Published in Spanish in 1972 as a translation of the 1971 French article, Morgenstern’s methodological essay challenges economics’ claims to numerical precision. Its target is not quantification itself, but the failure to distinguish numerical detail from reliable knowledge. Moving from statistical observation through national accounting and econometrics to computer calculation, he argues that economic science must acknowledge uncertainty at every stage. Scientific rigor means investigating errors and their consequences, not presenting uncertain results as exact.
The opening account of Laplace’s unsuccessful request that the French census publish an estimate of its possible error establishes the central contrast. Physics laboriously converts qualitative observations into quantities and therefore scrutinizes measurement. Social life already supplies prices, monetary totals, production figures, and population counts. These apparently ready-made numbers encourage economists to mistake careful transcription and aggregation for adequate scrutiny. Even classification involves discretionary decisions: assigning a mixed cotton-and-nylon fabric to natural or synthetic products changes the resulting statistics.
Morgenstern’s discussion of uncertainty does not demand perfect observations before theorizing. Newton and Einstein developed powerful theories while the relevant measurements still contained substantial errors. Their example establishes a constructive alternative to both spurious exactness and skepticism:
Lo que precede supone una lección importante, la de que teorías poderosas pueden formularse partiendo de nociones bastante poco satisfactorias, y verificarse por medio de esas mismas nociones, de manera suficientemente probatoria.
English translation: The foregoing offers an important lesson: powerful theories can be formulated on the basis of quite unsatisfactory information, and verified by means of that same information in a sufficiently convincing way.
The requirement is that theoretical claims fit the quality of available evidence. Morgenstern contrasts this discipline with national-product figures implying precision to tiny fractions of a percentage point. Discrepancies between trading partners’ records, revisions to American money-supply growth, and census undercounts illustrate how uncertain even carefully collected statistics remain. Such errors need not make the figures useless; treating them as nonexistent is the unacceptable step.
Sería deseable, por lo tanto, poder ver estadísticas en las cuales cada cifra estuviera acompañada por la indicación de su desviación típica.
English translation: It would therefore be desirable to see statistics in which each figure was accompanied by an indication of its standard deviation.
The obstacle is partly political. Governments dependent on legislative funding prefer not to display uncertainty, while parties defending either Marxism or capitalism claim conclusions more definitive than their evidence permits. Morgenstern reports that a German advisory body’s attempt to publish error margins was stopped by the government. Concealing uncertainty thus protects authority at the expense of scientific improvement and encourages unwarranted confidence in policy prescriptions.
The section on gross national product deepens the critique: defective concepts matter alongside inaccurate observations. A scalar total cannot represent the changing composition of an economy. Congestion, wasted fuel, and repairs after catastrophe can increase expenditure and therefore register as growth, turning failures into apparent achievements.
La economía es demasiado compleja como para que se la pueda representar por una cifra única.
English translation: The economy is too complex to be represented by a single figure.
This is not merely a complaint about insufficient data. Morgenstern questions what the aggregate measures and whether its movements warrant judgments about progress. More accurate calculation cannot by itself correct a misleading conceptual framework. He likewise separates observational precision from predictability: stock-market quotations can be precisely recorded while future movements remain deeply uncertain.
The econometrics discussion rejects the opposition between theory and measurement. Neither advances independently: empirical material enables new concepts, while theory guides meaningful observation. Yet increasingly refined mathematical methods often remain detached from applications, especially across countries with different and changing economic structures. The sophistication of a model must be assessed against the quality of its inputs. Successful prediction alone cannot settle the issue:
Aun en el caso de teorías falsas son posibles las previsiones exactas.
English translation: Even in the case of false theories, accurate predictions are possible.
Ptolemaic eclipse prediction supplies the example: predictive success establishes neither the truth of a theory nor the accuracy of its underlying observations. Morgenstern consequently opposes the unqualified presentation of economic knowledge in government and teaching.
The concluding sections make computation the decisive test of these arguments. Computers offer exceptional opportunities for simulation, but cannot manufacture reliability from inadequately examined inputs. Aggregating individuals into a few variables reduces computational burdens while creating further conceptual problems. Morgenstern distinguishes four sources of error: simplification in constructing the model, observational error in its parameters, numerical approximation of mathematical operations, and unavoidable rounding during calculation. He gives particular weight to parameter errors and the sensitivity of results to them.
The final paired systems of equations show how almost imperceptible coefficient changes can produce enormous changes in solutions. Together with an example in which a wage-coefficient error sharply alters estimated profits, they turn a general plea for caution into a question of model vulnerability. The article’s enduring relevance lies in this connection between measurement, conceptual adequacy, computational stability, and political judgment: greater mathematical elaboration warrants confidence only when the uncertainties it processes have themselves been examined.
This work was divided into 6 sections when it entered the library's research corpus—an apparatus for search and citation, not necessarily the author's own table of contents. Each title opens its summary.
Put a question to this work; the Librarian answers from its 6 sections and cites the passage.
Ask the Librarian