I have been waiting to get various discussions in before responding, but I think most comments are in so I will respond. 

Again: A warning--this discussion is a response to comments on an earlier post of mine, so if you haven't been following that discussion, you might not want to bother reading this. 

To those who are part of this discussion, I suspect we have worn the patience of many on the list who are not so involved in this, so I suggest that if we continue this discussion, we continue it off line.

Dave



Roger asks me whether two statements that "concern how the world is, yet they have obvious implications for policy are art or science. I reprint one in the interest of space.

The government has decided that it wishes to reduce inflation by a
specific amount, and that it wants to achieve this through policies
that do not interfere with certain other objectives. The economist
writes a report in which he or she shows that adopting a certain list
of policies will achieve that objective. Does it make any difference
whether the list of "other objectives" is comprehensive and clearly
specified or vague and even non-existent?

This is definitely art in my view. It is the engineering branch of art. No one says that art cannot, and should not be objective.  Why is this question not in science?  Because the issues are too vague--definitions of inflation involve many complications, and the policies that he or she will discuss will have implications that go way beyond economic science.  Ditto for his other example. 

So in answer to his question:

Is the difference that if the economist judges policy according to the
government's stated objectives, the economist is engaging in science;
but if the economist judges policy according to objectives that are
not explicitly stated, the economist is combining scientific knowledge
with judgments about the criteria that should be used to evaluate
policy, and is therefore engaging in art?

My answer is no--that is not the defining aspect. The defining aspect is whether the questions are formulated in a clear enough manner so that one can provide an answer based on a tested economic theory that does not raise other issues that would need to be answered if one is to act on the advice.  In my view most of what economists do is not science. It is a type of engineering/art. 

Don't get me wrong. The definition of science is not the point. The appropriate methodology is the point. The distinction doesn't matter except in terms of the methodology used. As J.N. Keynes said, the methodology for the policy art of economics is much broader than the methodology for the science of economics. The reason I think the distinction is worth raising is that to use the methodology of science to answer a question in the art of economics means that one is limiting one's consideration of issues and missing many connections, and is not doing a good job of policy advice. To do good policy advice you have to take in all relevant dimensions of the problem, or at least be aware of them and point them out. Most of our economic models don't do that, and thus leave policy issues unaddressed.  

This policy art can be highly technical, and or it can be looser--the nature of the art depends on the nature of the question being asked. So I am arguing for using science in economics in a highly restrictive way--to refer to pure science. I recognize that this is only a definition--we could distinguish realistic science and pure science as Pigou did, and say that what I am calling engineering/art is actually realistic science. That would be fine with me as long as (1) one did not claim that policy recommendations that follow from it had the backing of pure economic science, but were based on art, and (2) one allowed broad methodological discussions of the policy, and did not try to shut up critics with the argument that they are not doing science. 

Now let me turn to John's comments.  John writes the following:

 As I understand David's position (correct me if I am wrong), the economic scientist is merely looking at the naked, empirical facts, and let the chips fall where they may. And he can certainly have no purpose in looking at these facts, nor any ideological lens which might blur his vision.

I have no idea of what a naked, empirical, fact is (it sounds of bit lewd), but I fully agree that how one looks at something influences what one sees.  My argument is simply that trying to see the facts, and understand the lenses through which facts are seen is what a scientists should be doing. That means recognizing that if you have a policy view, that policy view is likely to influence your interpretation of the facts. Thus, in my proposed scientific methodological approach, you (1) should lean over backwards to let others know that you have this policy view  that might influence your interpretation so they can judge your statements accordingly, and  (2) should carefully watch yourself to see that you aren't pushing a particular view. That's why pluralism and really caring about what others who disagree with you believe are so important--they help keep you from dressing the facts in  a particular way. 
 In my testimony to congress on the DSGE Model (http://science.house.gov/sites/republicans.science.house.gov/files/documents/hearings/072010_Colander.pdf)  I developed this distinction and these arguments further and justified them more,  so I refer anyone interested to that.  

As for John's discussion of definitions of economics--he writes: 

 ' "Economics is the study of the allocation of scarce resources among alternate uses." Let me suggest that this "definition" is total nonsense.'

I'm not sure what total nonsense in a definition is, but I prefer descriptive definitions, not prescriptive definitions of economics. For me, economics is what economists do, and that changes over time.  We can debate what part of what they do is science and what part art, and discuss appropriate methodology for each, but the definition one uses should capture what they do, or at least see themselves as doing. I would point out that when Robbins put that definition that John doesn't like out there, it was as a definition of economic science, not of economics, and at the time studying that issue was what cutting edge economic scientists were doing. They have since moved on and the definition of economic science should move on as well. 

Finally turning to Steve. Steve writes that economics is a policy science, and that there is little interest in what I am calling economic science. I tend to agree with him--there would be less interest in the science of economics if economic scientists were honest about what the science of economics has to say about policy.  But there would still be interest, and I for one would support some really bright people thinking very abstractly and creatively about how economies function for no other purpose than to understand how the damn things work. And who knows, ultimately these economic scientists may well have a larger effect on policy than do more practical policy economists. But the goal of their work is simply to understand. 

What most   economists do is not science as described above; it relates to policy. My view is that that work should not pretend to have the imprimatur of science--and should recognize that the appropriate methodology for it does not come from science methodology, but is more related to art and engineering methodology. It is more of an all things go methodology. In my view training for that policy branch of economics needs to be much broader, more historically focused and less focused on economic science and more focused on economic institutions, history, and history of ideas than is most current training. Future policy economists need to be aware what economic scientists are doing, but they don't have to be trained to do it. That's been my complaint about graduate economic education over the last 25 years.