Chris Dillow has an interesting post on mechanisms and why economists cannot make good predictions. But first of all, he uses “mechanism” in a different way than I like to see it used. In my mind, a mechanism is as described in the Mechanisms page of this web site. Dillow follows Elster in using a more generic sense of the word: for him, a mechanism is a step-A-leads-to-step-B-to-step-C… story.
Economists have many mechanisms, in Elster’s sense, that may be behind what we see happening. However, the typical stories that these “mechanisms” embody are only good for giving some insight on what happened after it has. There are always too many dormant “mechanisms” waiting to be activated, and we cannot predict what triggers their activation, so we cannot make predictions safely.
A commenter identified as “Mr Art” says, in reference to Dillow’s post, that if a theory offers explanations but not predictions, it is not science. This is a tired criticism, hardly worth addressing, but I will try, briefly. A theory starts with some assumptions and then generates conclusions. You can test, in principle, whether these conclusions occurred or not while the assumptions held. This allows falsification of the theory, in principle. But who says that these conclusions have to do with the future, thereby making them predictions in the sense that Wall Street wants to have predictions?
I am reminded, for the Nth time, of the comment by, I believe, Kenneth Boulding (please correct me if I misremember) that economists predict because they are asked to, not because they can.
H/T to Mark Thoma for bringing me to the Dillow post.