From Diversity to Confusion? The Challenge of Biodiversity Footprint Quantification

Roeder, Raphaela; Utz, Sebastian Article — Published Version From Diversity to Confusion? The Challenge of Biodiversity Footprint Quantification Business Strategy and the Environment Provided in Cooperation with: John Wiley & Sons Suggested Citation: Roeder, Raphaela; Utz, Sebastian (2025) : From Diversity to Confusion? The Challenge of Biodiversity Footprint Quantification, Business Strategy and the Environment, ISSN 1099-0836, Wiley, Hoboken, NJ, Vol. 34, Iss. 5, pp. 5887-5900, https://doi.org/10.1002/bse.4215 This Version is available at: https://hdl.handle.net/10419/323859 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. http://creativecommons.org/licenses/by/4.0/ Business Strategy and the Environment, 2025; 34:5887–5900 https://doi.org/10.1002/bse.4215 5887 Business Strategy and the Environment RESEARCH ARTICLE OPEN ACCESS From Diversity to Confusion? The Challenge of Biodiversity Footprint Quantification RaphaelaRoeder | SebastianUtz Faculty of Business and Economics and Centre for Climate Resilience, University of Augsburg, Augsburg,Germany Correspondence: Sebastian Utz ([email protected]) Received: 22 December 2024 | Revised: 2 February 2025 | Accepted: 11 February 2025 Funding: This research was supported by the University of Augsburg x Green Research Network BRaVE. Keywords: biodiversity| biodiversity footprint| disagreement ABSTRACT This study documents a significant disagreement between the biodiversity footprints of three major providers. This disagreement mainly stems from fundamental disagreement on the underlying methods and data (measurement), while providers agree to a large part on which firm operations contribute to a loss in biodiversity and how they are aggregated (scope and weight). The disagreement is especially high for large firms with a high biodiversity footprint and firms from the industries of Energy, Consumer Staples, and Basic Materials. A transparent and detailed ESG disclosure can decrease the disagreement. The results highlight the importance of being careful when integrating biodiversity footprint into financial decisionmaking, regulations, and academic research. The results also underline the need for further standardized and transparent biodiversity disclosure on firm level. 1 | Introduction Humanity is facing an unprecedented loss in biodiversity.1 Recently, this loss has gained awareness among investors, decisionmakers, and firms, driven by increasing regulation around the disclosure and measurement of a firm's impact and dependencies on nature. In December 2022, the Global Biodiversity Framework (GBF) was signed by 196 countries, calling on firms to disclose their impact on biodiversity, and in September 2023, the Taskforce for Naturerelated Financial Disclosure (TNFD) published their final set of recommendations for firms and financial institutions to “identify, assess, manage and (…) disclose naturerelated issues” (TNFD2023b, p.7). Already in 2021, the French government obligated financial institutions in Article 29 of the Energy and Climate Law to disclose their exposure to biodiversityrelated risk as well as their impact on biodiversity, referring to socalled biodiversity footprints (Art. 29, LEC). In response to these regulatory developments, firms and investors are increasingly measuring and disclosing their impact and dependencies on nature. Biodiversity footprints—quantitative metrics describing the negative impact of a firm's operations on biodiversity—have emerged as influential tools used by firms, financial institutions, researchers, and more (TNFD2023a). A variety of data providers now offer firmlevel biodiversity footprints. However, measuring a firm's impact on nature is a complex and challenging process as the impact on biodiversity is multilayered and stems for example from resource use, waste and water management, and greenhouse gas emissions. Since the regulatory frameworks require reliable data sources to be effective, understanding the methods and the agreement of biodiversity footprints by different providers is key. Various studies assess the agreement of other sustainability ratings and find substantial disagreement between different providers for ESG scores (Berg, Kölbel, and Rigobon2022; Chatterji etal. 2016; Christensen, Serafeim, and Sikochi2022; Dorfleitner, Halbritter, and Nguyen 2015; Widyawati 2020), SDG ratings (Bauckloh et al. 2024), physical climate risk scores (Hain, Kölbel, and Leippold2022), and ITR ratings (Kathan, Utz, and Chmel2023). The low correlation of sustainability ratings raises This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2025 The Author(s). Business Strategy and the Environment published by ERP Environment and John Wiley & Sons Ltd. 5888 Business Strategy and the Environment, 2025 questions about their usefulness. In general, sustainability ratings have predictive power for future ESG news, but that relationship weakens for firms with a high disagreement between raters (Serafeim and Yoon2023). Moreover, stronger ESG score disagreement is associated with higher stock return volatility, larger absolute price movements (Christensen, Serafeim, and Sikochi2022), and higher stock returns, suggesting a risk premium for those firms (Avramov etal. 2022; Gibson, Krueger, and Schmidt2021). Similarly, a low level of agreement between SDG ratings affects the riskreturn characteristics of the rated firms (Bauckloh etal. 2024). Methods to assess biodiversity impact have been developed and discussed in the Life Cycle Assessment (LCA) literature for some time (e.g., Winter etal. 2017; Wilting etal. 2017). Recently, they are also gaining increasing relevance among scholars and decisionmakers in the business and finance communities, where biodiversity footprints are often used as a measurement for firmlevel biodiversity impact. It has been shown that investors are including biodiversity footprints in their investment decisions, as a biodiversity footprint premium began emerging after COP15 (Garel etal. 2024). Coqueret etal.(2025) additionally find that the same biodiversity footprint impacts both the expected and realized stock returns. Even when only focusing on the firmlevel exposure of land use, one of the main drivers of biodiversity loss, a link to stock returns can be found (Xiong 2024). While this research focuses on biodiversity impact, another stream of literature has investigated biodiversity risk, that is, both the physical risk arising from a loss in biodiversity and the transitional risk associated with new regulations and rules to stop the decline in biodiversity. Giglio etal.(2024) developed an approach to measure the firmlevel exposure to biodiversity risk from firms' 10K reports. This risk already affects equity prices (Giglio etal. 2024) and hinders firm performance (Bach, Hoang, and Le2024). Moreover, firms that better manage biodiversity risk and the closely connected water and pollution risks profit from better longterm refinancing conditions (Hoepner etal. 2023) and suffer less stock price crashes (Bassen etal. 2024). There is also a positive relationship between firm biodiversity disclosure and return on assets (Elsayed2023). Further literature in the biodiversity finance field focuses on the biodiversity finance gap and financing mechanisms that can contribute to conserve biodiversity, for instance, blended finance structures (Flammer, Giroux, and Heal2025; Karolyi and Tobinde la Puente2023). The above mentioned studies emphasize the increasing importance of biodiversity in firm and investment decisions, as well as the use of biodiversity footprints by decisionmakers and scholars. They thereby underline the need for a better understanding of the metrics, as well as the severity of potential consequences from its disagreement. This paper aims to contribute to this literature by analyzing the footprints of three major providers (i.e., ISS ESG, Iceberg Data Lab, and Impact Institute). After briefly describing the process of calculating a biodiversity footprint in general, we test the level of agreement between the biodiversity footprints of the three providers on a global set of 941 firms using different measures for pairwise and multiple provider comparison. We find substantial disagreement for all providers. Subsequently, we investigate the reasons for the disagreement by following Berg, Kölbel, and Rigobon(2022) and distinguishing scope, measurement, and weight disagreement between providers. We find that while providers generally focus on similar factors when calculating the footprint, there are strong differences in the measurement of those factors. Aggregating those factors into impact drivers that are responsible for the loss in biodiversity reveals that the firms' contribution to land use & pollution as well as climate change is a large factor on its overall contribution to biodiversity loss. Consequently, the providers' disagreement stems in a large part from the discrepancies on these factors. Using crosssectional regressions, we find that the disagreement is particularly strong for large firms and firms from the industries of Energy, Basic Materials, and Consumer Staples. In contrast, firms from the industries of Telecommunications and Financials exhibit a lower disagreement. Moreover, detailed and transparent ESG disclosure on firm level as well as a high share of institutional investors tend to decrease the disagreement by the providers. Our findings have important implications for businesses, regulators, investors, and other users of biodiversity footprints. The disagreement between the biodiversity footprints can harm the very purpose of them, making it difficult to evaluate the impact of a firm on biodiversity. If the users are not aware of the different methodologies or do not consider them appropriately when choosing and using a biodiversity footprint, the footprints do not contribute sufficiently to decreasing the information asymmetry between businesses and stakeholders about their impact on biodiversity. This has farreaching consequences: Efforts to reallocate capital into biodiversityfriendly firms are being mitigated, firms' uncertainty about the material factors of biodiversity performance increases, and their incentive to improve their biodiversity performance decreases. In addition, it reduces the reliability of the results of academic studies based on such metrics. The remainder of the paper is structured as follows. Section2 describes the theoretical framework that this study is based on. In Section3, we first describe how biodiversity impacts are measured and how a biodiversity footprint is calculated in general. Subsequently, we describe our sample. Section4 starts by showing the significant disagreement between the providers, then decomposing the disagreement in scope, measurement, and weight, and identifying the contribution of each to the overall disagreement. Section5 analyzes determinants of the disagreement, and Section6 documents the robustness of our main results. We conclude in Section7. 2 | Theoretical Background Institutional theory provides the theoretical foundation of why businesses and other stakeholders (such as investors) demand consistent biodiversity footprints. A key element of the institutional theory is the concept of institutional pressures, which, for example, describe a society's ethical expectations on an organization's behavior (Meyer and Rowan1977; DiMaggio and Powell 1983; Scott and Meyer 1994). The responses of a firm to institutional pressures are mainly driven by the motivation to ensure its legitimacy (Meyer and Rowan1977). Legitimacy is critical for a firm's survival as it ensures access to important resources such as capital and talent. It also decreases the 5889 probability of being targeted with retributions like fines or loss of sales (Deegan2002a). In that sense, sustainability reports can be seen as means to obtain approval from society, comply with the “community license to operate” and support its continued existence (Deegan 2002b). While reporting on climaterelated issues such as carbon emissions is increasingly being standardized, biodiversityfocused disclosure is found to be very limited, inconsistent, and highly variable between firms (Hassan, Roberts, and Atkins2020; Adler etal. 2017; Adler, Mansi, and Pandey 2018). Alarmingly, firms with a significant negative impact on biodiversity even adopt reporting strategies aimed at neutralizing stakeholder concerns rather than providing transparency (Boiral2016). This inconsistent reporting environment of biodiversity impact leads to a high level of uncertainty and information asymmetry regarding biodiversity impacts for stakeholders. Biodiversity footprints, like other thirdparty sustainability ratings, have the potential to reduce information asymmetry by acting as intermediaries between firms and stakeholders (Bauckloh etal. 2024; Chatterji and Toggel2010). Still, this potential can only be realized if the available footprint measures provide an accurate and consistent signal of a firm's biodiversity impact. To the best of our knowledge, there is no research around the reliability and consistency of biodiversity footprints yet. Since there are numerous models that can be used for calculating biodiversity impact (e.g., Damiani etal. 2023, with a review of 64 methods), and the usage of different models has been found to lead to different results (SanyMengual etal. 2023), the footprints by different providers might also exhibit significant differences. This observation would add to the finding of studies that raise concerns regarding the accuracy and consistency of other sustainability ratings. It has been found that firms implement superficial actions primarily to improve their ratings rather than to achieve substantive environmental improvements (Cornaggia and Cornaggia2023; Clementino and Perkins2020; Chelli and Gendron2013). Moreover, sustainability ratings exhibit significant disagreement among providers (Chatterji etal. 2016; Berg, Kölbel, and Rigobon 2022; Bauckloh et al. 2024; Hain, Kölbel, and Leippold2022; Kathan, Utz, and Chmel2023). This lack of clarity can hinder appropriate decisionmaking by investors and other stakeholders, which has severe asset pricing implications (Avramov et al. 2022; Gibson, Krueger, and Schmidt2021; Serafeim and Yoon2023; Christensen, Serafeim, and Sikochi2022). Discrepancies among biodiversity footprints could lead to similarly adverse consequences, especially considering that financial institutions often apply them too simplistic (TNFD2023a). Biodiversity disclosure and biodiversity risk are already affecting stock returns (Elsayed2023; Giglio etal. 2024). Biodiversity footprints of thirdparty providers have also been found to have implications for asset pricing (Garel etal. 2024; Coqueret etal. 2025). While the disagreement between ESG scores can be related to different perceptions of what sustainability is, the biodiversity footprint focuses clearly on the firm's impact on the diversity of species. Therefore, on the one hand, biodiversity footprints could provide a clearer signal and agree more strongly than other sustainability ratings, such as ESG scores. On the other hand, differences in ESG scores have been found to be strongly driven by differences in measurement (Berg, Kölbel, and Rigobon2022). Differences in measurement can also lead to disagreement between biodiversity footprints, in particular taking into account the multitude of possibly underlying methods and models (Damiani etal. 2023; SanyMengual etal. 2023). This variability might even be more pronounced, as the impact on biodiversity is multilayered, firm disclosures are unclear (Hassan, Roberts, and Atkins2020; Adler etal. 2017; Adler, Mansi, and Pandey2018), and the footprint is explicitly trying to capture the impact through the whole value chain of the firm and the entire life cycle of its products. Our study empirically tests whether biodiversity footprints are a suitable means for firms (stakeholders) to ensure (assess) legitimacy. For this purpose, we analyze whether the providers create a clear signal of firms' biodiversity impact. 3 | Sample and Data 3.1 | Measuring Biodiversity Impacts Research institutions (e.g., Oxford Biodiversity Network2023; University of Cambridge Institute for Sustainability Leadership 2020), firms (e.g., ASN Bank and PRé Sustainability2022; Kering2021), governments (e.g., Broer etal. 2021), and data providers have developed methodologies and metrics to measure the impact and dependencies of a firm on nature. Over time, the socalled biodiversity footprint has become the most commonly used method to measure and express a firm's impact on nature. While there is no clear agreement on a definition for a biodiversity footprint, the Institute for European Environmental Policy (IEEP) mentions that it has to be “measured in terms of biodiversity change as a result of production and consumption of particular goods and services” (IEEP2021, p. 12), and the Partnership for Biodiversity Accounting Financials (PBAF) emphasizes that the impact must be quantified (PBAF2022). The calculation of the biodiversity footprint is based on a LCA approach. LCA is used to quantify a product's or organization's impact on the environment over its full life cycle and in that context the impact on biodiversity has been explored already for over 20 years (Winter etal. 2017), long before biodiversity became a relevant topic in the finance industry. Today, there is a wide variety of methodologies to calculate a biodiversity footprint (Damiani etal. 2023). While the footprints of some providers are developed to analyze the impact of a firm on nature, others focus only on investments, products, and projects, as well as industries and countries. Moreover, the methodologies differ in the data used, the calculations, the outcome metric, and more. In this paper, we employ firmlevel footprints generated by thirdparty data providers. The first step to calculate the footprint of a firm is to identify environmental inputs and outputs that are linked to the firm's business activities. These aspects include the amount of water used, the emitted greenhouse gas emission, and the amount of land that a firm uses. Factors like the diversity of species and the amount of endangered species in a firm's operating areas also affect the biodiversity footprint. That information is 5890 Business Strategy and the Environment, 2025 usually based on geospatial data and often stems from further third party providers, for example, the Integrated Biodiversity Assessment Tool (IBAT). If a firm is not providing specific information but their peers do, some providers use a benchmarking approach, others developed sophisticated machine learning methods. In the second step, firmspecific input and output variables are translated into impact drivers which exert pressure on biodiversity by contributing to (1) habitat loss, (2) overexploitation, (3) climate change, (4) pollution, and (5) the introduction of invasive species (IPBES2019). Lastly, the impact of these pressures is measured and usually expressed in “Mean Species Abundance” (MSA) or “Potentially Disappeared Fraction of Species” (PDF) (PBAF 2022; Broer etal. 2021). Both MSA and PDF are measures of biodiversity intactness. The MSA compares the abundance of an original species in an area to the estimated abundance if the ecosystem would be undisturbed and is expressed as a percentage rate (Hertog, Bor, and de Horde2022; Schipper etal. 2020). The PDF shows the fraction of species lost in a specific area due to environmental pressures, without taking a decline in species population into account (Hertog, Bor, and de Horde2022). However, it is also possible to express the footprint in other measures, such as a loss of pristine biodiversity per hectare or in a monetized way. Besides calculating a biodiversity footprint, the impact and dependencies of a firm on nature can be expressed in other aggregated scores. For instance, the World Benchmark Alliance publishes its assessment of the impact on biodiversity of over 800 firms on a scale from 0 to 100.2 It is important to note that most biodiversity scores that are not biodiversity footprints focus only on partial aspects of biodiversity or use simplified approaches to capture biodiversity impacts and risks. Examples comprise the rating of the commitment of over 700 financial institutions to deforestation generated by the nonprofit organization Global Canopy,3 the percentage value stating to what level a firm complies with the “Do No Significant Harm” goal regarding the biodiversity objective of the EU Taxonomy of Bloomberg, and the scoring of the exposure of US firms to biodiversity risks based on a textual analysis of their 10K statements (Giglio etal. 2024; Bach, Hoang, and Le2024). Per definition, these scores are not constructed to capture the impact of a firm on biodiversity as a whole, while the marked goal of biodiversity footprints is to include the entire range of biodiversity aspects throughout the whole life cycle of a product or a firm. Therefore, we are limiting the analysis to footprint measures. 3.2 | Sample Description We obtain the most recently available biodiversity footprints in January 2024 from the three providers Iceberg Data Lab (IDL), Impact Institute (II), and ISS ESG (ISS). All three scores are recommended by the Finance for Biodiversity Foundation and the EU Business and Biodiversity Platform (Bailon, Bor, and Redn 2024); the scores by IDL and II are also recommended by the TNFD (TNFD 2023a). Moreover, the biodiversity footprints by IDL have already been used in different research (Coqueret etal. 2025; Garel etal. 2024). IDL discloses the footprint and the underlying impact drivers in MSA.km2, ISS both in PDF.km2 and MSA.km2, and II in MSA.ha and PDF.ha. To ensure comparability, we zstandardize these values for the empirical analysis. A higher number indicates a larger biodiversity footprint, meaning a stronger negative impact of a firm on biodiversity. Our sample per provider consists of 3388 firms (IDL), 1516 firms (II), and 17,931 firms (ISS). The final data set consists of all firms that have been rated by all three providers and are publicly traded, leaving a set of 941 firms. Henceforth, the providers will be anonymized and labeled only as “Provider1”, “Provider2”, and “Provider 3” to ensure discretion regarding their exact methodologies. Table 1 shows the average standardized footprint by industry and region per provider. Most of the analyzed firms' headquaters are located in North America (405 firms) and Europe (258 firms), operating in the Financials, Industrials, Consumer Discretionary, and Technology industry.Further, Table1 indicates a strong disagreement between providers on regionand industrylevel. Across regions, Latin America has the highest mean footprint for Providers 2 and 3 but the smallest mean footprint for Provider 1. Similar strong disagreements can be found when comparing the average footprints by industry per provider. There is no agreement on which industry has the most or the least average impact on biodiversity. Still, the Energy and Consumer Staples industries display a high average biodiversity footprint for all providers, with a standardized mean footprint ranging between 0.132 and 1.572, and 0.028 and 0.887, respectively. In contrast to that, the industries Real Estate, Technology, and Telecommunications tend to have a small impact on biodiversity, with a footprint below the mean of zero for all providers. Table2 presents the distribution of the standardized footprints per provider. It demonstrates that for all providers, very few firms have an extremely large footprint, with the maximum ranging between 19.369 for Provider 1 and 29.483 for Provider 4.4 A large share of the remaining firms has very similarsized footprints well below the mean. The median ranges between −0.248 for Provider2 and −0.050 for Provider3. 4 | The Disagreement of Biodiversity Footprints This section discusses the level of agreement between the biodiversity footprints of different providers. 4.1 | Association Measures Table1 indicates strong disagreement between all providers on regionand industrylevel. Moreover, anecdotal evidence also suggests disagreement on firmlevel. Walmart, for example, has a standardized footprint of 11.58 for Provider1 but −0.029 for Provider3. To further understand the agreement on firmlevel, we calculate metrics for pairwise (Pearson and Spearman correlation) and multiple (Krippendorff's alpha) providercomparison. While the Pearson correlation measures the linear and the Spearman correlation the rankbased relationship between two continuous variables, Krippendorff's alpha is commonly used to analyze interrater reliability. A value of 1 indicates perfect agreement, a value of 0 the complete absence of agreement, and 5891 a negative value indicates systematic disagreement. A value greater than 0.8 can be interpreted as sufficient agreement, while 0.667 is the lowest acceptable limit to conclude any agreement (Krippendorff2004). Table3 shows the correlation matrices of the standardized biodiversity footprint of our sample firms. The Pearson correlation ranges between 0.024 and 0.493 with an average of 0.206, and the Spearman correlation ranges between 0.619and 0.745 with an average of 0.670. The overall Krippendorff's alpha for all raters is 0.206. We also calculate Krippendorff's alphas for all combinations of two providers and find values in the range of 0.025 and 0.493 (unreported results). No pair of providers displays a Krippendorff's alpha higher than the threshold of 0.667, suggesting no systematic agreement between the different footprints. Overall, the disagreement on firms' biodiversity footprint measured by Krippendorff's alpha is substantially higher than the disagreement on ESG scores, which already has severe consequences (Serafeim and Yoon2023; Avramov etal. 2022; Christensen, Serafeim, and Sikochi 2022). ESG scores from different providers have a Krippendorff's alpha of 0.55 (Berg, Kölbel, and Rigobon2022). The average pairwise Pearson correlations between 0.45 (Gibson, Krueger, and Schmidt2021) and 0.54 (Berg, Kölbel, and Rigobon2022) are of similar size as the correlations we observe for the biodiversity footprint. Thus, we document indications that the biodiversity footprints of the TABLE 1 | This table reports the mean and standard deviation of the standardized biodiversity footprint by the different providers, distributed by region (Panel A) and ICB industry classification (Panel B). Provider1 Provider2 Provider 3 N Mean Std. dev. Mean Std. dev. Mean Std. dev. Panel A: Regions Asia Ex Japan 48 0.271 2.859 −0.069 1.066 −0.038 0.061 Europe 258 0.028 0.827 0.030 1.422 −0.043 0.026 Japan 184 −0.125 0.290 −0.165 0.423 −0.045 0.011 Latin America 6−0.220 0.093 0.157 0.904 4.872 12.057 North America 405 0.028 0.936 0.071 0.879 −0.015 0.414 Oceania 40 −0.177 0.235 −0.095 0.496 −0.048 0.005 Panel B: Industries Basic materials 45 0.271 1.246 −0.052 0.378 −0.041 0.011 Consumer Discretionary 148 0.025 1.148 −0.017 0.902 −0.046 0.008 Consumer Staples 82 0.887 2.415 0.468 1.213 0.028 0.324 Energy 37 0.132 0.447 1.572 3.552 0.986 4.982 Financials 146 −0.073 0.425 −0.173 0.304 −0.050 0.002 Health Care 91 0.011 0.926 −0.047 0.587 −0.044 0.013 Industrials 175 −0.173 0.192 −0.123 0.447 −0.047 0.009 Real Estate 54 −0.271 0.017 −0.347 0.045 −0.050 0.001 Technology 114 −0.240 0.071 −0.204 0.450 −0.048 0.009 Telecommunications 20 −0.271 0.016 −0.081 0.414 −0.049 0.001 Utilities 29 −0.218 0.060 0.105 0.488 −0.049 0.002 TABLE 2 | This table reports the distribution of the standardized footprints per providerby presenting the minimum, maximum, and the 1%, 10%, 50%, 90%, and 99% percentiles. Minimum 1% perc. 10% perc. 50% perc. 90% perc. 99% perc. Maximum Provider1 −0.281 −0.281 −0.281 −0.241 0.324 3.606 19.369 Provider2 −0.379 −0.377 −0.361 −0.248 0.496 3.374 20.923 Provider3 −0.051 −0.051 −0.051 −0.050 −0.035 0.073 29.483 Note:The mean standardized footprint for all providers is zero, due to the zstandardization. 5892 Business Strategy and the Environment, 2025 considered providers differ substantially. In the following sections, we analyze where the disagreement comes from and for which firms it is particularly pronounced. 4.2 | Reasons for Disagreement In the following section, we explore the reasons for the providers' disagreement. Following Berg, Kölbel, and Rigobon(2022)'s analysis of the disagreement between ESG raters, the causes for interrater disagreement can stem from divergence in scope, measurement, and weight. Like ESG scores, biodiversity footprints are based on different indicators. Consequently, also the disagreement in the biodiversity footprints can be decomposed into these factors. In our case, scope refers to the different impact drivers that the providers consider when calculating the footprint, measurement to differences in the calculations on impact driver level, and weight to the weight with which each impact driver contributes to the final score. Scope. Divergence in scope occurs if the providers are measuring different factors to capture a firm's biodiversity impact. As described above, the biodiversity footprints are calculated by identifying how a firm's activities contribute to different impact drivers. Beneath these impact drivers lie different indicators, for example, specific gas emissions contributing to air pollution or differentiation between freshwater and marine acidification, both contributing to water pollution. The aggregated number of considered indicators varies from 4 to 13 across the three providers. Comparing the number and depth of these indicators only gives insight into the granularity of the data that the providers share with their clients, but it would be false to assume that it can be translated into divergences in scope. Therefore, we summarize the indicators into impact drivers that they contribute to and compare them. Providers 1 and 2 consider the four impact drivers Climate Change, Air Pollution, Water Pollution, and Land Use & Pollution. Provider3 individually shares indicators for Climate Change, Land Use & Pollution, Water Pollution, and Water Extraction. The providers express the contribution of each indicator to the decline in biodiversity in the same metric that they use for the footprint (PDF or MSA). Aggregated, the values of all indicators add up to the entire footprint. This allows us to measure the contribution of each impact driver to the entire footprint by calculating the average share of the aggregated indicator values that are underlying the respective impact driver to the entire footprint: In Equation(1), f denotes one of the 941 investigated firms, di the impact driver by Provider i,indfi the value of the underlying indicator for firm f by Provider i and Ffi the complete footprint of firm f by Provider i . Naturally, the contributions of all impact drivers per provider sum up to 1. Table4 shows the contribution of each impact driver per provider in %. On average, the impact driver Land Use & Pollution contributes 67.17% to the overall footprint and is therefore the most influential impact driver, followed by Climate Change, which contributes on average 17.06%. Measurement. Divergences in measurement describe how providers use different approaches and data to measure the same factor. This can stem from different ideas about which firm activities contribute in which way to an impact driver, and from different approaches to solving the issue of missing data. As regulations on biodiversity disclosure are globally still in a very early stadium, many firms only provide very limited information about their impact on biodiversity (Adler etal. 2017; Adler, Mansi, and Pandey2018; Hassan, Roberts, and Atkins2020). While some providers use a benchmark approach to fill these data gaps, others have developed machinelearning models. (1) Contribution( di)=1 941 � f∈{1, …,941} ∑ ind∈di indfi Ffi TABLE 3 | This table reports the pairwise Pearson and Spearman correlation of each providerpair. Provider1 Provider2 Provider 3 Panel A: Pearson correlation Provider1 — 0.493 0.024 Provider2 0.493 —0.100 Provider 3 0.024 0.100 — Panel B: Spearman correlation Provider1 — 0.619 0.646 Provider2 0.619 —0.745 Provider 3 0.646 0.745 — TABLE 4 | This table shows the average contribution of each impact driver to the entire footprint (in %). Impact driver Provider1 Provider2 Provider3 Climate Change 22.13 28.18 0.87 Land Use & Pollution 51.53 51.56 98.41 Air Pollution 6.82 19.34 0 Water Pollution 19.52 0.92 0.71 Water Extraction 0 0 0.00 5893 The divergence in measurement is assessed by calculating the same agreement measures as for the overall footprint for the value of the different impact drivers. A low agreement on impact driver level would point out that the overall disagreement cannot only stem from discrepancies in the provider's perception about what contributes to biodiversity loss (scope) and to which degree (weight) but that they also in fact have different ways of measuring the same aspects. The results in Table5 show that the highest agreement between the providers is on the Land Use & Pollution of the firms, reaching a Krippendorff's alpha of 0.458. The agreement for the impact drivers Air Pollution and Climate Change is higher than for the overall footprint as well, with a Krippendorff's alpha of 0.442 and 0.211, respectively. On the other hand, there seems to be a lot of uncertainty around Water Pollution. Still, Krippendorff's alpha does not exceed the limit of 0.667 for any of the impact drivers, indicating no systematic agreement also on impact driver level. Weight. Berg, Kölbel, and Rigobon(2022) also study the weight of the indicators in their study on ESG scores. In our cases, the considered biodiversity footprints of IDL, II, and ISS follow an equally weighted aggregation scheme, that is, the weights are known. As all footprints follow an equalweighting in general, we expect weight not to be a substantial driver of the observed disagreement. A theoretical argument that weight should not drive the disagreement is based on the unit at which the biodiversity footprints are measured. The footprints are typically expressed in MSA or PDF units, which essentially quantify the number of species or individuals of a species that a firm displaces or eliminates within a particular region. Impact drivers provide more granular data, indicating that a firm's land use results in the extinction of x species/individuals, while water pollution causes the loss of y species/individuals, for instance. Naturally, impact drivers are simply summarized, and while different weighting makes sense for ESG scores, it does not for footprints in MSA and PDF units. 4.3 | Decomposition The previous section showed that the providers disagree both in scope and measurement. In this section, we follow Berg, Kölbel, and Rigobon(2022) and decompose the footprints to determine how much each factor contributes to the overall disagreement between the footprints. Let Ffi be the notstandardized footprint of a firm f , given by a Provider i∈{1;2;3} . For each two Providers i,j ( i,j∈{1;2;3},i≠j ), it then holds: where Dfi,com is the sum of the impact drivers by provider i that i has in common with provider j for firm f , measured in PDF or MSA depending on the provider, and Dfi,ex is the sum of the mutually exclusive impact drivers of provider i and j , by provider i for firm f . As the original footprints are given in different metrics depending on the provider(PDF or MSA), we have to standardize the footprints to the same unit when we analyze the differences in the footprints of two providers. It follows from the formula of the zstandardization that for the standardized footprint Fz fi of firm f by Provider i , it holds the following: where dfi is the value of a single impact driver by Provider i for firm f from either the set of common impact drivers Dfi,com or the set of exclusive impact drivers Dfi,ex,𝜇i is the crosssectional mean of all aggregated footprints by Provider i,𝜎i the standard deviation of the aggregated footprints by Provider i , and Ni the number of impact drivers (both exclusive and common) that Provider i is including. In our case, Ni is 4 for all providers.For easier readability, we define the first summand of Equation(3) as Dz fi,com and the second summand as Dz fi,ex . Consequently, we can decompose the difference Δz fi,j in the standardized footprints by Providers i and j for firm f in the following way: (2) Ffi =Dfi,com +Dfi,ex (3) F z fi =∑ dfi∈Dfi,com dfi − 1 Ni𝜇i 𝜎i ⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟ :=Dz fi , com +∑ dfi∈Dfi,ex dfi − 1 Ni𝜇i 𝜎i ⏟⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏟ :=Dz fi , ex TABLE 5 | This table shows the pairwise Pearson (Spearman in parentheses) correlations between the providers for each impact driver as well as Krippendorff's alpha measuring the agreement over all raters. Climate Change Land Use & Pollution Air Pollution Water Pollution Pearson (Spearman) correlation between Provider1 and Provider2 0.546 0.660 0.442 0.158 (0.667) (0.507) (0.651) (0.344) Provider3 0.002 0.311 − 0.006 (0.672) (0.553) (0.495) Pearson (Spearman) correlation between Provider2 and Provider3 0.083 0.403 0.039 (0.850) (0.718) (0.669) Krippendorff's alpha 0.211 0.458 0.442 0.064 5894 Business Strategy and the Environment, 2025 We define Δz fi,j,meas as the difference between Dz fi,com and Dz fj,com , stemming from divergence in the measurement of the common impact drivers, and Δz fi,scope as the difference between Dz fi,ex and Dz fj,ex , stemming from differences in the scope of each provider. As all ratings are normalized to have a mean of zero, the total difference between two ratings sums to zero as well. However, the firmspecific differences vary from zero. By using the variance as a measure of disagreement, we obtain summary statistics of these differences. Therefore, we take the variance over the sample of firms in Equation(4) and obtain the following: Now we can calculate the contribution of scope and measurement to the disagreement as Table6 presents the results of Equation(6) for each pair of providers, and the average for each provider. Overall, the contribution of measurement to the overall disagreement is 96.29% and thus considerably higher than that of scope (3.71%). As Providers 1 and 2 consider the same impact drivers, their disagreement stems only from divergence in measurement. Despite considering a unique set of impact drivers compared to the other providers, the average contribution of scope divergence to the overall disagreement is only 5.56% for Provider3. This can be explained by the contribution of each impact driver to the entire footprint that is presented in Table4. The impact driver Water Extraction, which is not considered by Providers 1 and 2, only has a very small impact on the overall footprint by Provider3, while the impact driver Land Use & Pollution, which is shared with bothother providers, contributes by far the most to the entire footprint. 5 | Determinants of the Disagreement In this section, we analyze how the disagreement between the impact driver and different firm characteristics is driving the overall disagreement. For that purpose, we follow Bauckloh etal.(2024) and calculate the two firmlevel disagreement measures sd and max−min. sd is the standard deviation of the three standardized footprints per firm, while max−min is the difference between the maximum and the minimum standardized footprint for each firm. The same disagreement measures are calculated at the impact driver level. First, we explain the variation in the overall sd and max−min with the respective disagreement measures at the impact driver level by running the following two regression models, using robust standard errors: where sd(impact driver)f and max − min(impact driver)f represent the standard deviation and range of the standardized assessments of the three providers for the respective impact driver for each firm f . 𝛼 is the constant term and 𝛽j(j=1, …, 4) are the coefficients estimated in the regression analysis. As the impact driver Water Extractionis exclusively considered by Provider3, it is not included in this part of the analysis. Table7 shows the results. The high R2 of 0.876 and 0.874 indicate that the overall disagreement can be explained to a large extent through the disagreement at the impact driver level, which underlines the results found in the decomposition (Section4.3). The disagreement measures of both impact drivers Climate Change and Water Pollution have coefficients significant at 1%- level. Climate Change (Water Pollution) displays a coefficient of 0.387 (0.352) for the sdregression and 0.382 (0.349) for the max− minregression. (4) Δz fi,j= Fz fi − Fz fj =Dz fi,com −Dz fj,com ⏟⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏟ :=Δz fi,j,meas +Dz fi,ex −Dz fj, ex ⏟⏞⏞⏞⏞⏟⏞⏞⏞⏞⏟ :=Δz fi,j,scope (5) Var( Δ z i,j ) = Cov( Δ z i,j , Δ z i,j ) =Cov(Δz i,j ,Δz i,j,meas )+Cov(Δz i,j ,Δz i,j,scope) (6) Contribution Measurement = Cov( Δ z i,j , Δ z i,j,meas ) Var(Δz i,j) Contribution Scope = Cov(Δz i,j,Δz i,j,scope ) Var(Δz i,j ) (7) sd f = 𝛼 + 𝛽1 ×sd(Land Use &Pollution) f +𝛽2×sd(Water Pollution)f +𝛽3×sd(Climate Change)f +𝛽4×sd(Air Pollution)f +𝜖 f (8) max − min f=𝛼+𝛽1× max − min(Land Use &Pollution) f +𝛽2×max−min(Water Pollution)f +𝛽3×max−min(Climate Change)f +𝛽4×max−min(Air Pollution)f +𝜖 f , TABLE 6 | This table presents the contribution of measurement and scope to the disagreement. Measurement Scope Panel A: Rater pairs Provider1 Provider2 100.00% 0.00% Provider1 Provider3 99.45% 0.55% Provider2 Provider3 89.43% 10.57% Average 96.29% 3.71% Panel B: Rater averages Provider1 99.73% 0.27% Provider2 94.72% 5.28% Provider3 94.44% 5.56%