Major International Review Finds Half of Social Science Results Cannot Be Replicated

Stockholm – September 11, 2026

A comprehensive international review of social science research has found that nearly half of published results cannot be independently replicated, raising critical questions about reproducibility in fields spanning economics, psychology, marketing, and organizational behavior.

The analysis, part of the SCORE (Systemizing Confidence in Open Research and Evidence) program, represents the largest investigation of its kind in social and behavioral sciences. The study examined 3,900 scientific articles published between 2009 and 2018 across 62 peer-reviewed journals.

Study Design and Scale

The research team included 865 researchers who employed three different methodological approaches to assess scientific reliability. The replication study tested the same research questions using fresh data. The reproducibility study re-analyzed original datasets using the same statistical methods. The robustness study had multiple researchers independently analyze the same data using their preferred analytical approaches.

Only 144 of the 600 papers in the random sample—roughly 24 percent—had shared their underlying data openly. When the team obtained data from authors or reconstructed datasets from source materials, they were able to assess 143 articles in depth.

Key Findings on Reproducibility

When data and code were publicly available, reproducibility rates climbed substantially. Approximately 74 percent of studies could be reproduced to an approximate degree, and 54 percent matched the original findings exactly. With full data and code sharing, these figures rose to 91 percent and 77 percent respectively.

The robustness study analyzed 100 articles where at least five researchers independently tested the same hypotheses. Only one-third of the new analyses produced results very close to the original reports. However, three-quarters of the analyses reached the same overall conclusions as the published studies. In roughly 25 percent of cases, no clear effects were detected at all.

"This shows that transparency is key to achieving credible research results," said Gustav Nilsonne, associate professor of neuroscience at Karolinska Institutet, who co-led the robustness study and is a co-author of all three papers.

Discipline and Journal Policy Differences

Reproducibility varied significantly across fields. Political science and economics papers showed higher rates of data sharing and reproducibility compared to other disciplines. This difference correlated with journal policies—journals that mandate data sharing saw substantially more transparent research practices.

In 2018, the final year of the sample period, just 27.4 percent of journals had data sharing requirements, 22.6 percent had code sharing requirements, and only 3.2 percent conducted reproducibility checks. By mid-2025, those figures had improved to 51.6 percent, 46.8 percent, and 19.4 percent respectively.

Open Data and Scientific Trust

The study highlighted open data as the most significant factor affecting reproducibility. The unavailability of author data was the primary barrier to verification attempts. When data was available but imperfectly shared, reproducibility was incomplete but measurable.

The collaboration was led by the Center for Open Science, Pennsylvania State University, TwoSix Technologies, University of Southern California, and Eötvös Loránd University. The program received funding from DARPA, the US Defense Advanced Research Projects Agency.

Implications for Research Practice

The three Nature papers collectively suggest that implementing verification measures is essential for maintaining trust in social science research. The authors note that reproducibility checks have become more common but remain far from universal.

"This is the world's largest research project to date investigating the reliability of reported scientific results, and an example of how large-scale collaborations can address questions that no single research group could answer alone," Nilsonne concluded.

The findings arrive as academic institutions and funding agencies increasingly prioritize open science practices. The study provides empirical evidence that data sharing policies directly correlate with research reproducibility.

Research Collaboration

The analysis brought together researchers from Karolinska Institutet, Stockholm University, and institutions worldwide. Lead authors included Olivia Miske, Balazs Aczel, and Andrew Tyner, who co-authored the three complementary papers examining reproducibility, robustness, and replicability.

Source: Karolinska Institutet News, September 11, 2026. Nature, "Investigating the reproducibility of the social and behavioural sciences," April 1, 2026. Nature, "Investigating the analytical robustness of the social and behavioural sciences," April 1, 2026. Nature, "Investigating the replicability of the social and behavioural sciences," April 1, 2026.