In a talk at the World Bank that I gave last week, I repeated a riff that I’ve used before. Suppose your internist told you:
The x-ray shows a mass that is probably cancer, but we don’t have any good randomized clinical trials showing that your surgeon’s recommendation, operating to remove it, actually causes the remission that tends to follow. However, we do have an extremely clever clinical trial showing conclusively that Botox will make you look younger. So my recommendation is that you wait for some better studies before doing anything about the tumor but that I give you some Botox injections.”
If it were me, I’d get a new internist.
To be sure, researchers would always prefer data from randomized treatments if they were available instantly at zero cost. Unfortunately, randomization is not free. It is available at low or moderate cost for some treatments and at a prohibitively high cost for other potentially important treatments. Our goal should be to recommend treatments and policies that maximize the expected return, not to make the safest possible treatment and policy recommendations.
I prefaced my talk by saying that I’m following a strategy of “show, don’t tell.” I work on high expected-return policies that can be implemented, with no concern about whether I will be able to publish the results from this work in the standard economics journals. (This updated post on urbanization and growth, which includes some of the data that I presented at in my talk at the bank, is an example of work that motivates one such policy, Urban Expansion.)
I have the good fortune of knowing that I can be a successful academic even if the journals will not publish results from the work I do. I realize that many other economists do not have this freedom. I understand that they have to respond to the incentives they face, and that the publication process biases their work in the direction of policies that are more like Botox than surgery.
But we can all work to change the existing equilibrium. It is good that economists pay careful attention to identification and causality. This inclination will be even more important as new sources of “big” non-experimental data become available. But it is not the only good thing we can do. We have to weigh the tradeoffs we face between getting precise answers about such policies as setting up women’s self-help groups (the example that Lant Pritchett uses as his illustration of what I am calling Botox for economic development) versus such other policies as facilitating urbanization or migration that offer returns that are uncertain but have an expected value that is larger by many orders of magnitude.
If economists can’t understand the tradeoff between risk and expected return, who can?