Social science reproducibility: Not great, but not as bad as reported?

(credit: wingedwolf)

Social science, including behavioral economics, has recently come under fire as failing to generate studies with reproducible results. The Reproducibility Project made waves in August of 2015 when it announced that reproducing social science experiments is very difficult—not because the original studies were difficult, but because published findings often aren’t as strongly backed by the data as the original authors claimed.

A recent issue of Science contains three articles that alternately contest and defend the reproducibility of social sciences. Critics argue that the true reproducibility rate of articles in these fields is much higher than was initially reported; defenders, including some researchers from the Reproducibility Project, say that the reproduciblity rate is truly less than half.

Though the authors don’t all agree with one another, for the most part they do agree on two things. One is that there were at least some issues with the initial report from the Reproducibility Project, which made a big splash in the media at the time. The second is that we'd probably still like to see higher rates of reproducibility.

Read 12 remaining paragraphs | Comments