Joseph A. Tainter, The Collapse of Complex Societies
Cambridge University Press
Fred Bateman (1969) has investigated changes in labor efficiency in the American dairy industry between 1850 and 1910. There was no major technological breakthrough in this interval, but other changes took place. One major shift was the widespread extension of dairying into the winter months. Another was improvements in feeding. Still a third was the addition of stricter sanitation requirements. All of these added to the labor requirements of dairying, although yields did not increase proportionately. The figures in Table 1 show that between 1850 and 1910 dairy output per unit of labor declined by 17.5 percent.
Gregory Johnson (1982) has shown graphically that as the size of a social group increases, the communication load increases even faster. Information processing increases in response until capacity is reached. After this point, information processing performance deteriorates, so that greater costs are allocated to processing that is less efficient and reliable. At this point information processing hierarchies may be expected to develop (Johnson 1982: 394-5).
Parkinson (1957) indicated bureaucratic self-serving to account for declining marginal returns on investment in hierarchical specialists. However comforting to some, this is far too simplistic an explanation. Bendix (1956) has compiled for private industry, in several nations, data similar to those Parkinson has uncovered in government. He was able to show that a pattern of increasing hierarchical specialization characterizes the private sector as strongly as Parkinson has demonstrated for the public (Fig. 15). Clearly in the private sector, where economic success depends on efficiency, this pattern cannot be attributed to self-serving inefficiency. The reason why complex organizations must allocate ever larger portions of their personnel and other resources to administration is because (as discussed in Chapter 2) increased complexity requires greater quantities of information processing and greater integration of disparate parts.
Although some authors (e.g. Schmookler 1962) believe that technical innovation responds to economic productivity, there are also data suggesting that technical innovation often occurs along a curve of declining marginal productivity. Fig. 17 shows reductions in fuel consumption of steam engines resulting from increases in thermal efficiency, from the early eighteenth to the middle twentieth centuries. In such a field, technical innovation slows down as returns for improvement diminish. For the steam engine, the remaining possibilities of fuel-saving were reduced as thermal efficiency was increased. A doubling of efficiency in this century would save much less fuel, per engine, than would a 10 percent increase in the eighteenth century, and the saving would be much harder to achieve (Wilkinson 1973: 144-5).
Zolotas has argued that the productivity of industrialism for producing social welfare is declining. In partial support of this assertion he points out that while U. S. per capita product increased 75 percent from 1950 to 1977, weekly work hours declined by only 9.5 percent (Zolotas 1981: 92-3).
Why does information processing often show a declining marginal return? Why do investment in education, and in research and development, result in decreasing productivity? The answers in both spheres are similar.
The case of education was touched on earlier. Reiterating in brief: general education, which occurs in the earliest years of life, is of the most lasting, widespread value. It is also attained at the lowest comparative cost. Later, more specialized training is considerably costlier. Its benefits may apply only to narrow segments of the society, while its costs are spread throughout the system. It may institutionalize rigidity where flexibility is called for. What is more, the benefits of specialized training are at least partly attributable to the generalized education will obtain greater marginal returns on its investment than a society dependent on specialized training.
The situation in research and development is similar. As with education, specialized scientific knowledge depends upon prior, general principles. Within a scientific field, early work develops the general parameters of the discipline, the nature of the subject matter, the scope of inquiry, broad research questions, and like scholars (Kuhn 1962, nevertheless there is also substantial derivation from it (Schwartz 1971: 43). Thus, the product of early, generalized work in a scientific field includes all knowledge derived from later, specialized research, and so -- again axiomatically—specialized work can never yield the benefits achieved by earlier, generalized research. It is no coincidence that the most famous practitioners historically in each field tend to be persons who were instrumental in developing the field, and in establishing its basic outline. No Einsteinian physicist, no Darwinian biologist, and no Marxist social scientist will ever achieve the fame and influence of these masters who revolutionized their fields.