The Breakthrough Findings from the University of Stirling Study

A groundbreaking investigation led by researchers including Dr. Arran Reader from the University of Stirling has delivered compelling evidence on the critical role of data sharing in ensuring scientific reproducibility within the social and behavioural sciences. Published on April 1, 2026, in the prestigious journal Nature, the paper titled "Investigating the reproducibility of the social and behavioural sciences" examined 600 quantitative research papers spanning fields like psychology, economics, sociology, political science, education, and business. This large-scale effort, part of the US Defense Advanced Research Projects Agency (DARPA)-funded Systematizing Confidence in Open Research and Evidence (SCORE) program, underscores a simple yet profound truth: when data and analysis code are openly shared, the vast majority of results hold up under scrutiny.

The study's core revelation is that approximately 74% of evaluated claims could be reproduced at least approximately (within 15% of original effect sizes or 0.05 of p-values), and 54% precisely matched the original numbers when data was accessible. However, access was the bottleneck—only 24% of papers provided data, and just 20% shared both data and code. When both were available, reproducibility soared to 88% approximate and 75% precise, plummeting to a mere 11% precise when analysts had to painstakingly reconstruct datasets from public sources.

Unpacking the Reproducibility Crisis in Social and Behavioural Sciences

The reproducibility crisis has loomed large over scientific fields for over a decade, with psychology's infamous 2015 replication project revealing that only about half of landmark studies could be replicated. Social and behavioural sciences, reliant on complex human data, surveys, experiments, and statistical modeling, face amplified challenges: small sample sizes, p-hacking (manipulating analyses for significance), publication bias favoring positive results, and opaque methods. In Europe, where social sciences inform EU policies on migration, inequality, and mental health, unreliable findings risk misguided decisions—from welfare reforms to pandemic responses.

Statistics paint a stark picture: pre-2020 replication rates in social sciences hovered around 40-60%, lower in psychology (36-50%) and sociology, higher in economics (60-70%) due to stricter econometrics standards. The Stirling-led study provides the most comprehensive audit to date, confirming that while errors occur, they are often fixable with transparent access—yet most papers remain locked away, hindering verification and cumulative progress.

How the Study Was Conducted: A Gold Standard Methodology

The SCORE project's reproducibility arm drew a stratified random sample of 600 papers from 62 high-impact journals published 2009-2018, ensuring representation across disciplines. Independent expert analysts—865 researchers worldwide—contacted authors for data and code. For non-responders or unavailable materials, some datasets were reconstructed from original sources like public surveys.

Reproducibility was rigorously defined: precise if statistical outputs matched exactly; approximate if close enough to rule out major discrepancies. Over 551 claims were tested, inverse-weighted by paper to avoid bias from multi-claim studies. This mirrors real-world verification, revealing not just accuracy but documentation quality. Failures stemmed from ambiguous code, software mismatches, or incomplete data cleaning steps—not necessarily flawed science.

Key Statistics: Data Sharing Dramatically Boosts Success Rates

These figures highlight data sharing's transformative impact. Political science and economics led with superior rates, thanks to journal mandates; psychology and sociology lagged. Positively, journal data policies doubled from 27% (2018) to 52% (2025), correlating with rising reproducibility in newer papers.

Field Variations and Temporal Trends in European Contexts

Economics benefits from data archives like the American Economic Association's; political science from mandates in American Journal of Political Science. In Europe, similar trends emerge: UK Economic and Social Research Council (ESRC) requires data deposit, boosting compliance at institutions like LSE and Oxford.

• Economics/Political Science: 60-80% reproducibility where data shared.
• Psychology/Sociology: 40-60%, hampered by sensitive human data.
• Post-2015 papers: 10-15% higher rates, reflecting open science push.

Dr. Reader notes: "Three-quarters precisely reproduced when data and code available highlights sharing's importance."

Photo by Ben Garratt on Unsplash

Implications for Trust, Policy, and Public Confidence

Reproducibility is foundational: unverifiable results erode trust, slow meta-analyses, and undermine evidence-based policy. In Europe, where social sciences guide €100bn+ EU cohesion funds and behavioral nudges in climate action, opacity risks inefficiency. The study reassures—most science is solid—but mandates verification tools like journals' code checks.

Olivia Miske (Center for Open Science) emphasizes: "Openness aids quality control, spotting mistakes without distrusting researchers." Enhanced practices build public faith amid misinformation eras.

European Higher Education's Open Science Landscape

Europe leads globally via Horizon Europe (2021-2027, €95.5bn), mandating "as open as possible" data under FAIR principles (Findable, Accessible, Interoperable, Reusable). The European University Association's (EUA) 2021 survey of 300+ universities found 80% have Research Data Management (RDM) policies, 60% data repositories, but social sciences lag due to privacy.

UK's Stirling launched Stirling Open Research and Scholarship (SORS) Network and ReproducibiliTea club for training. Netherlands' Utrecht University mandates OS plans; Germany's DFG funds RDM hubs. Yet, only 40% social science projects fully comply with data sharing.

Challenges: GDPR and Sensitive Data in Social Research

General Data Protection Regulation (GDPR) protects personal data, vital for behavioral studies (surveys, experiments). Overcompliance hampers sharing—e.g., anonymization fears block 30% potential deposits. Pseudonymization, federated access (query without download), and ethics waivers help, as in UK's UK Data Service (holds 7,000+ datasets).

• Legal bases: Consent, public interest for science.
• Tools: CESSDA (Consortium of European Social Science Data Archives) ERIC shares 50,000+ studies securely.
• Stats: 25% drop in shared sensitive data post-GDPR.

Case Studies: Successes from European Universities

At University of Stirling, SORS fosters workshops; ReproducibiliTea discusses preregistration. Oxford's Centre for Reproducible Science trains 1,000+ yearly. CESSDA's Metadata Aggregator enables cross-EU reuse, e.g., ESS data reused in 10,000+ papers.

Amsterdam's BIAS project audits social biases in data sharing, finding underrepresentation of Global South voices.

Future Outlook: Mandates, Training, and Cultural Shifts

Journal mandates rose to 52%; expect 70% by 2030. Horizon Europe's DMPs (Data Management Plans) evolve to enforce sharing. Universities invest: 65% plan RDM training per EUA. Incentives like badges, grants prioritize OS. AI tools automate checks, but human oversight key.

Dr. Gemma Learmonth (Stirling): "Reproducibility improves design, documentation, sharing." Actionable: preregister, use OSF.io, cite data DOIs.

Photo by Antoine Schibler on Unsplash

Actionable Insights for Researchers and Institutions

• Adopt GitHub/OSF for code/data.
• Train via ReproducibiliTea-style clubs.
• Leverage CESSDA/Horizon for compliance.
• Fund RDM roles—ROI via reusable data.

For Europe's 5,000+ universities, embracing data sharing fortifies social sciences against crises, accelerates discovery, and positions graduates for trustworthy careers. Explore research positions advancing open science at leading institutions.