The Competitiveness of Nations

in a Global Knowledge-Based Economy

May 2003

AAP Homepage

Nathan Rosenberg

On Technological Expectations [1]

Economic Journal

Volume 86, Issue 343

Sept. 1976, 523-535.


I – Introduction

II – Innovation: “… a series of footnotes upon Schumpeter.”

III - Conclusions

I - Introduction

One of the most important unresolved issues in the theory of the firm and in the understanding of productivity growth is the rate at which new and improved technologies are adopted. [2]  I will argue that expectations concerning the future course of technological innovation are a significant and neglected component of these issues, inasmuch as they are an important determinant of entrepreneurial decisions with respect to the adoption of innovations.

Recent work in various aspects of economic theory and measurement has pointed, once again, to the important role played by expectations of future change in influencing the behaviour of economic agents.  In a number of instances the expectation of future changes has led to quite different patterns of behaviour than might have been expected to have taken place if it had been anticipated that no changes in level and/or trend would occur.  What would appear to be aberrational or irrational behaviour on the basis of such expectations of no changes is often fully explained once allowance is made for a different set of expectations about the future.

In this paper I should like to draw a similar analogy for the study of the diffusion of technological innovations.  The timing and nature of the adoption decision on the part of individual business firms is a key question with major implications for both the micro and macro levels of analysis.  I will suggest that there are expectational elements in the adoption decision which have not been given the attention and explored as systematically as could be done to illuminate the diffusion process.  Often the explanation of specific rates and patterns of technique adoption seems difficult to comprehend under the implicit conditions assumed, i.e. of no future changes in the technological and economic spheres.  Yet, as in other parts of economic analysis, the introduction of attention paid to various types of expectations, not only of prices but, more interestingly, of expectations concerning the future rate of technological change itself, will provide some important insights.  Since the technological future is, inevitably, shrouded in uncertainty, it is not surprising both that different entrepreneurs will hold different expectations, and also that entrepreneurial behaviour will further differ due to varying degrees of risk aversion on the part of decision-makers.

1. I am heavily indebted to Stanley Engerman for his encouragement and frequent counsel in developing the central argument of this paper and in exploring some of its implications.  The paper has also benefited greatly from Paul David’s searching criticism of an earlier draft, and from valuable comments and suggestions by David Mowery, W. B. Reddaway, Ed Steinmuller, George Stigler and an anonymous referee.  I am also grateful to the National Science Foundation for financial support during the time this paper was being written.

2. Ed Mansfield’s work has forcefully called attention both to the general overall slowness as well as to the wide differentials in adoption rates among different innovations . See, for example, Ed Mansfield, Industrial Research and Technological Innovation (W. W. Norton and Co., New York, 1968), chapter 7.


In terms of historical issues there are two patterns of expectations with quite different implications which need to be examined.  One, which has been implicit in the comparisons of nineteenth century technological change in the United States and the United Kingdom, is the impact of steady but differing rates of technological change in the two economies.  As pointed out by Habakkuk, and more formally by Williamson, the effect of the expectation of higher rates of technological change in the United States should have led to a shorter optimal life for American machinery - i.e. in some sense a more rapid introduction of new techniques. [1]  More interesting, perhaps, are other patterns of expectations which may be more important in studying the diffusion process.  Specifically, at certain times it may be more plausible to anticipate an acceleration of the rate of technological change.  Similarly, there may be situations where large-scale improvements are confidently expected after the introduction of some major innovation. [2]  In such cases these expectations may lead to a surprising result of making rational a delay in the widespread diffusion of the innovation.  Therefore, in analysing any historical decision with respect to diffusion, one must be sensitive to the specific nature of the expectations held by entrepreneurs with respect to the future course of technology.


II – Innovation: “… a series of footnotes upon Schumpeter.”

If, as Alfred North Whitehead once asserted, the history of western philosophy may be adequately described as a series of footnotes upon Plato, it may equally be said of the study of technological innovation that it still consists of a series of footnotes upon Schumpeter.  Although the footnotes may be getting longer, more critical and, happily, richer in the recognition of empirical complexities, we still occupy the conceptual edifice which Schumpeter built for the subject.  Inevitably, therefore, Schumpeter’s concepts constitute our point of departure.

Schumpeter’s theory of capitalist development, it will be recalled, starts out from the circular flow of economic life where producers and consumers are all in equilibrium, and where all adjustments and adaptations have been made.  Schumpeter then introduces an innovation - a shift in the production function - into this circular flow.  The entrepreneurial response to this new profit prospect in turn generates a sequence of alterations in the behaviour of economic actors, beginning with an expansion of bank credit and including, eventually, a secondary wave of investment activity imposed on top of the primary wave as the expectations of the larger business community are affected by the evidence and by the consequences of business expansion. [3]

1. H. J. Habakkuk, American and British Technology in the Nineteenth Century (Cambridge University Press, 1962), and Jeffrey Williamson, “Optimal Replacement of Capital Goods: The Early New England and British Textile Firm”, Journal of Political Economy, Nov./Dec. 1971.

2. For earlier empirical studies dealing with the life-cycle of specific technological innovations, see Simon Kuznets, Secular Movements in Production and Prices (Houghton Muffin, 1930), chapter I, and Arthur F. Burns, Production Trends in the United States Since 1870 (National Bureau of Economic Research, 1934), especially chapter 4.

3. J. A. Schumpeter, The Theory of Economic Development (Harvard University Press, 1934), chapters and 2.


Schumpeter himself was so much persuaded of the large elements of risk and uncertainty inherent in the innovation decision that he down-played the role of rational calculation itself in the decision-making process.  The Schumpeterian entrepreneur is a distinctly heroic figure, prepared (unlike most mortals) to venture forth boldly into the unknown.  His decisions are not the outcome of precise and careful calculation and, Schumpeter emphasised, cannot be reduced to such terms.

The point to be made here is that there is a further dimension of uncertainty in the innovation decision of a sort not emphasised by Schumpeter in his stress on the discontinuous nature of technological innovation.  This is, quite simply, the uncertainty generated not only by technological innovations elsewhere in the economy, but by further improvement in the technology whose introduction is now being considered.  Schumpeter’s argument creates a presumption that the first innovator reaps the large rewards.  Nevertheless, the decision to undertake innovation X today may be decisively affected by the expectation that significant improvements will be introduced into X tomorrow (or by the firmly held expectation that a new substitute technology, Y, will be introduced the day after).  The possible wisdom of waiting is reinforced by observations, abundantly available to all would-be entrepreneurs, concerning the sad financial fate of innumerable earlier entrepreneurs who ended up in the bankruptcy courts because of their premature entrepreneurial activities. [1]  As soon as we accept the perspective of the ongoing nature of much technological change, the optimal timing of an innovation becomes heavily influenced by expectations concerning the timing and the significance of future improvements.  Even when a new process innovation passes the stringent test of reducing new average total costs below old average variable costs, it may not be adopted.  The reason for this is that the entrepreneur’s views about the pace of technological improvements may reflect expectations of a higher rate of technological obsolescence than that allowed for by conventional accounting procedures in valuing the investment.  Moreover, accounting formulae may not give adequate recognition to the “disruption costs” involved in introducing new methods, especially when such disruptions are frequent.  Thus, a firm may be unwilling to introduce the new technology if it seems highly probable that further technological improvements will shortly be forthcoming. [2]  This problem

1. Marx long ago called attention to “the far greater cost of operating an establishment based on a new invention as compared to later establishments arising ex suis ossibus.  This is so very true that the trail-blazers generally go bankrupt, and only those who later buy the buildings, machinery, etc., at a cheaper price, make money out of it” (Karl Marx, Capital (Foreign Languages Publishing House, Moscow, 1959), vol. III, p. 103).  He also called attention to the rapid improvements in the productivity of machinery in its early stages as well as the sharp reduction in the cost of its production.  “When machinery is first introduced into an industry, new methods of reproducing it more cheaply follow blow by blow, and so do improvements, that not only affect individual parts and details on the machine, but its entire build” (Karl Marx, Capital (Modern Library Edition, New York, no date), vol. I, p. 442).  In a footnote on that page, Marx cites approvingly Babbage’s statement: “It has been estimated, roughly, that the first individual of a newly invented machine will cost about five times as much as the construction of the second.”  For a discussion of related problems with respect to the growth of nations, see Ed Ames and Nathan Rosenberg, “Changing Technological Leadership and Economic Growth”, Economic Journal, March 1963.

2. Fellner’s discussion of what he calls “anticipatory retardation” is relevant here. See William Fellner, “The Influence of Market Structure on Technological Progress”, Quarterly Journal of Economics [(1951), pp. 556-77, as reprinted in R. Heflebower and G. Stocking (eds.), Readings in Indust Organization and Public Policy (Richard D. Irwin, 2958).  The discussion of “anticipatory retardation” appears on pages 287-8 of that volume.]

HHC: [bracketed] displayed on page 526 of original.

525 Index

of how the optimal timing of innovation is affected by expectations of dis­continuous technological change is an extremely significant one which has received relatively little attention in the theoretical literature.’ Nor have there been systematic empirical studies of the phenomenon.

In their earliest stages, innovations are often highly imperfect and are known to be so. Innumerable “bugs” may need to be worked out. [2]  If one anticipates significant improvements, it may be foolish to undertake the innovation now - the more so the greater the size of the financial commitment and the greater the durability of the equipment involved.  Whereas the Schumpeterian innovator experiences abnormally high profits until the “imitators” catch up with him, the impetuous innovator may go broke as a result of investing in a premature model of an invention.  This apparent distinction follows from the earlier-stated difficulties of Schumpeter’s concept of innovation with its emphasis on discontinuity and its implication that all problems in the introduction of a new product or process have already been completely solved.  Moreover, as Mansfield points out: “In cases where the invention is a new piece of equipment, both the firm that is first to sell the equipment and the firm that is first to use it may be regarded as innovators.  The first user is important because he, as well as the supplier, often takes considerable risk.” [3]

1. Of course attention has been paid to the optimal timing of the introduction (and scrapping) machinery, but these models generally do not deal with problems of expected future changes technology.  Rather, they are more often concerned with issues relating to relative factor prices, future product demands, or the relation between machine use and deterioration.  While the expectation relating to technology can no doubt be easily incorporated into such models, the specific problem virtually unexplored.  For a brief discussion, see Vernon L. Smith, Investment and Production (Harvard University Press, 2962), pp. 143-5.  The relationship between expectations concerning the rate technological progress and the rate of return on investment in the context of a model of embodied technological change is discussed by Robert Solow in Capital Theory and the Rate of Return (No Holland Publishing Co., 1963, pp. 62-4).  For an interesting treatment of a somewhat related problem, how the introduction of a new technology will be affected by expected future growth in demand under different forms of market structure, externalities and property rights, see Yoram Ban “Optimal Timing of Innovations”, Review of Economics and Statistics, August 1968.

2. This term should be taken to include a great many production problems involving the use new equipment which it is almost impossible to anticipate and which become apparent only a result of extensive use - e.g. metal fatigue in aeroplanes.  William Hughes has made this point well with respect to exploration of the scale frontier in electric power generation: “Even under the most favorable conditions for advancing the scale frontier the cost side of the equation imposes fairly strict upper limits on the economical pace of advance, and trying to force the pace could mean sharply rising cost of development.  The experience required for pushing out the scale frontier is related time and cannot be acquired by increasing the number of similar new units.  Perhaps the great uncertainties connected with units arise from problems that may not show up until the units have been in operation a few years.  For the industry as a whole, the socially optimal number of pioneer: units during the first two or three years of any major advance in scale, design, or steam conditions is probably rather small, most often ranging from perhaps two or three or half a dozen” (William Hughes, “Scale Frontiers in Electric Power,” in William Capron (ed.), Technological Change Regulated Industries (The Brookings Institution, Washington, D.C., 1971), p. 52).  One of the other virtues of the Hughes article is its forceful reminder of the intimate link which often exists between technological progress and economies of scale.  “The realization of latent scale economies is especially important form of technological progress in the utility industries” (ibid., p. 45).

3. Edwin Mansfield, Industrial Research and Technological Innovation, op. cit., p. 83.


De Tocqueville long ago pointed out a distinctive characteristic of the American scene: “...I accost an American sailor, and I inquire why the ships of his country are built so as to last but for a short time; he answers without hesitation that the art of navigation is every day making such rapid progress that the finest vessel would become almost useless if it lasted beyond a certain number of years.  In these words, which fall accidentally and on a particular subject from a man of rude attainments, I recognize the general and systematic idea upon which a great people directs all its concerns.” [1]  Similarly, expectations of future changes may have the effect, not of delaying innovation, but of determining some of the specific characteristics of the innovation chosen.  An adaptation to these expectations might be deliberately to construct cheaper and flimsier capital equipment.  Thus, for example, as between two economies with different expected rates of future technical change, it would be anticipated that optimal life would be shorter where expected changes are largest. [2]  And, correspondingly, an anticipated acceleration in the course of technological improvement could lead to the selection of equipment with expected optimal life shorter than would otherwise be chosen.

Central to the analysis of the problem of expectations with respect to further improvements in technology, is the question of the specific source of subsequent improvements.  It may be that these improvements can only be brought about by the process which we have come to call “learning by doing”, in which case the pace of improvement is itself determined by either the producers’ accumulated experience over time or by their cumulative output. [3]  Alternatively, improvements may be rigidly linked to the passage of time necessary to acquire information about the results of earlier experience. [4]  Consider the problems associated with the introduction of commercial jet aeroplanes.  Britain introduced the Comet I two years before the Americans began the development of a jet airliner.  Yet the Americans eventually won out.  In retrospect, it is apparent that the American delay was salutary rather than costly to them, and that Boeing and Douglas chose the moment to proceed better than did de Havilland.  “Their delay allowed them to offer airplanes that could carry

1 Alexis De Tocqueville (trans. Henry Reeve), Democracy in America, 2 vols. (D. Appleton and Company, New York, 290!), vol. I, p. 526.

2. For a discussion of optimal replacement in the ante bellum textile industries of the United States and the United Kingdom, see Williamson, op. cit.  For a criticism of the Williamson article which does not alter the relationship between expected rates of technical change and optimal life, see David Denslow, Jr., and David Schulze, “Optimal Replacement of Capital Goods in Early New England and British Textile Firms: A Comment”, Journal of Political Economy, May/June 2974.

3. See Paul David, “Learning by Doing and Tariff Protection: A Reconsideration of the Case of the Ante-Bellum United States Cotton Textile Industry”, Journal of Economic History, September 1970, and Tsuneo Ishikawa, “Conceptualization of Learning by Doing: A Note on Paul David’s ‘Learning by Doing and... The Ante-Bellum United States Cotton Textile Industry’ “, ibid., December 2973.  Note the importance of this distinction as well as that between learning which can be freely captured by others and that which accrues solely to the learner.  This latter distinction will have important implications for the number of firms in an industry as well as the optimal entry date for any one firm. See Barzel, op. cit.

4. This is not really “learning by doing” in the usual sense, which is restricted to learning by participation in the production process, even though it does involve the passage of time and the accumulation of experience.  However, the distinctions above concerning the “appropriability” of the information generated by the experience remain important for examining entry decisions.

527 Index

up to 180 passengers when the Comet IV carries up to 100, and a cruising speed of 550 m.p.h. instead of 480 m.p.h. - hard commercial advantages that they could offer because they were designing for later and more powerful engines.  But they were also aided by the delay of four years in making the Comet safe after its accidents from metal fatigue.” [1]  More generally, information concerning the useful life of metal components could only be derived from prolonged periods of use and experience.  Which forms of improvements may be expected to dominate, and the conditions under which they may become available to other firms, will have important implications not only for entrepreneurs making entry or adoption decisions, but also for public policy concerning efficient growth.

There are many possible reasons why waiting may be the most sensible decision.  Indeed, often there may be no real choice.  On the purely technological level, innovations in their early stages are usually exceedingly ill-adapted to the wide range of more specialised uses to which they are eventually put.  Potential buyers may postpone purchase to await the elimination of “bugs” or the inevitable flow of improvements in product performance or characteristics.  On the other hand, they may have to wait through the lengthy process of product redesign and modification before a product has been created which is suitable to specific sets of final users.  Thus, widely used products such as machine tools, electric motors and steam engines have experienced a proliferation of time-consuming changes as they were adapted to the varying range of needs of ultimate users.  In the case of machine tools, for example, there was a successful search for the application of specific machine tool innovations across a wide variety of industrial uses. [2]  In the case of final consumer goods, the redesigning is likely to be primarily concerned with the development of product varieties suitable to the financial resources of different income groups.  What is observed over their life cycles is a gradual expansion of their quality range to accommodate these final users. [3]  Today’s academics are keenly aware of this problem when deciding whether to purchase today or to defer the purchase of pocket calculators, given their expectations of ongoing improvement in their capacities and characteristics.  Similar problems have characterised the selection decision with respect to the last few

1. Ronald Miller and David Sawers, The Technical Development of Modern Aviation (Praeger Publishers, New York, 1970), p. 27.  For an American failure in the attempted development of the “Demon” fighter plane under the pressures of the Korean War, see Eighth Report of the Preparedness Investigating Sub-committee of the Committee on Armed Services, United States Senate (84th Congress, 2d Session), Navy Aircraft Procurement Program: Final Report on F3H Development and Procurement (United States Government Printing Office, Washington, 1956).  The specific failure here was that numerous airframes became available years before it was possible to develop a jet engine with the required performance characteristics.  In the lugubrious tone of the congressional investigating committee which was appointed to account for the resulting loss of several hundred million dollars: “What has somehow been overlooked is that the essential procurement practice employed with respect to the Demon fighter would inevitably result in some wastage of airframes if the engine were not forthcoming” (ibid., pp. 9-10).

2. Nathan Rosenberg, “Technological Change in the Machine Tool Industry, 1840-1910”, Journal of Economic History, December 1963.

3. For historical evidence in support of these assertions, see Dorothy Brady, “Relative Prices in the Nineteenth Century”, Journal of Economic History, June 1964, pp. 145-203.


generations of computers, and may also have influenced the choice between purchase and rental on the part of users. [1]

There are some areas, of course, where the pressures militate very heavily against delay - as in weapons acquisition.  Nevertheless, even when the costs of being late may be regarded as uniquely high, considerations of technological expectations cannot be ignored.  Peck and Scherer, in their very careful study of weapons procurement, rephrase two of the key questions.  First, “Should we begin developing this newly feasible weapon system immediately to ensure early availability, or should we wait a year or so until the state of the art is better defined so that fewer costly mistakes will be made during the development?”  Secondly, “... should funds be committed to long lead time production items early in the development program so that full-scale production can start as soon as the development is completed?” [2]

A recent important instance in which the combination of long lead time and rapid technological change led to an apparent large misallocation of resources is the case of automobile emission controls.  Under pressure from the federal government, American automakers, with production lead times of four or five years, were required to specify which anti-pollution technology would be incorporated into the 1975 automobiles.  The catalytic converter seemed at that time to be a proven and more certain solution to the problem than did alternative technologies, such as the still-unproven stratified-charge engine.  However, the subsequent development of this latter technology produced a solution to automobile emissions which was superior to the catalytic converter.  Foreign automakers, such as Honda, with shorter production lead times than the Americans, were able to adopt the stratified-charge technology, thus avoiding what now appears to be the costly and premature commitment of American auto firms.  (It is important to note, however, that some of the difficulty in this particular case is due to uncertainty concerning the future of the emission standards themselves as well as technological uncertainty.)

Problems such as these have, of course, a long history.  In his book on the early history of electrical equipment manufacturing, Harold Passer extensively documents the marketing difficulties which confronted makers before 1900.  During this period, expectations of rapid technical improvement were firmly entrenched in the minds of potential buyers, and such expectations served to reduce present demand. “The manufacturer has to convince the prospective buyer that no major improvements are in the offing.  At the same time, the manufacturer must continue to improve his product to maintain his com-

1. See William F. Sharpe, The Economics of Computers (Columbia University Press, New York, 1969), particularly chapter 7.

2. Merton J. Peck and Frederic M. Scherer, The Weapons Acquisition Process: An Economic Analysis (Division of Research, Graduate School of Business Administration, Harvard University, Cambridge, 1962), pp. 283, 318.  The authors also observe: “The risks of early investment in production are impressive - in some programs enormous quantities of special tools and manufactured material became worthless due to unexpected technical changes.  On the other hand, the gamble has been successful and valuable time savings have been achieved in other programs (such as in the prewar British radar effort and in the first U.S. ballistic missile programs) by preparing for production concurrently with development” (ibid., pp. 318-19).

529 Index

petitive position and to force existing products into obsolescence.” [1]  Passer’s statement neatly identifies the horns of the dilemma which threaten to impale the entrepreneur in the early stages of product innovation.  One must attempt to persuade potential buyers of product stability at the same time as one commits resources to the search for product improvement so as not to fall behind the pace of such improvement which is set by one’s competitors.

The implications of such considerations may be very great for the innovation process.  When these intertemporal considerations loom very large, a rapid rate of technological progress need by no means result in as rapid a rate of introduction of new technological innovation.  As Sayers has pointed out for Great Britain:

There were times, between the wars, when marine engineering was changing in such a rapid yet uncertain way that firms in the highly competitive shipping industry delayed investment in the replacement of old high-cost engines by the new low-cost engines.  In the middle ‘twenties progress was rapid in all three propulsion methods - the reciprocating steam-engine, the geared steam-turbine and the diesel motor.  Minor variations are said to have brought the number of possible combination types up to nearly a hundred.  For some classes of ships there was momentarily very little to choose between several of these combinations, and shipowners were inclined to postpone placing orders until a little more experience and perhaps further invention had shown which types would be holding the field over the next ten years.  Put in economic terms, the shipowners’ position was that, though total costs of new engines might already be less than running costs of old engines, the profit on engines of 1923 build might be wiped out by the appearance in 1924 of even lower-cost engines, the purchase of which would allow a competitor (who had postponed the decision) to cut freights further.  Also there was uncertainty as to which of two types of 1923 engine would prove to work at lower cost.  If shipowner A installed engine X, and shipowner B installed engine Y, whose costs in 1923 appeared to equal those of X, a year’s experience might show that in fact Y costs were much lower than X costs, in which event shipowner A would have done better to wait until 1924 before installing new engines. [2]

1. Harold Passer, The Electrical Manufacturers, 1875-1900 (Cambridge, Harvard University Press, 1953), p. 45.  The 1880s and the 1890s were, in fact, a period of continuous product improvement in incandescent lighting, and it was only with the advent of metallic filaments in the early years of the twentieth century that the incandescent lamp established its decisive superiority over gas lighting and arc lighting.  Even so, the record from 1880 to 1896 is one of startling improvements in product efficiency and reductions in cost.  Bright has stated: “The following figures show the approximate course of list prices per lamp for standard 16-candlepower lamps from 1880 to 1896: 1880-6, $1.00; 1888, 80 cents; 1891, 50 cents; 1892, 44 cents; 1893, 50 cents; 1894, 25 cents; 1895, 18-25 cents; 1896, 12-18 cents... In 1896 a dollar could buy approximately six standard carbon-filament lamps which would give more than twelve times as many lumen-hours of light as a single lamp of the same candlepower costing a dollar in 1880” (Arthur A. Bright, Jr., The Electric-Lamp Industry (New York, Macmillan, 1949, pp. 93 and 134).

2. R. S. Sayers, “The Springs of Technical Progress in Britain, 1919-39”, Economic Journal, June 1950, pp. 289-90.  These circumstances might also be described as a case of “technological uncertainty” as to which specific methods would yield the greatest long-run effect, but the impact [upon diffusion is the same.  Habakkuk has called attention to a similar experience in shipbuilding immediately after the opening of the Suez Canal: “It accelerated the technical perfection of the steamer; the rate of technical progress was so rapid that the steamers built in the early ‘seventies were unable to compete with those completed in the middle of the decade...” (H. J. Habakkuk, “Free Trade and Commercial Expansion, 1853-1870”, in The Cambridge History of the British Empire, vol. 2, p. 762).]

HHC: [bracketed] displayed on page 531 of original.


For a much more recent period, Walter Adams and Joel Dirlam call attention to the U.S. Steel Corporation’s concern over the imminence of future improvements as an explanation for their delay in introducing the oxygen steel-making process.

U.S. Steel conceded that “some form of oxygen steel-making will undoubtedly become an important feature in steelmaking in this country”, but it declined to say when or to commit itself to introducing this innovation.  Indeed, three years later, Fortune still pictured the Corporation as confronted by “painfully difficult choices between competing alternatives - for example, whether to spend large sums for cost reduction now [1960], in effect committing the company to present technology, or to stall for time in order to capitalise on a new and perhaps far superior technology that may be available in a few years”. [1]

Expectations of continued improvement in a new technology may therefore lead to postponement of an innovation, to a slowing down in the rate of its diffusion, or to an adoption in a modified form to permit greater future flexibility.  Moreover, one must consider expectations relating not only to possible improvements in the technology being considered, but also the possibility of improvement in both substitute and complementary technologies.  Further improvements in an existing product may be held up because of the expectation that a superior new product will soon be developed. [2]  At the same time expectations of continued improvement in the old technology, which the new technology is designed to displace, will exercise a similar effect.  There is much evidence to suggest that historically, the actual improvement in old technologies after the introduction of the new were often substantial and played a significant role in slowing the pace of the diffusion process, so that this provides a quite reasonable basis for such a set of expectations.  The water wheel continued to experience major improvements for at least a century after the introduction of Watt’s steam engine; and the wooden sailing ship was subjected to many major and imaginative improvements long after the introduction of the iron-hull steamship. [3]  During the 1920S the competition of the internal combustion engine is said to have been responsible for much technological improvement in steam engines, while in the same period the competition from the radio stimulated experiments which led to the new

1. Walter Adams and Joel Dirlam, “Big Steel, Invention, and Innovation “, Quarterly Journal of Economics, May 1966, pp. 18 1-2.  Emphasis by Adams and Dirlam.

2. Jewkes et al. have pointed out that “... improvements in the more traditional methods of producing insulin were held up by the widespread belief that a synthesized product would soon be found”.  See John Jewkes, David Sawers and Richard Stillerman, The Sources of Invention (Macmillan and Co., London, 1958), p. 232.

3. See Nathan Rosenberg, “Factors Affecting the Diffusion of Technology”, Explorations in Economic History, Fall 1972, for a more extended discussion.

531 Index

and improved type of cable which was introduced in 1924. [1]  The Welsbach mantle, perhaps the single most important improvement in gas lighting, was introduced after the electric utilities had begun to challenge the gas utilities over the respective merits of their lighting systems.  The Welsbach gas mantle brought about a dramatic increase in the amount of illumination produced by a standard gas jet. [2]  Not only the diffusion of technologies but also the effort devoted to the development of new technologies may be decisively shaped by expectations as to future improvements and the continued superiority of existing technologies.  One explanation for the limited attention devoted to the development of the electric motor for many years was the belief that the economic superiority of the steam engine was overwhelming and beyond serious challenge. [3]  The decision to neglect research on the electrically powered car in the early history of the automobile industry reflected the belief, justified at the time, in the total superiority of the internal combustion engine (this neglect may soon be repaired!).  Similarly, the limited shift to nuclear sources of power over the past quarter century has been influenced by continued improvements in thermal efficiency based upon the “old-fashioned” but still apparently superior fossil fuel technologies.  It seems equally clear that the recent growing pessimism about future fossil fuel supplies is likely to accelerate the concern with nuclear technologies, and lead to a more rapid rate of technical improvements there as experience accumulates.  This will first require, however, a careful sorting out and distinction between short-term and long-term phenomena.  Large-scale commitment of private resources to nuclear energy (or shale processing) is unlikely so long as investors in energy-supplying industries anticipate that the price of oil is likely to fall sharply from its post-October 1973 levels. [4]

1. Sayers, op. cit., pp. 284-5.

2. “The Weisbach mantle was a lacy asbestos hood which became incandescent when attached over a burning gas jet.  It increased the candlepower six-fold with a white light far superior to the yellowish flame of the bare jet... The Welsbach mantle, improved by many changes in the original design, extended the life of gas lighting nearly half a century.  It sustained the gas utilities while they were discovering other productive uses for gas which would permit them to give up the struggle against electric lighting” (Charles M. Coleman, P.G. and E. of California, the Centennial Story of Pacific Gas and Electric Company, 1852-1952), p. 81.

Even the old arc-lamp technology experienced substantial improvements in response to the competition of new forms of lighting. See Bright, op. cit., pp. 213-18.

It is interesting to note that, even though the gas industry could not indefinitely meet the competition of electricity in lighting, the gas industry as a whole continued to find new uses for its product and experienced no long-term decline.  “Despite the declining importance of gas in lighting, the manufactured-gas industry as a whole has grown continually.  The competition of the carbon lamp and the old open arc during the 1880s had encouraged the gas industry to spread its field to heating, and it was that use which permitted the industry to continue expanding when its lighting market was destroyed.  Cooking, space heating, water heating, and later refrigeration resulted in a steady growth in the value of products of the industry from $56,987,290 in 1889 to $512,652,595 in 1929” (Bright, op. cit., p. 213).

3. Kendall Birr, “Science in American Industry”, in David Van Tassel and Michael Hall (eds.), Science and Society in the United States (The Dorsey Press, Homewood, Illinois, 1966), p. 50.

4. Dependence upon price expectations has, of course, been extensively discussed elsewhere.  The purpose of the present article is confined to calling attention to the significance of expectations with respect to technological change itself.  Expectations about future changes in technology might themselves be produced by projecting the anticipated influence of relative price changes on the innovation process.  The willingness to explore for new techniques which involve factor combinations drastically [different from those which currently prevail, will in turn depend upon the magnitude of expected price shifts.]

HHC: [bracketed] displayed on page 533 of original.


The recent efforts by oil firms to gain control of competing technologies - coal and uranium - may be seen as attempts to assure long-term market control by minimising the potential threats arising from technological breakthroughs in the provision of substitute products. [1]  The earlier experience of gas companies in meeting the competition of electricity is also highly instructive in this regard.  The ubiquitousness of “Gas and Electric” utility companies in the United States today suggests something of their past success in making the transition from the old to the new technology.

Expected profitability will not only be affected negatively by expected improvements in substitute technologies; it will also be affected positively by expected technological improvements in complementary technologies.  Since innovation in steel-making will be affected not only by innovations in aluminum or pre-stressed concrete which provide potential substitutes for steel, but also by technological changes in petroleum refining which increase the size of the automobile-producing sector, the entrepreneur must consider expected developments in these other sectors.  The profitability of technological changes in electric power generation will be favourably affected by metallurgical improvements which provide power plant components with increased capacity to tolerate higher pressures and temperatures in the form of high-strength, heat-resistant alloys.  On the other hand, improvements in power generation would have a limited impact upon the delivered cost of electricity without improvements in the transmission network which reduce the cost of transporting electricity over long distances.  The point is that the need for and expected availability of complementary innovations will often affect the profitability and therefore the diffusion of an innovation. [2]  Therefore single technological breakthroughs hardly ever constitute a complete innovation. [3]  It is for this reason - the expected as well as realised changes in other sectors - that decisions to adopt an innovation are often postponed in situations which might otherwise appear to constitute irrationality, excessive caution, or over-attachment to traditional practices in the eyes of uninformed observers.

1. Such attempts, by providing a limited diversification for these firms, reduce the risks they face, and this may conceivably have some role in encouraging research activity.

2. See Fishlow’s discussion of the importance of expectations with respect to transport improvements and their impact upon American land settlement patterns.  Albert Fishlow, American Railroads and the Transformation of the Ante-Bellum Economy (Harvard University Press, 1965).  Fishlow explains why mid-western railroads were profitable immediately after their construction as a result of the process of anticipatory settlement.  More broadly, it would be of interest to examine the impact of the expectations generated by transport improvements upon agricultural practices in the older areas of the northeast.

Parker and Klein have argued that the mechanisation of American agriculture which so increased its productivity would not have occurred in the absence of those transport improvements which made it possible to introduce its products into world markets.  See William Parker and Judith Klein, “Productivity Growth in Grain Production in the United States, 1840-60 and 1900-10”, in Studies in Income and Wealth No. 30, Output, Employment and Productivity in the United States after 1800 (Columbia University Press, 1966).

3. Rosenberg, “Factors Affecting the Diffusion of Technology”, op. cit.

533 Index

III - Conclusions

Some significant (and superficially paradoxical) implications can be drawn from this discussion.  Specifically, the relationship between the rate of technological change and the rate of technological innovation and diffusion is by no means a simple one, and may well be the opposite of intuitive expectations.  A rapid rate of technological change may lead to a seemingly slow rate of adoption and diffusion, or to the introduction of machinery which fails to incorporate the most “advanced” technology, so long as it leads potential buyers to anticipate, by extrapolation, a continued or accelerating rate of future improvements.  A decision to buy now may be, in effect, a decision to saddle oneself with a soon-to-be-obsolete technology. [1]  Conversely, when the rate of technological change slows down and the product stabilises, the pace of adoptions may increase owing to the much greater confidence on the part of potential buyers that the product will not be superseded by a better one in a relatively short period of time.  Thus the lag behind the “best available” methods might appear less when technological change is at a slower rate or is decelerating than when it is more rapid.  In this sense, our argument may explain the failure of firms to function at the “best practice” frontier - such failure may owe a great deal to differences in entrepreneurial expectations about the future pace of technological improvement. [2]

There are two distinct issues involved here.  One is the rate of improvement of best practice technology.  The second is the rate of adoption of those best practice methods.  If the two were independent, any lag in adoption would seem to impose social costs.  However, as seems clear from historical experience, an important explanation of lagged adoption is the environment created by the high rate of improvement of best practice technology.  Thus, a lagged rate of adoption is the “price” paid by technologically dynamic economies for their technological dynamism.

Clearly, a further examination of the adoption of technological innovations must be conducted within an enlarged framework which includes expectations not only of own-improvements but of improvements in the range of closely-linked substitutes and complements as well.  Further research along such lines may considerably improve our understanding of the diffusion process.  For the present, I would suggest that decisions to postpone the adoption of an innovation are often based upon well-founded and insufficiently-appreciated expectations concerning the future time-flow of further improvements.  Even the most widely accepted justification for postponement, the elimination of conspicuous but not overwhelmingly serious technical difficulties, or “bugs”,

1. As discussed, the effect might also be to lead to the initial adoption of a more flexible set of techniques which more readily permits the adoption of future improvements.  For an analogous argument with respect to the benefits of flexibility in the design of plant and equipment, see George Stigler, “Production and Distribution in the Short Run “, Journal of Political Economy, June 1939.  Reprinted in American Economic Association, Readings in the Theory of Income Distribution (Philadelphia, 1946).

2. For a discussion of other reasons for the failure to operate at the “best practice” frontier, see W. E. G. Salter, Productivity and Technical Change (Cambridge University Press, 1960), chapter 4.


can reasonably be interpreted as merely a special case of expectations of future technological improvement.  In this case, the expectations approach complete certainty that the technical difficulties can shortly be eliminated by recourse to the application of ordinary engineering skills.  I suggest, finally, that entrepreneurs may be making appraisals of the future payoff to innovations of greater objective validity than are made by social scientists who invoke all sorts of extra-rational factors to account for the delay or “lag” in the adoption and diffusion of innovation.  Practical businessmen tend to remember what social scientists often forget: that the very rapidity of the overall pace of technological improvement may make a postponed adoption decision privately (and perhaps even socially) optimal.

Nathan Rosenberg

Stanford University



The Competitiveness of Nations

in a Global Knowledge-Based Economy

May 2003

AAP Homepage