Arnold [2] introduces the subject of structural stability in the following words: “In every mathematical investigation, the question will arise whether we can apply our mathematical results to the real world. Indeed, let us assume that the result is very sensitive to the smallest change in the model. In such a case, an arbitrarily small change in the model (say a small change of the vector field defining a differential equation) leads to another model with essentially different properties. A result like this cannot be transferred to the real process under consideration, because, when constructing the model, the real situation was idealized and simplified.. . Consequently the question arises of choosing those properties of the model of a process which are not very sensitive to small changes in the model, and thus may be regarded as properties of the real process.” Thus, structural stability of the models constructed is a question of prudence in scientific research. As an illustration Arnold takes the model of a pendulum. A model including friction is structurally stable, because a small change of the friction coefficient only modifies the approach to equilibrium. A model without friction, on the other hand, is structurally unstable, as it represents an isolated case between damped and explosive motion, and would be thrown into one of these classes by any change however small it is. Arnold’s point is that even if we knew nothing about friction in real life we should include it in order to produce a structurally stable model. An economic analog is the case of classical multiplier-accelerator models of the business cycle, which in the linear version can produce damped, explosive, or simple harmonic oscillations. If we wish to propose the model as an explanation of real business cycles, then we should choose the damped variant. It might be tempting to choose the case of simple harmonic motion, but we are advised by Arnold not to believe that the parameters should be in that exact relationship which produces the simple harmonic case. Even if we had estimated such a combination of parameters we should consider all the factors we have neglected by the abstraction process. The illustration from business cycle theory also serves to make clear the difference between