Dani Rodrik has some interesting thoughts on mainstream neoclassical economics (I only read his first 2 answers). I very strongly agree with two of those thoughts
"I love an old quote from Carlos Diaz-Alejandro who once said something along the lines of “by now any graduate student can come up with any policy conclusion he desires by building appropriate assumptions into his model.” And that was some thirty years ago! We have plenty more models that generate unorthodox conclusions now."
I heard a similar remark from Robert Barro. If Barro and Diaz-Alejandro agree on something, it is fairly likely that they have a point.
"A peculiar deformation of mainstream economics is the tendency to pooh-pooh the real-world relevance of all the theoretical reasons market fail and government intervention is desirable."
I think it is very common for economists to present the implications of decades old models as the truth. This is incorrect for two reasons. First a priori theory can't tell us about the universe in which we happen to find ourselves -- it can only generate hyptheses to be tested empirically. Second it is now known that the interesting implications of old models were implied by assumptions made for tractability (note the absense of any qualifications). Economics as often presented to non economists is a set of conclusions not implied by contemporary economic theory (which theory is taken seriously by many economists despite a notably lack of emprical success -- but that's another issue).
I think I disagree with other things which Rodrik said
1) Economists such as George Akerlof, Paul Krugman, and Joe Stiglitz are some of the names that come to mind who exemplify this tradition. Each of them has questioned conventional wisdom, but from within rather than from outside.
Akerlof and Krugman when ? Each won the Nobel prize for innovative work (innovative enough to challege conventional wisdom) done withing neoclassical economics. Each has moved on. Akerlof listens to sociologists, Krugman dismisses the dominant contemporary academic approach to macroeconomics. One can admire what they did then and what they do now, but they aren't still within the same school of thought that they were within.
One reaction I get when I say this is the following: “how can economics be useful if you have a model for every possible outcome?” [skip] The best we can do is to understand the structure of behaviour in each one of these cases, and then have an empirical method that helps us apply the right model to the particular context we are interested in.
Here Rodrik hasn't demonstrated (or even really argued) that there is any benefit from using models at all. deleting everything up to "have an empirical method that helps us [in] the particular context we are interested in" ?
What plausibility, appeal, record of success or any imaginable good feature of an argument could Rodrik's argument have and the modified argument for atheoretical empiricism lack ? Pulling a magic "empirical method" out of your hat *after* writing own models with every possible implication is no easier than cutting out the middlemodels and just using the magical empirical method to begin with.
Note I am claiming that Rodrik has to much faith in empirics unquided by theory. The problem is that if thery is consistent with "every possible outcome" then theory can not give us a specification, identifying restrictions, a prior or well anything that combined with a data set yields something more than that data set.
Of course the "empirical method" isn't all powerful. Rodrik does not admit the possibility that it might tell us to throw all our models away and start fresh.
Here I think that Rodrik has fallen for the typical economists equivocation. The core assumptions of neoclassical economics A have no implications (agreed) and B are useful to empirical research. A implies that B is false. What happens (always I think) is that when the validity of economic theory is questioned it is defined so that it can't be false. Reassured that it isn't false, economists try to apply it. To do that it must be re-defined so that it can be false.
I think Rodrik just hasn't understood the question. There is a vast difference between a set of theories such that any possible "policy conclusion" is implied by one and a set of theories such that "every possible outcome"
is implied by one. An "outcome" can be "random policy is optimal and the data collected from the dawn of time until now are big bang, ...., US 2009q1 GDP = X, the average price of tea in China on may 3rd 1893 was Y, ..., . Any possible outcome" means that we don't have any idea what will happen. It also means that in a trillion years, knowing so much more forgetting none of it and processing all of it optimally, if we try to predict what will happen next, we will have no idea what will happen from then on.
I think that Dani Rodrik might benefit from reading "Fact Fiction and Forecast" by Nelson Goodman. I'm pretty sure it's for sale at the COOP.