Read on Twitter
Good thread from @paulnovosad on data and code requirements for econ publications. These fairly new requirements are still a work in progress and certainly can be improved, but I am more optimistic than Paul: I do think they generate clear benefits. https://twitter.com/paulnovosad/status/1354803866710142979
Empirical papers can be wrong for lots of reasons. In my experience, requiring computational reproducibility is a good way to reduce common errors that would not have been detected from reading the manuscript or appendices.
The mere act of requiring computational reproducibility will cause many researchers to discover mistakes that they wouldn't have discovered otherwise. It also increases the public's faith in our research.
Yes this requires more work, but that's the cost of making sure research is right. A challenge is that these types of skills are not taught in most grad programs. Hopefully that will change. In the meantime, for @stata users, here is a guide that may help: https://julianreif.com/software/
For R users, I recommend following @grant_mcdermott, @paulgp and others on Twitter who frequently provide very helpful workflow advice.
