I think researchers should share their data and code, and that includes me. This page lists what my co-authors and I have posted publicly, paper by paper, and says what checking (if any) has been done on it.
I’m also starting to go back over my own published work, to see whether the numbers hold up when the original code is rerun. I’m using AI for this: Claude, running in Claude Code. It’s much faster than doing it by hand, and it means I can check old papers I’d otherwise never get back to. But it isn’t independent. I direct it, it works from my own files, and its output needs checking too. Where AI did the checking, I say so.
If you find a problem with any of this, please tell me.
Reinstein and Snyder (2005): expert reviews and movie demand
“The Influence of Expert Reviews on Consumer Demand for Experience Goods: A Case Study of Movie Critics” (with Chris Snyder). Journal of Industrial Economics 53(1): 27–51, 2005.
- Shared: the original data, our original Stata do-files and logs, a Python port of the code, and its outputs, on GitHub. The results are written up on a replication page, which takes Hypothes.is comments.
- Checked: yes, by AI (November 2026). Tables IV–VI, the main results, reproduce exactly. Tables I–III match apart from small print slips. Table VII matches our own Stata log, but the one estimate we printed as significant at 10% (two thumbs up reducing competing movies' revenue) has p = 0.137 in that log. Its significance also depends on an arbitrary sort order in our code: 8 of the 80 possible orderings give p < 0.10. So that star isn’t supported.
- Who checked it: Claude (Anthropic’s Claude Opus 5.5, running in Claude Code), guided by me. This is not an independent replication: I’m a co-author, I chose what to check, and it uses our own data and code. The numbers come from running the posted code; no person has re-derived them by hand.
Kellner, Reinstein and Riener (2019): “give if you win”
“Ex-ante Commitments to ‘Give if you Win’ Exceed Donations After a Win” (with Christian Kellner and Gerhard Riener). Journal of Public Economics 169: 109–127, 2019.
- Shared: the preregistration and pre-analysis plans for the later experiments, on the AEA RCT Registry (2017). The data and code aren’t public yet.
- Checked: no replication or re-check yet.
Bergh and Reinstein (2021): victim images and charity evaluations
“Empathic and Numerate Giving: The Joint Effects of Victim Images and Charity Evaluations” (with Robin Bergh). Social Psychological and Personality Science 12(3): 407–416, 2021.
- Shared: a GitHub repository with the R and SPSS analysis code, the Qualtrics survey files, the preregistration of Study 6, and the data for Study 6 (an online omnibus survey). Data from Studies 1–5 couldn’t be shared because of ethics-board (IRB) constraints. Two related 2019 preregistrations from this project are on OSF (April, May).
- Checked: no replication or re-check yet.
Reinstein and Riener (2012): desert and tangibility in giving
“Decomposing Desert and Tangibility Effects in a Charitable Giving Experiment” (with Gerhard Riener). Experimental Economics 15(1): 229–240, 2012.
- Shared: the analysis dataset, deposited by Gerhard Riener on GRO.data (Göttingen’s Dataverse; 2022, CC BY-NC-SA 4.0). No code there yet. The lab instructions are in the journal’s supplementary material.
- Checked: no replication or re-check yet.
Wild animal welfare attitudes scale (2025)
“Measuring Wild Animal Welfare Attitudes” (Sleegers, Moss, McAuliffe, Reinstein and Waldhorn). Psychology of Human-Animal Intergroup Relations, 2025. A Rethink Priorities project led by Willem Sleegers; I’m one of five co-authors.
- Shared: survey materials, data and R scripts for the pilot and all four studies, on OSF. Studies 1–4 were preregistered there.
- Checked: no replication or re-check that I know of.
EA Market Testing (2021–2024)
Trials of fundraising and messaging approaches with effective-giving organizations, run by a team I directed.
- Shared: the analysis code and the reports built from it. The trial data themselves are kept private and shared only with the partners' permission.
- Checked: no replication or re-check yet.
LLM research evaluation (work in progress)
“Just Ask the Model: One-Shot LLM Research Evaluation and Structured Expert Review” (with Valentin Klotzbücher, Lorenzo Pacchiardi and Tianmai Michael Zhang), 2026. An intermediate draft; much of the code, analysis and text was generated by AI agents under our direction.
- Shared: the full project, with data, prompts, model outputs, scripts and a data-provenance note, in Valentin Klotzbücher’s GitHub repository.
- Checked: no independent check yet.
Checking other people’s work
I’ve also helped reproduce other people’s papers, as part of Institute for Replication (I4R) events.
- Colantone et al. (2024, APSR), “The Political Consequences of Green Policies”: checked by a five-person team (people, not AI) at the 2024 Brown replication games. Our comment, the team’s code and our notebook are public.
- The many-team study (PNAS, 2026) comparing human-only, AI-assisted and AI-led reproduction teams, where I’m one of many co-authors. I4R’s reproducibility package and OSF project have the data, code and pre-analysis plan.
Preregistrations
My OSF profile has around fifteen public preregistrations from 2018 to 2024, mostly charitable-giving experiments. These include the field experiment on “impact per dollar” information in a large charity email campaign (2018), the charity-substitution experiment, the collective-action experiments with David Hugh-Jones and Hannes Titeca, and the studies on cost-effectiveness ratings in social-media donation ads. They are plans, not data, but they let you compare what we said we’d do with what we did. No one has done that comparison systematically yet.