Beyond resource use: XGBoost reveals unexpected drivers of corporate CO2 emissions

Clément, Alexandre; Robinot, Élisabeth; Trespeuch, Léo Article Beyond resource use: XGBoost reveals unexpected drivers of corporate CO2 emissions Contemporary Economics Provided in Cooperation with: VIZJA University, Warsaw Suggested Citation: Clément, Alexandre; Robinot, Élisabeth; Trespeuch, Léo (2025) : Beyond resource use: XGBoost reveals unexpected drivers of corporate CO2 emissions, Contemporary Economics, ISSN 2300-8814, VIZJA University, Warsaw, Vol. 19, Iss. 2, pp. 163-185, https://doi.org/10.5709/ce.1897-9254.560 This Version is available at: https://hdl.handle.net/10419/323489 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ www.ce.vizja.pl 163 This work is licensed under a Creative Commons Attribution 4.0 International License. This study aimed to determine the most relevant variables for carbon dioxide (CO2) intensity using an XGBoost artificial intelligence (AI) model's feature selection. From more than 600 variables used to assess businesses using extra-financial data and environmental, social, and governance (ESG) scores, the AI algorithm was used to identify the most influential variables for CO2 emissions. The study also explored theories to explain the relationships between the variables and conducted tests based on sectors of activities and geographic locations to display specific disparities. For researchers, the results reveal an interesting relationship between ESG information and CO2 emissions in business. For managers, they confirm assumptions about CO2 emissions and provide new insights into business emissions. This study found that energy, water, and renewable energy consumption significantly impact businesses’ CO2 emissions. Additionally, the proportion of women in the workforce was found to interact with a business's CO2 emissions in most tests, whereas corporate philanthropy showed a weaker relationship. 1. Introduction1. Introduction The effects of climate change are becoming more noticeable as temperatures increase, forests continue to burn at an alarming rate, and floods occur more frequently. Unfortunately, experts have warned that these changes will have a negative impact on our lives and will continue to worsen (Masson-Delmotte & IPCC, 2021). Human activities, notably the burning of fossil fuels, have caused an alarming increase in the concentration of CO2 (carbon dioxide) in the atmosphere (Dong et al., 2019). Excess CO2 in the air is harmful to the environment because of its heat-trapping ability and long-lasting effects (Bazzaz, 1990). Once CO2 is released into the atmosphere, it can remain there for hundreds or even thousands of years, implying that current emissions will continue to have a significant impact far into the future. Evidence shows that CO2 emissions are the primary driver of anthropogenic climate change, and unlike other metrics related to climate, CO2 intensity shows a direct and unambiguous connection between global warming and business as its key contributor, making it a clear and reliable measure (Masson-Delmotte & IPCC, 2021). The 26th United Nations Climate Change Conference, also known as COP26, took place in Glasgow in 2021; the conference was focused on climate change mitigation. It led to the creation of an international agreement called the Climate Pact, which sets ambitious goals for achieving net-zero emissions and encompasses approximately 90% of global emissions. Regarded as a significant step, the Climate Pact aims to cap the global temperature increase at 1.5°C through CO2 reduction (Lennan Beyond Resource Use: XGBoost Reveals Unexpected Drivers of Corporate CO2 Emissions ABSTRACT Q52, Q56, M14, O44. KEY WORDS: JEL Classification: Corporate Social Responsibility (CSR), CO2 Emissions, ESG Scores, Women’s Influence, XGBoost. 1 École des Sciences de Gestion de l’Université du Québec à Montréal (ESG UQAM) 2 École de Gestion de l’Université du Québec à Trois-Rivières (ESG UQTR) Correspondence concerning this article should be addressed to: Alexandre Clement, École des Sciences de Gestion de l’Université du Québec à Montréal (ESG UQAM) 320, rue Sainte-Catherine Est, Montréal, QC H2X 3X2, Canada. E-mail: [email protected] Alexandre Clement1 , Élisabeth Robinot1 , and Léo Trespeuch2 , Primary submission: 13.02.2025 | Final acceptance: 08.03.2025 164 Alexandre Clement, Élisabeth Robinot, Léo Trespeuch, 10.5709/ce.1897-9254.560DOI: CONTEMPORARY ECONOMICS Vol. 19 Issue 2 163-1852025 & Morgera, 2022). However, its success hinges on the commitment of participating nations, particularly major corporations in developed countries and the world's largest polluters. CO2 emissions are directly linked to climate change and represent a crucial metric that must be addressed to prevent catastrophic events and protect the planet. People can contribute positively in addressing climate change in different ways, such as lifestyle changes (Balderjahn et al., 2020; Gallo et al., 2023), consumption reduction (e.g., reducing electricity consumption (Van Den Broek et al., 2019), traveling less frequently (Kantenbacher et al., 2017), reducing waste production (Aschemann-Witzel et al., 2019)), product substitution (e.g., meat substitution (Lemken et al., 2019)), and the adoption of innovative technologies (e.g., electric cars (Thøgersen & Ebsen, 2019)). Although some solutions may have a greater impact than others, people have diverse options for addressing climate change (Thøgersen, 2021). In the near future, businesses that are underperforming with regard to their CO2 reduction goals will likely begin to lose customers because of their emissions. Thus, studying the multiple factors linked to the CO2 emissions of a business is crucial. CO2 emissions are important in the context of climate change, and businesses require a measurement metric to address climate change. CO2 has emerged as “the” metric businesses need to focus on to do their part in the decarbonization of the economy. Environmental, social, and governance (ESG) is a general term associated with quantitative metrics aligned with these three factors. In general, ESG scores are developed to evaluate a business's environmental impact. However, these scores have faced criticisms and have some limitations (Clément et al., 2022a, 2022b; Escrig-Olmedo et al., 2019). One limitation is that they try to capture various broad concepts related to corporate social responsibility (CSR), not just one concept such as climate. CSR involves integrating environmental and social concerns into a business model. The difference between ESG and CSR is that the former is an explicit measurement of environmental performance, whereas the latter is a management philosophy. Thus far, limited research has examined quantitative ESG metrics, their integration into CSR performance, and the relationships between metrics that can influence the reduction and measurement of CO2. Against this background, the current study aims to evaluate the ESG metrics that can influence CO2 emissions by businesses and support their emission reduction goals. The study is conducted using an extreme gradient boosting (XGBoost) artificial intelligence (AI) model that facilitates the assessment of the most important variables for CO2 emissions in business. 2.2. Theoretical Framework2.2. Theoretical Framework The theoretical framework includes a literature review that outlines the relationship between CSR and the use of CO2 intensity to measure a business's environmental performance. This is followed by a review of the XGBoost AI model, the database and index used, and three evaluations that were conducted—CO2 intensity, CO2 intensity by sector of activity, and CO2 intensity segmented by continents— with the aim of determining the most relevant variables for CO2 intensity using the XGBoost AI model's feature selection. 2.1. Relationship Between CSR and ESG Because business activity is an important driver of climate change, a vast body of literature has focused on business environmental impacts and sustainability in recent years (Ahmad et al., 2024). Two approaches can be used to assess business environmental impact and sustainability. The first is a qualitative or philosophical perspective that examines business environmental impact through the lens of CSR practices. This approach allows for a nuanced understanding of a company's ethical considerations and voluntary actions toward environmental stewardship. The second perspective employs a more quantitative method, utilizing ESG scores to assess and compare CSR practices across businesses numerically. These ESG metrics provide a standardized framework for evaluating a company's environmental performance, among other factors (Kaźmierczak, 2022). Importantly, these two approaches are not mutually exclusive, but rather complementary (Nugroho et al., 2024), offering a comprehensive view of a business's environmental stance and actions. www.ce.vizja.pl 165 Beyond Resource Use: XGBoost Reveals Unexpected Drivers of Corporate CO2 Emissions This work is licensed under a Creative Commons Attribution 4.0 International License. CSR is a management approach that incorporates social and environmental concerns into a company's operations and stakeholder relationships, which is distinct from the more quantitative ESG that is based on numbers. Specific to CSR, Igalens & Joras (2006) demonstrated that companies prioritizing CSR enjoy numerous benefits, including the development of a positive brand image (Ramesh et al., 2019), improved employee recruitment and retention (Supanti et al., 2015), and reduced tax and duty costs (Sprinkle & Maines, 2010). CSR is essential to a brand as it has the potential to significantly impact consumer behavior and brand equity (Kitchin, 2003). Studies have also shown that at least half of the consumers survey are willing to pay more for products that are sustainable and environmentally friendly (Y.-S. Chen, 2010). This indicates that good CSR practices can improve consumer loyalty and the perceived value of brand equity(Mahmood & Bashir, 2020). Moreover, today's consumers are willing to pay more and look for brands that do more than just pursue profits (Dangelico & Vocalelli, 2017). According to van Doorn et al. (2017), good CSR practices can positively influence consumer behavior, which is especially important for innovative companies with limited marketing budgets. In addition to improving brand reputation and consumer trust, CSR practices can help companies recruit and retain employees (S. Du et al., 2007; Islam et al., 2021; Mahmood & Bashir, 2020) as well as retain customers (Nugroho et al., 2024). By engaging in socially responsible practices, companies can create a more motivated and engaged workforce that is more invested in the organization's development. These benefits are further reinforced by consumers and governments preferring and supporting companies that perform well on environmental and social issues. While it is commercially advantageous to practice CSR to improve the performance of business operations (Peratz & Strønen, 2024) or marketing, its positive impact on climate change is less evident. Moreover, businesses associated with high greenhouse gas emissions or resistance to environmental regulations may suffer reputational damage, leading to decreased customer loyalty and even boycotts (Bansal & DesJardine, 2014; Kotsantonis & Serafeim, 2019). This suggests that the promotion of the advantageous elements of CSR practices can be diminished solely by poor CO2 performance. Businesses need to balance broad environmental and social initiatives while reducing the risk of being accused of greenwashing, by ensuring that they clearly understand the impact of business actions on the planet. ESG scores are sometimes used to quantitatively express CSR practices, as CSR practices are broad and difficult to measure. The objective of ESG scores is to provide a type of metric that allows the assessment of businesses' environmental sustainability and ultimately allows comparability. However, a growing body of literature employing various methodologies and datasets demonstrates that ESG scores may not be a comprehensive representation of a company's sustainability(Eccles et al., 2020; Eccles & Stroehle, 2018; Escrig-Olmedo et al., 2019; Gillan et al., 2021; Kotsantonis & Serafeim, 2019; Olmedo et al., 2010; Widyawati, 2020). These scores need to be more accurate in depicting organizations' beneficial impacts on the planet (Bernier-Monzon et al., 2019). Instead, the design of ESG scores is such that they primarily gauge the associated risks emanating from ESG factors, thereby minimizing the negative repercussions of a company's actions on its investment prospects (Berg et al., 2022; Bernier-Monzon et al., 2019; Cornell & Damodaran, 2020; Maiti, 2021; Taparia, 2021). Therefore, ESG scores are the predominant tool that investors use to assess a company’s environmental dimensions (Widyawati, 2020). Additionally, researchers have increasingly utilized ESG scores as a standard yardstick for comparing sustainability performance across corporations (Clément et al., 2022b). CSR and ESG scores are similar in their encompassing various elements, such as child labor policies, employee-customer relations, and global supply chain practices. They are both broad measures that track different subjects. They are not designed to track specific issues such as the increase in global CO2 emissions or business impacts on climate change. “The role of CSR in combating some of the world's most urgent environmental problems, such as greenhouse gas emissions, loss of biodiversity, etc., is even more vague” (Kudłak, 2019, p. 169). To combat climate change and assess the impact of organizations on the environment, an effective approach would be to focus on a single metric incorporated in companies' 166 Alexandre Clement, Élisabeth Robinot, Léo Trespeuch, 10.5709/ce.1897-9254.560DOI: CONTEMPORARY ECONOMICS Vol. 19 Issue 2 163-1852025 CSR practices and ESG scores, such as CO2 emissions. Reporting on CSR and being transparent in ESG assessments are effective ways for businesses to demonstrate to their customers and the public that they are taking action toward mitigating climate change and other social challenges. However, the tangible impact of such reporting on a global scale is still debatable (Walls et al., 2012). Yet, reporting on CO2 emissions can help companies show their progress toward meeting international climate commitments, such as those outlined in the Paris Agreement. CO2 and CSR practices are related; in fact, companies that prioritize CSR within their organizations, mainly through self-regulatory measures, tend to have lower emissions than their peers (Kudłak, 2019). The same applies to those maintaining transparency in their ESG assessments (Chouaibi & Affes, 2021). However, ESG scores have several limitations. ESG rating agencies have their own methodologies, which are relatively opaque, and the CSR practices they measure for a company can vary greatly from one agency to another (Kotsantonis & Serafeim, 2019; Olmedo et al., 2010; Serafeim et al., 2019). Prior research by Berg et al. (2022) identified a modest correlation among ESG scores from different agencies, a notable finding given the near-perfect correlation found between credit ratings such as Moody’s and Standard & Poor's. For instance, a 38% correlation was observed between MSCI and Refinitiv, and a 70% correlation between Sustainalytics and Vigeo Eiris, with an industry average of 60% correlation (Berg et al., 2022). Other previous studies have corroborated this discrepancy (Dimson et al., 2020. The issuing agency's objectives and target clientele critically influence the comprehensiveness and range of the ESG scores produced (Eccles et al., 2020). Moreover, the depth of qualitative data can fluctuate based on the geographic location of the companies being evaluated, thereby affecting the representativeness of the scores (Baldini et al., 2018; Daugaard & Ding, 2022). The voluntary nature of ESG disclosures frequently limits transparency levels, (Hazen, 2021) and the degree of transparency of the evaluated business is critical in determining the final ESG score (Chouaibi & Affes, 2021). Another key issue with CSR and ESG is the absence of standardized metrics or guidelines. This results in companies defining and reporting their CSR activities differently, making it challenging to compare their efforts (Drempetic et al., 2020; Friede et al., 2015; Gillan et al., 2021; Saadaoui & Soobaroyen, 2018). The limitations of ESG and CSR in addressing climate change are a matter of concern because both measures are linked to many variables that are distinct from climate change and are not measured using similar methods (Berg et al., 2022). ESG providers offer a quantitative representation of business CSR policies (Kudłak, 2019), but still fail to directly address climate change. Environmental concerns encompass several issues, not just those related to CO2; however, there is a strong consensus worldwide that CO2 is the main contributor to climate change. CO2 is one of the hundreds of variables that compose ESG and is not explicitly covered by CSR policies. However, because the problems of climate change are directly linked to CO2, shifting from ESG to CO2 to measure the impact of a business on the environment is crucial. 2.2. Carbon Intensity Regardless of the rapid incorporation and broad adoption of CSR and ESG, global emissions have not decreased or stopped; in fact, they have increased (Masson-Delmotte & IPCC, 2021). On the one hand, there is agreement that CO2 is the most important metric to express climate change, and reducing overall emissions is the key to limiting the impact of climate change (Masson-Delmotte & The Intergovernmental Panel on Climate Change [IPCC], 2021). Additionally, it is difficult for a company to know how to achieve CO2 emission reductions based on past environmental innovations (Albitar et al., 2023). As CO2 emissions are a link between human activities and climate change, more research on business sustainability could benefit from focusing on CO2 directly. To address climate change and avoid greenwashing, improving on the ESG metric is not enough. There is strong consensus that CO2 is the main metric that must be assessed by every stakeholder to fight climate change (MassonDelmotte & IPCC, 2021). Similar to the literature on CSR and ESG, CO2 emissions considerations from a business perspective can lead to many managerial advantages. According to Rahman et al.(2021), environmental innovation, emissions reduction, and the effective www.ce.vizja.pl 167 Beyond Resource Use: XGBoost Reveals Unexpected Drivers of Corporate CO2 Emissions This work is licensed under a Creative Commons Attribution 4.0 International License. use of natural resources are the three best areas to focus on for enhancing brand value. Some critics argue that CSR initiatives or ESG performance are often implemented as a public relations strategy rather than as a genuine commitment to social or environmental issues (Turker, 2009; Weber, 2008). Research indicates that lowering carbon emissions benefits the environment and can lead to significant business advantages, including cost savings, enhanced brand reputation, and compliance with regulatory standards (Kolk et al., 2008). Moreover, businesses that actively manage their CO2 emissions and are transparent about them perform better in terms of CO2 intensity relative to their peers (Haque & Ntim, 2022). Finally, lower carbon emissions through innovation generally result in better financial performance for a firm and better return on equity (Albitar et al., 2023; Gallego Alvarez, 2012). Focusing on CO2 emission instead of CSR or ESG could lessen the global assessment of business impacts. Nonetheless, it presents a more straightforward and transparent approach to addressing the urgent issue of mitigating climate change. Moreover, customers are becoming averse to greenwashing, and CSR and ESG are increasingly facing this critique. For comparability, a standardized version of carbon emissions from businesses is expressed as carbon intensity. Carbon intensity is the amount of CO2 emitted per unit of economic activity or energy and is used to help identify the sectors and activities that are most carbon-intensive (Q. Wang et al., 2017). Carbon emissions vary greatly among different industries, which is why carbon intensity provides a more comprehensive view of a company's environmental impact. The CO2 intensity metric considers not only the total amount of emissions but also the efficiency with which energy is used. For instance, two companies may have the same total emissions. However, if one produces more output, its carbon intensity will be lower. Therefore, carbon intensity provides a more nuanced and accurate picture of a company's environmental and economic performance. It also allows meaningful comparisons and benchmarking across different businesses, sectors, and countries (Canadell et al., 2007). Lower carbon intensity can serve as a competitive advantage for businesses, as it demonstrates their commitment to sustainability. As companies grow and evolve, their total emissions may also change. Tracking carbon intensity can help determine whether a company is becoming efficient, and whether its growth is sustainable from a climate perspective (Burgess et al., 2020). CO2 intensity is valuable for developing mitigation strategies and can be incorporated into broader CSR plans (P. Chen, 2023). Le Quéré et al. (2015) highlighted the complexities involved in accurately measuring carbon intensity, including the need for standardized methodologies and the challenges of comparability across different industries. Despite these challenges, carbon intensity has become the primary metric used to evaluate businesses in the finance sector (Berg et al., 2022). Traditionally, the energy (generation or energyintensive business) and transportation sectors have high carbon intensities (Ritchie et al., 2023). The energy production sector is one of the largest carbon emitters, because it relies heavily on fossil fuels (Burgess et al., 2020; Masson-Delmotte & IPCC, 2021). When these fuels are burned to generate electricity or heat, they release significant CO2 into the atmosphere. The transportation sector is another major contributor, mainly because of its use of nonrenewable resources. Industrial processes also have a high carbon footprint, particularly in the steel, cement, and chemical production industries, where fossil fuels are used for energy and manufacturing (Murshed et al., 2021). Yet, detecting high CO2 intensity levels may offer opportunities for companies to improve their efficiency, explore cleaner energy sources, and innovate in processes and technologies. Unlike other metrics, CO2 intensity offers a direct and unambiguous connection with key contributors to global warming, making it a clear and reliable measure for businesses and consumers (N. Stern, 2007, 2014). Assuming that CO2 emissions are the most important metric related to climate change, the objective of this study is to evaluate the drivers of business linked to its CO2 emissions. ESG scores 168 Alexandre Clement, Élisabeth Robinot, Léo Trespeuch, 10.5709/ce.1897-9254.560DOI: CONTEMPORARY ECONOMICS Vol. 19 Issue 2 163-1852025 are composed of more than 600 variables, known as extra financial variables. “Extra-financial variables” is a broad term that describes all the information regarding a business that is not financial in nature. The ESG score is derived from an aggregation of these non-financial variables, which evaluate many CSR aspects. CSR practices such as business values and management styles are sometimes difficult to quantify mathematically. Therefore, assessing extra-financial variables with CO2 emissions could point to interesting avenues for enhancing CSR management in business, with the ultimate objective of CO2 reduction. The primary metric chosen is CO2 intensity because of its direct link to climate change and comparability across industries. Unlike ESG scores, CO2 intensity provides a clear, quantifiable measure of a company's environmental impact relative to its economic output. This metric allows for meaningful cross-sector analysis, useful for diverse stakeholders to measure their impacts on the environment, improve their performance, and reduce their negative impacts. In essence, CO2 intensity acts as a universal measure for a universal problem. 2.3. Hypothesis Development Evidence shows that how corporations use resources is a major and immediate factor driving CO₂ emissions (Murshed et al., 2021). Companies that rely heavily on fossil fuels for energy inevitably produce greater greenhouse gas emissions, owing to the carbon-heavy nature of these sources. Similarly, high water usage often aligns with broader resource management inefficiencies that can increase a company’s carbon footprint, particularly regarding the energy required for water treatment, transportation, and heating. Meanwhile, businesses that proactively optimize or curtail resource use through conservation strategies, sustainable innovations, and renewable energy integration can significantly reduce their emission intensities. Collectively, these findings indicate that visible signs of resource inefficiency— high energy consumption, limited use of renewable energy, and excessive water usage—are usually associated with elevated CO₂ intensity levels, highlighting resource management as a crucial area for tackling climate change (Murshed et al., 2021). H1. Greater resource usage is associated with higher CO2 intensity. Research has consistently highlighted that philanthropic actions, often viewed as “good business citizenship,” reflect a company's commitment to social welfare (Kudłak, 2019; Wang et al., 2008). By voluntarily channeling resources into charitable causes, organizations demonstrate their intention to exceed basic legal and commercial commitments. This proactive approach often encompasses environmental efforts; if a company’s leaders are ready to invest in projects that benefit communities, they are also more likely to implement strategies that reduce their ecological impact. Engaging in philanthropic endeavors can help build stronger stakeholder networks, creating both social capital and reputational advantages that further promote environmentally responsible behavior. Consequently, initiatives such as donating a portion of revenue may indicate a broader, value-driven corporate culture that emphasizes the well-being of both social and ecological systems. H2. Companies that actively engage in philanthropy or charitable giving are more likely to exhibit lower CO₂ intensity, suggesting a positive spillover from social responsibility to environmental stewardship. Previous studies have shown that having women in top management positions positively affects a company's environmental performance. Research indicates that increased female representation in executive positions enhances stakeholder-oriented decision making (Bernardi & Threadgill, 2010), raises awareness of sustainability issues (SetóPamies, 2015), and cultivates a social responsibility culture (del Carmen Valls Martínez et al., 2022; Mehmood, 2022). These claims stem from broader evidence that suggesting that female leaders often demonstrate higher ethical standards and adopt more comprehensive governance approaches. In theory, these attributes should lead to tangible environmental results such as effective CO₂ reduction strategies. H3. Executive gender diversity does not, by itself, exert a significant direct effect on corporate CO₂ intensity. www.ce.vizja.pl 169 Beyond Resource Use: XGBoost Reveals Unexpected Drivers of Corporate CO2 Emissions This work is licensed under a Creative Commons Attribution 4.0 International License. Prior research consistently points to women’s heightened sensitivity to a broad spectrum of stakeholder interests and their greater propensity to integrate long-term social and environmental objectives into organizational strategies (Brammer et al., 2007). This pattern appears to reflect a stronger inclination toward balancing diverse stakeholder demands rather than pursuing narrow, profit-driven goals. Such stakeholderoriented approaches may, in turn, promote more comprehensive climate initiatives, given that effective CO₂-reduction strategies often entail cross-departmental collaboration, transparent communication with external parties, and a readiness to invest in potentially costly but beneficial long-term environmental measures (Glass et al., 2015). Building on this perspective, scholars have argued that women in the workforce are less likely to favor short-term earnings maximization at the expense of sustainable growth, often advocating policies that consider both ecological constraints and community well-being (Tronto, 2022). H4. Higher gender diversity in the workforce is associated with lower CO2 intensity. 3. Methodology 3. Methodology 3.1. Refinitiv dataset This study explores the connection between extra-financial information comprising ESG scores in the Refinitiv database. It proposes a hierarchy based on the importance of each variable to CO2 intensity. The Refinitiv database provides information about businesses, including aspects such as workforce diversity, management practices, supply chain, societal impact, and resource management. The Refinitiv dataset provides a comprehensive overview of ESG factors, each of which can influence a company’s CO₂ intensity in different ways. The “environmental” dimension evaluates the effect of businesses on natural ecosystems and resources. The key metrics encompass greenhouse gas emissions, energy consumption and efficiency, waste management strategies, water usage, and biodiversity. The “social” dimension focuses on an organization's engagement with its employees, suppliers, customers, and local communities. Important indicators include workforce diversity and inclusion, labor practices, and community involvement initiatives. Governance emphasizes effective leadership, board composition and independence, shareholder rights, and corporate transparency. Metrics indicating well-organized boards, ethical executive pay, strong risk management, and anti-corruption measures are essential for promoting climate-related initiatives. Strong governance bodies can implement environmental policies that align with business objectives. Simultaneously, motivated employees and strong community relationships can support the necessary operational changes to reduce emissions. Thus, each pillar has the power to influence both the direct and indirect factors affecting carbon intensity. Refinitiv’s dataset, which includes hundreds of variables across these pillars, supports rigorous cross-company comparisons. By applying a systematic selection or regularization approach to identify the most relevant ESG metrics, researchers can more accurately pinpoint the factors that significantly affect CO₂ intensity. In a broader context, understanding these key variables helps investors and stakeholders determine which companies are positioning themselves to meet evolving climate challenges, manage regulatory risks, and potentially create longterm shareholder value through more sustainable operations. The universe used for the analysis was the MSCI ACWI. This index includes a broad coverage of listed equity companies in both industrial nations and emerging markets. It is composed of 2802 companies from all industrial sectors, including retail, energy, and transport, among others. Therefore, this index allows a broad range of coverage, making it ideal for comparing data from various types of companies. The data were obtained through Refinitiv's Eikon database. Refinitiv calculates ESG scores through an exhaustive evaluation of a company's sustainability efforts and performance. The selection of ESG metrics used a regulation methodology to evaluate the available metrics and identify those to be utilized in the evaluation. Figure 1 illustrates this regularization. All the available metrics in the database were considered. The database consists of 630 variables, mostly Boolean and numerical variables. 170 Alexandre Clement, Élisabeth Robinot, Léo Trespeuch, 10.5709/ce.1897-9254.560DOI: CONTEMPORARY ECONOMICS Vol. 19 Issue 2 163-1852025 3.1.1. Metric Selection From April 20th to May 5th, 2024, the Refinitiv database was used to extract ESG data for MSCI ACWII. All the ESG data points available at the time were considered for this research. The Refinitiv database is a reliable source of information and has been used previously by scholars such as Filippou and Taylor (2021) for their analyses of European firms. During our analysis, we identified 630 variables, of which 4 were categorical, 268 were Boolean, and 358 were numeric. In the first step, variables with low coverage in the dataset were removed. Specifically, 277 variables with more than 50% missing values (NA) were excluded, reducing the NA in the dataset from 703 772 to 19 506 samples. There were 269 variables linked to the environmental pillar, 33 of which had no coverage of any listed companies in the MSCI ACWI universe. Of the remaining variables, 72 were Boolean, 161 were numeric, and 3 were categorical. For the social pillar, 279 variables were available that express the different social characteristics of the companies. Of these, 26 variables offered no coverage of any company present in the MSCI dataset. Of the remaining variables, 151 were numeric, 1 was categorical, and 101 were Boolean. Finally, the governance pillar consisted of 165 variables, of which 24 did not have any coverage for any company present in the MSCI ACWI universe. Of the remaining variables, 46 were numeric and 95 were Boolean. Figure 1 presents the procedure employed to obtain the final dataset. Some variables are unique to a specific industry, while others are closely related to one another. For instance, banks possess data on their assets under management related to ESG, which are not available for other industries. Moreover, there are numerous variables that measure the same metric using different methodologies, such as CO2 emissions tracked under 20 metrics, including CO2 tons to millions of USD and CO2 tons by CARG or EVIC, among others. To address this issue, financial analysts from the socially responsible investment team at an institutional bank collaborated with the authors in two separate meetings, each lasting 1 h, to review all the variables and select those that could be removed while retaining a representative dataset of all industries. Only variables related to X to revenue in a million were retained. All others linked to CARG and EVIC or other financial ratios were removed. The coverage of ESG issues linked to revenue was larger and easier to understand for further analysis. Similarly, all variables specific to an industry, such as “flaring gas,” were also removed. In this step, 75 variables were eliminated. The final dataset was classified based on conceptual congruence. Instead of a formal algorithmic method, a heuristic approach is used to propose a more segmented classification than the three pillars of ESG that are normally proposed. This approach is useful for qualitative problems such as this classification (Douglass & Moustakas, 1985), providing a more comprehensive classification. The first three categories are the same as the ESG score, but an additional category was created, internal social responsibility, which encompasses employee metrics. Table 1 presents the results of the study. The variance inflation factor (VIF) is useful for identifying multicollinearity among variables (Craney & Surles, 2007). The ultimate objective is to obtain a dataset in which none of the features are correlated with each other. Achieving this requires testing various parameters to determine the optimal combination. In this particular case, the aim was to retain as many variables as possible while ensuring that the VIF of each variable remained below 5. Consequently, 60 variables were eliminated. The dataset used in this study contained 218 variables representing ESG issues. These variables were carefully selected to ensure that they were not correlated and specific to any particular industry. Owing to the large number of variables, the dataset was divided into two parts. One part contained all the Boolean variables, whereas the other contained numeric and categorical variables transformed into numeric values. The dataset contained 36 quantitative variables and 185 Boolean variables. As the Boolean dataset did not provide useful information, it was removed from further analysis, and all analyses were conducted based on the numerical dataset. 3.2. Model and Feature Selection Feature selection is the process of selecting the most important variables that can affect the outcome. XGBoost is a model used for feature selec- www.ce.vizja.pl 177 Beyond Resource Use: XGBoost Reveals Unexpected Drivers of Corporate CO2 Emissions This work is licensed under a Creative Commons Attribution 4.0 International License. variables, particularly energy and water consumption, were the most significant predictors of CO2 intensity. The presence of women in the workforce was also highlighted, which was surprising. To explore this further, that is, understand the result of the variable "women employees," a second test was performed to analyze the dataset, but this time by segmenting CO2 intensity by industry sectors. This allowed the identification of sector-specific factors influencing emissions, demonstrating that energy-intensive industries, such as the industrial, information technology, and material industries, showed distinct patterns compared to other sectors. Finally, the third test examined CO2 intensity by geographic location. This test revealed that cultural and regional differences may play a role in the significance of certain ESG topics such as corporate philanthropy and gender diversity. The themes that emerged consistently throughout the tests can be split into categories consisting of resources used, business practices, and human management. The resources used directly affect CO2 emissions (Murshed et al., 2021), and the results of previous tests showed that these variables are highly important for CO2 emission in business. The second theme is corporate philanthropy. Corporate philanthropy can be linked to CO2 emissions because businesses that actively manage their CSR or promote sustainable practices are more inclined to do better in terms of CO2 emissions (Kudłak, 2019). Finally, CO2 emissions affect variables related to human resources, such as hiring women employees and net employment creation. 5.1. Linkage between CO2 Intensity and Resources Used Most sectors and their businesses use resources such as water, electricity, and minerals in the production process. Therefore, the first theme that emerges is resources used, which comprises three important factors regarding the CO2 intensity of a business: energy used, renewable energy used, and water used. The most important group of variables linked to CO2 intensity across industries and worldwide is “energy usage.” Energy used was highest in 7 of 9 industry sectors and 4 of the 5 continents tested. This result was expected, as it aligns with the literature indicating that the energy sector is the largest producer of CO2. Energy is used to power operational and production processes, and can be directly converted into GHG emissions. The same logic applies to the “renewable energy is used” metric. Businesses that use renewable energy will have positive outcomes through lesser CO2 emissions (Murshed et al., 2021). Water used can be seen as unrelated to CO2, but businesses that use more water generally use more resources and thus produce more CO2 (B. Du et al., 2022). Therefore, companies proven to be efficient in terms of water usage are generally more efficient with most of their resources. Technological innovation is the most effective method for reducing water use and businesses that use technological innovation generally have lower CO2 emissions (Du et al., 2022). The variables under the resources used are linked to the amount of activity present in the business and resource-intensive activities, and, as a result, produce more CO2. Resource usage is a crucial variable influencing CO2 levels. Energy consumption remains a key determinant of a company’s carbon intensity, supporting the existing literature that highlights the link between energy use and environmental harm. Thus, H1 (on resource usage) is validated. Consistent with H1, both total energy consumption relative to revenue and water usage relative to revenue were identified as primary predictors, indicating that increased dependence on essential resources is associated with elevated CO2 intensity. 5.2. Linkage Between CO2 Intensity and Corporate Philanthropy Corporate philanthropy can be interpreted as a good business practice that goes beyond regulations and obligations. The variable “total donations to revenue” expresses the good intention of the business resulting from their voluntary action and the firm's intention to remain close to social value. Implementing corporate philanthropic strategies improves a company's environmental performance. Wang et al. (2008) and Trespeuch and Robinot (2023) have shown that up to a certain threshold, corporate philanthropy has a positive impact on a company's performance. Through donations, busi- 178 Alexandre Clement, Élisabeth Robinot, Léo Trespeuch, 10.5709/ce.1897-9254.560DOI: CONTEMPORARY ECONOMICS Vol. 19 Issue 2 163-1852025 nesses express their awareness of an issue and are generally more aware of the other negative impacts they might have. Globally, good business practices and philosophy, improving business relationships with society through donations, and reducing the pay gap between upper management and the workforce show business social value and the ambition to do better than their peers. Business actions toward giving to society or advocating equality can be genuinely undertaken to promote CSR or to remain competitive with their peers (Cha & Rajadhyaksha, 2021). The link between the variables in this category and the CO2 emissions of a business is the result of the good intentions of the business globally, positively affecting CO2 emissions. Hence, corporate philanthropy correlates with CO2 emissions because companies that are aware of their environmental impact also consider their CO2output. Thus, H2 (on corporate philanthropy) is confirmed. Corporate donations in relation to revenue show an important connection with CO2 intensity, and indicate a relationship between these two concepts. This relationship aligns with the notion that philanthropic activities are generally accompanied by lower emissions, although the effect size was less pronounced compared to that for resource intensity validated by H1. 5.3. Linkage Between CO2 Intensity and Human Resources The management of human capital includes human resources within a company and the variables associated with employee satisfaction, development opportunities, diversity, and pay. Multiple internal social responsibility variables are found to have important links with CO2 intensity. Variables such as net employment creation can be explained as proxies for business growth. The hypothesis is that companies experiencing substantial growth, manifested through net employment creation, are likely to increase their operational scale and, consequently, their consumption of resources, thereby impacting their CO2 intensity. Nonetheless, the variable flagged most often in this study was the number of women employees in the workforce. Women employees remained at the top of the list, regardless of the tests conducted. Further tests conducted by industry confirm this trend. The results by country show that this variable was most important in Europe. It was expected that this variable would have a high score in North America and Europe, as both these regions are known to have good programs for women, such as child daycare and maternity leave. Unfortunately, these benefits are not universal in the United States, which could be one of the reasons why this variable did not score as high in the United States as in Europe. The inclusion of women employees as a top influential metric to CO2 intensity was not the initial expectation; the expectation was around another variable named “executive members’ gender diversity.” “Executive members’ gender diversity” tracks the gender diversity in upper management. This variable did not reveal important results, regardless of previous studies (Bernardi & Threadgill, 2010; del Carmen Valls Martínez et al., 2022; Mehmood, 2022; Setó-Pamies, 2015) showing that women in executive positions are more prone to environmental and social considerations of their organization. Previous findings have indicated that having women in top management is linked to a clear enhancement in environmental performance. Yet, this study could not confirm whether increasing diversity within executive teams leads to better carbon intensity management. Consequently, H3 is rejected. This is likely because the statistical significance attributed to female participation in the model is being captured by women employees. The variable of women employees emerged strongly in all the studies conducted in this research. This variable indicates the proportion of women employees in the workforce and remains important, especially in male-dominated industries. Based on the results of Test 2, it was observed that women play a crucial role in traditionally maledominated sectors, including industry, information technology, and materials, for CO2 emissions. This interesting finding can be attributed to the fact that women are naturally inclined to take into consideration the interests of various stakeholders. This tendency can significantly improve the environmental considerations in these industries. Women have been found to be better than men in balancing the www.ce.vizja.pl 179 Beyond Resource Use: XGBoost Reveals Unexpected Drivers of Corporate CO2 Emissions This work is licensed under a Creative Commons Attribution 4.0 International License. performance-driven concerns of shareholders with the diverse interests of other stakeholders such as communities, employees, suppliers, and customers (Brammer et al., 2007; Harrison & Coombs, 2012). The centrality of stakeholder-oriented leadership is key for advancing eco-friendly initiatives, given that corporate sustainability endeavors strive to satisfy both direct stakeholders, such as investors, and indirect stakeholders, such as clients and local communities (Dyllick & Hockerts, 2002). From a philosophical perspective, the ethics of care (Tronto, 2022) can help explain the effect of women employees in the workforce on business CO2 emissions. The ethics of care posits that moral decision-making is not solely a matter of applying abstract principles but involves a nuanced understanding of the specific needs and circumstances of individuals involved. This perspective contrasts with the individualism prevalent in many traditional ethical theories, which often prioritizes autonomy and impartiality. The ethics of care argues that ethical actions arise from an empathetic understanding of others' perspectives and needs, thus fostering a more compassionate and context-sensitive approach to moral dilemmas. This emphasis on care and connectedness shifts the ethical focus from abstract rights and duties to the concrete responsibilities and wellbeing of others. Moreover, the ethics of care (Tronto, 2022) highlights the importance of emotions in ethical deliberation. While traditional ethical theories often regard emotions as potential sources of bias, the ethics of care sees them as essential for understanding and responding to the needs of others. Empathy, compassion, and sensitivity to others' suffering play a crucial role in guiding moral actions and fostering ethical relationships. The role of emotion and purpose reflected in women’s morale could explain the link between CO2 intensity and the workforce. A simple understanding is that, statistically, women care more about others and results that are not monetary than men. It shows not only in their day-to-day ways of living but also their choices. According to this theory, women tend to be more concerned with environmental governance (Glass et al., 2015). Further, in line with the findings of Stern et al.(1993), women generally exhibit greater sensitivity toward the interconnectedness of environmental degradation and individual well-being than men. This observation is further corroborated by research conducted by Andreoni and Vesterlund (2021) regarding the gender dynamics of altruistic behavior. This study revealed that women are more inclined to act altruistically when the cost of such actions is high, whereas men are more prone to altruism when the associated costs are low. Therefore, their participation in organizational decision-making could serve as a catalyst for environmentally conscious policies, ultimately leading to reduced emissions. It can be argued that corporations that demonstrate proactive gender diversity initiatives may be more vigilant in adhering to existing environmental regulations (Lu et al., 2019). As a result, women could choose to work in businesses that are more in line with their environmental values, and businesses with good management of their CO2 may attract more women, especially in male-oriented industries. However, further exploration is needed to establish causality, but this is consistent with previous links established by Konadu et al. (2022) and Streimikiene, (2023), who call for more research on the role of women in low carbon economy transition. Thus, H4 (on women’s workforce representation) is confirmed. A noteworthy finding is the consistent importance of gender composition within the workforce. The percentage of female employees consistently emerged as a key variable influencing CO2 intensity, often surpassing other governance metrics. This indicates that companies with a higher proportion of women tend to exhibit lower carbon intensity, possibly reflecting a broader interest for them to work in organizations that perform well in terms of environmental performance. 5.4. International Social Responsibility: Implications GHG emissions continue to become an important issue as climate change awareness continues to increase. Managers must consider CO2 emissions in addition to CSR practices in the current business climate. This awareness has been amplified by the COVID-19 pandemic (Tosun & Köylüoglu, 2022; Trespeuch et al., 2021), as people have realized the role of human development in the emergence of the pandemic (Barouki et al., 2021). From a manage- 180 Alexandre Clement, Élisabeth Robinot, Léo Trespeuch, 10.5709/ce.1897-9254.560DOI: CONTEMPORARY ECONOMICS Vol. 19 Issue 2 163-1852025 rial perspective, the insight of the current research allows decision makers to understand how various practices can impact CO2 emission intensity. Furthermore, it enables the measurement of practices and ranks in importance the hundred extra-financial variables available, which allows us to focus on a subset of available data highly linked to CO2 emissions. Resources used is the most important group of variables affecting CO2 emissions. This theory is widely accepted, particularly regarding the energy consumption of businesses, which is generally produced from non-reusable sources (Murshed et al., 2021). Workforce gender diversity is another indicator that can be measured and tracked, with a significant relationship between emissions intensity and business impact. Research has shown that women are more interested in working in positive environments and in businesses with socially responsible objectives (Konadu et al., 2022), especially in male-oriented industries. Finally, businesses that engage in corporate philanthropy and, more broadly, promote social value externally through donations or internally by reducing the pay gap are more likely to achieve improved CO2 emissions performance. Corporate philanthropy allows businesses to express their values and shape their social profiles (Polonsky & Jevons, 2009). Reducing CO2 emissions is an additional benefit of the broader inclusion of CSR values in the business DNA. The results demonstrate that resource-use metrics, particularly energy and water consumption, are the most significant predictors of CO2 intensity. This finding is consistent with existing literature, which underscores the role of resource efficiency, especially energy use, in mitigating corporate carbon emissions. The analysis revealed a significant correlation between gender diversity within an organization and its carbon intensity, warranting further exploration to establish causality. This finding implies that companies that prioritize diversity may demonstrate a broader pattern of socially responsible behavior. This also suggests a complex relationship between social factors and environmental performance, indicating that diversity may play a role in promoting sustainable corporate behavior. Finally, firms that actively engage in philanthropic activities or demonstrate a broader commitment to societal well-being also exhibit superior environmental performance, suggesting synergy between socially responsible behavior and sustainability outcomes. This study has some limitations. Feature selection cannot identify missing variables that would better explain the outcome. In addition, variables with the same F-score are equally important for the model. In addition, the F-score method may not consider the combined effects of the variables on the target variable. Therefore, two variables with low individual F-scores may significantly affect the target variable when considered together. Finally, the F-score method needs to be calibrated; although it can rank variables, it does not provide an absolute measure of their importance. It is crucial to exercise caution when interpreting F-scores and using other techniques to gain a more comprehensive understanding of variable importance. Further research is necessary to establish a causal link between the variables before generalizing the findings. Notwithstanding these limitations, the main contribution of this study is that it sheds light on a subset of variables for further exploration and causality research. 6. Conclusion6. Conclusion The objective of this research was to use AI to identify the most important and relevant extra-financial variables linked to CO2 emissions in a business. Through the feature selection process applied to a large dataset encompassing more than 600 extra-financial metrics, this study systematically identified the key drivers of CO2 intensity across diverse sectors and geographical regions. This study offers intriguing insights into the relationship between the CO2 emissions of companies and additional financial data. The most important variables are classified into the following themes: resources used, human resources, and corporate philanthropy. This research provides a new perspective on business CO2 cause and management. Although the feature selection results must be interpreted and used cautiously, this technique makes it possible to find the most important variables for an outcome and rank them by relative importance. This provides a baseline for more research to explore other different variables www.ce.vizja.pl 181 Beyond Resource Use: XGBoost Reveals Unexpected Drivers of Corporate CO2 Emissions This work is licensed under a Creative Commons Attribution 4.0 International License. and the links between them. This would help reinforce the causality link and provide a clear understanding of the importance of these variables. Some links to the resources used and, to a certain extent, those related to corporate philanthropy, are easy to explain; however, there is a lack of literature on those related to the workforce. Nevertheless, this research provides a novel way to analyze the possible links between CO2 and business. For firms, efforts to reduce CO2 emissions should not only focus on optimizing resource use but also on the strategic incorporation of social governance practices, particularly those related to workforce diversity and corporate social engagement. In conclusion, mitigating the adverse impacts of climate change is urgent and requires the identification of meaningful metrics upon which companies can act. Although the CO2 intensity metric is comprehensive, an understanding of its determinants is crucial. This study contributes to addressing this issue by providing a mathematical perspective of the situation through the lens of AI. The results have implications for business practitioners seeking to improve their sustainability efforts and for those seeking to establish more robust environmental governance frameworks. The complex relationship between governance and environmental metrics emphasizes the multifaceted nature of CSR that companies must manage. This suggests that companies with strong governance frameworks are likely to be better equipped to implement successful sustainability initiatives. ReferencesReferences Abdurrahman, G., & Sintawati, M. (2020). Implementation of XGBoost for classification of Parkinson’s disease. Journal of Physics: Conference Series, 1538, 12024. https://doi.org/10.1088/17426596/1538/1/012024 Ahmad, H., Yaqub, M., & Lee, H. S. (2024). Environmental, social, and governance-related factors for business investment and sustainability: A scientometric review of global trends. Environment, Development and Sustainability, 26, 2965–2987. https:// doi.org/10.1007/s10668-023-02921-x Albitar, K., Borgi, H., Khan, M., & Zahra, A. (2023). Business environmental innovation and CO2 emissions: The moderating role of environmental governance. Business Strategy and the Environment, 32(4), 1996–2007. https://doi.org/10.1002/BSE.3232 Aschemann-Witzel, J., Giménez, A., Grønhøj, A., & Ares, G. (2019). Avoiding household food waste, one step at a time: The role of self-efficacy, convenience orientation, and the good provider identity in distinct situational contexts. Journal of Consumer Affairs, 54, 581–606. https://doi. org/10.1111/joca.12291 Balderjahn, I., Seegebarth, B., & Lee, M. S. W. (2020). Less is more! The rationale behind the decisionmaking style of voluntary simplifiers. Journal of Cleaner Production. https://doi.org/10.1016/j. jclepro.2020.124802 Baldini, M., Maso, L. D., Liberatore, G., Mazzi, F., & Terzani, S. (2018). Role of countryand firmlevel determinants in environmental, social, and governance disclosure. Journal of Business Ethics, 150(1), 79–98. https://doi.org/10.1007/s10551016-3139-1 Bansal, P., & DesJardine, M. (2014). Business sustainability: It is about time. Strateg. Organ., 12(1), 70– 78. https://doi.org/10.1177/1476127013520265 Barouki, R., Kogevinas, M., Audouze, K., Belesova, K., Bergman, A., Birnbaum, L., Boekhold, S., Denys, S., Desseille, C., Drakvik, E., Frumkin, H., Garric, J., Destoumieux-Garzon, D., Haines, A., Huss, A., Jensen, G., Karakitsios, S., Klanova, J., Koskela, I.-M., … Vineis, P. (2021). The COVID-19 pandemic and global environmental change: Emerging research needs. Environment International, 146, 106272. https://doi.org/10.1016/j.envint.2020.106272 Bazzaz, F. A. (1990). The response of natural ecosystems to the rising global CO2 levels. Annual Review of Ecology and Systematics, 21, 167–196. https://www.jstor.org/stable/2097022 Becchetti, L., Bobbio, E., Prizia, F., & Semplici, L. (2022). Going deeper into the s of ESG: A relational approach to the definition of social responsibility. Sustainability 2022, Vol. 14, Page 9668, 14(15), 9668. https://doi.org/10.3390/SU14159668 Berg, F., Koelbel, J., & Rigobon, R. (2022). Aggregate confusion: The divergence of ESG ratings. Review of Finance, 26(6), 1315–1344. https://doi. org/10.2139/ssrn.3438533 Bergstra, J., Yamins, D., & Cox, D. D. (2013). Making a science of model search: Hyperparameter optimization in hundreds of dimensions for vision architectures. In Proceedings of the 30th International Conference on Machine Learning (Vol. 28, pp. 115–123). PMLR. https://proceedings.mlr. press/v28/bergstra13.html Bernardi, R. A., & Threadgill, V. H. (2010). Women Di- 182 Alexandre Clement, Élisabeth Robinot, Léo Trespeuch, 10.5709/ce.1897-9254.560DOI: CONTEMPORARY ECONOMICS Vol. 19 Issue 2 163-1852025 rectors and Corporate Social Responsibility. Electronic Journal of Business Ethics and Organization Studies, 15(2). http://ejbo.jyu.fi/ Bernier-Monzon, S., Serafimov, V., & Couteaux, B. (2019). Les entreprises responsables les mieux notées le sont-elles vraiment [Are the highest-rated responsible companies really the best]? Étude de Cas. https://www.sigmagestion.com/wp-content/ uploads/2019/10/etude-de-cas-unilever.pdf Brammer, S., Millington, A., & Pavelin, S. (2007). Gender and ethnic diversity among UK corporate boards. Corporate Governance, 15(2). https://doi. org/10.1111/j.1467-8683.2007.00569.x Burgess, M. G., Ritchie, J., Shapland, J., & Pielke, R. J. (2020). IPCC baseline scenarios have overprojected CO2 emissions and economic growth. Environmental Research Letters. https://doi. org/10.1088/1748-9326/abcdd2 Canadell, J. G., Le Qué Ré, C., Raupach, M. R., Field, C. B., Buitenhuis, E. T., Ciais, P., Conway, T. J., Gillett, N. P., Houghton, R. A., & Marland, G. (2007). Contributions to accelerating atmospheric CO 2 growth from economic activity, carbon intensity, and efficiency of natural sinks. Proceedings of the National Academy of Science. www.pnas.org/cgi/ content/full/ Cha, W., & Rajadhyaksha, U. (2021). What do we know about corporate philanthropy? A review and research directions. Business Ethics, the Environment and Responsibility, 30(3), 262–286. https://doi. org/10.1111/BEER.12341 Chen, P. (2023). Corporate social responsibility, financing constraints, and corporate carbon intensity: New evidence from listed Chinese companies. Environmental Science and Pollution Research, 30(14), 40107–40115. https://doi.org/10.1007/ S11356-023-25176-5/TABLES/6 Chen, T., & Guestrin, C. (2016). XGBoost: A scalable tree boosting system. KDD ’16: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. https:// doi.org/10.1145/2939672.2939785 Chen, Y.-S. (2010). The drivers of green brand equity: Green brand image, green satisfaction, and green trust. Journal of Business Ethics, 93, 307–319. https://doi.org/10.1007/s10551-009-0223-9 Chouaibi, S., & Affes, H. (2021). The effect of social and ethical practices on environmental disclosure: evidence from an international ESG data. Corporate Governance, 21(7), 1293–1317. https://doi. org/10.1108/CG-03-2020-0087 Clément, A., Robinot, É., & Trespeuch, L. (2022a). Improving ESG scores with sustainability concepts. Sustainability, 14(20). https://doi.org/10.3390/ su142013154 Clément, A., Robinot, É., & Trespeuch, L. (2022b). The use of ESG Scores in academic literature, a meta analysis. Academy of Innovation, Entrepreneurship, and Knowledge (ACIEK). Cornell, B., & Damodaran, A. (2020). Valuing ESG: Doing good or sounding good? J. Impact ESG Invest., 1(1), 76–93. https://doi.org/10.3905/ jesg.2020.1.1.076 Craney, T. A., & Surles, J. G. (2007). Model-dependent variance inflation factor cutoff values. Quality engineering,14(3), 391-403. https://doi.org/10.1081/ QEN-120001878 Dangelico, R. M., & Vocalelli, D. (2017). “Green Marketing”: An analysis of definitions, strategy steps, and tools through a systematic review of the literature. Journal of Cleaner Production, 165, 1263–1279. https://doi.org/10.1016/j. jclepro.2017.07.184 Daugaard, D., & Ding, A. (2022). Global drivers for ESG performance: The body of knowledge. Sustainability, 14(4). https://doi.org/10.3390/ su14042322 del Carmen Valls Martínez, M., Manuel Santos-Jaén, J., Soriano Román, R., & Antonio Martín-Cervantes, P. (2022). Are gender and cultural diversities on board related to corporate CO2 emissions? Journal of Cleaner Production, 363, 132638. https:// doi.org/10.1016/j.jclepro.2022.132638 Dimson, E., Marsh, P., & Staunton, M. (2020). Divergent ESG RATING. The Journal of Portfolio Management, 47(1), 75–86. https://doi.org/10.3905/ JPM.2020.1.175 Dong, K., Dong, X., & Dong, C. (2019). Determinants of the global and regional CO 2 emissions: What causes what and where? Applied Economics, 51(46), 5031–5044. https://doi.org/10.1080/00036 846.2019.1606410 Douglass, B. G., & Moustakas, C. (1985). Heuristic inquiry. Http://Dx.Doi.Org.Proxy.Bibliotheques. Uqam.ca/10.1177/0022167885253004, 25(3), 39– 55. https://doi.org/10.1177/0022167885253004 Drempetic, S., Klein, C., & Zwergel, B. (2020). The influence of firm size on the ESG score: Corporate sustainability ratings under review. Journal of Business Ethics, 167(2), 333–360. https://doi. org/10.1007/s10551-019-04164-1 Du, B., Guo, X., Liu, G., Wang, A., Duan, H., & Guo, S. (2022). China’s water intensity factor decomposition and water usage decoupling analysis. Applied www.ce.vizja.pl 183 Beyond Resource Use: XGBoost Reveals Unexpected Drivers of Corporate CO2 Emissions This work is licensed under a Creative Commons Attribution 4.0 International License. Sciences 2022, Vol. 12, Page 7039, 12(14), 7039. https://doi.org/10.3390/APP12147039 Du, S., Bhattacharya, C. B., & Sen, S. (2007). Reaping relational rewards from corporate social responsibility: The role of competitive positioning. International Journal of Research in Marketing, 24(3), 224–241. https://doi.org/10.1016/j.ijresmar.2007.01.001 Dyllick, T., & Hockerts, K. (2002). Beyond the business case for corporate sustainability. Business Strategy and the Environment Bus. Strat. Env, 11, 130–141. https://doi.org/10.1002/bse.323 Eccles, R. G., Lee, L. E., & Stroehle, J. C. (2020). The social origins of ESG: An analysis of innovest and KLD. Organization & Environment, 33(4), 575– 596. https://doi.org/10.1177/1086026619888994 Eccles, R. G., & Stroehle, J. C. (2018). Exploring Social Origins in the Construction of Environmental, Social and Governance Measures. https://doi. org/10.2139/ssrn.3212685 Escrig-Olmedo, E., Fernández-Izquierdo, M. ángeles, Ferrero-Ferrero, I., Rivera-Lirio, J. M., & MuñozTorres, M. J. (2019). Rating the raters: Evaluating how ESG rating agencies integrate sustainability principles. Sustainability, 11(3). https://doi. org/10.3390/su11030915 Filippou, I., Taylor, M. P., & Olin, J. M. (2021). Pricing ethics in the foreign exchange market: Environmental, Social and Governance ratings and currency premia. Journal of Economic Behavior and Organization, 191, 66–77. https://doi. org/10.1016/j.jebo.2021.08.037 Friede, G., Busch, T., & Bassen, A. (2015). ESG and financial performance: aggregated evidence from more than 2000 empirical studies. J. Sustain. Financ. Invest. , 5(4), 210–233. https://doi.org/10.10 80/20430795.2015.1118917 Gallego Alvarez, I. (2012). Impact of CO 2 Emission variation on firm performance. Business Strategy and the Environment, 21, 435–454. https://doi. org/10.1002/bse.1729 Gallo, T., Pacchera, F., Cagnetti, C., & Silvestri, C. (2023). Do Sustainable Consumers Have Sustainable Behaviors? An Empirical Study to Understand the Purchase of Food Products. Sustainability 2023, Vol. 15, Page 4462, 15(5), 4462. https:// doi.org/10.3390/SU15054462 Gillan, S. L., Koch, A., & Starks, L. T. (2021). Firms and social responsibility: A review of ESG and CSR research in corporate finance. Journal of Corporate Finance, 66. https://doi.org/10.1016/J.JCORPFIN.2021.101889 Glass, C., Cook, A., & Ingersoll, A. R. (2015). Do women leaders promote sustainability? Analyzing the effect of corporate governance composition on environmental performance. Business Strategy and the Environmen, 25, 495–511. https://doi. org/10.1002/bse.1879 Haque, F., & Ntim, C. G. (2022). Do corporate sustainability initiatives improve corporate carbon performance? Evidence from European firms. Business Strategy and the Environment. https://doi. org/10.1002/bse.3078 Harrison, J. S., & Coombs, J. E. (2012). The moderating effects from corporate governance characteristics on the relationship between available slack and community-based firm performance. Journal of Business Ethics, 107, 409–422. https://doi. org/10.1007/s10551-011-1046-z Hazen, T. L. (2021). Social issues in the spotlight: The increasing need to improve social publicly-held companies’ CSR and ESG disclosures. Journal of Business Law, 23(3), 741–796. https://scholarship. law.unc.edu/faculty_publications Igalens, J., & Joras, M. (2006). Responsabilité sociale de l’entreprise [Corporate social responsibility]. In Responsabilité sociale de l’entreprise. https://doi. org/10.3917/dbu.rose.2006.01 Islam, T., Islam, R., Pitafi, A. H., Xiaobei, L., Rehmani, M., Irfan, M., & Mubarak, M. S. (2021). The impact of corporate social responsibility on customer loyalty: The mediating role of corporate reputation, customer satisfaction, and trust. Sustainable Production and Consumption, 25, 123– 135. https://doi.org/10.1016/j.spc.2020.07.019 Kantenbacher, J., Hanna, P., Miller, G., Scarles, C., & Yang, J. (2017). Consumer priorities: what would people sacrifice in order to fly on holidays? Journal of Sustainable Tourism, 27(2), 207–222. https:// doi.org/10.1080/09669582.2017.1409230 Kaźmierczak, M. (2022). A literature review on the difference between CSR and ESG. Scientific Papers Of Silesian University Of Technology. https://doi. org/10.29119/1641-3466.2022.162.16 Kitchin, T. (2003). Corporate social responsibility: A brand explanation. Journal of Brand Management, 10(4–5), 312–326. Kolk, A., Levy, D., & Pinkse, J. (2008). Corporate responses in an emerging climate regime: The institutionalization and commensuration of carbon disclosure. European Accounting Review, 17(4), 719–745. https://doi. org/10.1080/09638180802489121 Konadu, R., Sam Ahinful, G., Jeff Boakye, D., & Elbar- 184 Alexandre Clement, Élisabeth Robinot, Léo Trespeuch, 10.5709/ce.1897-9254.560DOI: CONTEMPORARY ECONOMICS Vol. 19 Issue 2 163-1852025 dan, H. (2022). Board gender diversity, environmental innovation and corporate carbon emissions. Technological Forecasting & Social Change, 174, 40–1625. https://doi.org/10.1016/j.techfore.2021.121279 Kotsantonis, S., & Serafeim, G. (2019). Four things no one will tell you about ESG data. Journal of Applied Corporate Finance, 31(2), 50–58. https://doi. org/10.1111/jacf.12346 Kudłak, R. (2019). The role of corporate social responsibility in predicting CO2 emission: An institutional approach. Ecological Economics, 163. https:// doi.org/10.1016/j.ecolecon.2019.04.027 Le Quéré, C., Moriarty, R., Andrew, R. M., Peters, G. P., Ciais, P., Friedlingstein, P., Jones, S. D., Sitch, S., Tans, P., Arneth, A., Boden, T. A., Bopp, L., Bozec, Y., Séférian, R., Segschneider, J., Steinhoff, T., Stocker, B. D., Sutton, A. J., Takahashi, T., … Zeng, N. (2015). Global carbon budget 2014. Earth Syst. Sci. Data, 7, 20. https://doi.org/10.5194/essd7-47-2015 Lemken, D., Spiller, A., & Schulze-Ehlers, B. (2019). More room for legume-Consumer acceptance of meat substitution with classic, processed and meat-resembling legume products. Appetite, 143. https://doi.org/10.1016/j.appet.2019.104412 Lennan, M., & Morgera, E. (2022). The glasgow climate conference (COP26). International Journal of Marine and Coastal Law, 37(1), 137–151. https://doi. org/10.1163/15718085-bja10083 Lu, J., Herremans, I. M., Jing Lu, C., & Lang, G. S. (2019). Board gender diversity and environmental performance: An industries perspective. Bus Strat Env., 28, 1449–1644. https://doi.org/10.1002/ bse.2326 Mahmood, A., & Bashir, J. (2020). How does corporate social responsibility transform brand reputation into brand equity? Economic and noneconomic perspectives of CSR. International Journal of Engineering Business Management, 12. https://doi. org/10.1177/1847979020927547 Maiti, M. (2021). Is ESG the succeeding risk factor? Journal of Sustainable Finance & Investment, 11(3), 199–213. https://doi.org/10.1080/20430795.2020.1 723380 Masson-Delmotte, V., & The Intergovernmental Panel on Climate Change. (2021). Climate change 2021: The physical science basis: Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. In Cambridge University Press. In Press. https://www. ipcc.ch/report/ar6/wg1/downloads/report/IPCC_ AR6_WGI_FullReport.pdf Mehmood, U. (2022). Investigating the linkages of female employer, education expenditures, renewable energy, and CO2 emissions: application of CS-ARDL. Environmental Science and Pollution Research, 1, 3. https://doi.org/10.1007/s11356-022-20275-1 Murshed, M., Ali, S. R., & Banerjee, S. (2021). Consumption of liquefied petroleum gas and the EKC hypothesis in South Asia: evidence from cross-sectionally dependent heterogeneous panel data with structural breaks. Energy, Ecology and Environment, 6, 353–377. https://doi.org/10.1007/s40974020-00185-z Nugroho, D. P. D., Hsu, Y., Hartauer, C., & Hartauer, A. (2024). Investigating the Interconnection between environmental, social, and governance (ESG), and corporate social responsibility (CSR) Strategies: An examination of the influence on consumer behavior. Sustainability 2024, Vol. 16, Page 614, 16(2), 614. https://doi.org/10.3390/SU16020614 Olmedo, E. E., Torres, M. J. M., & Izquierdo, M. A. F. (2010). Socially responsible investing: sustainability indices, ESG rating and information provider agencies. International Journal of Sustainable Economy, 2(4), 442. https://doi.org/10.1504/ ijse.2010.035490 Peratz, A., & Strønen, F. (2024). Corporate social responsibility as a source of competitive advantage - strategic contradictions in the food industry. Contemporary Economics, 18(2), 171–191. https:// doi.org/10.5709/ce.1897-9254.532 Rahman, M., Ángeles Rodríguez-Serrano, M., & Faroque, A. R. (2021). Corporate environmentalism and brand value: A natural resource-based perspective. Journal of Marketing Theory and Practice, 29(4), 463–479. https://doi.org/10.1080/10696679. 2021.1872387 Ramesh, K., Saha, R., Goswami, S., Sekar, & Dahiya, R. (2019). Consumer’s response to CSR activities: Mediating role of brand image and brand attitude. Corporate Social Responsibility and Environmental Management, 26(2), 377–387. https://doi. org/10.1002/csr.1689 Ritchie, H., Rosado, P., & Roser, M. (2023). CO₂ and Greenhouse Gas Emissions. Https://Ourworldindata.Org/Co2-and-Greenhouse-Gas-Emissions. Saadaoui, K., & Soobaroyen, T. (2018). An analysis of the methodologies adopted by CSR rating agencies. Sustainability Accounting, Management and Policy Journal, 9(1), 43–62. https://doi. org/10.1108/SAMPJ-06-2016-0031 Serafeim, G., Kramer, M., Porter, B. M. E., Serafeim, G., www.ce.vizja.pl 185 Beyond Resource Use: XGBoost Reveals Unexpected Drivers of Corporate CO2 Emissions This work is licensed under a Creative Commons Attribution 4.0 International License. & October, M. K. (2019). Where ESG Fails. Institutional Investor, 1–17. https://www.institutionalinvestor.com/article/b1hm5ghqtxj9s7/Where-ESGFails Setó-Pamies, D. (2015). The relationship between women directors and corporate social responsibility. Corporate Social Responsibility and Environmental Management, 22(6), 334–345. https://doi. org/10.1002/CSR.1349 Sprinkle, G. B., & Maines, L. A. (2010). The benefits and costs of corporate social responsibility. Business Horizons, 53(5), 445–453. https://doi. org/10.1016/j.bushor.2010.05.006 Stern, N. (2007). The economics of climate change. Cambridge University Press. https://doi.org/10.1017/ CBO9780511817434 Stern, N. (2014). Ethics, equity, and the economics of climate change, paper 2: Economics and politics. Economics and Philosophy, 30, 445–501. https:// doi.org/10.1017/S0266267114000303 Stern, P. C., Dietz, T., & Kalof, L. (1993). Value orientations, gender, and environmental concern. Environment and Behaviorismes, 25(3), 322–348. https://doi.org/10.1177/0013916593255002 Streimikiene, D. (2023). Low-carbon Energy Transition from the lens of feminist theories. Contemporary Economics, 17(4), 456–468. https://doi. org/10.5709/ce.1897-9254.522 Supanti, D., Butcher, K., & Fredline, L. (2015). Enhancing the employer-employee relationship through corporate social responsibility (CSR) engagement. International Journal of Contemporary Hospitality Management, 27(7), 1479–1498. https://doi. org/10.1108/IJCHM-07-2014-0319 Taparia, H. (2021). Change the Misleading ESG Ratings System to Fix ESG Investing. Stanford Social Innovation Review. https://doi.org/doi.org/10.48558/ PC0C-TV52 Thøgersen, J. (2021). Consumer behavior and climate change: Consumers need considerable assistance. Behavioral Sciences, 42, 9–14. https://doi. org/10.1016/j.cobeha.2021.02.008 Thøgersen, J., & Ebsen, J. V. (2019). Perceptual and motivational reasons for the low adoption of electric cars in Denmark. Transportation Research Part F: Traffic Psychology and Behaviour, 65, 89–106. https://doi.org/10.1016/j.trf.2019.07.017 Tosun, P., & Köylüoglu, A. S. (2022). The impact of brand origin and CSR actions on consumer perceptions in retail banking during a crisis. International Journal of Bank Marketing, 41(3), 485–507. https://doi.org/10.1108/IJBM-03-2022-0137 Trespeuch, L., & Robinot, É. (2023). Exploring the impact of corporate philanthropy on brand authenticity in the luxury industry: Scale development and empirical studies. Sustainability 2023, Vol. 15, Page 12274, 15(16), 12274. https://doi. org/10.3390/SU151612274 Trespeuch, L., Robinot, É., Botti, L., Bousquet, J., Corne, A., De Ferran, F., Durif, F., Ertz, M., Fontan, J. M., Giannelloni, J. L., Hallegatte, D., Kreziak, D., Lalancette, M., Lajante, M., Michel, H., Parguel, B., & Peypoch, N. (2021). Are we moving towards a more responsible society thanks to the Covid-19 pandemic? Natures Sciences Sociétés, 29(4), 479– 486. https://doi.org/10.1051/nss/2022005 Tronto, J. (2022). Moral boundaries: A political argument for an ethic of care. Routledge. Turker, D. (2009). Measuring corporate social responsibility: A scale development study. Journal of Business Ethics, 85(4), 411–427. https://doi.org/10.1007/ s10551-008-9780-6 Van Den Broek, K. L., Walker, I., & Klöckner, C. A. (2019). Drivers of energy saving behaviour: The relative influence of intentional, normative, situational and habitual processes. Energy Policy, 132, 811–819. https://doi.org/10.1016/j.enpol.2019.06.048 Van Doorn, J., Onrust, M., Verhoef, P. C., & Bügel, M. S. (2017). The impact of corporate social responsibility on customer attitudes and retention—the moderating role of brand success indicators. Marketing Letters, 28(4), 607–619. https://doi.org/10.1007/ s11002-017-9433-6 Wang, H., Choi, J., & Li, J. (2008). Too little or too much? Untangling the relationship between corporate philanthropy and firm financial performance. Organization Science, 19(1), 143–159. https://doi. org/10.1287/orsc.1070.0271 Wang, Q., Hang, Y., Su, B., & Zhou, P. (2017). Contributions to sector-level carbon intensity change: An integrated decomposition analysis. Energy Economics, 70, 12–25. https://doi.org/10.1016/j.eneco.2017.12.014 Weber, M. (2008). The business case for corporate social responsibility: A company-level measurement approach for CSR. European Management Journal, 26(4), 247–261. https://doi.org/10.1016/j. emj.2008.01.006 Widyawati, L. (2020). A systematic literature review of socially responsible investment and environmental social governance metrics. Bus. Strategy Environ., 29(2), 619–637. https://doi.org/10.1002/bse.2393