Knowledge that Transforms

10. LE BUT de cette communication est d'examiner les problkmes que l'on rencontre dans la construction des courbes statistiques de demande et d'offre, a cause des rapports d'interdependance existant entre les differents prix. En th6orie l'on peut admettre que tous les prix sont interdependants. En pratique toutefois l'on peut considerer interd6pendants (en telle mesure au moins qu'il vaille la peine d'en tenir compte) seulement les prix des biens qui se trouvent entre eux en relations 6conomiques directes. D'apres la nature de ces relations les diff6rents biens peuvent se diviser en groupes diff6rents, dont les plus importants sont les cinq suivants: (1) groupe de biens A production associ6e; (2) groupe de biens de remplacement; (3) groupe de biens compl6mentaires; (4) groupe de biens concurrents du cote de l'offre; (5) groupe de biens lies entre eux par des rapports d'instrumentalit6. Nous allons nous occuper des deux premiers groupes seulement. Mais les conclusions auxquelles nous arriverons sont suffisantes A nous montrer la complexit6 de ces relations et la necessit6 d'en tenir compte, si possible, dans la construction des courbes de demande et d'offre. 20. En ce qui concerne le premier groupe il faut d'abord distinguer les biens A production associee qui sont produits A rapport fixe, de ceux qui sont produits A rapport variable. Dans le premier cas l'4quilibre du march6 de ces biens et les changements de cet 6quilibre d&pendent des conditions techniques qui d6terminent le rapport des quantit6s des deux biens. Dans le deuxieme cas ils dependent des conditions qui d6terminent le rapport des quantit6s le plus avantageux pour les producteurs.' En r6gime de libre concurrence parfaite, l'6quilibre se r6alise toujours en relation au principe du coAt de production, mais avec cette difference, en comparaison avec les cas plus simples consider6s par les 6conomistes, que l'egalitW doit se r6aliser non pas entre le prix et le cott marginal de chaque marchandise, mais entre le coitt marginal global et la somme des prix des fractions d'unite des biens associ6s que l'on peut obtenir de chaque unite de production

INVESTIGATORS attempting to measure the forces responsible for the variations in the price of an individual commodity have almost always recognized that the supply and price of other commodities which might be used for the same purpose might influence both the price at which a given quantity of the commodity would be sold and the quantity of the commodity which would be purchased at a given price. Thus the production of corn has been considered as a factor in studying the price of oats;' the price of beef steers has been considered in studying the price of hogs ;2 hog prices have been considered in studying beef prices;3 supplies of other fresh fruits have been examined in studying cantaloupe prices;4 prices of hogs, steers, and veal, have been included in studying lamb prices;5 supplies of sweet potatoes and corn meal have been considered in analyzing movements in rice prices;6 and similar efforts have been made to recognize the influence of competing products in other studies of prices. In a recent brilliant paper, Bean and Thorne showed that changes in prices which previously had been ascribed to secular trends in demand were really related to changes in consumer purchasing power as indicated by the index of payrolls.7 This analysis included the price of beef as a factor in explaining changes in

THERE has been, to some extent, a divergence of purpose between the work of mathematical statisticians and mathematical economists. The former have studied variation, such as the form and measurement of frequency groups, and the correlation between two or more variables; the latter have studied equations of equilibrium, in which the data represent totals, margins, the representative, the average-single quantities in general-where heterogeneity does not enter into the argument. Econometricians have, indeed, worked on variation and covariation, but most attention has been given to examples where time is one of the variables, that is to time series; and the relationships obtained have been in the form of a demand curve in one group of studies, to periodic curves in time in another, and to correlation between quantities which are paired in time, but not variable at a given instant of time. I wish to direct attention to a field which has not been thoroughly explored, and in which, so far as I know, little work is being done. What are the forces which produce the variation found in common phenomena, how far are they economic, what changes in frequency distributions are produced by economic forces, how can they best be described or measured? To name one example, which has already been studied: Incomes in certain circumstances are distributed in accordance with Pareto's equation. What has produced this particular form; is the form modified by economic change, or rather by changes in the sociological or legal spheres; how can the effects of these best be measured? Variation has been studied in many fields and it may be well to mark off those which are not primarily economic from those which are. We will exclude anthropometry and biometry, and statistics of mortality and morbidity. Demographic statistics generally would enter only in their economic aspects, such as age distribution in occupations, and concentration of population. Within the economic sphere proper, there are many distributions which have been the subject of careful work: price changes in time, local variations in price, prices of securities, size of industrial firms, rates of profit and dividends, ratio of

THE econometric approach to statistics is a rather ambitious one. The econometrician is not content with mere description and historical or geographical comparison of economic data. He aims at obtaining invariant features of economic phenomena, invariant, at least, as long as the main institutional conditions of a given economic order do not substantially vary. He considers economic invariants as an aim worthy of scientific endeavor, even if they prove to be confined to some very general characteristics common to a loose family of curves. Of course, the numerical formulation of such is in itself again nothing more than a description of facts. But in the econometric approach an endeavor is made to reduce these facts to rules which can stand the test of applicability to, or of forecasting of, other facts. Among recent statistical works having immediate econometric interest, three more important groups may be discerned: (1) attempts towards a construction of demand and supply curves of certain commodities (including productive factors), (2) analysis of business cycles, and (3) numerical investigation into the branches of national spending, including saving or investment, and production (possibly connected with research into national stocks and cash), and into time changes in their proportions. Obviously, the three subjects are closely connected with one another. The proportions between the main economic currents depend on price ratios between commodities and/or productive factors; and the analysis of business cycles, if it is more than an enumeration of symptoms, is hardly possible without knowledge of the reactions to which the different commodities and productive factorsand consequently the different streams of production and spendingare likely to be originated by given changes of the circumstances. The three groups of research have, however, some differences in the technical apparatus employed, and these permit a separate review. We are concerned, in this review, with the third group only. On questions recently discussed in economic theory, such as the influence of profits or of capital accumulation on national productivity, some light may be thrown if we know the order of magnitude of the parameters concerned, and the intensity and velocity of their mutual reactions.

NOT long ago it was stated with some emphasis that a seeming impasse in value theory has been reached, that the systems and points of view of five leading contemporary Anglo-American economists, considered representative and outstanding in England and the United States, are rather bewildering in their contradictions and lack of concordance, and that way out is in sight.' The way out, if there is one, is assuredly the path pursued by other sciences in shaking off the incubus of pseudo-scientific method in the course of their development: by the geometers of ancient Greece who, following the analyses of Aristotle and Plato, freed themselves from Pythagorean mysticism in establishing the system of Euclid; by William Gilbert, who much later proved by painstaking experiment that neither goat's blood nor garlic had anything to do with the phenomena of magnetism; by Francis Bacon, who pled with his contemporaries to cease spinning laborious webs of learning out of no great quantity of matter, and test their ideas against the actual facts of experience; by Robert Boyle, who urged students of chemistry to write plainly and unostentatiously about their experiments and thereby keep men from being stunned, as it were, or imposed upon by dark and empty words. If an impasse has been reached in value theory, it may not seem wholly unwarranted to venture the suggestion that this phase of economic study may be in about the same position today that mathematics was in the days of Pythagoras, that physics was in the days of Gilbert, or that chemistry was in the days of Bacon and Boyle. These sciences have passed their periods of impasse and are making steady headway because they have learned to distinguish the scientific from the pseudo-scientific, to regard hypotheses as tentative and of existential worth unless or until verified, and to eschew dogma of whatever variety. Such reflections, it should be added, if pertinent at all, are meant to apply mainly to value theory with its subjective implications, and not