The discussion thus far has focused on R&D activity at a single mo ment in ti me. It is possible to explain continuing technological progress over time within a simple search model of the sort dis cussed above, under the assumption that tomorrow’ s ro und of pro j ects is independent of what happens today except for the fact that what is achieved today imposes a higher standard of success for tomorrow’s efforts. Under plausible assumptions, in the absence of increased knowledge there would then be diminishing returns to R&D. These can be offset, however, by growing demand for product. Thus, R&D may grow and technical advance be sustained. This char acterization may capture what is going on in some technologies, but it fails to explain the cumulative nature of technological advance in many sectors. In many technological histories the new is not just better than the old; in some sense the new evolves out of the old. One explanation for this is that the output of today’ s searches is not merely a new technology, but also enhances knowledge and forms the basis of new building blocks to be used tomorrow.
For example, today’s research and development may be searching for a new technology (say, a chemical compound) with certain eco nomic attributes and may be focusing on alternatives with certain technological attributes: the blue technologies. Tests prove that a particular blue- striped technology indeed does have desirable eco nomic attributes. As a result, the hypothesis that the set of blue tech nologies contains some good ones is reinforced, and th e subhy pothesis might be formulated that the subclass of blue- striped tech nologies is particularly attractive.
Or consider the set of technological possibilities defined by various mixes of ingredients. On the basis of past experience, the de cision maker believes that a significant increase in the amount of one of these relative to others will yield an economically superior prod uct. But he is not sure, and does not know how much more to add . A sensible R&D strategy might involve first testing the economic attri butes of a mix somewhat richer than the prevailing one, and, if the results are favorable, trying out an even richer mix, and so on-in ef fect, hunting for the top of the hilL In general, a good strategy will stop the R&D project somewhere short of the top of the hill because the economic attributes achieved are good enough and because the gains from varying the mix in one way or another are not expected to be worth the cost of performing another test. But the knowledge ac quired in the course of the project may have implications for the next round of R&D projects, perhaps involving a different ingredient. 7
In many hardware- producing industries such as the aircraft in dustry, R&D may be represented as a gradual filling- in of the details of an overall rough design idea, with the course of the design work being guided by a series of studies and tests. In the later stages these involve prototype versions of the actual new hardware. In endeavors of this sort the metaphor of “alternatives out there waiting to be found” is somewhat forced . Researchers are building a technological variant that was not in existence before and are finding out how it works. Information is being acquired not only in activities that are incidental to discovery but also in the course of creating and learning about something new. In general, the new design will i nvolve a large number of subdesign. elements or components . Regarding each of these there may be certain “design problems” to solve, in the sense that certain performance goals need to be achieved. Knowledge can facilitate the problem solving by guiding the effort toward promising design alternatives . And knowledge can facilitate overall design by indicating what problems may be hard or easy and by guiding strat egy toward configurations that do not require that the former kind be solved . As was the case with the search for a better chemical mix, a successful development project creates more than a discrete p ractical invention. Today’s new hardware represents a set of solutions to de sign problems and provides a new starting point for the next round of research and development efforts .
. In all of these examples, the result of today’s searches is both a successful new technology and a natural starting place for the searches of tomorrow. There is a “neighborhood” concept of a quite natural variety. It makes sense to look for a new d rug “similar to” but possibly better than the one that was discovered yesterday. One can think of varying a few elements in the design of yesterday’S s uc cessful new aircraft, trying to solve problems that still exist in the de sign or that were evaded through compromise.
This formulation appears to explain relatively satisfactorily certain aspects of what has become known as the “product cycle.” The his tory of many technologies seems to be characterized by occasional major inventions followed by a wave of minor ones. Part of what is going on is product design evolution. As Miller and Sawers (1968) tell the story, the original Douglas DC-3, the result of the confluence of a number of R&D strands, represented a radically new civil aircraft package: all-metal skin, low wing, streamlining of body and engine configuration, more powerful engines. Over the subsequ ent decade, the basic desi gn was improved in a variety of models, designed by other manufacturers as well as by Douglas. Each successive genera tion of plane was faster, had longer range, and was more comfort able. The original basic design was stretched to achieve additional performance and was differentiated to meet a variety of different de mands and conditions. The DC-4 represented the start of a s eries of four-engine versions. By the mid 1950s the potentialities of this de sign concept appear to have been largely exploited . The advent of the Boeing 707 and Douglas DC-8 represented the start of another tech n ological product cycle within the civil aircraft industry . Enos (1962) reported a similar patt ern in petroleum-refining technology. Again, technical change was marked by the periodic introduction of major new technologies (the batch thermal process in 1931, catalytic cracking in 1936, and so on), followed by a wave of improvements . The flow of subsequent improvements in petroleum refi ning appears to have been even more important than that in aviation. Enos reports that in many cases the first versions of the new technology tended to be only marginally superior to the most recent versions of the older technology and were sometimes not superior at all. The advantages of the new were achieved largely through the wave of improvements that were possible with the new design, compared with the difficulty of finding further major improvements in the old one.
As the product evolves, so do the p rocesses of p roduction. Hirsch (1952), in one of the earliest but still among the most illuminating of the studies of “learning curves,” pointed out three different kinds of mechanisms at work: workers are learning to do their jobs better, management is learning how to organize more effectively, and engi neers are redesigning the product to make the j ob easier and to re place labor where it is possible and economical to do so. Hirsch in his study of machinery and Asher (1956) in his research on aircraft noted that different kinds of costs are affected differently over the learning process. In particular, unit labor costs tend to be reduced dramatically, unit materials costs are reduced to a lesser degree, and unit capital costs may rise. This corresponds closely to events that Enos observed during the design improvement process for petroleum-refining equipment. In addition, the detailed studies of the “learning process” do not treat learning as somehow an inevita ble and uninfluenceable consequence of doing. Rather, learning is viewed more actively, and it is apparent that re sources can be ap plied to learning.
In some cases this hunting for marginal improvements involves exploring in a variety of different directions. But in some cases a few directions seem much more compelling of attention than others. Par ticularly in industries where technological advance is very rapid, ad vance seems to follow advance in a way that appears almost inevita ble. Rosenberg (1969) writes of “technological imperatives” as guiding the evolution of certain technologies: bottlenecks in con nected processes and obvious weak spots in products form clear targets for improvement. In other cases, the directions taken seem “straighter” than is suggested by Rosenberg’s emphasis on the shifting focus of attention. We term these paths “natural trajec tories . “
Natural traj ectories are specific to a particular technology or broadly defined “technological regime.” We use “technological regime” in much the same way as Hayami and Ruttan (1971) use “meta production function. ” Their concept refers to a frontier of achi evable capabilities, d efined in the relevant economic dimen sions, limited by physical, biologicaC and other constraints, given a broadly defined way of doing things. Our concept is more cognitive, relating to technicians’ beliefs about what is feasible or at least worth attempting. For example, in the case discussed by Miller and Sawers (1968) , the advent of the DC-3 in the 1930s defined a particular tech nological regime; metal skin, low wing, piston- powered planes. Engineers had notions regarding the potential of this regime. For more than two decades innovation in aircraft design essentially in volved better exploitation of this potential: improving the engines, enlarging the planes, making them more efficient .
In many cases the promising trajectories and strategies for tech nological advance, within a given regime, are associated with improvements of major components or aspects thereof. In aviation, engineers can work on improving the thrust-weight ratio of engines or on increasing the lift-drag ratio of airframes. General theoretical understanding provides clues as to how to proceed . In jet engine technology, thermodynamic under standing relates the performance of the engine to such variables as temperature and pressure at com bustion. This naturally leads designers to look for engine designs that will enable higher inlet temperatures and higher pressures . In airframe design, theoretical understanding (at a relatively unsophis ticated level) always has indicated that there are advantages to get ting planes to fly at higher altitudes, where air resistance is lower. This leads designers to think of pressurizing the cabin, demanding aircraft engines that will operate effectively at higher altitudes, and so forth.
Often there are complementarities among the various trajectories. Advances in engine power and the streamlining of aircraft are com plementary. The development of seeds that germinate at the same time and grow at the same rate facilitates mechanical harvesting.
While natural trajectories almost invariably have special elements associated with the particular technology in question, in any era there appear to be certain natural trajectories that are common to a wide range of technologies. Two of these have been relatively well identified in the literature: p rogressive exploitation of latent econ omies of scale and increasing mechanization of operations that have been done by hand.
In a wide variety of industries and technologies, the advance of equipment technology involves the exploitation of latent economies of scale. In chemical process industries, in power generation, and in other sectors where equipment of larger capacity will p ermit output expansion without a p roportional increase in capital or other costs, the objectives of cost reduction apparently lead designers to focus on making equipment larger. Hughes (1971) documented the way in which designers of electric power equ ipment have aimed progres sively to push forward the scale frontier. Levin (1974) p rovided a gen-eral theoretical discussion of the phenomenon as well as case studies of the process in the manufacture of sulfuric acid, ethylene, and ammonia and in petroleum refining. Exploitation of scale economies is an important part of the story of the improvement of refining equipment as told by Enos. In the development of aircraft technol ogy, designers long have understood that larger planes could in prin ciple operate with lower costs per seat- mile. Of course, in aviation, as in electric power, the possibilities for exploiting latent economies of scale are limited by the market as well as by engineering. In avia tion, high volumes and long hauls provide the market’s targets of opportunity . And historically these have tended to grow in impor tance over time . Th is has permitted engineers to follow their design instincts. As a rule, each generation of commercial aircraft has tended to be made up of larger vehicles than those in the preceding generation.
Another quite common natural trajectory is toward the mechani zation of processes previously done by hand. This shows up strik ingly in the Hirsch and Asher studies of learning. Mechanization seems to be viewed by designers of equipment as a natural way to re duce costs, increase reliability and precision of production, gain more reliable control over operations, and so on. This point has been stressed by Rosenberg (1972) in his study of nineteenth-century in novation in American industry. That this tendency to mechanize still exists has been suggested by Piore (1968) and documented in consid erable detail by Setzer (1974) in her work on the evolution of produc tion processes at Western Electric. Inventors and research and devel opment engineers , operating under a higher-order objective to look for inventions and design changes that will reduce costs, look for opportunities to mechanize. Engineers, through training and experi ence, apparently acquire heuristics that assist the design of machin ery. For this reason, hunting for opportunities for mechanization, like trying to exploit latent economies of scale, can serve as a useful focus for inventive activity.
David (1974), in a fascinating and important essay, proposed a dif ferent but complementary hypothesis. Whereas the studies above point to “easy invention” in directions that i ncrease the capital labor ratio, David suggested that in the late nineteenth century tech nologies that already were capital-intensive were easier to improve in a “neutral” direction than were technologies that involved a lower degree of capital intensity; at that time there was “a lot of room” for improving mechanized operations, and engineer- designers had some clever ways of moving in that direction.
Exploitation of latent economies of scale and opportunities for fur ther mechanization are important avenues for technological advance in many sectors at the present time, just as they were in the nine-teenth century. Many of the studies cited above are of relatively con temporary examples. However, there is no reason to believe, and many reasons to doubt, that the powerful general trajectories of one era are the powerful ones of the next. For example, it seems apparent that in the twentieth century two widely used natural traj ectories opened up that were not available earlier: first, the exploitation of an understanding of electricity and the resulting creation and improve ment of electrical and later electronic components, and, second, simi lar developments regarding chemical technologies. As with the case of mechanization during the 1800s, these developments had several different effects. For example, a greater understanding of electrical phenomena and growing experience with electrical and electronic equipment led to a substitution of these kinds of components for others . And technologies that had many and important electronic components were better able to benefit from the improvements in these components than were other technologies.
It is apparent that industries differ significantly in the extent to which they can exploit the prevailing general natural trajectories, and that these differences influence the rise and fall of different in dustries and technologies. During the nineteenth century, cotton gained ascendancy over wool in large part because its production processes were easier to mechanize. Quite possibly both Rosenberg trajectories and David trajectories were involved. In the twentieth century, Texas cotton drove out southeastern cotton mainly because the area was amenable to mechanized picking. In the current era, where conside rable skill has developed regarding the design and imp.rovement of synthetic products, synthetic fibers have risen in importance relative to natural ones.
One aspect of natural trajectories, whether specific to a particular technology or more general, whether of the nineteenth century or contemporary, is that underlying the movement along them is a body of knowledge held by the technicians, engineers, and scientists in volved in the relevant inventive activity. The knowledge may be quite specific, such as an understanding of the tactics for hybrid development of seeds or an understanding of the operating charac teristics of jet engines. The knowledge may involve more art and intuition than science; this certainly was so of the knowledge behind the mechanization and scale economies trajectories duri ng the 1800s. But in the middle to late twentieth century, many scholars have been strongly tempted by the hypothesis that underlying the technologies that have experienced the most rapid advance (or built into a key component of these) is a relatively well- articulated scientific knowl edge. This does not mean that the “inventors” are active scientists or that “inventing” exploits knowledge produced by recent science. But the fact that college-educated scientists and engineers now comprise the dominant group doing applied research and development indi cates that, at the very least, scientific literacy is an important back ground factor.
The interpretation given here of product cycles and of trajectories within classes of technology is useful for organizing thinking about certain irregularities in the pace and pattern of technical progress. Consider a set of technological possibilities that consists of a number of quite different classes of technology- say, engines employing dif ferent thermodynamic cycles, or different technologies for the gener ation of electric power. Within any of these classes of technology, technological advance may follow a particular trajecto ry. At any given time all the R&D may be focused on one class of technologies (the blue ones), with no attention being paid to the yellow tech nologies because the structure of knowledge (the ability effectively to explore within that subset) is weak in that area. Along the prevailing trajectory there will be a tendency for returns to fall. Assume, how ever, that knowledge occasionally is created (perhaps from basic re search done at universities) that significantly improves the structure of understanding regarding portions of the set in which knowledge previously had been weak and hence that applied research tended to ignore (stri ped yellow technologies tend to be very effective, dotted y ellow ones ineffective) . Then one would expect that a significant shift would occur in the nature of the R&D that goes on and that old experience and knowledge would become obsolete. The R&D game would become very different, perhaps requiring people of different kinds of backgrounds, different kinds of firms, and so on. And tech nical progress would surge forward as solutions appeared to problems suddenly made relatively easy by the strengthening of the knowledge base-only to slacken again as the new areas of search become, in their turn , relatively well explored.
Source: Nelson Richard R., Winter Sidney G. (1985), An Evolutionary Theory of Economic Change, Belknap Press: An Imprint of Harvard University Press.