Paul Hoyningen-Huene (Leibniz Universität Hannover)
In a series of papers from 2000 on, Robert Sugden has analyzed the epistemic role of theoretical models in economics. His view is that these models describe a counterfactual world that is separated from the real world by a gap. This gap has to be filled if the model should have an epistemic function for our understanding of the real world. According to Sugden, this gap “can be filled only by inductive inference”. The putative inductive inference that Sugden constructs leads “from the world of a model to the real world”, based on “some significant similarity between these two worlds”. In philosophy, the “significant similarity” that Sugden correctly adduces for the legitimacy of inductive steps has been spelled out as common membership to a natural kind. However, for Sugden’s inductive step to be legitimate, the union of the appropriate set of models with the appropriate set of real target systems should form a natural kind, which is certainly not the case. For instance, with respect to causality model cities are utterly different from real cities, contrary to Sugden: models may at best represent the real causality.
In fact, the inferential step from models to reality is abductive, as Sugden correctly notes. However, he misunderstands abduction as a sub-category of induction. Yet, abduction does not lead to generalizations as induction, but to risky explanatory hypotheses.
The abductive inference from a model to reality has the following form:
(i) x has property Z (empirical finding),
(ii) Situations of type A have property Z (model),
therefore
(H) x is a situation of type A.
If (H) is true, then (H) together with (ii) explain (i). However, the abductive step to (H) is risky, because it may also hold:
(ii*) Situations of type B have property Z, with B ≠ A.
Based on (i) and (ii*), one gets by abduction the alternative explanatory hypothesis
(H*) x is a situation of type B, with B ≠ A.
Thus, all one gets by an abductive step is a potential explanation (sketch). The only way to obtain the actual explanation is by showing that the model situation is sufficiently similar to reality and by excluding all alternative explanations. Thus, the real explanation is not distinguished from alternative explanations by an intrinsic property of high credibility, as Sugden assumes, but by its comparative advantage against competitors.
The upshot is that a theoretical model in economics (like Schelling’s) never directly explains any particular empirical case (this resolves Reiss’ “explanation paradox”). Instead, a model allows for abductive generation of a sketch of a potentially (perhaps surprising) explanatory hypothesis. In order to transform this potential explanation sketch into an actual explanation, the sketch must be elaborated and its empirical adequacy be shown. The latter crucially contains the exclusion of alternative potential explanations. This may be accomplished by showing that the empirical conditions necessary for plausible alternative mechanisms to work do not obtain.
