Our understanding of reparations builds on the definition developed by the Office of the United Nations High Commissioner for Human Rights (OHCHR), but extends it beyond a strictly human rights framework. We include a broader range of measures addressing “unjust harms”, including cases where reparations are not explicitly framed as responses to rights violations or are not situated within transitional justice contexts.

At the same time, we emphasize the exceptional nature of reparations. This reflects the widely shared moral view that reparations should not have to exist and the harms they address should never have occurred.

We therefore define reparations as: “exceptional measures that aim to redress past wrongs by improving the situation of those who have been harmed”.

Our definition of reparations payments draws on the UN’S Basic Principles and Guidelines on the Right to a Remedy and Reparation. These identify several forms of reparations, including restitution, compensation, rehabilitation, satisfaction, and guarantees of non-repetition. Across these categories, we focus specifically on financially quantified transfers of wealth, which we call “reparations payments”. We do not distinguish between payments labelled as “compensation” or “satisfaction”, because their economic and symbolic meanings are often inseparable.

The RPD collects cases of reparations payments pledged and/or paid from 1 January 1945 onwards. This starting date was chosen to reflect the changing nature of reparations in the second half of the twentieth century. The database includes cases identified during the data collection process up to 31 December 2024.

Data collection for the database was conducted in two main steps:

1. integrating information from existing databases, and

2. conducting semi-systematic literature review of online media sources.

Integration of existing databases

As a first step, we reviewed eight existing databases and repositories that document reparations, transitional justice measures, or peace agreements. Relevant entries from these sources were reviewed and integrated into the RPD. These databases are:

• The Political Apologies and Reparations Archive

• The Transitional Justice Evaluation Tools (TJET)

• Reparations in the United States

• The Nuhanovic Foundation’s Reparations Database

• The Transitional Justice Data Base (TJDB)

• The Peace Agreements Database (PA-X)

• The ICC Legal Tools Database

• The Peace Accords Matrix (PAM)

Literature review

To identify additional cases, we conducted a semi-systematic literature review, adapting the “Online Media Search Guide” developed for the Political Apologies Database (Schaafsma & Zoodsma, 2021) to fit the scope of the RPD.

First, we searched for key terms related to reparations payments - including “paid reparations”, “received reparations”, “reparations agreement”, “reparation payments”, and “reparations payments” - using Google News and Lexis Nexis (Nexis Uni).

Second, to improve global coverage, we conducted targeted searches for every country and economy listed by the World Bank. Each country or economy name was combined with the term “reparation payments” using Google and Google News. This step aimed to ensure that cases from all world regions were captured.

Additional cases and verification

In addition to the structured searches, further cases were identified incidentally during the research process through related literature and references.

All identified entries were then reviewed and screened for duplicates, and additional searches were conducted where necessary to verify information and complete missing data.

During the data collection phase, cases were included based on the project’s working definition of reparations payments, alongside a set of temporal, conceptual, and evidentiary criteria.

Inclusion Criteria

Cases were included in the database when they met the following conditions:

• Definition of reparations payments: The database is guided by the project’s definition of reparations payments. This includes financially defined measures, primarily cash transfers, as well as certain in-kind measures when they are economically structured or clearly monetized as forms of compensation or reparation. The dataset is not limited strictly to compensation in a narrow legal sense.

• Defined scope of harms: Inclusion was guided by a collaboratively developed list of harms compiled by the research team. This list served as a reference framework to ensure consistency in identifying cases and assessing whether payments were intended to address historical injustices, violations, or other forms of unjust harm within the scope of the project’s definition of reparations. A complete list and definitions of the included harms are provided in the RPD codebook.

• Temporal scope: The RPD covers reparations cases committed between 1 January 1945 and 31 December 2024.

• Actors: The identification process encompassed a broad range of actor types, including states, individuals, and institutional actors, reflecting the diversity of entities involved in reparations processes.

• Victim- and survivor-centred approach: In cases where classification was ambiguous - particularly when distinguishing between routine compensation and exceptional reparative measures - entries were included only when at least one credible source explicitly framed the measure as reparations.

• Case-by-case assessment: When uncertainty remained, inclusion decisions were made through careful case-by-case evaluation, based on available evidence and alignment with the project's definition of reparations payments. These decisions were guided by collective team judgment and remain open to revision as new information becomes available.

Exclusion Criteria

Standard or routine compensation schemes not explicitly framed as reparations were excluded from the database, as they do not reflect the exceptional nature of reparations as defined within the scope of the RPD.

While efforts were made to identify and include private or corporate payments where relevant, such cases were subject to careful review and included only when clearly documented and explicitly described as reparations in reliable sources, such as existing reparations-focused databases and other authoritative references.

This approach - and the analytical challenges encountered - reflects both the increasing role of private actors in contemporary reparative measures and the growing blurring of boundaries between routine compensation and exceptional reparations, which can make classification complex.

To support the construction of the RPD, as well as data analysis and visualization, the team explored the use of automation scripts from the early stages of methodology development, in consultation with reparations scholars and members of the Advisory Board.

A mixed approach was ultimately adopted. All entries were first reviewed and assessed individually by the research team. Selected data fields were subsequently populated automatically in cases where interpretation was minimal (e.g. regional assignment and currency conversion).

This approach increased efficiency and reduced time-consuming manual tasks, while maintaining careful human oversight of all classification decisions.

Automation processes were implemented in Python, with the assistance of AI tools (such as ChatGPT) to support script development and troubleshooting. For more information on the variables implemented through automation and the scripts used, please consult the RPD Codebook and the GitHub repository.

The team will continue to explore and refine the use of automation in future versions of the RPD, with the aim of further improving efficiency while maintaining methodological rigor and careful human oversight.

The database aims to be as comprehensive as possible and, to our knowledge, represents the most extensive repository of reparative payments compiled to date. However, full comprehensiveness is not achievable. The RPD reflects the most thorough coverage possible within the scope and constraints of this project.

The dataset and its analysis should therefore be interpreted in light of the following limitations:

• Scope limitations: Private compensation schemes, settlements, court-ordered payments, and contributions by corporations or other non-state actors could not be systematically included. This is largely due to limited transparency, inconsistent reporting, and challenges in accessing reliable information.

• Language and search limitations: Data collection was conducted primarily in English and relied heavily on Google-based searches and accessible digital sources. As a result, sources in other languages, non-digitized archival materials, and information from regions where Google is less widely used or accessible may be underrepresented.

• Data availability and quality: The availability and quality of data vary significantly across cases, countries, and historical periods. In some instances, key information—such as payment amounts, numbers of beneficiaries, or implementation dates—was incomplete, inconsistent, or not publicly available.

• Approximations and interpretative decisions: In certain cases, the available information required approximations or interpretative judgments when reporting data. Where precise figures or details were unavailable or inconsistent across sources, the research team relied on the best available evidence and applied consistent methodological criteria to ensure comparability across entries.

• Currency conversion, valuation, and reported amounts: Financial amounts reported across different currencies and time periods required conversion to enable comparison. These conversions rely on available exchange rate data and may not fully capture differences in purchasing power, inflation, or broader economic conditions. In addition, some figures are estimates or partial totals, meaning that the overall value of certain reparations programs may remain uncertain.

• Evolving information: Reparations programs continue to evolve, with new initiatives emerging and existing ones expanding or changing over time. While the RPD seeks to document these developments as comprehensively as possible, it functions as a repository rather than a continuously updated observatory. Entries therefore reflect the best available information at the time of compilation and may not capture subsequent developments.