World Bank headquarters in Washington, DC.Credit: J. David Ake/Getty
Reproducibility has become a major challenge in the social and behavioral sciences. Research findings can be difficult to replicate, meaning that independent researchers are sometimes unable to reach the same conclusions as the original study team—even when they use the same data and analytical methods. When results cannot be reproduced, confidence in the strength and reliability of the evidence is weakened. Comprehensive analyses of reproducibility across scientific disciplines published earlier this year have reinforced these concerns (O. Miske et al., Nature 652, 126–134; 2026; A. Brodeur et al., Nature 652, 151–156; 2026).
Reproducibility problems affect every area of science. But when research in the social and behavioral sciences cannot be independently verified, the consequences can extend directly to public health, economic development and human well-being.
The World Bank provides a useful example. Based in Washington, DC, the international financial institution aims to reduce poverty and promote shared prosperity around the world. Each year, it distributes approximately US$120 billion in loans and grants, mainly to low- and middle-income countries. The bank also employs researchers whose studies and reports help shape policy. A separate evaluation department assesses whether the bank’s programmes and funding achieve their intended goals.
Reliable research in society, behavior, and science
Researchers at the World Bank, along with external scientists who evaluate its operations, recognize the importance of making research reproducible. Beginning in 2023, the bank introduced measures to strengthen research transparency. From 2025 onwards, all published research results are expected to include a “reproducibility package”.
These packages explain how researchers reached their conclusions using the original dataset. They also provide the underlying code, data and other supporting materials when those materials can be shared. Each package is citable and includes instructions for rerunning the analysis and checking the results.
Before publication, a dedicated in-house team reviews the materials. Packages that pass the internal assessment receive a badge confirming that the research is reproducible. This designation can help journal editors and peer reviewers assess the quality and transparency of submitted research. For public institutions, openly sharing the materials behind research is an important way to build trust and demonstrate accountability.
In June, the World Bank announced that more than 500 reproducibility packages had been released since 2023 (go.nature.com/4gqjen8). “While the first 500 packages took nearly three years to arrive, we expect the next 500 to arrive even faster,” the bank’s researchers wrote. The bank has also launched a repository dedicated to hosting reproducible research packages (go.nature.com/4ytc3he).
Beyond the World Bank
The practice of publishing reproducible research materials is gradually spreading beyond the World Bank. The Inter-American Development Bank, which provides about $35 billion in loans and grants each year to countries in Latin America and the Caribbean, has begun releasing reproducibility packages alongside technical notes, working papers and reports.
High standards for research transparency should not be limited to development banks. Nature spoke with staff members at other international development organizations, including those within the United Nations system, that publish research-based reports. Some said they were working to verify evidence internally or conduct source checks. However, none appeared to be taking an approach as systematic as that of the development banks.
International organizations should consider adopting similar standards. Preparing and reviewing reproducibility packages requires time and money, which may seem difficult to justify while many institutions face budget reductions. Yet publishing research that cannot be independently checked can be far more costly. For organizations that influence public policy and allocate billions of dollars, weak evidence should not be an acceptable alternative.
More introspection in research can lead to better science
Some research findings—such as recommendations about standards or policy principles—may not require extensive supporting materials. But whenever it is relevant and technically possible, international organizations should make the data, code and documentation behind their analyses publicly available. They should also allocate staff time to reviewing those materials so that independent scientists can reproduce the work successfully.
Creating these norms would improve the reliability of evidence used by international institutions. It could also lead to better outcomes for organizations whose missions affect people around the world, including efforts to control disease, reduce poverty and respond to climate change.
New technologies create additional challenges for research transparency. A representative of the Inter-American Development Bank told Nature that officials were concerned artificial-intelligence tools could scan and misinterpret code or data released through reproducibility initiatives. The bank is considering how to address that risk and whether data and code should continue to be released at an early stage.
Regardless of the approach institutions ultimately choose, the principle is clear: lending and policy decisions based on robust, reproducible research are preferable to decisions built on uncertain evidence. Every organization that produces or relies on research has a responsibility to protect its quality, transparency and credibility.
Source: www.nature.com