Reproducibility and robustness of economics and political science research
5 hours ago
- #reproducibility
- #social-science
- #robustness
- Over 85% of published claims from 110 articles in economics and political science were computationally reproducible, highlighting the reliability of findings under mandatory data and code sharing policies.
- In robustness checks, 72% of statistically significant estimates remained significant with the same direction, and the median reproduced effect size was 99% of the original, indicating strong robustness of results.
- Six independent teams examining hypotheses on robustness determinants found that more experienced researchers reported lower robustness levels, while robustness did not correlate with author characteristics or data availability.
- The study emphasizes the importance of computational reproducibility as a diagnostic tool for strengthening science, promoting self-correction, and informing policy-making, though large-scale reproducibility efforts in social science remain limited.