|
Items in cart [1] |
Questions? Call 888-624-8373 |
| Copyright © 2009. National Academy of Sciences. All rights reserved. Terms of Use and Privacy Statement |
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 24
The Sources and Rate of Technological
Change in the U.S. Economy
Technological change is often difficult to predict, and its employment
and productivity consequences usually are felt gradually rather than
suddenly. Although the pace of technological change affects employment
and productivity growth, the impact of new technologies also is affected
heavily by organizational, institutional, and social factors. A central
reason for the complex, gradual character of the employment, productiv-
ity, and other economic effects of such change is that these impacts are
felt only through the adoption of new technologies by individuals and
firms. In light of this fact, we devote considerable attention in this chapter
to the process of adopting new technologies.
DEFINING TECHNOLOGICAL CHANGE
Technological change has two major effects: (1) it transforms the
processes by which inputs (including labor and materials) are converted
into goods and services, and (2) it enables the production of entirely
new goods and services. Process innovation is technological change that
improves the efficiency with which inputs are transformed into out-
puts; product innovation results in the production of new goods. The
distinction between process and product innovation often is hazy.
New products, such as the transistor, frequently require significant
process innovations before they can be produced economically. Con-
versely, the potential cost reductions offered by many new manufactur-
ing processes may be realized only after the products to which they are
24
OCR for page 25
SOURCES AND RATE OF TECHNOLOGICAL CHANGE 25
applied have been redesigned. In addition, an innovative product developed
by one firm for example, computer numerically controlled machine tools-
may be transformed into a process innovation when it is adopted by another
firm.
Product innovations may serve entirely new markets; consequently,
their effects are notoriously unpredictable. (Chapter 4 explores the
uncertainty surrounding predictions of the employment consequences of
process innovations.) Repeatedly, technological forecasts have failed to
foresee the size and nature of the markets for new products. Computers
are a classic example. Howard Aiken, one of the developers of the
electronic computer in the 1940s, was skeptical about the plans of J.
Presper Eckert and John Mauchly to launch commercial computer
production; Aiken predicted that the total U.S. market would be no more
than four or five machines. Internal IBM studies conducted prior to the
firm's decision to begin computer production were equally pessimistic;
according to the studies, the market for the "tape processing machine"
would amount to roughly 25 units (Ceruzzi, 19861. The record of techno-
logical advance contains many such examples (Rosenberg, 1983~.
Invention, Innovation, and Diffusion
The history of scientific discoveries like penicillin or x rays contributes
to a popular perception that technological change is a process of dramatic
breakthroughs. In fact, it might better be described as incremental and
consisting of several stages, extending well beyond the moment of
scientific discovery. The invention stage includes the discovery of a
scientific or technological advance and its translation into a prototype-
for example, a working model. Invention, which subsumes basic re-
search, must be distinguished from innovation, which includes the
processes of advanced development (e.g., "scaling up" a pilot plant for
commercial-volume production). In the case of the transistor, an impor-
tant product innovation that has been fundamental to modern technolog-
ical advance, invention spanned the period from the late 1930s, when Bell
Telephone Laboratories inaugurated its program of basic research in
solid-state physics, through 1947, when the first model of a point-contact
transistor was produced by Bardeen, Brattain, and Schockley (see Braun
and MacDonald, 1978; Mowery, 1983; Nelson, 1962; Tilton, 1971~. The
innovation stage that saw the translation of this crude invention into a
commercially marketable product occurred during 1947-1954. This stage
included significant advances in the theory of semiconductors and in
materials refining and processing. Advances in both the theory of mate-
rials and in production techniques for making pure silicon crystals
OCR for page 26
26 TECHNOLOGY AND EMPLOYMENT
contributed to the introduction by Texas Instruments of the silicon
junction transistor in 1954.
The diffusion of an innovation (discussed in detail later in this chap-
ter) refers to the period of its adoption by users. Once again using
transistors as an example, the diffusion that began once the product was
introduced commercially in 1954 has continued to the present day;
moreover, during this period, transistors have undergone considerable
modification in design and production. Chaudhari (1986) described the
dramatic advances since 1960 in the miniaturization of transistor com-
ponents, focusing on the shrinkage in the width of "lines" that connect
the transistor to other electronic components: "A typical line width in
1960 was 30 micrometers.... Today line widths are commonly on the
order of one micrometer. . ." (p. 1371. Among other significant ad-
vances during the diffusion stage was the development of the planar
process for manufacturing integrated circuits and other solid-state
components.
Each of these stages invention, innovation, and diffusion consists
of a series of interacting phases; within the invention stage, for example,
basic research often is heavily influenced by applied research findings
(see Kline and Rosenberg, 1986; Rosenberg, 1983~. Moreover, the
invention, innovation, and diffusion processes themselves are linked in
a complex fashion, which can be seen in the extensive modifications that
are often made to an innovation during its diffusion. In the case of the
transistor, the innovation stage of its development required fundamental
research, just as its application to new uses during the diffusion stage
has required investments in applications engineering and fundamental
research.
Influences on Invention, Innovation, and Diffusion
Despite the close links among them, the invention, innovation, and
diffusion stages of a technology appear to respond to different influences
that are not always easy to distinguish. In the case of invention, for
example, the factors affecting individual genius simply are not well
understood. These stages also may be carried out by different individuals
or organizations. In many instances, the inventor of a new product or
process does not develop and market it. The original inventors of the
computer, for example, were not employees of the firm that proved most
successful at developing, improving, and marketing the device. Another
case is that of DuPont. Many of the most significant innovations com-
mercialized by DuPont prior to the invention of nylon during the 1920s
and 1930s were based on patents purchased from other firms and
individuals, rather than on the inventions of its employees.
OCR for page 27
SOURCES AND RATE OF TECHNOLOGICAL CHANGE 27
Compared with invention, innovation is a far more costly stage of
technological change. ~ It is likely therefore to be affected by such
economic factors as the investment climate, rates of capital formation, or
expectations of the location and size of future markets. The diffusion of
innovations, which is discussed later in this chapter, appears to be
influenced by cost considerations, uncertainty, and other factors unique
to specific markets, such as regulations that affect the structure of the
market for an innovation. For example, the regulation of pharmaceuticals
and air transportation and the availability of third-party payments for
medical services have affected the speed and the extent of new technol-
ogy diffusion in those industries. Moreover, inasmuch as the diffusion of
new technology is the result of decisions to invest in machinery or
products that embody a technology, the rate of diffusion of innovations is
affected by factors that determine the rates of net investment within an
economy, including the domestic savings rate, the cost of capital,
depreciation practices, and price stability.
The Interaction of Technological and
Organizational Change
Technological change creates new options for the performance of
specific functions. Yet the precise organization of these functions or the
skill requirements associated with them are seldom determined solely by
the characteristics of the technology. Organizational factors strongly
influence the implementation of new technologies and their effects on skill
requirements, quality of worklife, productivity, and profitability. Indeed,
the potential improvements offered by many innovations often can be
realized only if there are complementary organizational changes.
For example, redesigning products often allows more profitable use of
many new computer-based manufacturing technologies. After installing
equipment for computer-integrated production of lawn and garden trac-
tors, Deere and Company realized substantial savings by redesigning
its products to allow a single component design to be used in eight
different tractors.2 Other firms have redesigned their products for easier
automated assembly; a recent IBM desktop printer has been so simplified
for automated assembly that it can be manually assembled in minutes.
"'Development" alone, which is the portion of innovation incorporating most of the
activities of production engineering and tooling, typically accounted for more than 65
percent of privately financed U.S. R&D investments annually during 1960-1985 (National
Science Foundation, 1985, Appendix Tables 2-3 and 2-9).
2Remarks by G. R. Sutherland of Deere and Company at a meeting of the Panel on
Technology and Employment, April 25, 1986.
OCR for page 28
28 TECHNOLOG Y AND EMPLO YMENT
(Lehnerd, 1987, discusses similar changes in the design of Black and
Decker power tools.)
Such integration of production engineering and product design often
demands extensive organizational change. The National Research
Council's Committee on the Effective Implementation of Advanced
Manufacturing Technology (1986) noted in its report that computer-
integrated manufacturing (CIM) requires that "Edjecisions once made by
people in functions that were relatively independent must now be made
jointly. Efforts to design the product and process simultaneously, for
example, require product engineering and manufacturing engineering to
work closely together" (p. 29~. Although CIM has not yet been widely
adopted in U.S. manufacturing, its requirements for organizational adap-
tation are by no means unique. A number of other computer-aided
manufacturing technologies impose similar organizational demands.
In many cases, once a new technology has been adopted, the resulting
improvements in the quality of a firm's manufactures and its productivity
come as much from the reorganization of production and other activities
required by the adoption as they do from the technologies themselves.
For example, management personnel interviewed by panel members and
staff in the course of this study argued that the organizational changes
necessary to adopt computer-aided manufacturing processes yielded
savings as great as those realized from this new production technology
itself (the IBM printer described previously is one example). In most
cases, these organizational changes were necessary to introduce com-
puter-aided technologies. The converse was not true, however the
reorganization of design, engineering, and production processes did not
require new technologies.
The value and importance of attention to the organizational dimensions
of technological change, then, cannot be overstated. Indeed, without such
attention, the potential profitability or product quality benefits of new
technology may not be realized. Prior to the extensive use of advanced
computer-aided or computer-integrated manufacturing technologies, Jap-
anese automotive firms, for example, achieved great advances in produc-
tivity and product quality mainly through organizational techniques. The
best-known of these successes, the Toyota production system, was
developed during the 1960s and 1970s, prior to the development of CIM
and robotics; it used production technologies that did not differ signifi-
cantly from those of U.S. automobile manufacturers at the time (Abeg-
glen and Stalk, 19861.
Within the U.S. automotive industry, General Motors (GM) offers
dramatic plant-level contrasts in productivity and product quality that
illustrate the importance of organizational factors in realizing the potential
of new technology. In Fremont, California, the joint venture between GM
OCR for page 29
SOURCES AND RATE OF TECHNOLOGICAL CHANGE 29
and Toyota (known as New United Motor Manufacturing, Inc.
NUMMI) uses modest levels of factory automation that are embedded in
the Toyota production system manned by a unionized work force; thus
far, NUMMI has been extraordinarily successful in meeting production
and quality targets. By contrast, GM's factory in Hamtramck, Michigan,
which uses advanced factory automation technologies, operates at
roughly 50 percent of its planned capacity and has experienced serious
quality problems (Nag, 1986; Womack, in press).
Qualitative and anecdotal evidence suggests that, in the past, U.S.
management and labor have been insufficiently attentive to the need to
reorganize design and work processes to support technological change.
Jaikumar (1986) presented data that illustrate this point in his analysis of
35 "flexible" manufacturing systems (i.e., systems that use computer-
aided machinery and "work cells" to produce a wide variety of products
at low cost) in the United States and 60 such installations in Japan. He
concluded that:
Rather than narrowing the competitive gap with Japan, the technology of
automation is widening it further....
With few exceptions, the flexible manufacturing systems installed in the United
States show an astonishing lack of flexibility. In many cases, they perform worse
than the conventional technology they replace. The technology itself is not to
blame; it is management that makes the difference. Compared with Japanese
systems, those in the U.S. plants produce an order-of-magnitude less variety of
parts. Furthermore, they cannot run untended for a whole shift, are not integrated
with the rest of their factories, and are less reliable. Even the good ones form, at
best, a small oasis in a desert of mediocrity. (p. 69)
U.S. managers and workers must understand that the "rules of the
game" of international competition and technology's role within that
competition have changed. Automation and firm and factory reorgani-
zation are means to the end of higher-quality, lower-cost products.
Achieving this goal requires attention to production technology, product
design, and work organization. Without such attention, the payoffs from
the adoption of new technologies will be realized slowly or not at all.
Measuring Technological Change
If we could measure the rates of invention, innovation, and diffusion in
the U.S. economy, we could simplify greatly the analysis of technology's
impact on employment. Such measurements, however, are far from
simple. The United States and most other industrial nations do not collect
systematic time series data on the rates of diffusion of specific new
technologies. As a result, there are few reliable data or indices with which
OCR for page 30
30 TECHNOLOGY AND EMPLOYMENT
to measure such rates. Measuring the rates of invention or innovation also
is hampered by the fact that the outputs of these stages are extraordinarily
difficult to measure. Those indices that have been used-the number of
patents, publications, or expert tabulations of technologically and com-
mercially significant innovations have serious shortcomings. For exam-
ple, a widely used gauge of inventive or innovative activity, R&D
investment, measures only inputs into invention and innovation, rather
than outputs. Without output measures, we cannot assess the efficiency
with which investments in science and technology are translated into
inventions or innovations. A further inadequacy of R&D investment as a
measure is that it includes development expenditures; such expenditures
affect both invention and innovation, as well as diffusion, and thus do not
allow for separate measurement of these stages.
Other commonly used proxies for the rate of technological change
include increases in the joint productivity of capital and labor- that is,
"total" or "multifactor" productivity growth. Multifactor productivity
growth measures improvements in the efficiency with which inputs are
translated into outputs and thus should be responsive to changes in the
rates of new technology generation and adoption. As a gauge of techno-
logical change, however, this index has several defects. Empirically,
multifactor productivity growth is derived as a residual that is, after
adjusting for contributions made to greater output by increases in the
quality and quantity of capital and labor. As a residual, it is a measure of
ignorance, an index of the contributions to output growth of unmeasured
influences rather than a direct measure of technological change. In
addition, like all productivity indexes, measures of multifactor produc-
tivity are sensitive to fluctuations in the level of economic activity. To
reduce the influence of such fluctuations, multifactor productivity growth
typically is measured across business cycles. An alternative productivity
measure that does not account for improvements in the productivity of
capital inputs is labor productivity growth, measured as growth in output
per hour. During most of the postwar period, these two measures have
exhibited similar trends; since 1973 rates of growth in both labor and
multifactor productivity have been much lower than in the 1950s and
1960s (Gullickson and Harper, 1986~.
Using productivity growth as an index of the rate of technological
change has other drawbacks. Many factors other than technology influ-
ence investment and diffusion, the processes that underpin productivity
advance. The low savings rate in the United States, for example, may
increase the cost of capital to private firms, thus lowering net investment
and impeding diffusion and productivity growth. Furthermore, in measur-
ing productivity change, it is often difficult to adjust measures of physical
output for changes in the quality of products. Should a modern computer
OCR for page 31
SOURCES AND RATE OF TECHNOLOGICAL CHANGE 31
be treated as identical in quality to the machines of the mid-1950s? In
theory, quality adjustments should be made frequently, but the data
requirements for such a task are so great that until recently, the output
data compiled by the Bureau of Economic Analysis (BEA) of the U.S.
Department of Commerce, which historically have been used by the
Bureau of Labor Statistics (BLS) for productivity measurement, did not
incorporate adjustments for improvements in product quality. In 1985
BEA developed a computer price index that adjusted computers for
quality; the new index resulted in dramatic declines in the estimated costs
of such equipment after 1972, and it also will affect measured productivity
growth for this period (Cole et al., 1986; Slater, 19861. This is merely one
example of the complexity involved in analyzing and measuring the
relationship between technological change and productivity growth.
Much of the current concern over the effects of technological change on
employment is based on the belief that the rate of such change whether
it is defined as innovation or diffusion has increased in recent years.
Although specific technologies (e.g., office automation) may be experi-
encing more rapid change or diffusion now than in the past, aggregate
indicators suggest that there has been no across-the-board increase in the
rates of innovation or diffusion of technologies. The rate of growth in the
number of patents granted within the United States (i.e., the number of
inventions deemed novel and therefore patentable by the U.S. Patent
Office) was lower during the early 1980s than during the late 1960s.3 The
average annual rate of growth in patent grants was 3.7 percent during
1965-1970, -0.1 percent during 1970-1975, 0.1 percent during 1975-1980,
and 1 percent during 1980-1984 (National Science Foundation, 1985,
Table 4-81. In another study, Baily (1986) examined technological change
in several industries, including the research-intensive chemicals industry,
and concluded that innovation actually may have slowed in these indus-
tries in the past decade, resulting in lower rates of productivity growth.
Measures of diffusion rates are, if anything, even more difficult to
obtain than measures of the rates of invention or innovation. What work
there has been in this area lends support to the conclusion that diffusion
rates are not increasing. Mansfield (1966), for example, found little or no
support for the hypothesis that the rate of diffusion had increased during
the post-World War II period. The National Research Council's Panel on
Technology and Women's Employment (1986) also expressed skepticism
about the claim that diffusion rates of information technologies are likely
to increase: "In the panel's judgment, diffusion will not accelerate over
3To avoid deceptive, short-run fluctuations as a result of changes in the length of time
required to process patent applications, patents were dated by the year in which they were
applied for rather than the year in which they were granted.
OCR for page 32
32 TECHNOLOGYAND EMPLOYMENT
the next ten years: deliberate rather than headlong speed seems likely"
(p. 641. As noted previously, measures of multifactor or labor productiv-
ity growth, which incorporate the impacts of changes in rates of innova-
tion and diffusion, also have been lower since 1973. (See Chapter 3 for an
extended discussion of productivity.)
The rates of invention, innovation, and diffusion within the U.S.
economy thus do not appear to have increased during the past two
decades. Nevertheless, the widely perceived increases in the employ-
ment-displacing effects of technological change on the U.S. economy,
which have generated increased concern over the employment impacts of
technological change, may reflect shifts in the geographic location of
innovative activity.
For much of the 1950s and 1960s, the United States commanded a
considerable technological lead over European industrial nations and Japan.
Since then, the technological dominance of the United States has declined
somewhat (see "The Diffusion of Technology" later in this chapter). Foreign
governments and enterprises now are important sources of new technology
as well as leaders in its adoption (see the next section). As a result, there is
an increased likelihood that innovation and diffusion will occur either initially
or more rapidly in other countries, enhancing the competitiveness of foreign
producers. As the sources of new technologies and the location of their initial
application continue to broaden internationally, the displacement of U.S.
workers due to more rapid foreign technology adoption or innovation may
occur more frequently and more quickly- although there may be no change
in the underlying worldwide rate of innovation. Moreover, the pace at which
technologies are transferred within the international economy and thus
become available to foreign firms now appears to be more rapid than in
previous decades (Abramovitz, 1986; Baumol, 1986; Mansfield and Romeo,
1980; Organisation for Economic Co-operation and Development, 19791.
Indeed, Baumol suggests that the increased speed of international technol-
ogy transfer is partly responsible for convergence in productivity growth
rates among industrialized nations.
SOURCES OF TECHNOLOGICAL CHANGE
Although individual inventors continue to play a role in the U.S.
innovation system, their importance as a source of new technology has
declined considerably over the course of this century (Schmookler, 1957~.
Broadly speaking, there are now three main sources of U.S. technological
change that is, three sources of financial support for the development
and application of new technologies within the U.S. economy: (1)
industrially financed R&D; (2) R&D financed by the federal government and
performed in industry, university, and government laboratories; and (3)
OCR for page 33
SOURCES AND RATE OF TECHNOLOGICAL CHANGE 33
55
In 50
-
c,
Cal
-
o
In
o
._
=
._
m
-
~n
LU
fir:
cat
is
x
45
40
35
30
25
20
15, .
O .
1 960
f
1965 1970 1975
YEAR
1980 1 985-86
Estimates
FIGURE 2-1 Industry expenditures on R&D, 1960-1986. SOURCE: National Science
Foundation (1985, 1986a).
foreign R&D, both privately and publicly funded. The relative importance of
these three sources has shifted over time and changed substantially during
the postwar period. Two significant changes since the 1960s include reduc-
tions in the importance of federally financed defense R&D for commercial
innovation and an increase in the amount of foreign R&D.
Industrially Funded Research and Development
in the United States
A large share (3~50 percent during the postwar period) of the total U.S.
R&D investment is industrial research expenditures (National Science
Foundation, 19851. Figure 2-1 depicts trends (in 1982 dollars) during
1960-1986 in industrially funded R&D.After growing throughout the 1960s
at an annual rate of more than 6 percent, industrial R&D spending scarcely
grew at all during the early 1970s; after 1975 it began to climb again.4
4The deflator (i.e., the index used to convert these figures into 1982 dollars, which is the
implicit gross national product deflator) used in Figure 2-1 may understate growth in the costs
of R&D somewhat (Mansfield, 1984). This means that some of the apparent rebound in real
R&D spending after 1974 may be illusory. In addition, as Cordes (1986) notes, industrial R&D
spending as a share of sales declined from 1970 to 1978; after 1978, growth resumed.
OCR for page 34
34 TECHNOLOGY AND EMPLOYMENT
Empirical research suggests that industrially funded R&D yields
significant improvements in productivity. Mansfield's (1972) summary of
a number of industry studies concluded that productivity growth was
directly related to the level of R&D investment. Griliches (1985)
conducted a statistical analysis of a large sample of firms, concluding
that higher levels of privately financed R&D were associated with higher
rates of productivity growth. Mansfield (1980b) found that the share of
"long-term" or basic R&D within privately financed R&D was associ-
ated positively with productivity growth within both industries and
firms. (See also Mansfield, 1980a, for a summary of this research.) These
and other studies suggest that the benefits of R&D investment are
realized only after a lag of 3-6 years (the lag is greater for basic research
investments), which reflects the length of time needed to embody R&D
results in innovations and market or adopt the innovations. Thus, the
detrimental effects of the slowdown in industrial R&D spending during
the early 1970s have been felt within the past S-10 years; the benefits of
the renewed growth in R&D investment after 1975 have probably been
realized only since 1980.
Neither the slowdown in industrial R&D investment during the early
1970s nor its resurgence in the late 1970s and early 1980s have been
satisfactorily explained. For example, there is little evidence that the
lower U.S. R&D investment of the early 1970s was the result of less
favorable tax treatment. Neither can we explain the recent resurgence of
growth in R&D investment by the more lenient treatment of R&D under
the tax code; the resurgence in R&D investment substantially predates
the passage of the R&D tax credit in 1981. (See Cordes, 1986, for a
summary of the evidence for these conclusions.)
Clearly, the recent recovery in U.S. R&D growth is a positive eco-
nomic development, but when measured as a share of the gross national
product (GNP), privately financed U.S. R&D lags behind that of such
nations as Japan and West Germany. In 1984, the last year for which
comparable data are available, the GNP shares for industry-financed
R&D were 1.3, 1.S, and 1.7 percent, respectively, in the United States,
West Germany, and Japan. For the GNP share of privately financed U.S.
R&D to match the GNP share of privately financed Japanese R&D
investment, U.S. industry would have to increase its 1984 R&D spending
(roughly $49 billion) by approximately $15 billion more than 30 percent
of privately financed U.S. R&D in 1984 (National Science Foundation,
1986a). Although some scholars (e.g., Brooks, 198S) have criticized the
use of GNP shares as a basis for comparing national R&D investments,
this measure captures the concept of R&D investment as a necessary cost
of competing in the modern world economy as a developer or adopter of
new technologies. In contrast to its competitors, U.S. industry appears to
OCR for page 40
40 TECHNOLOG Y AND EMPLO YMENT
2.75
2.50
~ 225
Z
O 2.00
A:
As
Do:
CL
1.75
West Germany ~
~United States
Japan ~
1.00
' United Kingdom
France
1985 U.S. NONDEFENSE R&D
EXPENDITURES = $74.6 BILLION
1 1
1971 1975 1980 1985 1986
YEAR
(est.)
FIGURE 2-5 Nondefense R&D expenditures as a percentage of gross national product
(GNP) by country. SOURCE: National Science Foundation (1987).
Although R&D investment is an important source of technological
innovation, as the previous discussion noted, the firm or nation under-
taking such investment does not always receive a majority or (in some
cases) any of the profits from its investment. As scientific and technical
data and research results spread throughout the world more quickly, the
ability of a single firm or nation to "appropriate" all the financial or
competitive fruits of its R&D investment has declined. Sustained support
for the generation of new knowledge remains critically important in the
current world economic environment. What is now of equal importance,
however, is the ability of a firm to move rapidly from invention to
commercial application and the ability of a national economy to adopt
new technologies quickly, thus narrowing the gap between current and
"best" practices. R&D investment positively influences the adoption and
rapid exploitation of new technologies; these activities are discussed in
the next section.
THE DIFFUSION OF TECHNOLOGY
The economic effects of new technology, whether revealed in produc-
tivity growth, creation or loss of jobs, or changes in wages and profits, are
realized only through its adoption. Therefore, no analysis of the effects of
OCR for page 41
SOURCES AND RATE OF TECHNOLOGICAL CHANGE 41
new manufacturing and office technologies on U.S. economic perform-
ance and employment is complete without considering technology diffu-
sion that is, the factors that affect the speed and extent of adoption of
innovations.
Perhaps the most striking aspect of diffusion, and the factor that most
complicates the task of forecasting the employment and economic im-
pacts of new technologies, is its gradualness.6 It can take decades for all
of the members of a given industry, firm, or sector to adopt an innovation.
Enos (1962) found that the period between the invention of a new process
or product and its initial application (in other words, substantially prior to
its extensive utilization) averaged 14 years for one sample of inventions;
in another study, Mansfield (1961) found that, for 9 of 12 innovations,
adoption by all of the large firms in the coal mining, railroad, brewing, and
iron and steel industries took more than 10 years.
There are several reasons why diffusion is such a lengthy process.
Prospective adopters often find it difficult to evaluate new technologies;
as a result, they are uncertain about the benefits and costs involved and
may be reluctant to adopt a new technology rapidly. Moreover, the
transmission and absorption of the information necessary to adopt an
innovation require considerable time. Adopting a specific innovation may
also demand extensive complementary investments in new plants and
equipment and in work force training and retraining. Finally, the age and
other characteristics of the existing capital stock in potential adopter firms
affect the attractiveness of investing in a new technology.
Factors Affecting the Diffusion of Technology
Theoretical and empirical studies of technology diffusion suggest that
two broad factors influence the rate of diffusion of technologies: (1)
uncertainty surrounding the characteristics of a new technology and the
payoffs from adopting it, and (2) the actual profitability of its adoption.
Sociologists such as Rogers (1983) and economists such as Griliches
(1957, 1960) and Mansfield (1961, 1963b, 1966) have defined the charac-
teristic s-shaped curve describing the diffusion of an innovation: plotted
against time, the proportion of adopters within a population increases
slowly, then accelerates, and finally levels off (Figure 2-61. These re-
searchers suggest that the adoption of a technology by a growing number
of firms or individuals progressively reduces uncertainty and increases
6The fact that the economic and employment consequences of new technologies
frequently are felt more gradually than economic change induced by other causes, such as
currency fluctuations or natural disasters, should simplify the development of policies to aid
worker adjustment (see Chapter 7).
OCR for page 42
42 TECHNOLOG Y AND EMPLO YMENT
o
1
o
CL
O I
1~ ,~
Z
~ _
Z ~
Lll a
c' O
O: cat
-
-
/
/
TIME
FIGURE 2-6 Time path of the diffusion of a "typical" innovation.
the amount of information available to potential adopters, thereby accel-
erating adoption, until a large fraction of the relevant population has
adopted the innovation. Firm size also affects the speed with which an
innovation is adopted. Mansfield (1963a) found that large firms adopted
innovations more rapidly than small firms and attributed this difference to
the larger in-house engineering and scientific staff and financial resources
of the bigger firms, among other factors.
Diffusion rates vary across industries and technologies as a result of
structural and other factors that affect the profitability of adoption and the
level of uncertainty about such profitability. For example, government
regulation can play a role either in increasing or slowing diffusion.
Regulation of U.S. commercial air transportation prior to 1978 supported
rapid diffusion of new commercial aircraft among the passenger airlines
by encouraging competition based on service quality rather than on
price (Jordan, 1970; Mowery, 19851; in another case, more stringent
regulation since 1962 appears to have slowed the introduction and
OCR for page 43
SOURCES AND RATE OF TECHNOLOGICAL CHANGE 43
diffusion of new pharmaceuticals in the United States (Schwartzman,
1976~. There are limits, however, to what can be determined about
technology diffusion from the available data. Most empirical studies of
diffusion focus on cross-sectional differences in the adoption of a single
innovation, which restricts the ability to predict diffusion rates for
multiple innovations over time. Thus, little is known about the determi-
nants of aggregate trends in diffusion rates within an economy.
Any analysis of the diffusion of innovations is further complicated by
the fact that an innovation often is greatly modified in the course of its
diffusion (Rosenberg, 1976~. Examining the diffusion of computers during
the past four decades, for example, involves analyzing the diffusion of a
number of very different products, each of which has been modified
drastically since its introduction the capabilities of the original personal
computers differed greatly from those of subsequent microcomputers,
and these products bear little if any resemblance to the mainframe
behemoths of the 1950s and 1960s.
Another limitation in any analysis of diffusion rates is that empirical
studies have focused on manufacturing, health care, or agriculture there
are few studies of the diffusion of innovations within the services sector
outside of health care. The service industry diffusion studies that have
been performed (e.g., Stoneman, 1976, who considered the diffusion of
computers within British banks) confirm the importance of profitability
and information as key determinants of the rate at which productivity-
enhancing innovations are adopted. Although the specific impediments to
diffusion within the service industries may differ somewhat from those
observed within manufacturing, the general determinants of the rate of
diffusion appear to be quite similar across the two sectors.
How do the diffusion rates of specific technologies in manufacturing
and services compare? U.S. industry's use of advanced manufacturing
technologies, including robotics and computer numerically controlled
machine tools, seems to be increasing at a rate comparable to the rates of
diffusion of earlier process innovations such as mainframe computers.
The number of robots in U.S. industry, for example, grew at an average
rate of roughly 40 percent per year during 1981-1985, although this growth
appears to have slowed recently.7 The number of robots per 1,000
manufacturing employees (a figure including white-collar workers) grew
from 0.1 in 1976 to 1.3 in 1986 (J. Bernstein, Robotic Industries Associ-
ation, personal communication, 1987; Flamm, 19861. Moreover, and of
greater significance for the long-run employment impacts of technological
7Flamm (1986); see also "GM Throws a Monkey Wrench Into the Robot Market,
Business Week, August 25, 1986.
OCR for page 44
44 TECHNOLOGY AND EMPLOYMENT
change in U.S. manufacturing, both the level of use and the rate of
adoption of such productivity-enhancing innovations as robotics and
computer numerically controlled machine tools within U.S. industry
appear to be lagging behind those of many industrial competitors, notably
Japan, Sweden, and West Germany (Flamm, 1986; Mowery, 1986;
Technology Management Center, 19851. Jaikumar (1986) estimates that
"[iln the last five years, Japan has outspent the United States two to one
in automation. During that time, 55% of the machine tools introduced in
Japan were computer numerically controlled (CNC) machines, key parts
of FMSs [flexible manufacturing systems]. In the United States, the figure
was only 18~o" (p. 701.
The differences that can be observed among the United States and
other nations in the rates of diffusion and use of robotics are not well
explained by differences in wage rates, capital costs, or industry mix in
U.S. and foreign economies (Flamm, 19861. The empirical evidence on
rates of adoption (Mansfield, 1963b) also suggests that small U.S. firms
are likely to be even further behind the technological "frontier" than
large firms. This is a matter of some concern; the competitive and
technological vitality of smaller firms is important for overall U.S.
employment and competitiveness because of the roles such firms play as
employers (see Chapter 6) and as suppliers to larger manufacturing firms.
Data on rates of investment by U.S. firms in office automation and
information technologies suggest that diffusion of these technologies may
be occurring somewhat more rapidly than the diffusion of some new
manufacturing technologies. In the early 1980s, the rate of growth in the
use of computer workstations (on-line terminals and workplace personal
computers), which are predominantly found in nonmanufacturing settings
(Harris, 1983), was higher than that for robots. As Figure 2-7 indicates, the
number of U.S. workstations has increased from approximately 675,000 in
1976 to roughly 28 million in 1986, an average growth rate of 47 percent per
year.8 The number of workstations has grown from 15.4 for every 1,000
white-collar employees in 1976 to 450 per 1,000 white-collar workers in 1986.
Obstacles to the Diffusion of Technology
Before a firm can adopt many of the computer-based office and
manufacturing technologies of interest to this panel, it must overcome a
number of obstacles, which reflect the factors mentioned earlier as
important influences on the diffusion of technology. The obstacles a firm
8Letter of January 8, 1987, to Dennis Houlihan from Donald C. Bellomy, editor of
International Data Corporation's computer industry report The Gray Sheet.
OCR for page 45
SOURCES AND RATE OF TECHNOLOGICAL CHANGE 45
28
26
24
an
o
-
._
- 20
O 18
16
5e 1 4
o
Em:
LU
m
I
22
12
10
8
6
4
2
O4
f
of'
/
1 1 1 1 1 1 1 1
1975 1977 1979 1981 1983 1985
YEAR
FIGURE 2-7 Growth in the number of U.S. workstations (on-line terminals and nonhome
personal computers), 1975-1986. SOURCE: Donald C. Bellomy, International Data Corpo-
ration, personal communication, January 8, 1987.
faces can be grouped into three broad and overlapping categories: (1)
adoption costs, (2) product standards, and (3) the availability and evalu-
ation of relevant information.
The adoption costs associated with computer-based technologies that
integrate numerous separate operations are in many cases greater than
those associated with discrete innovations with less demanding integra-
tion requirements. Often, the technologies that underpin many of these
computer-based innovations are new to the industries and firms faced
with an adoption decision a factor that heightens the uncertainties about
the technology and increases the costs of acquiring the necessary exper-
tise for its evaluation and operation. Uncertainty and hence costs are
also increased by rapid changes in these technologies. The substantial
costs of the applications engineering necessary for adoption are likely to
be particularly onerous for smaller firms, which may have few or no
specialized technical personnel on their payrolls.
A related impediment contributing to higher adoption costs stems from
the fact that higher-level skills are often required for successful adop-
tion in the early stages of the introduction of new technologies. A
number of scholars (Barter and Lichtenberg, 1987; Nelson and Phelps,
OCR for page 46
46 TECHNOLOGYAND EMPLOYMENT
1966; Nelson et al., 1967) argue that the installation and "debugging" of
complex machinery for which little operating experience has been com-
piled are frequently lengthy processes, requiring specialized skills and in
many cases extensive scientific or technical training:
The observers of the early production of transistors remarked on the high
percentage of physicists and engineers required to control the processes. As
experience accumulated, however, it became possible to design machines to do
some of the jobs formerly requiring highly educated talent, and to develop training
programs to teach less educated workers the special things that they needed to
know to be effective workers. (Nelson et al., 1967, p. 106)
A highly skilled work force can adopt new technologies more rapidly.
Nonetheless, the high costs of training may impede the diffusion of
technologies in the United States; this is especially true if firms and
workers are unable to develop contractual agreements to share the costs
and benefits (in terms of higher productivity, higher wages, or product
quality) of retraining investments (Bendick and Egan, 19821. Moreover,
these retraining costs may place heavy burdens on small firms. Sweden
and Japan have been leaders in the adoption of computer-based manu-
facturing technologies and robotics, and both have labor market institu-
tions and practices that may support higher levels of investment in
training for their blue-collar work forces (see Chapter 71. Such invest-
ments may aid the faster adoption of some key manufacturing technolo-
gies in these nations.
Product standards play a central role in the development and adoption
of information and computer technologies. Within the United States,
standards in information technologies historically have been set by
market forces rather than by a governmental or industry-wide group. For
example, standards for computers were largely established by IBM,
reflecting its dominance of the market. For other technologies (e.g., office
automation or computer-based manufacturing), no single vendor domi-
nates the market; as a result, standards have been slower to emerge,
despite the activities of the Corporation for Open Systems and the
American National Standards Institute.
Because standards lessen the need for large investments in applications
engineering to modify interfaces among incompatible pieces of hardware
or software, they lower adoption costs and aid the adoption of new
technologies. In view of the salience of these costs for small firms,
standards are likely to be particularly useful in helping small firms adopt
new technologies. Once established, however, a product or process
standard may have an extremely powerful influence over the future
direction of technological change. Uninformed or hasty standardization
may effectively "lock in" an inefficient technology (David, 19851. The
OCR for page 47
SOURCES AND RATE OF TECHNOLOGICAL CHANGE 47
lack of standards thus retards diffusion, whereas their premature or
ill-informed establishment increases the risks of technological suboptimi-
zation.
Prospective adopters of computer-based technologies in manufacturing
and services often face problems in evaluating the cost consequences of
adoption. Many of the essential areas in which these process technologies
yield significant cost savings are not incorporated in conventional invest-
ment analyses because of conceptual flaws in these analytic frameworks.
For example, reductions in inventory or work in progress have been
singled out by several researchers as important dimensions of resource
savings that are ignored by accounting systems developed for the evalu-
ation of discrete investment decisions (see Ettlie, 1985, 1986; Kaplan,
1986; Technology Management Center, 19851. In some instances, U.S.
managers are not sufficiently familiar with a new technology to evaluate
its performance effectively. Improving management education may be
one way to provide the familiarity and analytic skills necessary for
informed evaluations of new technologies.
KEY TECHNOLOGY "CLUSTERS"
The preceding sections of this chapter have discussed technology in
general terms. What specific technologies will affect employment and the
workplace in the next 10-15 years? Brief descriptions of several salient
technologies follow; our discussion of them focuses on trends in technolog-
ical development and adoption and their employment implications. Four
technology "clusters" are considered: information technologies; computer-
aided manufacturing technologies (robotics, CIM, and flexible manufactur-
ing systems); materials; and biotechnology. Many of the important innova-
tions in all four of these technology clusters are well beyond the invention
stage and are now undergoing development for commercial applications.
This list is not comprehensive, nor are the items on the list mutually
exclusive-information technologies, for example, are critical to CIM, and
innovations in materials underpin both information technologies and
computer-aided manufacturing processes. The panel considered these tech-
nologies to be worthy of particular attention because of the widespread
notice each has received as well as their potential for widespread application
within the U.S. economy in the near future.
Information Technologies
One of the most important structural changes in the U.S. economy, a
change that affects both the manufacturing and services sectors, has been
the rapid development and application of information technologies (i.e.,
OCR for page 48
48 TECHNOLOGY AND EMPLOYMENT
technologies that store, retrieve, analyze, or transmit information). Com-
puters, telecommunications equipment, and the microelectronic compo-
nents on which they rely are included in this cluster. Within many sectors
of the modern economy, information is an increasingly important input to
the production of goods and often reduces the amount of labor and the
quantity of other inputs required per unit of output.9 Information has also
become an increasingly valuable commodity in its own right. The evi-
dence (e.g., U.S. Bureau of Labor Statistics, 1986b) suggests that the
development of these technologies should enhance the demand for
workers who manipulate and analyze information, relative to the demand
for workers who enter and collate data.
Computer-Aided Manufacturing Technologies
The incorporation of computer- and microelectronics-based technologies
within manufacturing has transformed the work environment in some
industries and firms while simultaneously contributing to public concern
over job displacement. These technologies include robotics, computer-aided
design and manufacturing, and microelectronics-based, machine-controlled
technologies such as computer numerically controlled machine tools.
Current estimates of the rates of development and diffusion of these
technologies in a wide range of functions suggest that they are unlikely to
produce mass displacement of workers during the next decade or two.
Moreover, according to some analysts (Cyert, 1985; Sanderson, 1987),
computer-aided technologies could support growth rather than reductions in
U.S. manufacturing employment: the reduced direct labor costs made
possible by these technologies may allow some U.S. firms to move assembly
and fabrication operations back to the United States from low-wage areas of
the world. Most public concern about these technologies focuses on the
displacement of production workers. Widespread adoption of computer-
aided manufacturing technologies, however, is also affecting middle-level
engineers and managers, as Chapter 6 notes.
Advanced Materials
Fundamental to progress in microelectronics and information technol-
ogies, as well as to many areas of manufacturing and the services sector,
are advances in such materials as ceramics (including high-temperature
9Freeman and Soete (1985) argue that "it is this feature which distinguishes IT
[information technology] so clearly from 'old-fashioned' automation. Some of the most
significant productivity gains linked to the introduction of IT relate to more efficient
inventory control, as well as significant energy, materials, and capital savings" (p. 55).
OCR for page 49
SOURCES AND RATE OF TECHNOLOGICAL CHANGE 49
superconducting materials), nonmetallic composites, and polymers. In-
novations in materials technology affect employment in several ways.
First, they may reduce markets for the materials they replace. On the
other hand, markets for the new materials may expand and create new
employment. The net employment effect of such a substitution is deter-
mined by the comparative labor requirements per unit of output for the
two materials, as well as the relative size and rates of growth in the
respective markets. Materials innovations also may affect labor require-
ments for processing and fabricating materials. Currently, however, the
magnitude and even the direction of these employment effects are
uncertain.
Biotechnology
The U.S. Congress's Office of Technology Assessment (1984) defines
biotechnologies as technologies that use living organisms to modify plants
or animals and develop microorganisms for specific purposes.~° Biotech-
nology arguably is the least advanced of the four clusters, reflecting its
recent development and the impediments to its rapid diffusion. The
sectors in which these technologies initially will be introduced the
pharmaceutical and chemical industries, agriculture, and environmental
protection do not employ large numbers of people, leading us to
conclude that the near-term aggregate employment impacts of biotech-
nologies will be modest and will primarily influence shifts within profes-
sional and technical occupations.
SUMMARY
The pace of technology diffusion governs the rate at which the
economic and employment effects of new technologies are realized. The
data discussed in this chapter suggest that within the U.S. manufacturing
sector, the pace of adoption of some new technologies is slower and the
levels of utilization lower than in some other industrial nations. This
slower rate of adoption within U.S. manufacturing may allay concerns
over the job-displacing impacts of rapid technological change, but it
actually carries a false assurance. Because foreign firms are adopting
Patois definition is by no means universally accepted. The National Research Council's
Board on Agriculture (1987) defines biotechnology as "the use of technologies based on
living systems to develop commercial processes and products. . . [including] the tech-
niques of recombinant DNA, gene transfer, embryo manipulation and transfer, plant
regeneration, cell culture, monoclonal antibodies, and bioprocess engineering" (p. 3). Other
analysts (Miller and Young, 1987) reject any effort to develop a definition of biotechnology.
OCR for page 50
50 TECHNOLOGY AND EMPLOYMENT
these technologies more rapidly than U.S. firms and are expanding their
shares of the U.S. and world markets, job displacement from the slow
adoption by U.S. firms of these productivity-increasing manufacturing
technologies is likely to be more serious than any displacement resulting
from rapid adoption; the recent surge in import penetration of many U.S.
manufacturing industries provides support for this assertion (see Chapter
3 for additional discussion). U.S. industry must operate closer to the
technological frontier if this nation is to maintain high employment levels
and living standards.
Representative terms from entire chapter:
manufacturing technologies
