Wednesday, September 19, 2012

Something I think critics of mainstream economists need to be more careful about

"Optimization" is a misleading thing to talk about, because common language suggests the output of an optimization process is "optimal". Talking about "optimization" has the tendancy to make people think that mainstream economics has a Pollyannaish view of the economy (this is particularly common among critics on the left).

What is critical to remember is that "optimization" refers to the decision making process, not a characteristic of the outcome. We'll opt for the good over the bad and the better over the good. That's what is implied by "optimization", not that the outcome is always wonderful.

It's also critical to remember what is being optimized. An objective is always what is being optimized.

And to make matters more confusing, that objective is subjective, as are all the other arguments of the optimization problem (including the constraints).

If you think people have imperfect subjective views of the world (but not so erratic that you can't get them to be behave mathematically), that they choose good over bad, and that the outcome of this process can definitely be "wrong" relative to some standard outside of the individual's subjective objective function, then you shouldn't get all that worked up about optimization.

Even heuristics jive with everything in that previous paragraph if we remember that the whole reason we use heuristics is that it is better than not using them (from a cognitive resources perspective).


  1. There is a difference between optimizing and satisficing.

    1. Is there really so much of a difference?

      Optimization doesn't lead to the optimal, it can't be expected to. In some cases it's assumed that an optimal point is reached, optimal according to the preferences of some set of individuals. Satisficing was an idea produced to deal with this issue, as I understand it it doesn't deal with it in a very different way to what Daniel writes here. The argument was put forward that assuming things are in "optimal" states at any given time is ridiculous, which sometimes it is. But, often that's not really the argument that's been made, the underlying point-of-view replies on optimization not on achievement of the optimal. It's maybe useful to introduce the term "satisficing" but really it's an old idea.


All anonymous comments will be deleted. Consistent pseudonyms are fine.