The Review of Economics and Statistics Vol. 43 No. 2 1961
Innovation, Diffusion, and Productivity Changes
Abstract
FORECASTS of productivity changes are usually made by extrapolating time series. For individual industries productivity fluctuates widely from decade to decade1 and the extrapolation method is vulnerable. An alternative is to use leading series. In an earlier study it has been shown that a well-defined time lag exists in the cotton textile industry among the estimates of productivity from engineering data, plant data, and industry data.2 Leading best-practice series, unfortunately, are hard to come by. But the results have suggested a third alternative: the forecast of productivity changes in industries by studying the diffusion of more advanced technology among plants.3 The present paper is an attempt to develop and test a framework by which productivity changes may be deduced from cross-section plant data. The cross section provides the initial conditions concerning the technological mix before changes. A simple set of rules on innovation and diffusion, also suggested by the crosssection information, then yields the expected changes of the mix. The cross-section approach has many virtues. It is unconstrained by the existence and quality of historical series, and moreover a suitably designed sample also catches the peculiar characteristics of an industry at a particular time or in a particular region. The possibility of refinement is virtually unlimited. From a theoretical point of view the opportunity afforded for testing the behavior of individual plants is also invaluable. These merits are ranged against some equally conspicuous difficulties, the most important of which are probably the difficulties in introducing time variables and in interpreting the results.4 In this paper a simple method for ordering technologies is suggested. After a tag is attached to each technology indicating its place on the scale running from obsolete to advanced, a rule of technology diffusion is introduced. In the third section the rule is applied to each of the two-digit Census Standard Industrial Classification (SIC) manufacturing industries in New England. The results forecasts of productivity, inputs, and outputs are evaluated in the last section of this paper. Although it would be desirable to use the model to predict the known data of some past year, the information at hand has not permitted such an endeavor without gross assumptions. It will therefore be argued only that the long-range forecasts based on the present model are reasonable and consistent in view of historical and present conditions.
- DOI
- 10.2307/1928668
- Volume
- 43
- Issue
- 2
- Pages
- 175
- Sources
- openalex crossref