05.05.2026 |
News
Article in Nature quantifies the robustness and replicability of articles from leading Economics and Political Science journals
A large-scale analysis of 110 articles from leading economics and political science journals shows that approximately 85% of published findings are computationally reproducible, while 72% of statistically significant estimates are robust to alternative specifications. These findings indicate strong, though not perfect, reliability in empirical research, with high-quality journals’ mandatory data-sharing policies facilitating verification.
Martín Brun from Tampere University is one of co-authors in the project.
A recent volume of Nature focuses on the replication, robustness and reproducibility of results across different social and behavioral sciences. Findings reveal encouraging strengths and persistent vulnerabilities in how scientific knowledge is built, and highlight how progress depends not only on discovery, but on willingness to question, test and rebuild previous knowledge.
Science can move forward only when it is clear which findings persist and which do not. Systematic examinations of previous research prove to be much valuable, as they not only strengthen their reliability but also promote self-correction in future research. Despite their aggregate benefits, these exercises have rarely been conducted on a scale in the social sciences.
Brodeur et al. (2026) is one of the studies featured in a recent volume of Nature dedicated to the reproducibility, replicability and robustness of research in social and behavioral sciences. These studies provide a snapshot of the strengths and weaknesses in fields such as Economics, Political Science, Sociology, Psychology, and Education.
Martín Brun (Tampere University) is one of the co-authors in this project resulting from the collective effort of a large network of researchers around the world. He collaborates in various projects within this network, with the goal of advancing the robustness of empirical research.
Brodeur et al. (2026) focuses on 110 articles published in top journals in Economics (i.e., Quarterly Journal of Economics, Journal of Political Economy, Review of Economic Studies, Economic Journal, American Economic Review and other American Economic Association journals) and Political Science (i.e., American Journal of Political Science, American Political Science Review, Journal of Politics) in 2022 and 2023.
The sample of articles is composed by those that comply with data and code availability policies. The researchers chose among these articles to analyze, based in subfield of specialization and methods used among others. Researchers got familiar with the original articles, its data and codes, and produced a reproduction report. These reports focus on the key findings discussed in the original articles, and contain a previously agreed set of statistics. For each original article, the researchers reported whether the original results are computationally reproducible, how the effect sizes change with different types of robustness checks, as well as various characteristics of the research design. This harmonized dataset enables to conduct cross-article statistical analyses, quantify the overall rates of reproducibility and robustness, and examine potential determinants of variation across studies.
The analysis finds that more than 85% of the articles analyzed are computationally reproducible, which means that the results originally reported can be reproduced. Moreover, 72% of the key statistically significant results maintain significance and direction in the robustness checks performed. These figures stand tall among the fields covered in the same Nature volume (Miske et al., 2026), and contrast previous studies documenting low reproducibility in the field (Gertler et al., 2018). This may reflect the effectiveness of journal editorial policies that have introduced data editors and mandatory availability of replication packages.
Although the journal sample is selective, the findings are encouraging and suggest a high level of computational reproducibility. The increasing adoption of common editorial policies that encourage more rigorous methodologies, deter questionable research practices, and promote strong data-sharing norms, provide optimism about the strengthening of trust in findings in the field.
More information
Martín Brun
Postdoctoral Research Fellow
Faculty of Management and Business
Tampere University
E-mail: martin.brun@tuni.fi
The article ‘Reproducibility and Robustness of Economics and Political Science Research’ was published in Nature in April 2026.
Brodeur, A., Mikola, D., Cook, N. et al. Reproducibility and robustness of economics and political science research. Nature 652, 151–156 (2026). https://doi.org/10.1038/s41586-026-10251-x
