Policy Research Working Paper 11036 The Emergence and Diffusion of Green Technologies Firm-level Evidence from Textual Analysis of Patents and Earnings Calls Paulo Bastos Lucio Castro Development Research Group & Prosperity Vertical January 2025 Policy Research Working Paper 11036 Abstract This paper uncovers new stylized facts on the emergence and mentioning keywords associated with them. Among ini- diffusion of green technologies across countries, sectors, and tially high-emissions firms, those that mentioned green firms. It draws on the textual analysis of patents and corpo- technologies in earnings calls tended to observe a decline rate earnings calls matched with multi-country, firm-level in carbon emissions in subsequent years. Buyer-supplier panel data for 2012–2021. The paper documents the grow- relationships and innovation partnerships with these firms, ing importance of green technologies after 2019, as revealed especially when they had high-emissions intensity, are sys- by a rapid increase in the share of earnings calls transcripts tematically linked with the diffusion of green technologies. This paper is a product of the Development Research Group, Development Economics and the Prosperity Vertical. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank. org/prwp. The authors may be contacted at pbastos@worldbank.org and lcastro@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team The Emergence and Diffusion o f G reen Technologies: Firm-level Evidence from Textual Analysis of Patents and Earnings Calls∗ Paulo Bastos† Lucio Castro‡ Keywords: Green technologies, earnings calls, patents, carbon emissions, firm-to-firm linkages. JEL Classification: F6, O3, Q5. ∗ We are grateful to Juan Tapia Montenegro, Samuel Asuquo Edet and Matias Garibotti for excellent research assistance. We are also grateful to Juan Pablo Rud and seminar participants for very useful comments. The authors gratefully acknowledge financial support from the Umbrella Facility for Trade, which is funded by the governments of The Netherlands, Norway, Switzerland, Sweden and the United Kingdom, and the Research Support Budget of the World Bank. † Development Research Group, World Bank; CEPR; and REM/UECE. E-mail: pbastos@worldbank.org. ‡ World Bank. Email: lcastro@worldbank.org 1 Introduction There is now wide agreement that progress towards the decarbonization of the global economy will require a significant increase in the utilization of technologies that contribute to reducing CO2 emissions. While these green technologies are becoming more ubiquitous, there is still little systematic evidence on their emergence and diffusion across firms, sectors, and countries, and the economic mechanisms underlying their diffusion. In this paper, we draw on the textual analysis of patents and earnings conference calls matched with multi-country, firm-level panel data for 2012-2021 to provide systematic evidence on these issues. We begin by establishing new stylized facts about the emergence and dissemination of green technologies over the last two decades using textual analysis of patents. We identify a set of business and technical-relevant keywords for technologies aimed at protecting the environment and making a positive impact on climate change mentioned in the World Intellectual Property Organization (WIPO) green patents database. This approach allows us to identify in a sys- tematic way green technologies. The analysis of patent data for the period 2000-2021 reveals that green innovation was primarily driven by new technologies from the construction, energy, and transportation sectors, including technologies such as biofuels, geothermal energy, LED technology, drones, and hybrid/electric vehicles. We then provide evidence on patterns of adoption and diffusion of these green technologies across firms and countries, drawing on the textual analysis of earnings conference calls tran- scripts (ECTs). We follow a growing literature seeking to infer firm-level technology adoption using general-text classification methods for identifying business-relevant keywords associated with specific technologies; and by counting the number of times these words are mentioned in conference calls that publicly listed firms host with financial analysts. This approach has been increasingly used by researchers to measure firm uncertainty (Hassan, Hollander, van Lent, and Tahoun, 2019), firm sentiment (Hassan, Hollander, van Lent, and Tahoun, 2020), and the dif- fusion of disruptive technologies (Bloom, Hassan, Kalyani, Lerner, and Tahoun, 2021). The rationale behind it is that earnings calls are a venue in which senior management responds directly to questions from market participants about the firm’s prospects. They consist of a management presentation followed by a Q&A session, which requires management to comment on subjects they might not otherwise have voluntarily proffered. Thus, the extent to which the adoption of green technologies is discussed in earnings conference calls serves as a proxy of their importance for both the firm’s management and market participants. Using unique firm identifiers, we were able to merge these firm-level textual-based indicators 1 with information on inter-firm relationships and firm-level information for the period 2012 to 2021. These multi-country firm-level data cover buyer-supplier relationships, thereby allowing us to inspect for evidence of dissemination of green technologies across global value chains. Furthermore, they contain information on innovation partnerships between firms. The textual analysis of earnings conference calls transcripts points to a rapidly growing im- portance of green technologies after 2019, as revealed by a rapid increase in the share of ECTs mentioning keywords associated with specific green technologies, as well as in the set of firms that have ECTs mentioning these green technologies. This rise is primarily driven by a few green technologies, including construction material, carbon capture technology, hybrid/electric vehicles, and wind turbines, which tend to be mentioned together in the same ECT. It is concen- trated in firms operating in a few industries—notably manufacturing and utilities—and among larger firms within sectors. We also find evidence of an increased geographical spread in the location of these green firms, with a growing presence in large emerging countries in Asia, Latin America, and the Middle East and North Africa, in addition to North America and Europe. Importantly, we find that high-emissions firms that mention green technologies in their ECTs are significantly more likely to observe a subsequent reduction in their carbon emissions, which supports the hypothesis that ECTs capture firm investments in green technologies among these firms. We then examine the extent to which the diffusion of green technologies between firms is shaped by firm-to-firm linkages, notably buyer-supplier relationships and innovation partner- ships. We find robust evidence that suppliers and buyers of firms that mention green technology keywords in their earnings calls are significantly more likely to mention them as well in their own ECTs. These effects are especially relevant when firms are linked to high-emissions firms that mention green technologies in their own ECTs. Our data consists mainly of multinational enterprises (MNEs), but we also find some evidence of diffusion of green technologies through firms-to-firm linkages to (and from) firms that are not MNEs. In addition to the papers cited above, this paper contributes to several strands of literature. Prior research on low-carbon technologies has often confined its attention to the diffusion of a few technologies in a specific sector or country (see, e.g., Costantini, Crespi, and Palma (2017)). In addition, the literature on low carbon technologies has focused mostly on their invention etre, H´ and production (Popp, 2002, 2022; Aghion, Dechezleprˆ emous, Martin, and Van Reenen, etre, Glachant, and M´ 2016; Dechezleprˆ ere, 2013) or on the costs of adopting them (Way, eni` Ives, Mealy, and Farmer, 2022; Grubb, Drummond, Poncia, McDowall, Popp, Samadi, Penasco, etre, and Pavan, 2021), Gillingham, Smulders, Glachant, Hassall, Mizuno, Rubin, Dechezleprˆ 2 but has had relatively little to say about whether, how and to whom they have diffused across countries, industries, and firms. An important exception is a recent paper by Bastos, Greenspon, Stapleton, and Taglioni (2024), who draw on textual analysis of online job postings to document the diffusion of low carbon technologies across 35 countries, and examine econometrically the role of the 2022 global energy crisis in accelerating their diffusion across countries and firms. The current paper provides complementary stylized facts on both the emergence and diffusion of green technologies by drawing on the textual analysis of patents and ECTs in a wide range of countries; examining the link between green technology adoption and subsequent changes in emissions at the firm-level; and documenting the role of buyer-supplier links in the diffusion of green technologies. Finally, our paper is also related to the broader literature on technology diffusion, including Comin and Hobijn (2004), Comin and Hobijn (2010), Comin and Mestieri (2018), Bloom, Hassan, Kalyani, Lerner, and Tahoun (2021) and Bastos, Stapleton, Taglioni, and Wei (2024). Our approach for examining the role of firm-to-firm linkages in the diffusion of low carbon technologies is related to that adopted in Bastos, Stapleton, Taglioni, and Wei (2024), who use data on online job postings and information on firm-to-firm linkages to examine the spread of new technologies, but do not focus on green technologies. The remainder of the paper is organized as follows. Section 2 describes the various sources of data, and our approach to identify green technologies and associated keywords. Section 3 presents stylized facts on innovation and dissemination of green technologies based on patent data from the WIPO green database, which allows us to assess the dissemination of the most novel and influential green technologies. Section 4 presents evidence about the diffusion of green technologies among firms, sectors, and countries using textual analysis of earnings calls transcripts. Section 5 reports evidence of patterns of diffusion of green technologies through firm-to-firm relationships. Section 6 concludes the paper with the main takeaways. 2 Data sources and textual analysis methodology We identify a set of business and technical-relevant keywords for technologies aimed at protecting the environment and making a positive impact on climate change mentioned in the World Intellectual Property Organization (WIPO) green patents database. Using these keywords and textual analysis of patents and earnings conference calls, we build a multi-country firm-year panel data set of exposure to green construction technologies spanning the period 2012-2022 and covering 10,000 firms in 79 countries. This section describes the data sources and our approach in more detail. 3 2.1 WIPO Green Patents We use information from the WIPO green patents data set for the period 2000-2021. This data set contains detailed information on green technology patents, inventions, and technologies from the WIPO Patentscope database, covering seven sectors: Energy, Water, Farming and Forestry, Pollution Waste, Transportation, Construction, and Products, Materials and Processes (PMP). Green technologies are defined by WIPO as “climate friendly technologies that protect the environment, are less polluting, use resources in more sustainable manner, recycle their wastes and products and handle residual waste in a more sustainable manner”. These technologies include know-how procedures, goods and services, and equipment as well as organizational and managerial procedures. WIPO systematically gathers patent data from 193 national Intellectual Property (IP) offices worldwide. The database provides information about the year of filling of the patent, the country where the IP office and the inventor are located, and technical description and status of the technology (e.g. in development, commercialization, etc.). 2.2 Earnings calls transcripts We use data from Refinitiv Eikon (formerly Thompson Reuters) on the transcripts from quarterly shareholder earnings calls meetings of publicly listed firms from 2012 to 2021, covering 10,554 firms in 82 countries. Earnings calls are key corporate events on the investor relations agenda in which senior management responds directly to questions from financial analysts and other market participants about the firm’s financial performance over the past quarter and, more broadly, discuss current developments (Hollander, Pronk, and Roelofsen, 2010; Hassan, Hollander, van Lent, and Tahoun, 2020). They consist of a management presentation and, importantly, a Q&A session which requires management to comment on subjects they might not otherwise have voluntarily proffered. The extent to which the adoption of green technologies is discussed in earnings conference calls serves as a proxy of their importance for both the firm’s management and market participants. Given that green technologies potentially account for a large share of firms’ investment decisions, they would be expected to feature in conference call discussions. In fact, these data have been increasingly used by researchers to measure firm uncertainty (Hassan, Hollander, van Lent, and Tahoun, 2019), firm sentiment (Hassan, Hollander, van Lent, and Tahoun, 2020), and proxies for whether new technologies are “disruptive” as inferred by being widely discussed in boardrooms (Bloom, Hassan, Kalyani, Lerner, and Tahoun, 2021). 4 2.3 Factset data on firm linkages and carbon emissions We use data on firm relationships from FactSet, a private company specializing in market intel- ligence and firm networks. FactSet collates data from various sources, including annual reports, investor presentations, press releases, and company websites. Among FactSet’s offerings, the FactSet Fundamentals provides consolidated financial statements and firm information for pub- licly traded companies, while FactSet Relationships covers supply chain relationships, including suppliers, customers, and partnerships, as well as firm-to-affiliate relations. The FactSet Revere Supply Chain Relationships database is regarded as one of the most comprehensive sources for global firm supply chain information (Huang, Lin, Liu, and Tang, 2023). This dataset integrates information from official firm filings (10-K, 10-Q, and 8-K reports submitted to the SEC), investor presentations, press releases, and company websites (FactSet, 2021).1 FactSet analysts normalize relationship types by cross-referencing a firm’s records with its partners’ (“reverse” links), providing a comprehensive coverage of supply chain linkages (Huang, Lin, Liu, and Tang, 2023). As of March 2024, FactSet actively monitored over 54,000 “source” companies, resulting in a dataset of more than 2 million relationship links involving approximately 360,000 entities. Each link includes information on start and end dates, with historical data available from 2003 and global coverage starting in 2014, expanding to Latin America in 2016.2 The database specifies the nature of each relationship. For our study, we focus on buyer and supplier relationships. At any given time, we observe the parent and ultimate parent firm of a company, and any subsidiaries, if applicable.3 The dataset also includes firm name, industry affiliation, headquarters country, and entity type, which we use in our analysis. Annual sales data, converted to constant 2020 U.S. dollars, are available but not used in our study. We combine the two aforementioned FactSet data modules to construct a final dataset of 169,420 entities with bilateral relationship information from 2012 to 2022. These entities are defined as de jure corporate institutions (e.g., firms and affiliates), with 27% (46,000 entities) having available information from FactSet Fundamentals for at least one year within the sample period. To measure carbon emissions at the firm level, we draw on the ISS ESG Carbon and Climate Impact (ICC) data from FactSet. This data set provides comprehensive coverage of emissions 1 In financial reporting, a 10-Q is a company’s quarterly report with unaudited financial statements, while the 10-K is an annual report with audited financials. An 8-K updates shareholders on significant unscheduled events. U.S. listed firms must detail any customer accounting for more than 10% of revenue (Gofman and Wu, 2022), with some firms voluntarily disclosing additional customers (Huang, Lin, Liu, and Tang, 2023). 2 Many Latin American records from 2014 suggest backward updating for these firms. 3 The parent or ultimate parent is identical to the company itself for self-owned firms. 5 data, accounting for all the emissions of the firm (including subsidiaries), spanning from 2012 to 2021. The ICC includes detailed information on companies’ Scope 1, Scope 2, and Scope 3 emissions and emissions’ intensity, sourced from official reports such as Sustainability and ESG reports, the Carbon Disclosure Project (CDP), as well as estimated emissions from ISS ESG when official data is not available. For the analysis, we focus specifically on Scope 1 emissions since these are direct GHG emissions from sources owned or controlled by the firm. 2.4 List of green technologies and associated keywords In line with the approach of Bloom, Hassan, Kalyani, Lerner, and Tahoun (2021), we use the WIPO green patent data and the ECTs to identify biagrams or a set of two words that represent each green technology. We follow several steps. First, we extract the summary of the green technologies included in each patent in the WIPO database. Second, we transform the summary’s text into biagrams by applying a text recognition algorithm, removing numbers, special characters, and stop words (e.g. of, the, from, etc.) and converting each plural word into its singular form (eg. batteries, battery). Third, we drop bigrams that are not included in at least 20 patents and prior to the year 2000. The rationale behind this choice is to include not only the most influential but also the most novel technologies in our analysis. Fourth, we perform a human audit of the resulting biagrams. Fifth, we train a word embedding algorithm (word2ovec) with the patent’s summary to identify similar biagrams to each biagram and dropped non- technical bigrams using Google Bigrams. Sixth, we perform a human audit of the new set of biagrams. Seventh, we train the word embedding algorithm with ECTs to capture business-like language referring to technical bigrams. Finally, we restrict the focus to technologies mentioned in more than 50 ECTs during the 2012-2021 period. Table A1 in the Appendix contains the final set of green technologies considered and the corresponding bigrams searched in the textual analysis of patents and earnings calls. Table A2 reports the number of ECTs that mention each technology in the period 2012-2021. As can be observed in this Table, the technologies that appear more frequently in ECTs include construction material, air conditioning, wind turbines, LED technology, hybrid/electric vehicles, Li-ion batteries, carbon capture technologies, transgenic crops, biofuels, and solar energy. 6 3 Evidence on the emergence and dissemination of green tech- nologies identified in patents In this section, we present the main findings of our analysis of innovation and diffusion of green technologies based on patent data from the WIPO green database. This analysis allows us to assess the dissemination of the most novel and influential green technologies. As discussed above, we follow a recent literature seeking to infer innovation and dissemination of technologies using general-text recognition methods for identifying patents associated with specific technologies. Figure 1 reports the share of each WIPO green sector in the total mentions of green tech- nologies in patents. The evidence in this figure reveals that green technologies belonging to the WIPO green sectors energy and construction account for the vast majority of mentions in patents over the period 2000-2021, accounting for about 80 percent of total mentions in most years. Transportation is the third most important sector, and its relative importance experi- enced significant growth in the mid-2000s, remaining relatively stable thereafter. Technologies from the PMP and Farming and Forestry sectors are the least important, accounting together for less than 10 percent of total patent mentions in most years. Figure 2 reports the percentage of patents mentions by technology within each WIPO green sector over the same period. Among green technologies from the construction sector, we observe a clear downward trend for the relative importance of air conditioning, and a growing importance of LED technology, sensors, and construction material. Since 2005, LED technology accounted for about 40 percent of total mentions in patents within the construction sector, while the other technologies stabilized at around 20 percent each. In the energy sector, we observe a growing importance of biofuels, geothermal energy, and inverter technology. In transportation, drones accounted for a progressively higher share of total mentions, reaching about 80 percent in the early 2000s and stabilizing around 70 percent between 2015 and 2020. Among the PMP sector, carbon capture became the most important technology since 2005, reaching about 60 percent of total mentions at the end of the sample period. In Farming and Forestry, transgenic crops became the most important technology (as measured by mentions in patents) since 2012, reaching about 70 percent of total mentions in this sector at the end of the sample period. 7 4 Evidence on the diffusion of green technologies as revealed by ECTs We now turn to the evidence on patterns of adoption and diffusion of these green technologies across firms and countries drawing on the textual analysis of earnings conference calls tran- scripts. As discussed above, we follow a growing literature seeking to infer firm-level technology adoption using general-text classification methods for identifying technical-relevant keywords associated with specific technologies; and by counting the number of times these words are men- tioned in the quarterly earnings conference calls that publicly listed firms host with financial analysts. The rationale behind this approach is that earnings calls are a venue in which senior management responds directly to questions from market participants about the firm’s prospects. They consist of a management presentation followed by a Q&A session, which requires manage- ment to comment on subjects they might not otherwise have voluntarily proffered. Thus, the extent to which the adoption of green technologies feature in ECTs serves as a proxy of their importance for both the firm’s management and market participants. The evidence in Figure 3 reveals that green technologies have featured more prominently in ECTs and firms since 2015. In 2015, green technologies were mentioned in about 5 percent of ECTs and 8 percent of firms. By 2021, the last year of our sample period, green technologies were mentioned in 10 percent of ECTs and by 18 percent of firms. This increase in the relevance of green technologies in quarterly earnings conference calls was especially marked after 2019, rising from about 6 percent to 10 percent of ECTs and from 12 to 18 percent of firms. Panel A in Figure A1 in the Appendix reveals that a similar pattern is observed when looking at the economic relevance of firms mentioning green technologies in their ECTs, which increased from around 20 percent of sales in 2019 to almost 35 percent of sales in 2021. This suggests that firms mentioning green technologies in ECTs tend to be relatively large, a pattern to which we will return below. Looking across WIPO green sectors in Figure 4, we observe that this marked increase ob- served after 2019 was accompanied by a rise in the relative importance of technologies from the PMP sector, whose relative importance became similar to those from the energy sector. Mentions of technologies from the construction sector accounted for the majority of mentions in 2021, but their relative importance observed a downward trend since 2014. Figure 5 reveals that a few technologies drove these trends. In the PMP sector, carbon capture experienced a marked increase after 2019. In the energy sector, we observe a significant increase in the relative importance of Wind Turbines and Li-ion Batteries after 2019. In the construction sector, we ob- 8 serve a significant rise in mentions of green construction material. Finally, in the transportation sector, we observe a marked increase in mentions of hydro-electric vehicles after 2019. Interestingly, Figure 6 reveals that green technologies are often not mentioned in isolation in ECTs. The figure depicts a network representation of the relationship between the green technologies identified in ECTs over the period 2012-2021. The position of the nodes indicates how connected each technology is in the entire network, with technologies in the center being the most frequently mentioned in boardroom discussions. The closeness between technologies indicates how related they are between them, i.e. how often they are mentioned together in the same ECT. The results reveal that Construction Material, LED technologies, Air conditioning, hybrid-electric vehicles wind turbines, and Li-ion batteries are the most central technologies, and show a high degree of closeness among themselves. Figure A2 in the Appendix depicts the distribution of ECTs mentioning green technologies across the NAICS sectors to which the corresponding firm belongs. Panel A in this figure reveals that manufacturing and utilities are the main sectors in which these firms operate. In 2021, manufacturing firms accounted for about 60 percent of ECTs mentioning green technologies, while firms in the utilities sector accounted for more than 25 percent of total mentions. Panel B reveals that, in these two sectors, the proportion of ECTs mentioning green technologies was clearly the largest among all sectors, and observed a significant rise after 2019. As discussed above, by the end of the sample period firms mentioning green technologies in their ECTs account for a relatively large share of total sales. This is in part because they tend to be relatively large firms. The evidence in column (1) of Table 1 reveals for technologies within each WIPO Green sector, the probability of mentioning green technologies is positively and significantly associated with firm sales. These regressions include year, sector and country fixed effects and hence these positive correlations are observed within these cells. The coefficients tend to be stronger for technologies from construction and energy, but are also positive and significant for PMP and Transportation. Only for Farming and Forestry we do not observe a significant correlation with firm size. Th estimates in column (2) reveal that a broadly similar pattern can be observed for firms’ operating income, which also displays a positive correlation with the probability of mentioning green technologies in construction, energy and transportation. Figure A4 in the Appendix reveals that a positive association between firm sales and the likelihood of mentioning green technologies in ECTs is also observed within each NAICS sector. The sole exceptions are for firms in accommodation and food services and from finance and insurance. Figure A6 in the Appendix shows the geographic distribution of green firms per capita in the 2012-2017 period and the 2018-2021 period. The evidence in this figure can be compared 9 with the distribution of firms from the full sample in the same period (Figure A5). Taken together, the analysis of these figures reveals that green technologies are mainly used in high- income countries and large developing countries such as China, India, the Russian Federation and Brazil. It also reveals evidence of some diffusion to these and other emerging economies in the period 2018-2021, as revealed by a growing presence of firms with ECTs mentioning green technologies in these countries. Although the existing literature suggests that the mention of green technologies in ECTs is likely to be a suitable proxy for the adoption of green technologies, this approach may be subject to some measurement error, for example, if firms mention green technologies in ECTs but do not invest in them, or if these investments are relatively small to have a meaningful impact on carbon emissions. In Table 2, we provide more direct evidence on these concerns. We examine the extent to which the presence of keywords associated with green technologies in ECTs is followed by a decline in carbon emissions in the corresponding firms. The dependent variable is the log of Scope 1 emissions of the firm, while the independent variable is a dummy variable that equals 1 in the year the firm has an ECT mentioning at least one green technology and in all subsequent years. We also include interactions between this variable and a dummy variable for firms that in 2012 had emissions above the sample mean. Our prior is that these firms would have greater incentives to make significant investments in green technologies that reduce their carbon emissions. The regressions include year-fixed effects, firm-fixed effects, and time-varying controls (notably firm sales). The estimates in columns (1) and (2) of Table 2 reveal that having a green ECT leads to a systematic reduction in carbon emissions among firms that had high-emissions to begin with – and hence had a greater incentive to make significant investments in green technologies. These results are observed in the full sample, as well as in sub-samples referring to ECTs mentioning technologies from the energy sector, farming and forestry, and transportation. For construction and PMP, we also observe a negative coefficient on the interaction term of interest, but this coefficient is not statistically significant. In Table 3, we examine the robustness of this finding, considering an alternative definition of high-emissions firm. In particular, we adopt a dummy variable for firms that in 2012 had emissions above the 90th percentile of the sample. Once again, the estimates in columns (1) and (2) of Table 3 reveal that having a green ECT leads to a systematic reduction in carbon emissions among firms that had high-emissions to begin with. These results are observed in the full sample, as well as in sub-samples referring to ECTs mentioning technologies from most sectors. The coefficients of interest are larger than in Table 2, consistent with a larger reduction 10 in emissions for firms that had higher emissions to begin with. 5 Evidence on the diffusion of green technologies through firm- to-firm linkages The evidence in the previous section revealed that green technologies featured more prominently in ECTs after 2019, especially among large firms from the manufacturing and energy sectors, and across a wide range of high-income and large emerging economies. We have also shown that the mentioning of green technologies in ECTs tends to be followed by a reduction in scope 1 emissions of the corresponding firms that had high emissions to begin with. In this section, we examine the extent to which the diffusion of green technologies across firms and countries was systematically linked with firm-to-firm relationships. We consider four types of firm-to-firm linkages: innovation partnerships with a green firm, customer of a green firm, supplier of a green firm, and customer and supplier of a green firm. Table 3 reports the estimates on the relationship between having a linkage with a green firm in year t (i.e. with a firm that has an ECT mentioning keywords related to green technology) and being a green firm in year t. We also consider differential effects if the linkages were with a high-emission firm, i.e. with a firm that had emission intensity (GHG emissions per million USD) in 2012 above the sample mean. The data set for these regressions is at the related firm pair-year level, where each related firm pair must have had at least one active relationship (customer, supplier or tech partner) at any point between 2012-2021. The yearly coverage for each related firm pair ranges from 2012 to 2021, conditional on having relationship information of both firms for each year. The estimates in column (1) of this table reveal that firms that had an innovation partnership with a green firm in year t were significantly more likely to also have an ECT mentioning a green technology in year t. Column (2) shows that this relationship is considerably stronger if the linkage was with a firm that had high emissions in 2012 (and for which the evidence above shows a significant decline in carbon emissions after mentioning a green technology in their ECTs). Columns (3)-(4), (5)-(6) and (7)-(8) show a similar association for the other types of linkages considered: respectively, customer of a green firm, supplier of a green firm, and customer and supplier of a green firm. Taken together this evidence suggests that these firm-to-firm linkages are systematically associated with the diffusion of green technologies across firms and countries, especially if the green firm with which the linkage exists was a firm with high-emissions. 11 Table 4 reports analogous estimates, but allowing for a 1-year lag in the transmission of green technologies (as measured by mentions in the corresponding ECTs). The evidence shows a broadly similar pattern with positive and significant coefficients for the variables of interest. The sole exception for the coefficient on the interaction term in column (8), which suggests that when firms are simultaneously buyers and suppliers of a green firm, they are significantly likely to become green irrespective of the level of emissions of the green firm to which they were linked. Since firms may have various types of linkages at the same time, it may be difficult to isolate the role of each type of linkage when considering the full sample. For robustness, in Table 5 we examine if the baseline results are sensitive to the exclusion of firms with multiple linkages. Once again, we observe that the results remain qualitatively very similar. The majority of firms in our sample have affiliates in other countries and hence are multi- national enterprises. In Table 6 we restrict the sample to include only linkages with MNEs. Once again we observe qualitatively similar results. In Table 7, we consider only linkages with non-MNEs. In this case, we are left with a relatively small number of observations, but we still observe positive and significant coefficients for most linkages considered. In Table 8 we consider instead a sub-sample in which the firm that is linked to the green firm is a MNE, while the green firm could be or not an MNE. Once again, the baseline results remain valid. Finally, in Table 9 we consider a sub-sample in which the linked firm is not an MNE, while the green firm may or not be an MNE. In this case, we are again left with a relatively small number of observations, but we observe once again a positive and significant coefficient for most linkages considered. Taken together, these estimates suggest that firm-to-firm linkages play an important role in the transmission of green technologies across multinational enterprises, especially when the green firm had high emissions to begin with. We also observe some evidence that these linkages matter when one of the firms involved is not an MNE, but in this case the relationships of interest are estimated from a small sample and are less precise. 6 Concluding remarks While there is a growing recognition that green technologies are key for reducing the carbon footprint of economic activity at the global level, there is still little systematic evidence on their emergence and diffusion across firms, sectors, and countries, and on the economic mechanisms underlying this process. In this paper, we have used textual analysis of patents and earnings conference calls matched with multi-country, firm-level panel data for 2012-2021 to provide 12 systematic evidence on these issues. We first uncovered new stylized facts on the emergence and dissemination of green tech- nologies over the last two decades using textual analysis of patents. We identified a set of business and technical-relevant keywords for technologies aimed at protecting the environment and making a positive impact to climate change mentioned in WIPO green patents data set over the period 2000-2021. The textual analysis of these data suggests that green innovation was primarily driven by new technologies from the construction, energy and transportation sec- tors, including technologies such as biofuels, geothermal energy, LED technology, drones and hybrid/electric vehicles. We have then provided evidence on patterns of adoption and diffusion of these green tech- nologies across firms and countries drawing on the textual analysis of ECTs. We followed a growing literature seeking to infer firm-level technology adoption using general-text classification methods for identifying business-relevant keywords associated with specific technologies; and by counting the number of times these words are mentioned in conference calls that public listed firms host with financial analysts. Using unique firm identifiers, we were able to merge these firm-level textual based indicators with information on inter-firm relationships and firm-level information for the period 2012-2021. These multi-country firm-level data cover buyer-supplier relationships, thereby allowing us to inspect for evidence of dissemination of green technologies across global value chains. They also contain information on innovation partnerships between firms. The textual analysis of earnings conference calls transcripts reveals a growing importance of green technologies since 2019, as revealed by a rapid increase in the share of ECTs mentioning keywords associated with specific green technologies, as well as in the set of firms that have ECTs mentioning these green technologies. This increase is primarily driven by a few green technolo- gies, including construction material, carbon capture technology, hybrid/electric vehicles, and wind turbines, which tend to be mentioned together in the same ECT. It is concentrated in firms operating in a few industries— especially manufacturing and utilities and among larger firms within industries. We also find evidence of an increase in the spatial spread of these green firms, with a growing presence in large emerging countries in Asia, Latin America and the Middle East and North Africa, in addition to North America and Europe. Importantly, we find that firms that initially had a high degree of emissions intensity that mention green technologies in their ECTs are significantly more likely to observe a subsequent reduction in carbon emissions. This evidence supports the hypothesis that ECTs capture firm investments in green technologies among these firms. 13 We then studied if the diffusion of green technologies between firms is mediated by firm- to-firm linkages, notably buyer-supplier relationships and innovation partnerships. We report robust evidence that suppliers and buyers of firms that mention green technologies in their earnings calls are significantly more likely to mention them as well in their own ECTs. These effects are especially relevant when firms are linked to green high-emissions firms. Although our data consists mainly of multinational enterprises (MNEs), we also find some evidence of green technology diffusion through firms-to-firm linkages to (and from) firms that are not MNEs. References Aghion, P., A. Dechezlepre ˆtre, D. He ´mous, R. Martin, and J. Van Reenen (2016): “Carbon Taxes, Path Dependency, and Directed Technical Change: Evidence from the Auto Industry,” Journal of Political Economy, 124(1), 1–51. Bastos, P., J. Greenspon, K. Stapleton, and D. Taglioni (2024): “Did the 2022 global en- ergy crisis accelerate the diffusion of low-carbon technologies?,” World Bank Policy Research Working Paper. Bastos, P., K. Stapleton, D. Taglioni, and H. Wei (2024): “Firm networks and Global Technology Diffusion,” World Bank Policy Research Working Paper 10905. Bloom, N., T. A. Hassan, A. Kalyani, J. Lerner, and A. Tahoun (2021): “The Diffusion of New Technologies,” Working Paper 28999, National Bureau of Economic Research. Comin, D., and B. Hobijn (2004): “Cross-country technology adoption: Making the theories face the facts.,” Journal of Monetary Economics, 51, 39–83. (2010): “An exploration of technology diffusion,” American Economic Review, 100, 2031–59. Comin, D., and M. Mestieri (2018): “If technology has arrived everywhere, why has income di- verged?,” American Economic Journal: Macroeconomics, 10(3), 137–178. Costantini, V., F. Crespi, and A. Palma (2017): “Characterizing the policy mix and its impact on eco-innovation: A patent analysis of energy-efficient technologies,” Research Policy, 46(4), 799–819. Dechezlepre ˆtre, A., M. Glachant, and Y. Me `re (2013): “What Drives the International ´nie Transfer of Climate Change Mitigation Technologies? Empirical Evidence from Patent Data,” Envi- ronmental & Resource Economics, 54(2), 161–178. FactSet (2021): “Supply Chain Relationships - Data and Methodology Guide,” https://www.factset. com/marketplace/catalog/product/factset-supply-chain-relationships, documentation ac- cessed on 2023-02-06. (2023): “FactSet Supply Chain Relationships [Dataset],” https://www.factset.com/ marketplace/catalog/product/factset-supply-chain-relationships, data accessed from server on: 2024-03-20. Gofman, M., and Y. Wu (2022): “Trade credit and profitability in production networks,” Journal of Financial Economics, 143(1), 593–618. Grubb, M., P. Drummond, A. Poncia, W. McDowall, D. Popp, S. Samadi, C. Penasco, K. T. Gillingham, S. Smulders, M. Glachant, G. Hassall, E. Mizuno, E. S. Rubin, A. Deche- zlepre ˆtre, and G. Pavan (2021): “Induced innovation in energy technologies and systems: a review of evidence and potential implications for COsub2/sub mitigation,” Environmental Research Letters, 16(4), 043007. Hassan, T. A., S. Hollander, L. van Lent, and A. Tahoun (2019): “Firm-Level Political Risk: Measurement and Effects,” Quarterly Journal of Economics, 134(4), 2135–2202. 14 Hassan, T. A., S. Hollander, L. van Lent, and A. Tahoun (2020): “The Global Impact of Brexit Uncertainty,” (26609). Hollander, S., M. Pronk, and E. Roelofsen (2010): “Does silent speak? An empirical analysis of disclosure choices during conference calls.,” Journal of Accounting Research, 48(3), 531–563. Huang, Y., C. Lin, S. Liu, and H. Tang (2023): “Trade networks and firm value: Evidence from the US-China trade war,” Journal of International Economics, p. 103811. Popp, D. (2002): “Induced Innovation and Energy Prices,” American Economic Review, 92(1), 60. (2022): “Environmental policy and innovation: a decade of research,” . Way, R., M. C. Ives, P. Mealy, and J. D. Farmer (2022): “Empirically grounded technology forecasts and the energy transition,” Joule, 6(9), 2057–2082. 15 Figure 1: Percentage of patents mentions by WIPO green sector, 2000-2021 100 80 % of patents mentions 60 40 20 0 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 Construction Energy Farming and Forestry PMP Transportation Notes: The figure depicts the share of patents that belong to each of the 5 WIPO Green sectors in the period 2000-2021. 16 Figure 2: Percentage of patents mentions by technology within each WIPO green sector, 2000- 2021 Construction Energy Farming and Forestry 80 Air Conditioning 80 Biofuels 80 Soiless Agriculture LED Technology Geothermal Energy Transgenic Crops Sensors Inverter Technology Water Saving Irrigation Construction Material Solar Energy 60 60 60 Li-ion Battery 40 40 40 20 20 20 % of patents mentions 0 0 0 00 05 10 15 20 00 05 10 15 20 00 05 10 15 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 PMP Transportation 80 Carbon Capture 80 Drones Insulating Glass Hybrid/Electric Vehicles 60 60 40 40 20 20 0 0 00 05 10 15 20 00 05 10 15 20 20 20 20 20 20 20 20 20 20 20 Notes: The figure depicts the share of patents that belong to each of the selected technologies within each WIPO Green sector in the period 2000-2021. 17 Figure 3: Percentage of green ECTs and firms, 2012-2021 A. ECTs 10 % of ECs mentioning green tech 8 6 4 12 13 14 15 16 17 18 19 20 21 20 20 20 20 20 20 20 20 20 20 B. Firms 18 % of firms with at least one EC mentioning green tech 16 14 12 10 8 12 13 14 15 16 17 18 19 20 21 20 20 20 20 20 20 20 20 20 20 Notes: Panel A depicts the share of ECTs mentioning at least one green technology over the period 2012-2021. Panel B depicts the share of firms that mentioned at least one green technology in the corresponding ECTs. 18 Figure 4: Percentage of green ECTs and firms by WIPO sector, 2012-2021 A. ECTs 100 80 % of green ECs by technology sector 60 40 20 0 12 13 14 15 16 17 18 19 20 21 20 20 20 20 20 20 20 20 20 20 Construction Energy Farming and Forestry PMP Transportation B. Firms 100 80 % of green firms by technology sector 60 40 20 0 12 13 14 15 16 17 18 19 20 21 20 20 20 20 20 20 20 20 20 20 Construction Energy Farming and Forestry PMP Transportation Notes: Panel A depicts the proportion of ECTs that mention green technologies by WIPO green sector over the period 2012-2021. Panel B depicts the percentage by WIPO green sector over the period 2012-2021. 19 Figure 5: Percentage of ECTs mentioning technologies from WIPO green patents, 2012-2021 Construction Energy Farming and Forestry 2.5 Construction Material 2.5 Wind Turbines 2.5 Transgenic Crops Air Conditioning Li Ion Battery Soilless Agriculture Led Technology Biofuels Water Saving Irrigation 2 Sensors 2 Solar Energy 2 Inverter Technology 1.5 1.5 1.5 % of ECs mentioning green patents 1 1 1 .5 .5 .5 0 0 0 12 13 14 15 16 17 18 19 20 21 12 13 14 15 16 17 18 19 20 21 12 13 14 15 16 17 18 19 20 21 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 PMP Transportation 2.5 Carbon Capture Technology 2.5 Hybridelectric Vehicle Insulating Glass Drones 2 2 1.5 1.5 1 1 .5 .5 0 0 12 13 14 15 16 17 18 19 20 21 12 13 14 15 16 17 18 19 20 21 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 Notes: The figure depicts the share of ECTs that belong to each of the selected technologies within each of WIPO Green sector in the period 2021-2021. 20 Figure 6: Relationship between green technologies in ECTs, 2012-2021 Soilless Agriculture Geothermal Energy Transgenic Crops Water Saving Irrigation CCUS Construction Material LED Technology HEV Insulating Glass Air Conditioning Biofuels Wind Turbines Sensors Li-ion Battery Drones Solar Energy Inverter Technology Notes: The network depicts the relationship between the green technologies identified in the ECTs over the period 2012-2021. The position of the nodes indicates how connected each technology is in the entire network, with technologies in the center being the most mentioned in boardroom discussions. Size of the nodes is proportional to the number of times each technology is mentioned in the ECTs. The closeness between technologies indicates how related they are between them (i.e., mentioned together in the same ECTs). 21 Table 1: Relationship between mentioning green technologies in ECTs and firm performance, 2012-2021 Dep. var.: green firm = 1 (1) (2) Construction log sales 0.021*** (0.005) log operating income 0.018*** (0.004) Energy log sales 0.013*** (0.004) log operating income 0.007** (0.003) Farming and Forestry log sales 0.001 (0.001) log operating income 0.002 (0.001) PMP log sales 0.005*** (0.002) log operating income 0.002 (0.001) Transportation log sales 0.010*** (0.003) log operating income 0.006*** (0.002) Year FE Yes Yes Sector FE Yes Yes Country FE Yes Yes Observations 21,490 19,388 Notes: Table reports OLS estimates on the relationship between firm attributes and the probability of having an ECT mentioning keywords related to green technology adoption by WIPO Green sector. The dependent variable takes the value of 1 in the year the firm mentioned keywords related to a green technology of the WIPO green sector, and in all subsequent years. Standard errors clustered at the firm level in parentheses. Significance levels at the 1, 5, and 10 percent levels are denoted by ***, **, and *, respectively. All models include controls for the country’s GDP per capita (USD). 22 Table 2: Relationship between mentioning green technologies in ECTs and CO2 emissions, 2012-2021 Dep. var.: log Scope 1 emissions All sectors Construction Energy Farming and Forestry PMP Transportation (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Green tech firm 0.052 0.090** 0.083** 0.095** 0.121** 0.218*** 0.026 -0.072 0.002 -0.004 -0.053 -0.014 (0.035) (0.040) (0.041) (0.048) (0.053) (0.069) (0.134) (0.116) (0.078) (0.117) (0.066) (0.077) Green tech firm × High-emission firm -0.210** -0.123 -0.352*** -0.244** -0.087 -0.249* (0.084) (0.085) (0.103) (0.115) (0.158) (0.151) 23 Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Time varying controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 20,523 19,339 20,523 19,339 20,523 19,339 20,523 19,339 20,523 19,339 20,523 19,339 Notes: Table reports OLS estimates on the relationship between having an ECT mentioning keywords related to green technology and CO2 emissions. The variable ”green tech firm” takes the value of 1 in the year the firm mentions keywords related to a green technology and in all subsequent years. High-emission firms had emission intensity (GHG emissions per million USD) in 2012 above the sample mean. Standard errors clustered at the firm level in parentheses. Significance levels at the 1, 5, and 10 percent levels are denoted by ***, **, and *, respectively. Log sales included as time varying control. Table 3: Relationship between mentioning green technologies in ECTs and CO2 emissions, alternative measure of emissions intensity, 2012-2021 Dep. var.: log Scope 1 emissions All sectors Construction Energy Farming and Forestry PMP Transportation (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Green tech firm 0.052 0.095** 0.083** 0.107** 0.121** 0.204*** 0.026 -0.072 0.002 0.015 -0.053 -0.012 (0.035) (0.038) (0.041) (0.045) (0.053) (0.062) (0.134) (0.116) (0.078) (0.105) (0.066) (0.075) Green tech firm × High intensive firms -0.375*** -0.278*** -0.446*** -0.244** -0.182 -0.352** 24 (0.096) (0.101) (0.113) (0.115) (0.153) (0.147) Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Time varying controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 20,523 19,339 20,523 19,339 20,523 19,339 20,523 19,339 20,523 19,339 20,523 19,339 Notes: Table reports OLS estimates on the relationship between having an ECT mentioning keywords related to green technology and CO2 emissions. The variable ”green tech firm” takes the value of 1 in the year the firm mentions keywords related to a green technology and in all subsequent years. High-emission firms had emission intensity (GHG emissions per million USD) in 2012 above the 90th percentile. Standard errors clustered at the firm level in parentheses. Significance levels at the 1, 5, and 10 percent levels are denoted by ***, **, and *, respectively. Log sales included as time varying control. Table 4: Firm linkages and the probability of becoming green, 2012-2021 Dep. var: green firm = 1 (1) (2) (3) (4) (5) (6) (7) (8) Green tech partner 0.047*** 0.022*** (0.003) (0.004) Green tech partner × High-emission firm 0.096*** (0.009) Green customer 0.031*** 0.026*** (0.002) (0.003) Green customer × High-emission firm 0.031*** (0.006) Green supplier 0.029*** 0.018*** (0.002) (0.003) Green supplier × High-emission firm 0.041*** (0.005) Green customer and supplier 0.038*** 0.031*** (0.007) (0.008) Green customer and supplier × High-emission firm 0.034* (0.020) Year FE Yes Yes Yes Yes Yes Yes Yes Yes Firm pair FE Yes Yes Yes Yes Yes Yes Yes Yes Mean dep.var. 0.195 0.196 0.216 0.216 0.199 0.199 0.179 0.180 Observations 155,216 151,287 311,344 297,047 311,344 300,090 34,168 32,677 Notes: Table reports OLS estimates on the relationship between having a linkage with a green firm in year t (i.e. with a firm that has an ECT mentioning keywords related to green technology) and being a green firm in year t. The dummy variables for green firm take the value of 1 in the year the firm mentioned keywords related to a green technology and in all subsequent years. High-emission firms had emission intensity (GHG emissions per million USD) in 2012 above the sample mean. The dataset for these regressions is at the related firm pair-year level. Each related firm pair must have had at least one active relationship (customer, supplier or tech partner) at any point between 2012-2021. Yearly coverage for each related firm pair ranges from 2012 to 2021, conditional on having relationship information of both firms for each year. Standard errors clustered at the firm level in parentheses. Significance levels at the 1, 5, and 10 percent levels are denoted by ***, **, and *, respectively. 25 Table 5: Firm linkages and the probability of becoming green, lagged one year, 2012-2021 Dep. var: green firm = 1 (1) (2) (3) (4) (5) (6) (7) (8) Green tech partner (t-1) 0.037*** 0.022*** (0.004) (0.004) Green tech partner (t-1) × High-emission firm 0.064*** (0.009) Green customer (t-1) 0.016*** 0.014*** (0.002) (0.003) Green customer (t-1) × High-emission firm 0.018*** (0.007) Green supplier (t-1) 0.023*** 0.014*** (0.002) (0.003) Green supplier (t-1) × High-emission firm 0.034*** (0.006) Green customer and supplier (t-1) 0.025*** 0.017** (0.008) (0.008) Green customer and supplier (t-1) × High-emission firm 0.032 (0.021) Year FE Yes Yes Yes Yes Yes Yes Yes Yes Firm pair FE Yes Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 0.197 0.198 0.219 0.219 0.199 0.199 0.180 0.181 Observations 139,654 136,135 280,138 267,298 280,138 270,047 30,736 29,402 Notes: Table reports OLS estimates on the relationship between having a linkage with a green firm in year t − 1 (i.e. with a firm that has an ECT mentioning keywords related to green technology) and being a green firm in year t. The dummy variables for green firm take the value of 1 in the year the firm mentioned keywords related to a green technology and in all subsequent years. High-emission firms had emission intensity (GHG emissions per million USD) in 2012 above the sample mean. The dataset for these regressions is at the related firm pair-year level. Each related firm pair must have had at least one active relationship (customer, supplier, or tech partner) at any point between 2012-2021. Yearly coverage for each related firm pair ranges from 2012 to 2021, conditional on having relationship information of both firms for each year. Standard errors clustered at the firm level in parentheses. Significance levels at the 1, 5, and 10 percent levels denoted by ***, **, and *, respectively. 26 Table 6: Firm linkages and the probability of becoming green, unique linkages, 2012-2021 Dep. var: green firm = 1 (1) (2) (3) (4) (5) (6) (7) (8) Green tech partner 0.049*** 0.014** (0.005) (0.006) Green tech partner × High-emission firm 0.110*** (0.012) Green customer 0.029*** 0.025*** (0.003) (0.003) Green customer × High-emission firm 0.024*** (0.007) Green supplier 0.027*** 0.015*** (0.002) (0.003) Green supplier × High-emission firm 0.042*** (0.006) Green customer and supplier 0.024** 0.036*** (0.011) (0.013) Green customer and supplier × High-emission firm -0.034 (0.026) Year FE Yes Yes Yes Yes Yes Yes Yes Yes Firm pair FE Yes Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 0.206 0.207 0.225 0.224 0.204 0.204 0.184 0.185 Observations 63,044 61,522 240,242 228,234 240,242 231,588 15,864 14,873 Notes: Table reports OLS estimates on the relationship between having a linkage with a green firm in year t (i.e. with a firm that has an ECT mentioning keywords related to green technology) and being a green firm in year t. The dummy variables for green firm take the value of 1 in the year the firm mentioned keywords related to a green technology and in all subsequent years. High-emission firms had emission intensity (GHG emissions per million USD) in 2012 above the sample mean. The dataset for these regressions is at the related firm pair-year level. Each related firm pair must have had at least one active relationship (customer, supplier or tech partner) at any point between 2012-2021. Only firm pairs with a single type of relationship are considered. Yearly coverage for each related firm pair ranges from 2012 to 2021, conditional on having relationship information of both firms for each year. Standard errors clustered at the firm level in parentheses. Significance levels at the 1, 5, and 10 percent levels are denoted by ***, **, and *, respectively. 27 Table 7: Firm linkages and the probability of becoming green, linkages with MNEs, 2012-2021 Dep. var: green firm = 1 (1) (2) (3) (4) (5) (6) (7) (8) Green tech partner 0.046*** 0.023*** (0.003) (0.004) Green tech partner × High-emission firm 0.094*** (0.009) Green customer 0.031*** 0.026*** (0.002) (0.003) Green customer × High-emission firm 0.028*** (0.007) Green supplier 0.029*** 0.018*** (0.002) (0.003) Green supplier × High-emission firm 0.041*** (0.005) Green customer and supplier 0.036*** 0.031*** (0.008) (0.008) Green customer and supplier × High-emission firm 0.025 (0.021) Year FE Yes Yes Yes Yes Yes Yes Yes Yes Firm pair FE Yes Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 0.194 0.195 0.217 0.217 0.199 0.199 0.176 0.177 Observations 151,606 148,057 298,102 285,259 298,292 287,828 33,098 31,697 Notes: Table reports OLS estimates on the relationship between having a linkage with a green firm in year t (i.e. with a firm that has an ECT mentioning keywords related to green technology) and being a green firm in year t. The dummy variables for green firm take the value of 1 in the year the firm mentioned keywords related to a green technology and in all subsequent years. High-emission firms had emission intensity (GHG emissions per million USD) in 2012 above the sample mean. The dataset for these regressions is at the related firm pair-year level. Each related firm pair must have had at least one active relationship (customer, supplier or tech partner) at any point between 2012-2021. Only firm pairs in which the green firm is an MNE are considered. Yearly coverage for each related firm pair ranges from 2012 to 2021, conditional on having relationship information of both firms for each year. Standard errors clustered at the firm level in parentheses. Significance levels at the 1, 5, and 10 percent levels are denoted by ***, **, and *, respectively. 28 Table 8: Firm linkages and the probability of becoming green, from non-MNEs, 2012-2021 Dep. var: green firm = 1 (1) (2) (3) (4) (5) (6) (7) (8) Green tech partner 0.071*** 0.027 (0.024) (0.032) Green tech partner × High-emission firm 0.112** (0.051) Green customer 0.033*** 0.009 (0.010) (0.013) Green customer × High-emission firm 0.071*** (0.024) Green supplier 0.047*** 0.030** (0.011) (0.014) Green supplier × High-emission firm 0.031 (0.023) Green customer and supplier 0.100** 0.077 (0.042) (0.094) Green customer and supplier × High-emission firm 0.027 (0.107) Year FE Yes Yes Yes Yes Yes Yes Yes Yes Firm pair FE Yes Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 0.230 0.229 0.205 0.210 0.208 0.205 0.250 0.258 Observations 3,610 3,230 13,242 11,788 13,052 12,262 1,070 980 Notes: Table reports OLS estimates on the relationship between having a linkage with a green firm in year t (i.e. with a firm that has an ECT mentioning keywords related to green technology) and being a green firm in year t. The dummy variables for green firm take the value of 1 in the year the firm mentioned keywords related to a green technology and in all subsequent years. High-emission firms had emission intensity (GHG emissions per million USD) in 2012 above the sample mean. The dataset for these regressions is at the related firm pair-year level. Each related firm pair must have had at least one active relationship (customer, supplier or tech partner) at any point between 2012-2021. Only firm pairs in which the green firm is not an MNE are considered. Yearly coverage for each related firm pair ranges from 2012 to 2021, conditional on having relationship information of both firms for each year. Standard errors clustered at the firm level in parentheses. Significance levels at the 1, 5, and 10 percent levels are denoted by ***, **, and *, respectively. 29 Table 9: Firm linkages and the probability of becoming green, to MNEs, 2012-2021 Dep. var: green firm = 1 (1) (2) (3) (4) (5) (6) (7) (8) MNE Green tech partner 0.047*** 0.021*** (0.003) (0.004) MNE Green tech partner × High-emission firm 0.099*** (0.009) MNE Green customer 0.031*** 0.025*** (0.002) (0.003) MNE Green customer × High-emission firm 0.030*** (0.007) MNE Green supplier 0.029*** 0.018*** (0.002) (0.003) MNE Green supplier × High-emission firm 0.041*** (0.005) MNE Green customer and supplier 0.036*** 0.030*** (0.008) (0.008) MNE Green customer and supplier × High-emission firm 0.029 (0.020) Year FE Yes Yes Yes Yes Yes Yes Yes Yes Firm pair FE Yes Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 0.196 0.197 0.220 0.220 0.201 0.201 0.179 0.181 Observations 151,606 147,741 298,292 284,715 298,102 287,366 33,098 31,687 Notes: Table reports OLS estimates on the relationship between having a linkage with a green firm in year t (i.e. with a firm that has an ECT mentioning keywords related to green technology) and being a green firm in year t. The dummy variables for green firm take the value of 1 in the year the firm mentioned keywords related to a green technology and in all subsequent years. High-emission firms had emission intensity (GHG emissions per million USD) in 2012 above the sample mean. The dataset for these regressions is at the related firm pair-year level. Each related firm pair must have had at least one active relationship (customer, supplier or tech partner) at any point between 2012-2021. Only firm pairs in which the firm linked to a green firm is an MNE are considered. Yearly coverage for each related firm pair ranges from 2012 to 2021, conditional on having relationship information of both firms for each year. Standard errors clustered at the firm level in parentheses. Significance levels at the 1, 5, and 10 percent levels are denoted by ***, **, and *, respectively. 30 Table 10: Firm linkages and the probability of becoming green, to non-MNEs, 2012-2021 Dep. var: green firm = 1 (1) (2) (3) (4) (5) (6) (7) (8) Non-MNE Green tech partner 0.054*** 0.060*** (0.018) (0.023) Non-MNE Green tech partner × High-emission firm 0.003 (0.038) Non-MNE Green customer 0.044*** 0.035*** (0.011) (0.012) Non-MNE Green customer × High-emission firm 0.046* (0.027) Non-MNE Green supplier 0.033*** 0.018 (0.011) (0.013) Non-MNE Green supplier × High-emission firm 0.045** (0.023) Non-MNE Green customer and supplier 0.078** 0.050 (0.033) (0.034) Non-MNE Green customer and supplier × High-emission firm 0.129 (0.086) Year FE Yes Yes Yes Yes Yes Yes Yes Yes Firm pair FE Yes Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 0.152 0.153 0.139 0.138 0.154 0.153 0.157 0.146 Observations 3,610 3,546 13,052 12,332 13,242 12,724 1,070 990 Notes: Table reports OLS estimates on the relationship between having a linkage with a green firm in year t (i.e. with a firm that has an ECT mentioning keywords related to green technology) and being a green firm in year t. The dummy variables for green firm take the value of 1 in the year the firm mentioned keywords related to a green technology and in all subsequent years. High-emission firms had emission intensity (GHG emissions per million USD) in 2012 above the sample mean. The dataset for these regressions is at the related firm pair-year level. Each related firm pair must have had at least one active relationship (customer, supplier or tech partner) at any point between 2012-2021. Only firm pairs in which the firm linked to a green firm is not an MNE are considered. Yearly coverage for each related firm pair ranges from 2012 to 2021, conditional on having relationship information of both firms for each year. Standard errors clustered at the firm level in parentheses. Significance levels at the 1, 5, and 10 percent levels are denoted by ***, **, and *, respectively. 31 A.1 Appendix Figures and Tables 32 Figure A1: Economic relevance of firms mentioning green technologies in ECTs, 2012-2021 A. Overall 35 % of sales by firms with ECTs mentioning green keywords 30 25 20 15 12 13 14 15 16 17 18 19 20 21 20 20 20 20 20 20 20 20 20 20 B. By WIPO Green sector 15 % of sales by firms with ECTs mentioning green keywords 10 5 0 12 13 14 15 16 17 18 19 20 21 20 20 20 20 20 20 20 20 20 20 Construction Energy Farming and Forestry PMP Transportation Notes: Panel A depicts the share of total sales of firms mentioning green technologies in ECTs over the period 2012-2021. Panel B depicts the share of total sales of firms mentioning green technologies in ECTs over the period 2012-2021 by WIPO Green sector. 33 Figure A2: Percentage of green ECTs and firms by NAICS sector, 2012-2021 A. Distribution of ECTs mentioning at least one green technology across sectors 100 80 % of green ECs by NAICS sector 60 40 20 0 12 13 14 15 16 17 18 19 20 21 20 20 20 20 20 20 20 20 20 20 Information Manufacturing Professional Services Utilities Other Sectors B. ECTs mentioning at least one green technology within each sector 30 Manufacturing Information Professional, Scientific, and Technical Services Utilities % of ECTs mentioning green tech Other Sectors 20 10 0 12 13 14 15 16 17 18 19 20 21 20 20 20 20 20 20 20 20 20 20 Notes: Panel A depicts the proportion of ECTs that mention green technologies by NAICS sector over the period 2012-2021. Panel B depicts the percentage of ECTs that are green within each NAICS sector over the period 2012-2021. 34 Figure A3: Economic relevance of firms mentioning green technologies in ECTs by sector, 2012- 2021 A. By WIPO Green sector 15 % of sales by firms with ECTs mentioning green keywords 10 5 0 12 13 14 15 16 17 18 19 20 21 20 20 20 20 20 20 20 20 20 20 Construction Energy Farming and Forestry PMP Transportation B. By NAICS sector 80 Manufacturing Information % of sales by firms with ECTs mentioning green keywords 70 Professional, Scientific, and Technical Services Utilities Other Sectors 60 50 in each sector 40 30 20 10 0 12 13 14 15 16 17 18 19 20 21 20 20 20 20 20 20 20 20 20 20 Notes: Panel A depicts the share of total sales of firms mentioning green technologies in ECTs by WIPO green sector over the period 2012-2021. Panel B depicts the share of total sales of firms mentioning green technologies in ECTs by NAICS sector over the period 2012-2021. 35 Figure A4: Average sales of firms mentioning green technologies in ECTs versus other firms by NAICs sector, 2012-2021 Accommodation and Food Services Construction Finance and Insurance Manufacturing 20000 8000 10000 20000 8000 15000 6000 15000 6000 4945.28 (*) 10000 4000 3426.27 10000 7894.91 (***) 2516.98 4000 3177.77 (*) 5447.52 5920.77 2000 5000 5000 2000 2473.20 (***) 0 0 0 0 Mean sales (Millions USD) Green Firms Other Firms Green Firms Other Firms Green Firms Other Firms Green Firms Other Firms Mining, Oil and Gas Professional Services Real Estate Retail Trade 30000 8000 3000 40000 6000 30000 20000 2000 14187.96 (***) 4000 20000 18153.31 (***) 1279.69 (**) 2980.41 (**) 10000 1000 807.66 (**) 2000 1604.81 (**) 10000 6139.00 (***) 1141.92 (***) 0 0 0 0 Green Firms Other Firms Green Firms Other Firms Green Firms Other Firms Green Firms Other Firms Transportation and Warehousing Utilities Waste Management Wholesale Trade 15000 20000 8000 40000 15000 6000 30000 10000 4976.86 (***) 7488.16 (***) 10000 4000 20000 8147.48 (**) 15351.19 (**) 5000 2005.18 (***) 2887.83 (***) 5000 3599.71 (**) 2000 10000 7019.37 (**) 0 0 0 0 Green Firms Other Firms Green Firms Other Firms Green Firms Other Firms Green Firms Other Firms Notes: Figure depicts average sales of firms mentioning green technologies in ECTs versus the average sales of other firms by NAICS sector in the period 2012-2021 36 Figure A5: Geographical distribution of firms in 2012-2017 and 2018-2021, full sample A. 2012-2017 B. 2018-2021 Notes: Panel A depicts the geographical distribution of all firms per capita in the period 2012-2017. Panel B depicts the geographical distribution of all firms per capita in the period 2018-2021. 37 Figure A6: Geographical distribution of green firms in 2012-2017 and 2018-2021 A. 2012-2017 B. 2018-2021 Notes: Panel A depicts the geographical distribution of green firms per capita in the period 2012-2017. Panel B depicts the geographical distribution of green firms per capita in the period 2018-2021. Size of the circles are proportional to the average level of sales for the firm in the corresponding period. 38 Table A1: List of bigrams searched by technology Technology Bigrams Air Conditioning air condition, air conditioner, air conditioning, air cooler, conditioning compressor, conditioning heating, conditioning hvac, conditioning industrial, conditioning sys- tem, conditioning ventilation, efficient hvac, fridge air, heat system, hvac heating, hvac system, power conditioner, refrigeration system Biofuels material biodiesel, biodiesel also, biodiesel blend, biodiesel feedstock, biodiesel glyc- erin, biodiesel plant, biodiesel process, biodiesel product, biodiesel using, biofuel bioethanol, biofuels biodiesel, biofuels ethanol, ethanol biofuel, generation biofuel, produce biodiesel, produce biofuel, producing biofuels, use biodiesel, advanced bio- fuels, ddgs distiller, distiller dried, distiller grain, dried distiller, dried grain, dry distiller, grain ddg, solubles ddgs Carbon Capture capture carbon, capture co, capture sequestration, capture storage, captured co, Technology (CCT) capturing co, carbon capture, carbon sequestration, carbon storage, co capture, co captured, co could, co sequestered, co sequestration, co storage, combustion cap- ture, dioxide sequestration, extracting co, mineral carbonation, recovered co, se- quester carbon, sequester co, sequestered carbon, sequestering carbon, sequestering co, sequestration carbon, sequestration co, storage co Construction Mate- alkali metal, alternative cementitious, arc furnace, ash basin, ash cloud, ash coal, ash rial concrete, ash disposal, ash fly, ash pond, ash remediation, ash silica, blast furnace, calcium carbonate, carbon dioxide, carbonate talc, cement concrete, cement opc, cement slurry, coal ash, concrete cement, electric arc, fly ash, furnace blast, furnace slag, hydraulic binder, hydraulic cement, magnesium carbonate, micro silica, ordi- nary portland, readymix concrete, recycling coal, sand fly, silica fume, silica sand, slag cement, slag fly, supplementary cementitious, volcanic ash Drones aerial drone, aerial vehicle, aircraft drone, uav unmanned, uavs drone, unmanned aerial, vehicle drone, vehicle uavs, vehicle unmanned Geothermal Energy cycle orc, efficiency geothermal, geothermal energy, heat geothermal, system egs Hybrid/Electric Ve- automobile hybrid, autonomous electric, bev hybrid, charger obc, charging ev, cng hicle hybrid, electric hybrid, ev charger, ev charging, ev electric, ev hybrid, ev motor, hev hybrid, hev vehicle, hv hybrid, hybrid car, hybrid electrical, hybrid ev, hybrid vehicle, mild hybrid, motor hybrid, phev electric, phev vehicle, plug hybrid, plugin hybrid, pure ev, suv electric, suv hybrid, transmission hybrid, v hybrid, vehicle bevs, vehicle charger, vehicle ev, vehicle hybrid, vehicle robot Insulating Glass insulated glass, strengthened glass, vacuum insulated Inverter Technology bidirectional inverter, energy inverter, igbt device, inverter battery, inverter con- verter, micro inverter, module inverter, mosfet igbt, pfc controller, svc static 39 Led Technology bulb led, connected led, controlling led, converter led, current leds, diode led, diode semiconductor, emitting diode, led application, led bulb, led chip, led converter, led device, led diode, led display, led electronics, led fixture, led fluorescent, led illumi- nation, led lamp, led light, led lighting, led luminaires, led module, led replacement, led retrofit, led semiconductor, led string, led tube, led tv, light led, lighting led, lighting module, organic leds, retrofit bulb, retrofit lighting, semiconductor led Li Ion Battery battery lithium, ion cell, li battery, lithium battery, lithium ion Sensors daylight sensor, sensor gesture, sensor led, sensor lighting, temperature sensor, ther- mal sensor Soilless Agriculture farming vertical, grow pod, horticulture lighting, hydroponic farming, hydroponic gardening, hydroponic greenhouse, hydroponics system, use hydroponic, vertical farm, vertical farming, vertical garden, vertical grow Solar Energy cell pv, concentrated photovoltaic, cpv module, cpv system, cpv technology, csp con- centrated, csp plant, csp technology, generation photovoltaic, inverter solar, module photovoltaic, module pv, module solar, photovoltaic laminate, photovoltaic pv, pho- tovoltaic receiver, photovoltaic string, photovoltaics cpv, power photovoltaic, pv array, pv module, pv panel, pv solar, roof photovoltaic, system photovoltaic, tech- nology photovoltaic, thinfilm photovoltaic, tower generator Transgenic Crops biotech trait, codon optimized, crop trait, gene editing, gene modification, gene trait, genome dna, genome editing, interfering rna, rna interference, rna sirnas, rnai process, silencing complex, targeted genome, trait crop Water Saving Irriga- agricultural irrigation, irrigation pump, irrigation technology tion Wind Turbines blade nacelle, component nacelle, energy blade, generator blade, nacelle blade, na- celle generator, nacelle hub, nacelle tower, system nacelle, turbine nacelle, turbine offshore, wind blade, wind energy, wind turbine 40 Table A2: Number of Earnings Call Transcripts associated with each green technology belonging to each WIPO Green sector in the period 2012-2021 WIPO Green sector Technology ECTs Construction Construction Material 2,225 Construction Air Conditioning 2,215 Energy Wind Turbines 1,918 Construction LED Technology 1,655 Transportation Hybrid/Electric Vehicle 1,440 Energy Li-ion Battery 1,058 PMP Carbon Capture Technology 1,011 Farming and Forestry Transgenic Crops 404 Energy Biofuels 337 Energy Solar Energy 300 Transportation Drones 183 Construction Sensors 160 Energy Inverter Technology 107 Farming and Forestry Soilless Agriculture 87 Energy Geothermal Energy 61 PMP Insulating Glass 53 Farming and Forestry Water Saving Irrigation 49 41 Table A3: Green firms per NAICS sector, 2012-2021 NAICS Sector Number of green firms Accommodation and Food Services 39 Administrative and Support and Waste Management and Remediation Services 47 Construction 75 Finance and Insurance 171 Information 142 Manufacturing 1271 Mining, Quarrying, and Oil and Gas Extraction 202 Professional, Scientific, and Technical Services 122 Real Estate and Rental and Leasing 138 Retail Trade 124 Transportation and Warehousing 135 Utilities 201 Wholesale Trade 53 Notes: The table presents the number of green firms (that mentioned at least one green technology in ECTs) for each NAICS sector from 2012 to 2021 and the percentage of green firms (that mentioned at least one technology in ECTs) in each WIPO green sector. 42