I. Introduction
Over the past two decades, the field of Business and Human Rights (BHR) has experienced rapid normative expansion. From the endorsement of the United Nations Guiding Principles on Business and Human Rights (UNGPs) in 2011 to the emergence of mandatory human rights and environmental due diligence legislation across multiple jurisdictions, expectations of corporate responsibility for human rights and environmental impacts have increased. Despite some recent political and regulatory setbacks (e.g., the EU’s Omnibus Package), this broader trajectory reflects a multi-year trend that, while not without difficulties and reversals, demonstrates the steady consolidation of human rights within business and regulatory discourses.
Advancing this trajectory requires more systematic evidence on corporate human rights conduct—not only regarding the policies and commitments firms formally adopt, but especially concerning documented instances of misconduct. Such evidence is essential to assess whether and how corporate behaviour changes in response to private standards, public regulation and broader societal pressures. Beyond case-based evidence, research in this area has so far been largely based on environmental, social, and governance (ESG) indicators provided by private data vendors (e.g., Sustainalytics, Asset4/Eikon-Refinitiv, RepRisk and KLD/MSCI ESG) that are not designed to capture discrete abusive eventsFootnote 1 and present important data limitations that raise concerns about their reliability for BHR analysis.Footnote 2 Other BHR data sources either focus on corporate human rights disclosure, policies and practices, rather than on abusive events (e.g., the Corporate Human Rights Benchmark) or they are not openly available—as is the case of the Corporations and Human Rights Database, developed by Tricia Olsen and collaborators,Footnote 3 representing one of the most systematic efforts to document allegations of corporate human rights abuse and access-to-remedy mechanisms in Latin America.Footnote 4
Against this background, scholarship has often called for open-access BHR data specifically designed to document firm-level abusive events.Footnote 5 The REBALANCE Business-related Human Rights and Environmental (BRAVE) database responds to this call by providing a systematically coded, firm-level, event-based dataset documenting reported corporate involvement in human rights and environmental abuses. The BRAVE database is an output of the Horizon Europe REBALANCE project (# 101061342) and it is openly accessible via its website.Footnote 6
II. The REBALANCE Business-Related Human Rights and Environmental Database (BRAVE)
A. Scope and selection of monitored firms
The database covers all publicly listed European companies that appeared at least once in the annual ranking of the Top 100 Non-Financial Multinational Enterprises by Foreign Assets published by the United Nations Conference on Trade and Development (UNCTAD) in its World Investment Report between 2010 and 2020. This criterion yielded a final sample of 83 companies.
B. Codification procedure
To construct the database, we identified, validated and coded reported business-related human rights and environmental abuses (hereinafter, abuses), involving the monitored firms.Footnote 7
Our coding relied on a comprehensive body of qualitative materials (including news, reports, videos, etc.) collected and curated by the Business & Human Rights Centre (BHRC), a widely recognized and credible knowledge hub in the field.Footnote 8 The BHRC systematically collects and curates alleged business-related abuses involving more than 10,000 companies in over 180 countries, while excluding unfounded or unverifiable claims. Since the BHRC has been operational since 2000, our dataset covers the 2000–2020 period.
We conducted data collection between 2023 and 2025 and limited the dataset to reported abuses up to 2020 to account for the observed average three-year lag between the occurrence of an abuse and its public reporting, as identified by prior practice. BHRC provided direct access to a raw dataset comprising the full archive of documents related only to the firms in our sample, which served as our starting point for the codification process. The codification process involved a dedicated team of five researchers and over 50 research assistants engaged in both systematization and cross-validation activities.Footnote 9
Our team manually reviewed over 27,000 documents, including CSO/NGO reports, news articles, journal publications and even video documentaries. While most materials were in English, we employed translators for documents in other languages when necessary. The team assessed each document and then systematically coded the relevant allegations to ensure consistency and comparability. Importantly, we developed a codification protocol through dedicated consultations and focus groups with different human rights experts and stakeholders.Footnote 10 In the codification process, we included all allegations reported in the BHRC data, excluding only incidents that were unlikely to be attributable to firm-level decision-making, while retaining cases indicative of corporate neglect. For example, we omitted abuses stemming from workers’ misconduct for personal reasons, as well as accidents resulting solely from individual negligence or natural disasters. By contrast, incidents suggesting corporate neglect—such as failures in plant maintenance that resulted in harm—were included. Each event captures a distinct type and location of harm, allowing multiple concurrent abuses involving the same firm to be recorded. To avoid duplication, identical events reported across multiple sources were coded only once per company. Based on manual codification, BRAVE researchers have begun developing an AI-enabled codification process for future versions of the BRAVE dataset; however, this was not yet implemented at the time of writing.
C. Codification of reported abuses
i. General information
We assigned each company in our sample a unique identifier and name (labelled as Company ID and Company name) and, based on the BHRC and related sources, coded each discrete abusive event with a unique Abuse ID, together with a brief description of the incident (Brief description). Moreover, we recorded the country where each abuse allegedly occurred (Country) and the time frame reported in the sources, including the year of alleged onset (Year of alleged start) and the year of alleged termination (Year of alleged end). Additionally, we noted the year in which the abuse was first reported or denounced by the considered source (Year first reported), and whether it was part of a legal case brought against the company (Legal vs reported). Finally, we specified the type of source from which the information used to codify the abuse was first reported by the BHRC or other secondary sources (Type of source).
ii. Type of abuse
We classified human rights and/or environmental abuses into 19 macro-categories (Type), each comprising several subtypes (Subtype). These typologies have emerged through bottom-up coding, and each corresponds to a class of human rights, as defined by international human rights law.Footnote 11 The term ‘abuse’ is used in an umbrella sense to capture reported instances of corporate conduct with adverse human rights and/or environmental implications. The coding framework includes both outcome-based abuses (direct harms to rights-holders, such as deprivation of life or torture) and conduct-based abuses (corporate actions or omissions associated with such harms, such as corruption, complicity or tax avoidance). Although interrelated, these dimensions are recorded within a common event-based framework and remain separable for analysis.
Next, we present an overview of the 19 macro-categories of abuses. A detailed description of the associated subtypes is provided in Appendix A, which is available online. The first group of abuses of fundamental human rights concerns practices that violate individual dignity, integrity and freedom. These include child labour (1), which undermines children’s development and persists in global supply chains, including those linked to the energy transition; slavery (2), reflected in forced labour, debt bondage and coercive recruitment; and torture (3), involving severe physical and psychological abuse. We also capture trafficking (4), defined as the recruitment, transportation, transfer, harbouring or receipt of persons for the purpose of exploitation through coercion or abuse of vulnerability—and other illicit commercial practices.
Additional categories include deprivation of life (5) and violence and harassment (6), covering killings, sexual violence, repression of protest and fatal workplace incidents. Discrimination (7) captures structural inequalities based on gender, ethnicity, religion, caste, health status or sexual orientation. Next, labour rights abuses (8) include denial of leave, precarious employment, unjust dismissals, excessive or unpaid overtime, unsafe working conditions, recruitment fees and wage violations. Environmental harms (9) cover pollution, climate impacts, biodiversity loss and unsafe resource extraction. Health impacts (10) include pollution-related illness, restricted access to basic needs and unsafe products. Heritage impacts (11) refer to the destruction or appropriation of cultural and spiritual heritage. Displacement (12) involves forced removals without adequate consultation or compensation. Governance-related abuses include corruption (13) and complicity (14) with state or paramilitary actors, as well as involvement in weapons production (15) and tax avoidance or evasion (16), which undermine states’ fiscal capacity. Finally, civic and political rights abuses include intimidation (17), restriction of civil and political rights (18) and abusive legal action (19), such as Strategic Lawsuits Against Public Participation (SLAPPs), which are used to silence or deter accountability efforts.
iii. Type of victims
We identified the victims impacted by the abuse (Type of victims), grouped into five macro categories: (1) children, (2) workers and trade unionists, (3) communities, (4) consumers and clients and (5) activists and journalists.
iv. Severity of abuse
We assessed the severity of abuses using two dimensions: (1) whether the abuse involves physical integrity, and (2) whether it breaches non-derogable norms under international law. The online Appendix B provides the full classification of abuses by severity, which is summarised below.
Regarding physical integrity (1) we drew on international law and BHR scholarship,Footnote 12 and distinguished between physical and non-physical integrity abuses. Physical integrity abuses (1.a)—such as torture, enforced disappearance, extrajudicial killing and other forms of violence—are generally considered among the most serious violations of fundamental rights, as they violate jus cogens norms, which occupy the highest status in international law,Footnote 13 and directly undermine the essence of human dignity, an inviolable value at the core of human rights protection.Footnote 14 By contrast, non-physical integrity abuses (1.b)—for example, restrictions on freedom of expression, movement or assembly—typically entail less immediate harm to individual integrity,Footnote 15 though they may systematically erode civic and political freedoms.
Concerning the codification of abuses as violations of non-derogable norms (Non-derogable right) (2), we relied on a foundational principle of international human rights law whereby certain rights—such as the right to life, the prohibition of torture and freedom from slavery—are classified as non-derogable (2.a) under international law, meaning they cannot be suspended under any circumstances, including in states of emergency.Footnote 16 Conversely, derogable rights (2.b) are rights that individuals normally enjoy, but that may be temporarily restricted by authorities in exceptional circumstances, such as a state of emergency and serious security threats. In coding terms, these are treated as lower-severity rights.
v. Type of company involvement
Our classification of company involvement distinguishes different forms of corporate connection to adverse human rights impacts. Rather than adopting the UNGPs’ typology of corporate involvement (i.e., cause, contribute to or be directly linked to), we employed an international business perspective that is better suited to facilitating quantitative analyses of multinational enterprises and their governance challenges across diverse institutional contexts. Specifically, we distinguished between direct human rights abuses, which originate within the firm’s organizational boundaries—including headquarters and subsidiaries—and indirect human rights abuses, which arise through relationships with external actors connected to the firm via value chains, governmental interactions or other business relationships. This distinction, consistent with insights from prior research,Footnote 17 enables the systematic examination of how different organizational configurations, governance mechanisms and cross-border relationships shape firms’ exposure to, and management of, human rights risks. Accordingly, direct involvement (Company direct involvement) (1) occurs where a company—including the parent company and its subsidiaries—causes an adverse human rights or environmental impact through its own activities, including both actions and omissions attributable to the company. By contrast, indirect involvement (Company indirect involvement) (2) arises when an adverse impact is caused by a third party—including an entity in the company’s global value chain (Company indirect involvement through GVC), a local government or government agency (Company indirect involvement through government agencies) or other entities such as armed groups or other non-state actors (Company indirect involvement through other entities)—but is connected to the company’s operations, products or services through a business relationship.
III. User guidance and data limitations
Beyond its value as a standalone source of information on corporate misbehaviour, BRAVE can be integrated with a wide range of firm-level datasets, including financial, ownership, governance, internationalization and ESG data. This enables researchers to examine the organizational, institutional and economic determinants and consequences of corporate misconduct across firms, industries, countries and time. By providing systematic, longitudinal and comparable evidence on alleged abuses, BRAVE fills a critical data gap in the BHR field and creates new opportunities for quantitative, multi-level and cross-disciplinary research on the drivers, dynamics and outcomes of corporate misconduct.
The BRAVE database adheres to the FAIR principles (Findable, Accessible, Interoperable, and Reusable) in line with Horizon Europe guidelines. It is openly available for download online, together with its codebook, and updated versions of the dataset are regularly uploaded to the same website. We recommend the following guidelines when using BRAVE data for empirical analysis.
First, like all datasets built on publicly available allegations, BRAVE necessarily captures only observed and reported abuses. It is therefore subject to both false negatives (unreported or undetected abuses) and false positives (allegations that, despite careful scrutiny, may still misrepresent underlying events). Although we rely on a high-quality and credible primary source (i.e., BHRC) and systematically cross-check information against additional sources, measurement error cannot be fully eliminated. Users—including in econometric applications—should therefore maintain appropriate caution and transparency regarding these inherent limitations. Moreover, we do not code companies’ responses to allegations, given the difficulty of validating their reliability.
Second, limitations are accentuated by structural features of the underlying reporting environment. While the BHRC methodology improves reliability by filtering out unfounded claims, it does not eliminate asymmetries in information availability. Shifts in thematic priorities—such as labour rights, natural resources, technology accountability and human rights defenders—may affect the visibility of different types of abuses over time. Additionally, the gradual expansion of country expert networks over time implies that earlier periods are likely less comprehensively documented, generating potential time trends in reporting. We therefore recommend that analysts account for these reporting dynamics using appropriate econometric controls, such as year fixed effects or time trends.
Finally, reporting biases are a pervasive feature of all ESG and BHR datasets, including BRAVE: the likelihood that an abuse enters the public domain is itself systematically shaped by firm, sector and country characteristics. If used in quantitative analyses, BRAVE data require empirical specifications to control for these differences in reporting intensity—for example through firm size, media visibility, country-level conditions and industry fixed effects—so as to better isolate variation in underlying abusive conduct rather than variation in its observability.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/bhj.2026.10058.
Attribution of the Dataset, Disclaimers and Contacts
The BRAVE dataset can be downloaded from here: https://rebalanceproject.org/data/. Users of the dataset are requested to cite this article as the primary reference. While every effort has been made to ensure the accuracy and reliability of the data, the database may contain omissions or errors. The coders have relied on available secondary sources to verify reported information to the best of their ability. As noted, the dataset codifies reported allegations of human rights and/or environmental abuses. Researchers seeking further information are invited to contact: rebalanceprojecteu@gmail.com.
Acknowledgements
The authors would like to thank the Business & Human Rights Centre (BHRC) for providing access to the data that constitute the foundation of the BRAVE database. Thanks go to Toby Rubenstein; Amelie Freund; Jordi Vives Gabriel; Santiago Calvo Lopez; Saskia Wilks; Johannes Blankenbach and Jennifer Teufel. Suggestions by Rebalance team members Flaviano Bianchini; Andrew Crane; Chiara Macchi; Nadia Bernaz; Olena Uvarova; Francesco Zirpoli; Tim Bartley and Antonella Angelini are also gratefully acknowledged. We are also grateful to the network of research assistants whose contributions made systematic coding of allegations possible, including Guillermina Actis; Elisa Aiello; Giulio Arioti; Laura Aryn; Pietro Bagnoli; Elisa Baselli; Ginevra Bertini; Gulnaz Bolatova; Benedetta Bonato; Pietro Capaccioli; Martina Checchi; Pietro D’Ercole; Aldiron Degjoni; Roberto Deglon; Arianna Della Rosa; Martina Dinucci; Margherita Ferracci; Francesca Frigato; Elena Fuschi; Federico Gallo; Alessandra Geri; Fidan Hasanzade; Anik Sofia Lammel; Leonardo Lista; Alessia Malacarne; Carolina Mancuso; Aurora Maneschi; Oliver James Manfrè; Carolina Menconi; Aruzhan Mirasheva; Luca Tobia Monaco; Federico Musumeci; Viola Mucci; Francesca Orsi; Maria Costanza Palmisano; Sofia Panu; Ginevra Papini; Polina Patiuk; Sabrina Kathryn Pedigo; Daryl Perry; Giovanni Piccoli; Mafalda Piegaia; Alessandro Pozzetti; Federico Raponi; Eleonora Rexhepaj; Giulia Romano; Giulia Rossello; Venera Serkina; Niccolò Spampani; Sara Lucia Tedesco; Sorin Toma; Alice Tortelli; Lisa Wu; Leila Yow; Kule Yosia; Riccardo Zizzi, among others. Any errors or omissions remain the responsibility of the authors. This research is funded by the European Commission under the Horizon Europe Framework (Rebalance Project GA No 101061342).
Open access
