Wantrup's Work 3

Recent readers may have gathered that Wantrup [prior posts] was not a big fan of formal economics. To him, they suffered the defect of unrealistic assumptions. In this 1961 paper [pdf], he critiques linear optimization methods (a popular means of finding the optimal solution when multiple constraints and/or activities are possible) as useful but perhaps misleading, i.e.,

I should like to submit that optima of social welfare are constructs in the sense of useful scientific fictions. Such constructs are not operational policy objectives. There is danger that the two are confused when formal programming is used to obtain quantitative optima for large social aggregates.

On the other hand, as fictional constructs, social optima are useful as organizing principles for the great number of controllable and uncontrollable variables and kinds of relations that must be considered in welfare economics -- to decide which ones to bring into a particular investigation explicitly, which ones to neglect, which ones to combine with others, and which ones to take into account as constraints. In policy investigations, special care is needed to decide whether social institutions should be treated as constraints.

[snip]

It may be concluded from this sketch of three important limitations that formal programming is most useful when the model is little concerned with values, when the treatment of social institutions as constraints is logical, and when the influences of time and uncertainty are small. Stated differently, formal programming is more useful in engineering than economics, more for private than for policy decisions, and more under static than under dynamic assumptions. Usefulness for natural resource policy is limited on all three counts.

Note that he also takes the time to differentiate between risk and uncertainty -- making the very relevant point that programming models have a hard time including uncertainty (because it cannot be quantified in a probability distribution). The typical response -- to ignore and omit uncertainty -- is not acceptable, since the model will fail to capture important features of reality.

Bottom Line: Listen to Wantrup! Pay attention to your modeling assumptions, lest your garbage in produce garbage out.