The archive.

Mathematical precision does not settle whether a model captures the economic problem it claims to solve. In this review of Samuel Karlin’s two-volume Mathematical Methods and Theory in Games, Programming and Economics, Oskar Morgenstern praises lucid proofs and valuable applications while challenging the emphasis on two-person zero-sum games and standard equilibrium models. His objection concerns control: economic participants make decisions under conditions partly determined by other participants, not simply against a fixed background. Readers can discover why he regards n-person, nonzero-sum games as essential to economic analysis—and why his admiration for Karlin’s mathematical achievement coexists with reservations about its modelling choices. His interest in poker as a model of negotiation gives that distinction a concrete, suggestive endpoint.

A single correlation or average lag can conceal different relationships between short-term fluctuations and long-term movements. In this 1961 Princeton research memorandum, Oskar Morgenstern asks what economists could discover by separating those temporal scales rather than imposing familiar business-cycle classifications. Drawing on communication engineering, he advocates spectral analysis as a way to detect rhythms within aggregate data and examine how series interact at different frequencies. His example of New York money-market rates makes the stakes concrete: longer-term components can show greater lags than shorter-term ones. Yet finer measurement is not economic explanation. Morgenstern’s proposal turns on a demanding tension: longer records improve frequency analysis, but also span institutional and technological changes that may undermine its assumptions. Readers encounter a research program in which statistical discovery challenges theory without claiming to replace it.

A faster aircraft or a larger economy does not automatically confer strategic power. In this published address to a naval aviation meeting, Oskar Morgenstern challenges technical forecasts that overlook qualitative change and military calculations that neglect political purpose. His perspective as an economist and game theorist gives the critique its edge: conflict involves an opponent who conceals, bluffs, and deliberately frustrates plans, not merely conditions to be optimized. Ballistic-missile submarines make the distinction concrete—destroying their bases cannot neutralize weapons already at sea. His proposal to measure national power rather than aggregate growth similarly questions what impressive numbers actually establish. The address offers a pointed encounter between technical confidence and strategic uncertainty, including Morgenstern’s contentious Cold War argument for nuclear capabilities in limited war.

We have to make decisions not against nature, but against a consciously interfering adversary.

A symmetric game need not have a fully symmetric solution. In this brief research abstract, Oskar Morgenstern reports incomplete joint work with the late John von Neumann on games whose minimal winning coalitions contain all but one player. He isolates solutions that treat every player symmetrically except one, then examines the set of payoffs available to that exceptional player. Two announced, optimal bounds constrain both the lowest attainable payoff and the gaps between attainable payoffs. The abstract offers a compact example of how symmetry can restrict an asymmetry without eliminating it; the proofs are reserved for a separately announced RAND paper.

A game can treat every player alike without requiring equal payoffs—or even fully symmetric solution sets. In this RAND paper, published in 1961 from a 1946 manuscript, John von Neumann and Oskar Morgenstern investigate a cooperative game in which a coalition wins only if it contains all but at most one player. They impose symmetry on the first n−1 players while leaving the last player outside that requirement. This deliberate asymmetry makes a precise question accessible: how low can the remaining player’s minimum payoff be, and how large can gaps in that player’s possible payoffs become? By redistributing payoffs and testing domination, the authors derive bounds rather than a complete classification. The paper offers a concrete encounter with how symmetry can constrain stable solutions without determining them outright.

Railroad share battles, the advertising war between Lestoil and Procter & Gamble, Coca-Cola against Pepsi, European cartel threats and compensations — business life, this essay argues, is a tissue of strategic situations that ordinary maximization cannot describe. Conceding that game theory stands as an autonomous branch of mathematics, Morgenstern presses the harder claim that economic reality is itself frequently a game: the moment a rival controls a decisive variable and reacts, the normal maximum assumptions fail. Walrasian equilibrium and the crossing of demand and supply curves become special constructions that hide number, timing, and coalition. His strongest preference is for cooperative n-person theory, which can model mergers, wage bargains, proxy fights, and above all 'Macht' — power that marginal-productivity theory has no place for. Where power exists, he concludes, the prevailing theory must give way.

Es gibt kein Geschäft ohne Geschäftsgeheimnis.

English translation: “There is no business without a business secret.”

A satellite can gather weather data and military intelligence with the same instruments: how, then, can states agree to separate peaceful exploration from strategic rivalry? In this 1962 essay from Space Flight Report to the Nation, Oskar Morgenstern tests hopes for international space cooperation against the demands of enforcement. He disputes the usefulness of Antarctica as a precedent and asks who could restrain the very powers capable of policing an agreement. His Cold War distrust of Soviet conduct shapes the argument, but its sharpest distinction is between technical capability and political authority: being able to destroy a satellite does not establish when doing so is justified. The essay offers a compact examination of how discovery unsettles sovereignty—and why scientific cooperation alone cannot supply workable political controls.

Morgenstern’s brief, signed preface places two advance chapters at a precise moment in the revision of On the Accuracy of Economic Observations: on July 9, 1962, he expected the rewritten second edition to reach the printer within weeks. Its interest is documentary rather than argumentative. Readers can establish the chapters’ relation to the larger book and trace the acknowledged contributions of Econometric Research Program collaborators and typist Lois A. Crooks. The preface records publication circumstances and working relationships, not a thesis about statistical accuracy.

Can stock-price changes look random while price levels retain a structure that a random walk cannot explain? In this 1962 research memorandum, Clive W. J. Granger and Oskar Morgenstern use spectral analysis to separate movements at different time scales. Their distinctive finding is not a dependable market cycle, but a methodological tension: differencing prices can obscure slow fluctuations, making a model appear more adequate than it is. Short-run movements largely support the random-walk benchmark; longer-run variation exceeds its predictions. Yet weak seasonal and business-cycle components offer little support for cyclical investment rules. The memorandum shows readers why evidence against a simple model need not yield a usable forecast—and how statistical transformations can conceal precisely the features an inquiry seeks to understand.

Destroying an enemy’s communications might unleash the weapons it was meant to disable. This possibility gives Oskar Morgenstern’s 1962 memorandum, a lightly revised invited address, its central tension: nuclear security depends not only on the capacity to strike but on preserving political control on both sides. Morgenstern treats command as an organizational problem involving human judgment, selective information and resilience under attack—not simply faster signals or more reliable electronics. Safeguards against unauthorized action can impede legitimate orders; economical networks can prove dangerously fragile. His case for survivable communications available to adversaries offers a concrete way to examine why military advantage and strategic safety may diverge, and why an opponent’s ability to restrain its forces and negotiate can be a condition of one’s own security.

Are there limits to the use of mathematics in economics? The question, this essay answers, is wrongly posed: the obstacle lies not in the subject but in whether economists understand what mathematics is and formulate their problems well. Morgenstern dismisses the usual objections — psychology, non-quantitative data, expectations, unmeasurable utility — by noting that mathematics is not merely quantitative, that no deep gulf separates a simple addition from an integration. Economics erred not by too much formalism but by shallow model-building, borrowing equilibrium from mechanics and casting agents as solitary maximizers under fixed conditions. Game theory marks the conceptual break, supplying a mathematics fitted to strategic interdependence; expected utility, axiomatized, replaces cumbersome indifference curves. Since natural science once remade mathematics, he expects social science to do the same, leaving no fixed boundary that can honestly be drawn.

Nichts ist leichter, als die eigene Beschränktheit für die der Methode oder des untersuchten Gegenstandes zu halten.

English translation: “Nothing is easier than to mistake one's own limitations for those of the method or of the subject under investigation.”

When an economic model fails, has mathematics reached its limit—or has the economist posed the wrong problem? In this 1963 research memorandum, Oskar Morgenstern defends mathematical economics while challenging the substitution of formal elegance for economic understanding. Drawing on his work with von Neumann, he argues that uncertainty and strategic interaction demand more than techniques borrowed from mechanics: they may require new concepts and new mathematics. His examples make the distinction concrete. Counting equations does not prove that an equilibrium exists; treating agents as isolated maximizers can miss their dependence on one another’s choices. Readers encounter a defence of rigorous reasoning that grants no automatic prestige to symbols or axioms, and insists that historical, experimental, and statistical inquiry remain indispensable.