Knowledge that Transforms

ITFALLS encountered in the use of Soviet statistics have been the subject of numerous articles on the Soviet Union. At the same time, however, there has been a considerable amount of agreement that data obtained from official pronouncements and statistical publications have meaning and significance.1 At least it is widely held that the Russians do not engage in outright falsification. Some additional evidence concerning the reliability of Soviet statistical data is contained in a recently released docu-

PREDICTION usually requires counting, dating, and measuring, and so plunges us at once into semantics, the perpetual bugbear of statistics, the endemic disease that more or less infects all our figures. For statistics begins with counting, and counting begins with definition of the thing to be counted -and inventions can hardly be defined. Patents are not quite so bad, and we shall take them up later. Inventions, like many things and notably people, don't stay the same, but are perpetually evolving, growing, squirming out of their definitions. You may have heard of the Irishman who complained that his pig wouldn't stay counted. That is supposed to be a joke on the Irishman; but he was dead right it was the pig that was false and unstatistical. For that pig began as a little shoat, and ate and grew night and day until he became a 5oo-pound hog, a creature utterly different in every pertinent aspect from the little suckling piglet. So he couldn't stay counted Similarly in invention, suppose we consider television. That sounds simple, since we all know what television is. Or do we? Was it already television in I847 when Souvestre satirically predicted it? Or did it begin in I877 when the first apparatus was built, or in I882 when the scanning disk was added, or in I9OI when Fessenden designed a wireless system? Or was it Zworykin's modern cathode ray receiver of I929 that constituted the invention of television, with the kinescope, and some experimental broadcasts the next year? Or is our date I928 or I937 when regular broadcasting began? Or should it be some future date when with color, three-dimensional vision, binaural hearing, and worthy programs, the art will at last enter the prodigious destiny of the home theater? And is this television, at whatever stage, to be counted always as invention? If so, our statistics mean no more than that one pig. Definition of invention

W E regard economic phenomena as resulting from an interaction of human actions within a field of nonhuman environmental conditions and restraints. Of the various phases of economic phenomena that may interest social scientists there is one which is of particular importance to policy-makers whether they are in business or in government. One of their special needs is for knowledge concerning the consequences or impacts of the various actions which they are able to take and which they consider taking. This paper deals with methods that more adequately attempt to meet this need. We assume that the policy-makers know what they are seeking to achieve. We also assume that they have a number of actions at their disposal and that they wish to know which of these will achieve the desired objectives in as satisfactory a manner as possible. We take it for granted that it will usually be necessary to observe discrepancies between the desired situation and the actual situation in order to guide whatever actions are taken, but we will not concern ourselves here with problems of this sort. Rather we will limit our attention to certain aspects of the problems involved in discovering and specifying the consequences of actions.

E VER since the General Theory, continued experimentation with various sets of independent variables has produced numerous consumption functions. The bulk of this work was completed during or immediately following World War II, and therefore little of it has been tested with respect to the postwar period. Some may object to the use of the postwar years for testing purposes, on the grounds that the economic system had not returned to normal conditions, especially in the sphere of consumers' expenditures, owing to the existence of the deferred demand which had grown to immense proportions as the result of five years of war production. But, as we are now engaged in a staggering rearmament program coupled with a foreign war, these immediate postwar years of I946-50 appear as close an approach to the normal period as may legitimately be expected in the near future. Thus, it seems opportune to take stock of some of the existing contributions to see if we may find a consumption function suitable for use as a tool for prediction. The functions which we shall endeavor to test for the postwar period fall into three groups.

T HE index of output per unit of total input that is discussed in this article is intended to describe the pattern and magnitude of technical change for the United States as a whole from I869 to I938. Based on national output and input, the index possesses the following attributes: I. Output equals gross national product in I929 prices (the Kuznets series). 2. Input equals the sum of labor, land, capital, and enterprise inputs in I929 prices. a. Labor input, involving separate weighting of agricultural and non-agricultural labor by I929 wage rates, is presented in two variants one measuring labor in manhours, the other, in man-years. Two alternate total input measures result. b. Other than capital consumption, property input (including enterprise) is determined by weighting the given-period's combined value of land and reproducible producers' wealth in I929 prices by the I929 rate of return on property. c. Changes in the amount of farm acreage dominate the land component. The land valuation weight, however, reflects the I929 value per acre of both urban and farm land. d. Reproducible producers' wealth in I929 prices is derived from one of Kuznets' series. e. Capital consumed in I929 prices is included in nonlabor input. 3. Input and output are presented in overlapping decade averages. We may now note the leading merits and shortcomings of the indexes. Formally our statistics measure output and input for the period in I929 prices. Our objective is, however, to measure them as though valued by baseperiod buyers. Since relative quantities of different products and resources varied over the period, we can assume that base-period buyers would have valued each ingredient otherwise than in the base period. Because of these variations, the precise relation of our indexes to the true valuation of aggregate inputs and outputs is indemonstrable. This flows from the permanent and insoluble defect of index numbers. For the index exactly to fulfill its intended purpose, the measurement of technical change, linearity should pervade transformation relationships between goods, substitution relationships between factors, consumer indifference relationships between goods, and production relationships between factors and goods (including constant returns to scale). Though these requirements are not met, so long as the defects are not destructive of all significance of results, a defective tool is preferable to none at all. For shorter-period comparisons, an index number can give fairly valid results. For longer-term comparisons, the uncertainties mount, but the need for at least a rough approximation remains. Linked by a common language, government, and monetary system. America's economy with all its manifold diversity has been a fairly coherent whole. Averages of attributes of its parts are not only computationally possible: the averages themselves in a real sense significantly influence the parts. For this reason aggregative measures, conceptual flaws notwithstanding, are indispensable. That the secular and cyclical position of the economy exerts great force upon the corresponding characteristics of its components cannot be gainsaid. In this paper adumbrations of one.more such influence will appear. As we shall also see, our indexes describe patterns of technical change corroborated by independent vidence. We shall discuss first the components of the indexes, and then compare the aggregative efficiency index with similar indexes for major industrial segments.

M OST economic cost studies have been concerned primarily with the relation of cost to rate of output. Shortrun costs are usually said to be those associated with variation in the utilization of fixed plant or other facilities, whereas longrun cost encompasses changes in the size and kind of plant. Strictly, then, the distinction is based upon the degree of adaptation of all input factors to rate of output.1 However, cost may vary because of changes in technical knowledge.2 Economists have explicitly excluded all irreversible changes in technology. Most longrun cost theories, for instance, are timeless; one future point in time is selected at which output rate and facilities are permitted to change. That such a cost function, particularly its height, will be affected by improvements in technical knowledge is beyond doubt. It is convenient to clarify the issue of the different cost functions by referring to production functions, which express the net relation between the input of variable productive factors and output during a given production period, under the assumption of a given plant and technical knowledge. From the production function we can derive a static shortrun cost function, which also assumes a given plant and technical knowledge. Longrun cost permits changes in the size and kind of plant, but assumes stability in technical knowledge. Thus, a longrun cost function is related to points on different production functions, each point involving a different plant while using the same technical knowledge. There can be a cost function which permits changes in technical knowledge but not in plant and other facilities. In a sense this is a dynamic cost function. If direct labor is the cost we consider, we shall speak about a (unit) learning or progress function. This expresses the net relation between the amount of direct labor needed to produce one product-unit and the cumulative units produced in a given facility. The progress function thus permits us to estimate the amount of direct labor needed to manufacture the nth unit, from n, the cumulative number of the product-unit. The function is related to a number of points on different production functions involving successive changes in technical knowledge in a given facility. In this study we will analyze data of a large United States machine builder, who has been one of the largest machine tool 3 manufacturers for about three-quarters of a century. Since the end of World War II the company has also manufactured textile and construction machinery. Of its twenty products, seven have been selected for an empirical analysis. They are either new products or new models. Machine i is a new semi-automatic machine tool and machines 2-5 are new automatic machine tools which differ greatly in the functions they perform, weight, labor requirement, etc. Machine 6 is a textile machine and machine 7 is a multipurpose construction machine. Production is carried out in one large plant. During the period under consideration, I946-50, the monthly direct labor force varied from 930 to 3,600 men. The production process is very complicated, calling for high professional skills. It is divided into machining of parts and their assembly. Each month top management decides on a production schedule which instructs the production manager to have ready for shipment by a *The author would like to express his gratitude to the Bureau of Business and Economic Research of the University of California for financial support of this study. 1J. Dean, Managerial Economics (New York, I951), p.