Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 307
Microeconom~cs of Technological Innovation EDWIN MANSFIELD We are still very far from having a satisfactory understanding of the innovation process, the determinants of the rate of innovation, the measurement of the rate and direction of technological change, and the effects of changes in technology. In view of the enormous difficulties that studies in these areas face, it is not surprising that existing knowledge remains limited. Nonetheless, steadily progress has been made in this area. Until the 1950s, the economics profession generally was woefully deficient in its treatment of technological change. With few exceptions, like Joseph Schumpeter (1934), economists failed to recognize the central importance of industrial innovation. In the past 30 years. however, a great deal of research, much of it financed by the National Science Foundation, has been carried out by economists to extend the* limited understanding of the nature, determinants, and effects of technological change. Without question, si~,- nificant progress has been made. New models have been constructed and new kinds of data have been assembled. Yet we are far from a satisfactory understanding of this very difficult topic. This chapter describes briefly some of the principal work that has been done to help answer the following questions: (1) What has been the effect of research and development (R&D) on the rate of productivity growth? (2) What has been the rate of return from investments in industrial innovation? (3) What have been the size, determinants, and effects of imitation costs? (4) How much effect have patents had on imitation costs and the rate of innovation? (5) How great has been the rate of inflation in R&D? (6) What factors deterTnine the rate of diffusion of an innovation? (7) To what extent has the rate of international technology transfer increased? 307
OCR for page 308
308 EDWIN MANSFIELD While these are not the only questions that microeconomists have dealt with, they clearly are among the most important. Since there is a considerable literature on each of these questions, my treatment is selective, and it will tend to focus on the kinds of work that my students and I have been doing in recent years. RELATIONSHIP BETWEEN R&D AND PRODUCTIVITY GROWTH For both analytical and policy purposes, it is important to investigate the relationship between the amount spent by an industry or firm on R&D and its rate of productivity increase. Dunng the past 30 years a number of studies of this kind have been carried out. They are by no means free of problems, however, as I have indicated in detail elsewhere (Mansfield et al., 19821. Perhaps the most important of their findings is that R&D has a very significant effect on the rate of productivity increase in the industries and the time periods that have been studied. In one of the earliest studies of this topic, Minasian (1969) found that the rate of productivity increase in chemical firms was directly related to their expenditures on R&D. His results indicated that, during the period to which his data pertained, the marginal rate of return— that is, the rate of return from an additional dollar spent was about 50 percent for R&D in chemicals. A study of my own indicated that the marginal rate of return from R&D was about 40 percent or more in the petroleum industry and about 30 percent in the chemical industry (Mansfield, 19651. In agnculture, Gnliches (1964) found that output was related in a statistically significant way to the amount spent on research and extension. Assuming a 6-year lag between research input and its returns, his results indicated a marginal rate of return from agncultural R&D of 53 percent. Another study, by Evenson (1968), used time-genes data to estimate the marginal rate of return from agricultural R&D, which it found to be about 57 percent. Peterson's study ( 1971 ) of R&D in poultry indicated a marginal rate of return of about 50 percent. Throughout the 1970s these studies were extended. Gnliches (1980) used data for almost 900 manufacturing fibs to examine He relationship between R&D and the rate of productivity growth. The results indicated that the amount spent by a firm on R&D was directly related to its rate of productivity growth. Also, he found that the private rate of return from R&D was about 17 percent. It seemed much higher than this in chemicals and petroleum and lower in aircraft and electrical equipments Terleckyj (1974) studied the effects of R&D expenditures on productivity change in 33 manufacturing and nonmanufactunng industries during the years 1948 to 1966. In manufacturing, the results suggested about a 30 percent rate of return from an indus~y's R&D on its own productivity. In addition, his findings showed a very substantial effect of an indus~y's R&D on pro-
OCR for page 309
MICROECONOMICS OF TECHNOLOGICAL INNOVATION 309 ductivity growth in other industries, resulting in a social rate of return greatly exceeding that of 30 percent. Nadin and Bitros (1980) constructed an econo- metric model in which output was treated as an exogenous variable and R&D, labor, and capital inputs were regarded as functions of input prices, sales, the rate of capacity utilization, and the lagged dependent variables. They found that labor productivity is significantly affected in both the short run and the long run by the level of a firms R&D expenditures. During the 1980s, studies by Griliches, Link, Nadiri, Scherer, Terleckyj, and myself, among others, provided still further evidence concerning the rates of return from R&D.2 In interpreting the rates of return obtained, both in these studies and in those done before, it is important to distinguish between private rates of return and social rates of return. The private rate of return is the rate of return to the Finn carrying out the R&D; He social rate of return is the rate of return to society as a whole. Since the Grin Hat carries out the R&D frequently cannot appropriate many of the benefits of doing so, the social rate of return may be considerably in excess of the private rate of retum. Thus, many of the rates of return cited above, because they are social rates of rehem, are likely to be higher than the private rate of return to the firm carrying out the R&D. How do the results of the studies of the 1980s compare with those of the studies of the 1970s and 1960s? In general, they are quite consistent, in the sense that they continue to indicate that the level of R&O seems to be closely related to the rate of productivity grown, and that the marginal rate of return from investment in R&D is high, although perhaps not as high as in earlier years. The high marginal social rate of return from R&D is important because it suggests that there may be an undennvestment in R&D, a phenomenon that many economists attribute partly to the differences between private and social rates of return from innovative activities. SOCIAL AND PRIVATE RETURNS FROM SPECIFIC INNOVATIONS The types of econometric studies just described have been employed by economists to estimate the social rates of return from investments in new technology. But they are by no means He only type of study carried out by economists for this purpose. A number of microeconomic studies of the returns from specific innovations have been carried out as well. To estimate He social benefits from an innovation, economists have often used a model of the following sort. If the innovation results in a shift downward in the supply curve for a product, they have used the Yea under the product's demand curve between the preinnovation and postinnovation supply curves as a measure of the social benefit from the innovation during the relevant time period. If all other prices remain constant, this area equals the social value of He additional quantity of the product plus the social value of the
OCR for page 310
310 EDWIN MANSFIELD resources saved as a consequence of the innovation. Consequently, if one compares the stream of R&D (and other) inputs relating to the innovation with the stream of social benefits measured in this way, it is possible to estimate the social rate of return from the investment in the new technology. The first such studies concerned only agricultural R&D. Although only a few studies of this sort were conducted, notably by Gnliches (1958), Peterson (1971), and Schmitz and Seckler (1970), the results were quite consistent in the sense that Hey all indicated that the rate of return from agricultural R&D in the United States has tended to be high. Until the 1970s, no such estimates were made for ~ndustnes other than agriculture. In an attempt to help fill this gap, my co-workers and I estimated the rate of return from the investment in 17 industrial innovations, which occurred in a variety of industries and which stemmed from fins of quite different sizes (Mansfield et al., 1977b). Most of these innovations were of average or routine importance, not major breakthroughs. Although this sample cannot be regarded as randomly chosen, there is no obvious indication that it was biased toward very profitable innovations socially or privately—or relatively unprofitable ones. To estimate the social rate of return from the investment in each of these innovations, we extended the model described above to include the pricing, behavior of the innovator, the effects on displaced products, and the costs of uncommercialized R&D and of R&D done outside the innovating orga- nization. The results indicate that the median social rate of return from the investment in the innovations studied was 56 percent, a very high figure. On the other hand, the median private rate of return was 25 percent. (In interpreting the latter figure, it is important to note that innovation is a risky activity; see Mansfield et al., 1971.) Information also was obtained concerning the returns from the innovative activities of one of America's largest fins, from 1960 to 1972. For each year this fin had made an inventory of the technological innovations re- sulting from its R&D and estimated its effects on its profit stream in detail. When the average rate of return from this f~n's total investment in innovative activities was computed, the result was 19 percent, which is not too different from the median private rate of return (25 percent) noted above. Also, lower bounds were computed for the social rate of return from this firm's invest- ment, which was about double that for its private rate of return (see Mansfield et al., 1977a). To extend our sample and replicate our analysis, the National Science Foundation commissioned two studies, one by Robert R. Nathan Associates and one by Foster Associates. Their results, like ours, indicated that the median social rate of return tends to be very high and much higher than the median private rate of return. Based on its sample of 20 innovations, Nathan Associates (1978) found the median social rate of return to be 70 percent and the median private rate of return to be 36 percent. Foster Associates
OCR for page 311
MICROECONOMICS OF TECHNOLOGICAL INNOVATION 311 (1978), based on its sample of 20 innovations, found the median social rate of return to be 99 percent and the median private rate of return to be 24 percent. These findings pertain to the average rate of retum. As pointed out in the previous section, econometric investigations indicate that the marginal rate of return has also tended to be high. In sum, practically all of the studies carried out to date indicate that the average social rate of return from in- vestments in new technology in both agriculture and industry has tended to be very high. Moreover, the marginal social rate of return also seems high, generally at least 30 percent. As I have stressed repeatedly elsewhere, there are very important problems and limitations inherent in each of these studies. Certainly, they are frail needs on which to base policy conclusions. But recognizing this fact, it nonetheless is remarkable that so many independent studies based on so many types of data result in so consistent- a set of conclusions. As noted above, many economists view these conclusions as evidence of an underinvestment in civilian technology. BASIC RESEARCH AND PRODUCTIVITY While the foregoing studies provide valuable information concerning the relationship between total R&D input and productivity change, they tell us nothing about the effect of the composition of an industry's or firm's R&D on its rate of productivity change. In particular, they tell us nothing about the role of basic research in promoting productivity increase. Basic research is defined by the National Science Foundation as "original investigation for the advancement of scienuf~c knowledge . . . which [does] not have im- mediate commercial objectives." Does basic research, as contrasted win applied research and development, make a significant contribution to an industry's or firm's rate of technological innovation and productivity change? Although the studies cited above indicated that an indusmy's or f~rrn's R&D expenditures have been directly related to its rate of productivity change, they were unable to shed light on this question because no attempt was made to separate basic research from applied research and development. About 5 years ago, an econometric study was carried out to determine whether an industry's or f~rrn's rate of productivity change in recent years had been related to the amount of basic research it perfo~ed, when other relevant variables (such as its rate of expenditure on applied R&D) were held constant. This study (Mansfield, 1980, 1981) has various limitations, but its results are of interest, particularly since so little research has been done on this subject. The findings indicate that there was a statistically significant and direct relationship between the amount of basic research carried out by an industry or firm and its rate of increase of total factor productivity, when its expenditures on applied R&D were held constant. To
OCR for page 312
312 EDWIN MANSFIELD some extent, this may reflect a tendency for basic research findings to be exploited more fully by the industries and Grins that were responsible for them. Or it may reflect a tendency for applied R&D to be more effective when carried out in conjunction with some basic research. Whether the relevant distinction is between basic and applied research is by no means clear: there is some evidence that basic research may be acting to some extent as a proxy for long-term R&D. Holding constant the amount spent on both applied R&D and basic research, an industry's rate of pro- ductivity increase seems to be directly and significantly related to the extent to which its R&D is long term. This constituted the first systematic evidence that the composition, as well as the size, of an industry's or fien's R&D expenditures affect its rate of productivity increase. However, the study was really just a beginning. Much more work is required in this area, since the composition of R&D is for many purposes as important as its total amount. CENTRAL ROLE OF IMITATION COSTS AND TIMES Economists have long recognized that the costs of imitating new products have an important effect on the incentives for innovation in a market econ- omy. As Arrow (1962) and others have pointed out, if fins can imitate an innovation at a cost that is substantially below the cost to the innovator of developing the innovation, there may be little or no incentive for the innovator to carry out the innovation. In Heir discussions of the innovation process, economists frequently have called attention to the major role played by the costs of imitation, but there has been little or no attempt to measure those costs, to test various hypotheses concerning the factors influencing them, or to estimate their effects. In the first empirical study of this topic (Mansfield et al., 1981) data were obtained from fimns in the chemical, drug, electronics, and machinery in- dustries concerning the cost and time of imitating (legally) 48 product in- novations.3 By imitation cost is meant all costs of developing, and introducing the imitative product, including applied research, product specification, pilot plant or prototype construction, investment in plant and equipment, and manufacturing and marketing start-up. (If there was a patent on the inno- vation, the cost of inventing around it is included.) By imitation time is meant the length of time that elapses from the beginning of the imitator's applied research (if there was any) on the imitative product to the date of its commercial introduction. The sample of firms for this study was chosen more or less at random from among the major firms in these four industries in the Northeast, and He new products were chosen more or less at random from among, Hose introduced recently by these firs. For 34 of the products, the data are based on acn}a1 experience, because the new product had already been imitated.
OCR for page 313
MICROECONOMICS OF TECHNOLOGICAL INNOVATION 313 For the remaining 14 products, no imitator had appeared as yet, but the innovating Finn provided detailed estimates that we regarded as reliable. Also, for all 48 products, data were obtained from the innovating fib con- ce~ning the costs of the innovation, as well as the time it took to bring the innovation to market (from the beginning of applied research to the date of its commercial introduction). The ratio of the imitation cost to the innovation cost averaged about 0.65, and the ratio of the imitation time to the innovation time averaged about 0.70. There was considerable variation about these averages, however. For about half of the products, the ratio of imitation cost to innovation cost was either less than 0.40 or more than 0.90. For about half of the products, the ratio of imitation time to innovation time was either less third 0.40 or more than 1.00. Products with a relatively high (low) ratio of imitation cost to innovation cost tended to have a relatively high (low) ratio of imitation time to innovation time. It is worth noting that the imitation cost was no smaller than the innovation cost for about one-seventh of the products. This was not due to any superiority of the imitative product over the innovation. Instead, in a substantial per- centage of the cases, it was due to the innovator's having, a technological edge over its rivals in the relevant field. Often this edge was due to superior "know-how" that is, better and more extensive technical information based on highly specialized experience with the development and production of related products and processes. Such know-how is not divulged in patents and is relatively inaccessible (at least for a period of time) to potential Imitators. Thus, these data indicate that innovators routinely introduce new products despite the fact Mat other firms can imitate those products for about two- thirds (often less) of the cost and time expended by the innovator. In some cases, this is because, although other firms could imitate these products in this way, there are other barriers to entry (for example, lack of a well-known brand name) that discourage potential imitators. But to a greater extent (at least in this sample), it seems to be due to a feeling on the part of the innovators that, even if imitators do begin to appear in a relatively few years, the innovation will still be profitable. Patents and Imitation Costs In recent years, economists have also begun to study more systematically We effects of patents. Of particular interest is the question: To what extent is the ratio of imitation cost to innovation cost affected by whether the innovator has patents on the new product? Contrary to popular opinion, patent protection does not make entry impossible, or even unlikely. Within 4 years of their introduction, 60 percent of the patented successful innovations in
OCR for page 314
314 EDWIN MANSFIELD the above sample were imitated. Nonetheless, patent protection generally increased imitation costs. To obtain information concerning the size of this increase, my co-workers and I asked the finns in the sample described above to estimate how much the imitation cost for each patented product increased because it was patented. The median estimated increase was 11 percent.4 We also asked the firms to estimate how much the imitation cost for each unpatented product would have increased if it had been patented. The median estimated increase was only about 6 percent. (Indeed, for two of these prod- ucts, patent protection would have reduced the money and time required for imitation because without patent protection the innovator was able to keep secret the essential information underlying the product, whereas if the product had been patented, some of the inflation would have had to be disclosed. ) The fact that a patent resulted in a larger increase in the imitation costs of the patented products than of the unpatented products was, of course, a major reason why some products were patented and others were not. Patents had a larger impact on imitation costs in ethical drugs than in the other industries sampled, which helps to account for survey results indicating that patents are regarded as more important in ethical drugs than elsewhere. The median estimated increase in imitation cost due to patent protection was about 30 percent in ethical drugs, in contrast to about 10 percent in chemicals and about 7 percent in electronics and machinery Without patent protection, it frequently would have been relatively inexpensive (and quick) for an imitator to determine the composition of a new Mug and to begin producing it. These results are in accord with the conclusion of Taylor and Silberston (1973) that the lack of patent protection would reduce the rate of expenditure on innovative activity to a greater extent in drugs than in other industnes. Imitation Costs, Entry, and Concentration Let us turn now from the determinants of imitation costs to their effects on market entry and industry concentration. Holding constant the discounted profit (gross of the imitation cost) that the imitator expects to earn by imitating a new product, the new product is more likely to be imitated if the imitation cost is small. To discourage market entry, We innovator may adopt pricing (and other) policies to reduce the imitator's expected discounted gross profit if the imitation cost is low. Taking this into account, is it still true that the probability of entry is inversely related to the size of the imitation cost? To find out, my co-workers and I determined whether each product innovation in the sample described above was imitated within 4 years of its introduction. (Innovations that had been on the market fewer than 4 years and unsuccessful innovations clearly had to be omitted.) We then carried out a logit analysis5 to detennine whether We ratio of imitation cost to innovation cost influenced the probability that entry of this sort occurred within ~ years. Based on the
OCR for page 315
MICROECONOMICS OF TECHNOLOG1C~ INNOVATION 315 results, imitation cost seemed to be related in the expected way to whether entry occurs. Imitation cost may also affect an industry's level of concentration. One would expect an industry's concentration level to be relatively low if its members' products and processes can be imitated easily and cheaply. Existing econometric findings, which are based on limited data, have been entirely consistent with this hypothesis. Given the large number of factors influencing an industry's concentration level, it is interesting that this relationship is relatively close. Differences among industries in the technology transfer process (including transfers that are both voluntary and involuntary from the point of view of the innovator) may be able to explain much more of the nterindustry variation in concentration levels than is generally recognized. PATENTS AND THE RATE OF INNOVATION One of the most important and controversial questions concerning the patent system is: What proportion of innovations would be delayed or not introduced at all if they could not be patented? To shed light on this question, economists have camed out carefully designed surveys to determine the proportion of their patented innovations that Grins report they would have introduced (with no appreciable delay) if patent protection had not been available. Although answers to such questions have obvious limitations and must be treated with caution, they should shed some light on this topic, about which so little is known. According to the Grins in one such study (Mansfield et al., 1981), about one-half of the patented innovations would not have been introduced without patent protection. The bunk of these innovations occurred In the drug industry. Excluding drug innovations, the lack of patent protection would have affected less than one-fourth of Me patented innovations in the sample. One important reason why patents frequently are not regarded as crucial is that Hey often have only a limited effect on the rate at which imitators enter the market. For about half of the innovations, the firms in this study believed Hat patents had delayed the entry of imitators by less than a few months. Although patents generally increased the imitation costs, they did not increase the costs enough in these cases to have an appreciable effect on the rate of entry. But although patent protection seems to have only a limited effect on entry in about half of the cases, it seems to have a very important effect in a minority of ~em. For about 15 percent of the innovations, patent protection was estimated to have delayed the time when the first imitator entered the market by 4 years or more. In another study (Mansfield, 1985) based on a random sample of 100 firms from 12 industries (excluding very small firms) in the United States, the results indicate that patent protection was judged by He fimns to have been
OCR for page 316
316 EDW N MANSFIELD essential for the development or introduction of 30 percent or more of the inventions commercialized in 2 industries pharmaceuticals and chemicals. In another 3 industries (petroleum, machinery, and fabricated metal prod- ucts), patent protection was estimated to be essential for the development and introduction of about 10 to 20 percent of their inventions. In the remaining 7 industries (electrical equipment, office equipment, motor vehicles, instru- ments, primary metals, rubber, and textiles), patent protection was estimated to be of much more limited importance in this re Hard. Indeed, in office equipment, motor vehicles, rubber, and textiles, the firms were unanimous in reporting that patent protection was not essential for the development or introduction of any of their inventions during the years 1981 to 1983. This does not mean, however, that firms make little use of the patent system. On Me contrary, even in those industries in which practically all inventions would be introduced without patent protection, the bulk of the patentable inventions seem to be patented. And in such industries as phar- maceuticals and chemicals, in which patents are important, over 80 percent of the patentable inventions are reported to have been patented. Clearly, firms generally prefer not to rely on trade secret protection when patent protection is possible. Even in industnes, like motor vehicles, in which patents are frequently said to be relatively unimportant, about 60 percent of the patentable inventions seem to be patented. Despite We frequent assertions Nat films are making less use of the patent system than in Me past, the evidence does not seem to bear this out. Even in electronics, where "potting" (that is, black boxing, such as the encap- sulation of products in epoxy resin to deter imitation) is said to have cone into prominence, and patents are claimed to be less important, the firms in our sample reported no such trend. This is important because it is the first systematic evidence concerning Me extent to which the reduction in Me patent rate during the 1970s was due to a shift away from patents and toward trade secrets and over forms of protection. If, as some responsible observers have claimed, "the so-called patent decline may be merely a patent bypass" (Shapley, 1978:848-~49), it is important that policymakers be aware tI}at this is the case. Based on these results, there is no indication Mat this is - true. PRICE INDEXES FOR R&D INPUTS Economists, policymakers, and analysts are interested in the changes over time in real R&D expenditure- that is, the changes in the amount of real resources devoted to R&D. To estimate such changes, it is necessary to have price indexes for R&D inputs. Unfortunately, Imt~l very recently no such price indexes existed. Official government statistics in the United States use the GNP deflator to deflate R&D expenditures. The relevant government
OCR for page 317
MICROECONOMICS OF TECHNOLOGICAL INNOVATION 317 agencies are well aware that the GNP deflator is only a rough approximation, but little has been known about the extent to which the results would change if price indexes for R&D inputs were constructed and used instead of the GNP deflator. In the late 1970s, Goldberg (1978) and Schankerman (1979) constructed price indexes for R&D inputs. These price indexes were based on the use of proxies (that is, series that were thought to be highly correlated with the relevant input prices). For example, Goldberg used data from the National Survey of Professional, Technical, and Clerical Pay to represent the changes over time in the level of wages for R&D engineers and scientists, and Schank- e~an used the Index of Cost of Matenals of the Bureau of Economic Analysis as a proxy for R&D materials prices. Although indexes based on proxies are interesting and useful, there are obvious advantages in constructing price indexes for R&D inputs based on data obtained from firms regarding actual prices and expenditures. Many observers have urged that such indexes be constructed. Recently, a study financed by the National Science Foundation was camed out along these lines, based on detailed data obtained from about 100 fins in 12 industries (Mansfield et al., 1983; Mansfield, 1985~. The study results indicate that, if one is interested in making short-term comparisons of total real R&D expenditure in the nation as a whole, the GNP deflator Is reasonably adequate. For example, in comparisons of successive years, the percentage change in real R&D expenditure based on the GNP deflator is generally within a per- centage point of that based on the R&D price index produced by the study. However, for long-term comparisons of national R&D expenditure, the use of the GNP deflator can result in substantial errors. Thus, whereas real R&D expenditure went up by 26 percent in the years 1969 to 1981, based on the GNP deflator, it really went up by only 15 percent, according to He R&D price Index. The reason why the GNP deflator performs worse in He long run than in the short run is that, for He vast majority of years for which we have data, it has tended to underestimate the rate of inflation in R&D. This problem is especially severe in particular industries. For example, based on the GNP deflator, real R&D expenditure in the chemical industry grew by about 42 percent during 1969-1981, but based on the R&D price index, it grew by only about 22 percent during this period. Similar errors occur in the oil, primary metals, fabricated metal products, rubber, automobile, instruments, and "other" industries. In all of these industries, the GNP deflator results in an overestimation of the 1969-1981 growth of real R&D expenditure of 15 percentage points or more. Given the obvious importance of the R&D figures to economic analysis in this area, the availability of these improved R&D price indexes should be a significant step forward. This is the sort of data improvement that tends
OCR for page 318
318 EDWIN MANSFIELD to be invisible to people who are not involved closely with empirical work. But such improvements can be important. THE DIFFUSION OF INNOVATIONS Whereas R&D puce indexes have attracted the attention of relatively few economists, the diffusion of innovations has been the focus of a considerable amount of work. Although we are far from having completely satisfactory models of the diffusion process, substantial progress has been made in this area. In general, the diffusion of a major new technique tends to be a slow process. For example, measuring from the date of first commercial appli- cation, it generally took more than 10 years for all of the major American firms in the bituminous coal, steel, railroad, and brewing industries to begin using a sample of important new techniques. (Among the innovations in- cluded in this sample were the shuttle car, trackless mobile loader, and continuous mining machine in the bituminous coal industry, and the by- product coke oven, continuous annealing, and continuous wide strip mill in the steel industry.) More recently, similar findings have been reported for the chemical and other industries. Also, Me rate of diffusion vanes widely. Sometimes it took decades for Fisons to install a new technique, but in other cases they imitate the innovator very quickly. To some extent, these differ- ences may reflect a tendency for the diffusion process to go on more rapidly in more recent times than in the past (Mansfield, 1961, 19681. Based on the available evidence, the rate of diffusion of an innovation depends on the average profitability of the innovation; the variation among firms in the profitability of He innovation; the size of He investment required to introduce the innovation; and the number of firms in the industry, their average size, the inequality in their sizes, and the amount that they spend on research and development. Using these variables, one can explain a large proportion of the variation among innovations in the rate of diffusion. More- over, this seems to be the case in a wide vaneW of industries and in other countries as well as in the United States. Econometric models using these variables seem to be useful devices for technological forecasting (Mansfield, 19776~. According to studies of a number of industries, fibs in which the expected returns from the innovation are highest tend to be quickest to introduce an innovation. Also, holding constant the profitability of the innovation, large firms tend to introduce an innovation more quickly than do small firms. In some industries, this may be due to the fact Hat larger firms although not necessarily the largest ones are more progressive than small fins. But even if the larger fimns were not more progressive, one would expect them to be quicker, on the average, to begin using a new technique, for reasons
OCR for page 319
MICROECONOMICS OF TECHNOLOGICAL INNOVATION 319 discussed elsewhere (Mansfield, 196Sb). Also, holding other factors con- stant, firms with younger and better-educated managers tend to be quicker to introduce new techniques—or at least, this seems to be the case in in- dustnes in which the relevant data have been collected. Companies also differ greatly with regard to the intrafirm rate of diffu- sion the rate at which, once it has begun to use the new technique, a firm substitutes it for older methods. A considerable amount of this variation can be explained by differences among firms in the profitability of the innovation, the size of the fin, and the fimn's liquidity. Also, there is a tendency for late starters to catch up. That is, firms that are slow to begin using an innovation tend to substitute it for older techniques more rapidly than those that are quick to begin using it. It is also relevant to note that the same sort of process occurs on the international scene: countries that are slow to begin using, an innovation tend to substitute it for older techniques more rapidly than countries that are quick to begin using it. The reasons for this tendency, both at the company and national levels, seem clear enough (Mansfield, 1968a; Nasbeth and Ray, 19741. Sociologists have studied the nature and sources of information obtained by managers concerning new techniques. The sources of information some- times vary depending on how close the manager is to adopting the inno- vation. For example, in agriculture, mass media are most important sources at the very early stages of a manager's awareness of the innovation, but friends and neighbors are most important sources when a manager is ready to try the innovation. Also, there is evidence of a ''two-step flow of communication." The early users of an innovation tend to rely on sources of information beyond their peer group's experience; after they have begun using the innovation, they become a model for their less expert peers, who can imitate their performance (Rogers, 1962, and subsequent pub- lications). Turning to other factors that also influence the rate of diffusion of an innovation,- the diffusion process may be slowed by bottlenecks in the pro- duction of Me innovation as in the case of the Boeing 707. Also, the extent of advertising and other promotional activities used by producers of the new product or equipment will have an effect. So, too, will the innovation's requirements with respect to knowledge and coordination. The diffusion process will be impeded if the innovation requires new kinds of knowledge on the part of the user, new types of behavior, and the coordinated efforts of a number of organizations. If an innovation requires few changes in sociocultural values and behavior patterns, it is likely to spread more rapidly. Also, the policies adopted by relevant labor unions influence the rate of diffusion. For example, some locals of the painters' union have refused to use the spray gun In addition, the users' willingness to take risks can have an important influence on an innovation's rate of diffusion. Nonetheless,
OCR for page 320
320 EDWIN MANSFIELD while the diffusion process is probably better understood than many other aspects of technological change, much more research is needed.7 INTERNATIONAL TEClINOLOGY TRANSFER One aspect of the diffusion process that has received considerable attention in recent years is international technology transfer. Economists have long recognized Nat the transfer of technology is at the heart of the process of economic growth, and that the progress of both developed and developing countries depends on the extent and efficiency of such transfer. In recent years, economists also have come to realize (or rediscover) the impact of international technology transfer on the size and patterns of world trade. The work of Hufbauer (1966), Tilton (1971), Schwartz, myself, and others indicates that technology is being transferred across national boundaries more rapidly than in the past. Based on a sample of chemical, semiconductor, and pharmaceutical innovations, this was found to be true even when a variety of other relevant factors were held constant. In considerable part, this is due to the growing influence of multinational Finns, many of which are heavily involved in transfemng technology. U.S.-based multinational fimns are trans- femng their technology to their foreign subsidiaries much more quickly than in the past. One study of technology diffusion found that in 1969 to 1978 about 75 percent of the technologies that were transferred to subsidiaries in developed countries were less than 5 years old, in 1960 to 196S, the proportion was about 27 percent (Mansfield and Romeo, 19801. Nations that spend relatively large amounts on R&D (in the relevant in- dustry) tend to be relatively quick to begin producing a new product, even if they are not the innovator. This finding is analogous to the finding (cited above) that firms that spend relatively large amounts on R&D tend to be quick adopters of new technology developed by others. Both for entire nations and individual firms, R&D provides a window on various parts of the en- vironment, and it enables the nation or firm to evaluate external developments and react more quickly to them. In some economic models, R&O is viewed as an invention-producing or innovation-producin~, activity. While correct as far as it goes, this view misses much of the point of R&D, which is that it is also aimed at a quick response to rivals and at clever modification, adaptation, and improvement of their results. In many important industries, like pharmaceuticals, international technol- ogy transfer is being promoted by the fact that companies have been carrying on increasing shares of their R&D overseas. About 10 percent of the R&D camed out by U.S. firms is performed outside die United States. In some industries, again like pharmaceuticals, this percentage is much larger. When compared with me total R&D expenditures in various host countnes, the size of overseas R&D is perhaps even more striking. In the early 1970s, about
OCR for page 321
MICROECONOMICS OF TECHNOLOGICAL INNOVATION 321 one-half of the industrial R&D performed in Canada and about one-seventh of the industrial R&D performed in the United Kingdom and West Ge~any was done by U.S.-based films (Mansfield, 19841. International technology transfer has also been promoted by the fact that, in many areas, the process of innovation has been internationalized. For example, in the pharmaceutical industry, a new ding is no longer discovered, tested, and commercialized, all within a single country. Instead, the discovery phase often involves collaboration among laboratories and researchers located in several different countries, even when they are within the same firm. And clinical testing generally becomes a multicoun~y project. Even in the later phases of drug development, such as dosage formulation, work often is done in more Man one country. In contrast, older economic models of He process of international technology transfer often tended to assume that innovations were camed out in a single country, generally the United States, and Hat the technology resided exclusively within that country for a considerable period after the innovation's initial commercial introduction. Economists are in the process of replacing these models with others that conform more closely to current conditions. Etht;CTS ON OTHER COUNTRIES OF THE OIJTELOW OF U.S. TECHNOLOGY As pointed out earlier, technology transfer lies at He heart of the process of economic development. Innovations are primarily responsible for many increases in output per capita. How rapidly innovations spread and thus raise per capita output in countries other than the innovating nation depends on the process of technology transfer. The multinational firm is, of course, a major agent in the process of international technology transfer, but its role is highly controversial. Many host countries, although eager for modern technology, are suspicious of the activities of multinational firms. One of the most important unanswered questions concerning He transfer of technology via the multinational firm is: How big have the economic benefits to countries outside the United States been from technology transfer of this sort? Put somewhat more precisely, bow much lower would total output outside the United States have been if technology transfer of this sort had not occurred, but if He relevant technology and goods were perhaps available from He United States or elsewhere? Economists have begun to assemble some quantitative evidence bearing on this question. Based on the limited data available, it has been suggested Hat total annual output of countries outside He United States would have been at least $35 billion, or at least 1 percent, less if technology transfer of this sort had not occurred (Mansfield and Romeo, 19801. (Note Hat this estimate is a lower bound.) However, much more work is required in this area. International technology
OCR for page 322
322 EDWIN MA~SFIEI~:) transfer, like so many other aspects of technological change, is very imper- fectly understood. CONCLUSIONS Microeconomists have been making steady pro press in the 30 years that have elapsed since the economics profession began to direct a substantial proportion of its energies and resources to the study of technological change. About 15 years ago, I was asked by the National Science Foundation to write a report (which subsequently appeared in Science) describing what we thought we knew in this area and what types of research needed to be done (Mansfield, 19721. It is gratifying to say that significant progress has been made on practically all of the topics identified there as being in great need of fumier work. In a chapter of this length, it is impossible to do more than sample some of the many kinds of research that microeconomists have carried out in the past 30 years. A wide variety of econometric models, empirical investiga- tions, computer simulations, and other kinds of output has resulted from the extensive work (largely financed by the National Science Foundation) that has been done. I have made no attempt to cover all, or even most, of the relevant results. Among the interesting contributions in recent years have been Paul David's (1975) diffusion studies, George Eads's (e.g., 1980) policy research, Zvi Gnliches's (1958, 1964, 1980, 1984) and Dale Jorgenson's econometric investigations (e.g., Gollop and Jorgenson, 1980), the theoret- ical work of Kamien and Schwartz (1982), the dynamic models of Burton Klein (1977), the computer simulation models of Nelson and Winter (1982), the empirical studies of Pavitt and his associates at Sussex University's Science Policy Research Unit (e.g., 1974), Nathan Rosenberg's historical investigations (1982), Richard Levin's (1984) and F. M. Scherer's statistical studies (e.g., 1982), and Vernon Ruttan's innovation models (e.g., Evenson et al., 19791. Many of these investigators are contributors to this volume, so I have not attempted to summarize their results, since they obviously have both an absolute and comparative advantage in this regard.8 Finally, although we clearly know a great deal more about Me economics of technological change than we did 30 years ago, it is evident that we are still very far from a satisfactory understanding of the innovation process, the determinants of Me rate of innovation, the measurement of Me rate and direction of technological change, and Me effects of changes in technology. In view of the enormous difficulties Mat studies in these areas face, it is not surprising that existing knowledge remains limited. The extent to which economists are able and willing to work with, and learn from, technologists and scientists may play a significant role in deter- mining how successful we are in chipping away at the many perplexing
OCR for page 323
MICROECONOMICS OF TECHNOLOGICAL INNOVATION 323 problems that remain To my mind, economists frequently have been far too insular and divorced from technological realities. This volume is a very welcome step in the right direction. I hope that it will set Me stage for more extensive and effective collaboration between economists and technologists in attempts to deal win We many fundamental problems in this area that concern us all. NOTES 1. Also, he found that the returns from R&D seemed to be lower in industries in which much R&D is federally funded. 2. Many of these studies are con~ned in or referred to in Griliches (1984). 3. Much of this section and the next two sections draw heavily on Mansfield et al. (1981). 4. The empirical results presented in this section are largely from Mansfield et al. (1981). 5. A logit analysis is a statistical technique used to estimate the effects of independent variables on a dependent variable that assumes values of zero or one. 6. Blac}cmar~ (1971) and others have found this model useful in their forecasting studies, and Hsia (1973) and others have found it useful in studies of countries other than the United States. ;'. Models devised by economists to represent the diffusion process have been used (with some success) by a variety of funs and government agencies. Although the applicability and power of such models should not be exaggerated, they have proved to be reasonably helpful, if crude, devices, when used with proper caution. 8. This list does not include many of the major figures in this field, let alone the promising younger economists on which the future development of the field will largely depend. Several dozen economists are currently doing interesting work in this field; I have named primarily those who participated in the Symposium on Economics and Technology held at Stanford University, March 17-19, 1985. REFERENCES Arrow, K. 1962. Economic welfare and the allocation of resources for invention. In National Bureau of Economic Research, The Rate and Direction of Inventive Activity. New York. Blackman, A. 1971. The rate of innovation in the commercial aircraft jet engine market. Techno- logical Forecasting and Social Change. David, P. 1975. Technical Choice, Innovation, and Economic Growth. New York: Cambridge IJniversity Press. Eads, G. 1980. Regulation and technical change: Some unexplored issues. American Economic Review. Evensong R. 1968. The Contribution of Agricultural Research and Extension to Agnculn~al Pro- duction. Ph.D. dissertation, University of Chicago. Evenson, R.. P. Waggoner and V. Ruttan. 1979. Economic benefits from research: An example from agriculture. Science. September 14, 1979. Foster Associates. 1978. A Survey on Net Pates of Return on Innovations. Washington, D.C.: National Science Foundation. Goldberg, L. 1978. Federal policies affecting industrial research and development. Presented at Me meetings of the Southern Economic Association, November 9, 1978. Gollop, F.' and D. Jorgenson, 1980. U.S. productivity growth by industry, 1947-73. In J. Kendrick and B. Vaccara, eds., flew Developments in Productivity Measurement and Analysis. Chicago: University of Chicago Press.
OCR for page 324
324 EDWIN MANSFIELD Gnliches, Z. 1958. Research COStS and social retains: Hybnd corn and related innovations. Journal of Political Economy. Gnlic}~es, Z. 1964. Research expenditures, education, and the aggregate agricultural production function. American Economic Review. Gnliches, Z. 1980. Retunes to research and development expenditures in the private sector. In J. Kendnck and B. Vaccara, eds., New Developments in Productiviry Measurement and Analysis. Chicago: University of Chicago Press. Gnliches, Z.. ed. 1984. R&D, Patents, and Productivity. Chicago: University of Chicago Press. Hsia. R. 1973. Technological change in the industrial growth of Hong Kong. In B. Williams, ea., Science and Technology in Economic Growth. New York: Macmillan. Hufbauer, G. 1966. Synthetic Materials and the Theory of International Trade. Cambndge, Mass.: Harvard University. Kamien, M., and N. Schwartz. 1982. Marker Structure and Innovation. New York: Cambridge University Press. Klein, B. 1977. Dynamic Economics. Cambridge, Mass.: Harvard University Press. Levin, R., and P. Reiss. 1984. Tests of a Schumpeterean model of R&D and market structure. In Z. Griliches, ea.. R&D, Patents, and Productiviry. Chicago: University of Chicago Press. Mansfield, E. 1961. Technical change and the rate of innovation. Econometrical Mansfield, E. 1965. Rates of return from industrial research and development. American Economic Review. Mansfield, E. 1968a. The Economics of Technological Change. New York: W. W. Norton. Mansfield, E. 1968b. Industrial Research and Technological innovation. New York: W. W. Norton for He Cowles Foundation for Research in Economics at Yale University. Mansfield, E. 1972. Contribution of R&D tO economic growth in the United States. Science, February 4, 1972. Mansfield, E. 1980. Basic research and productivity increase in manufactunag. American Economic Review. Mansfield, E. 1981. Composition of R&D expenditures: Relationship to size of find, concentration, and innovative output. Review of Economics and Statistics. Mansfield. E. 1984. Technological change and the international diffusion of technology. To be published by the Royal Commission on the Economic Union and Development Prospects for Canada. Mansfield, E. 1985a. Patents and Innovation: An Empirical Study. Philadelphia: University of Pennsylvania. Mansfield, E. 1985b. Price Indexes for R&D Inputs, 1969-83. Philadelphia: University of Penn- sylvania. Mansfield, E., and A. Romeo. 1980. Technology transfer to overseas subsidiaries by U.S.-based fimns. Quarterly Journal of Economics, December. Mansfield, E., et al. 1971. Research and innovation in the Modern Corporation. New York: W. W. Norton. Mansfield, E., et al. 1977a. The Production and Application of New Industrial Technology. New York: W. W. Norton. Mansfield, E., et al. 1977b. Social and private rates of return from industrial innovations. Quarterly Journal of Economics. Mansfield. E., M. Schwartz, and S. Wagner. 1981. Imitation COStS and patents: An empirical study. Economic Journal. Mansfield, E., et al. 1982. Technology Transfer, Productivity, and Economic Policy. New York: W. W. Norton. Mansfield, E., A. Romeo, and L. Switzer. 1983. R&D price indexes and real R&D expenditures in the United States. Research Policy.
OCR for page 325
MICROECONOMICS OF TECHNOLOGICAL INNOVATION 325 Minasian. J. 1969. Research and development, production functions. and rates of retune. American Economic Review. Nabseth. L.. and G. Ray. 1974. The Diffusion of New Industrial Processes. London: Cambridge University Press. Nadin, ME and G. Bitros. 1980. Research and development expenditures and labor productivity at the find level. In J. Kendrick and B. Vaccara, eds., New Developments in Productivity Mea- surement and Analysis. Chicago: University of Chicago Press. Robert R. Nathan Associates. 1978. Net Rates of Return on Innovations. Washington, D.C.: National Science Foundation. Nelson. R., and S. Winter. 1982. An Evolutionarv Theory of Economic Change. Cambridge. Mass.: Belknap. Pavitt. K. 1974. In National Science Foundation. The Effects.of International Technology Transfers on U.S. Economy, Washington, D.C.: Government Printing Office. Peterson, W. 1971. The returns to investment in agricultural research in the United States. In Resource Allocation in Agricultural Research. Minneapolis: University of Minnesota. Rogers, E. 1962. Diffusion of Innovations. New York: The Free Press. Rosenberg, N. 1982. Inside the Black Box: Technology and Economics. New York: Cambridge University Press. Schankerrnan, M. 1979. Essays in the Economics of Technical Change. Ph.D. dissertation. Harvard University. Scherer, F. M. 1982. Inter-indusby technology flows and productivity growth. Review of Economics and Statistics. Schmitz, A.. and D. Seckler. 1980. Mechanized agriculture and social welfare: The case of the tomato harvester. American Journal of Agricultural Economics. Schumpeter, J. 1934. The Theory of Economic Development. Cambridge. Mass.: Harvard University Press. Shapley, D. 1978. Electronics industry takes to potting the products for market. Science, November 24: 848-849. Taylor, C.. and Z. Silberston. 1973. The Economic Impact of the Patent System. New York: Cambridge University Press. Terlec}cyj, N. 1974. Effects of }~&D on the Productivity Growth of Industries. Washington, D.C.: National Planning Association. Tilton, J. 1971. International Diffusion of Technology: The Case of Semiconductors. Washington, D.C.: The Brookings Institution.
OCR for page 326
Representative terms from entire chapter: