…is from Armen Alchian’s 1950 address to the Pacific Coast Economic Association, as reprinted on page 635 of the Liberty Fund’s 2006 volume of Alchian’s work: Choice and Cost Under Uncertainty (emphasis added):
The real issue [of the debate about economic methodology and mathematical models] relates to the economic substance – the applicability to real economic problems, or what I call here economic validity – of the mathematical models. And with regard to this, it is submitted here that the mathematical analysis reveals very clearly the true scope and structure of the analysis, thereby facilitating an evaluation of its economic validity. One may well argue that the empirical invalidity, or unrealism, of so many mathematically expressed theories and hypotheses is simply a result of their clarity, which reveals quickly the applicability of the system, whereas in literary theorists it is difficult to make such an appraisal so readily. In any event, the mathematical approach must not be treated as the new delivery truck was treated by the delivery man who became so absorbed in the operating characteristics of the truck that he neglected to deliver the goods.
JMM: As is characteristic of Alchian, he delivers a lot of economic insight into a small area. Mathematical models have an important role to serve in economics: they can sharpen our thinking by pointing out logical flaws in the reasoning. They can help solidify the pattern predictions we make by testing them against real data. But, as Alchian warns, we should not become so obsessed with the working of our models that we economists forget that our primary duty is to provide insight into economic matters. If our models do not do that, then they are failing in their job, as are we.