Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
The Imp act of Technological Innovation: A Histoncal View NATHAN ROSENBERG Perhaps the reason we do so poorly at predicting the impact of technological change is that we are dealing with an extraordinarily complex arid interdependent set of relationships. We should, how- ever, be able to do a somewhat better job of it irz the fixture, if only by developing a better appreciation of some of the reasons why we have done so badly in the past. It is reasonable to assume that readers of this volume believe, at the very least, that technological change plays a major role in shaping our economy and society generally. I had originally intended to devote this chapter to a historical look at technological change, but the more I thought about it, the more intrigued I became by a different but closely related question, a question that ought to be an urgent and persistent concern to precisely such an au- dience. The question is, quite simply: Why do we consistently do such a poor job of anticipating the effects of technological change? Why is our intellectual framework for thinking about the way technology transforms our lives so obviously inadequate? This question, which has occurred to me many times over the years, reasserted itself recently when I encountered a piece of filturology given to President McKinley in 1899 by a Mr. Charles H. Duell. Mr. Duell was the commissioner of the Patent and Trademark Office at the time and, rather uncharacteristically for a public officeholder, was encouraging the President to close down his agency. His reason was quite startling in its simplicity: "Everything that can be invented has been in- vented." I am happy to report, 86 years and approximately 3.8 million patents later, that the President did not heed Mr. Duell's advice. We are all quite properly disdainful of Mr. Duell's total bankruptcy of 17
18 NATHAN ROSENBERG imagination. I resurrect him here, for a brief cameo appearance, only because he managed to express in extreme foIlll, but with epigrammatic precision, a widely held view regarding the future impact of technological change, a view that systematically underrates its future contributions. In retrospect, it is apparent that we have persistently underestimated the contribution of tech- nological change to the growth of the economy. As part of the same bias, we have failed to anticipate the contribution that technological change would make to alleviating or eliminating certain future problems that earlier gen- eraiions regarded as both serious and intractable. It is time that I did a certain amount of intellectual unpacking. In particular, who is the "we" that I have been invoking so far? That in itself turns out to be an interesting question. Much of what I want to say will focus on the question of who has been responsible for shaping the most influential view of the future and the role played by technological forces in that future. In examining the views of various professional groups with respect to how accurately they anticipated certain aspects of technological change, it turns out that no group covers itself with glory. What is intriguing is that they all come off badly, although the reasons for the poor performance vary consid- erably. For many decades, as far back as the writing of Malthus and Ricardo at the beginning of the nineteenth century, it almost seemed that economists had a stranglehold on the expression of deeply pessimistic views of the future. Malthus, in particular, made clear in the very title of the first edition of his essay, first published in 1798, his intention of rejecting naive Enlightenment views on the future prospects for improvement in the human condition. The full title was Essay on the Principle of Population as it affects the future improvement of Society, with remarks on the Speculations of Mr. Godwin, M. Condorcet and Others. More recently, however, there has been an in- terestin=, reversal of roles. Very pessimistic forecasts have emanated from other circles, in particular from systems analysts, biologists, ecologists, and other natural scientists who have become concerned with a set of social issues beyond their more narrowly defined professional spheres. Economists, by contrast, have been cast in the role of explaining why these pessimistic forecastsin some cases, prophecies of doom were unwarranted. In fact, I cannot resist the opportunity of pointing out how well economists have come off in the past 15 years or so. I know economists are not accus- tomed to being praised for Me accuracy of their predictions, but that is because the public is so obsessed with predictions about the performance of next year's GNP, or what interest rates will look like in 6 months' time, that it pays little attention to some other, fundamental aspects of economic reasoning and the predictions that they generate. During the 1970s, public discussion was preoccupied with visions of im- minent natural resource exhaustion (as well as pollution-induced ecological
THE IMPA~ OF TECHNOLOGICAL INNOVATION: A HI=ORIC~ VIEW 19 disasters) that were bound to bring economic growth to an end. The dominant theme, readers will recall, was The Limits to Growth, the title of a well- known book published in 1972. The argument was that there were inexorable natural resource constraints that placed a rigid upper limit on economic growth possibilities. Simple extrapolation of recent rates of utilization of key natural resources was thought sufficient by some futurologists to generate fairly precise predictions of an apocalypse awaiting mankind not too far down the roadperhaps around the middle of the twenty-f~rst century. The whole exercise was remarkably Malthusian. Indeed, it often seemed to me exactly as if Malthus had resumed to earth in the 1970s in the guise of a slightly off-the-rails computer programmer. These purely intellectual preoccupations were powerfully reinforced in late 1973 and 1974 by the Arab oil embargo and by the first of two drastic increases in world petroleum prices orchestrated by the Organization of Pe- troleum Exporting Countries (OPEC). The sudden sharp rise in energy prices was widely interpreted as powerful confirmation of the "limits to :,~owth" view of the future. But, in this period, economists were no longer the leading spokesmen for an essentially Malthusian view of mankind's dreary future prospects. Rather, economists, far more than other professionals, continually called attention to a number of adjustment mechanisms that market forces could reliably be expected to generate. These mechanisms, they argued, would, at the very least, drastically mitigate the dreariest aspects of the pessimistic predictions that were so rampant at the time. When a specific resource becomes increasingly scarce, they pointed out, one may reasonably expect a number of adjustment mechanisms, generated by its higher price, to come into play" substitution, conservation, and more intensive explora- tion. Less emphasis was placed on technological change as such an adjustment mechanism, although the historically minded readily pointed to earlier his- torical experiences as suggesting that new technologies were a highly prob- able, and highly powerful, response to rising energy prices. The surprising fact, however, is that the main tradition in economics has never paid extensive attention to technological change. Classical economics, the economics of Malthus and Ricardo, was very much concerned with the long-tenn prospects for economic growth. Nevertheless, classical economists devoted their main energies to demonstrating the limitations on such growth prospects that were imposed by the niggardliness of nature and the inevitably diminishing returns to capital and labor when the supply of land is fixed. Technological innovation might indeed offset such diminishing returns, but one can only record that classical economists simply did not seem to attach much weight to that possibility. Rather, they placed far greater emphasis on the potential benefits of a policy of free trade between more advanced, densely populated countries and countries where the man:land ratio was more fa- vorable. Although Malthus and Ricardo were writing in the midst of one of
20 NATHAN ROSENBERG the genuine discontinuities of human history the Industrial Revolution- they could only understand progress in the context of painfully slow change in an agrarian society. It would have taken a major flight of imagination to escape from the preconceptions imposed on them by their knowledge of past history. ~ Although Marx, who may be considered the last of the great classical economists, did recognize the enormous potential for technolo;,ical change under capitalism, that potential was, in his view, increasingly frustrated by the internal contradictions of capitalism. For Marx, the inherent laws of motion of capitalism, as he called them, are such that the labor-savin:, bias of technological chan Be has the primary effect of generating increasing un- employment. The notion that technological change may generate unemploy- ment of a kind that is not quickly corrected by market forces is traceable back to Ricardo, but it was Marx who brought the unemployment-creatin=, aspects to center stage. Thus, ironically, the one major economist of the nineteenth century who fully foresaw the great product~vity-increasing po- tential of technological change also argued that such change occurred too rapidly. For, under capitalist institutions, a rapid rate of labor-savin, tech- nological change created increases in productivity that capitalism was in- capable of absorbing. The eventual result, for Marx, was that rapid technological change under capitalist institutions would bring about a collapse in the in- stitutional system itself. The neoclassical tradition in economics, beginning in the late nineteenth century, turned away from the classical concern with long-term economic grown prospects and concentrated instead on examining the implications of maximizing behavior in a static framework. A main concern, which has dominated this tradition up to the present day, is to analyze how a market economy generates forces bringing about a return to equilibrium after some force has disturbed that equilibrium. Considerable attention has been devoted to analyzing the conditions determining the stability and the efficiency of the equilibrium state to which the economic system gravitated. However, neoclassical economics consists largely of a comparison of successive equi- libnum states and does not incorporate an analysis of the adjustment process per se. Technological change, when it is considered at all, is usually treated as some exogenous, once-and-for-all, cost-reducing process innovation to which the economy subsequently adjusts. Or, within the firm, the decision maker might be assumed to be confronted with a range of exogenously determined technologies among which he has to choose In fact, He act of choice itself is a very time-intensive and resource-intensive process of search, wherein the alternatives being explored are not well defined and have a number of uncertainties attached to them. Much of what we label "R&D" consists of expenditures on precisely this search process. Thus, the main traditions of classical and neoclassical economics have
THE IMPACT OF TECHNOLOGICAL INNOVATION: A HISTORICAL VIEW 21 devoted very little attention to the analysis of technological change. Although classical economists were centrally concerned with the determinants of long- tenn economic growth and might therefore have been expected to devote considerable attention to technological change, they assumed that the contours of long-term growth would be dominated by natural resource constraints and diminishing returns resulting from the pressures exerted by a growing pop- ulation against those constraints. Their conclusions were that diminishing returns would swamp the growth in productivity resulting from technological change. Neoclassical economists, on the other hand, focused on short-run problems of optimal resource allocation within a static framework, from which technological change had usually been explicitly excluded. When the question of economic growth was considered by neoclassical economists, it was regarded primarily as involving a rise in the capita!: labor ratio. The great strength of neoclassical reasoning is that it powerfully focuses attention on a wide variety of adjustment mechanisms that permeate all economically motivated behavior. The economic world is indeed full of potential substitution possibilities and, in my view, it is the glory of neo- classical economics that it has provided a profound understanding of such mechanisms and their implications for optimal resource allocation. Thus, shifts in important economic parametersa sudden rise in petroleum pnces- give rise to a range of possible adjustments, and neoclassical economics offers a systematic analysis of the directions in which rational economic agents may be expected to move in response to such shifts. On the other hand, it is fair to say that economists since the time of Malthus, classical and neoclassical alike, have paid insufficient attention to what has been, in the long run, the most powerful single mechanism of adjustment, technological change. Although in recent years much attention has been devoted to the role of technological change as a determinant of productivity improvement (largely as a result of the work of Ab~novitz and Solow in the mid-19SOs, which highlighted some of the deficiencies of the neoclassical approach to growth, economists, like visually everyone else, continue to reason as if the supply of natural resources is permanently fixed. At least, they do not typically address this issue explicitly. Within an ana- lytical framework, technological change is regarded as generating a greater output from a given input of resources. Although this is a perfectly acceptable procedure for many short-run purposes, it is extremely misleading for longer- term issues when it gives rise to the assumptions that natural resource inputs can be unambiguously defined, independently of the state of technical knowl- edge, and that the quantity of such inputs is fixed in amount. Although it is obviously true that nature imposes certain constraints on resource supplies, it is also true, and of fundamental importance, that many technological improvements, when they occur, have the effect of vastly enlarging the resource base, that is, they constitute materials-au~nenting -
22 NATHAN ROSENBERG technological change. By making it possible to exploit resources that could not be exploited before, technological change is making additions to the resource base of the economy, in economic ted Acts if not in geological terms. Thus, it should not be regarded as paradoxical, although many will think it is, to state that the United States has a far larger quantity of iron ore deposits within its borders today than it did 50 years ago. In the past few decades, new processing techniques that prepare the iron ore for the blast fi~rnace- pelletization and benef~ciation have made it possible to exploit the immense deposits of hard, low-grade taconite ores. Such low-grade ores were ignored as long as the high-grade iron ores of the Mesabi Range were still abundant. The development of these new processing techniques has been fully equiv- alent to a gigantic expansion of resource supplies. In fact, pelletization and benef~ciation have brought such great economues in transport costs and blast- furnace efficiency that the energy cost of a finished ton of steel has declined substantially even Tough the iron content of the taconite ores is very low.2 Similar developments had occurred in the nineteenth century. The intro- duction of the Gilchrist-Thomas basic steel-making process in the late 1870s changed the course of European history by making possible, for the first time, the exploitation of the enormous high-phosphorus iron ore deposits of Western Europe. The "low quality" deposits were simply not usable with the earlier iron-making technology. The release of energy from the atom during World War lI meant a vast expansion of energy supplies, although obviously there had been no changes In the natural environment or in the physical characteristics of uranium. The invention of the internal-combustion engine toward the end of the nineteenth century converted petroleum deposits into an energy source for the first time. Until Mat engine was developed, it will be recalled, petroleum seated pn- marily as an illuminant. As recently as the 1930s, natural gas was Sull regarded as an unavoidable and dangerous nuisance that needed to be safely disposed of. Unless there happened to be some urban markets nearby, it was typically treated as a waste material and flared, as it still Is in some parts of the world. Eventually, the perfection of a technique for producing high-pressure pipelines trans- formed natural gas from a waste product into our most attractive household fuel a fuel that currently plays a major role in many industrial markets and constitutes a large fraction of total energy supplies. Thus, the point has been systematically ignored, or systematically under- appreciated since the time of Malthus, that natural resources possess eco- nomic significance only as a function of technological knowledge, and that increases in such knowledge are fully equivalent to an expansion of the resource base of the economy. The best that can be said for the widespread intellectual parlor game of calculating how long it will take to exhaust the supply of a particular strategic raw material, at recent or current rates of
THE IMPACT OF TECHNOLOGIC~ INNOVATION: A HI=OMC~ VIEW 23 utilization, is that the long division is usually carried out correctly. Such calculations are of very limited relevance to a technologically dynamic econ- omy, however. Technological innovation has been a method for overcoming specific natural resource scarcities by vastly expanding the number and the quality of resources that are capable of being economically exploited. In this sense, technological innovation has been the most efficient of all adjustment mechanisms for dealing with growing natural resource scarcity. THE LIMITATIONS OF THE EXPERT I turn now to another category of explanation in accounting for the dif- f~culties in dealing with the future impact of technological change. Much of the difficulty is attributable to the high degree of specialization of techno- logical knowledge that characterizes modern industrial societies. Economists have been at the forefront, at least since the time of Adam Smith, of those who have emphasized the gains resulting from increasing specialization and division of labor. Such gains, from Adam Smith's eighteenth-century pin factory to the research activities of a modern university (Stanford Linear Accelerator Center), have been immense. But there is another side to the specialization coin because, while there are indeed great benefits to spe- cialization, there are also drawbacks. Experts of any kind tend to look at the world in terms of a very limited number of variables indeed, that is a reasonable definition of what it means to be an expert. The training and experience of experts equip them to deal with movement along some very particular trajectories, but not others. The old aphorism that an expert is a person who knows more and more about less and less conveys an important truth, one that has serious implications for the understanding of technological change (`'When all that you have is a hammer, everything looks like a nail. "I. A specialist is typically capable of extending and improving the methods of his or her expertness and applying them to new uses. However, the very nature of an expert's education and professional experience is likely to dis- qualify that person from developing new technologies based on different principles or even from appreciating their potential significance. The industrial history of the past century is replete with evidence for these assertions. Carriage makers played a negligible role in the development of the automobile (although Fisher Body did make the transition), and the makers of stagecoaches played no role in the development of the steam locomotive. Ike makers of steam locomotives, in turn, made no contribution to and showed no interest in the new technology that displaced their invention, the diesel locomotive. It is hardly surprising. No amount of expertise in the operation and improvement of steam locomotives would equip an engineer with the capabilities required to develop an engine based on such different principles. Similarly, many of the manufacturers of piston-driven aircraft
24 NATHAN ROSENBERG engines could not negotiate their way into the jet age. The makers of vacuum tubes failed to transfer their dominance to the semiconductor market. Nor were nylon or rayon introduced by experts who knew a great deal about silkworms and mulberry leaves. Western Union showed no initial interest in that newfangled device, the telephone. In fact, Western Union turned down the opportunity to purchase Alexander Graham Bell's telephone patent when it was offered to them for a mere $100,000! My point is that there are discontinuities as well as continuities in the course of technological change. The failure of industrial firms to make the transition in these episodes of discontinuity is not due to some inherent failing or unavoidable human conservatism. Rather, it reflects the limitations of technical expertise. Whereas experts in an existing field are obviously in- dispensable for generating improvements that draw on their accumulated technical skills, those very skills may become bamers during periods of discontinuity. At such points those skills are no longer relevant for a tech- nology based on different skills or methodologies. Although the existing set of technical skills in an industry may be of no use during the transition to a drastically different technological base for that industry, technical skills in other industries may be very useful. For example, although the transition from propeller-dnven aircraft engines to jet engines represented a genuine discontinuity for makers of propeller-driven engines, it represented much less of a discontinuity for manufacturers of steam tur- bines. Manufacturers of steam turbines already possessed designing and man- ufacturing skills Mat gave them a great comparative advantage in the exploi- tation of the new aircraft power plant. Thus, it is not surpnsing that General Electric, Amenca's largest manufacturer of steam turbines, entered the busi- ness of making aircraft engines when jet propulsion was introduced. When drastically new technologies make their appearance, thinking about their eventual impact is severely handicapped by the tendency to think about them in terms of the old technology. It is difficult even to visualize the complete displacement of an old, long-dominant technology, let alone ap- prehend a new technology as an entire system. Thus, time and again, new technologies have been thought of as mere supplements to offset certain inherent limitations of the old. In the early years, railroads were thought of merely as feeders into He existing canal system, to be constructed In places where the terrain had rendered canals inherently impractical. In precisely the same fashion, the radio was thought by its originators and proponents to have potential applications mainly where wire communication was impractical, for example, ships at sea and remote mountain locations. (The old teen "wireless," still employed in Britain, effectively perpetuates that early per- spective.) The extent to which the old continues to dominate Winking about the new is nicely encapsulated in Thomas Edison's practice of regularly referring to his incandescent lamp as `'~e burner." Rather more seriously,
THE IMPACT OF TECHNOLOGICAL INNOVATION: A HISTORICAL VIEW 25 in his work on an electric meter, a biographer reports, Edison for a long time attempted to develop a measure of electricity consumption in units of cubic feet!3 In the case of aircraft engines, the time intervals between overhaul of jet engines were originally based on the earlier practices with piston engines. As a result, a major economic benefit of jet engines their much greater reliability and durability and, consequently, lower maintenance re- quirementswas nowhere near fully exploited. It was only after some years that airlines extended the time between overhaul of jet engines to intervals that reflected the performance characteristics of the new power plant. (The time between overhaul of piston engines was 2,000 to 2,500 hours of service. The time interval was extended to as long as 8,000 hours for jet engines.) But if thinking about the future impact of new technologies is handicapped by the force of conceptualizations based on the old, that form of thinking receives substantial reinforcement from the opposite direction: inventions typically enter the world in very primitive form compared with the shape that they eventually acquire. Thus, a basic reason for underestimating the impact of a new technology is that new technologies often appear distinctly unpromising at the outset. Their dominating characteristics are often high cost and poor performance standards, including an infuriating degree of unreliability ("Get a horse!". The difficulty here seems to be in predicting the trajectory of improvements that will occur in die course of the life cycle of the new product. A disinterested observer who happened to be passing by at Kitty Hawk on that fateful day in 1903 might surely be excused if he did not walk away with visions of 747s or C-SAs in his head. However, I think there is a deeper issue at stake here. Although, as ~ argued earlier, existing technical expertise is not very useful in an encounter with genuine technical discontinuities, it is rather different when technical continuities are involved. I believe that technical experts are reasonably good at anticipating the kinds of performance improvements that can be teased out of a given technology, once it has been established and its working principles are reasonably well understood. Technical specialists usually have a good appreciation of likely improvement trajectories. Their work is guided by an informed sense of probable directions and rates of future improvement. Why, then, the poor performance in dealing win the future impact of new technologies? I believe the answer to this question takes us back to the central concern of this volume. The impact of new or improved technologies is not just a matter of improved technical performance. It is, rather, a matter of translating such information into its potential economic and social significance. Doing this requires something much more than purely technical expertise. It is, in fact, an extraordinarily difficult exercise. Understanding the technical basis for wireless communication, which Marconi did, was a very different matter from understanding the possibilities for a vast new entertainment broadcasting
26 NATHAN ROSENBERG industry that would reach into every household (and automobile), which Marconi completely failed to envisage. The point is that social change or economic impact is not something that can be extrapolated out of a piece of hardware. New technologies are, rawer, building blocks. Their eventual impact will depend on what is subsequently designed and constructed with them. New technologies are unrealized potentials that may take a very large number of eventual shapes. What shapes they actually take will depend on a wide range of social priorities and values, on the way the demand for particular goods and services changes in response to rising incomes or declining pnces. I will return to this issue shortly. An additional reason why it is so difficult to anticipate He effects of technological change is that most inventions have the* origin in the attempt to solve very specific, even narrowly defined problems. It is very common, however, that a solution, once found, has important applications in totally unintended contexts. In this sense, the eventual impact of new technologies is very difficult to anticipate, because much of the impact is realized through the intersectoral flows of technology that are so characteristic of modern industrial economies. Inventions increasingly have serendipitous life histo- nes. The steam engine, for example, was invented in the eighteenth century specifically as a device to pump water out of flooded mines. It was, for a long time, regarded exclusively as a pump. A succession of improvements later rendered it a feasible source of power for textile factories, iron mills, and an expanding array of industrial establishments. In the course of the early nineteenth century the steam engine became a generalizable source of power and had major applications in transportation railroads, steamships, and steamboats. In fact, in the United States before the Civil War, the main use of die steam engine was in transportation Later in the nineteenth century the steam engine was used to produce a new and even more generalizable source of power electricity which, in turn, satisfied innumerable final uses to which steam power itself was not efficiently applicable. Finally, the steam turbine displaced the steam engine in the generation of electric power, and the special features of electricity its ease of transmission over long distances, the capacity for making power available in "fractionalized'' units, and the far greater flexibility of electacity-powered equipment- spelled the eventual demise of the steam engine Thus, the life history of the steam engine was shaped by forces that could hardly have been foreseen by inventors who were working on ways of re- moving water from increasingly flooded coal mines. Its subsequent history was shaped by unanticipated applications to industry and transportation and eventually by the systematic exploitation of new technologies that were 1ln- dreamed of at the time the steam engine itself was invented, such as elec-
THE IMPACT OF TECHNOLOGICAL INNOVATION: A HlSIORlCAL VIEW 27 eicity. Nevertheless, the very existence of We steam engine, once its operating principles had been thoroughly understood, served as a powerful stimulus to other inventions. Thus, major innovations, such as the steam engine, once they are estab- lished, have the effect of inducing furler innovation and complementary investments over a wide frontier. Indeed, Me ability to induce such furler innovations and investments Is a reasonably good definition of what consti- tutes a major innovation. It is a useful way of distinguishing between tech- nological advances that are merely invested with great novelty from advances that have the potential for a major economic impact. But this also highlights the difficulties in foreseeing, the eventual impact, since Mat will depend on the size and the direction of these future complementary innovations and investments. In the twentieth century there is an additional, increasingly significant relationship that complicates the ability to foresee the eventual impact of technological change. This relationship is partially obscured by the prevailing linear model, which looks on innovation as originating in '`blue-sky" basic research, which feeds downstream to applied research and, eventually, to new product development. In fact, to an increasing degree, it is the needs of the technological realm that direct scientific research. Increasingly, the needs of an expanding technological system shape and mobilize scientific research in specific directions. This is what the term "mission-oriented basic research" is all about. More to the point, this is what one of the most important institutional innovations of the twentieth century is all about: the industrial research laboratory. These labs have been specifically established to facilitate the exploitation of scientific knowledge for industrial purposes. But, to an increasing degree, the best of these labs generate much of the scientific knowledge that they exploit. At the same time, the problems en- countered by sophisticated industrial technologies, and the anomalous ob- servations or unexpected difficulties they produced, have served as powerful stimuli to scientific research, in the academic community as well as the industrial research lab. In these ways the responsiveness of scientific research to economic needs and opportunities has been powerfully reinforced. Thus, solid-state physics, presently the largest subdiscipline of physics, attracted only a few physicists before the advent of the transistor. In fact, the subject was not even taught at most universities. The training in solid- state physics that Shockley received at Me Massachusetts Institute of Tech- nology in the 1930s was probably unavailable at any other university in America at the time, with Me exception of Princeton. The situation was transformed, of course, by the invention of the transistor in 1948. The tran- sistor demonstrated the potentially high payoff of solid-state research and led to a huge concentration of resources in that field. It is important to note that the rapid mobilization of resources in solid-state research after 1948
28 NATHAN ROSENBERG occurred in the university as well as in private industry. Thus, transistor technology was not building on a vast earlier research commitment. Rather, it was the initial breakthrough of the transistor that gave rise to a subsequent large-scale commitment of scientific resources. Similarly, the advent of the laser as a potentially important new mode of transmission has served as a powerful focusing device In shaping the direction of scientific research. However, this scientific research has gen- erated a vast array of unanticipated applications, including optic surgery, precision measurement, navigational instruments, military applications in outer space, and the shaping of a wide assortment of materials in manufac- turing, ranging from clothing to aircraft composites. At the same time, a high-payoff application of laser technology was clearly anticipated and suc- cessfully consummated. It was the development of laser technology that suggested the feasibility of using optical fibers for transmission purposes. This possibility, in tum, pointed to the field of optics, where advances in scientific knowledge can now be expected to have high potential payoffs. As a result, optics as a field of scientific research has experienced ~ great resurgence in recent years. It has been converted by changed expectations, based on past and prospective technological innovations, from a relatively quiet intellectual backwater to a burgeoning field of scientific research. It is likely that this scientific activity, in turn, will yield a new array of unanti- cipated applications. The research system within modern industry Bus affects technological predictability in two offsetting ways. On the one hand, certain high-payoff applications (as in He case of He laser) can be realized more rapidly and predictably through the application of scientific research to technological breakthroughs. On He other hand, these very same scientific research aciiv- ities have ~emseives generated a large number of unanticipated applications. The overall impact of laser and fiber optics technologies is highly uncertain, even as He realization of certain applications appears to have become more predictable. TECHNOLOGICAL CHANGE AND UNEMPLOYMENT I turn now to a final category of the dialogue over the impact of tech- nological change. It involves a concern Hat has been particularly prominent in He past few years, but that has deep intellectual roots going back to Mam and even to Ricardo. That is He concern Hat He primary impact of tech- nological change will be increased levels of unemployment. There was and there remains a widespread tendency to attribute the higher unemployment levels that emerged during the 1970s to the in~oduchon of new technologies, especially electronic technologies, Hat purportedly had a strong labor-saving bias. Moreover, Here is widespread apprehension Hat we are now poised at
THE IMPACT OF TECHNOLOGICAL INlIOVATION: A HISTORICAL VIEW 29 one of those great discontinuities in history and that new technologies are going to have unprecedentedly large effects in generating a permanent pool of unemployed. This is because a number of new technologies robotics, computer-aided design/computer-aided manufacture (CADICAM), the grow- ing capacity of the electronic chip, automation are expected to have strong labor-saving effects. It is, of course, always impossible to prove anything about the future by looking at the past. It is impossible to prove that we are not poised at some genuine discontinuity in history. Moreover, although Mere is some evidence of recent improvement, it is paisley clear that the American economy has, in some important respects, been performing poorly for more than a decade. Prodllctivity improvement has been particularly dismal, and the "natural" rate of unemployment seems to have been increasing during He 1970s and l980s. It is far from clear, however, that high unemployment has been primarily due to the character of technological change, nor are there compelling reasons to believe that new technologies will have an unusually job-reducing bias in the future. Some categories of employment will, of course, suffer. Tech- nological change has always reduced specific categories of employment- e.g., farm workers, railroad workers, coal miners, lumberjacks. The electric light bulb displaced the candle maker, and the automobile put saddlers and whip makers out of business. The crucial question is whether the thrust of technological change is to reduce total employment, not whether it eliminates specific jobs. Three points are worth making in this regard. First, a simple empirical observation. Although unemployment levels in He American economy were indeed high, by h~stoncal standards, during He 1970s, that decade can hardly provide persuasive evidence that new tech- nologies have been reducing aggregate employment opportunities. In fact, during Hat decade the number of employed persons rose by a remarkable 20 million from 80 million in mid-1970 to 100 million in mid-1980. What- ever job-reducing forces may have been at work within the nature of He new technologies were, at the very least, swamped by mechanisms working in the opposite direction. Second, labor-saving innovations are not the same as job-reducing inno- vations. The reductions in cost and price associated win labor-saving in- novations may bring in their wake vast increases in specific kinds of employment, and in fact have often done so. When Henry Ford introduced the progressive assembly line into the American automobile industry in 1914, the result was a huge reduction in the number of labor hours required to produce a car. But the resulting ability to sell a Model T Ford for only $400 was a revolutionary event that resulted in an immense increase in employment in the automobile industry. The demand for cars turned out to be highly elastic. On the other hand, when demand is inelastic, labor-saving innovations reduce He demand for labor in that sector but shift demand elsewhere. The
30 NATHAN ROSENBERG final employment impact of technological change cannot be confined to the sector (nor to the country) where it occurredit is a problem in general equilibrium analysis, not partial equilibrium analysis.4 Third, as a more general matter, it seems to be much easier to anticipate the employment-displacing effects of technological change than the employ- ment-expanding ones. Partly this is because we do not have a good technique for dealing with the impact of product innovation. The anticipation of the employment-expand~ng consequences of innovations seems to require a much greater exercise of the social imagination, an ability to foresee uses in entirely new social contexts. As I suggested earlier, no one seems to have correctly anticipated Me immense commercial, educational, and entertainment uses to which the radio would be put in the twentieth century. In the 1950s, when the computer was still in its infancy, it was authoritatively predicted that all of America's future needs would be adequately catered to by fewer than a dozen computers. Even earlier, Thomas J. Watson, Sr., president of IBM and perhaps the most experienced person in the business, believed that a single computer (the Selective Sequence Electronic Calculator, built in 1947 and in operation at IBM's New York headquarters) "could solve all Me important scientific problems in the world involving scientific calculations." He was reported to believe that computers had no commercial possibilibes.5 Even Thomas Edison, a true inventive genius, is said by one of his biog- raphers to have anticipated that the phonograph would be used primarily to record the deathbed wishes of elderly gentlemen! The point is that it is extremely difficult to anticipate the impact of new innovations, because that impact is not obvious from the hardware itself. It depends, rawer, on social uses and cultural contexts, on how society chooses to mobilize and to exploit the potential of a piece of hardware. No one seems to have anticipated the astonishing amount of information processing that would take place in our society when the productivity of the calculating technology was increased by a couple orders of magnitude. Thus, there appears to be a systematic bias in perceptions about the future. This bias sharpens the awareness of possible job-reducing consequences of technological change but at the same tune fails to identify the prospects for enlarged employment opporh~ruties that flow from the ability to produce certain products more cheaply or to invent entirely new products with quite unanticipated uses and applications. A distinctive feature of western capi- talism seems to have been the ability to produce very cheap variants of products that, in an earlier age, were consumed only by a small elite nylon stockings for sink ones, ballpoint pens for Parker 51s, recorded stereophonic music for court musicians. In fact, we are still insufficiently aware of the extent to which sustained high rates of aggregate economic grown have depended on He continual introduction of new products to offset He retar- dation resulting from He slower rates of grown of older industries.
THE IMPACT OF TECHNOLOGICAL INNOVATION: A HISI ORICAL VIEW 31 Equally important, discussions of the future impact of technological change that emphasize its net unemployment-generating effects have systematically ignored what has been perhaps the single most conspicuous feature of struc- tural change in the American economy for several decades: the expansion of the service sector. That sector is now far and away the largest sector of the economy, employing more workers than the entire commodity-producing sector. In 1982, service occupations accounted for 74 percent of all em- ployment. Indeed, the growth in employment in the United States since World War lI has been overwhelmingly a growth in the service sector 86 percent of all employment growth has occurred in the service-producing sector. (There are more musicians in the American labor force than coal miners, and several times as many real estate agents.) Although certain aspects of the service sector have received a great deal of attention, e.g., the apparently much slower growth of productivity, far less attention has been given to the connections between technological change and service employment. The story is a complicated one; indeed, a large part of the problem is that it is a great many stories, since the service sector is a huge portmanteau comprising many very different kinds of activities. Moreover, the growth of that sector is not just a matter of technological change. Rather, it is a matter of tech- nological change interacting in subtle ways with changes in the composition of consumers' expenditures as their incomes rise, and with changes associ- ated, not only with the cheapening of certain services, but with highly sig- nificant changes of a qualitative nature as well. There is no quick and easy way to summarize the changes that have occurred in the service sector in the past several decades. However, if we are to come to grips with the impact of technological change, we need to examine the very diverse experiences in the delivery of health care, education, recreation, retailing, insurance, finance, and government, bow at federal and state and local levels. ~ only want to suggest to you, for the moment, the total arbi~anness of assuming that the outcome of these experiences is likely to be declining employment opportunities in the future. CONCLUSION Perhaps the end result of my discussion has been simply to persuade you that the reason we do so poorly at predicting the impact of technological change is that we are dealing win an extraordinarily complex and interde- pendent set of relationships. ~ would certainly not want to resist that con- clusion. ~ would, however, want to insist that we should be able to do a somewhat better job of it in the future, if only by developing a better ap- preciation of some of We reasons why we have done so badly in the past. hope this discussion has pointed in some helpful directions.
32 NATHAN ROSENBERG NOTES 1. Although Malthus certainly won the debate at the time, some of his opponents took positions that, in retrospect, seem remarkably farsighted. For example, William Godwin, in his book, Of Population, made the following fascinating observation in 1820: "Of all the sciences, natural or mechanical, which within the last half century have proceeded with such gigantic strides, chemistry is that which has advanced the most rapidly. All the substances that nature presents, all that proceeds from earth or air, is analyzed by us into its original elements. Thus we have discovered, or may discover, precisely what it is that nourishes the human body. And it is surely no great stretch of the faculty of anticipation, to say, that whatever man can decompose, man will be able to compound. The food that nourishes us, is composed of certain elements; and wherever these elements can be found, human art will hereafter discover the power of reducing them into a state capable of affording corporeal sustenance. No good reason can be assigned, why that which produces animal nounshment, must have previously passed through a process of animal or vegetable life. And, if a certain infusion of attractive exterior qualities is held necessary to allure us to our food, there is no reason to suppose that the most agreeable colours and scents and flavours may not be imparted to it, at a very small expense of vegetable substance. Thus it appears that, wherever Earth, and water, and the other original chemical substances may be found, there human art may hereafter produce nourishment: and thus we are presented with a real infinite series of increase of Be means of subsistence, to match Mr. Malthus's geometrical ratio for the multiplication of mankind." [William Godwin, Of Population (London: Longman, Hurst and Company, 1820), pp. 499-501.] Peter Kakela, Iron ore: Energy, labor and capital changes with technology, Science, Decem- ber 15, 1978. The electric meter, it should be pointed out, was an extremely important complementary in- vention within the emerging electric power system. Before a satisfactory meter was developed, around 1900, meters were likely to be both very expensive and highly unreliable. As a result, flat-rate contracts were common, and consumers had no incentive to economize on the use of electricity. Moreover, in the absence of a meter, electric utilities had to undertake excessively large investments in generating and transmitting equipment. 4. It may be added that, in the present international context, new steps forward in automation may increase U.S. employment by repatriating activities that have moved offshore. Thus, the au- tomation of a variety of labor-intensive assembly-line work may well bong back to the United States jobs that have recently migrated overseas in search of cheaper labor. Currently, in Silicon Valley, a number of industrial sectors are confronting the choice between robotics or overseas assembly. 5. Barbara G. Katz and Almarin Phillips, We computer industry, in Richard Nelson ea., Gov- ernment and Technical Progress (New Yorlc: Pergamon Press, 1982), p. 171. See also William F. Sharpe, The Economics of Computers (New York: Columbia University Press, 1969), p. 185.
