Oskar Morgenstern, Wirtschaftsprognose: Eine Untersuchung ihrer Voraussetzungen und Möglichkeiten

Oskar Morgenstern’s 1928 monograph examines whether scientific forecasting of an entire economy is possible and whether it could serve economic stabilization. Its target is the emerging apparatus of forecasting institutes, commercial prediction services, and statistical business barometers. Morgenstern separates the theoretical possibility of prediction from its statistical technique, then distinguishes both from the effects of publishing and applying forecasts. His central contention is stronger than a complaint about inadequate data: economic forecasts concern historically particular outcomes constituted by human actions, and their publication changes those actions.

The book develops this argument in three parts. The first classifies business-cycle theories according to the regularity they attribute to economic movements. The second examines forecasting’s conceptual foundations and then its effects upon the economy. The third extends the inquiry beyond economics to social prediction and economic policy. A concluding discussion proposes a different role for research institutes.

Morgenstern’s initial principle is that forecasting presupposes business-cycle theory. Statistical techniques cannot independently establish which series characterize an economy, which movements constitute phases, or why observed relationships should persist. His classification runs from erratic fluctuations to strict periodicity, with several forms of recurrent but variable cycles between them. Pure periodicity initially appears to permit exact prediction; the more plausible intermediate cases introduce uncertainties about duration, amplitude, causal combinations, and interruptions. Yet even the apparently favorable periodic case remains vulnerable to the effects of prediction itself.

The conceptual center is the distinction between individual forecasts and “total” forecasts. Consumers and entrepreneurs necessarily anticipate future circumstances when acting, although consumers orient their decisions toward overall utility and entrepreneurs toward monetary profit:

Jeder Akt ist bereits Prognose¹).

English translation: Every act is already a forecast¹).

This constitutive anticipation does not establish the possibility of forecasting the whole system. A total forecast concerns the future configuration of interdependent series, themselves composed of actions incorporating individual forecasts. It cannot be obtained simply by adding those forecasts together. Unlike individual anticipation, total forecasting is not necessary for an economy to function; introducing it supplies an additional, potentially transformative datum.

Morgenstern consequently treats prices and statistical series as expressions of behavior, not autonomous objects governed by mechanical movement laws. Trends must be studied in their mutual relationships, rather than extrapolated separately or reduced to a uniform coefficient of progress. Structural changes and broken trends undermine historical comparisons. His distinctive concept of the “time-quality” of prices further complicates the statistical picture: identical nominal prices can occupy different causal positions, while earlier prices remain effective through contracts and unfinished production processes.

Die ganze Wirtschaft ist entschieden in die Zukunft gewendet, hängt aber mit festen Ketten an der Vergangenheit (und: kennt keine Gegenwart).

English translation: The whole economy is decidedly turned toward the future, but is bound by firm chains to the past (and: knows no present).

A statistical cross-section therefore combines quantities with different histories and unfinished effects. Recorded magnitudes do not exhaust the relationships determining subsequent movement. Likewise, observed time-lags differ fundamentally: some reflect technical constraints, others adjustable production arrangements, and others changeable behavior or social power. The lags most useful as forecasting signals are often the least securely fixed.

His attack on frequency-based statistical induction follows from this account. Insurance statistics predict numbers of interchangeable occurrences; total forecasting must identify a particular historical development in a causally connected sequence. Morgenstern argues that the former procedure cannot establish the latter. Comparisons among completed cycles also encounter insufficient numbers, heterogeneous institutions, and changing relationships between symptoms and causes. These are arguments against treating statistical regularities as self-sufficient predictive warrants, not against statistical research.

The decisive test concerns an authoritative published forecast. If rising prices are announced and believed, buyers advance purchases, sellers retain stocks, and entrepreneurs accelerate orders. The resulting movement differs in timing, intensity, and structure from the announced development:

Wenn zwischen einer mit Autorität bekanntgegebenen Prognose und dem vorausgesagten Geschehen ein Kausalzusammenhang herstellbar ist, kann solche Prognose prinzipiell nicht wahr werden.

English translation: If a causal connection can be established between a forecast announced with authority and the predicted events, such a forecast cannot in principle come true.

Correcting the forecast does not resolve the difficulty: the correction becomes another behavioral datum requiring further correction. The Holmes–Moriarty pursuit illustrates this regress of mutually anticipated decisions. Competing forecasts, partial acceptance, and restricted circulation add uncertainties about who believes what. Publication can thus destroy the comparability of the historical evidence supporting prediction. Morgenstern’s stringent requirement that a forecast actually come true is crucial to his sweeping negative conclusion.

The third part argues that economic prediction would also require anticipating political, religious, natural, and other changes treated as external data by economic theory. Enlarging forecasting into general social prediction merely expands the problem of historically novel combinations. Nor does anticipation necessarily stabilize the economy: it may accelerate or intensify movements. Secret forecasts escape publication’s feedback, but applying them reintroduces the question of their effects.

The conclusion explicitly distinguishes practical recommendations from the preceding theoretical argument. Institutes should abandon total prediction while improving timely information about present conditions:

Maximalisierung der Publizität ist die Forderung.

English translation: Maximization of public disclosure is the demand.

Public knowledge of inventories, orders, employment, capacity, and costs should improve decentralized individual anticipation. Morgenstern acknowledges that securing disclosure may require coercion; this is no simple doctrine of nonintervention. The book’s enduring relevance lies in its account of expectations and predictive feedback: information can improve coordination without making the economy’s future mechanically calculable.