I’ve seen a lot of defenses of economics as a science in the past year. The best of the bunch is by Robert Shiller, one of the winners of this year’s “Nobel Prize” in Economics.

This all takes place in the context of recent developments like the controversy over the Reinhart-Rogoff paper, “Growth in the Time of Debt”, that had been used to argue for fiscal austerity, or the work that shows that free trade harms labor and advantages capital (contrary to conventional economic wisdom).

Shiller’s position is that economics has some science as well as a lot of non-science, which he hopes will eventually be crowded out. I think that’s true of a lot of fields, including my own field of computer science. Economics has the additional handicap of politics, that is, policy makers want the veneer of “economics” to validate their policy choices, regardless of what economic science says. There’s a market for anti-science economics.

What I would like to see is for economists to put science directly in policy. Policy should be used to conduct economic science.

For example, in the US tech industry, immigration policy is controversial. Some people say that we don’t have enough native talent and we need more immigrants, and others say that the purpose of the H-1B program is just to get cheap labor. A scientific immigration policy should try to answer these questions. It would be constructive to modify the current H-1B program to do a randomized trial: after awarding visas by the usual process, convert half of them to green cards at random, and study the results. Do H-1B holders stay with companies longer than green card holders? Who gets paid more, over time? Is there a company-wide or industry-wide effect?

I’m sure that a real economist could design a better experiment. If economics as a whole wants to be seen as a science, at a minimum they should have testable hypotheses, and they should actively pursue tests via policy.