42. Böhm-Bawerk Lecture

Anna Dreber Almenberg is a professor of economics at the Stockholm School of Economics, and was previously an affiliated professor with the University of Innsbruck. Her research is mainly in meta science, exploring topics such as the replicability and generalizability of scientific results and the potential use of prediction and decision markets for replications. Anna also does work in behavioral and experimental economics on topics such as hormones and decision making and economic preferences more broadly. Anna is an editor at the Journal of Political Economy. She is also a Wallenberg Scholar, a member of the Royal Swedish Academy of Sciences and a member of the Royal Swedish Academy of Engineering Sciences.


Abstract

(Predicting) replications and generalizability

Which results replicate or generalize? In this talk, Anna Dreber Almenberg will discuss several large replication projects in mainly psychology and economics, where her coauthors (many from the University of Innsbruck) and she have redone experiments published in high impact journals with new and larger samples to see whether the main result replicates. She will also cover their studies on the forecasting of scientific results, and discuss their recent work on the possibility of using decision markets to select which studies to replicate. In the prediction markets they have set up, researchers are given monetary endowments to bet on whether results picked for replication will replicate with the definition of successful replication being a statistically significant effect in the same direction as the original study. In decision markets, they use market prices to decide which studies to replicate. She will also briefly discuss work on analytical heterogeneity and design heterogeneity. These are projects where researchers are asked to test the same hypotheses on the same experimental or non-experimental data, or where researchers are asked to design experiments to test the same hypothesis. The results suggest that a large share of results do not replicate and that this is to some extent predictable. The results also suggest that there is substantial analytical and design heterogeneity, limiting the generalizability of individual results. She will also discuss potential ways forward for tests of new hypotheses.