
A recent Twitter conversation I had with other social scientists made me reflect upon the state of my own discipline when it comes to standards of evidence. That is, about when we can say that we are fairly certain about a specific conclusion drawn from analyzing data.
The conversation began when Raul Pacheco-Vega said “I love experimental methods, but being obsessed with it is unhealthy.”
I responded that what is unhealthy is the attitude according to which “the only good research questions are the ones that can be randomized.” That is, the mode of thinking – dominant among certain empirical economists – according to which if a research question does not lend itself to randomization, it is not worth one’s time to try answering that research question.