Oskar Morgenstern, Descriptive, Predictive and Normative Theory

Oskar Morgenstern’s article, first published in 1972 and reprinted in 1976, examines the connected tasks of describing economic activity, predicting outcomes, and guiding action. Its central methodological concern is the adequacy of theories to a reality shaped by expectations, strategic decisions, and changing institutions. Formal rigor is essential, but neither mathematical presentation nor predictive success alone establishes a satisfactory economic explanation.

Morgenstern begins by questioning methodology detached from scientific practice:

It is a serious question whether much new can be said about methodology by approaching this area solely from a philosophical standpoint.

Axiomatization offers a means of making reasoning explicit and rigorous, not a vocabulary that automatically validates a theory. The proof underlying von Neumann–Morgenstern numerical utility supplies a benchmark for distinguishing demonstrated results from the mere appearance of rigor. Yet axiomatic development does not eliminate historical observation or preliminary conceptual work:

It is by now fortunately and finally clear that preliminary description and rigorous theory interact, that both are needed to make a science progress.

This interaction also governs measurement. Economic quantities depend on theoretical definitions; collecting more data cannot by itself correct inadequate categories. Morgenstern criticizes concepts inherited from the analysis of physical goods for their limited grasp of an economy increasingly organized around services. Aggregate measures such as GNP compress complex relations and may register malfunction as growth, while deterministic models insufficiently acknowledge measurement error. His objection reaches beyond statistical technique:

The economy is never described adequately by a mere listing, no matter how detailed and sophisticated, of physically tangible objects or entities or even prices.

Economic description must capture the decisions, expectations, plans, and valuations that move physical objects and give them economic significance. Retrospective accounting can represent this network only partially. The criticism also informs Morgenstern’s treatment of Walrasian competition: models that exclude agents’ influence on one another suppress the strategic relations central to economic activity. Game theory addresses situations structured by cooperation, conflict, and anticipation, although logical feasibility remains insufficient without realism.

The discussion of prediction distinguishes general implications of a theory from forecasts of particular historical events. False theories may predict successfully, while sound theories may lack the information needed for detailed forecasts. Comparing weather prediction by historical resemblance with prediction through mathematical mechanisms, Morgenstern favors structural explanation over extrapolation. Economic prediction encounters an additional difficulty: agents can learn a theory and change their conduct because of it. The dissemination and acceptance of economic knowledge consequently become part of the economy being explained.

Feedback does not make every economic proposition self-defeating. The disappearance of exploitable stock-market regularities and the protection supplied by appropriately mixed strategies illustrate results that can withstand being known and applied. Nevertheless, concrete forecasts remain dependent on political, legal, moral, and ideological circumstances. Social inventions, including new monetary practices and organizations, alter the mechanisms that theory must explain. Economics cannot therefore become a completed deductive system. Education should prepare students for unfamiliar situations, while controlled experiments—especially those encouraged by game theory—offer opportunities to improve predictive understanding.

The final discussion explains how descriptive theory can guide action without erasing the logical distinction between facts and values. Normative force is conditional on accepted aims, institutions, or theoretical principles. Thermodynamics can rule out a proposed perpetual-motion machine without choosing an agent’s purposes; expected-utility theory can reveal behavior inconsistent with an accepted preference system. Similarly, accepting private property supports a prohibition on theft but does not determine an appropriate punishment.

Morgenstern thus connects description, prediction, and prescription without collapsing their differences. Theory guides conduct through accepted premises, yet its practical application remains constrained by uncertain information, imprecisely defined institutions, and historical change. The article joins formal ambition to methodological restraint: economics must analyze agents whose learning and strategic conduct can transform the conditions under which its own propositions apply.