What do scientists think about economic modeling?
Similar to models used in physics or engineering, economic models are formal models. However, they cannot be easily tested and improved in experiments and thus often have limited power for making quantitative predictions. There has been a long scientific debate about what economic models are and for what purposes they are useful.
Gibbard and Varian: Models as stories with a structure
Gibbard and Varian have made an important contribution to this debate in 1978 (Journal of Philosophy, Vol. 75, pp. 664–677). They explain that:
‘A [theoretical] model […] is a story with a specified structure. […] The structure is given by the logical and mathematical form of a set of postulates, the assumptions of the model. […] The theorems that follow from the postulates tell us things about the structure that may not be apparent from an examination of the postulates alone. Although the term ‘model’ is often applied to a structure alone, we shall use it in another sense. In economists’ use of models, there is always an element of interpretation: the model always tells a story.’ (p. 666)
Thus, in their view, economic models are a combination of formal work (deriving results from equations) and a story that interprets these results and relates them to the actual world.
A somewhat surprising element of their view is that the application of the model to the real world can be done in a rather casual way. Furthermore models are often only approximations:
‘Much of economic theorizing consists not of an overt search for economic laws, not of forming explicit hypotheses about situations and testing them, but of investigating economic models. […] The hypothesis may be that the conclusions of an applied model are approximately true, and that that is because its assumptions are sufficiently close to the truth.’ (p. 676).
Sugden: Modeling credible worlds
Sugden (2000, Journal of Economic Methodology, Vol. 7, pp. 1–31) uses this as a starting point and asks why we should believe that a model is indeed a good approximation, if we cannot ensure this by using controlled experiments.
He uses two examples to highlight Gibbards and Varians view of modeling but then states that:
‘[…] Gibbard and Varian have disappointingly little to say about how a casual model explains an aspect of the real world, or how it allows us to investigate the likely effects of real-world factors on real world phenomena.’ (p. 13)
Sugden then argues that for applying economic models to reality, we often need an ‘inductive leap’: In our model some result X is caused by some economic mechanism Y. We observe that X holds in some cases in reality and thus – using an inductive step – conclude that the mechanism Y is at work in the real world. As the last step is not a logical conclusion (we cannot go from the specific to the general using only logic), we need to justify it.
Sugden provides several ways for such a justification. Among them is robustness. If a result is robust (it holds not only for some specific assumptions, but in a wide range of settings), it is obviously much easier to argue that this result is likely to hold in reality.
However, robustness does not suffice. In addition, we need to be convinced that the model used to derive the result is credible, in the following sense:
‘Credibility in models is, I think, rather like credibility in ‘realistic’ novels. In a realistic novel, the characters and locations are imaginary, but the author has to convince us that they are credible – that there could be people and places like those in the novel. As events occur in the novel, we should have the sense that these are natural outcomes of the way the characters think and behave, and of the way the world works. We judge the author to have failed if we find a person acting out of character […]. But we do not demand that the events of the novel did happen, or even that they are simplified representations of what really happened.
For a model to have credibility, it is not enough that its assumptions cohere with one another; they must also cohere with what is known about causal processes in the real world.’ (p. 25–26)
Thus in Sugden’s view, models need to be a credible description of the real world.
Summing up, economic models are neither accurate descriptions of the real-world nor are they purely formal exercises without any link to the world we live in. Economic models are used to tell consistent stories in a model world that is a simplified but credible description of our real world. This is what makes them so useful for exploring possible futures.
© University of Basel