IPA’s weekly links

Guest post by Jeff Mosenkis of Innovations for Poverty Action.

  • With the debate over deworming in danger of overtaking actual worms in terms of lost productivity, a reminder that much arguing about what analysis methods were chosen can be solved with a pre-analysis plan. The Journal of Economic Perspectives has two helpful articles:
    • Olken’s Promises and Perils of Pre-analysis Plans goes over a checklist of advantages, such as allowing for choosing unconventional tests without accusations of cherry picking, and tips for compromises, like making some of the initial data cleaning choices “blinded” (without separating treatment/control).
    • Coffman and Niederle argue Pre-analysis Plans Have Limited Upside, Especially Where Replications Are Feasible, and the problem is really that there’s little incentive for academic researchers to replicate.
    • One thought: A stats prof of mine who was a former physicist in his 80’s (at least), used to say back when stats were done by hand or after waiting days for a turn at the university’s basement-sized mainframe computer, people chose their stats tests and planned their analysis far in advance and much more carefully. When there was a “cost” to each analysis, the process was slower & more deliberative – essentially right in between “pre-analysis plan before study starts” and “test as you go.”

And Nigerian lawyer and satirist Elnathan John offers the Gospel According to Aid.

Comments are closed.