Changes in Scale of Production in United States Manufacturing Industry, 1904-1947

Abstract

T HERE is general agreement that nineteenth century was unparalleled in growth of large-scale production. However, fate of scale of production in twentieth century seems to be lost in a limbo of uncertainty. One investigator comments that the movement towards large-scale production is largely a nineteenth century phenomenon and had run its course by i890. 1 Another commentator holds that size is growing with great rapidity. 2 Yet again we read that long-term, general and pervasive increase in plant size throughout most industries has come to an end. I It is purpose of this paper to present some empirical evidence on changes in scale of production in United States manufacturing industry. By scale of production we refer to size of plant rather than size of firm. The very notion of scale of production implies a relationship between volume of output and unit costs. Where economies of scale bear upon questions of monopoly, problem is one of control of output. Hence, scale of production is measured in this paper by physical output per establishment. Indexes of physical output are available for United States manufacturing industry, permitting construction of indexes of scale of production as measured by an index of output per establishment.4 To measure scale by number of employees would tend to underestimate industrial expansion linked with labor-saving innovations.5 Similarly, trends in ratio of value-added or capital per establishment may diverge significantly from movement of scale. In addition, data from which indexes of value-added can be constructed are not available for a sufficient period of time to be useful while records of capital value are flagrantly unreliable in that they are subject to judgment of person making estimate and to vagaries of longand short-term fluctuations in prices. The interpretation of long-term movements in indexes of output per establishment as changes in optimal plant size need not be vitiated by assumption of an optimum range of output rather than an optimum point.