perform better compared to firms only serving the domestic market. Richter and Schiersch (2017) find evidence supporting an exporter’s environmental efficiency premium - measured as CO2productivity. The used German firm-level data span the period 2003 to 2011 and contain detailed information on a firm’s fuel consumption that allows for a calculation of direct CO2emissions from production. They find that an increase in export 13 intensity by one percentage point, leads to average CO2emission efficiency gains by 0.22%. Forslid et al. (2018) theoretically and empirically shed light on the mechanism as for how a firm’s export activity relates to CO2efficiency when emissions are subjected to an environmental tax. After export entry, firms scale-up production in order to meet foreign demand. This scalingup of production enables firms to spread fixed production costs across an increased number of units. Incentivised by exposure to environmental policy regulations, the fixed costs of abatement can also be allocated among more units, increasing the attractiveness of investments in CO2 efficiency. They back up their theoretical predictions with empirical evidence using data of Swedish manufacturing firms for the time period 2004 to 2011. Holladay (2016) employs data concerning toxic chemical emissions from U.S. manufacturing establishments, for the period 1990 to 2006. He finds that exporters generate less pollution emissions compared to non-exporters from the same industry once output levels are considered in the equations. In an additional finding, he reveals that import competition reduces the domestic prices of varieties and drives the least productive and more pollution-intensive establishments out of business. Shapiro and Walker (2018) show that between 1990 and 2008 the air pollution emissions from the U.S. manufacturing sector decreases by 60% - primarily driven by within-product changes in emission intensity induced by environmental regulation rather than on changes in productivity or trade. However, Cherniwchan (2017) focuses on effects related to trade liberalisation and finds empirical evidence for a clean-up of the U.S. manufacturing induced by the North American Free Trade Agreement (NAFTA). For the period 1994 to 1998, he finds that nearly two-thirds of the reduction of particular matter (PM10) and sulphur dioxide (SO2) emissions can be attributed to the trade liberalisation. The effect is driven by two main sources, namely improved export opportunities to the Mexican market and improved access to imports from Mexico. The latter effect is relatively larger in magnitude and indicates a high importance of import competition for the clean-up of firms’ environmental performance. Kong et al. (2022) point to a bias in the literature towards using firm-level data from developed countries and investigate the firm-level environmental effects related to the Chinese WTO accession in 2001 – a date that allows them to label China as a developing country and emphasising substantial differences between developing and developed countries. They find that the trade liberalisation significantly reduced firms’ toxic emissions - including sulfur dioxide (SO2), soot and dust, nitrogen oxide (NOx) - through the scale of production, export opportunities, technology upgrading, intermediate input and within-firm mechanisms. Guti´errez and Teshima (2018) build on Cherniwchan (2017) and manifest the positive environmental effect of import competition. They combine Mexican plant-level and satellite imagery data for the time period 2000 to 2003 and emphasise that firms in developing countries can benefit disproportionately from trade-related positive impacts on technological choice. This is because these firms are on average further removed from the technological frontier or best management practices. Their key results reveal that import competition increases energy efficiency, but is likely to induce a decrease in abatement expenditures. This latter prediction suggests a technology uptake, reducing the need for expenditure on abatement. They also argue that the literature on global value chain participation and environmental efficiency is biased towards 14 developed countries. A key reason for this bias is the availability of data containing concrete and reliable measures of environmental performance at the firm-level. The Indian firm-level Prowess data represent a popular source for investigations into several empirical aspects of international economics (e.g. De Loecker et al. (2016); Goldberg et al. (2010a,b); Topalova and Khandelwal (2011)). As a rare feature in the context of developing countries, the data cover panel information on a firm’s energy inputs. This allows researchers to calculate the direct CO2emissions embodied in production, similar to the approach used by Richter and Schiersch (2017) or Forslid et al. (2018) for firms from developed countries. Barrows and Ollivier (2018) even apply the methodology at the product-level, investigating how a firm’s product-mix affects the intensity of CO2emissions. They find that emission efficiency decreases with distance to the core competency, where core competency is proxied by the product with the highest sales rank. Exporters are likely to sell their core competency on foreign markets, but more productive firms are also able to diversify their foreign sales and these products may indeed be dirtier compared to products characterised by the core competency. To this end, firm-level analysis on environmental efficiency that fails to account for changes in the product-mix, can be misleading. In line with this, Barrows and Ollivier (2021) contribute a further study to shed light on within-product CO2efficiency gains. They find that increasing foreign demand induces exporters to scale-up production and induces efficiency gains at the product-level. The latter finding can be interpreted as a technique effect, strongly backing prior findings that export activity links to environmental gains following the adoption of clean technologies. Mart´ınez-Zarzoso et al. (2020) analyse the relation between exports and energy consumption of Indian firms between 1987 and 2016. This period covers India’s 1991 economic liberalisation and accession to the WTO in 1995. In line with prior analyses, they find that exporters spend more on energy, but their production can be characterised as energy efficient. In particular, firms from high energy consuming industries can incorporate trade-related efficiency gains. These high energy users are more likely to be located in an upstream position along the global value chain, bringing them relatively closer to the raw material of a value chain (Semrau, 2022; Shapiro, 2021). A negative correlation between energy consumption and selling finished goods indicates such a pattern at the Indian firm-level Mart´ınez-Zarzoso et al. (2020). However, global value chain positioning does not comprise the core of the main analysis. Additionally, the measure of global value chain positioning does not account for the global value chain positioning of products sold as intermediates. Antr`as et al. (2012) and Antr`as and Chor (2018) show how World Input-Output tables can be used to calculate a country’s industry global value chain positioning in a respective year. This measure considers sales directly sold to final consumers and the proximity to final consumption of sales sold as intermediates. By applying the respective global value chain measure, Shapiro (2021) reveals an environmental bias of trade policy: Upstream industries emit substantially higher levels of CO2per unit of output, while facing lower trade tariffs. The pattern between CO2and upstreamness is somewhat monotone and can be observed over a wide range of different countries. Generally, upstream processes are more energy and capital demanding, e.g. the extraction of raw materials from the ground and downstream stages disproportionately use labour as a direct input, which is generally cleaner. International trade policy serves as 15 an indirect subsidy for these dirty industries, since upstream industries successfully lobby for reduced tariffs in order to reduce input costs. Based on Shapiro’s findings, Copeland et al. (2022) state as a stylised fact on trade and the environment the pattern that dirty industries are more likely to be upstream. Semrau (2022) analyses the relation between upstreamness and dirty production using the Indian Prowess database. By doing so, he applies the global value chain measure of Herkenhoff et al. (2021), who take the industry-level global value chain measure to the firm-level. He finds that once controlling for different firm characteristics, upstream firms tend to require more energy, emit higher levels of CO2and demonstrate reduced efficiency for both measures. There are alternative explanations for the pattern going beyond Shapiro (2021). First, consumer awareness for environmentally friendly energy and CO2efficiency production might increase with proximity to final consumption (Zhu et al., 2012). Second, once firms observe that prior stages have already made investments in the production of inputs perceived as ethical, firms are incentivised to contribute to a cleaner final product (Herkenhoffet al., 2021). Interestingly, the negative relation between upstreamness and clean production reduces with increased exposure to environmentally demanding export markets. In line with Guti´errez and Teshima (2018), the latter links to higher distance from the technological frontier or best management practises for firms occupying an upstream position. Foreign ownership represents another direct form of global value chain participation, characterised as a particular form of a firm’s export strategy. Horizontal foreign direct investments can serve as a substitute for exports. Additionally, vertical foreign direct investments can be viewed as a sub-set of a firm’s internationalisation strategy, fragmenting production across different countries (Antr`as and Yeaple, 2014). Foreign ownership is seen as a powerful force for convergence in productivity and production technologies (Arnold and Javorcik, 2009). The literature on foreign ownership and environmental performance dates back to Eskeland and Harrison (2003), who analyse whether multinationals flock to developing countries, characterised as pollution havens due to lax environmental policy stringency. However, they find weak evidence that foreign investors are more likely to target sectors with high levels of air pollution. Conversely, they find that foreign firms in Mexico, Morocco, Cˆote d’Ivoire and Venezuela produce environmentally more efficiently, compared to domestic firms.3Generally, empirical results on the relationship between firms’ environmental performance and ownership are rather mixed (Cole et al., 2017). Cole et al. (2008) analyse Ghanaian firm-level data to investigate if training-related capacity building for key employees, transfers from parent companies to their foreign affiliates. They find that training of key decision makers in the affiliates does improve efficiency in fuel, energy and electricity use. However, there is no direct link between foreign ownership and fuel use nor energy use, except some indication for a switch to a higher share of electricity consumption, on average being a cleaner input compared to alternative fuels used. 3The following analysis limits to the latter observed effect on the relation of foreign ownership and a firm’s environmental performance. However, it should be noted that these neglects possible exploitation of parent firms based in advanced economies facing stringent environmental regulation and might exploiting lax environmental regulations in less-advanced economies. The latter is for instance shown by Hanna (2010) by employing panel data of US firms. 16 In a more recent study, Brucal et al. (2019) apply a difference-in-difference approach combined with propensity score matching to Indonesian firm-level data. They show that acquired plants increase energy demand, but reduce energy intensity by about 30% two years following acquisition. Additionally, they find similar results for a firm’s CO2intensity. They interpret the results as suggesting an association between foreign acquisition and improvements to production processes arising through the adoption of better technologies or better management. The studies discussed so far find evidence suggestive of an environmental premium from global value chain participation or exposure. Export activity, import competition, downstream position and to some extend foreign ownership relate to environmental efficiency gains. However, the research reviewed so far, predominantly focuses on efficiency measures of energy, CO2 and other emissions. But the concept of environmental innovation covers a diverse set of environmental domains (Kemp and Pearson, 2007). The next section connects these results with a relatively independent literature on the drivers of environmental innovation adoption by a firm. 2.3 Firms’ global value chain participation and environmental innovation adoption Based on the Oslo Manual, most scholars investigate the drivers of environmental innovations defined as innovations that result in a reduction of the environmental impact in consumption or production and are novel to the firm (OECD and Eurostat, 2005; Kemp and Pearson, 2007; Del R´ıo et al., 2016). Building on Rennings (2000) and Jaffe et al. (2005), it is well-established that the drivers of environmental innovation and standard innovation adoption differ. Horbach (2008) categorises the determinants of environmental innovation adoption into a supply side technology-push, a demand-pull, environmental regulation-push and other firm specific factors (Barbieri et al., 2016; Ghisetti and Pontoni, 2015). Variables that capture the extent of a firm’s internationalisation, e.g. foreign ownership or export activity, are contained in the list of firm-specific drivers. However, the results for these variables are ambiguous and remain controversial. Empirical studies that capture a firm’s export activity as a determinant of environmental innovation adoption, to a large extent, do not report any positive correlation (Cainelli et al., 2012; De Marchi and Grandinetti, 2012; Horbach et al., 2013; Pe˜nasco et al., 2017). The results by Chiarvesio et al. (2015) confirm that export activity does not induce any environmental upgrading. Additionally, they also control for the direction of trade and do not find a positive impact if exports target non-emerging countries. The authors claim that lack of a significant correlation between export activity and environmental innovation adoption might be an artefact of the data - namely the geographic restriction of their study and other similar prior studies to advanced European countries such as Italy, Spain, Germany or France. These countries are characterised by a high level of consumer awareness and environmental policy stringency. Keshminder and del R´ıo (2019) partly address this research imbalance, even though their analysis is restricted to the chemical sector in Malaysia. Their findings suggest that foreign stakeholders influence a firm’s environmental strategy, including efforts to comply with foreign buyers’ demand and environmental regulation. Moreover, firms can benefit from knowledge 17 spillovers. Sanni (2018) runs a comprehensive analysis on the drivers of environmental innovation adoption for Nigerian firms, including a discussion on the importance for firms from developing countries to comply with international standards. However, the empirical analysis does not include measures of a firm’s global value chain participation. Besides the empirical bias towards single-country studies in advanced countries, Ghisetti et al. (2015) point to a strong focus on innovative active firms that also affects the generalisability of any study findings. Kafouros et al. (2008) claims that a firm’s export activity allows the firm to tap into knowledge created in different markets, pushing the firm’s innovative capacity. Accordingly, and in line with the theoretical literature of Melitz (2003) and Bustos (2011), the sub-sample of innovative firms is more likely to be active in foreign markets as well as to report higher productivity that itself links to a better environmental performance (Cainelli et al., 2015; Hanley and Semrau, 2022; Shapiro and Walker, 2018). Girma and Hanley (2015) use an instrumental variable approach to assess the drivers of environmental innovation adoption for firms based in the United Kingdom. They do not restrict the sample to innovative firms and find robust evidence for a firm’s export activity and its adoption of environmental innovation. Ghisetti et al. (2015) restrict the sample to innovative firms and exploit a multi-country dataset, covering a broad range of Eastern European countries. Interestingly, the reported correlation between a firm’s export status and the likelihood of environmental innovation adoption is positively significant, but they do not investigate this correlation in more detail. However, Ghisetti and Pontoni (2015) argue that export activity does not directly relate to a firm’s environmental innovation adoption but rather moderates R&D activity as part of a firm’s technology-push. Conversely, Hanley and Semrau (2022) argue for a direct relationship, pointing to the literature on trade induced knowledge-flows and learning-by-exporting (De Loecker, 2007). Generally, the adoption of environmental innovation requires firms to source knowledge beyond their own boundaries (De Marchi, 2012; Cainelli et al., 2015). The argument is that trade induced knowledge flows are relatively more important for firms based in catch-up economies, facing reduced access to knowledge due to deficiencies in the domestic knowledge base or in labour force skills. By exploiting a cross-country sample and running a split focusing on Eastern European countries, Hanley and Semrau (2022) show that exporters are more likely to adopt process-based environmental innovations. These results are in line with Horbach (2016) who reveals empirical indication for knowledge flows from West to East Europe. Additionally, the results echo Pe˜nasco et al. (2017) who claim that international drivers are more important for firms based in countries with lower levels of domestic demand for environmental upgrading. Emphasising the importance of regulation for environmental innovation adoption (Ambec et al., 2013; Borghesi et al., 2015; Brunnermeier and Cohen, 2003; Porter and Van der Linde, 1995), Hanley and Semrau (2022) also investigate the role of the direction of trade, exploiting the OECD Environmental Policy Stringency index to proxy foreign demand for environmental quality (Botta and Ko´zluk, 2014). They find that higher export exposure to more environmentally stringent markets increases both, the likelihood of any environmental innovation adoption as well as the number of environmental innovations adopted. The positive correlation is observable for productand process-based environmental innovations and indicates that firms incorporate 18 foreign stakeholders’ demand into their production decision. Del R´ıo et al. (2016) point to lack of studies in the innovation literature focusing on foreign ownership as driver of environmental innovation adoption. This is somehow surprising, since foreign ownership is shown to foster a firm’s process and product innovation activities. The latter effect goes beyond a mere selection effect, the cherry-picking of foreign investors who might be tempted to acquire the most productive firms within any industry (Guadalupe et al., 2012). Moreover, and similar to export activity, foreign direct investments relate to direct knowledge spillovers, an important driver for firms’ environmental innovation adoption (Javorcik, 2004; Arnold and Javorcik, 2009; De Marchi, 2012). However, Pe˜nasco et al. (2017) do not find that Spanish firms with foreign equity are more likely to adopt environmental innovations. The missing finding for foreign direct investment is similar to evidence produced by Inoue et al. (2013) who, for a sample of Japanese firms, find no positive correlation between having a head office in a foreign country and the introduction of quality standards. Similarly, Cainelli et al. (2012) use Italian firm-level data and do not find evidence for a significantly positive correlation between foreign direct investment and environmental innovation adoption. However, they find that foreign-owned firms are more likely to implement environmental innovations if they are embedded in local production systems, e.g. by cooperation to suppliers or universities. Conversely, Cainelli et al. (2015) find at least some indication of a positive correlation between firms experiencing foreign investment flows and the adoption of environmental innovations. However, the effect is only significant at the ten-percent level. As similar to the remarks relating to the interface between export activity and environmental innovation adoption, the composition of the samples used by researchers might explain the different results. Albornoz et al. (2014) use Argentinean firm-level data to investigate if foreignownership influence firms’ environmental actions. They find foreign-owned firms show compared to domestic firms’ higher activities in the adoption of six out of eight environmental measures. Also, Kannen et al. (2021) analyse data from non-advanced countries. They extend the data coverage to a cross-country sample containing firms from Eastern Europe, Africa and Asia at different stages of development. They find that foreign direct investment positively correlates with the implementation of a green management, measured as the adoption of a green manager or green strategy, and show that green management go hand in hand with better environmental performance, e.g. higher environmental innovation adoption rates. By directly measuring the link between foreign ownership and green management, they pin down the mechanism of improved management practices suggested by Brucal et al. (2019) on how a foreign ownership relates to a firm’s environmental performance. Such a relation between foreign ownership and green management is also observable for other environmental-related management routines. For instance, Zhu et al. (2012) explore Chinese survey data from 2006 and find that ISO 14001 environmental management certification is more likely adopted in foreign owned firms. Summing up, there is a distinct lack of consensus in the literature on foreign ownership and environmental innovation, suggesting more work remains to be done. The final section summarises the main results of the literature review and connects the findings to gaps in the literature. 19 Javorcik, B. S. (2004): “Does foreign direct investment increase the productivity of domestic firms? In search of spillovers through backward linkages,” The American Economic Review, 94, 605–627. Kafouros, M. I., P. J. Buckley, J. A. Sharp, and C. Wang (2008): “The role of internationalization in explaining innovation performance,” Technovation, 28, 63–74. Kannen, P., F. O. Semrau, and F. Steglich (2021): “Green gifts from abroad? FDI and firms’ green management,” Working Paper 2200, Kiel Institute for the World Economy, Kiel. Kemp, R. and P. Pearson (2007): “Final report MEI project about measuring ecoinnovation,” Report, UM Merit, Maastricht. Keshminder, J. and P. del R´ ıo(2019): “The missing links? The indirect impacts of drivers on eco-innovation,” Corporate Social Responsibility and Environmental Management, 26, 1100–1118. Kong, D., G. Ma, and N. Qin (2022): “Trade liberalization and firm toxic emissions,” Review of International Economics. Mart´ ınez-Zarzoso, I., S. Roy-Mukherjee, F.-O. Semrau, and A. M. Voicu (2020): “Pollution reduction by rationalization in Indian firms,” Working Paper 20020.01, INFER, International Network for Economic Research. Melitz, M. J. (2003): “The impact of trade on intra-industry reallocations and aggregate industry productivity,” Econometrica, 71, 1695–1725. OECD and Eurostat (2005): “Oslo Manual: Guidelines for collecting and interpreting innovation data, 3rd Edition,” The measurement of scientific and technological activities, OECD Publishing, Paris. Pe˜ nasco, C., P. del R´ ıo, and D. Romero-Jord´ an (2017): “Analysing the role of international drivers for eco-innovators,” Journal of International Management, 23, 56–71. Porter, M. E. and C. Van der Linde (1995): “Toward a new conception of the environmentcompetitiveness relationship,” Journal of Economic Perspectives, 9, 97–118. Rennings, K. (2000): “Redefining innovation—eco-innovation research and the contribution from ecological economics,” Ecological Economics, 32, 319–332. Richter, P. M. and A. Schiersch (2017): “CO2emission intensity and exporting: Evidence from firm-level data,” European Economic Review, 98, 373–391. Sanni, M. (2018): “Drivers of eco-innovation in the manufacturing sector of Nigeria,” Technological Forecasting and Social Change, 131, 303–314. Semrau, F. O. (2022): “On the drivers of clean production: Firms’ global value chain positioning,” in Beitr¨age zur Jahrestagung des Vereins f¨ur Socialpolitik 2022: Big Data in Economics, Kiel, Hamburg: ZBW - Leibniz Information Centre for Economics. 26 Shapiro, J. S. (2021): “The environmental bias of trade policy,” The Quarterly Journal of Economics, 136, 831–886. Shapiro, J. S. and R. Walker (2018): “Why is pollution from US manufacturing declining? The roles of environmental regulation, productivity, and trade,” The American Economic Review, 108, 3814–3854. Topalova, P. and A. Khandelwal (2011): “Trade liberalization and firm productivity: The case of India,” Review of Economics and Statistics, 93, 995–1009. Veugelers, R. (2012): “Which policy instruments to induce clean innovating?” Research Policy, 41, 1770–1778. Zhu, Q., J. Cordeiro, and J. Sarkis (2012): “International and domestic pressures and responses of Chinese firms to greening,” Ecological Economics, 83, 144–153. 27 Chapter 3 Stepping up to the mark? Firms’ export activity and environmental innovation in 14 European countries∗† Aoife Hanley Kiel Institute for the World Economy (IfW) Kiel Centre for Globalization (KCG) Christian-Albrechts-Universit¨at zu Kiel [email protected] Finn Ole Semrau‡ Kiel Institute for the World Economy (IfW) Kiel Centre for Globalization (KCG) finn-ole.semr[email protected] Abstract We investigate the ability of exports to trigger the adoption of environmental innovation (EI) in firms, shedding light on the determinants of convergence in environmental standards for Europe’s catch-up economies. To analyse this question empirically, we measure the latter as the 1) probability a firm adopts EI and the 2) breadth of EI adoption in firms. Applying Logit and Poisson pseudo-maximum likelihood estimations to firm-level data for 14 European countries, we find that particularly Eastern European exporters report higher EI adoption propensities, an effect exclusively driven by process-based EI and not observable for product-based EI. Additionally, we reveal that regardless of a firm’s origin, exposure to importing countries with high market-related environmental policy stringency, is linked to the adoption of EI. We conclude that learning-by-exporting, regulation-push and demandpull mechanisms may help to explain these findings, with foreign markets being characterised by a wider diversity of stakeholder preferences. JEL Classification Numbers: F18, O31, Q52, Q56 Keywords: Green-/Eco-/Environmental innovation; Circular economy; Environment and trade; Export activity; Green Deal ∗The chapter has been published as: Hanley, A.; Semrau, F.O. (2022). “Stepping up to the mark? Firms’ export activity and environmental innovation in 14 European countries.” Industry and Innovation, 29(5), 672700. The article is accessible via https://doi.org/10.1080/13662716.2021.2021865. In addition an earlier version of the manuscript has been published as: Semrau, F. O.; Hanley, A. (2019). “Stepping up to the mark? Firms’ export activity and environmental innovation in 14 European countries.” Beitr¨age zur Jahrestagung des Vereins f¨ur Socialpolitik 2019: 30 Jahre Mauerfall - Demokratie und Marktwirtschaft, Kiel, Hamburg: ZBW - Leibniz Information Centre for Economics. †We would like to thank Richard Franke, Holger G¨org, Jens Horbach, Beata Javorcik, Thilo Kroeger, Jakob Lehr, David Maddison, Giovanni Marin, Massimiliano Mazzanti, Horst Raff, Christina Raasch, Katrin Rehdanz, Claas Schneiderheinze, Frauke Steglich, Ruijie Tian, Julian Vehrke, other researchers of the KCG and the highly helpful comments by two anonymous referees. Additionally, we thank seminar/conference/workshop participants at the Trinity Business School Seminar Series 2021, the Annual Conference of the Verein f¨ur Socialpolitik 2019, the EEARE 2019, the DRUID 2019, Annual CERCIS Workshop 2019 in Ferrara, the INFER Workshop in Coimbra 2019, the G¨ottinger Workshop in International Economics 2019, the Kiel Centre for Globalization Workshop 2018 the Kiel Institute Young Researchers Seminar 2018 and the Aarhus-Kiel Workshop 2017. Moreover, we acknowledge financial support from the Leibniz Association through the Leibniz Science Campus KCG (Grant number is SAS-2016-IfW-LWC). ‡Corresponding author 28 3.1 Additional appendix The following appendix supplements the published manuscript accessible via https://doi. org/10.1080/13662716.2021.2021865. Description of dependent variable As described in Section 3.1, the harmonised Community Innovation Survey (CIS) waves 2008 and 2014 include an expanded list of environmental innovation (EI) variables. However, there are slight differences between both waves. Figures 3.1 and 3.2 present the exact questions used to create the dependent variables. Figure 3.1: List of questions related to EIs covered in the harmonised CIS 2008 29 Figure 3.2: List of questions related to EIs covered in the harmonised CIS 2014 We define five process-based EIs by using the following variables: •ECOMAT •ECOENO, we combine ECOEN and ECOCO into a single variable in the CIS 2008 •ECOPOL, does also include water in CIS 2014 •ECOSUB •ECOREC In addition, we define three product-based EIs by using the following variables: •ECOENU, does also include CO2footprint in CIS 2014 •ECOPOS •ECOREA Finally, we exclude two domains introduced in CIS 2014 from the analysis, namely ECOREP and ECOEXT. 30 Description of main independent variables Section 3.2 describes the econometric strategy, including a discussion of the main independent variables: export activity and EPS exposure. Next, more details of these variables are presented. Export activity Figures 3.3 and 3.4 include the respective question used to define a firm’s export activity. Both questions are widely similar between both survey waves. Figure 3.3: Question used to define a firm’s export activity in the harmonised CIS 2008 Figure 3.4: Question used to define a firm’s export activity in the harmonised CIS 2014 We denote three categories of a firm’s export status. The first category applies if a firm does not serve foreign markets, so only MARLOC is selected. The second category exporter denotes a firm serving foreign markets but where the bulk of the firm’s turnover are generated domestically. Accordingly, MARNAT, MAREUR or MAROTH are selected but no foreign geographic market is selected as largest market in terms of turnover (LARMAR). Finally, export dominates denotes that a foreign geographic market is selected as the largest market. 31 Environmental policy exposure The OECD EPS Index is used to construct a proxy of a firm’s exposure to foreign EPS. In so doing, we distinguish between market-related and non-market related EPS exposure. Overall, the OECD EPS Index includes 14 environmental policy instruments. Nine instruments are market-related and five instruments are non-market related (Botta and Ko´zluk, 2014). The raw data already includes variables capturing the respective market-related and non-market related EPS of a country. These variables are a weighted mix of the different policies. Market-related EPS allocates a weight of 0.33 to taxes, trading schemes and feed-in tariffs, respectively. The respective instruments considered are: •Tax CO2 •Tax Diesel •Tax NOx •Tax SOx •Trading schemes CO2 •Trading schemes green certificates •Trading schemes white certificates •Feed-in tariffs solar •Feed-in tariffs wind Non-market related EPS includes standards and R&D subsidies. To each policy a weighting factor of 0.5 is allocated and the respective instruments are: •Emission limit values NOx •Emission limit values PMx •Emission limit values SOx •Sulphur content limit for diesel •Renewable energy public RD&D budget Table 3.1 includes the overall EPS ranking of the countries covered in the OECD EPS Index. The table is sorted by the EPS scores in 2006. 32 Table 3.1: OECD EPS scores sorted by the score in 2006 Year/Country 2006 2007 2008 2009 2010 2011 2012 France 3.28 2.86 2.90 3.69 3.15 3.70 3.57 Denmark 3.16 2.83 2.96 4.07 4.03 3.98 3.85 Finland 3.15 2.82 3.08 3.25 3.21 3.48 3.43 Sweden 3.03 2.70 2.92 3.34 3.09 3.23 3.10 Germany 3.00 2.67 2.64 3.06 3.02 3.14 2.92 Korea 2.96 2.96 3.38 3.52 3.52 3.44 2.63 Spain 2.96 2.75 2.70 3.00 2.72 2.85 2.22 Czech Republic 2.88 2.55 2.72 2.89 2.89 2.37 2.38 Austria 2.82 2.44 2.91 3.33 3.33 3.08 2.95 Netherlands 2.80 2.64 3.23 3.69 4.13 3.51 3.63 Italy 2.72 2.34 2.60 2.73 2.84 2.79 2.77 Portugal 2.71 2.21 2.26 2.47 2.54 2.27 2.13 Hungary 2.59 2.30 2.55 2.66 2.77 2.68 2.63 Belgium 2.40 2.20 2.34 2.58 2.60 2.53 2.47 United Kingdom 2.29 1.95 2.40 2.58 3.62 3.47 3.29 Poland 2.26 2.08 2.26 2.96 2.96 2.96 2.58 Ireland 2.23 1.71 2.05 2.16 2.22 2.43 2.05 Canada 2.17 3.27 3.31 3.85 3.35 3.67 3.42 Switzerland 2.13 2.13 2.67 3.19 3.33 3.29 3.29 Norway 2.13 2.05 2.34 3.19 3.19 3.19 3.26 United States 2.13 2.34 2.47 2.93 2.68 2.47 3.17 Australia 2.01 2.01 2.26 2.69 2.50 3.34 3.72 Greece 1.84 1.92 1.83 2.08 2.33 2.33 2.13 Slovak Republic 1.78 1.40 1.53 2.39 2.30 3.05 2.99 Japan 1.63 1.69 1.69 1.73 2.03 2.96 3.50 Turkey 1.50 1.50 1.50 1.54 2.06 2.21 1.83 China (People’s Republic of) 0.77 0.77 0.81 0.98 1.10 1.35 2.04 India 0.67 0.63 0.63 1.13 1.20 1.26 1.30 Russia 0.65 0.65 0.60 0.60 0.60 0.60 0.60 South Africa 0.52 0.52 0.48 1.52 1.75 1.71 0.71 Indonesia 0.50 0.50 0.50 0.50 1.17 1.17 1.17 Brazil 0.42 0.42 0.42 0.42 0.42 0.38 0.38 Slovenia n.a. n.a. 1.64 1.85 2.46 2.42 2.28 Description: The OECD Environmental Policy Stringency (EPS) index is a country-specific and internationally-comparable measure of the stringency of environmental policy. Stringency is defined as the degree to which environmental policies put an explicit or implicit price on polluting or environmentally harmful behaviour. The index ranges from 0 (not stringent) to 6 (highest degree of stringency). The index covers 28 OECD and 6 emerging countries for the period 1990-2012. The index is based on the degree of stringency of 14 environmental policy instruments, primarily related to climate and air pollution (Botta and Ko´zluk, 2014). Version: 24.05.2018 33 Description of other independent variables This part of the additional appendix complements Section 3.2 by providing more information how the independent variables are constructed using the CIS 2008 and CIS 2014 waves. MNC captures a foreign ownership dummy. The dummy determines whether a firm belongs to an enterprise group with a headquarter located outside of the firm. For the construction, the following questions of the CIS 2008 and CIS 2014 are used: Figure 3.5: Question used to define MNC in the harmonised CIS 2008 Figure 3.6: Question used to define MNC in the harmonised CIS 2014 We capture the number of external cooperation partners involved in the innovation progress by including ext. coop and the respective squared term. In total, we consider seven different cooperation partners based on the survey questions presented in Figures 3.7 and 3.8. In the CIS 2014, information about cooperation with clients and customers is split into the private and public sector. Both categories are matched into a single category to improve the comparability across survey waves. 34 Figure 3.7: Question used to define ext. coop in the harmonised CIS 2008 Figure 3.8: Question used to define ext. coop in the harmonised CIS 2014 35 Chapter 4 On the Drivers of Clean Production: Firms’ Global Value Chain Positioning∗† Finn Ole Semrau Kiel Institute for the World Economy (IfW) Kiel Centre for Globalization (KCG) finn-ole.semr[email protected] Abstract Industries that occupy upstream positions in global value chains (GVCs) - being positioned closer to the raw product - produce proportionately more CO2-intensive. However, firms are heterogeneous, even in narrowly defined industries. In this paper, I empirically investigate whether the relationship between upstreamness and CO2emissions, measured in absolute and relative terms, holds within industries at the firm level. Using granular data of Indian manufacturing firms and controlling for established drivers of clean production, I reveal that firms producing products closer to final consumption produce less CO2-intensive. I corroborate the finding by using a 2-SLS instrumental variable approach. Interestingly, I find that exposure to importing countries with stringent environmental policies attenuates the link between upstreamness and dirty production. The latter finding suggests the imperative of technology upgrading for dirty upstream producers aiming to remain competitive in international markets. JEL Classification Numbers: F14, F18, O13, Q56. Keywords: Environment and trade; environment and development; CO2intensity; global value chains; sustainable development. ∗An earlier version of the manuscript has been published as Semrau, F. O. (2022). “On the Drivers of Clean Production: Firms’ Global Value Chain Positioning.” Beitr¨age zur Jahrestagung des Vereins f¨ur Socialpolitik 2022: Big Data in Economics, Kiel, Hamburg: ZBW - Leibniz Information Centre for Economics. In addition, a version of the manuscript has been submitted to a Special Issue on sustainable value chains in World Development in July 2022. It has been one of eight papers part of the proposal for the Special issue by the “Research Network Sustainable Global Supply Chains”. †I thank Eckhardt Bode, Carolyn Fischer, Andreas Fuchs, Robert Gold, Holger G¨org, Aoife Hanley, Tobias Heidland, Thilo Kroeger, Jakob Lehr, Chiara Lodi, Stefan Pahl, Sonja Peterson, L´eo Picard, Niklas Potrafke, Martin Quaas, Katrin Rehdanz, Claas Schneiderheinze, Frauke Steglich, Rainer Thiele and Norihiko Yamano for their valuable comments and discussions on earlier versions. Furthermore, I thank participants of the Annual Conference of the Verein f¨ur Socialpolitik in Basel 2022, the Research Conference “Sustainability in Global Value Chains” 2021 jointly organised by the United Nations Industrial Development Organization (UNIDO) and the “Research Network Sustainable Global Supply Chains”; the 26th Annual Conference of the European Association of Environmental and Resource Economists (EAERE) 2021; the Seminar of the University Ferrara 2022, the Kiel Institute for the World Economy (IfW) and the Kiel Centre for Globalization Research Seminar 2022; the G¨ottinger Workshop on International Economics 2020; the Aarhus-IfW Workshop 2018; the Brown Bag Workshop of the Kiel University 2018; and the HSU-IfW Workshop 2018. Moreover, I acknowledge financial support from the Leibniz Association through the Leibniz Science Campus KCG (Grant number is SAS-2016IfW-LWC 42 Highlights •In this paper, I empirically investigate the relation between firms’ CO2emissions and GVC positions •Ifind that within industries firms that occupy upstream GVC positions produce more CO2-intensive •Exposure to export markets with strong environmental regulation moderates (negatively) the relation between upstreamness and dirty production •Afirm’s export status does not significantly moderate the relation between upstreamness and dirty production •Accelerating the environmental abatement of firms in upstream GVC positions is key reaching global goals on CO2emissions 43 4.1 Introduction Promoting the green transition of industries is a pressing challenge of our time and it is covered in the United Nations Sustainable Development Goals 9 and 12. Even in the light of energy efficiency gains over the last decades, economic growth strongly correlates with energy consumption and energy consumption itself strongly correlates with greenhouse gas (GHG) emissions (Costa-Campi et al., 2015). Global warming, a consequence of GHG emissions, represents a serious threat for future generations through increasingly volatile weather conditions, changed agricultural productivity, threatened biodiversity and elevated sea levels. Emerging and developing countries are especially vulnerable to the economic risks related to global warming. At the same time, with their economic rise, most of the global increase in GHG emissions originates from these countries (Copeland et al., 2022). Accordingly, beyond the urgency of a significant decoupling between economic growth and GHG emissions in industrialised countries, major successes in GHG emission reduction of developing and emerging countries are crucial if there is any hope of attaining the Paris Agreement’s target of limiting global warming to well below 1.5°Celsius. This paper, contributes to the understanding of the determinants of firms’ clean production, measured as reductions of firms’ CO2emissions and CO2intensity, in developing countries. In so doing, I complement prior studies on the relationship between a firm’s global value chain (GVC) participation and its environmental performance - the focus being a firm’s GVC positioning, rather than foreign ownership, export, or import status (Batrakova and Davies, 2012; Brucal et al., 2019; Guti´errez and Teshima, 2018). To this end, I empirically analyse data of firms based in India, the third largest global emitter of CO2and a country with historically high rates of emissions growth. A main contribution to the literature is whether Shapiro’s (2021) finding of a negative relationship between industries’ upstreamness and clean production can be confirmed at the firm level. Specifically, this means confirming whether firms located more upstream along the GVC - being positioned closer to the raw product than their peers producing a similar core product - demonstrate a worse environmental performance. Additionally, I investigate whether a firm’s exports and exposure to export markets characterised by stringent environmental policies, negatively moderates the relationship for upstreamness. The negative relationship between industries’ upstreamness and clean production is monotonous and somewhat linear across industries and countries, at all stages of development (Shapiro, 2021). In line with this finding, Copeland et al. (2022) list as a stylised fact that upstream industries do indeed carry a heavier environmental footprint. However, firms are heterogeneous, even within narrowly defined industries (Melitz, 2003). Accordingly, relationships observed at the industry level cannot comment on firm level phenomena such as productivity, exports or product mix. These and other firm level idiosyncrasies, which potentially relate to the environmental performance of a firm, motivate this paper. There are strong intuitions why these firm level phenomena matter. Firstly, differences in the product mix are expected to connect both to a firm’s GVC positioning as well as the environmental impact of its production. A manufacturing firm that also sells mining products is likely to report worse CO2intensity than its peers because mining products require more energy, on average, during the production process. At the same time, mining products are more predominant in upstream GVC positions. 44 Second, productivity varies substantially between firms within an industry and productive firms are more likely to participate in global trade (Melitz, 2003). A firm’s trade participation can induce additional productivity gains, technology upgrading, and improvements in its environmental performance (Bustos, 2011; Cherniwchan et al., 2017; Copeland et al., 2022; Shapiro and Walker, 2018). Guti´errez and Teshima (2018) argue that developing countries, in particular, can benefit environmentally from trade-induced technology adoption because firms based in these countries are, on average, further removed from the technology frontier. I argue that a similar claim might hold for the relationship between upstreamness and cleaner production. If firms in upstream value chain positions are characterised by a higher distance to the technology frontier, they might benefit disproportionately from trade-related drivers of cleaner production. These benefits from trade could arise from learning-by-exporting, demand-pull, and regulation-push (Brunnermeier and Cohen, 2003; De Loecker, 2007; Horbach, 2008). Additionally, they are more pronounced for firms exposed to countries with higher degrees of environmental policy stringency (EPS) (Hanley and Semrau, 2022). The latter might link to a negative moderating role of EPS exposure on the relationship between upstreamness and dirty production. However, strong environmental regulation in import destinations can cause firms in developing countries to fully specialise in the production of (dirty) inputs not conforming to these standards (Acemoglu et al., 2014). Outsourcing of dirty production to countries with lax environmental regulation is in line with the pollution haven hypothesis, as outlined by Antweiler et al. (2001) and Eskeland and Harrison (2003). Similar to Barrows and Ollivier’s (2021) recent contribution on the nexus between the GVC participation of firms and their environmental performance, I use the Prowess database. I denote clean production by firms’ CO2emissions and CO2emissions per unit of sales. Additionally, I follow Dardati and Saygili (2021), measuring firms’ CO2intensity by using CO2emissions per unit of value added. In this way, value added is calculated by subtracting spending on intermediates from the sales of each firm, which covers the value of a good acquired in prior stages of the GVC. Producing these intermediates is linked to indirect emissions along the GVC, emissions that are emitted during the production of the intermediates and transportation. Accordingly, using CO2per value added provides a measure for a firm’s environmental performance at a specific stage of the value chain. Exploiting firm level information, the CO2emissions of a firm are calculated using data on the quantities of energy consumed, for each energy sources used. Such detailed firm level information is rare in the context of a developing country and makes it possible to regard differences in CO2emissions between different sources of energy, in a way that is not captured if energy costs are applied as a proxy for clean production (Barrows and Ollivier, 2018, 2021). To measure the GVC position of a firm, I apply the industry level GVC measure of Antr`as et al. (2012) and Antr`as and Chor (2018). Specifically, I follow the methodology of Herkenhoff et al. (2021), calculating the Indian industry-year-specific GVC positions by employing the WIOD available for the period from 2000 to 2014 (Timmer et al., 2015). I then combine the industry information with the product level information component of Prowess to calculate firm-specific GVC positions (Herkenhoffet al., 2021). Empirically, I apply an OLS regression, 45 controlling for established drivers of clean production. By including industry fixed effects, the empirically exploited variation in GVC positions originates from differences in the product mix of firms within an industry (Herkenhoffet al., 2021). In two alternative specifications, I include respective interaction terms - first, between a firm’s upstreamness and exports, and second, between a firm’s upstreamness an its exposure to markets characterised by a high EPS. The main empirical finding confirms the stylised fact that firms occupying upstream GVC positions report higher CO2emissions from their production. The effect is statistically significant and economically meaningful. A change in the upstreamness measure by one unit, such as the equivalent of switching from the average GVC position of firms in the textiles industry to the basic metals industry, increases the estimated CO2emissions by 28.15%, CO2per unit of sales by 26.87% and CO2per unit of value added by 35.39%, respectively. A comparison of firms in very downstream positions, with firms in very upstream positions further illustrates the economically meaningful difference. The estimated CO2emissions increase by 63.23%, CO2per unit of sales by 60.0%, and CO2per unit of value added by even 82.21%. The latter corresponds to an increase from 0.76 kg to 1.39 kg CO2emissions per US dollar value added. Beyond the main finding, I reveal that the relationship between upstreamness and dirty production decreases with exposure to strong environmental regulation. For firms in upstream positions exposed to strong environmental regulation in export markets, the relationship between upstreamness and dirty production even disappears completely. This finding links to the adoption of more efficient technologies by firms aiming to remain competitive in environmentally demanding export markets. Learning-by-exporting, regulation-push, and demand-pull help firms to adoption of these technologies and are especially relevant for dirtier producers in upstream positions, arguably more remote from the technology frontier. For downstream firms, environmental performance deteriorates with exposure to strong environmental regulation in export markets. The finding suggests that, on average, producers in downstream positions specialise in the production of dirty inputs, while producers in upstream positions demonstrate environmental gains from exposure to strong environmental regulation. Interestingly, I do not find any moderating effect for exports on the relationship between upstreamness and clean production. The latter result suggests that firms located in more upstream positions do not benefit disproportionately from the trade-induced technique effect. The key results on the relationship between upstreamness and clean production are robust to a variety of alternative specifications. These include the use of an alternative measure of environmental performance by firms, extending the number of covered firms in the estimation sample and the application of a 2-SLS instrumental variable approach. In the latter, I use advertising intensity (advertising expenditure to sales) among firms producing a similar core product. The relevance of the instrument is motivated by a higher incentive for advertising expenditure for firms in downstream positions, such as targeting final consumers. By excluding afirm’s advertising spending and sales from the measure, the instrument is exogenously given toafirm and cannot be influenced by its managers or other stakeholders. The underlying analysis takes place against a backdrop in which policy-makers, keen to foster the transition towards a green economy, are seeking to understand the determinants of clean production. I focus on India, a country that faces multiple challenges. Apart from striving 46 to deal with environmental degradation, India faces other challenges in its efforts to achieve sustainable development, including poverty elimination and energy security. The twin targets of transformation towards a green economy and social development are strongly intertwined. It is imperative to break the detrimental dynamic between economic development and GHG emissions to achieve the ultimate goal of sustainable development. The present study covers the period 2000 to 2014, during which India’s share of global CO2emissions increased from 4.2% to 6.2%. Significantly, India’s absolute CO2emissions more than doubled within this period, and in absolute terms, India being the third largest emitter in the world, even though its CO2 emissions per capita are around 9.5 times less than the equivalent value for the United States (The World Bank, 2019).1 One approach fostering a global green economic transition is to design trade policy in such a way that rewards clean production along the GVC and fosters a green economic transition. A CO2border adjustment tax could represent such an instrument, as it penalises dirtier producers along the GVC by targeting dirtier production with higher tariffs (Aichele and Felbermayr, 2015). However, Shapiro (2021) uncovers an environmental bias in trade policy, where industries in downstream positions can be confronted with higher tariffs vs. industries in upstream positions, even if their production is, on average, more environmentally efficient. Such a policy works against the desired outcome because it subsidises dirty production in upstream industries, thereby undermining the overall target of reducing global CO2emissions. My results lend strong support at the firm level to Shapiro’s conclusion of trade-induced distortions, pointing to evidence of an environmental bias between relatively dirtier upstream, and cleaner downstream producers. The remainder of this paper is structured as follows. The next section summarises studies connected to GVCs and the environment, based on which I formulate the testable hypotheses. Section 4.3 presents the database used in the analysis, the empirical methodology, a description of the variables used, and the descriptive statistics. Next, the empirical results, the instrumental variable approach, and robustness checks are presented. I conclude with a summary of the results and their implications. 4.2 Firm’s GVC Positioning and Clean Production Significant efficiency gains in GHG emissions per unit of output are needed in order to further decouple economic growth and GHG emissions. However, firms cannot fully appropriate the welfare gains of investments in cleaner production techniques, largely because of the doubleexternality characteristic of these investments (Jaffe et al., 2005; Rennings, 2000). Accordingly, firms invest less than the social optimum in GHG emissions reduction. Key determinants in correcting market failures, which boost environmental innovation and investment in cleaner production, are regulation-push, market-pull, technology push, and firm-specific factors (e.g.: Barbieri et al. (2016); Brunnermeier and Cohen (2003); Cainelli et al. (2012); del R´ıo Gonz´alez (2009); De Marchi (2012); Ghisetti and Pontoni (2015); Ghisetti et al. (2015); Horbach (2008); Horbach et al. (2012); Horbach (2016); Kesidou and Demirel (2012); Rennings (2000)). 1Downloaded on 20 December 2019 47 Building on Shapiro (2021), I focus on GVC positioning as a driver of clean production - a driver that has not been considered in prior firm level studies. Shapiro (2021) finds that upstream industries are characterised by reduced environmental efficiency. The reasons suggested for this pattern are multiple. First, firms in upstream industries take raw materials extracted from the ground and transform them. This is more energy-demanding than the production of downstream firms, which depend more on relatively cleaner inputs such as labour (Shapiro, 2021). Second, fossil energy inputs such as coal are disproportionately used by the heaviest industries, typically those occupying more upstream positions such as electricity generation, cement manufacturing, and steel blast furnaces (Shapiro, 2021). There are also other mechanisms, which underpin the relationship between firms’ upstreamness and dirty production, beyond those outlined by Shapiro (2021). Market-pull factors affect afirm’s decision to invest in clean production (Horbach, 2008). Zhu et al. (2012) emphasise that downstream stakeholders - in particular - push firms to adopt environmental measures. For these stakeholders, downstream firms are relatively more visible than upstream firms; for instance, downstream products are a ubiquitous part of the daily routine of many consumers. If downstream stakeholders push for sustainable production, this urgency might diminish for less visible products, positioned further upstream the value chain. Herkenhoffet al. (2021) theoretically and empirically analyse the relationship between the upstreamness and corporate social responsibility (CSR), which serves as another dimension of firms’ sustainability in production. They reveal that firms in downstream positions spend comparatively more on CSR. A key reason is that CSR is characterised by incomplete contracts and CSR activities are observable, but not verifiable. As a result, firms have a higher incentive to invest in CSR once their suppliers invest in CSR. Herkenhoffet al. (2021) empirically confirm the prediction of amplifying CSR spending along the GVC, using the Indian Prowess database. It can be concluded from the Herkenhoffet al.’s study that a firm occupying a more downstream position might face a higher incentive to invest in environmentally friendly production techniques - an incentive underpinned by the prioritisation of environmental outcomes by suppliers along the GVC. These environmental-related expenditures share some similarities with CSR, because they are for buyers of intermediates observable, but not verifiable. Contrary to the mechanisms described earlier, which help to trigger a negative relationship between upstreamness and clean production, a positive relationship can also be argued. Shapiro (2021) points to an environmental bias of trade policy due to lower tariffs for upstream industries, but higher CO2emissions per US dollar of output. Bustos (2011) introduces technology change into the Melitz (2003) model and finds that Argentinian firms that face relatively lower tariffs increase revenues by entering foreign markets and are therefore better able to absorb costs of technology upgrading. Kong et al. (2022) show that such technology upgrading is also linked to the improved environmental performance of firms. They find that China’s trade liberalisation, following its World Trade Organization accession, significantly reduced the toxic emissions of firms. Combining the results by Shapiro (2021), Bustos (2011), and Kong et al. (2022), firms in upstream positions face lower tariffs and therefore have better access to foreign markets. This, in turn, encourages a firm to introduce technology upgrading and to improve its environmental performance. Nonetheless, viewing the evidence collectively and building on Shapiro (2021), I 48 would expect a negative relationship between upstreamness and clean production to prevail. However, firms’ exports might serve to moderate the negative relationship between upstreamness and clean production. Recent firm level studies have found that exports positively relate to firms’ environmental performance (Batrakova and Davies, 2012; Forslid et al., 2018; Girma and Hanley, 2015; Hanley and Semrau, 2022; Richter and Schiersch, 2017). Guti´errez and Teshima (2018) emphasise the positive impact of trade on technological choice can play a large role, especially for firms in developing countries, because such firms are generally further removed from the technology frontier. Arguably, firms in upstream positions produce in an environmentally less efficiently manner and are more remote from the technology frontier. In line with this, the expected relationship between upstreamness and dirty production might diminish with exports, because firms occupying upstream positions can benefit - in environmental terms - disproportionately from exports. In general terms, studies investigating the effect of a firm’s trade participation on its environmental outcomes in developing or emerging countries are scarce. An exception is Barrows and Ollivier (2021), who similarly use the Prowess database and focus on the direct CO2emissions of firms. The methodology used - calculating the CO2emissions embodied in the physical energy consumed - is also applied in Barrows and Ollivier (2018) and is novel in the context of an emerging country. However, the concept follows in the spirit of related studies by Richter and Schiersch (2017) and Forslid et al. (2018) who use German and Swedish data, respectively. Barrows and Ollivier (2021) reveal that foreign demand growth increases CO2emissions and improves CO2efficiency. Interestingly, they also find evidence for technological upgrading at the product level. In sum, foreign demand increases the level of total output, but simultaneous improvements in efficiency, on average, mitigate the scale effect, to the extent of about a half of this effect. Apart from the expected effect for exports, I expect the magnitude of the relationship between GVC positioning and dirty production to diminish with increased exposure to EPS. Firms in upstream positions that face tough competition in international markets have a disproportionate need to adopt new technologies, in order to improve efficiency and to remain competitive in environmentally strongly regulated countries. In addition, Hanley and Semrau (2022) reveal that firms exposed to strong environmental regulation in import destinations demonstrate higher environmental innovation adoption rates. On average, consumers in countries with stringent environmental regulations have strong preferences for sustainable production and firms are more likely to be equipped with state-of-the-art technologies and skilled labour. Hanley and Semrau (2022) interpret this finding as a mixture of export-induced learning-by-exporting, demand-pull, and regulation-push transmitted through the value chain and augmented by high sustainability standards in import destinations (Brunnermeier and Cohen, 2003; De Loecker, 2007; Horbach, 2008; Newman et al., 2018). However, Acemoglu et al. (2014) emphasise that strong environmental regulation in import destinations may induce specialisation in the production of CO2-intensive inputs by firms in developing countries, where the latter face relatively lax domestic EPS. In line with this, the pollution haven hypothesis predicts a shift of dirty industries to developing and emerging countries with less rigorous EPS (Antweiler et al., 2001; Eskeland and Harrison, 2003). The latter can lead to worse environmental performance of these 49 firms. However, it might be the case that these firms still produce in a more environmentally friendly way than their non-exporting domestic peers. Summarising the literature review how GVC participation relate to the environmental performance of firms, I empirically investigate the following hypotheses: H1: Afirm’s CO2emissions increases with upstreamness. H2: Afirm’s CO2intensity increases with upstreamness. H3: Exports negatively moderate the link between upstreamness and dirty production. H4: Exposure to stringent environmental policy in export markets negatively moderates the link between upstreamness and dirty production. The next section presents the methodology, starting with a description of the data used. 4.3 Methodology 4.3.1 Data Description I use the firm level Prowess database, merged with the WIOD to test the four hypotheses derived in the previous section. Comprising data on the financial performances of Indian firms, the Prowess data is provided by the Centre for Monitoring the Indian Economy.2Designed for academia (and other users), the data primary source information from income statements and balance sheets. Total production in Prowess covers more than 80% of India’s GDP (Bos and Vannoorenberghe, 2018). The empirical set-up spans the period between 2000 and 2014. Prowess data is also available before and after 2014. However, I apply the methodology of Herkenhoffet al. (2021), creating a measure of a firm’s GVC position using the WIOD (Timmer et al., 2015).3Unfortunately, the latter database is only available for the aforementioned period. Antr`as et al. (2012) describe how World Input-Output Tables can be used to calculate a country-industry specific continuous measure of upstreamness. The approach by Herkenhoffet al. (2021) links the industry-specific GVC measure with the product level data of the Prowess database, the latter including information on the capacity, production, and sales, disclosed by firms. Such product level information is rare in the context of an emerging country, going back to the Indian 1956 Companies Act, which mandated to disclose such information in the annual report (Goldberg et al., 2010b). Apart from such essential and not often accessible information at the product level, the coverage of energy consumption is unusual in the context of an emerging country. Firms report on their physical energy consumption, such as quantities of natural gas, electricity, coal, diesel or agricultural by-products consumed. The data availability goes back to a 1988 amendment in the Indian Companies Act, which forces firms to report on energy consumption by energy type. This level of detail makes it possible to calculate the heat content of energy consumption measured in MMBTU. The respective MMBTU of an energy source can be reformulated in 2The data were downloaded from https://prowessdx.cmie.com/ on the 26th of November 2019. I downloaded the data in US dollars and deflated all monetary variables using the World KLEMS data. The KLEMS data were downloaded from http://www.worldklems.net/data.htm, using the sheet GO P , which takes 2010 prices as the reference price. Last download on 21 September 2018. 3I also employ the WIOD Release 2016, including Input-Output Tables for 28 European countries and 15 other major economies (Timmer et al., 2015), downloaded from http://www.wiod.org/home 50 terms of its CO2content (Barrows and Ollivier, 2018, 2021). Given its unique data coverage, the Prowess database is widely used in the economics literature, including international economics (e.g., Barrows and Ollivier (2018, 2021); De Loecker et al. (2016); Goldberg et al. (2010a,b)). Similar to other studies analysing the determinants of a firm’s clean production, I limit the sample to the manufacturing base (Barrows and Ollivier, 2018, 2021; Richter and Schiersch, 2017). The manufacturing base is of particular interest in the context of GVCs and clean production because manufacturing products are comparatively high in CO2emissions, compared to service products. Additionally, manufacturing products are more frequently traded (Copeland et al., 2022). The next section presents the empirical strategy used in the analysis. 4.3.2 Empirical Analysis Equations 4.1 and 4.2 illustrate the two main OLS estimations, with CO2emissions and CO2intensity used as dependent variables, respectively. ln(CO2emissionsft) = α0+β1Upstreamnessft +β2Xft +γt+δf+�f+ζf,(4.1) ln(CO2intensityft) = α0+β1Upstreamnessft +β2Xft +γt+δf+�f+ζf,(4.2) CO2emissionsft is the natural logarithm of a firm’s fCO2emissions in kilotonnes (kt) in year tderived from its energy consumption during production. CO2intensityft is the natural logarithm of a firm’s CO2emissions in kt over its sales and value added in US dollars, respectively. The construction of the dependent variables is presented more comprehensively in Section 4.3.2.1. Upstreamnessft is the main independent variable of interest. The value increases with the upstreamness of a firm fin year t. The measure varies over time with changes in a firm’s product mix. This is described in more detail in Section 4.3.2.2. Xfcaptures different idiosyncrasies of the firm, potential co-determinants of the decision to apply a clean production technology. These drivers can be grouped into regulation-push, market-pull, technology-push, and firm-specific factors.4 EP S measures the exposure of a firm to foreign markets characterised by strong environmental regulation. On one hand, EPS exposure might connect to the transmission of regulationpush, demand-pull, and learning-by-exporting mechanisms through the value chain (Hanley and Semrau, 2022). On the other hand, domestic firms confronting weaker EPS can specialise in the production of dirty products (Acemoglu et al., 2014). The construction of EPS is presented in more detail in Section 4.3.2.3. Firms that are active in dirtier industries are more likely to trade their outputs (Copeland et al., 2022). However, firm level studies reveal that exporters demonstrate higher energy and emission intensities (Barrows and Ollivier, 2021; Batrakova and Davies, 2012; Forslid et al., 4For a literature review on the drivers of environmental innovation check Barbieri et al. (2016) and del R´ıo Gonz´alez (2009). 51 Table 4.2: Industry statistic about CO2emissions and upstreamness Industry CO2CO2CO2/SA CO2/SA CO2/VA CO2/VA Upstr. Obs. average median average median average median Chemicals 52.76 6.33 1.09 0.47 2.07 0.98 3.19 2,500 Coke & refined petroleum products 624.97 0.54 0.57 0.12 1.68 0.36 2.75 93 Basic metals 305.77 11.97 1.89 0.60 3.73 1.31 2.67 2,545 Paper 116.43 29.08 4.62 4.00 9.34 7.71 2.62 977 Other non-metallic mineral 332.66 15.48 3.66 1.57 4.90 2.24 2.34 720 Rubber & plastics products 32.40 2.97 0.68 0.33 1.36 0.73 2.30 1,112 Fabricated metal products 15.24 4.88 0.66 0.41 1.44 0.81 2.28 646 Wood 4.46 1.15 1.92 0.91 3.79 1.61 2.17 23 Printing & reproduction of recorded media 0.06 0.06 0.09 0.09 0.09 0.09 2.12 1 Computer electronic & optical products 1.40 0.82 0.31 0.11 0.53 0.18 2.11 171 Electrical equipment 6.30 0.80 0.27 0.09 0.60 0.17 2.00 292 Machinery & equipment 11.41 0.74 0.66 0.12 1.05 0.20 1.95 538 Other manufacturing 49.24 0.90 0.94 0.10 1.41 0.23 1.93 72 Other transport equipment 20.72 1.67 0.15 0.06 0.30 0.17 1.82 46 Motor vehicles 8.50 4.16 0.28 0.17 0.51 0.33 1.78 379 Textiles 22.92 7.72 1.57 1.01 2.97 1.80 1.68 3,061 Leather 2.35 1.76 0.62 0.33 1.69 0.48 1.65 28 Wearing apparel 2.73 1.03 0.32 0.09 0.40 0.21 1.65 82 Beverages 41.79 5.96 0.89 0.29 1.48 0.39 1.60 236 Tobacco 94.17 1.05 0.09 0.04 0.12 0.04 1.59 18 Food 45.24 5.59 1.89 0.70 3.54 1.15 1.57 2,713 Pharmaceuticals 33.07 2.91 0.88 0.18 1.31 0.29 1.57 1,289 product mix of firms producing in the same main industry. Columns (1) and (4) focus on the determinants of firms’ absolute CO2emissions, while the remaining columns focus on firms’ CO2intensities. Table 4.3: OLS estimation: Upstreamness and clean production (1) (2) (3) (4) (5) (6) VARIABLES CO2CO2/SA CO2/VA CO2CO2/SA CO2/VA Upstreamness 0.334*** 0.289*** 0.445*** 0.248** 0.238** 0.303*** (0.055) (0.053) (0.055) (0.104) (0.108) (0.107) EPS 0.062 0.135 0.039 -0.052 0.115 0.022 (0.273) (0.269) (0.270) (0.238) (0.233) (0.228) Export status -0.462*** -0.615*** -0.581*** -0.242*** -0.427*** -0.368*** (0.060) (0.059) (0.058) (0.055) (0.054) (0.054) Foreign ownership -0.423*** -0.455*** -0.662*** -0.284*** -0.337*** -0.514*** (0.118) (0.120) (0.122) (0.109) (0.110) (0.112) R&D dummy -0.201*** -0.320*** -0.448*** -0.009 -0.171** -0.254*** (0.075) (0.078) (0.080) (0.070) (0.070) (0.072) SOE 1.544*** 1.332** 1.353*** 1.372*** 1.195** 1.191*** (0.552) (0.578) (0.477) (0.465) (0.471) (0.366) (ln) productivity 0.083*** -0.179*** -0.331*** 0.037* -0.208*** -0.384*** (0.022) (0.021) (0.021) (0.021) (0.020) (0.020) (ln) assets 0.998*** 0.077*** 0.124*** 0.947*** 0.037** 0.074*** (0.021) (0.020) (0.020) (0.019) (0.018) (0.019) (ln) age 0.290*** 0.112*** 0.038 0.200*** 0.036 -0.047 (0.038) (0.037) (0.037) (0.034) (0.033) (0.033) Agricultural products 0.720*** 0.565*** 0.763*** 0.349*** 0.216** 0.428*** (0.115) (0.119) (0.120) (0.105) (0.109) (0.111) Mining products 0.880*** 1.104*** 0.926*** 0.648*** 0.873*** 0.693*** (0.199) (0.197) (0.196) (0.192) (0.198) (0.190) Service products -0.034 -0.203* -0.128 0.030 -0.126 -0.058 (0.114) (0.119) (0.119) (0.106) (0.109) (0.109) Exclusive manufacturing 0.172 0.154 0.160 0.093 0.096 0.091 (0.123) (0.128) (0.127) (0.113) (0.116) (0.116) Observations 17,542 17,542 17,542 17,541 17,541 17,541 R-squared 0.535 0.172 0.202 0.627 0.341 0.365 Year dummy Yes Yes Yes Yes Yes Yes Industry dummy No No No Yes Yes Yes State dummy Yes Yes Yes Yes Yes Yes Robust standard errors in parentheses (clustered at the firm level) *** p<0.01, ** p<0.05, * p<0.1 Upstreamness is positively significant in all six specifications. Interestingly, the coefficients are higher and standard deviations are lower for the specifications without industry fixed effects. However, the coefficients remain significant once industry fixed effects are added. This strongly backs the results and policy conclusions of Shapiro (2021) in the context of an applied microlevel approach, controlling for several firm characteristics. Accordingly, firms producing in upstream industries show a dirtier production, indicated by higher CO2emissions and CO2 emission intensities. Notably, comparing the results of CO2intensity per unit of sales with CO2 58 intensity per unit of value added reveals stronger results for using value added as a measure of CO2intensity. A change in the upstreamness measure by one unit, such as switching from the average GVC position of firms in the textiles industry to the basic metals industry - increases the estimated CO2emissions by 28.15%, CO2per unit of sales by 26.87%, and CO2per unit of value added by 35.39%.10 Figure 4.4 illustrates the estimated level of absolute CO2emissions and CO2intensities for different GVC positions, once all other variables are held constant at their means. The specifications depicted are Columns (4), (5), and (6) of Table 4.3. 1.4 1.6 1.8 2 2.2 2.4 Linear Prediction of (ln) CO2 p1p10 p20 p30 p40 p50 p60 p70 p80 p90 p99 Upstreamness (percentiles) Predictive Margins with 95% CIs -1 -.8 -.6 -.4 -.2 0 Linear Prediction of (ln) (CO2/SA) p1p10 p20 p30 p40 p50 p60 p70 p80 p90 p99 Upstreamness (percentiles) Predictive Margins with 95% CIs -.4 -.2 0 .2 .4 .6 Linear Prediction of (ln) (CO2/VA) p1p10 p20 p30 p40 p50 p60 p70 p80 p90 p99 Upstreamness (percentiles) Predictive Margins with 95% CIs Figure 4.2: Estimated relationship between GVC positioning and CO2emissions and CO2 intensities, respectively (Columns (4), (5), and (6) of Table 4.3) At the first percentile, the estimated level of CO2emission is ln(1.65) at a upstreamness of 1.43, while the effect at the 99th percentile is ln(2.14) at a upstreamness of 3.39. Taking the exponential function to obtain the value in kt leads to a magnitude of 5.21 kt and 8.50 kt, respectively. This corresponds to an increase of 63.23%, which is arguably a big difference. A similar pattern is observable for the estimated level of CO2emissions per unit of sales at the respective percentiles. The estimated level at the first percentile is ln(-0.81) and at the 99th percentile ln(-0.34). Taking the respective exponential function of the values leads to 0.44 kg CO2and 0.71 kg CO2emissions per US dollar sold. This corresponds with 60.0% also to an economically meaningful increase. Using value added as a measure of CO2intensity reveals a difference of even 82.21%. The estimated level at the first percentile is ln(-0.27) and at the 99th percentile ln(0.33). Taking the 10The derived change in percentage is calculated by solving: (exp(coefficient)-1)*100. 59 respective exponential function of the values leads to 0.76 kg CO2and 1.39 kg CO2emissions per US dollar value added. All in all, the results statistically and economically back Hypotheses 1 and 2 and show that also within industries, firms in upstream positions produce dirtier than firms in downstream positions.11 The following section focuses on an instrumental variable approach to deal with possible sources of endogeneity. 4.4.2 Instrumental Variable Approach The above-presented results on H1 and H2 are biased if a firm’s environmental outcome affects its GVC position or if an omitted variable simultaneously affects a firm’s upstreamness and the environmental outcome in production. An example of reverse causality is where a firm introduces a clean product to the product mix to avoid reputational risks. Generally, consumers might be more likely to claim sustainable production on products they are familiar with, typically in a more downstream GVC position. To this end, the introduction of a clean product paves the way to get closer to customers, while reducing reputational risks. A relatively similar threat to identification would arise if an omitted variable simultaneously affects GVC positioning and clean production. For instance, a firm’s management has social preferences and favours simultaneously B2C (business to consumers) products in downstream positions and introduces a green strategy to improve the environmental footprint in production. In this case, the revealed effect occurs due to the social preferences of the management that are not covered in the firm level data. I employ an instrumental variable estimation to overcome such endogeneity concerns. By construction, a firm can only influence its GVC position by changes in its product mix. The GVC positioning of each product is already beyond the scope of the management because it is measured at the industry level. However, to exclude any influence by the management, I instrument upstreamness with the natural logarithm of the advertising spending over sales among firms producing a similar main product measured at the five-digit level ((ln)advertising/SA). The instrument does not include the respective spending and sales of the firm. The instrument should meet two criteria to be valid. First, the instrument needs to be relevant, so the correlation between upstreamness and the instrument should be sufficiently high. I argue that the incentive of a firm for advertising increases along the GVC. Accordingly, firms producing the main product that is closer to final consumption spend more on advertising, on average. Generally, if a correlation is sufficient cannot finally be stated, but the correlation should be significant and a high F-statistic above 104.7 serves as a very conservative rule of thumb for validation (Lee et al., 2022). Second, a valid instrument meets the exclusion restriction. In line with the exclusion restriction, the instrument should not affect the environmental outcome of a firm, except through the GVC position. The latter is objectively difficult to evalu11Going beyond upstreamness and taking Column (6) as a benchmark to measure the drivers of firms’ clean production reveals that exporters, foreign-owned firms, more productive firms and firms active in R&D show lower CO2intensity. Conversely, state-owned firms and firms producing mining or agricultural products produce more CO2-intensive. No effect is observable for purely manufacturing firms or service firms. Lastly, exposure to EPS does not relate to any changes in a firm’s clean production. 60 ate. However, I argue that the exclusion of a firm’s spending on advertising and sales fulfils the exclusion restriction because a firm does not have any influence on the value of the instrument. Moreover, if a firm decides to extend or reduce the product mix, this normally does not affect the main product. Accordingly, short-term changes due to management preferences do not enter the estimation and the respective instrumental variable remains exogenously given for a firm. Table 4.4 includes the results of a 2-SLS estimation. The first column reports on the first stage, estimating the respective value of a firm’s � upstreamness. The specification does not include industry fixed effects because there is no sufficient correlation between the instrument and upstreamness at the first stage once the estimation focuses on the within industry variation. However, this is not completely surprising, because it is quite restrictive in the light of a high correlation of average advertising spending over sales estimated at the five-digit product level within an industry. Accordingly, the 2-SLS estimation does closely link to the specifications covered in Columns (1)-(3) of Table 4.3, overall linking to the results at the aggregated level by Shapiro (2021). Columns (2)-(4) of Table 4.4 refer to the main results of the 2-SLS estimation, including the estimated values of � upstreamness as an instrument for firms’ GVC positions. Table 4.4: 2-SLS Estimation (1) (2) (3) (4) � Upstream. CO2 CO2/SA CO2/VA VARIABLES OLS 2-SLS 2-SLS 2-SLS (ln) advertising/SA -0.123*** (0.007) � Upstreamness 1.632*** 1.215*** 1.854*** (0.169) (0.152) (0.172) EPS -2.874*** 3.905*** 2.889*** 4.162*** (0.102) (0.573) (0.520) (0.587) Export status 0.075*** -0.540*** -0.670*** -0.663*** (0.023) (0.070) (0.066) (0.071) Foreign ownership 0.166*** -0.509*** -0.512*** -0.759*** (0.046) (0.136) (0.133) (0.144) R&D dummy -0.005 -0.060 -0.209** -0.300*** (0.031) (0.090) (0.087) (0.096) SOE 0.178 1.387*** 1.218** 1.181** (0.132) (0.519) (0.549) (0.523) (ln) productivity 0.018** 0.029 -0.215*** -0.388*** (0.008) (0.026) (0.024) (0.026) (ln) assets -0.012* 0.990*** 0.073*** 0.116*** (0.007) (0.023) (0.022) (0.024) (ln) age -0.029** 0.342*** 0.154*** 0.103** (0.013) (0.044) (0.041) (0.044) Agricultural products -0.294*** 1.124*** 0.864*** 1.202*** (0.037) (0.132) (0.128) (0.137) Mining products 0.103** 0.720*** 0.988*** 0.740*** (0.043) (0.188) (0.185) (0.187) Service products 0.053 -0.060 -0.225* -0.163 (0.032) (0.115) (0.116) (0.120) Exclusive manufacturing -0.032 0.274** 0.221* 0.254* (0.036) (0.125) (0.126) (0.129) Observations 16,803 16,803 16,803 16,803 R-squared 0.410 F-stat 283.69 Year dummy Yes Yes Yes Yes Industry dummy No No No No State dummy Yes Yes Yes Yes Robust standard errors in parentheses (clustered at the firm level) *** p<0.01, ** p<0.05, * p<0.1 The results of the first stage reveal that (ln)advertising/SA significantly negatively correlates with a firm’s upstreamness. Also, the F-statistic is 283.69 clearly above standard thresholds. Both patterns provide a high level of confidence that the instrument is relevant. The positive relationship between firms’ upstreamness and dirty production remains unchanged in all specifications. In all specifications, the positive effect is significant at the onepercent level, which again strongly supports H1 and H2. As stated in Hypotheses 3 and 4, the negative effect of upstreamness on the environmental 61 outcome might differ with a firm’s export status and exposure to EPS. The next section analysis the validity of these hypotheses. 4.4.3 The Moderating Roles of Export Status and Exposure to EPS Columns (1) to (3) of Table 4.5 show the results of an interaction term between upstreamness and export status. Columns (4)-(6) include the interaction term between upstreamness and exposure to EPS. Table 4.5: OLS estimation: Moderating effects of export status and EPS on upstreamness (1) (2) (3) (4) (5) (6) VARIABLES CO2 CO2/SA CO2/VA CO2 CO2/SA CO2/VA Upstreamness 0.225** 0.206* 0.270** 0.988*** 0.768*** 0.886*** (0.112) (0.114) (0.114) (0.298) (0.289) (0.299) Export status -0.364** -0.595*** -0.543*** -0.242*** -0.427*** -0.368*** (0.181) (0.174) (0.177) (0.055) (0.054) (0.054) EPS -0.045 0.124 0.032 1.614*** 1.307** 1.335** (0.237) (0.233) (0.228) (0.609) (0.581) (0.592) Upstreamness x Export status 0.056 0.077 0.080 (0.081) (0.077) (0.079) Upstreamness x EPS -0.809*** -0.579** -0.637** (0.298) (0.277) (0.288) Foreign ownership -0.284*** -0.337*** -0.514*** -0.282*** -0.336*** -0.513*** (0.109) (0.110) (0.111) (0.109) (0.110) (0.112) R&D dummy -0.011 -0.174** -0.256*** -0.011 -0.172** -0.255*** (0.070) (0.070) (0.072) (0.070) (0.070) (0.072) SOE 1.374*** 1.198*** 1.195*** 1.370*** 1.194** 1.190*** (0.460) (0.463) (0.359) (0.468) (0.473) (0.367) (ln) productivity 0.037* -0.207*** -0.383*** 0.037* -0.208*** -0.384*** (0.021) (0.020) (0.020) (0.021) (0.020) (0.020) (ln) assets 0.947*** 0.038** 0.075*** 0.947*** 0.038** 0.074*** (0.019) (0.018) (0.019) (0.019) (0.018) (0.019) (ln) age 0.199*** 0.035 -0.048 0.201*** 0.036 -0.047 (0.034) (0.033) (0.033) (0.034) (0.033) (0.033) Agricultural products 0.351*** 0.219** 0.431*** 0.353*** 0.219** 0.430*** (0.106) (0.109) (0.111) (0.106) (0.109) (0.111) Mining products 0.651*** 0.878*** 0.698*** 0.637*** 0.866*** 0.685*** (0.192) (0.198) (0.190) (0.192) (0.197) (0.190) Service products 0.032 -0.123 -0.054 0.034 -0.123 -0.054 (0.106) (0.110) (0.110) (0.106) (0.109) (0.109) Exclusive manufacturing 0.097 0.101 0.096 0.094 0.096 0.092 (0.114) (0.116) (0.116) (0.113) (0.116) (0.116) Observations 17,541 17,541 17,541 17,541 17,541 17,541 R-squared 0.627 0.341 0.365 0.627 0.341 0.365 Year dummy Yes Yes Yes Yes Yes Yes Industry dummy Yes Yes Yes Yes Yes Yes State dummy Yes Yes Yes Yes Yes Yes Robust standard errors in parentheses (clustered at the firm level) *** p<0.01, ** p<0.05, * p<0.1 While the overall positive relationship between upstreamness and dirty production remains significant in all specifications, the interaction term of upstreamness and export status is not significant in any specification. Accordingly, there is no significant different slope for the relationship between upstreamness and dirty production depending on the export status of a firm observable. This finding does not back H3, pointing to possible differences in the magnitude of the technique effect along the value chain. Conversely, the results of Columns (4)-(6) reveal a moderating role of demanding export markets on a firm’s environmental performance. The interaction term between upstreamness and EPS is significant in all specifications, varying between five-percent and one-percent significance. This finding indicates that the main relationship between upstreamness and dirty production is conditional on the interaction because the positive slope diminishes with an increasing level of exposure to strong environmental regulation in foreign markets. Figure 4.3 illustrates the moderating role of EPS on the relationship between upstreamness and CO2emissions per unit of value added (Column (6)).12 The upper graph illustrates the 12The respective graphs for CO2emissions (Column (4)) and CO2emissions per unit of sales (Column (5)) 62 predicted CO2per unit of value added depending on upstreamness for given levels of EPS. The lower graph illustrates the predicted CO2intensities depending on EPS for given levels of upstreamness. -.5 -.4 -.3 -.2 -.1 0 .1 .2 .3 .4 .5 .6 Predicted (ln) (CO2/VA) 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 Upstreamness EPS = 0.7 EPS = 0.9 EPS = 1.1 EPS = 1.3 EPS = 1.5 Predictive Margins -.5 -.4 -.3 -.2 -.1 0 .1 .2 .3 .4 .5 .6 Predicted (ln) (CO2/VA) .7 .8 .9 1 1.1 1.2 1.3 1.4 1.5 1.6 EPS Upstr. = 1.4 Upstr. = 1.8 Upstr. = 2.2 Upstr. = 2.6 Upstr. = 3.0 Upstr. = 3.4 Predictive Margins Figure 4.3: Predicted (ln) CO2per unit of value added for different combinations of upstreamness and EPS An EPS score of 1.5 links to very strong environmental regulation in export destination markets. Given that firms face such a strong regulation, the slope of upstreamness is slightly negative, even if it is not significant. Conversely, an EPS score of very low 0.7 links to a strongly positive slope. The respective coefficient of upstreamness is 0.44 significant at the one-percent level. Holding the EPS score constant, it is also possible to derive the economic significance. For example, for firms exposed to weak environmental regulation in foreign markets (EPS = 0.7), a one-unit change in the upstreamness measure is associated with an increase of CO2emissions of 52.46%, CO2emissions per unit of sales of 43.72%, and CO2emissions per unit of value added of 55.29%.13 Interestingly, the relationship between EP S and dirty production becomes significantly positive once the interaction term between upstreamness and EPS is added. Also, the relationship between EPS and clean production is conditional on firms’ GVC positions. As the lower graph of Figure 4.3 illustrates, firms’ CO2intensity increases with EPS for firms in very downstream positions (upstreamness = 1.4). However, the derived coefficient of 0.44 is only significant at the 10-percent level. For firms in very upstream positions (upstreamness = 3.4), the coefficient are part of the Appendix (Figures 4.5 and 4.6). 13For example, the respective coefficient of CO2emissions is calculated by solving: 0.988-0.809*0.7=0.42. The effect is given by: (exp(0.42)-1)*100=52.46. 63 turns negative to 0.83, again being significant only at the 10-percent level. In line with the only slight significance and given that a one-unit change of EPS relates to a difference in exposure going slightly beyond the difference of the minimum and maximum value observed in the sample, the economic significance should not be over-interpreted.14 Still, these relationships indicate some specialisation in dirty production in downstream positions. All in all, the results revealed in Table 4.5 strongly back Hypothesis 4 and its economic relevance. This findings indicates that the relationship between upstreamness and dirty production diminishes with EPS for firms in upstream positions. Accordingly, firms in upstream positions seem to environmentally benefit outstandingly from exposure to strong environmental regulations in foreign markets. This insight links to the adoption of more efficient production techniques to remain competitive with firms in environmentally demanding markets, tradeinduced learning-by-exporting effects, demand-pull, and regulation-push especially important for firms being further removed from the technology frontier in upstream positions. Conversely, firms in downstream positions show dirtier production with increased exposure to strong environmental regulation in foreign markets. The latter links to a specialisation in dirty production in line with Acemoglu et al. (2014). The next section focuses on the robustness of the key results. 4.4.4 Robustness Unfortunately, not all firms report on their energy consumption in physical energy units. Accordingly, it is only possible to calculate annual CO2emissions for a subset of firms covered in Prowess. However, most firms report on their energy spending in US dollars.15 Energy costs are widely used in the literature to measure firms’ environmental performances, for example, Eskeland and Harrison (2003) and Batrakova and Davies (2012). Cole et al. (2008) also use energy costs to proxy firms’ environmental performance. However, they emphasise that their dependent variable should ideally capture a firm’s emissions. I follow this suggestion in the main specifications because there are two main drawbacks of using energy costs. First, price changes in energy costs result in changes on the environmental performance measure, even if it does not affect a firm’s environmental performance. Second, switching to cheaper energy sources results in an improvement of the measure, even though the newly used energy source can show a worse or better CO2footprint. Nonetheless, both proxies are highly correlated, which motivates this robustness check. Table 4.6 reports on the results once energy costs are used to calculate firms’ clean production in absolute and relative terms. Columns (1)-(3) connect to the main specifications on the relationship between upstreamness and dirty production discussed in Section 4.4.1. Columns (4)-(6) link to the main results outlined in Section 4.4.3 on the moderating role of exposure to EPS. 14For example, for firms in very downstream positions (upstreamness = 1.4), a one-unit change in the EPS measure is associated with an increase of CO2emissions of 61.83%, CO2emissions per unit of sales of 64.28%, and CO2emissions per unit of value added of 55.77%. Turning to very upstream positions, the sign of the relationship between EPS and clean production shifts positively. For example, for firms in very upstream positions (upstreamness = 3.4), a one-unit change in the EPS measure is associated with a decrease of CO2emissions of 67.91%, CO2emissions per unit of sales of 48.4%, and CO2emissions per unit of value added of 56.43%. 15In addition, using energy costs as dependent variable does not require any assumption about the CO2 conversion factor, e.g., the emissions per kwh can vary substantially between regions and time. 64 Table 4.6: OLS estimation: Robustness checks (1) (2) (3) (4) (5) (6) VARIABLES ENE ENE/SA ENE/VA ENE ENE/SA ENE/VA Upstreamness 0.344*** 0.340*** 0.424*** 1.089*** 0.829*** 1.012*** (0.069) (0.064) (0.072) (0.210) (0.195) (0.214) EPS 0.141 0.236** 0.207* 1.723*** 1.274*** 1.455*** (0.127) (0.119) (0.123) (0.402) (0.370) (0.395) Upstreamness x EPS -0.798*** -0.524*** -0.630*** (0.194) (0.180) (0.193) Export status 0.145*** -0.152*** -0.076** 0.145*** -0.152*** -0.075** (0.034) (0.032) (0.032) (0.034) (0.032) (0.032) Foreign ownership 0.154** -0.025 -0.167*** 0.155** -0.024 -0.166*** (0.062) (0.060) (0.062) (0.062) (0.060) (0.062) R&D dummy 0.069* -0.121*** -0.191*** 0.069* -0.122*** -0.191*** (0.039) (0.038) (0.039) (0.039) (0.038) (0.039) SOE 0.746*** 0.678*** 0.779*** 0.747*** 0.678*** 0.779*** (0.264) (0.220) (0.166) (0.265) (0.221) (0.166) (ln) productivity -0.014 -0.301*** -0.487*** -0.014 -0.301*** -0.486*** (0.014) (0.012) (0.014) (0.014) (0.012) (0.014) (ln) assets 0.812*** -0.121*** -0.073*** 0.812*** -0.121*** -0.073*** (0.011) (0.010) (0.010) (0.011) (0.010) (0.010) (ln) age 0.212*** 0.066*** 0.003 0.213*** 0.067*** 0.004 (0.019) (0.018) (0.019) (0.019) (0.018) (0.019) Agricultural products 0.561*** 0.432*** 0.614*** 0.565*** 0.435*** 0.617*** (0.077) (0.072) (0.077) (0.077) (0.072) (0.077) Mining products 0.218** 0.409*** 0.280*** 0.206** 0.402*** 0.270** (0.104) (0.102) (0.105) (0.104) (0.103) (0.105) Service products 0.082 0.006 0.034 0.086 0.009 0.038 (0.080) (0.077) (0.078) (0.080) (0.077) (0.078) Exclusive manufacturing 0.027 0.135* 0.115 0.027 0.135* 0.116 (0.082) (0.080) (0.080) (0.083) (0.080) (0.080) Observations 42,762 42,762 42,762 42,762 42,762 42,762 R-squared 0.632 0.375 0.416 0.632 0.375 0.416 Year dummy Yes Yes Yes Yes Yes Yes Industry dummy Yes Yes Yes Yes Yes Yes State dummy Yes Yes Yes Yes Yes Yes Robust standard errors in parentheses (clustered at the firm level) *** p<0.01, ** p<0.05, * p<0.1 The relationship between upstreamness and environmental performance is statistically significant at the one-percent level in all six specifications. In addition, the interaction between upstreamness and EPS is significant in all three specifications. To summarise, the robustness checks back prior results on H1, H2, and H4. Accordingly, firms in downstream positions show significantly cleaner production and exposure to strong environmental regulation in export markets diminishes the negative effect of upstreamness on clean production. The last section concludes the overall results of the study. 4.5 Conclusion A large share of India’s population still lives in poverty. Economic growth is essential to eliminate poverty but corresponds to a higher energy demand. Higher energy demand strongly correlates to higher CO2emissions, a nexus that has not been sufficiently broken. Even though the per capita CO2emissions of India are relatively low on global scale, the country is the third largest global emitter of CO2. Although industrialised countries are responsible for the majority of historical CO2emissions, successes in developing countries, including India, are crucial if there is any hope of fulfilling the Paris Agreement’s target of limiting global warming to well below 1.5°Celsius. This is mainly because most emissions growth originates from this subset of world’s economies (Copeland et al., 2022). To this end, regional and global decision makers need to understand the determinants of firms’ environmental efficiency in production. Shapiro (2021) reveals that industries in downstream GVC positions emit less CO2per unit of output. However, even in narrowly defined industries, firms differ substantially, for instance in their productivity and product mix. I empir65 ically contribute to the literature on the understanding of the drivers of clean production, more precisely on the question of how firms’ GVC positioning relates to its environmental outcome. The derived results show that Indian manufacturing firms occupying upstream GVC positions emit significantly more CO2and report worse CO2intensities, measured as CO2per unit of sales, and as CO2per unit of value added. A change in the upstreamness measure by one unit - such as switching from the average GVC position of firms in the textiles industry to the basic metals industry - increases the estimated CO2emissions by 28.15%, CO2per unit of sales by 28.87%, and CO2per unit of value added by 35.39%. These key findings are robust against a 2-SLS estimation. In addition, the empirical results reveal that exposure to foreign EPS negatively moderates the relationship between upstreamness and dirty production. Firms in upstream positions are, on average, further removed from the technology frontier and have to adopt more efficient production techniques to remain competitive in environmentally demanding markets abroad. These upstream firms can also disproportionately benefit from trade-related positive drivers of their clean production such as learning-by-exporting, regulation-push, and demand-pull. Interestingly, for firms in downstream positions, environmental performance deteriorates with increased exposure to EPS in export markets - a finding linked to specialisation in dirty production (Acemoglu et al., 2014). Notably, the negative relationship between upstreamness and clean production does not differ with the firm’s export status. My study contributes to the understanding of the determinants of firms’ clean production in developing countries. Copeland et al. (2022) emphasise that the relationship between upstreamness and dirty production is observable in countries at all stages of development. I would expect similar results using firm level data for advanced countries. However, I expect that the moderating role of EPS in foreign markets might distinguish between studies using data for advanced countries. Firms occupying upstream positions and already facing strong domestic environmental regulation might not have to step up to new production techniques, even if they encounter strong EPS abroad. Such an analysis could represent part of a future research agenda. Moreover, in this paper, I have not touched on the role of indirect emissions. The relative importance of indirect emissions amplifies with downstreamness and it would be interesting to investigate whether the revealed upstreamness is also observable for indirect emissions (Copeland et al., 2022). The key results have different implications for a firm’s management and political decision makers. First, management can improve the overall environmental performance by introducing relatively clean products in downstream positions. A switch of the product mix to include cleaner products does not require any technological adoption and might be an attractive tool for the management to improve the general perception of stakeholders on sustainable production, such as when publishing statistics relating to CO2emission intensity, for outside investors. Arguably, the switch to downstream products is a form of greenwashing because it does not require the firm to make any fundamental changes in its processes - merely, in its product mix. Accordingly, this strategy is open to criticism by more sophisticated kind of stakeholder - those able to monitor the firm’s performance at the product level. Second, regional decision makers have an incentive to attract downstream industries to 66 improve their regional CO2balance sheet. However, on a global scale, the negative effect of GHG emissions on global warming does not depend on the location of emissions. Accordingly, and in line with the environmental bias of trade policy revealed by Shapiro (2021), tariffs for downstream industries should be reduced, bolstering support for them relative to upstream industries. Currently, the pattern is completely reversed, leading to an implicit subsidy for the relatively CO2-intensive upstream industries. Third, regulators should position the spotlight on the CO2intensity of upstream industries, where more energy is required to produce one unit of value added. The issue is how to incentivice consumers and other stakeholders interested in sustainable production, to expend higher efforts to monitor these environmental outcomes for upstream GVC positions, where end-consumers may be less directly invested in the products. Some of these upstream industries are tagged as hard to decarbonise, e.g. chemicals, cement or steel production. However, this does not preclude regulators from applying specific measures. One approach would be to reduce or eliminate the implicit subsidy for upstream industries by decreasing tariffs for downstream positions. Another, would be to establish an appropriate CO2price mechanism and to incentivise environmentallyfriendly inputs such as renewable energies, renewable natural gas or green hydrogen. A judicious use of these measure could accelerate the green transition of these industries. Lastly, monitoring global efforts on CO2reduction, multilateral institutions should consider differences in the average GVC positions of individual countries. Upstream industries generally report reduced environmental efficiency, but they are integral to GVCs, whose final production stages are often located in other – more developed - countries. The latter depend on these inputs. Accordingly, these countries are indirectly contributing to global pollution, when the GVC is viewed in its entirety. A comprehensive assessment of environmental footprints - consumption-based evaluation vs. production-based evaluation - could help overcome this distortion. Additionally, the application of a price which would correct for the environmental externality, for example CO2border adjustment taxes, would provide an incentive to downstream firms to push for sustainable production along the entire GVC. 67 Forslid, R., T. Okubo, and K. H. Ulltveit-Moe (2018): “Why are firms that export cleaner? International trade, abatement and environmental emissions,” Journal of Environmental Economics and Management, 91, 166 – 183. Ghisetti, C., A. Marzucchi, and S. Montresor (2015): “The open eco-innovation mode. An empirical investigation of eleven European countries,” Research Policy, 44, 1080–1093. Ghisetti, C. and F. Pontoni (2015): “Investigating policy and R&D effects on environmental innovation: A meta-analysis,” Ecological Economics, 118, 57–66. Girma, S. and A. Hanley (2015): “How green are exporters?” Scottish Journal of Political Economy, 62, 291–309. Goldberg, P. K., A. K. Khandelwal, N. Pavcnik, and P. Topalova (2010a): “Imported intermediate inputs and domestic product growth: Evidence from India,” The Quarterly Journal of Economics, 125, 1727–1767. ——— (2010b): “Multiproduct firms and product turnover in the developing world: Evidence from India,” The Review of Economics and Statistics, 92, 1042–1049. Guti´ errez, E. and K. Teshima (2018): “Abatement expenditures, technology choice, and environmental performance: Evidence from firm responses to import competition in Mexico,” Journal of Development Economics, 133, 264–274. Hanley, A. and F. O. Semrau (2022): “Stepping up to the mark? Firms’ export activity and environmental innovation in 14 European countries,” Industry and Innovation, 29, 672–700. Herkenhoff, P., S. Krautheim, F. O. Semrau, and F. Steglich (2021): “Corporate Social Responsibility along the global value chain,” Working Paper 9498, CESifo, Munich. Horbach, J. (2008): “Determinants of environmental innovation - New evidence from German panel data sources,” Research Policy, 37, 163 – 173. ——— (2016): “Empirical determinants of eco-innovation in European countries using the community innovation survey,” Environmental Innovation and Societal Transitions, 19, 1–14. Horbach, J., C. Rammer, and K. Rennings (2012): “Determinants of eco-innovations by type of environmental impact - The role of regulatory push/pull, technology push and market pull,” Ecological Economics, 78, 112–122. Jaffe, A. B., R. G. Newell, and R. N. Stavins (2005): “A tale of two market failures: Technology and environmental policy,” Ecological Economics, 54, 164–174. Kannen, P., F. O. Semrau, and F. Steglich (2021): “Green gifts from abroad? FDI and firms’ green management,” Working Paper 2200, Kiel Institute for the World Economy, Kiel. Kesidou, E. and P. Demirel (2012): “On the drivers of eco-innovations: Empirical evidence from the UK,” Research Policy, 41, 862–870. 74 Kong, D., G. Ma, and N. Qin (2022): “Trade liberalization and firm toxic emissions,” Review of International Economics. Lee, D. L., J. McCrary, M. J. Moreira, and J. Porter (2022): “Valid t-ratio Inference for IV,” Working Paper 29124, National Bureau of Economic Research Working Paper, Cambridge, MA 02138. Levinsohn, J. and A. Petrin (2003): “Estimating production functions using inputs to control for unobservables,” The Review of Economic Studies, 70, 317–341. Manj´ on, M., J. Manez, et al. (2016): “Production function estimation in Stata using the Ackerberg-Caves-Frazer method,” Stata Journal, 10, 288–308. Melitz, M. J. (2003): “The impact of trade on intra-industry reallocations and aggregate industry productivity,” Econometrica, 71, 1695–1725. Newman, C., J. Rand, F. Tarp, and N. Trifkovic (2018): “The transmission of socially responsible behaviour through international trade,” European Economic Review, 101, 250– 267. Olley, G. S. and A. Pakes (1996): “The dynamics of productivity in the telecommunications equipment industry,” Econometrica, 64, 1263–1297. Rennings, K. (2000): “Redefining innovation - eco-innovation research and the contribution from ecological economics,” Ecological Economics, 32, 319–332. Richter, P. M. and A. Schiersch (2017): “CO2emission intensity and exporting: Evidence from firm-level data,” European Economic Review, 98, 373–391. Shapiro, J. S. (2021): “The environmental bias of trade policy,” The Quarterly Journal of Economics, 136, 831–886. Shapiro, J. S. and R. Walker (2018): “Why is pollution from US manufacturing declining? The roles of environmental regulation, productivity, and trade,” The American Economic Review, 108, 3814–54. The World Bank (2019): World Bank Development Indicators. Timmer, M. P., E. Dietzenbacher, B. Los, R. Stehrer, and G. J. Vries (2015): “An illustrated user guide to the world input–output database: the case of global automotive production,” Review of International Economics, 23, 575–605. Zhu, Q., J. Cordeiro, and J. Sarkis (2012): “International and domestic pressures and responses of Chinese firms to greening,” Ecological Economics, 83, 144–153. 75 Chapter 5 Corporate Social Responsibility along the Global Value Chain∗† Philipp Herkenhoff University of Mainz philipp.herkenhoff@uni-mainz.de Sebastian Krautheim‡ University of Passau CESifo fsebastian.kr[email protected] Finn Ole Semrau Kiel Institute for the World Economy (IfW) Kiel Centre for Globalization (KCG) finn-ole.semr[email protected] Frauke Steglich Kiel Institute for the World Economy (IfW) frauke.ste[email protected] Abstract Locating substantial parts of the production process in developing and emerging economies, many firms face an increasing demand by stakeholders for Corporate Social Responsibility (CSR) along their value chains. Contractual incompleteness between firms and their suppliers at different stages of production can exacerbate the ability to meet these demands. We analyze a model of sequential production with incomplete contracts where CSR by independent suppliers differentiates the final product in the eyes of caring consumers. Progressing down the value chain, our model predicts an increasing CSR profile from upstream suppliers with low CSR to downstream suppliers with higher CSR. We confirm this prediction using Indian firm-level data – computing a firm’s value chain position by combining its product-level sales information with the World Input-Output Database. We find that more downstream firms report higher CSR expenditures as measured by a combination of staffwelfare spending and social community spending. ∗The chapter is an earlier version of: “Herkenhoff, P., Krautheim, S., Semrau, F. O., & Steglich, F. (2024). “Corporate Social Responsibility along the global value chain”. Journal of Development Economics, 167, 103236.” The published article is accessible via https://doi.org/10.4337/9781802200065.00016. An earlier version of the manuscript has been published as: Herkenhoff, P.; Krautheim, S.; Semrau, F.O. and Steglich, F. (2021). “Corporate Social Responsibility Along the Global Value Chain.” CESifo Working Paper No. 9498. Please note that the writing style of the chapter differs due to submission requirements, such as writing in American English. †We would like to thank Julia Cajal-Grossi, Ron Davies, Ben Faber, Gabriel Felbermayr, Sourafel Girma, Robert Gold, Holger G¨org, Aoife Hanley, Roza Khoban, Krisztina Kis-Katos, Carol Newman, Marcelo Olarreaga, Gianmarco Ottaviano, Philip Saur´e, Jan Schymik, Rainer Thiele, Farid Toubal, Gonzague Vannoorenberghe, Yoto Yotov, participants of the HSU-IfW Workshop 2019, the PEGNet Conference 2019, the EGIT 2020, the launch event for the Research Network Sustainable Global Supply Chains 2021, the CESifo Area Conference on Global Economy 2021, the Workshop ‘Shaping Globalization’ at JGU Mainz 2021, the ETSG 2021, the G¨ottingen Workshop on International Economics 2022, the TRISTAN Workshop at the University of Bayreuth 2022, as well as seminar participants at Trinity College Dublin, University of Passau, University of Nottingham, JGU Mainz, Kiel Institute for the World Economy (IfW), Kiel Centre for Globalization (KCG), and the University of G¨ottingen. Krautheim gratefully acknowledges financial support by the German Research Foundation (DFG) grant no. 421074032; Semrau and Steglich acknowledge financial support by the Leibniz Association through the KCG grant no. SAS-2016-IfW-LWC. ‡Corresponding author 76 JEL Classification Numbers: F12, F61, F63, L23, M14, O12 Keywords: Corporate Social Responsibility; global value chains; incomplete contracts; sustainable development; ethical sourcing; India 77 5.1 Introduction Participation in global value chains (GVCs) is often considered a path to development for low and middle-income countries (OECD/WTO, 2013; World Bank, 2020). Concerns about negative social and environmental impacts of GVCs in those countries are, however, widespread and boosted by tragedies such as the Rana Plaza collapse in 2013. Consumers have responded to these concerns by becoming increasingly vocal about their demand for ethical standards along the entire value chain. With large parts of today’s value chains located in developing and emerging economies characterized by high leniency with regard to environmental or labor regulation and/or enforcement, this requires that suppliers invest in Corporate Social Responsibility (CSR). CSR investments – by definition – extend beyond local legal requirements (e.g., McWilliams and Siegel, 2001) and are therefore key for meeting the demands of an ever growing number of ethically concerned consumers and, more broadly, for the general acceptance (or not) of globalized production in the public debate. Persistent and sizable compliance issues concerning labor and environmental practices in the value chains of a large number of firms indicate, however, how difficult it can be for firms to impose CSR levels on their international suppliers, e.g., through legally binding codes of conduct (e.g., Locke, 2013b). At the same time, relying on the individual self-interest of a large number of independent suppliers at different positions along the value chain is unlikely to deliver optimal levels of CSR (from the perspective of the firm). With this paper, we contribute to a better understanding of the factors that drive CSR along the global value chain. More specifically, we ask the following questions: Which role does the value chain position of suppliers play in their CSR investments? What are the resulting patterns of CSR along a value chain, from upstream to downstream production stages? We make two main contributions. First, we analyze a model of non-contractible CSR investments by suppliers in a sequential production process (a value chain). In our model, CSR investments of suppliers in the Global South provide horizontal and vertical product differentiation when selling the final product to caring consumers in the Global North. We find that the equilibrium CSR investments are low in upstream stages, increase along the value chain, and peak at the most downstream stages. Second, we test this prediction employing Indian firm-level data. Combining detailed information on a firm’s product mix with the World Input-Output Database (WIOD), we construct a measure of a firm’s GVC position to empirically investigate how it affects a firm’s monetary CSR spending. Our empirical analysis supports the theoretical prediction: CSR expenditure increases along the value chain. In our model, firms and consumers are located in the Global North, while suppliers are located in the Global South and operate under lenient regulation. We introduce a group of ‘caring’ consumers who appreciate a product’s ‘ethical quality’ embodied in environmental and welfare improvements (CSR) implemented by suppliers along the value chain.1In line with the findings in Bartling et al. (2015), in our model, CSR along the value chain provides product differentiation for the final product. Higher CSR investments by suppliers therefore allow the firm to charge a higher markup.2 A central premise of our analysis is that suppliers make their decisions on CSR in an environment characterized by incomplete contracts: the firm selling the final good to consumers cannot write binding contracts prescribing production standards in supplier factories. We argue in detail in Section 5.2.1 that for the case of CSR, contracts are incomplete as the supplier’s action is “observable, but not verifiable” 1Surveys (e.g., O’Rourke, 2005, and Loureiro and Lotade, 2005) as well as field experiments with real purchasing decisions (e.g., Hiscox and Smyth, 2011, and Hainmueller et al., 2015) suggest that consumers do care about the ethical content of their consumption and in fact have a higher willingness to pay for ‘ethical’ products. 2This implies that we abstract from massive consumer boycotts or well-mediatized scandals, which depress demand of targeted firms. Such events may play a role for generating or expanding the group of caring consumers. We take the existence of caring consumers as given and rather model CSR investments along the value chain as opportunities to increase the perceived (ethical) quality of the product by incrementally contributing CSR from production stage to production stage (see Section 5.2.2.1). 78 (Hart and Moore, 1999, p. 118). Central to the argument is the difference between circumstantial evidence on a supplier’s infringements of ethical standards (observability) on the one hand and, on the other hand, evidence that will hold in a court of law (verifiability). While circumstantial evidence can be sufficient for caring consumers to re-assess their valuation of the product and therefore demand, it is insufficient in a court of law, especially so in countries with weak legal institutions.3 We introduce the above elements – caring consumers and incomplete contractability of CSR – into a model of sequential production along the lines of Antr`as and Chor (2013). We find an increasing profile of the suppliers’ equilibrium CSR investments along the value chain: while CSR investments at upstream stages are low, they increase along the value chain towards more downstream stages. This pattern arises because under incomplete contracts, the cumulative CSR investment in previous stages provides the supplier with an incentive to choose a higher CSR level. We derive the complete industry equilibrium with free entry and find that endogenizing the number of firms has important implications for identifying testable implications. While in the partial equilibrium (taking aggregate variables as given), CSR investments increase in the total expenditure on ethically differentiated goods, this effect vanishes in the industry equilibrium. This leaves us with a single and clear testable implication of our model: the increasing profile of CSR along the value chain, which we take to the data in Section 3. Several policy implications directly emerge from our theory. First, as higher CSR levels at upstream stages encourage CSR investments at more downstream stages (sequential complementarity), policy interventions targeting CSR at upstream stages may benefit CSR along the entire value chain. Second, for governments aiming at decent work and economic growth (as specified in the UN’s Sustainable Development Goal 8), the fact that the more upstream stages of production feature the lowest CSR levels, provides an incentive to strategically focus industrial policy towards specialization in more downstream production stages. Third, policy initiatives like the EU Directive on Corporate Sustainability Due Diligence, the French Duty of Vigilance Law or the German Due Diligence Law shift the responsibility for supplier actions to firms. Our analysis hints at the limited ability of a firm to impose ethical standards on its suppliers and highlights that this problem is particularly severe for suppliers at upstream stages of production. Future due diligence regulation should take this asymmetry along the value chain into account. We test our model’s prediction using the Indian firm-level dataset Prowess provided by the Centre for Monitoring Indian Economy (CMIE), which has been used by Goldberg et al. (2010a,b); De Loecker et al. (2016); Bos and Vannoorenberghe (2018) as well as Barrows and Ollivier (2021), among others. It is a special feature of this dataset that it contains information on the firms’ staffwelfare spending as well as social community spending. We use these items to construct a measure of a firm’s CSR spending in monetary units, which we observe for 15,512 medium-size and large firms between 2000 and 2013 in India. Our approach to computing a firm’s value chain position parallels the procedure in Chor et al. (2021). First, we use the World Input-Output Database (WIOD) and apply the methodology from Antr`as et al. (2012) and Antr`as and Chor (2018) to compute the yearly upstreamness of each Indian industry. Second, to each product in a firm’s portfolio, we assign the value of annual upstreamness of its industry and calculate the firm’s value chain position, its upstreamness, as the sales-weighted average of its products’ upstreamness values. We estimate the relationship between a firm’s value chain position on its CSR expenditure. In our analysis, we control for a large set of firm-level determinants of CSR identified in the literature (e.g., 3Crifo and Forget (2015) survey a literature on CSR in management sciences and industrial organization. They list imperfect competition and contract incompleteness as central drivers of CSR (see, e.g., their Figure 1, p. 114). Schiller (2018) motivates his purely empirical work alluding to the same notion of incomplete contracts: the inability of a firm to fully control CSR investments of its suppliers. Moreover, there is a recent and quickly expanding literature on relational contracts in developing countries surveyed in Macchiavello (2021), which lends support to our modeling on incomplete contracts in value chains involving developing and emerging economies. 79 Newman et al., 2018; G¨org et al., 2018; Schiller, 2018). These include a measure of internationalization (share of exports in total sales), exposure to more demanding customers (measured by the fraction of exports to OECD countries), local embeddedness (share of domestically sourced inputs), firm size (total sales), age and firm ownership (dummies for state owned and foreign owned firms). Importantly, we control for the possibility that wage levels may vary systematically along the value chain using a firmspecific wage measure that combines industry-level information on labor compensation with a firm’s product mix.4 In our estimations, we rely on two different types of variation. First, in a cross-sectional specification, we control for several established firm-level drivers of CSR and industry-year as well as state fixed effects. This approach exploits variation in the value chain position of firms that share the same main industries. Second, in a more demanding specification, we exploit the time dimension of our data. By including firm fixed effects, we identify the effect from changes in a firm’s GVC position over time. In both cases, we find strong support for our model’s prediction: more downstream firms have higher CSR levels. Comparing firms at the 5th and the 95th percentile, our preferred specification indicates that predicted CSR spending is 55.3% higher for the downstream firm compared to the upstream firm. Our results also hold in several robustness checks. In one of them, we address the possible concern that more downstream producers might simply be more visible to final consumers and therefore face higher pressure to invest in CSR. We control for visibility by adding marketing and advertising expenses as an additional variable. We also smooth CSR spending over two years to control for lumpy CSR spending. Our work relates to the recent literature studying the “globalization backlash”, the massive skepticism economic globalization encounters in many countries (Pavcnik, 2017), which is surveyed in Colantone et al. (2022). We are interested in an additional and complementary source of skepticism towards globalization: the ongoing violation of ethical standards in international production - attracting criticism from consumers and actors in civil society. Moreover, while the literature above focuses on public regulation, we are interested in patterns of voluntary self-regulation along international value chains. There are several strands of literature that focus on the (consequences of) the demand for ethical minimum standards in international value chains. In the absence of effective international regulation (Battaglini and Harstad, 2020), this demand manifests itself as pressure on firms to self-regulate international production, a process Baron (2003) terms private politics. In Baron (2012), he argues that key actors of private politics are advocacy NGOs which use both confrontational strategies (like consumer boycotts) as well as cooperative approaches (like labels) to directly influence the actions of firms (instead of affecting government policies in the case of public politics). Our analysis therefore relates to the literature on labels, with the fair trade label being the best documented and most researched case in point.5The other strand of literature is a recent and expanding literature on the role of international advocacy NGO activity for international production and development.6The latter is more focused on NGO campaigns and consumer boycotts. Our object of study, CSR along the value chain, is closer to the former, as it emphasizes the opportunity for firms to increase demand and markups, rather than the risk 4Controlling for firm-specific wages addresses the following concern: If staffwelfare spending is a part of the overall compensation package, downstream firms might offer systematically higher pay due to higher skill intensity, for example. A negative correlation between upstreamness and CSR spending (as measured in part by staffwelfare spending) could then be interpreted as actually reflecting the negative relationship between upstreamness and employee compensation. 5The fair trade label is studied empirically, among others, by Hainmueller et al. (2015), de Janvry et al. (2015), and Dragusanu et al. (2022), and theoretically by St¨ahler and Richardson (2014) and Podhorsky (2015). See Dragusanu et al. (2014) for a survey. For studies of labels more generally, see Fischer and Lyon (2014), Heyes and Martin (2017), as well as Poret (2019). 6See, e.g., Aldashev and Verdier (2009), Aldashev and Verdier (2010), Aldashev et al. (2015), Krautheim and Verdier (2016), Koenig et al. (2021), Fontagn´e and Limardi (2020). 80 of facing a devastating boycott. In fact, the provision of labels by NGOs can be seen as a mechanism to strengthen the positive effect of CSR investments on demand. In our paper, we take this positive effect as a given and study how it shapes CSR investments along the value chain.7 There is a recent and quickly expanding literature on relational contracts in developing countries surveyed in Macchiavello (2021), which supports our notion of the relevance of incomplete contracts in sourcing relationships especially involving developing and emerging economies.8Most informative for our modeling (see Section 5.2.3.1), Cajal-Grossi et al. (2020) show that suppliers in the Bangladeshi garment sector obtain higher prices when selling to relational buyers than when selling to spot buyers. While these papers provide valuable insights into the contracting relationships to overcome contractual incompleteness in an environment with weak legal institutions, they do not analyze the relation to CSR investments. Our theory is rooted in the large literature studying the role of incomplete contracts for the international boundaries of the firm, based on the seminal contributions of Antr`as (2003) and Antr`as and Helpman (2004), surveyed in Antr`as and Yeaple (2014) and Antr`as (2016). We build our modeling on Antr`as and Chor (2013) where a firm’s value chain is modeled as a sequential production process. Like most of this literature, they focus on the integration vs. outsourcing decision along the value chain. Costinot et al. (2013) also model international sourcing as a sequential production process, but without incomplete contracts. They focus on the endogenous fragmentation of the value chain across countries. We abstract from both the international make-or-buy decision and the allocation of tasks across countries. Our main focus, in contrast, is on CSR investments along a sequential value chain characterized by incomplete contracts. Our modeling opens up both the theoretical and the empirical toolkit of the literature on international sourcing with incomplete contracts to the analysis of social and environmental issues in globalized production that feature prominently in the public debate. Closest to our paper in this respect is recent work by Herkenhoffand Krautheim (2022), who analyze a model featuring caring consumers, consumer boycotts and incomplete contracts to study the effects on the international make-or-buy decision in a setting with a single supplier. We, in contrast, study the determinants of CSR investments based on a model that features a continuum of suppliers along a sequential value chain. This allows us to study the effect of a supplier’s value chain position on its CSR investments and therefore the profile of CSR along the value chain. There is an extensive empirical strand of this literature, with Alfaro et al. (2019) being a recent example. These papers directly relate to the theoretical underpinnings, where incomplete contracts lead to underinvestment in a relationship-specific non-contractible variable (inputs, quality-adjusted inputs, etc.). As these non-contractible variables are notoriously difficult to observe in the data, the empirical studies turn to the observable optimal response of the firm to the underinvestment: the integration vs. outsourcing decision. Antr`as and Chor (2013), for example, investigate the role of upstreamness of a supplier for its – empirically unobservable – quality-adjusted investment. To generate testable implications, they therefore turn to the observable optimal response of the firm to the underinvestment: depending on the upstreamness of the supplier, either integration or outsourcing delivers the right incentives minimizing the negative consequences of the underinvestment. Our empirical analysis complements the existing literature as we, in contrast, directly observe the non-contractible variable of interest: CSR spending. This allows us to investigate more directly the role of the GVC position for supplier investments under 7Closest to our model in terms of the type of policy considered is Alfaro-Ure˜na et al. (2021) who develop a quantitative general equilibrium model to study the effects of responsible sourcing in origin countries based on data on affiliates of multinationals in Costa Rica. 8Macchiavello (2010) uses a model with incomplete contracts to study the effect of vertical integration on the pervasive use of subcontracting arrangements among small firms in developing countries. Casaburi and Macchiavello (2019) show how imperfect contract enforcement generates barriers to entry in the Kenyan dairy sector. Machiavello and Morjaria (2021) highlight the importance of long-term relational contracts in the Rwandan coffee sector. Brugu´es (2020) analyzes the role of imperfect contracts in the textile, pharmaceutical and cement sectors in Equador depending on market power. 81 incomplete contracts. There is a large literature on CSR at the intersection of the fields of management science and industrial organization. Kitzmuller and Shimshack (2012), Crifo and Forget (2015), and Pisani et al. (2017) provide surveys of this literature. This includes a large body of empirical studies on CSR which mainly focuses on the effect, rather than the determinants, of CSR. We take a different angle, contributing to a better understanding of the determinants of CSR by analyzing the impact of a firm’s value chain position. Moreover, most of this literature uses indices combining qualitative measures obtained by mandatory CSR reporting of listed firms in developed rather than emerging economies (e.g., Marano and Kostova, 2016; Pisani et al., 2017). We, in contrast, use data from India, a key emerging economy, which not only allows us to construct a continuous monetary measure of CSR, but also to study the behavior of non-listed medium-sized and large firms. Our empirical analysis complements a small number of empirical studies assessing the role of global value chain relationships for CSR. Noteworthy contributions are Schiller (2018) and Newman et al. (2018). Schiller (2018) uses data restricted to large publicly listed companies to show that corporate environmental and social policies of firms are transmitted to suppliers in their value chain and studies the effect of these policies on firm performance. He provides evidence that is consistent with the interpretation that despite incomplete contracts firms can still affect their supplier’s CSR choices to some extent. We, in contrast, are interested in how the limited control of a firm over its suppliers shapes their CSR choices. Most related to our empirical analysis is Newman et al. (2018). They exploit CSR indicators that capture compliance with labor standards, community and management-related CSR practices to analyze the transmission of socially responsible behavior through trade in Vietnamese firms. They find that a firm’s CSR increases with exporting and importing, varying across export destinations. The internationalization of supply chain relationships therefore seems to matter for CSR activities of firms. Moreover, Newman et al. (2018) conjecture that in addition to the internationalization of buyer-seller relationships, the supplier’s position along the value chain may matter and control for it using firm fixed effects. We, in contrast, use a time-varying, firm-specific measure of a firm’s value chain position that allows us to directly measure the GVC position of a firm and to analyze its relationship to the firm’s CSR spending. The remainder of the paper is structured as follows. Section 2 presents our theoretical analysis of CSR along the global value chain. Section 3 outlines our empirical analysis. Section 4 concludes. 5.2 A Model of CSR along the Global Value Chain In this section, we incorporate consumer valuation for ethical production practices and CSR investment into a model of sequential production with incomplete contracts based on Antr`as and Chor (2013). We assume that each firm in the Global North has a value chain with unit measure of required inputs, which are all provided by independent suppliers in the Global South, where environmental and labor regulation is insufficient from the perspective of caring consumers in the Global North. We derive the industry equilibrium of the model and show that our theory implies an increasing CSR profile along the value chain – a prediction which we take to the data in Section 3. 5.2.1 Incomplete Contractibility of CSR Before turning to the outline of our model, we first plead the case for the relevance (and the type) of incomplete contracts for CSR along global value chains. Assuring sufficient CSR levels by suppliers is notoriously difficult to achieve. Nike’s response to the “sweatshop campaign” in the 1990s is probably the best documented and most researched case in point. Protests against work conditions in Indonesian supplier factories led Nike to implement compulsory, legally binding codes of conduct for all its sup82 pliers.9Harrison and Scorse (2010) show that the campaigns caused an increase in the real wages in Indonesian factories manufacturing for Nike. Locke (2013a), however, shows that more than a decade after the introduction of legally binding codes of conduct combined with an auditing system, substantial compliance problems persisted in Nike’s supply chain. He analyzes data from factory audits of working conditions in more than 900 of Nike’s suppliers located across fifty countries and finds that, despite the fact that all suppliers are obliged to sign the codes of conduct and despite large investments into monitoring, there still is a substantial number of suppliers that are ranked as “noncompliant”. In Asia, the noncompliant suppliers even constitute the majority. While Nike is a particularly well-documented case in point, these problems are by no means specific to Nike or the footwear and apparel sector. Based on data from one of the world’s largest supply chain auditing firms containing 16,795 audits of 5,819 factories in 13 industries across 66 countries over a period of six years, Short, Toffel and Hugill (2016, 2020) document widespread violations of codes of conduct in areas like child labor, forced labor, working hours, occupational health and safety, the minimum wage and disciplinary practices. In our view, this calls for a model of CSR along GVCs that explicitly accounts for incomplete contracts between a firm and its suppliers. One standard way of motivating contractual incompleteness is that it is prohibitively costly - if possible at all - to cover all contingencies in a contract. This does not seem to be a major concern in our context: codes of conduct and similar contracts can specify minimum social, labor and environmental standards. In the case of CSR investments, the contractual incompleteness stems from another standard argument: the actions of the supplier are “observable but not verifiable” (e.g., Hart and Moore, 1999). This means that it is costless (or at least relatively cheap) to determine what is going on in a supplier’s factory but is prohibitively costly to hold the supplier accountable for the induced losses in a court of law. There are quite general and well-known problems with verifying labor and environmental standards implemented in production plants, especially so in low-regulation countries. Safety equipment may be distributed one day, but not the other, fire exits may be accessible one day, but blocked the other, workers may fear the loss of their jobs, if they report forced and excessive overtime, toxic waste may be disposed into a river when no-one is watching.10 It is nevertheless relatively cheap for, say, an advocacy NGO to talk to workers, measure toxins in waste water or send an undercover agent into a factory to find out about the actual production conditions. Such circumstantial evidence, possibly obtained by breaking domestic laws, is insufficient in a court of law. However, in many cases it is sufficient to convince caring consumers to re-assess their valuation of the product and to reduce demand.11 Producing the type of evidence that would lead to a conviction of the supplier - though in principle possible - will in many cases be prohibitively costly to produce. It would require a massive and continuous presence of auditors on site and even then, there is no guarantee that they will win the cat-and-mouse game with local management. This highlights that the notion in the incomplete contracts literature that the action is “observable but not verifiable” (e.g. Hart and Moore, 1999) applies to CSR investments along GVCs.12 9In May 1998, Nike CEO Phil Knight famously made the following statement: “The Nike product has become synonymous with slave wages, forced overtime, and arbitrary abuse. I truly believe the American consumer doesn’t want to buy products made under abusive conditions.” (New York Times, 1998). 10For a vivid description of the typical problems of auditors in the field, see Locke (2013b), p. 35-37. 11This is all the more so as NGOs are repeatedly reported to be the most trusted institutions among consumers before governments and the media. See, for example, the Edelman Trust Barometer for 2020, at https://cdn2.hubspot.net/hubfs/440941/Trust%20Barometer%202020/2020%20Edelman% 20Trust%20Barometer%20Global%20Report-1.pdf. 12In our modeling, we assume for simplicity that the firm and consumers can observe the actual level of CSR by the supplier - but that it is not verifiable by a court. In case that some readers take issue with the notion that the actual CSR level is observable but non-verifiable, note that this assumption can be relaxed - further broadening the gap between the cost of observing and the cost of verifying. For the mechanics of our model to function, we do not need the firm and consumers to observe the actual CSR investments. The firm only needs to know the CSR level which consumers observe or perceive (however misguided their perception may be) and which they use to form their opinion on the ethical quality, ε(ω). This observable/perceived CSR level is what 83 previous CSR levels. Under the maintained assumption discussed in Section 5.2.2.1 that α∈(0,ρ), CSR investment at stage mis increasing in previous CSR levels. This allows us to state the following lemma: Lemma 1 With α∈(0,ρ), CSR investments along the value chain are sequential complements, i.e., higher levels in previous stages lead to a higher optimal level of CSR in stage m. Proof: In the text. When the incentives for CSR investment of suppliers depend on previous investments by other suppliers, the position in the value chain relative to other suppliers can become a crucial determinant of CSR investment. We therefore now turn to the sequence of CSR investments along the complete value chain. Combining equation (5.11) with equation (5.9), we obtain r�(m) = 1 α�1−β cs�α 1−α�ρAα(1−ρ) ρq(ω)α�1 1−αr(m)ρ−α ρ(1−α).(5.12) Using the initial condition r(0) = 0, solving this differential equation delivers r(m) = �1−ρ 1−α�ρ(1−α) α(1−ρ)�1−β cs ρ�ρ 1−ρ A q(ω)ρ 1−ρmρ(1−α) α(1−ρ).(5.13) We can now plug this into equation (5.11), which delivers CSR investments at stage mas s(m) = �1−ρ 1−α�ρ−α α(1−ρ)�1−β cs ρ�1 1−ρ A q(ω)ρ 1−ρmρ−α α(1−ρ).(5.14) This expression for CSR investments at stage maccounts for CSR investments in all previous stages, which affect the optimal choice at stage maccording to Lemma 1. Equation (5.14) shows that the standard implemented at stage mdepends on the size of the order q(ω) placed by the firm, which we determine next.27 Total physical output q(ω) is chosen by the firm in the first stage of the game. The firm knows that it will either have to purchase inputs on the input market or has to compensate its suppliers for their production costs. In either case, costs are given by �1 0cxq(ω)dj =cxq(ω). The problem of the firm is therefore given by: maxq(ω)πF(ω) = βr(ω)−cxq(ω)−f. (5.15) Based on equation (5.13) evaluated at m= 1, we can write total revenue as r(ω) = �1−ρ 1−α�ρ(1−α) α(1−ρ)�1−β cs ρ�ρ 1−ρ Aq(ω)ρ 1−ρ.(5.16) Differentiating πF(ω) with respect to q(ω) and setting the derivative equal to zero gives the optimal quantity q(ω) as q(ω) = �1−ρ 1−α�ρ(1−α) α(1−2ρ)�1−β cs ρ�ρ 1−2ρ�ρ 1−ρ β cx A�1−ρ 1−2ρ .(5.17) 27Note that the fact that s(m) increases in the size of the order q(ω) placed by the firm implies that the firm has an incentive to order the entire quantity of input mfrom one single supplier, as this maximizes the level of the standard under which production takes place and therefore also maximizes the level of perceived ethical quality of the final product. 90 We can combine equations (5.14) and (5.17) and use the definition of A=EP ρ 1−ρto get the optimal CSR investment at stage m, s(m) = �1−ρ 1−α�ρ−α(1−ρ) α(1−2ρ)�1−β cs ρE�1−ρ 1−2ρ�ρ 1−ρ β cx P�ρ 1−2ρ mρ−α α(1−ρ).(5.18) We can now state the following lemma: Lemma 2 Equation (5.18) characterizes how suppliers along a firm’s value chain optimally choose their CSR levels taking aggregate variables as given (partial equilibrium). The production standard is higher in more downstream stages, i.e., suppliers that are closer to the final consumer invest more in CSR. With the price index Ptaken as given, CSR investments also increase in the total expenditure on ethically differentiated goods E. Proof: This follows directly from equation (5.18) and ρ<1/2. Lemma 2 shows that the supplier’s CSR investment increases in m, i.e., CSR is higher in more downstream stages. The result reflects the fact that CSR investments are sequential complements along the value chain, as pointed out in Lemma 1. In addition, the overall level of CSR investment across all stages is increasing in aggregate expenditure on ethically differentiated goods. We will see below that Edisappears from the equation when we solve for the industry equilibrium of our model. There, a supplier’s position in the value chain will emerge as the key determinant of its CSR expenditures in our model. 5.2.5 Equilibrium CSR We now derive the industry equilibrium of the model with free entry of final goods producers. This allows us to analyze the equilibrium determinants of CSR investments at the different stages of production. 5.2.5.1 Industry Equilibrium To solve for the industry equilibrium, we assume free entry and final good producers with homogeneous levels of productivity normalized to 1 so that it must hold that πF(ω)=0∀ω. Setting equation (5.15) equal to zero, plugging in the optimal q(ω) from equation (5.17), and using the fact that A=EP ρ 1−ρgives an expression for the price index Pas P=�1−α 1−ρ�ρ−α αρ �1−α βE�1−ρ ρcscx ρ2(1 −β)�f 1−2ρ�1−2ρ ρ .(5.19) Next, we solve for the equilibrium number of firms using the optimal pricing rule and the definition of the ideal price index from demand.28 From this, we get a second expression for the price index as a function of the number of firms nas P=n−1−ρ ρ cxcs ρ2(1 −β) 1−2ρ f�1−α 1−ρ�1−α α .(5.20) Combining the two results for the price index, one from free entry, equation (5.19), and one from demand, equation (5.20), gives the equilibrium number of firms as n=βE 1−ρ 1−2ρ f.(5.21) 28Details of the derivations are provided in Appendix 5.5.1. 91 Combining equation (5.19) with equation (5.17) on the one hand, and the definition of A=EP ρ 1−ρ on the other hand, yields expressions for order size q(ω) and the market size parameter Ain industry equilibrium as q(ω) = ρ cx f 1−2ρ(5.22) and A=�1−α 1−ρ�ρ−α α(1−ρ)1−α β�cscx ρ2(1 −β)�ρ 1−ρ�f 1−2ρ�1−2ρ 1−ρ .(5.23) This allows us to state the following lemma: Lemma 3 In industry equilibrium with free entry in the final goods sector, final goods producer size q(ω) and the equilibrium market size parameter Aare both independent of exogenous aggregate expenditure E. Proof: Follows directly from equations (5.22) and (5.23). Market and firm size are both independent of the aggregate level of expenditure Ein the economy. The reason is that in our setup of free entry with homogeneous final good producers, any increase in aggregate spending will be fully compensated by additional entry into the final goods sector. 5.2.5.2 Equilibrium CSR along the Global Value Chain Combining equations (5.18) and (5.19), the CSR level implemented at stage mcan be written as s(m) = f 1−2ρ (1 −α)ρ cs 1−β βmρ−α α(1−ρ).(5.24) It only depends on parameters and the production stage m. We can now state the following proposition: Proposition 1 – Equilibrium CSR along the Global Value Chain. With CSR investments across production stages being sequential complements, CSR expenditures s(m)in industry equilibrium are increasing along the value chain from upstream to downstream stages. Proof: This follows directly from equation (5.24). This constitutes the main theoretical result of our paper: equilibrium CSR levels increase along the value chain as suppliers get closer to the final consumer. The result is driven by the fact that investments in CSR are sequential complements, as established in Lemma 1: the larger CSR investments at a given stage m, the higher the incentive for more downstream suppliers to invest more. Comparison of the partial equilibrium expression (5.18) and the equilibrium expression in (5.24) shows that the increasing profile, our central finding, persists in industry equilibrium. The effect of total expenditure Eestablished in Lemma 2, however, does not survive endogenizing the price index P. We argue below that we use this difference to draw the line between the core testable implication of our model, which holds in industry equilibrium, and variables we merely control for in the empirical analysis, as they play a role in the partial equilibrium analysis only. Comparison of Proposition 1 to the outcome under complete contracts (see Appendix 5.5.2) shows that the increasing profile of CSR results from the setting of incomplete contracts. We argued above that CSR investments are in many cases observable but non-verifiable (Hart and Moore, 1999). Our model therefore hints at a general problem that may well be a typical feature of CSR along global value chains: when CSR investments are sequential complements (the case for which we argued in Section 5.2.2.1), the early stages of the production process are the ones that are crucial for the choices along the entire 92 value chain; at the same time, these are the stages with the lowest CSR investments. This implies that for both headquarters and policy makers who seek to increase overall CSR levels along the value chain, the upstream stages of production are the most important ones to target. 5.2.6 Discussion and Link to the Data Before we turn to the empirical analysis, we pause to first discuss how we link our theoretical results to the data. We then place our analysis in the context of the literature on incomplete contracts in international economics and explain why our theory-informed empirical analysis closes a gap in this literature. 5.2.6.1 Testable Implications: Partial vs. Industry Equilibrium As pointed out above, comparison of the partial equilibrium expression for CSR (5.18) and its industry equilibrium counterpart (5.24) shows that the central prediction of our theory – the increasing profile of CSR along the value chain – also holds in industry equilibrium. We bring this prediction to the data in Section 5.3, where we also provide a detailed description of the data. At this stage, simply note that we use Indian firm-level data together with input-output tables to compute a firm-level measure of upstreamness. Moreover, we observe CSR investments at the firm level. This allows us to investigate the negative relationship between upstreamness and CSR investments established in Proposition 1. If we had ended the theoretical analysis with the partial equilibrium expression for CSR investments in equation (5.18), we could have deduced additional testable implications from the theory. Specifically, taking the price index Pas given, Lemma 2 states that the CSR investments by the supplier at stage m of the production process increase in total expenditures on ethically differentiated goods, E. One could test this implication from partial equilibrium empirically by taking, e.g., the share of sales going to OECD countries as a proxy for being exposed to stronger demand for ethical quality. Proposition 1, however, shows that when the equilibrium price index is accounted for, the effect of E vanishes. This is an implication of the simple Krugman (1980) structure of our modeling of the ethically differentiated homogeneous final good producers, which leads them to having a constant equilibrium size. Adjustments to changes in aggregate variables are exclusively driven by adjustments in the number of firms (see equations (5.21) and (5.22)). As the impact of demand for ethical products does not survive the computation of the industry equilibrium, we do not consider it a core testable implication of our model. The partial equilibrium analysis nevertheless points to its relevance, and we therefore include in our regressions a proxy for OECD export exposure as a control variable. 5.2.6.2 A Direct Measure of the Non-Contractible Variable As indicated above, our data allow us to take a very direct approach to linking our model to the data. In our opinion, this approach fills a gap in the empirical literature on the role of incomplete contracts in international economics. The literature is surveyed in Antr`as and Yeaple (2014) and Antr`as (2016), with Alfaro et al. (2019) being a recent example. The empirical studies are grounded in theoretical models, where incomplete contracts lead to underinvestment in a relationship-specific non-contractible variable (typically quality-adjusted inputs). In our model, this non-contractible variable is CSR investment, which has some similarities with the quality dimension in the mentioned literature. It goes without saying that such non-contractible variables are difficult to measure. The literature therefore typically resorts to the observable organization of production as a dependent variable, which – according to the theory – is driven by the unobservable non-contractible variable. Nunn (2007) sheds empirical light on this theoretical mechanism showing that relationship-specificity shapes the patterns of international trade. Based on Nunn (2007), Nunn and Trefler (2008) construct a country and industry specific measure 93 of contractual completeness. They find that improved contracting of the inputs provided by the supplier favors integration over outsourcing. Our analysis adds an additional and complementary dimension: To our knowledge, ours is the first paper to use a direct measure of the non-contractible variable. This becomes possible as we shift the attention from some technical, but by construction externally unobservable, product or service characteristic to CSR investments, for which we do have a firm-level measure in our data. We can therefore directly study the effect – in our case of a supplier’s value chain position – on the non-contractible variable without recurring to the organization of production as a substitute observable outcome for the unobservable (under)investment decision.29 5.3 GVC Position and CSR in India In this section, we take the model to the data by testing whether CSR spending is larger when firms are positioned more downstream in global value chains. We begin with the description of the data. 5.3.1 Data Description For our analysis, we combine the World Input-Output Database (WIOD, Timmer et al., 2015) with the Prowess database.30 Our sample covers the years 2000 to 2013. The Prowess database contains a wide range of Indian firm-level information. Its coverage of the Indian economy is comprehensive, with total production of all companies in Prowess accounting for more than 80% of India’s GDP (Bos and Vannoorenberghe, 2018). The data are collected by the Centre for Monitoring Indian Economy (CMIE), which mainly sources information from annual reports of firms, stock exchanges and regulators. India is particularly well-suited to bring the theoretical predictions of our model to the data.31 In our theory, CSR is any action by a supplier that leads to an increase in the perceived ethical quality of the final product. This can be any action that makes the supplier appear more ethical in the eyes of consumers. We therefore use details on firms’ spending on staffwelfare and expenses for the benefit of society or the community in general as our measure of a firm’s CSR activities, our dependent variable. It is a unique feature of the data, and especially rare in an emerging economy context, that we can observe this information at the firm level. We describe the variable in more detail in Section 5.3.2.1. Moreover, the data include detailed information on general firm characteristics such as annual sales, export activity, ownership, input sourcing, age and the product mix. The latter is key for our empirical strategy.32 We exploit differences in the product mix across firms as well as within firms over time to construct a firm-specific measure of the global value chain position. To do so, we combine the Prowess data with industry-level information from WIOD. In a first step, we follow Antr`as et al. (2012) and Antr`as and Chor (2018) and use WIOD to construct a time-varying measure of Indian industries’ upstreamness. 29We do acknowledge that the analysis of the determinants of the integration vs. outsourcing decision of the firm is an important endeavor in its own right. Here, we simply want to highlight that our approach allows us to investigate the non-contractible variable directly and thereby closes a gap in the previous literature. 30We use the Prowess vintage of December 2017. Data were downloaded from https://prowessdx.cmie. com on February 1, 2018. The WIOD release 2016 was downloaded from http://wiod.org/home. Data were downloaded on July 22, 2019. 31First, the reduction in trade barriers following India’s accession to the WTO in 1995 substantially increased India’s integration into GVCs. Second, compared to most developed countries, firms in India are much less constrained by labor and environmental standards leaving ample room for voluntary CSR investments. Finally, related to our assumption of incomplete contracts, the congested court system (Boehm and Oberfield, 2020) amplifies the problem that contracts between the supplier and the headquarter are observable but not verifiable (Hart and Moore, 1999). 32Goldberg et al. (2010b) emphasize that it is a rare feature of the Prowess database that it captures annual changes in the product mix at the firm level. This feature stems from the Indian 1956 Companies Act requiring firms to disclose product-level information on capacities, production and sales in their annual reports. Several previous studies have used this information, e.g., Goldberg et al. (2010a), Goldberg et al. (2010b), De Loecker et al. (2016), Barrows and Ollivier (2018), Bos and Vannoorenberghe (2018) and Barrows and Ollivier (2021). 94 Linking WIOD industries to the firms’ products from Prowess, we use the product-level sales shares as weights in the calculation of an average, firm-specific upstreamness, our measure of a firm’s value chain position. We describe the calculation of our main independent variable in more detail in Section 5.3.2.2. Notably, multi-product firms account for 47% of Indian manufacturing firms and 80% of manufacturing output (Goldberg et al., 2010b, p.1043). Therefore, a key feature of our approach is that even for firms within the same main industry, measured GVC positions may differ across firms, either because of differences in the product mix or because of differences in the sales weights of identical product mixes. Our measure also varies over time and within firms if a firm’s product sales weights change or if it adds or drops products. In addition, a firm’s value chain position may change over a longer horizon if the position of its products changes. Section 135 of the 2013 Indian Companies Act obliges firms above a certain turnover and profitability threshold to spend 2% of their average net profits on CSR. Compliance with the law changes the CSR decision of a firm and might also change its reporting behavior.33 To avoid such a discontinuity in their CSR spending, we limit the analysis to the period between 2000 and 2013, although our data are available up to 2014.34 In the next section, we describe our key variables of interest, CSR spending and a firm’s GVC position, in more detail. 5.3.2 Measuring CSR Spending along GVCs A complex set of motives can drive a firm’s strategy on CSR spending (B´enabou and Tirole, 2010; Crifo and Forget, 2015). Crifo and Forget (2015) describe a firm’s CSR decision as a response to market imperfections with the aim of satisfying social preferences of at least one stakeholder. In the context of GVCs, satisfying these social preferences becomes even more complex as firms face not just domestic but also foreign stakeholder demands (Newman et al., 2018). Consumers in developed countries increasingly demand that minimum standards for working conditions and environmental aspects of production are observed also at production sites abroad. Before we focus on the empirical specification, we explain our measure of CSR spending. 5.3.2.1 Measuring a Firm’s CSR Spending There is no general agreement about the definition of CSR nor a common way of quantifying CSR spending (Crifo and Forget, 2015; Newman et al., 2020). Shirodkar et al. (2018) claim that in the context of developing and emerging countries, CSR primarily refers to the commitment to behave ethically and to contribute to economic development. McWilliams and Siegel (2001, p.117) define CSR more strictly as “actions that appear to further some social good, beyond the interests of the firm and that which is required by law”. In this paper, we follow this concept and a large literature by defining our dependent variable as social spending beyond what is required by law (B´enabou and Tirole, 2010; Crifo and Forget, 2015; Newman et al., 2018, 2020). We argue that this approach is especially well-suited in the context of GVC integration of firms originating from developing or emerging countries because these firms are operating in a relatively lax social and environmental regulatory setting. Facing lenient domestic regulation, firms respond to market failure with higher CSR spending to satisfy stakeholder demands (B´enabou and Tirole, 2010; Crifo and Forget, 2015). In the context of GVC integration, the CSR decision of firms located in developing countries is determined by relatively higher demand for CSR by foreign stakeholders (Newman et al., 2018). 33See Dharmapala and Khanna (2018) for more details on the legislation and an evaluation of the effects on firm’s CSR activities. 34To control for potential anticipation effects, we carry out a robustness test (available upon request) where we exclude the years 2011 to 2013. We find that our results are unaffected. 95 Our data allow us to directly observe monetary CSR spending on own staffwelfare and expenses for the benefit of the society or the community. We use this information to measure what we call production standard or CSR investment in the theoretical model. Staffwelfare spending refers to various amenities that are made available to the employees for their general welfare and go beyond regular compensation in the form of salaries. Staffwelfare spending includes free or subsidized medical treatment, transportation facilities, recreation facilities, stafffood, and canteen expenses. These aspects all cover labor welfare, which is the predominant component of CSR spending in emerging economies (Newman et al., 2018). Society or community spending refers to expenses on building or maintaining public parks, garden maintenance, building temples, constructing roads or contributing to social occasions, etc. These community-related expenses closely refer to the catalog of community-related CSR activities applied by Newman et al. (2018) and Newman et al. (2020).35 Our ability to observe the money amount of CSR spending allows us to investigate the continuous relationship between CSR spending and a firm’s position in GVCs. Accordingly, we can directly exploit variation in the level of CSR spending across firms and within a firm over time. We argue that our focus on observable data of the monetary expenses related to CSR, but not labeled CSR, substantially reduces a possible bias due to self-reporting of firms as discussed in Newman et al. (2020). 5.3.2.2 Measuring a Firm’s Value Chain Position We now turn to the key explanatory variable of interest: a firm’s GVC position. Our model predicts that a firm’s CSR spending depends crucially on its position in the value chain. Without recurring to a theoretical model, Newman et al. (2018) control for the role of the value chain position by using a time-invariant firm fixed effect. Our data allow us to not only control for a possible impact of the value chain position of CSR but to estimate the effect. To do so, we calculate a time-varying and firm-specific measure of a firm’s position in GVCs based on a combination of value chain information of industries from the World Input-Output Database and the Prowess information on product-level sales of firms. In a first step, we calculate Indian industries’ position in GVCs based on contributions by Antr`as et al. (2012) and Antr`as and Chor (2018). They show how World Input-Output Tables can be used to calculate an industry’s upstreamness, measuring the distance of an industry’s sales from final consumption. In line with this approach, we start by computing upstreamness Uv kof industry vin country k(India) by computing Uv k=1Fv k Yv k + 2�W w=1 �L l=1 avw kl Fw l Yv k + 3�W w=1 �L l=1 �W x=1 �L m=1 avw kl awx lm Fx m Yv k +..., (5.25) where Fv k/Y v kis the share of global final consumption in industry v’s gross output in country k. The following terms capture the proximity to final consumption of industry vin country kthrough its sales in intermediates. For instance, the second summand includes the share of output from industry vin country k, which is consumed by industry win country land sold to final consumers, multiplied by avw kl , which measures the dollar amount of industry v’s output in country kneeded to produce one dollar worth of industry w’s output in country l. The upstreamness of an industry is the weighted average of the number of stages that separate the typical unit of output in industry vfrom final consumers. The weighting scheme gives higher weight to 35Another important CSR component is spending related to environmental issues. Although we observe expenses on environmental causes in our data, we do not include this dimension in our main specification. The reason is that we cannot be sure that these expenses are truly voluntary. This variable could also capture, e.g., compulsory investment in environmentally friendly technologies required by law. The results for the extended CSR measure including environmental expenses stay robust and are available upon request. 96 the more indirect sales. If industry v’s output in country kis completely sold to final consumers, Uv k equals unity. If a part of the output is sold as intermediates, Uv kis above one and a higher value of Uv kassociates with higher upstreamness. Accordingly, larger values of Uv kare associated with increased distance to final consumption. We use the WIOD to apply this methodology to the Indian economy. We exclude non-tradable industries in the construction of our key dependent variable.36 Table 5.1 shows summary statistics of the derived industry upstreamness measure for 34 tradable 2-digit industries based on the ISIC-4 classification. In our period, the average upstreamness of Indian industries varies between 1.22 (water Table 5.1: Industry upstreamness measure for India ISIC-4 Industry mean sd min max Mining and quarrying 3.80 0.18 3.52 4.08 Manufacture of chemicals and chemical products 3.33 0.06 3.21 3.40 Manufacture of coke and refined petroleum products 2.79 0.20 2.52 3.05 Manufacture of basic metals 2.71 0.07 2.60 2.80 Manufacture of paper and paper products 2.63 0.08 2.48 2.74 Manufacture of other non-metallic mineral products 2.34 0.03 2.29 2.42 Manufacture of rubber and plastic products 2.33 0.03 2.28 2.40 Manufacture of fabricated metal products, except machinery and equipment 2.27 0.04 2.20 2.32 Telecommunications 2.24 0.10 2.08 2.42 Manufacture of wood and of products of wood and cork, except furniture; manufacture of articles of straw and plaiting materials 2.17 0.06 2.05 2.24 Manufacture of computer, electronic and optical products 2.14 0.20 1.93 2.58 Wholesale and retail trade and repair of motor vehicles and motorcycles 2.10 0.06 2.02 2.20 Warehousing and support activities for transportation 2.07 0.07 1.96 2.18 Retail trade, except of motor vehicles and motorcycles 2.07 0.05 1.99 2.15 Wholesale trade, except of motor vehicles and motorcycles 2.07 0.05 1.99 2.15 Manufacture of electrical equipment 2.00 0.10 1.88 2.17 Printing and reproduction of recorded media 2.00 0.07 1.88 2.12 Legal and accounting activities; activities of head offices; management consultancy activities 2.00 0.07 1.86 2.08 Land transport and transport via pipelines 1.95 0.05 1.87 2.03 Manufacture of machinery and equipment n.e.c. 1.94 0.07 1.83 2.05 Manufacture of furniture; other manufacturing 1.90 0.10 1.77 2.08 Forestry and logging 1.88 0.05 1.81 1.97 Manufacture of other transport equipment 1.84 0.21 1.58 2.19 Crop and animal production, hunting and related service activities 1.78 0.06 1.65 1.83 Architectural and engineering activities; technical testing and analysis 1.71 0.15 1.51 1.97 Other service activities 1.66 0.07 1.52 1.80 Manufacture of textiles, wearing apparel and leather products 1.63 0.05 1.55 1.71 Manufacture of motor vehicles, trailers and semi-trailers 1.61 0.09 1.49 1.75 Manufacture of food products, beverages and tobacco products 1.54 0.06 1.42 1.61 Manufacture of basic pharmaceutical products and pharmaceutical preparations 1.51 0.03 1.47 1.56 Air transport 1.51 0.14 1.26 1.71 Computer programming, consultancy and related activities; information service activities 1.46 0.09 1.29 1.59 Construction 1.34 0.04 1.28 1.44 Water transport 1.22 0.05 1.17 1.37 Across all industries 2.05 0.54 1.17 4.08 Source: Own calculation based Antr`as and Chor (2018) and WIOD data. Note: Only tradable industries are considered. transport) and 3.80 (mining and quarrying). Across all industries, the average upstreamness is 2.05. The maximum value for upstreamness (4.08) is measured in the mining and quarrying industry in 2008 and the minimum (1.17) is measured in the water transport industry in 2012. Looking at manufacturing industries only, the most downstream manufacturing industry is manufacture of basic pharmaceutical products and pharmaceutical preparations (1.51), while manufacture of chemicals and chemical products (3.33) is the most upstream one. Next, we combine the industry-level upstreamness measure with product-level sales information to obtain a measure of upstreamness at the firm-level. For each firm-year observation, we use the productlevel sales shares as weights to calculate the weighted average of upstreamness of the firms’ product portfolio as illustrated in equation (5.26), so that Uft = V � v=1 salesfvt salesft Uvt,(5.26) 36We base the classification of industries as non-tradable on the export pattern revealed in WIOD. To avoid that an industry with negligibly small exports is defined as tradable in one year and as non-tradable in the next year, we define a threshold, above which an industry is classified as tradable. The threshold is defined as the lowest amount that avoids any switching between being classified as tradable or non-tradable during our covered period. The exclusion of non-tradable industries affects 20 industries in WIOD, e.g., public administration and defence; compulsory social security,education and accommodation and food service activities. 97 where Uft is the level of upstreamness of firm fat time tselling products from Vdifferent industries.37 The measure accounts for the product mix by weighting a product’s GVC position Uvt with its share of total sales in year t.38 Our measure of a firm’s GVC position Uft has three sources of variation. First, the measure is affected by changes in a firm’s product portfolio, i.e., if a firm drops or adds new products. Second, for a given product mix, the relative importance of the different products may change. Third, the GVC position of the different products in a firm’s portfolio changes over time (e.g., because some of the firm’s products are increasingly used in more downstream industries). Figure 5.1 illustrates these sources of variation for a firm in the electrical equipment industry. This firm serves as an example of an average firm in our sample in terms of CSR spending, upstreamness and number of products. It has several Figure 5.1: A firm’s GVC position change over time 1 1.5 2 2.5 3 3.5 4 Industry Upstremaness 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Year product sales firm-specific upstreamness Notes: Average firm in terms of upstreamness, number of products and CSR spending was chosen for illustrative purposes. Firm from electrical equipment industry. Example: A firm's change in GVC position over time products that belong to industries that differ in terms of the upstreamness position. Over time, the firm’s product portfolio changes as a new product is added and one is dropped and the importance of products measured by their sales share changes as well. The firm’s upstreamness position (black dashed line) varies over time in line with the changes in the product portfolio. 5.3.3 Empirical Specification We start the discussion of the empirical methodology with a description of the main empirical setup, including the presentation of other independent variables, which may drive CSR decisions of firms. After discussing descriptive statistics, we present the empirical results. 37Our approach is comparable to the approach by Chor et al. (2021), who also combine an industry GVC measure with a firm’s product information. They use firms’ exports (imports) in an industry in total exports (imports) as weights to derive a firm-specific GVC position for Chinese firms. Similar to our derived GVC measure, the respective GVC measure varies over time within a firm. 38For some firms, there is no product-level information available. In these cases, we take the upstreamness of the main industry of the firm as the measure of the GVC position. This affects 8.49% of the observations, which enter our empirical estimation. 98 5.3.3.1 Empirical Set-Up Proposition 1 predicts that CSR expenditures increase along the value chain from upstream to downstream stages. To empirically test this prediction, we estimate the following two equations. Yfvt =α0+β1Uft−1+β2Xft−1+γvt +δs+�fvt,(5.27) Yfvt =α0+β1Uft−1+β2Xft−1+ζf+τt+�fvt,(5.28) where Yfvt measures CSR spending of firm factive in main industry vin year t. In our main specifications, we apply two different estimators, OLS and Poisson Pseudo-Maximum Likelihood (PPML). To apply OLS, we take the natural logarithm of a firm’s CSR spending (log(CSR spendingft)). This removes firms with zero CSR spending from our sample. Our alternative specification applying PPML with CSR spendingfvt as the dependent variable allows us to keep zeros in the data.39 Our main explanatory variable of interest is a firm’s GVC position at time t−1, Uft−1. In all regressions, we expect β1<0, which indicates lower CSR spending for more upstream firms. In our choice of fixed effects, we use two different approaches to analyze the relation between CSR spending and upstreamness. First, equation (5.27) is a repeated cross-section in which we include state δsand industry-year γvt fixed effects. We include state fixed effects δsto capture differences between Indian states which may affect the level of CSR spending in all firms alike, for example, differences in labor regulation. The industry-year fixed effects γvt capture shocks common to all firms in a particular industry and year. Some industries may be under tighter public scrutiny in general (e.g., tobacco), or there might be industrial disasters that affect certain industries in specific years. Accordingly, equation (5.27) exploits variation between firms within the same state, main industry, and year. Second, equation (5.28) exploits time-series variation within firms. We only include firm fixed effects ζfand year fixed effects τt.ζfcaptures unobservable firm characteristics, which are time-invariant. τt captures year-specific shocks to CSR spending common to all firms. In this specification, we thus exploit within-firm variation over time. In our data, we observe a firm 6.8 years on average. Xft−1includes firm-level control variables to capture alternative drivers of CSR spending. Similar to Uft−1, the control variables enter with a one-year time lag. We next present these control variables in more detail. The variable log(Wages) is a proxy for firm-level wage payments to control for the possibility that wages vary systematically with the value chain position. One could consider staffwelfare spending a part of the overall compensation package for employees. If downstream firms offer better compensation packages (e.g., due to higher skill intensity), a negative relationship between upstreamness and CSR spending (partly measured by staffwelfare spending) could then be interpreted as actually reflecting the negative relationship between upstreamness and employee compensation. We control for this by taking the hourly compensation paid in Indian industries from WIOD and using the product-level information of firms to generate a firm-level control variable that is similar in spirit to our firm-level upstreamness measure.40 In particular, we calculate wages as an industry’s average hourly compensation to employees 39PPML is widely used in the literature of international economics and does not require taking the natural logarithm of the dependent variable. Thus, using PPML also allows us to explore within firm variation over time for those firms that start spending on CSR during our analyzed period. Moreover, it outperforms its alternatives in the presence of many zeros and heteroskedasticity (Silva and Tenreyro, 2006). 40The wage bill is available for a subset of firms in Prowess. However, since there are many zeros and because the provision of the number of employees is not compulsory for firms, we decide not to include the wage bill and prefer to use our combined measure. 99 organization of production to the analysis of CSR and ethical production in global value chains involving developing and emerging economies. In this context, future research should further investigate the relevance of incomplete contracts for CSR in global production. To this end, new theories could be developed to neatly link existing or novel empirical measures of contractual incompleteness to firm’s CSR investments. 106 5.5 Theory Appendix 5.5.1 Industry Equilibrium with Incomplete Contracts We use equation (5.19) to calculate the values of q(ω) and s(m) in industry equilibrium from equations (5.17) and (5.14). q(ω) = ρ cx f 1−2ρ(5.29) s(m) = 1−α β (1 −β)ρ cs f 1−2ρmρ−α α(1−ρ)(5.30) Using the definition of ε(ω) we then get ε(ω) = ��1 0 s(m)αdm� 1 α =1−α β (1 −β)ρ cs f 1−2ρ�1−ρ 1−α�1 α .(5.31) Plugging this, the solution for q(ω), as well as the expression for Pinto inverse demand gives the pricing rule as p(ω)=A1−ρq(ω)−(1−ρ)ε(ω)ρ=cx ρ 1−ρ β.(5.32) Using ε(ω) and p(ω) in the definition of the price index from demand, P−ρ 1−ρ=�n 0�ε(ω) p(ω)�ρ 1−ρdω, gives equation (5.20), a second expression for the price index as a function of the number of firms, n. Combining the two expressions gives the equilibrium number of firms from equation (5.21), n=βE 1−ρ 1−2ρ f.(5.33) 5.5.2 Complete Contracts In this section, we solve the model for the case of complete contracts. Under complete contracts, the headquarter can offer fully specified contracts to each supplier jand will approach them in the right order. The headquarter offers a contract consisting of a physical quantity x(j) and an according payment w(j), as well as an implemented production standard s(j) with the according payment t(j). To make comparison with the baseline model feasible, we denote all endogenous variables in the complete contracts case with a tilde. The headquarter then maximizes its profits with respect to the four choice variables above, subject to two conditions: max {˜x(j),w(j),˜s(j),t(j)}j∈[0,1] ˜πF(ω) = ˜q(ω)ρ˜ A1−ρ��1 0 ˜s(j)αdj� ρ α −�1 0 w(j)dj −�1 0 t(j)dj −f s.t. w(j)−cx˜x(j)≥0 t(j)−cs˜s(j)≥0 Due to the large number of available suppliers, the headquarter can lower the payments w(j) and t(j) so that the constraints are binding and the supplier walks away with zero profits. Because ˜x(j) = ˜q(ω), we have �1 0 w(j)dj =cx�1 0 ˜x(j)dj =cx˜q(ω).(5.34) and because of the binding constraints, �1 0t(j)dj =cs�1 0˜s(j)dj. 107 This implies that we can restate the headquarter’s problem in terms of ˜q(ω) and ˜s(j) as max ˜q(ω),{˜s(j)}j∈[0,1] ˜πF(ω) = ˜q(ω)ρ˜ A1−ρ��1 0 ˜s(j)αdj� ρ α −cx˜q(ω)−cs�1 0 ˜s(j)dj −f(5.35) Differentiating with respect to the choice variables gives the first order conditions ˜q(ω) = ρ˜r(ω) cx ˜s(j)1−α=˜r(ω)ρ cs��1 0 ˜s(j)αdj�−1 Expressing ��1 0. . . dj�in terms of revenue ˜r(ω) gives ˜s(j) conditional on a level of output ˜q(ω) as ˜s(j) = ˜s= ˜r(ω)ρ−α ρ(1−α)˜ Aα(1−ρ) ρ(1−α)˜q(ω)α 1−α�ρ cs�1 1−α ∀j The difference to equation (5.11) is that here stage jstandard depends on total revenue ˜r(ω) instead of revenue up to that stage r(j). Plugging this back into the revenue expression from the maximization problem gives equilibrium revenue ˜r(ω) as a function of ˜q(ω) as ˜r(ω) = ˜ A�˜q(ω)ρ cs�ρ 1−ρ (5.36) Plugging the result for ˜r(ω) back into ˜sfrom above gives ˜s=˜ A˜q(ω)ρ 1−ρ�ρ cs�1 1−ρ ∀j(5.37) This gives a symmetric equilibrium investment in the standard ˜sat each stage jas a function of exogenous parameters and conditional on the scale of production ˜q(ω). We know from the first FOC that ˜q(ω) = ρ˜r(ω) cx and using the result for ˜r(ω) from equation (5.36) gives that the optimal scale of production is given by ˜q(ω) =  ρ�˜ A cx�1−ρ c−ρ s  1 1−2ρ (5.38) Combining the results of equations (5.37) and (5.38) gives equilibrium CSR investment at each stage 108 jas ˜s= ρ�˜ A cs�1−ρ c−ρ x  1 1−2ρ ∀j(5.39) Using the fact that with free entry, profits are zero for all (homogeneous) firms, and applying this to equation (5.35) gives the first expression for the equilibrium price index as ˜ P=ρ−2cscx�f 1−2ρ�1−2ρ ρ E−1−ρ ρ(5.40) Plugging this into equation (5.39) for ˜ A=E˜ Pρ 1−ρgives ˜s=ρ cs f 1−2ρ.(5.41) 5.5.2.1 Level of CSR with Complete Contracts and Comparison to Incomplete Contracts Under complete contracts, the optimal CSR investment is uniform across all production stages jand is given by ˜s=ρ cs f 1−2ρ. The uniformity result arises because all stages in the production process enter symmetrically into the production of ethical quality. Therefore, a headquarter able to write complete contracts has no incentive to assign different levels of investments to suppliers at different stages along the value chain. It is instructive to compare this result with our main result from equation (5.24). In fact, it is possible to express s(m) as a function of the complete contracts result ˜sas s(m) = (1 −α)1−β βmρ−α α(1−ρ)˜s. (5.42) Under incomplete contracts, CSR levels are smaller than the CSR investments under complete contracts for all - even the most downstream - stages of production. To see this, consider the following: Under our assumption that α<ρand because s(m) is increasing in m, showing that s(1) <˜swill be sufficient. This holds when β>1−α 2−α. Because 1/2>1−α 2−α, this condition is always satisfied under our maintained assumption of β≥1/2.44 It is apparent that there are two reasons for underinvestment in the incomplete contracts case relative to complete contracts. First, there is a term that applies equally to all stages, which results from individual profit maximization of suppliers along the value chain. Second, mdenotes the position in the value chain. So while there is underinvestment at each stage, the underinvestment is more severe for smaller m, i.e., further upstream. 5.5.2.2 Remaining Industry Equilibrium Results For completeness, we also report the remaining variables of the industry equilibrium with complete contracts. Using ˜ A=E˜ Pρ 1−ρand equation (5.40) in equation (5.38), gives the production quantity in industry equilibrium as ˜q(ω) = ρ cx f 1−2ρ. 44Recall that β≥1/2 directly stems from Antr`as (2003) and 0 <α<ρis discussed in Section 5.2.2.1. 109 Using this result in inverse demand ˜p(ω) = ˜ A1−ρ˜ε(ω)ρ˜q(ω)ρ−1 as well as the definition of ˜ε ˜ε(ω) = ��1 0 ˜s(j)αdj� 1 α = ˜s gives the optimal pricing rule as ˜p(ω) = cx ρ.(5.43) With the definition of the optimal price index ˜ P=��˜n 0�˜p(ω) ˜ε(ω)�−ρ 1−ρ dω�−1−ρ ρ we get a second expression for the price index, ˜ P=cscx ρ2 1−2ρ f˜n−1−ρ ρ,(5.44) which, combined with (5.40) gives the equilibrium number of firms as ˜n= (1 −2ρ)E f(5.45) 5.5.3 The Extended Model In this section, we show that our main result continues to hold in an extended version of the model, in which we introduce an additional parameter ν∈(0,1) that captures the scope of ethical quality differentiation. The extension generalizes the model outlined in Section 5.2 and collapses to the baseline version for ν= 1. This extension allows us to relax the assumption of ρ<1/2. In particular, for arbitrarily small ν, the maximum possible value ρapproaches one. See Footnote 15 for a discussion. For expositional simplicity, we maintain the same notation as in the main text. It is well understood that the expressions in this section and in the main text are identical for ν= 1 only. In the model setup, the only change is in the utility function, which now reads as U=��ω∈Ω [ε(ω)νq(ω)]ρdω�1 ρ .(5.46) Utility maximization subject to the budget constraint gives inverse demand for variety ωas p(ω)=ε(ω)νρq(ω)−(1−ρ)A1−ρ,(5.47) where A=EP ρ 1−ρand P−ρ 1−ρ=�ω∈Ω�ε(ω)ν p(ω)�ρ 1−ρdω. Using the definition of ε(ω) given in equation (5.3), firm revenue can be written as r(ω) = ��1 0 s(j)αdj� νρ α q(ω)ρA1−ρ(5.48) 110 5.5.3.1 Incomplete Contracts With incomplete contracts, inverse demand and firm revenues are still given by equations (5.47) and (5.48). Revenue secured up to stage mis given by r(m) = q(ω)ρA1−ρ��m 0 s(j)αdj�νρ α (5.49) and the incremental contribution of stage mto overall revenue then follows as ∂r(m) ∂m=r�(m) = νρ αs(m)αr(m)νρ−α νρ q(ω)α νA(1−ρ)α νρ .(5.50) As in the baseline model, each supplier maximizes its profits, composed of the share 1 −βof the incremental contribution to final revenue net of expenditure for the production standard, max s(m)πs(m) = (1 −β)r�(m)−css(m).(5.51) Differentiating these profits with respect to s(m) gives the analog of equation (5.11) from the baseline model as s(m) = �(1 −β)νρ cs q(ω)α νA(1−ρ)α νρ �1 1−α r(m)νρ−α νρ(1−α).(5.52) Plugging this back into equation (5.50) yields a differential equation analogous to equation (5.12) in the baseline model. The solution gives an expression for revenue secured up to stage mas a function of parameters, A, and q(ω) analogous to equation (5.13) as r(m) = �1−νρ 1−α�νρ(1−α) α(1−νρ)�1−β cs νρ�νρ 1−νρ A1−ρ 1−νρ q(ω)ρ 1−νρ mνρ(1−α) α(1−νρ).(5.53) Plugging this back into equation (5.52) gives the production standard implemented at stage mas a function of parameters, q(ω), and Aas s(m) = �1−β cs νρ�1 1−νρ �1−νρ 1−α�νρ−α α(1−νρ) q(ω)ρ 1−νρ A1−ρ 1−νρ mνρ−α α(1−νρ),(5.54) analogous to equation (5.14) in the baseline model. To choose the optimal scale of production the firm has to solve max q(ω)πFω=βr(ω)−cxq(ω)−f. (5.55) Using equation (5.53) evaluated at m= 1, the solution to the maximization problem yields q(ω) = �1−νρ 1−α�νρ(1−α) α[1−ρ(1+ν)] �1−β cs νρ�νρ 1−ρ(1+ν)�A1−ρ�β cx ρ 1−νρ�1−νρ�1 1−ρ(1+ν) ,(5.56) analogous to equation (5.17). Turning to the industry equilibrium, using equation (5.53) evaluated at m= 1 as well as equation (5.56) in equation (5.55) and setting it to zero gives a first expression for the price index, analogous to 111 equation (5.19) as P=�f 1−ρ(1+ν)�1−ρ(1+ν) ρcx ρ�cs (1 −β)νρ�ν E−1−ρ ρβ−1−νρ ρ(1 −α)1−νρ ρ�1−α 1−νρ�νρ−α αρ .(5.57) We can use this result to calculate the values of q(ω) and s(m) in industry equilibrium using equations (5.56) and (5.54). q(ω) = ρ cx f 1−ρ(1+ν)(5.58) s(m) = 1−α β (1 −β)νρ cs f 1−ρ(1+ν)mνρ−α α(1−νρ)(5.59) The above equilibrium expression for q(ω) shows that q(ω) remains positive for a larger range of values for ρ. More precisely, while in the case of ν= 1, we had to impose ρ<1/2, we can now impose a weaker condition, ρ<1/(1 + ν). As νgrows arbitrarily close to zero, the upper bound of the admissible range for ρapproaches unity. The equilibrium expression for s(m) shows that our main result continues to hold also in this extended version. For our model to predict an upward sloping profile of CSR along the value chain, we need to assume that α<νρ. The remaining equilibrium variables can be calculated as follows. Using the definition of ε(ω) we then get ε(ω) = ��1 0 s(m)αdj� 1 α =1−α β (1 −β)νρ cs f 1−ρ(1+ν)�1−νρ 1−α�1 α .(5.60) Plugging this, the solution for q(ω) as well as the expression for Pinto inverse demand gives the pricing rule as p(ω)=A1−ρq(ω)−(1−ρ)ενρ =cx ρ 1−νρ β(5.61) Using ε(ω) and p(ω) in the definition of the price index from demand, P−ρ 1−ρ=�n 0�ε(ω)ν p(ω)�ρ 1−ρdω, gives a second expression for the price index as a function of the number of firms, n. Combining the two expressions gives the equilibrium number of firms as n=βE 1−νρ 1−ρ(1+ν) f.(5.62) 5.5.3.2 Complete Contracts In the case of complete contracts, we can state the firm’s problem analogous to equation (5.35) as max ˜q(ω),{˜s(j)}j∈[0,1] ˜πF(ω) = ˜q(ω)ρ˜ A1−ρ��1 0 ˜s(j)αdj� νρ α −cx˜q(ω)−cs�1 0 ˜s(j)dj −f, (5.63) with the only difference being the parameter νattached to ethical quality. Following steps analogous to those outlined in Appendix 5.5.2 gives the following results in industry equilibrium. The production standard ˜s(j) = ˜s∀jis given by ˜s=νρ cs f 1−ρ(1+ν).(5.64) 112 Because ˜sis identical across stages, it also holds that ˜ε(ω) = ˜s. The optimal quantity of final good production is given by ˜q(ω) = ρ cx f 1−ρ(1+ν),(5.65) while the optimal price is given by ˜p(ω) = cx ρ(5.66) and the equilibrium number of firms is given by ˜n= [1 −ρ(1+ν)] E f.(5.67) 5.6 Empirical Appendix This section provides details on two control variables, namely, the measure of wages and the measure of export exposure to OECD countries. 5.6.1 Wages: We control for the possibility that wages vary systematically with the value chain position. In Prowess, the wage bill is available for a subset of firms. However, there are many zeros and the provision of the number of employees is not compulsory for firms so that we cannot compute wages per employee. We therefore decide not to include the wage bill from the firm-level data and prefer to use a proxy for hourly wages taken from WIOD. We take the hourly compensation paid in Indian industries from WIOD and use the product-level information of firms to generate a firm-level control variable that is similar in spirit to our firm-level upstreamness measure. To this end, we calculate Wages as an industry’s average hourly compensation to employees weighted by a firm’s product sales in the respective industry and year. This delivers our control variable of interest, which is computed in the following way: Wagesft = V � v=1 salesfvt salesft compensationvt.(5.68) We thus compute Wagesft taking industry v’s average hourly compensation to employees in year t and weighting it with a firm’s product sales in the respective industry. Splitting the analyzed sample by the median value chain position shows a median wage level of 24.77 Rupees for the more upstream firms and of 21.65 Rupees for the more downstream firms. However, this does not control for other firm characteristics. 5.6.2 OECD export exposure: The Prowess data do not cover information on the export destinations of firms. In order to control for heterogeneity in destination markets, we construct a proxy for exposure to OECD markets based on information taken from WIOD. This measure is constructed similar to our upstreamness and wage 113 measures based on WIOD industry data as OECD export exposureft = V � v=1 salesfvt salesft exports OECDvt exports totalvt ,(5.69) where we take an industry v’s exports to OECD countries relative to industry v’s total exports in year tand weight this according to a firm’s product portfolio. 114 Bibliography Aghion, P., R. B´ enabou, R. Martin, and A. Roulet (2020): “Environmental Preferences and Technological Choices: Is Market Competition Clean or Dirty?” NBER Working Paper 26921. Aldashev, G., M. Limardi, and T. Verdier (2015): “Watchdogs of the Invisible Hand: NGO Monitoring and Industry Equilibrium,” Journal of Development Economics, 116, 28–42. Aldashev, G. and T. Verdier (2009): “When NGOs Go Global: Competition on International Markets for Development Donations,” Journal of International Economics, 79, 198–210. ——— (2010): “Goodwill Bazaar: NGO Competition and Giving to Development,” Journal of Development Economics, 91, 48–63. Alfaro, L., P. Antr` as, D. Chor, and P. Conconi (2019): “Internalizing Global Value Chains: A Firm-Level Analysis,” Journal of Political Economy, 127, 509–559. Alfaro-Ure˜ na, A., B. Faber, C. Gaubert, I. Manelici, and J. P. Vasquez (2021): “Responsible Sourcing? Theory and Evidence from Costa Rica,” mimeo, UC Berkeley. Antr` as, P. (2003): “Firms, Contracts, and Trade Structure,” Quarterly Journal of Economics, 118, 1375–1418. Antr` as, P. and D. Chor (2013): “Organizing the Global Value Chain,” Econometrica, 81, 2127–2204. ——— (2018): “On the Measurement of Upstreamness and Downstreamness in Global Value Chains,” in World Trade Evolution: Growth, Productivity and Employment, Taylor & Francis, 126–194. Antr` as, P., D. Chor, T. Fally, and R. Hillberry (2012): “Measuring the Upstreamness of Production and Trade Flows,” American Economic Review, 102, 412–16. Antr` as, P. and E. Helpman (2004): “Global Sourcing,” Journal of Political Economy, 112, 552–580. Antr` as, P. and S. R. Yeaple (2014): “Multinational Firms and the Structure of International Trade,” in Handbook of International Economics, ed. by G. Gopinath, E. Helpman, and K. Rogoff, Elsevier, vol. 4, 55–130. Antr` as, P. (2016): Global Production: Firms, Contracts, and Trade Structure, Princeton University Press. Baldwin, R. and J. Harrigan (2011): “Zeros, Quality, and Space: Trade Theory and Trade Evidence,” American Economic Journal: Microeconomics, 3, 60–88. Baron, D. P. (2012): “The Industrial Organization of Private Politics,” Quarterly Journal of Political Science, 7, 135–174. 115 Einhaltung der DFG-Richtlinien Titel des Aufsatzes: Stepping up to the mark? Firms’ export activity and environmental innovation in 14 European countries Erscheinungsort: Hanley, A.; Semrau, F.O. (2022). “Stepping up to the mark? Firms' export activity and environmental innovation in 14 European countries." Industry and Innovation, 29(5), 672-700. Autoren und Autorinnen: Aoife Hanley Finn Ole Semrau Hiermit bestätigen die Autoren und Autorinnen, die Einhaltung der DFG-Richtlinien. Alle Autoren und Autorinnen haben einen wesentlichen Beitrag zu der wissenschaftlichen Veröffentlichung geleistet. Alle Autoren und Autorinnen haben zu allen Kapiteln einen wesentlichen Anteil geleistet. Ort, Datum, Unterschrift: Prof. Aoife Hanley Ort, Datum, Unterschrift: Finn Ole Semrau Kiel, 23.8.2022 Kiel, 30.8.2022 EinhaltungderDFG-Richtlinien  TiteldesAufsatzes:  CorporateSocialResponsibilityalongtheGlobalValueChain Erscheinungsort: Herkenhoff,P.;Krautheim,S.;Semrau,F.O.andSteglich,F.(2021). “CorporateSocialResponsibilityAlongtheGlobalValueChain.” CESifoWorkingPaperNo.9498. AutorenundAutorinnen: PhilippHerkenhoff SebastianKrautheim FinnOleSemrau FraukeSteglich HiermitbestätigendieAutorenundAutorinnen,dieEinhaltungderDFGRichtlinien. AlleAutorenundAutorinnenhabeneinenwesentlichenBeitragzuderwissenschaftlichen Veröffentlichunggeleistet.ImWesentlichenhabenPhilippHerkenhoffundSebastianKrautheimam Kapitel „A Model of CSR along the Global Value Chain“ und Finn Ole Semrau und Frauke Steglich am Kapitel„GVC Position and CSR in India“ gearbeitet. Alle Autoren und Autorinnen haben zuden Kapiteln„Introduction“ und „Conclusion“ beigetragen.  Ort,Datum,Unterschrift: Dr.PhilippHerkenhoff Ort,Datum,Unterschrift: Prof.SebastianKrautheim Ort,Datum,Unterschrift: FinnOleSemrau Ort,Datum,Unterschrift: FraukeSteglich Hamburg, 8.8.2022 Kiel, 8.8.2022 Frankfurt, den 12.08.2022 Mainz, 22.08.22 Acknowledgement And I would walk 500 miles. And I would roll 500 more. The lyrics of the Proclaimers song capture many of my feelings during the time writing this dissertation. The key feeling might be the happiness of the song which feels somehow naive given that this is around walking distance between the Kiel Institute and Freiburg. Difficult to imagine that you can keep all your positivity if you know that you are walking such a distance. Especially, once you arrive in Freiburg, you realise that you have to roll around 805 km more. Looking at the map, the way back is even worse, you have to roll up. However, I think I can say that I managed somehow to keep the motivation and a positive attitude over the complete distance. Thinking back on the last years, without a doubt, has not been possible without the support of great friends, my colleagues (some even got best friends) and family. With this acknowledgement, I try to name all of them, however, I am sure the list will be incomplete and I am very sorry for that. Let me first go back in time before I started working as a doctoral student at the Kiel Institute in October 2016. The year is 2009 and I am aware that deciding what to study is a crossroads. In the end, I have two opposing options: Oceanography in Hamburg or Economics in Kiel. Even though I love the sea and it sounds counter-intuitive, I made the right decision. During this time, I was obsessed with understanding the mechanisms behind economics, which affect all of our lives and led to the financial crisis just last year. Compared to the great field trips you can go on while studying oceanography, this sounds lame even to me. But the basic confidence that I could do some good by understanding how the economy works motivated me. I had no idea what that meant. The eye-opening moment was together with Claas Schneiderheinze around 2011. We both read the book ”The Bottom Billion” by Jeffrey Sachs, which was part of a great discussion with William Easterly on how to alleviate global poverty. From that point on, we became immersed in the development economics literature. Now Claas is a colleague at the Kiel Institute, I was his best man at his wedding to Marion and he was my best man at my wedding to Nicola. When Rainer Thiele advertised a position at the Kiel Institute in 2012 in his lecture on development economics, I applied directly. The job interview was warm and inviting. This positive atmosphere continued over the next few years, during which I worked with Marcus B¨ohme, Tobias Heidland, Toman Barsbai, Linda Kleemann, Kacana Khadjavi, Esther Ademmer, Fred Goretzko and Christiane Geb¨uhr. They all helped me to take my first steps as an economist. Rainer supervised my Master’s thesis and an updated version of one chapter of it was even my first publication in a peer-reviewed journal. This support is already far from normal, but in addition, he enabled me to spend three months at the UFRJ in Rio de Janeiro in 2015. At the end of my Master’s degree, I worked as an intern at the GFA Consulting Group and learned how to apply my (quantitative) knowledge as a practitioner in development cooperation. Linda and Tobias asked me to participate in the scientific monitoring of household surveys in ten developing countries. A challenging task and not usual in German development cooperation at that time and on that scale. I learned a lot from the team and even got the opportunity to apply what I had learned to an upcoming data collection in Western Kenya. I will never forget the first moment on a boda boda and the view of Lake Victoria. It is definitive proof that working as an economist is not necessarily lame and the job I want to do. When a position was advertised at the Kiel Centre for Globalization dealing with the interface of 124 global value chains and ecological and social consequences, it was immediately clear to me that I wanted to return to the Kiel Institute. I am grateful to Sonja Peterson, who wrote me a letter of reference and supported my application. It is outstanding that there were only women in my job interview, including my two supervisors Aoife Hanley and Katrin Rehdanz. Again, I felt comfortable from the first moment, even though the team was largely new to me. At this stage, it paid offthat I besides development economics also specialised in environmental economics during my master’s degree. The link between environment and development continued to be a big motivation for my research. Social development is not possible without success in combating climate change. A key to sustainable development is the diffusion of knowledge and technology throughout the world. With my dissertation, I am trying my best to contribute to this understanding. I will summarise my time a little faster at this point because it fits with the feeling that time has flown by. Without a doubt, my two supervisors Aoife Hanley and Katrin Rehdanz deserve an extra scoop of thanks and appreciation. Thanks for all the comments and feedback you have given me over the last years. I would like to especially thank my amazing team at the research centre “Innovation and International Competition” that has given me great support in several areas, such as methodology, proofreading, discussion of the topics and technical questions: Thanks to Carmen Anderson, Dirk Dohse, Sophia Fehrenbacher, Wan-Hsin Liu, Robert Gold, Ulrich Stolzenburg, Eckhardt Bode, Christina Raasch, Frank Bickenbach, Michaela Rank and Julian Vehrke. I would like to thank the great team of young researchers at the IfW, KCG, Kiel University and the departments: Malte Becker, Sonali Chowdhry, Richard Franke, Anke Jacksohn, Anna Jakobs, Levke Jessen-Thiesen, Thilo Kr¨oger, Finja Kr¨uger, Jakob Lehr, Hendrik Mahlkow, Saskia Meuchelb¨ock, Leonie Meißner, Alina Mulyukova, Galina Potjagailo, Christoph Schweickhardt, Vincent Stamer, Sneha Thube, and my friend Tim Viergutz. The discussions with you are always enlightening. But more importantly, without you (coffee, Mobys and after-work activities) it would not be possible to keep the atmosphere so great. A special thanks go to Frauke Steglich, we started together, and are co-authors and best friends. I want to thank my senior colleagues: Martin Ademmer, Lena Detlefsen, Olivier Godart, Julian Hinz, Cecilia Hornok, Menusch Khadjavi, and Nils Jannsen. They all played different roles. Be it as supervisors or simply as fantastic colleagues. The KCG also opened the door to feedback from many great professors who supported me at all stages of the dissertation. Holger G¨org, Horst Raffand Till Requate deserve special mention. Likewise, I would like to thank my two co-authors Philipp Herkenhoffand Sebastian Krautheim. Both are excellent economists and we found at ETSG 2019 in Bern that we share a similar passion for understanding what makes a company invest in corporate social responsibility along the global value chain. In addition to my academic work, I would like to thank Stefan Fett and Bj¨orn Hecht from the Gesellschaft f¨ur Internationale Zusammenarbeit for working together on monitoring and evaluation projects in development cooperation, including great field trips to the Beira Corridor in Mozambique. A special thanks go to Linda Kleemann and her great team at the GFA Consulting Group. It is always a pleasure to work together. A balanced private life with great hobbies, friends and family might be the most important ingredient for the stamina you need in science. I would like to highlight my two sisters Lisa and Indra, my father Jossie, my aunt Guni and my grandfather Gerd. They all kept me grounded, which is perhaps due to living on the quiet and pleasant west coast of Schleswig-Holstein. A very special thank goes to my wife Nicola Posniak. We met right at the beginning of our studies, which alone is enough to determine that I made the right decision to study economics. Since then, we have supported each other in all areas, including discussions about our professions. You gave me strength over all the years and your view from a business perspective is always enlightening for my research. Furthermore, we share an interest 125 in sports, especially kitesurfing and running. Together, this gives us the right balance to keep our heads up whatever life throws at us. It’s great to have you by my side. Indeed, life is not always on the bright side and I would like to dedicate part of this work to Moritz Bothe, Martin Grimm, my grandma Christa and my mother Carmen. My memories are still very much alive, even though it is hard to accept that you cannot celebrate the completion of this milestone with me. Where do I stand now? Honestly, I still don’t think I understand all facets of economics. So I will continue on my path as an economist and just walk another 500 miles. 126