Science, Invention and Economic Growth
The Economic Journal
Volume 84, Issue 333
NOT too many years ago most economists were content to treat the process of technological change as an exogenous variable. Technological change - and the underlying body of growing scientific knowledge upon which it drew - was regarded as moving along according to certain internal processes or laws of its own, in any case independently of economic forces. Intermittently, technological changes were introduced and adopted in economic activity, at which point the economic consequences of inventive activity were regarded as interesting and important - both for the contribution to long-term economic growth and to short-term cyclical instability. Schumpeter, for example, saw the engine of capitalist development as residing in this innovative process in the long run, and at the same time he developed a business cycle theory which centred upon the manner in which the capitalist economy absorbs and digests its innovations. In Schumpeter’s model, exogenous technological changes stimulated investment expenditures, the variations of which, in turn, generated cyclical instability.
In the years after the Second World War the economist’s attitude gradually changed. The vast expenditures on Research and Development made it increasingly obvious that inventive activity was - or could be made to be - responsive to economic needs (or even to non-economic needs if such needs received sufficient financial support). Clearly much of the search activity of R and D was highly purposive: business firms were looking for new techniques in specific categories of products, they spent much money upon this search, and they were sometimes highly successful. Similarly, government agencies had long directed research into specific problem areas and in some cases had achieved conspicuous successes - as in agriculture.
In addition, the growth of interest in technological change after the Second World War was closely connected with the increasing concern over the prospects for economic growth in underdeveloped countries. When economists turned their attention to this range of problems, they brought with them an intellectual apparatus which placed overwhelming emphasis upon the role of saving and the growth in the stock of capital goods as the engine of economic growth. But it soon became clear that long-term economic growth had taken place at rates far beyond what could plausibly be accounted for by mere growth in the supply of conventionally-measured inputs. It became increasingly obvious that economic growth could not be adequately understood in terms of the use of more and more physical inputs,
1. The author is grateful to Professors S. Engerman, W. B. Reddaway and E. Smolensky, and to an anonymous referee for their helpful comments on earlier drafts of this paper. They are, however, accorded the usual absolution for all remaining deficiencies.
but rather that it had to be understood in terms of learning to use inputs more productively. With this realisation came, of course, a renewed interest in technological change as the source of rising resource productivity.
The growing interest in the role of technological change as a contributor to economic growth led to a considerable amount of empirical research on technological change, particularly in two areas: (1) attempting to quantify the contribution of technological change to the growth in long-term resource productivity; and (2) attempting to study the rate at which new inventions, once made, were diffused throughout the economy, since clearly inventions exert an impact upon resource productivity only to the extent that they are actually adopted in the productive process. The work of Griliches was particularly important in showing that one could explain the diffusion process in considerable detail as a response to economic forces - i.e., on the basis of profit expectations as shaped by market size. 
Increasingly, therefore, economists have become more and more confident of their ability to deal with technological events in economic terms. This growing confidence was capped by the publication of a major book by Jacob Schmookler in 1966, called Invention and Economic Growth (Cambridge: Harvard University Press). Schmookler argued, quite persuasively, not only that one could explain the diffusion of existing inventions in economic terms - a la Griliches - but that one could even explain the pattern of inventive activity itself.
As a result of these developments, the attitude of the economics profession toward technological change seems to be coming full circle. Whereas technological change was once regarded as an exogenous phenomenon moving along without any direct influence by economic forces, it is now coming to be regarded as something which can be entirely explained by economic forces. Indeed, factors on the technological and scientific levels are increasingly coming to be regarded as not constituting very interesting problems, because we already “know” that we can explain their particular timing in economic terms. 
Schmookler’s book is obviously very appealing to the economist because it argues that inventive activity is an essentially economic phenomenon, and that it can be adequately understood in terms of the familiar analytical apparatus of the economist. Perhaps I should anticipate my conclusions by saying that I propose to start off from Schmookler’s analysis, not because I am in search of a convenient straw man, but rather because I am in sub-
1 Zvi Griliches, “Hybrid Corn: An Exploration in the Economics of Technological Change,” Econometrica, October 1957.
2. The issue is not just whether the scientific and technological spheres are autonomous or not, although that has been a much-debated issue. Even if one were satisfied, for example, that the scientific realm is an autonomous sphere, it need not follow that events in that sphere are unpredictable. They may not be directly influenced by economic variables, but they may be moving subject to an internal logic or an external set of forces which can be identified and then used, by economists, to explain sequences of inventive activity.
stantial agreement with much that he has to say. Moreover, Schmookler’s analysis is so rich and so suggestive that it has to be the starting point for all future attempts to deal with the economics of inventive activity and its relationship to economic growth.
Schmookler’s ultimate interest is, to quote the opening sentence of his book: “What laws govern the growth of man’s mastery over nature?” His book represents an attempt to supply building blocks for the answer to that very big question by systematically studying two smaller questions: (1) how to explain the variations in inventive activity in any particular industry over time; and (2) how to explain different rates of inventive activity between industries at a given moment of time. Schmookler’s fundamental answer to these questions involves the attempt to link up inventive activity with the structure of human wants and therefore with changes in the composition of demand which are associated with rising per capita incomes and other related aspects of economic growth.
The empirical core of Schmookler’s book is an attempt to demonstrate, through the study of several American industries, that demand-side considerations are the major determinant of variations in the allocation of inventive effort to specific industries. In examining the railroad industry, for which comprehensive data are available for over a century, Schmookler found a close correspondence between increases in the purchase of railroad equipment and components, and slightly lagged increases in inventive activity as measured by new patents on such items. The lag is highly significant because, Schmookler argues, it indicates that it is variations in the sale of equipment which induce the variations in inventive effort. Schmookler finds similar relationships in building and petroleum refining, although the long-term data on these industries are less satisfactory.
Furthermore, and no less important, in examining cross-sectional data for a large number of industries in the years before and after the Second World War, Schmookler finds a very high correlation between capital goods inventions for an industry and the volume of sales of capital goods to that industry. These data support the view that inventors perceive the growth in the purchase of equipment by an industry as signalling the increased profitability of inventions in that industry, and direct their resources and talents accordingly.  Thus, Schmookler concludes that demand considera-
1. Schmookler draws the implication from his data on inter-industry variations in capital goods invention that “... inventive activity with respect to capital goods tends to be distributed among industries about in proportion to the distribution of investment. To state the matter in other terms, a 1 per cent increase in investment tends to induce a 1 percent increase in capital goods invention.” Schmookler, op. cit., p. 144. Emphasis Schmookler’s. It is important to note that Schmookler’s results “... depend critically on the fact that our capital goods inventions were classified according to the industry that will use them, not according to the industry that will manufacture the new product or the intellectual discipline from which the inventions arise.” Ibid., p. 164. See also p. 166.
tions, through their influence upon the size of the market for particular classes of inventions, are the decisive determinant of the allocation of inventive effort.
Far from being an exogenous variable as most economists had earlier believed - an activity which, although it had important economic consequences was not controlled by economic forces - Schmookler concludes that we can treat invention just like any other economic activity. Just as we can ana1yse production and consumption in terms of revenues and costs and the desire to maximise some relevant magnitude, so we can analyse inventive activity in precisely the same terms.
Schmookler not only attempts to incorporate inventive activity into an economic framework. Within that framework he attaches overwhelming importance, as already indicated, to demand forces, and regards supply side considerations as relatively subordinate and passive. Thus, in discussing consumer goods inventions, Schmookler argues that it is the changes in consumer demand over time which are the primary determinant of shifts inthe direction of inventive effort.
… (I)f we start out at a given point of time with relative outlay on the different classes of goods given, and allow capital accumulation, technical progress, education, and so on, to occur, then per capita income will gradually rise. In consequence the proportion of income spent on different classes of goods will also gradually change. As different classes of goods become relatively more important than before the yield to inventive effort in different fields will tend to change correspondingly. And if we further grant that inventive effort is influenced by prospective yield, the direction of inventive activity will shift. Thus even under the extreme assumption that the structure of generic wants is permanently fixed, economic progress will bring successive sections of that structure into play over time, thereby altering the reward structure confronting inventors and rechannelling their efforts accordingly. This is why, for example, American inventors concentrated on food production in the first part of the nineteenth century but gave much more attention in the twentieth century to the requirements of leisure, by creating motion pictures, radio, television, and so on. 
Schmookler’s argument, as presented so far, would seem to be subject to the fatal objection that its overwhelming emphasis upon demand simply ignores the whole thrust of modern science and the manner in which the growth of specialised knowledge has shaped and enlarged man’s technological capacities. Such growing technological sophistication, surely, suggests that at least some of the initiative in the changing patterns of inventive activity lies on the supply side and not on the demand side where Schmookler has placed it.
1. Schmookler, op. cit., pp. 180-1. Of course Schmookler is well aware that consumer expenditure on particular classes of goods is not entirely a function of prices and incomes. Such factors as age structure of a population, climate, geography, and extent of urbanisation, will also play an important role.
Schmookler has anticipated this objection, and his answer is in fact an ingenious one. He argues that the commodity classes towards which inventors direct their efforts are determined by expectations concerning financial payoffs which, in turn, are shaped by the familiar considerations of demand and market size. Developments on the side of science and technology are highly relevant to the inventive process, but only in determining the technical realms - mechanical, electrical, chemical, biological - upon which the inventor will draw. While the growth in knowledge at the scientific and technological levels will thus influence the specific characteristics of inventions, the purposes for which inventions are undertaken will depend upon the state of the market for classes of final commodities.
The point is that, while a marketable improvement in envelope-making equipment is probably about as easy to make as one in glass making, it may be easier today to make an improvement in either field via electronic means than through some mechanical change... If differences exist in the richness of the different inventive potentials of the product technologies of different supplying industries, the pressure to improve an industry’s production technology tends to be met by the creation of relatively more new products in supplying industries with richer product inventive potentials. For example, if new electrical machines are easier to invent than are non-electrical machines, then the aggregate demand for new machinery tends to induce relatively more electrical than non-electrical machinery inventions. In brief, inventors tend to select the most efficient means for achieving their ends, and at any given moment, some means are more efficient than others. 
Schmookler thus argues for the primacy of demand side considerations, not by suggesting that shifts on the supply side have been unimportant. Quite the contrary. Science and technology have brought about a great transformation in man’s capacity to pursue his material ends. But it is precisely because of the versatility of man’s enlarged inventory of scientific and technical skills that demand side forces retain their primacy.
Oddly enough then, science and technology play a subordinate role in influencing the direction of inventive activity within Schmookler’s analysis, not because his analysis downgrades their historical significance, but rather because he regards science and technology in the modern age as being, in a significant sense, omnicompetent. Schmookler looks upon the body of modern science and technology as constituting a kind of “putty clay” out of which almost anything can be shaped. As he states, “... mankind today possesses, and for some time has possessed, a multi-purpose knowledge base. We are, and evidently for some time have been, able to extend the technological frontier perceptibly at virtually all points.” 
Now this is precisely the aspect of Schmookler’s argument which seems to be most inadequate. If Schmookler is right, then economists need not
1. Ibid., pp. 2 10-11.
2. Ibid., p. 218. Emphasis Schmookler’s.
pay too much attention to the internal histories and structures of the sciences and technologies in order to understand the direction of inventive activity. If he is right, then science and technology have not functioned as major independent forces in shaping the timing and the direction of the inventive process. If economic forces can so powerfully shape, not only technology, but science as well, in the achievement of its own ends, then these subjects retain little interest for the economist or economic historian.  On the other hand, if Schmookler is wrong in this respect, then his analysis needs to be supplemented by a more careful examination of the manner in which the state of knowledge at any time shapes and structures the possibilities for inventive activity.
To establish the independent importance of supply side considerations, it is necessary to demonstrate several things: (1) That science and technology progress, in some measure, along lines determined either by internal logic, degree of complexity or at least in response to forces independent of economic need; (2) that this sequence in turn imposes significant constraints or presents unique opportunities which materially shape the direction and the timing of the inventive process; and (3) that, as a result, the costs of invention differ in different industries.
As soon as one speaks of the “costs of invention” it is necessary to recognise that the economic analysis of inventive activity is seriously handicapped by our present inability to specify the production function for inventive activity with any pretence of precision. Inventions, unfortunately, do not come in units of equal size, whether considered from the point of view of their usefulness or their costs of production. Both the inputs and the outputs in the production of invention are appallingly difficult to measure. Schmookler’s basic unit of measurement is, in fact, not an “invention” but a “patent” which serves as a surrogate for an invention. Schmookler’s primary interest is in illuminating the process through which society allocates resources to inventive activity. The extreme heterogeneity which is the essence of inventive output is, Schmookler believes, less serious a problem for his interests than it would be in an attempt to link up the number of inventions with the larger phenomena of technological progress and economic
1. “Thus, independently of the motives of scientists themselves and with due recognition of the fact that anticipated practical uses of scientific discoveries still unmade are often vague, it seems reasonable to suggest - without taking joy in the suggestion - that the demand for science (and, of course, engineering) is and for a long time has been derived largely from the demand for conventional economic goods. Without the expectation, increasingly confirmed by experience, of ‘useful’ applications, those branches of science and engineering that have grown the most in modern times and have contributed most dramatically to technological change - electricity, electronics, chemistry and nucleonics - would have grown far less than they have. If this view is approximately correct, then even if we choose to regard the demand for new knowledge for its own sake as a non-economic phenomenon, the growth of modern science and engineering is still primarily a part of the economic process.” Ibid., p. 177.
growth.  Schmookler appears content to regard inventive output as adequately measured by the mere number of inventions since, it is important to note, he is not attempting a direct link-up between the inventive process and the larger question of the historical growth in resource productivity. His results, he is careful to point out, “… apply only to the number of inventions made, not to their importance... One of the problems of research now is to establish the nature of the connection between the number of inventions in a field and the rate of technological progress.”  Within this framework the attempt to compare a unit of invention in one industry with a unit of invention in another industry (or even two inventions in the same industry) is obviously fraught with difficulty. Schmookler is content to observe that the prospective value of inventive output is likely to be greater in industries undertaking large amounts of investment than in industries where such investment is smaller. An industry’s volume of investment activity, in other words, is the primary determinant of the profitability of a unit of invention.
This leaves us very much in the dark in attempting to attach a larger significance to a unit of invention. It would be most convenient, for analytical purposes, if there were an identifiable unit of invention which lowered the cost of production in a plant by, say, 1 %. This would enable us to assess the importance of a unit of invention by relating it to the size or to the rate of growth of the adopting industry. Unfortunately, the extreme heterogeneity of inventive output simply does not allow us to assume any simple relationship between the number of inventions and the number of such units of invention or productivity growth.  Schmookler does, however, hold the view that the cost of invention is likely to be the same in all industries. He points out that “... the very high correlations obtained… between capital goods invention and investment levels in different industries, and the substantial similarity in the patent-worker ratio of durable and nondurable goods industries indicate that a million dollars spent on one kind of good is likely to induce about as much invention as the same sum spent on any other good. Hence, doubling the amount spent on one kind of good is likely to induce about as much invention as the same sum spent on any other good.”  This position raises serious difficulties to which we will shortly return.
Although Schmookler’s treatment of the relationship between demand
1. See ibid., chapter 2, for a searching examination of the problems involved in using patent statistics as a surrogate for inventions and also for Schmookler’s justification for his belief that the deficiencies in the patent data and the problems posed by vast qualitative differences in inventions are less than is generally supposed. For a careful discussion of the measurement problems involved in the economics of inventive activity, see Simon Kuznets, “Inventive Activity: Problems of Definition and Measurement,” in R. R. Nelson (ed.), The Rate and Direction of Inventive Activity, Princeton, 1962, pp. 19-43.
2. Schmookler, op. cit., p. 163. See also p. 208, footnote 1.
3. It is, of course tautologically true to say, as Schmookler does, that “A given percentage improvement in productivity is more valuable in a large than in a small industry.” Ibid., p. 91.
4. Ibid., p. 172. Emphasis Schmookler’s. See also pp. 209 and 212.
forces and invention is, in general, highly illuminating, his conceptual apparatus even here contains some disturbing gaps. This is apparent when he states that, “From a broader point of view, demand induces the inventions that satisfy it.” [l] One wishes to rush in at once with qualifications: some demand induces the inventions that satisfy it. But which, and when? As soon as these questions are raised we are compelled to consider the different rates at which separate branches of science have progressed. Many important categories of human wants have long gone either unsatisfied or very badly catered for in spite of a well-established demand. It is certainly true that the progress made in techniques of navigation in the sixteenth and seventeenth centuries owed much to the great demand for such techniques in those centuries, as many authors have pointed out. But it is also true that a great potential demand existed in the same period for improvements in the healing arts generally, but that no such improvements were forthcoming. The essential explanation is that the state of mathematics and astronomy afforded a useful and reliable knowledge base for navigational improvements, whereas medicine at that time had no such base. Progress in medicine had to await the development of the science of bacteriology in the second half of the nineteenth century. Although the field of medicine was one which attracted great interest, considerable sums of money, and large numbers of scientifically-trained people, medical progress was very small until the great breakthroughs of Pasteur and Lister. Improvements in the treatment of infectious diseases absolutely required progress in a highly specific discipline – bacteriology - and the main thrust of medical “inventions” in the past one hundred years would be difficult to conceive without it. Indeed, it is highly doubtful that, with the single exception of vaccination against smallpox, medical progress was responsible for any significant contribution to the decline in human mortality before the twentieth century. 
The point at issue here is one of general importance to Schmookler’s argument. The role of demand side forces is of limited explanatory value unless one is capable of defining and identifying them independently of the evidence that the demand was satisfied. It would not require a very lively imagination, as the references to medical progress suggest, to compile an extensive list of “high priority” human needs which existed for many centuries, which would have constituted highly profitable commercial activities, but which yet remained unsatisfied. Schmookler’s formulation is such that it is capable of being fitted to almost any conceivable set of historical observations. For his argument to be non-tautological, however, it would have to be formulated in such a way that the component elements
1. Ibid., p. 184.
2. This is the judgment recently delivered by medical historians. See Thomas McKeown and R. G. Brown, “Medical Evidence Related to English Population Changes in the 18th Century,” Population Studies, 1955-66, pp. 119-41, and Thomas McKeown and R. G. Record, “Reasons for the Decline of Mortality in England and Wales During the 19th Century,” Population Studies, 1962, pp. 94-122.
of demand could be identified independently of our observations concerning inventive activity. Until this is done it is difficult to conceive of any set of observations which could directly refute Schmookler’s hypothesis. In the absence of a reasonably clear, independent specification of the composition of demand, one can never demonstrate either that important components of demand have gone unsatisfied or that supply side factors played an important role in laying down the time pattern of inventive activity.
In fact, the argument of this paper is that, if we want to explain the historical sequence in which different categories of wants have been satisfied via the inventive process, we must pay close attention to a special supply side variable: the growing stock of useful knowledge. Historical evidence confirms that inventions are rarely equally possible in all commodity classes. The state of the various sciences simply makes some inventions easier (i.e., cheaper) and others harder (i.e., more costly). In considering the manner in which the stock of scientific knowledge has grown, and the manner in which this growth has, in turn, shaped the possibilities for inventive activity, one basic fact stands out: The world of nature contains many sub-realms, which vary enormously in their relative complexity. If one considers the broad sweep of scientific progress over the past 300 or 400 years, the timing and sequence of the growth of knowledge in these separate disciplines is closely related to the relative complexity of each - as well as to the complexity of the technology upon which scientific research in the discipline depends. For example, the microbial world and to a great extent the biological world could not be examined without the assistance of the microscope, and the contemporary study of the atomic structure of giant molecules awaited the technique of X-ray crystallography. On the other hand, it is not surprising that the disciplines which were carried to the most advanced state in antiquity were astronomy, mathematics, mechanics and optics. These were each disciplines which could be carried far on the evidence of unassisted human observations, with little or no reliance upon complex instruments or experimental apparatus.  Thus, a mastery of the principles underlying the mechanical world was attained long before a similar mastery was achieved over the principles of chemistry - almost 200 years, if we use as our benchmark dates the publication of Newton’s Principia on the one hand and Mendelejeff’s periodic table of the elements on the other. Similarly, within the discipline of chemistry itself, progress was more rapid in inorganic than in organic chemistry. Even though it had long been apparent that there were huge economic benefits to be reaped throughout the vegetable and animal worlds from a greater knowledge of organic chemistry, such knowledge persistently lagged behind the growing knowledge of inorganic chemistry. Organic chemistry long remained intractable and unresponsive to an obvious and compelling demand. Even after it had become apparent
1. See T. S. Kuhn, The Structure of Scientific Revolutions, Chicago, 1962, chapter VIII and the same author’s article, “The History of Science,” in the International Encyclopedia of the Social Sciences.
that all organic substances are composed of small numbers of elements - mainly carbon, hydrogen, oxygen and nitrogen - science quite simply remained baffled at the mysteries of the organic world. Progress in organic chemistry, we now know, lagged far behind inorganic chemistry because of a basic and unyielding datum of the natural world: the far greater size and structural complexity of organic molecules.  Similar considerations underlie a broad range of research activities and go far towards explaining the timing with which commercially marketable results are extracted from such activities. Thus, the molecular structure of vitamin B12, essential in the treatment of pernicious anaemia, is much more complex than vitamin B1 or C and, as a result, it took far longer to isolate, synthesise and place in commercial production. Similarly, the comparative lateness of the organic chemist’s successful assault upon the structure of protein molecules is largely attributable, we now know, to their great complexity. Amorphous materials, as a group, are much more complicated in their atomic structure than crystalline solids and have therefore required a much greater research effort to understand. Progress in the treatment of diabetes has long been held up by the inability to decipher the insulin molecule. Recent research utilising X-ray crystallography has finally revealed a remarkably complex three dimensional structure consisting of no less than 777 atoms. This finding goes a long way towards explaining why a more effective medical programme has taken so much longer to launch in the case of diabetes than in the relatively “simple” diseases such as malaria, syphilis or cholera. Much scientific research at the micro-biological level is, in fact, preoccupied with mapping out the highly complex structural arrangement of the component atoms of organic molecules. 
Thus, while I believe that Schmookler has supplied an essential corrective to an earlier, widely-held view which looked upon the scientific enterprise as not only totally exogenous to the economic sphere but even as a completely autonomous force, propelled by a purely internal logic, I also believe that
1. The great nineteenth century breakthroughs in organic chemistry in turn laid the basis for the subsequent twentieth century revolution in biology. As Bernal points out: “The new organic chemistry had another essential part to play in the history of science - it was to lead to a fuller understanding of biological processes. In fact, the beginning of any deeper understanding than the microscope could provide was totally impossible without a knowledge of the laws of combination and the types of structure actually to be met with in biological systems. The nineteenth-century development of organic chemistry had to precede logically any attempt to formulate a fundamental biology.” J. D. Bernal, Science in History, Cambridge, Mass., 1971, 4 vols., vol. 2, p. 633.
2. On the great inherent complexity of biological studies Bernal makes the following interesting observations: “... (T)he same degree of complexity of even the simplest forms of life is something of an entirely different order from that dealt with by physics or chemistry. What we had admired before in the external aspects of life, in the symmetry and beauty of plants and flowers, or in the form and motion of the higher organisms, now appear, in the light of our wider knowledge, relatively superficial expressions of a far greater internal complexity. That internal complexity is itself a consequence of the long evolutionary history through which living organisms have raised themselves to their present state.” Ibid., vol. 3, p. 868. The notion that scientific progress has moved in an orderly sequence from the less complex to the progressively more complex aspects of the physical universe is clearly expressed in Frederick Engels, The Dialectics of Nature, Moscow, 1954.
he has overstated his case in some important aspects. Although economic forces and motives have inevitably played a major role in shaping the direction of scientific progress, they have not acted within a vacuum, but with the changing limits and constraints of a body of scientific knowledge growing at uneven rates among its component sub-disciplines. The shifting emphasis of inventive activity over the past two centuries - mechanical, chemical, electrical, biological - is deeply rooted in the history of science, and it is difficult in the extreme to visualise how any plausible set of social and economic forces could have brought about a total reversal of that order.  Given that sequence in the development of science, inventive activity in some commodity classes was much easier than in others. Furthermore, although Schmookler is doubtless correct that we have an increasingly multi-purpose knowledge at our disposal, it is easy to exaggerate the extent to which separate sub-realms of knowledge offer genuine options in the satisfaction given categories of human wants, in the sense of presenting methods which are substitutes for one another. Such substitution is frequently non-existent and usually highly imperfect. Moreover, in many cases the inventive process confronts relationships of complementarity rather than substitution. Thus the great twentieth century transformation in world agriculture is largely a product of biological knowledge - the mastery of the principles of heredity which have made it possible to develop entirely new, highly productive strains such as hybrid corn in the 1930s and 1940s and, more recently, new wheat and rice varieties. But a fundamental characteristic of these life-science “ inventions “is their high degree of complementarity with chemical inputs. Indeed, the new high-yielding rice varieties recently introduced into south-east Asia are often no more productive than the traditional varities if they are grown under the old techniques of crop and soil management. Their unique feature is a high degree of fertiliser-responsiveness brought about by genetic manipulation. A much better name than “miracle” rice would be “fertiliser-responsive.” There are no miracles. In fact, the sharp increases in output per acre, which superficially suggest massive improvements in resource productivity, are really the result of large increases in fertiliser and other chemical inputs combined with rigorous attention to techniques of water management.  Thus, these biological inventions require for their success, large doses of chemical inputs: fertiliser on the one hand and pesticides to protect them from the many pests to which they are peculiarly vulnerable, on the other.  In this critical area of agricultural technol-
1. For a brief but highly perceptive treatment of some of the underlying problems, see William] Parker, “Economic Development in Historical Perspective,” Economic Development and Cultural Change, October 1961, pp. 1-7.
2. The complexity and costliness of water management methods in the growing of rice is major reason why the new wheat varieties have so often been introduced more rapidly and with greater success than the new rice varieties. This has been the case, for example, in India.
3. “We know from experience in the U.S. that the rapid introduction and widespread use new crop varieties accelerates the biological dynamics of crop disease-host plant relationships.” [Albert H. Moseman, Building Agricultural Research Systems in the Developing Nations, N.Y., Agricultural Development Council, 1970, p. 97.]
HHC – [bracketed] displayed on page 101 of original.
ogy, then, and in other areas as well, the dominant relationships are those of complementarity and not substitution. In this respect, therefore, our freedom of choice in drawing upon different realms of science and technology for ways of increasing food production is largely illusory. The range within which we can exercise genuine options in the achievement of specific goals is, in fact, severely attenuated.
When we move from the realm of science to that of technology, we enter a world where economic motives are much more direct, immediate and pervasive. Since technological concerns are dealt with primarily within a matrix of profit-seeking business firms, one would expect to find, as one does, a high degree of responsiveness to conditions of market demand and profit expectations generally. But here too it is abundantly clear that an understanding of demand forces alone provides only very limited insight into the direction and the timing of inventive activity. Here, too, differences in the inherent complexity at the technological level shed a flood of light on the inventive process as it has occurred in historical time. If this is correct, then the Schmookler position that technological problems will be solved (one way or another) when the demand for such a solution is sufficiently pressing (i.e., profitable) is seriously incomplete, and needs to be supplemented by a careful scrutiny of supply side variables.
Consider one of the central events of the industrial revolution: the substitution of a mineral fuel for wood in industrial activities. The growing scarcity of wood and the desirability of substituting coal became increasingly clear in Great Britain as early as the second half of the sixteenth century, during which time the price of firewood rose far more rapidly than prices generally. By 1600 the growing pressure upon the limited supplies of firewood and timber had already produced numerous attempts to introduce coal into individual industries. And yet, in spite of strong and pervasive economic inducements, it took over 200 years before this substitution was reasonably complete. But what is particularly interesting from our present vantage point is that, in some industries, the transition to the new fuel was effected very rapidly, whereas in others, including some of the most important such as metallurgy, a span of 200 years was required.
Why? A complete answer would be long and complex, but a major part of the answer is that the substitution presented no technical problems at all in some industries, while it created very serious problems in others. No major problems arose in using coal in the evaporation of salt water in salt production, or in lime-making or in brick baking. But in other industries the use of the new fuel seriously reduced the quality of the final product
- as in glass-making, the drying of malt for breweries and most importantly, in the smelting of metallic ores. Throughout the seventeenth century considerable effort and experimentation were devoted to these problems. The problems of glass production were solved relatively early by the use of closed crucibles which protected the glass from the destructive effects of the mineral fuel (although, significantly, the method could be used only to produce a coarse cheap glass). In malt production a more palatable beer was being produced by mid-century by first reducing coal to coke and thus eliminating some of the offending elements. Later in the century a reverberatory furnace was introduced which was eventually successfully employed in the smelting of lead, tin and copper. The coke-smelting of iron was first achieved by Abraham Darby in 1709, but the method produced only a very inferior quality of iron. As a result the use of coke pig iron was restricted to the small, cast-iron branch of the iron industry, and charcoal pig iron continued to be used for almost another century for all high quality purposes. It was only after Henry Cort’s introduction of the puddling process in the 1780s for the refining of pig iron that the transition to mineral fuel was finally completed. 
Thus the timing of a whole series of inventions connected with the introduction of coal can be understood only in terms of a protracted effort at maintaining quality control while introducing coal into industrial uses. The use of coal created a series of new problems, of varying degrees of complexity, in different industries. Moreover, the fuel itself varied considerably in its chemical composition from one region to another. Since the nature of the chemical interchanges between the new fuel and the various raw materials with which it was employed were not understood, a great deal of time was required (in some cases hundreds of years) before crudely empirical methods finally sorted out the economic opportunities presented by the new fuel. Moreover, the sequence in which solutions were found to the problems of different industries varied considerably, depending upon the technical difficulties involved. Indeed, it may be confidently asserted that the solution came last in precisely that industry where the economic payoff was greatest: the iron industry.2
1. See T. S. Ashton, Iron and Steel in the Industrial Revolution, Manchester 1924; John Nef, “The Progress of Technology and the Growth of Large-Scale Industry in Great Britain, 1540-1640,” Economic History Review, 1934, pp. 3-24; John Nef, “Coal Mining and Utilization,” in Charles Singer et al., A History of Technology, London, 1957, 5 vols., vol. 3, pp. 72-88; E. A. Wrigley, “The Supply of Raw Materials in the Industrial Revolution,” Economic History Review, August 1962, pp. 1-16.
2. It is interesting to note that the historic links between coal and the iron and steel industry persist even today, in spite of extensive attempts to sever the links. As a matter of fact, one of the reasons for the relatively large size of the coal industry today in the face of strong competition from other fuels has been the inability thus far, in spite of prolonged exploration, to develop a satisfactory technique for producing iron without the use of high-grade coal. Although other fuels have been readily substituted for coal in many uses, the substitution in metallurgical processes poses unique and so far intractable difficulties.
The burden of my argument here is that the allocation of inventive resources has in the past been determined jointly by demand forces which have broadly shaped the shifting payoffs to successful invention, together with supply side forces which have determined both the probability of success within any particular time frame as well as the prospective cost of producing a successful invention. But even if one were to accept the proposition, which I do not, that demand side forces alone determine the allocation of inventive resources, it would still remain true that supply side forces exercise a pervasive influence over the actual consequences of such resource use: i.e., the output of successful inventions, and the timing of these inventions. The explanation of the nature and composition of inventive output necessarily requires an understanding of the operation of supply side forces. These supply side forces determine whether the output is of the kind associated with the medieval alchemist or the modern scientific metallurgist, the medical quack and patent medicines or broad spectrum antibiotics. Even if knowledge of demand forces alone yielded sensible predictions about the direction of inventive effort, such knowledge, in the absence of further information about supply side forces (the state of scientific knowledge, the prevailing levels of technological skills, the specific characteristics of raw material inputs, etc.) is likely to provide only limited insight into the flow of inventive output.
If we turn to the sequence of invention in textiles, the first major industry to experience full mechanisation, one overriding fact stands out: mechanisation at all stages in the productive process came much earlier to the new cotton branch of the industry than to the older woollen branch. There were several economic reasons for this, which were rooted in the underlying conditions determining the supply of the basic raw materials on the one hand, and the nature of the demand for each of the final products on the other. But, in addition, there was again a fundamental technological fact: cotton production lent itself to mechanisation far more easily than did wool production for reasons intrinsic to the nature of the two materials. As Landes has aptly pointed out:
… (C)otton lent itself technologically to mechanization far more readily than wool. It is a plant fibre, tough and relatively homogeneous in its characteristics, where wool is organic, fickle, and subtly varied in its behaviour. In the early years of rudimentary machines, awkward and jerky in their movements, the resistance of cotton was a decisive advantage. Well into the nineteenth century, long after the techniques of mechanical engineering had much improved, there continued to be a substantial lag between the introduction of innovations into the cotton industry and their adaptation to wool. And even so, there has remained an element of art - of touch - in wool manufacture that the cleverest and most automatic contrivances have not been able to eliminate.’
1. David Landes, The Unbound Prometheus, Cambridge, 1969, p. 83. Landes also points out that, even after machinery was introduced into the wool industry, the machines could be operated only much more slowly than in cotton. Ibid., pp. 87-8.
If we consider the sequence in which machine technology was introduced into separate operations in American agriculture, the relative difficulty of applying machine methods to different operations again looms up as a critical variable. Why did the reaping and threshing of wheat come so much earlier than mechanisation in cotton picking, corn picking and husking, and milking? Here again, conditions affecting the demand for such individual inventions spring readily to mind. The harvesting of wheat was especially constrained by weather conditions in a way that the other crops were not. The peculiar history of the cotton-growing South provided that region with more abundant labour than other parts of the country and thus considerably weakened the incentive to introduce labour-saving machinery. Yet, as Parker has pointed out, milking operations were also subject to a very strong time constraint and were concentrated in labour-scarce regions of the country where the incentive to invent labour-saving machinery should have been correspondingly strong. Moreover, there is abundant evidence - e.g., from the Patent Office - that considerable, if unsuccessful, inventive effort had been directed toward these operations in the nineteenth century.
Surely the most plausible single answer,” Parker suggests,
is that these operations were all inherently difficult to mechanize without radical alteration and improvement of basic elements in the prevailing technology. In the case of the corn harvester, the problem of harvesting the ear separately from the stalk, while preserving the stalk for forage, was hard to solve. In cotton picking, the need to make several passes over the field as the bolls ripened prevented a crude solution. The possibility of mechanical milking was hardly dreamed of, except by cranks, before the gasoline engine and electric power. It is no accident that in all three cases, the mechanical problem was to imitate complex motions of the human hand rather than the simple sweeping actions of the arm required in reaping and threshing. 
A large part of the economic history of the past 200 years is; in fact, the story of an enormous outward shift in industrial man’s capacity to solve certain kinds of production problems. This growing capacity has been fitful and highly selective. For most of the nineteenth century it involved the exploitation of new power sources and an increasing mastery over the use of large masses of cheap metal (iron and, later, steel). These techniques became available with no fundamental accretions to basic knowledge. They nevertheless were developed slowly because it took time to develop and then to diffuse new techniques in the precision working of metals and to devise the innumerable small improvements and adaptations which were often required to enable them to operate successfully. There is always a gap, moreover, between the ability to conceptualise a mechanism or technique and the capacity to bring it into effect. Thus, da Vinci’s notebooks are full
1. William N. Parker, “Agriculture,” in Lance Davis, et al., American Economic Growth, New York, Harper and Row, 1971, p. 385.
of sketches for novel machinery which could not be realised, with the primitive metal-working techniques at his disposal. Breech-loading cannon had been made as early as the sixteenth century, but could not be used until precision in metal working in the nineteenth century made it possible to produce an air-tight breech and properly fitting case. (Without the air-tight breech, a breech-loading cannon was likely to present far greater danger to the persons engaged in firing it than it did to those at whom the fire was being directed.) Christopher Polhem, a Swede, devised many techniques for the application of machinery to the quantity production of metal and metal products, but could not successfully implement his conceptions with the power sources and clumsy wooden machinery of the first half of the eighteenth century. Although the principle of compounding was embodied in a patent in 1781, compound steam engines were not introduced into ocean-going vessels until the 1880s, a full century later, in spite of strong economic incentives. Not until major breakthroughs in steel-making technology was it possible to provide high quality components such as boiler plates and boiler tubes upon which the operating efficiency of the compound engine depended. Charles Babbage had conceived of the main features of the modern calculator over a century ago, and had incorporated these features in his “analytical engine,” a project which was even favoured with a large subsidy from the British Exchequer. Babbage’s failure to complete this ingenious scheme was due to the inability of the technology of his day to deliver the components which were essential to the machine’s success.
The purpose of this recitation of frustrations and failures is simply to argue that, given the state of purely scientific knowledge, society’s technical competence at any point in time constitutes a basic determinant of the kinds of inventions which can be successfully undertaken. Of course it is possible to argue, as it has been with respect to the long delay in the introduction of a mechanical cotton picker, that if factor prices and/or cotton prices had been significantly different, a practical machine would have been introduced much earlier. If, for example, the available labour supply had been much more expensive, more inventive effort would presumably have been devoted to solving the complex technical problems of a cotton picking machine much sooner. While this is probably true, it is also incomplete. Because it is also true that, given the set of factor and commodity prices which actually prevailed, the cotton picking machine would also have been developed more quickly if the technical problems which had to be overcome were less serious. These technical problems and their relative complexity stand independently of demand considerations as an explanation of the timing and direction of inventive activity. Therefore any analytical or empirical study which does not explicitly focus upon both demand and supply side variables is seriously deficient.
Where has this analysis taken us? I have argued that the central weakness of Schmookler’s approach is his treatment - or, rather, his neglect - of the supply-responsiveness of technology and invention.
Essentially, Schmookler is saying that, given the state of science (and regardless of “how we got here”) the supply of inventions is, in effect, perfectly elastic, and at the same price, in all industries. At any moment in time it is possible to get as many inventions as wanted in any industry at a constant price. Therefore the observed composition of inventions is entirely a demand side phenomenon, reflecting the manner in which inventive resources have been allocated between industries (or, better, commodity classes) in response to the structure of (demand-induced) profit expectations.
The main objection which I have raised is that inventions are not equally possible in all industries. This is because there is a crucial intervening variable: the differential development of the state of sub-disciplines of science and bodies of useful knowledge generally at any moment in time. Indeed, I think it is very important that we cease talking about “the state of science” and begin thinking in terms of “sciences.” A central problem is to trace out carefully the manner in which differences in the state of development of individual sciences and technologies have influenced the composition of inventive activities. Let me suggest further that one way of getting at this is to pay more attention to historical failures.
Our understanding of inventive activity (and perhaps of social change generally) is excessively rooted in success stories. We study the history of successful inventions but devote little attention to inventions which were not made. Yet it is highly relevant to ask why it took so long to do certain things, and why inventors failed for so long at some inventive efforts while they succeeded quickly at others. It is certainly possible to study past patterns of research expenditure and inventive effort, and to seek the reasons for unsuccessful as well as successful outcomes, for very long gestation periods in the development of new inventions as well as for shorter periods.  In short, if we want to probe the relations between science, technology and inventive activity more deeply, we must learn much more about what was not possible as well as what was possible. We need to understand what scientific and technological discoveries were needed for key breakthroughs in invention. For knowledge not only permits - it also constrains. For this reason we can learn much from the study of unsuccessful attempts to invent something
1. It is worth mentioning here that our lack of interest in the study of failures may also have contributed in an important way to an under-estimation of the costs of invention. In our preoccupation with success stories we inevitably ignore the substantial commitment of resources to unsuccessful inventive efforts, and recognise only those which were connected with a successful outcome.
for which the market was perceived to be ready. In this respect, the study of failure is essential to a determination of the precise role of supply side variables in the inventive process. After all, the demand for higher levels of food consumption, greater life expectancy, the elimination of infectious disease, and the reduction of pain and discomfort, have presumably existed indefinitely in the past, but they have been abundantly satisfied only in comparatively recent times. It. seems reasonable to suppose that the explanation is to be found in terms of supply side considerations. It is unlikely that any amount of money devoted to inventive activity in 1800 could have produced modern, wide-spectrum antibiotics, any more than vast sums of money at that time could have produced a satellite capable of orbiting the moon. The supply of certain classes of inventions is, at some times, completely inelastic - zero output at all levels of prices. Admittedly, extreme cases readily suggest arguments of a reductio ad absurdum sort. On the other hand, the purely demand-oriented approach virtually assumes the problem away. The interesting economic situations surely lie in that vast intermediate region of possibilities where supply elasticities are greater than zero but less than infinity!
The perspective which I am suggesting, therefore, states that, as scientific knowledge grows, the cost of successfully undertaking any given, science-based invention declines - from infinitely high, in the case of an invention which is totally unattainable within the present state of knowledge, down to progressively lower and lower levels. Perfectly inelastic supply curves of invention gradually unbend and flatten out. (To what extent they flatten out is, of course, an empirical question, on which Schmookler has adopted the arbitrary and implausible extreme assumption of perfect elasticity.) Thus, the growth of scientific knowledge means a gradual reduction in the cost of specific categories of science-based inventions. The timing of inventions therefore needs to be understood in terms of such shifting supply curves which gradually reduce the cost of achieving certain classes of inventions. More precisely, we need to think in terms of a number of supply curves for individual industries, depending upon the knowledge bases upon which inventive activity in that industry can draw, and we need to understand more clearly the extent to which different “ pools” of knowledge are potential substitutes in the inventive process. Schmookler’s hypothesis states, in effect, that there is one supply curve for all industries and that the extent of substitution renders it unnecessary to look at supply conditions in individual industries. It seems to me that a clear articulation of the relations between science, invention and economic growth requires a critical examination of this assertion. The basic economic question, of course, is not an “either or” proposition telling us whether a particular technological achievement is or is not possible at a particular point in time. The economic question is: Given the state of the sciences, at what cost can a technological end be attained? How does the state of individual sciences differentially structure
the cost of society’s technological options?  Answers to these questions will carry us a long way towards a deeper understanding of both the nature of inventive activity and the process of economic growth by providing further insight into the economy’s changing capacity to respond to economic needs.
1. Note that my emphasis upon supply and cost considerations does not imply any sort of scientific or technological determinism. More costly inventions can always precede less costly ones in time if demand conditions are sufficiently strong.