I have previously discussed Milton Friedman’s infamous 1953 essay, ‘The Methodology of Positive Economics.’ The basic argument of Friedman’s essay is the unrealism of a theory’s assumptions should not matter; what matters are the predictions made by the theory. A truly realistic economic theory would have to incorporate so many aspects of humanity that it would be impractical or computationally impossible to do so. Hence, we must make simplifications, and cross check the models against the evidence to see if we are close enough to the truth. The internal details of the models, as long as they are consistent, are of little importance.
The essay, or some variant of it, is a fallback for economists when questioned about the assumptions of their models. Even though most economists would not endorse a strong interpretation of Friedman’s essay, I often come across the defence ‘it’s just an abstraction, all models are wrong’ if I question, say, perfect competition, utility, or equilibrium. I summarise the arguments against Friedman’s position below.
The first problem with Friedman’s stance is that it requires a rigorous, empirically driven methodology that is willing to abandon theories as soon as they are shown to be inaccurate enough. Is this really possible in economics? I recall that, during an engineering class, my lecturer introduced us to the ‘perfect gas.’ He said it was unrealistic but showed us that it gave results accurate to 3 or 4 decimal places. Is anyone aware of econometrics papers which offer this degree of certainty and accuracy? In my opinion, the fundamental lack of accuracy inherent in social science shows that economists should be more concerned about what is actually going on inside their theories, since they are less liable to spot mistakes through pure prediction. Even if we are willing to tolerate a higher margin of error in economics, results are always contested and you can find papers claiming each issue either way.
The second problem with a ‘pure prediction’ approach to modelling is that, at any time, different theories or systems might exhibit the same behaviour, despite different underlying mechanics. That is: two different models might make the same predictions, and Friedman’s methodology has no way of dealing with this.
There are two obvious examples of this in economics. The first is the DSGE models used by central banks and economists during the ‘Great Moderation,’ which predicted the stable behaviour exhibited by the economy. However, Steve Keen’s Minsky Model also exhibits relative stability for a period, before being followed by a crash. Before the crash took place, there would have been no way of knowing which model was correct, except by looking at internal mechanics.
Another example is the Efficient Market Hypothesis. This predicts that it is hard to ‘beat the market’ – a prediction that, due to its obvious truth, partially explains the theory’s staying power. However, other theories also predict that the market will be hard to beat, either for different reasons or a combination of reasons, including some similar to those in the EMH. Again, we must do something that is anathema to Friedman: look at what is going on under the bonnet to understand which theory is correct.
The third problem is the one I initially honed in on: the vagueness of Friedman’s definition of ‘assumptions,’ and how this compares to those used in science. This found its best elucidation with the philosopher Alan Musgrave. Musgrave argued that assumptions have clear-if unspoken-definitions within science. There are negligibility assumptions, which eliminate a known variable(s) (a closed economy is a good example, because it eliminates imports/exports and capital flows). There are domain assumptions, for which the theory is only true as long as the assumption holds (oligopoly theory is only true for oligopolies).
There are then heuristic assumptions, which can be something of a ‘fudge;’ a counterfactual model of the system (firms equating MC to MR is a good example of this). However, these are often used for pedagogical purposes and dropped before too long. Insofar as they remain, they require rigorous empirical testing, which I have not seen for the MC=MR explanation of firms. Furthermore, heuristic assumptions are only used if internal mechanics cannot be identified or modeled. In the case of firms, we do know how most firms price, and it is easy to model.
The fourth problem is related to above: Friedman is misunderstanding the purpose of science. The task of science is not merely to create a ‘black box’ that gives rise to a set of predictions, but to explain phenomena: how they arise; what role each component of a system fills; how these components interact with each other. The system is always under ongoing investigation, because we always want to know what is going on under the bonnet. Whatever the efficacy of their predictions, theories are only as good as their assumptions, and relaxing an assumption is always a positive step.
Consider the following theory’s superb record for prediction about when water will freeze or boil. The theory postulates that water behaves as if there were a water devil who gets angry at 32 degrees and 212 degrees Fahrenheit and alters the chemical state accordingly to ice or to steam. In a superficial sense, the water-devil theory is successful for the immediate problem at hand. But the molecular insight that water is comprised of two molecules of hydrogen and one molecule of oxygen not only led to predictive success, but also led to “better problems” (i.e., the growth of modern chemistry).
If economists want to offer lucid explanations of the economy, they are heading down the wrong path (in fact this is something employers have complained about with economics graduates: lost in theory, little to no practical knowledge).
The fifth problem is one that is specific to social sciences, one that I touched on recently: different institutional contexts can mean economies behave differently. Without an understanding of this context, and whether it matches up with the mechanics of our models, we cannot know if the model applies or not. Just because a model has proven useful in one situation or location, it doesn’t guarantee that it will useful elsewhere, as institutional differences might render it obsolete.
The final problem, less general but important, is that certain assumptions can preclude the study of certain areas. If I suggested a model of planetary collision that had one planet, you would rightly reject the model outright. Similarly, in a world with perfect information, the function of many services that rely on knowledge-data entry, lawyers and financial advisors, for example-is nullified. There is actually good reason to believe a frictionless world such as the one at the core of neoclassicism leaves the role of many firms and entrepreneurs obsolete. Hence, we must be careful about the possibility of certain assumptions invalidating the area we are studying.
In my opinion, Friedman’s essay is incoherent even on its own terms. He does not define the word ‘assumption,’ and nor does he define the word ‘prediction.’ The incoherence of the essay can be seen in Friedman’s own examples of marginalist theories of the firm. Friedman uses his new found, supposedly evidence-driven methodology as grounds for rejecting early evidence against these theories. He is able to do this because he has not defined ‘prediction,’ and so can use it in whatever way suits his preordained conclusions. But Friedman does not even offer any testable predictions for marginalist theories of the firm. In fact, he doesn’t offer any testable predictions at all.
Friedman’s essay has economists occupying a strange methodological purgatory, where they seem unreceptive to both internal critiques of their theories, and their testable predictions. This follows directly from Friedman’s ambiguous position. My position, on the other hand, is that the use and abuse of assumptions is always something of a judgment call. Part of learning how to develop, inform and reject theories is having an eye for when your model, or another’s, has done the scientific equivalent of jumping the shark. Obviously, I believe this is the case with large areas of economics, but discussing that is beyond the scope of this post. Ultimately, economists have to change their stance on assumptions if heterodox schools have any chance of persuading them.