How Scientific Are the Social Sciences?

According to Science.org, the findings from about half of all social science studies can’t be confirmed through independent checks. The large-scale SCORE project (Systematizing Confidence in Open Research and Evidence) analyzed more than 100 papers published before 2018 in leading journals in business, economics, education, political science, psychology, and sociology. The independent replication success rate came out to just 49%. The SCORE results, published in Nature, showed that only 49% of the 164 papers assessed were successfully reproduced.

According to the Karolinska Institute, 865 researchers took part in SCORE and examined nearly 3,900 scientific papers published between 2009 and 2018. The studies were evaluated across three dimensions: replicability, reproducibility, and robustness. A study is considered replicable if analysis of new, subject-matter-relevant data produces a similar result. Reproducibility means other scientists can reach the same conclusion using the authors’ original data and methods. Robustness is confirmed when a different analytical method is applied to the original data, but the answer remains unchanged.

The exact reproducibility rate was 54%. As for robustness, independent analysis of each selected paper by at least five specialists found that in only 34% of cases did all experts fully agree on the correctness of the original conclusions.

The project team hoped to identify clear markers that could be used to judge a paper’s credibility in advance. As Science.org reports, those efforts were not notably successful: no reliable universal indicator was found. Even artificial intelligence algorithms were unable to predict with high accuracy which studies would pass a follow-up check.

The only factor that showed a strong correlation with successful reproducibility was data availability. Papers whose authors provided open access to their underlying data and code—only about one-third of the sample—showed a much higher rate of result confirmation.

The SCORE leadership emphasizes that a single replication failure does not indicate fraud or professional negligence. The reasons for such failures are complex, and an unconfirmed result should be treated as new evidence that requires further study.

Still, independent experts agree on one point: the scientific community needs systemic reform. To build greater trust in science, the culture of research and the grant-funding system must change, shifting the focus away from the number of papers published and toward methodological rigor and transparency.