Investigating reproducibility in the social and behavioral sciences
Publication id: miske-2026-score-reproducibility
Status: verified
Citation: Miske, O., ..., Shakya, S., ... et al. (2026). Investigating the reproducibility of the social and behavioural sciences. Nature. Download PDF
Facts¶
Policy hook¶
Beyond replicability (can results be reproduced with new data?), reproducibility (can analyses be reproduced from original data?) matters for assessing research reliability. How reproducible are social and behavioral science analyses?
Main finding¶
Systematic assessment of reproducibility in social and behavioral sciences examines whether published analyses can be reproduced from original datasets and code.
Data and setting¶
Reproducibility project examining code and data availability for published studies in social and behavioral sciences; Varieties of Democracy and related databases.
Research design (plain language)¶
Systematic audit of published papers assessing whether researchers can reproduce original analyses using available code, data, and documentation.
One caveat¶
Reproducibility assessment requires researcher effort and cooperation; some failures may reflect inadequate documentation rather than fundamental analytical errors.
PDF or DOI¶
Why it matters¶
Reproducibility is a foundational principle of science: others should be able to verify published analyses by running the same code on the same data. Yet many published papers lack available data or code, making verification impossible. This systematic audit of reproducibility in social and behavioral science is crucial because policy often relies on these findings—if analyses can’t be verified, confidence in policy is undermined. For researchers, the audit provides accountability: shared data and code enable peer scrutiny and build trust. For funding agencies and journals, reproducibility metrics can inform publication standards and funding priorities. For policymakers, the results indicate which studies’ findings can be independently verified. The finding that many analyses lack reproducibility doesn’t necessarily discredit results—missing data might reflect privacy concerns or institutional constraints—but it does signal the need for transparency standards. Open science practices (data sharing, code availability) directly enable reproducibility. Combining reproducibility assessment (can analyses be verified?) with replicability testing (do new data confirm results?) gives a complete evidence quality picture.