This repository contains data from a census of papers published in the American Economic Review in 2021 that use a bootstrap, as described in Andrews and Shapiro (2025b).
For each paper, the directory output/derived includes two CSV files (and their corresponding log files with summary statistics):
NAME_Estimates.csvcontains the authors' original point estimates, bootstrap standard errors and (where reported) confidence intervals for selected objects of interestNAME_Replicates.csvcontains the bootstrap replicates for each object of interest
datastore/raw: original information (bootstrap replicates, replication archives, and correspondence) for each paper in censussource/raw/: metadata on papers and their objects of interestderived/: code to clean or extract replicates for each paper
output/: estimates and replicates for each paper's objects of interest
The structure and requirements of the repository follow JMSLab/Template. If you wish to run the code in the repository, please see the documentation there.
- See
source/raw/papers.csvfor details on the original source for each set of replicates.- For those replicates obtained directly from a paper's author(s), the code in
/source/derived/performs minimal cleaning and validation. - For those replicates obtained from a paper's replication archive, the code in
/source/derived/adapts the original code to extract the replicates. The original replication archives can be obtained by following the DOI link insource/raw/papers.csv.
- For those replicates obtained directly from a paper's author(s), the code in
If using the census in published work, please cite the following:
- Andrews, Isaiah, and Jesse M. Shapiro. 2025a. "BootstrapCensus". Code and data repository at https://github.com/JMSLab/BootstrapCensus.
- Andrews, Isaiah, and Jesse M. Shapiro. 2025b. “Communicating Scientific Uncertainty via Approximate Posteriors.” NBER Working Paper Series No. 32038. https://doi.org/10.3386/w32038.
We thank our dedicated research assistants for contributions to this project.