Knowledge that Transforms

THE PERIOD SINCE World War II has witnessed the rapid development of sample surveys of businessmen's short-term plans and anticipations. Such surveys are now conducted periodically in most of the countries of the Western world that can be classed as industrially advanced. Their common characteristics are (1) the querying of enterprises concerning their expectations about a number of variables entering into their operating plans; (2) the registration of the reports in qualitative rather than quantitative form; and (3) the compilation of corresponding qualitative reports on recent trends of the same and related variables. The principal compiler of survey data of this type in the United States is Dun & Bradstreet, Inc., with main headquarters in New York City. The present essay represents one phase of a research project aimed at explicating the nature and usefulness of such data, with particular reference to the Dun & Bradstreet surveys. The essay begins with an effort to relate such data, referred to as series, to more familiar economic aggregates of the type long compiled by goverment and private research agencies. The argument is adduced that diffusion data are closely related to rates of change in corresponding economic aggregates-for example, that the distribution of reported directions of change in individual-firm sales tends to vary as the rate of change in aggregate sales of the same firms. On the basis of this result, the main burden of the essay is an investigation of the determinants of business plans and anticipations as these are revealed in diffusion data. Somewhat different basic models of the process of expectation-formation are proposed for manufacturers, on the one hand, and for wholesalers and retailers, on the other; and various modifications of these models are explored. Since all these models involve simultaneous economic relations, a fundamental problem of the essay is to investigate the technical complications encountered in the estimation of such relations when expressed in terms of diffusion data. For a combination of methodo-