Consider the covering law model, also known as the deductive-nomological view of scientific methodology. In this view, scientists formulate universal laws, in some sense. In an application of a scientific law, the hypotheses or antecedents are asserted to be true. That is, the statement of scientific law is conjoined with initial conditions. One then checks that the consequent holds. If observation is inconsistent with the consequent and one is sure that the initial conditions are true, the law is refuted.
Milton Friedman advocates instrumentalism, in which the assumptions of a scientific theory are false. (Actually, his famous essay, "The Methodology of Positive Economics", is so incoherent, Friedman can be interpreted as advocating almost any methodology you care to name. But let's stick with a widely argued view.) In Friedman's view the antecedents are always false in a significant theory:
"Truly important and significant hypotheses will be found to have 'assumptions' that are wildly inaccurate descriptive representations of reality, and, in general, the more significant the theory, the more unrealistic the assumptions" -- Milton FriedmanThus, if one holds that economic theories state covering laws and that economists are and should be instrumentalists, economic theories cannot be refuted by observation. The logical implications of false antecedents need not be true.
Can economics be a science if it is practiced in keeping with Friedman's strictures?
Posts
4 comments:
Steve Keen, in Debunking Economics has a whole chapter debunking Friedman's ideas on this matter. And it is on-line as a sample chapter:
There is madness in their methodEssential reading!
I've always wondered what would happen is a creationist used Friedman's criteria when discussing evolution and biology. After all, you cannot get a more "wildly inaccurate descriptive representations of reality" as god for a starting assumption "and, in general, the more significant the theory, the more unrealistic the assumptions."While in biology the people putting forward such arguments are laughed at, in economics they are given the so-called Nobel prize...
Iain
An Anarchist FAQ
Although I did not link to that chapter, I mentioned Keen on Friedman some time ago. C. B. Macpherson had a nice take-down of Friedman too.
When constructing mathematical model in any field there is a need to make these assumptions to simplify the model so it can be worked. This is in all scientific fields. Even in physics. It is why you make the assumption that there is no air resistance when modeling on objects velocity. Which is a crazy and unrealistic assumption. The whole point is in any mathematical model when simplifying them is to assume away variables that do not effect the underlying outcomes of the model. To make a completely realistic economic model would require millions and maybe billions of variables. Also when making these models you want to keep everything the same so that you can isolate the effect on the model of certain variables. If you did not make the crazy assumption that the rest of the economy stays the same then it would be almost impossible to tell what is actually causing the change in the model. These models themselves are not made to give a prediction of the current economy. They are used to understand the underlying forces in the economy. For predictive models you need to look at the statistical model that are used in econometrics. A lot of these theoretical models once formulated tend to go through stages where economist add variables such as bounded rational actors and asymmetric information to make them more realistic to see if that changes the way the model acts. If it does it is kept in the model if not than the model goes back to more simplified form. Milton Friedman has said some things that i do not agree with and he does not represent the mainstream thinking except for his consumption function and the quantity theory of money this time I think what he said was completely reasonable and the unrealistic assumptions are not only present in economic math models but in every math models in any field of science even hard sciences.
The points Chris Kevlin made are all standard defences of Friedman's notions and in fact, one can find them in a form or another in Friedman's paper.
On the surface, one may find some of them compelling. The devil, however, lies in the details.
Let's give a clear example, using the first few sentences in Chris' post:
"(1) When constructing mathematical model in any field there is a need to make these assumptions to simplify the model so it can be worked.
(2) This is in all scientific fields. Even in physics.
(3) It is why you make the assumption that there is no air resistance when modeling on objects velocity. Which is a crazy and unrealistic assumption."
Truly, a model is a simplified representation of reality. That's the definition of a model. So one must conclude that (1) is a reasonable proposition.
Now, let's move on to the two remaining sentences:
Simplified mathematical models (I am aware of the tautology here, but bear with me, I just want to emphasize that models are simplified representations of reality) have worked pretty well in much of physics, too. Newtonian mechanics, for instance.
The same strategy hasn't worked that well with other natural sciences. Meteorology and ecology come to mind.
So, why have simplified mathematical models worked so well in Newtonian mechanics, while they are much less effective in other fields?
I venture the answer that the movement of bodies studied in Newtonian mechanics is an intrinsically simpler problem. And without denying their genius, Galileo and Newton just happened to stumble upon it: they chose the right problem and the right variables. On this they had a head start, too: since at least Aristotle the problem had been studied and the set of relevant variables had been largely identified.
So, in reality, the assumption of no air resistance is not so crazy, for a wide range of values. But even in physics, assumptions need to be made with utmost care: air resistance is indeed very relevant, for instance, when considering a feather falling in an atmosphere.
By contrast, in economics the number of variables is much larger, there is no universally accepted and reliable guide as to which variables are not relevant and under which circumstances they are not relevant.
The feather example leads to a second strategy that physicists have been able to employ to great effect: experimentation and measurement. The discrepancy between the predicted speed of a feather calculated under the assumption of no-air resistance versus its actual speed can and has been precisely measured.
No exact comparison can be made in economics. And, on top, the effort to measure economic phenomena (intrinsically more difficult, as experimentation is largely ruled out) has been given low priority, since the times of Adam Smith. It's important to remember that it was Malthus who started the serious consideration of statistics.
But what is perhaps the worst limitation, as it is self-imposed: the use of not directly observable variables like utility.
So, one must conclude that sentences (2) and (3) are highly inaccurate. The comparison physics-economics is misleading and breaks down from here on.
In practice, the use of Friedman's "methodology" (and I don't mean to be controversial or disrespectful here) has induced many economists to fall into the fallacy of affirming the consequent: Unemployment is rising? Oh well, people really want to take their time off.
What Chris said about adding or excluding variables according to the de facto performance of a model (not treated here up to now) would be frowned upon by Popper as anti-scientific and revealing that the models themselves cannot be falsified. If the model really sucks, I can always add another variable (and women's skirt length, day of the week and number of sunspots have been suggested as statistically significant variables, at one moment or another, in different contexts).
Post a Comment