I've read this article, which is quite famous, but I didn't know it. And it's never too late to read good stuff.
What I found in the speech of Richard Feynman just expresses, much better than I could, some of the doubts sometimes I have on the way economics is going on. Are we sometimes close to Cargo Cult?
Cargo Cult Science (R. Feynman, 1974)
Economics isn't easy, altough some of the main concepts are intuitive.
If you have a model, you can test it but you can never be sure of your results, because of the data you had, because of the empical methodology, because maybe you've been lucky etc.
The fact is simply that we can not make proper experiments, so often people try to figure out something by looking at stylezed facts and using logical reasoning and math to derive consequences. Then you may want to know if all the reasoning drives to conclusions which are consisent with the data.
And many other problems arise at this point.
First, "consistent" doesn't mean that our reasoning is correct, it just means that it might be. As it might be that we got the right results from the wrong causes (which may be wrong no matter how appealing they are). A theoretically perfect model then should be the only possible one driving some results, so that if results are indeed verified then we are sure the model is right. But it is never like that in reality.
Second, consistent to what data? That's a major problem. We can not have the data we want, we have the data we have. How we do collect data is affected by the way we think about economics, which might be wrong. Then within a model we might have a certain variable, but we only have proxies in practice, which it is not the same thing.
Third, how do we test a model? We have stastistics and econometric techniques. Often we need to reason in probabilistic terms and we need a way to test causality and deal with reverse causality issues. Then the same model can be tested in many ways, which is usaually done; and getting the same results with different methodologies is usally seen as a sign of rubustness. The fact is that we need this kind of robustness because we can never be really sure of what we get.
Suppose I see that GDP and Education (misured by some proxy!) correlate. Our understanding is that education increases productivity, which means higher GDP, and that probably higher Income also means more investment in Education. Is that all? No, we can think of many other different ways the two variables interact each other. How do we know which causation is the right or the most important one? We don't even know if we thought about all the possible relationships and we don't have a proper way to test which is the prime cause of everything.
Now suppose we decide to build new schools because we interpret the correlation as a causal link. GDP might then grow, but it might not. If it works people say it's because we correctly interpreted the data, if it doesn't they can say it's because the school quality matters too and similar stuff, so that no one can actually say that the model was wrong: it might have been incomplete, but hey it's a model, it was the practical aspect which was mistaken...
How similar is that to Cargo Cult? You follow your witch doctor and do strange stuff, if by chance it works, the doctor was right, if it doesn't then it's you who did something wrong...
Again, suppose I see Money (what is money?) and Interest Rate (which one?). It is intuitive to think of the interest rate as the opportunity cost of money, so it is the price of money and we easily think of a downward sloping money demand. Intuitively it is good and it kind of works in practice. But do we really know that interest rate is actually the price of money, that people behave in this terms? Maybe there are more complex relationships that make everthing consistent to what we think. So then in practice we can effectively think in terms of our model, but actually wrongly.
Isn't that thinking that the thunder causes the lightining? It works, you can effectively foresee a lightining by a thunder, but you are thinking in the wrong way.
You might say, that's a model, of course it makes things simple, it is designed for that. Right, but it is designed to help our understanding, not just to reproduce reality. And if the model makes me think in the wrong way, it is useless no matter how close to reality its results are.
___________
Note: I know people have thought about this stuff much deeper than me and that many techniques exist to try to deal with those issues. And I should know so much more about that... I just used the Feynman article as a good excuse to think a bit again about that and ask myself and you:
where would you put Economics between hard science and cargo cult science?
