Research Policy 29 Ž2000.

587–602
www.elsevier.nlrlocatereconbase

Standardization in technology-based markets

)
Gregory Tassey
Program Office, National Institute of Standards and Technology, ADMIN-1000, Gaithersburg, MA 20899, USA

Abstract

The complexity of modern technology, especially its system character, has led to an increase in the number and variety of
standards that affect a single industry or market. Standards affect the R & D, production, and market penetration stages of
economic activity and therefore have a significant collective effect on innovation, productivity, and market structure.
Standards are classified into product-element and nonproduct categories because the two types arise from different
technologies and require different formulation and implementation strategies. Because standards are a form of technical
infrastructure, they have considerable public good content. Research policy must therefore include standardization in
analyses of technology-based growth issues. q 2000 Elsevier Science B.V. All rights reserved.

Keywords: Standardization; Innovation; R & D; Economic growth; Industry structure

Through R & D-performing industries and the ef- by this sector on the overall economy, but also that
fect of new technologies on other parts of the econ- extensive diffusion of new technology must take
omy, technology accounts for one-third to more than place if adequate productivity growth rates are to be
one-half of US GDP growth and at least two-thirds achieved by the entire economy.
of productivity growth. However, the so-called Standardization affects both innovation and tech-
‘‘high-tech’’ sector only contributes approximately nology diffusion. It also can influence industry struc-
7% of US GDP ŽTassey, 1999.. 1 This relatively ture and thereby help determine which firms benefit
small direct contribution implies substantial leverage and which do not from technological change. Thus, a
concern of R & D policy should be the evolutionary
path by which a new technology or, more accurately,
certain elements of a new technology become stan-
)
E-mail: [email protected] dardized. Over a technology’s life cycle, standardiza-
1
The high-tech sector is defined here as consisting of four tion can affect economic efficiency. However, these
major categories: high-tech manufacturing ŽIT-related plus indus-
trial electronics., communication services, software and
effects can be both positive and negative. For exam-
computer-related services, and pharmaceuticals. For alternative ple, standardization can increase efficiency within a
definitions of IT-related high-tech industries, see American Elec- technology life cycle, but it also can prolong existing
tronics Association Ž1997. Žp. 128. and Department of Commerce life cycles to an excessive degree by inhibiting in-
Ž1998. ŽAppendix, pp. A1–A2.. The AEA definition results in a
vestment in the technological innovation that creates
6.1 GDP estimate for 1996 and the Commerce definition yields
about 8% for 1998. To either of these definitions should be added
the next cycle.
pharmaceuticals, which brings the AEA-defined high-tech sector’s Standardization can and does occur without for-
GDP contribution to 7%. mal promulgation as a ‘‘standard’’. This distinction

0048-7333r00r$ - see front matter q 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 0 4 8 - 7 3 3 3 Ž 9 9 . 0 0 0 9 1 - 8
588 G. Tasseyr Research Policy 29 (2000) 587–602

between de facto and promulgated standards will be systems can actually facilitate product variety and
made apparent and discussed in the following sec- hence choice for the customer.
tions. In one sense, standardization is a form rather Fig. 1 depicts the multiple functions performed by
than a type of infrastructure because it represents a standards. These functions transcend the three major
codification of an element of an industry’s technol- stages of technology-based activity — R & D, pro-
ogy or simply some information relevant to the duction, and market penetration, and are difficult to
conduct of economic activity. On the other hand, the construct and implement because many important
selection of one of several available forms of a technologies have both an intrinsic complexity and a
technology element as ‘‘the standard’’ has poten- ‘‘systems’’ character. Such characteristics demand
tially important economic effects. more sophisticated technological foundations for
standards and imply the need for technically compe-
tent standards setting processes.
1. Economic functions of standards The greater complexity of technologies and the
associated networks of firms and supporting infras-
A standard can be defined generally as a construct
tructure that develop and disseminate these technolo-
that results from reasoned, collective choice and
gies mean that supply chains are becoming the most
enables agreement on solutions of recurrent prob-
important level of policy analysis. Greater distribu-
lems. Looked upon in this way, a standard can be
tion of R & D among materials and equipment suppli-
viewed as striking a balance between the require-
ers, manufacturers of products, and providers of
ments of users, the technological possibilities and
services increasingly characterize high-tech supply
associated costs of producers, and constraints im-
chains. The consequent increase in market transac-
posed by government for the benefit of society in
tions involving technology also demands standards
general ŽGermon, 1986..
to reduce the associated transaction costs.
More functionally, an industry standard is a set
Technology consists of a number of discrete ele-
of specifications to which all elements of products,
ments that tend to evolve in different institutional
processes, formats, or procedures under its jurisdic-
settings. These elements have distinctly different
tion must conform. The process of standardization
character and require different types and combina-
is the pursuit of this conformity, with the objective
tions of standards to effect efficient development and
of increasing the efficiency of economic activity.
utilization ŽTassey, 1992, 1997.. Many variations
1.1. Nature and scope of impacts exist but broadly defined they fall into the three
major categories shown in Fig. 1:
Standards played an important role in the indus- 1. The fundamental or generic technology base of
trial revolution. They allowed factories to achieve the industry on which subsequent market applica-
economies of scale and enabled markets to execute tions Žproducts and services. are based.
transactions in an equitable and efficient manner.
Standardization of parts made supplier specialization
possible and increased efficiency over the entire
product life cycle by facilitating part repair or re-
placement.
In a modern economy, standards constitute a per-
vasive infrastructure affecting the technology-based
economy in a number of important and relatively
complex ways. Some of these impacts even appear
contradictory. For example, whereas the traditional
economic function of standards in production can
restrict product choice in exchange for the cost ad-
vantages of economies of scale, other types of stan-
dards common to advanced production and service Fig. 1. Roles of standards in a technology-based industry.
 G. Tasseyr Research Policy 29 (2000) 587–602 589

2. A set of infratechnologies that provide varied and point of departure for competition in an industry. For
critical technical infrastructure support for devel- example, a case study by Putnam et al. Ž1982. points
opment of the generic technology and subsequent out that when an automobile manufacturer develops
market applications. a new engine, the company specifies the minimum
acceptable lubrication attributes. This specification
3. The market applications Ž proprietary technolo- then becomes the basis for competition among
gies .. petroleum companies, who either compete on price
Because the market use of each element differs at the minimum specified level of quality or by
significantly, so do private-sector investment incen- offering motor oil with a level of performance above
tives with the result that underinvestment varies the minimum.
across the different elements of an industrial technol-
ogy. The greater the infrastructural character of a
technology, the more underinvestment is likely to 1.2.2. Information standards
occur. Standards and thus their technical underpin- Standards help provide evaluated scientific and
nings have a strong infrastructure character, so that engineering information in the form of publications,
underinvestment is common. electronic data bases, terminology, and test and mea-
surement methods for describing, quantifying, and
1.2. Basic functions of standards evaluating product attributes. In technologically ad-
vanced manufacturing industries, a range of mea-
To analyze the economic roles of standards in a surement and test method standards provide informa-
technology-based economy, a taxonomy is required tion, which, by virtue of being universally accepted,
that classifies standards by functions having unique greatly reduce transaction costs between buyer and
economic characteristics. For the purpose of eco- seller. In their absence, especially for complex, tech-
nomic impact assessment, the functions of standards nology-based products, considerable disagreement
are classified into four categories. 2 The positive will often ensue over verification of performance
effects of each of the four functions are described claims. These disputes raise the cost of consummat-
below. However, as discussed in later sections, stan- ing a marketplace transaction, which is reflected in
dards also can have negative economic consequences higher prices charged. The economic impact is to
or simply fail to achieve their maximum potential slow market penetration.
economic benefit. Measurement methods are also essential to con-
duct state-of-the-art research. In today’s semiconduc-
1.2.1. Qualityr reliability tor R & D, scientists and engineers must be able to
Standards are developed to specify acceptable measure the distances between individual atoms
product or service performance along one or more Ždopants. that are added to silicon to achieve the
dimensions such as functional levels, performance desired millions of high-density electronic functions
variation, service lifetime, efficiency, safety, and on a single chip. Standardization of some of these
environmental impact. A standard that specifies a methods is essential for the efficiency of R & D itself.
minimum level of performance often provides the For example, being able to replicate and verify re-
search results is often critical to obtaining follow-on
research funding or commitment to commercializa-
tion. Standardized scientific and engineering data Žin
2
This taxonomy follows Tassey Ž1982; 1992; 1997. and Link the sense of having been critically evaluated and
and Tassey Ž1987.. David Ž1987. proposes a similar taxonomy verified for accuracy. and standardized equipment
based on three kinds of standards Žreference, minimum quality, calibration techniques are also essential for efficient
and compatibility.. Other taxonomies have been developed based R & D.
on the process by which a standard comes into existence. For
example, David and Greenstein Ž1990. provide a framework to
Finally, the typical manufacturing process is
classify standards as de facto Ž‘‘unsponsored’’ or ‘‘sponsored’’. increasingly measurement intensive because of
or promulgated Žvoluntary or dejure.. growing demands for quality and real-time process
590 G. Tasseyr Research Policy 29 (2000) 587–602

control. Traditional manufacturing processes tested proprietary component designs to coexist — that is,
products after a production run. The inefficiency of they enable innovation at the component level by
this approach is large, not only because of the wasted being competitively neutral with respect to design. In
material and labor when a production run must be effect, competitors can innovate on ‘‘either side’’ of
scrapped, but also because of down time and hence the interface while the consumer of the product
unused capacity incurred while a problem is identi- system can select the particular components that
fied and fixed. 3 The availability of computers makes optimize system design. They also allow substitution
possible the real-time monitoring and control of a of more advanced components as they become avail-
process, potentially enabling instant adjustment of able over time, thereby greatly reducing the risk of
process variables. obsolescence of the entire system. Widespread fac-
When fully implemented, real-time control can tory automation as it is currently evolving in ad-
virtually eliminate waste and increases product mix vanced economies likely will not proceed without
flexibility. However, real-time control of a produc- these standards.
tion process is a complex systems technology, requi- Without interface standards, large companies of-
ring large numbers of sensors, computers, and ten supply ‘‘turnkey’’ systems where proprietary
software. Standardization of certain performance interfaces link components. However, the cost to the
measurements for, say, a sensor facilitates design of user can be high because the system is not optimized
the equipment. Equally large efficiency gains can for the user’s particular needs Žcompetitors will typi-
occur from standardization of certain elements of the cally offer components that are superior to some of
actual methods or techniques of process control, as those in the turnkey system. and price competition
adherence to these standards often removes the need will not be a factor when system components need
for much of the post production testing. Finally, replacement.
whatever production strategy is adopted, the equip- In such situations, system design can still be
ment that helps execute the strategy must be periodi- optimized. However, the cost of modifying physical
cally calibrated by standards in order to ensure maxi- and functional interfaces to allow components from
mum efficient performance. different vendors to work together Ži.e., to ‘‘interop-
erate’’. is usually prohibitive. Moreover, full func-
1.2.3. Compatibilityr interoperability tionality is often not obtained by reengineering pro-
Standards specify properties that a product must prietary Žnonstandard. interfaces.
have in order to work Žphysically or functionally.
with complementary products within a product or 1.2.4. Variety reduction
service system. This function of standards has been Standards limit a product to a certain range or
the most intensively studied by economists. 4 Com- number of characteristics such as size or quality
patibility or interoperability is typically manifested levels. The fourth function of standards is the tradi-
in the form of a standardized interface between tional one of reducing variety to attain economies of
components of a larger system. An effective inter- scale. The majority of standards perform this func-
face standard does not affect the design of the com- tion. However, variety reduction is no longer simply
ponents themselves, such as numerically controlled a matter of selecting certain physical dimensions of a
machine tools or the components of these tools, product for standardization Žsuch as the width be-
including controllers. In fact, interface standards pro- tween threads of a screw.. Variety reduction is now
vide ‘‘open’’ systems and thereby allow multiple commonly applied to nonphysical attributes such as
data formats and combined physical and functional
attributes such as computer architectures and periph-
3
eral interfaces.
Scrap can result when periodic testing of product reveals an The process of setting variety reduction standards
attribute flaw. All units produced between the sample product
tested and the previous test are equally defective.
also varies significantly. Many standards of this type
4
See, for example, David and Steinmueller Ž1994., Kahin and are viewed as infrastructure and thus adopted by an
Abbate Ž1995., and Link and Scott Ž1998.. industry consensus process. However, standardiza-
 G. Tasseyr Research Policy 29 (2000) 587–602 591

tion of some attribute or element of a product is just government intervention are very different in the two
as often achieved through the marketplace by one cases.
firm gaining control of the underlying technology Simply put, standardization of one or more at-
and using this control to force other manufacturers tributes of a product Žor service. can convey direct
with whom that firm competes to adopt its version competitive advantage to the ownerrcontroller of the
the technology. The product element then becomes a technology producing those attributes. They, there-
de facto Žnonconsensus. standard. 5 fore, get considerable attention from economic
Conceptually, the variety reduction function is the growth strategists. Conversely, nonproduct standards
most difficult category of standardization to analyze tend to be competitively neutral, at least within an
because of its ability to either enhance or inhibit industry or trading block. Hence, they tend to get
innovation. Variety reduction typically enables less attention. Yet, this latter category can be critical
economies of scale to be achieved, but larger produc- to the entire industry’s efficiency and its overall
tion volumes tend to promote more capital-intensive market penetration rate.
process technologies. This common evolutionary pat-
tern of a technology over a number of product life 2.1. Product-element standards
cycles usually reduces the number of suppliers and
increases their average size. Such trends may or may Product-element standards typically involve one
not reduce competition, but often progressively ex- of the key attributes or elements of a product, as
clude small, potentially innovative firms from entry opposed to the entire product. Government, espe-
due to increased minimum efficient scale thresholds. cially where large economies of scale are present or
when early market entry is considered an essential
part of a national economic strategy, mandates some
2. Types of standards product-element standards.
In most cases, at least in the US economy, market
Standards have been classified by the form they dynamics determine a de facto standard. Alternative
take. To varying degrees all four functions of stan- technologies intensely compete until the dominant
dards specified above can be described alternatively version gains sufficient market share to become the
in terms of design specifications or performance single standard. Market control by one firm can
levels. Design-based standards are much more re- truncate this competitive process. Such control is
strictive and can inhibit innovation to a greater de- particularly effective in cases of increasing returns
gree than performance-based standards. The latter and can quickly force acceptance of the monopolist’s
type allows flexibility in product or service design proprietary technology element as the standard.
while still meeting the performance requirements of However, the globalization of high-tech markets with
the standard. Thus, standards generally work more shorter product life cycles is making single-firm
efficiently when they are performance-based. dominance more difficult. In response, various com-
The economic functions of standards aggregate binations of vertical and horizontal consortia are
into two basic categories — product and nonproduct promulgating product-element standards by consen-
— delineated by their relationship to product Žor sus, at least within a single large economy or trading
service. structure and their public good content. Dis- block.
tinctions between the two types are important be- One of the more visible examples of the competi-
cause their economic roles are distinctly different tive effects of de facto standardization is the ‘‘archi-
and hence so are the processes by which they are set. tecture’’ of personal computers. 6 Elements of com-
The differences in public good content have impor-
tant implications for policy because the rationales for 6
Architecture is the scheme by which the function of a product
is allocated to physical components ŽUlrich, 1995, p. 419.. Archi-
tecture is important because it provides a set of standardized
5
Intel’s microprocessor architecture and Microsoft’s operating product attributes and the rules or protocols for their interaction
system are well-known examples. with other product elements.
592 G. Tasseyr Research Policy 29 (2000) 587–602

puter architecture such as the operating system, the browser. and applications programs that run on that
‘‘bus’’, the graphical user interface, and the applica- operating system. Such charges were pursued in a
tions programming interface, have been the focus of 1998–1999 US government antitrust lawsuit.
intense competition by firms seeking to gain suffi- In contrast, the Japanese PC industry has never
cient market control to ‘‘set’’ the de facto standard been able to establish a single standard operating
for the particular product element. 7 system. Reasons for this include Ž1. an oligopolistic
In this regard, Apple Computer made a brilliant industry structure Ž keiretsu., which has resulted in
move when it forced third-party software developers factionalism within the industry and thereby pre-
to adopt a standard ‘‘graphical user interface’’ so vented a single standard Ždomestic or foreign. from
that all programs running on Apple’s computers gaining dominance; Ž2. a mainframe computer orien-
presented the user with the same screen format and tation, which for a long time relegated PCs to basi-
command structure. 8 However, Apple kept its hard- cally terminal status, thereby slowing market growth;
ware operating system proprietary Ži.e., it did not Ž3. a reluctance to move away from customized
adopt an open systems architecture strategy. and, software; and Ž4. a language barrier — the inability
thus, it had no chance to become the industry stan- to export kanji-based software ŽCotrell, 1994..
dard. This decision explains Apple’s initial prosper- Over a technology’s life cycle, additional ele-
ity and subsequent competitive decline. In contrast, ments of the product’s technology become standard-
Sun Microsystems opened the microprocessor archi- ized so that the product takes on a ‘‘commodity’’
tecture for its workstations in order to obtain help in character. Competition among suppliers of the
gaining market share. It was willing to share its ‘‘standardized’’ product then becomes increasingly
markets in return for the increased probability that its based on price and service-related aspects of the
architecture would become a standard and thereby product’s acquisition and use. Dell Computer, for
create a large and stable customer base. example, has succeeded in the PC industry by being
The policy issue is thus: to the degree that certain the low-cost producer and offering excellent before
product elements must become standardized to en- and after sales service. This evolutionary pattern was
able economies of scale and network externalities to noted decades ago by the famous Austrian economist
be realized, the potentially large benefits that accrue Joseph Schumpeter Ž1950., who observed that one of
to the firm owning the technology element are ac- the essential dynamics of capitalism is assuring that
ceptable. However, this is the case as long as compe- the ‘‘silk stockings’’ initially purchased only by the
tition is not constrained in the other elements that rich would eventually be items of mass consumption.
make up the overall product or system technology.
This caveat with respect to economic efficiency
was the basis for charges by software firms that 2.2. Nonproduct standards
Microsoft was using its monopoly position as owner
of the dominant operating system standard to gain Nonproduct standards, as the name implies, de-
unfair advantage in marketing its own versions of rive from a different technical base than that upon
new operating system elements Žsuch as a web which the attributes of the product itself depend.
Industry organizations often set these standards using
consensus processes. The technical bases Žin-
fratechnologies. for such standards have large, al-
7
The business literature continues to debate the efficacy of
though not total, public good content, so that their
competition to supply the product elements that constitute an provision frequently depends upon a combination of
architecture. See Morris and Ferguson Ž1993.. industry and government investment. Examples of
8
The term graphical user interface refers to the use of infratechnologies frequently embodied in nonproduct
‘‘menus’’ of images instead of characters to indicate instruction standards include measurement and test methods,
options for the computer user. Apple pioneered this much-pre-
ferred format and also made this approach even more attractive to
interface standards, scientific and engineering
users by sticking to a standardized format for all application databases, and artifacts such as standard reference
programs. materials ŽTassey, 1997, Chaps. 8 and 9..
 G. Tasseyr Research Policy 29 (2000) 587–602 593

Most infratechnologies and therefore the resulting

industry standards are derived from basic standards.
Basic standards represent the most accurate state-
ments of the fundamental laws of physics and have
such diverse applications that they qualify as pure
public goods and hence are provided entirely by
government. Basic standards are relatively few in
number and are not easily transported to or used by
industry. Thus, as indicated in Fig. 2, such standards
are converted into working and transfer standards
that convey the standardized information to industry.
Large numbers of industry standards are based on
Žtraced back to. basic standards. With respect to
form, most industry standards are either method, Fig. 2. Hierarchy of nonproduct standards ŽTassey, 1997, Chap.
procedural, or normative. 9 9..
Fig. 2 provides an example of the hierarchy
through which basic standards are utilized to develop
infratechnologies upon which semiconductor indus- order to ensure widespread and accurate transfer of
try standards are based. The production of semicon- the infratechnology.
ductor components is a highly demanding process. The information transferred by a physical stan-
The densities of today’s circuits are such that each dard is itself determined or certified by a so-called
conducting path Ž‘‘line’’. on a chip is a small frac- working standard, which is laboratory-based and
tion of the width of a human hair. These widths must more accurate but not readily transportable. In this
be consistent with respect to design specifications to example, the working standard is a laser interferome-
avoid thermal, electrical, and other problems. The ter, which measures and certifies the physical dimen-
semiconductor producer therefore needs to be able to sions of the line reticule prior to transfer to industry.
measure the widths of circuit lines that make up a Finally, the laser interferometer is itself dependent
‘‘chip’’. Particularly important are line widths on the for calibration on a ‘‘basic’’ standard for length. 10
masks that are used to inscribe the multiple layers of The four basic functions of standards are all
circuit patterns on the chip itself. Such masks are represented in the range of standards for a single
used to make tens of thousands of chips. Their industry like semiconductors. Many standards, like
quality greatly affects performance of the chips pro-
duced and hence a semiconductor manufacturer’s
production yield. 10
The line-width measurement equipment must be Line-width measurement is just one of many infratechnolo-
gies that a competitive semiconductor industry must utilize. For
calibrated against a physical standard, which has a example, current state-of-the-art chips consist of multiple layers of
pattern of lines whose thickness and spacing have circuits. The circuits in each layer must be connected to adjoining
been determined to a specified level of accuracy. layers. Accomplishing this very difficult manufacturing step re-
This determination is done by an authoritative source, quires a precise alignment of the ‘‘mask’’ Žcircuit pattern. for
such as the National Institute of Standards and Tech- each layer. Until recently, the alignment process required multi-
million-dollar optical equipment. NIST, however, developed a
nology ŽNIST. in the United States. The physical or procedure that allows semiconductor manufacturers to ensure
‘‘transfer’’ standard used by industry must be easily proper alignment of successive layers of an integrated circuit with
transportable Ža reticule in the above example. in a precision better than ten nanometers. This new calibration
standard represents a more than fivefold improvement over cur-
rent alignment calibration methods and is much less expensive.
The cumulative economic impact of such advances in infratech-
9
A normative standard is one in which a particular value Žsize, nologiesrstandards is substantial and can greatly affect price as
performance, quality, or design. is selected from a range of well as quality, and hence competitive position for the domestic
alternative values. industry.
594 G. Tasseyr Research Policy 29 (2000) 587–602

line-width measurement, provide information, while are multiple sites — hospitals, labs, clinics, pharma-
others affect variety reduction, quality, reliability, or cies, home-care settings, and so on — that conduct
compatibility. The collective economic impact of numerous clinical and administrative activities. Cur-
these standards is greatly magnified when an entire rently, the operational computing infrastructures
supply chain is considered. Semiconductors are a across these organizations are rarely the same. In this
component of computers and communications equip- environment, incomplete standardization is an in-
ment, which, in turn, constitute an information net- creasingly serious problem. User-defined attributes
work. Each level in the supply chain requires an for specific applications make plug-and-play data
elaborate infrastructure of standards. 11 interchange impossible. Even where two applications
are built according to an existing standard, frequent
user-defined modifications to the applications often
3. Market structure effects of standards prohibit ‘‘seamless’’ communication between them.
The solution to such a problem, where standards
The market structure effects of standardization exist but customization of applications creates in-
have important effects on the achievement of eco- compatibility with the existing standardized inter-
nomic growth objectives. ‘‘Open systems’’ allow face, is to develop and reach agreement on true open
small and medium companies to participate in mar- systems. Achieving complete interoperability elimi-
kets for system technologies by supplying compo- nates levels of complexity in implementing limited
nents in which they have a comparative advantage. or partial standards. In the healthcare area, Hewlett-
This diversification on the supply side of the market Packard took the lead in organizing an industry
makes system optimization by users Žthe demand consortium ŽThe Andover Working Group. to build
side. possible and increases price competition. an ‘‘enterprise communication framework ŽECF.’’
Decisions by large companies to enter systems that provides a common interface for data communi-
markets at the component level are also affected by cations. HP was motivated by its desire to sell
the availability of interface standards. These firms multiple products into a single large market. With
typically target larger markets where they can benefit interoperability among system components, such a
from economies of scale. A strategy of selling into market retains the advantage of diversification at the
markets characterized by proprietary turnkey systems component and system levels and also achieves the
requires a high degree of product segmentation in efficiency advantages of interoperability.
order to service each turnkey system integrator. This Such interoperability is a demanding challenge. It
situation increases costs and constrains market growth is achieved by encapsulating existing standard mes-
projections, even with a superior technology. The sage profile specifications into software objects called
more distributed the participants in a market, the enterprise communicators. These communicators
more critical to technological innovation are open provide a comprehensive set of rules that make it
systems. possible for developers to achieve interoperability
A good example is medical services, which has for their applications without having to completely
traditionally suffered from excessive customization. understand the existing complex standards and
Moreover, healthcare environments are becoming ex- rewrite them into their applications. This software
tremely distributed. In every local community, there code is packaged within the ECF and embedded into
third-party applications as a software component.
In practice, attaining truly open computer systems
11
In communication networks, for example, there are published has been extremely difficult, with achievement of
standards and interface protocols that allow hardware components compatibility being partial at best. One of the major
and software from many vendors to operate as a single product or reasons for this difficulty is the complexity of sys-
as a system of products Ži.e., as a network.. To achieve efficient or tems technology. Each of a number of layers in the
‘‘seamless’’ integration, the standards and protocols define what
rules hardware components must adhere to in order to exchange
hierarchical structure of a typical information system
signals between applications software and operating systems at must have its own architecture and these architec-
different levels in the network. tures must be integrated into a larger network. To-
 G. Tasseyr Research Policy 29 (2000) 587–602 595

day, systems technologies not only have complex At the product or component level, increasing
hardware and software platforms, but require rules returns can result in a monopoly position for one
and protocols to introduce, process, store, move, and supplier. If the product happens to be critical in some
retrieve information. way, that monopoly can be extended into related
The existing systems architectures are typically component markets. In other cases, multiple stan-
controlled by a handful of firms. Vendors often dards exist for extended periods of time resulting in
claim to conform to open systems’ requirements, but a segmented market structure. The latter situation is
in reality they conform only to a portion of the set of caused by high switching costs relative to perceived
standards to which compliance is necessary to sell benefit gains by users. These switching costs can be
their systems. Moreover, even when true open sys- increased when suppliers with monopolistic control
tems compatibility is offered, users often find that of portions of the market for the main product Žsuch
mobility from their current proprietary system to the as a minicomputer. fail to agree on a single standard
open system does not exist, even when the same for a complementary product Žsuch as the operating
vendor provides both. That is, mobility products are system.. The UNIX operating system is an important
not made available, leaving the user with a severe example.
transition problem. Such partial conformity standards
tend to segment the market structure.
In summary, lack of conformity to interface stan- 4. ‘‘Systems’’ aspects of standards
dards by the supply side of a market can be a
conscious strategy or a reflection of the poor quality Product and nonproduct standards within a prod-
of the standard. 12 Large firms with market strategies uct or service system interact with each other to
focusing on turnkey systems have an incentive to create a systems character to the overall economic
resist open systems to protect the market shares effects of standardization. The interactions among
resulting from their horizontally integrated strategy. different types of standards that affect a single indus-
In cases of increasing returns, such strategies may try are summarized in Fig. 3. Each of the two
result in a monopoly position being attained by one industries shown here depends on standardized for-
supplier or prolonged market segmentation by sev- mats and evaluations of scientific and technical in-
eral competing suppliers. Either situation has the formation from supporting science and technology
potential to constrain economic efficiency. infrastructures. They use this information to select
Conversely, large users are likely to want to have and conduct R & D projects.
access to several different classes or brands of a
particular product, or at a minimum have a second
source of supply for a critical component. Such
product users have substantial need for effective
interfaces. The ability to integrate equipment and
software from different vendors Žor even to integrate
different classes of equipment from a single multi-
product vendor. has several economic benefits: sys-
tem costs can be reduced and performance optimized
thereby increasing productivity, competition among
vendors for each element of the system lowers prices,
and responding to technological or functional obso-
lescence is facilitated by replacing individual compo-
nents at relatively low cost.

12
See David and Greenstein Ž1990. for an excellent review of Fig. 3. The structure of industrial standardization ŽTassey, 1997,
the early literature on these topics. Chap. 9..
596 G. Tasseyr Research Policy 29 (2000) 587–602

These industries then set additional standards spe- standards. The huge estimated cost of next genera-
cific to their internal needs for production efficiency. tion fabrication plants and shorter product life cycles
Finally, each industry must also be concerned with finally have made the risk too great for individual IC
standards that affect their ability to sell their prod- manufacturers to incur the considerable burden of
ucts in specific markets. Meeting procurement speci- undertaking the standards development efforts alone.
fications in domestic markets or complying with As a result, an international consortium of semicon-
certification requirements for exporting to foreign ductor firms is developing of standards for the gener-
markets are examples. Moreover, as many products ation of semiconductors based on a 300-mm wafer.
now become components of larger product systems,
part of meeting specifications is demonstrating com-
patibility with the performance attributes of other
5. Potential negative impacts of standardization
products.
Interaction among interface standards occurs be-
cause of the need to optimize product systems. For Both types of standardization described here can
example, a numerically controlled machine tool is a greatly increase overall economic efficiency and
system of components such as the controller and hence rates of growth in technology-based industries.
numerous sensors. Link and Tassey Ž1987. Žpp. 88– However, through their structure or through the tim-
96. point out that the multiple interface standards ing of their implementation, they can also have
involved in a machine tool allow modularization of negative economic impacts.
the product, which permits custom design and pre-
vents obsolescence. However, these standards must
5.1. Product-element standards
be linked by a common data format, if all compo-
nents are to function together as an efficient system.
Moreover, the structure of the components affects Once a product-element standard is set, network
the interface standards. The technical difficulty and externalities are typically realized. However, the
hence cost of constructing an interface standard in- marketplace dynamics that result in one firm’s ver-
creases when all components are completely propri- sion of the technology becoming the standard do not
etary. A frequent solution is to standardize certain guarantee that this version is the optimal one. A
attributes or elements of key components. In the case ‘‘lock-in’’ effect ensues, whereby many developers
of the machine tool, a standard architecture for the of related products conform to the standard and
machine tool controller has been agreed upon. This purchasers of the standardized technology invest
step has significant efficiency advantages, such as substantial resources in learning to absorb and use it
allowing upgrades to the controller without costly as well as complementary technologies and infras-
reengineering of the interfaces between the controller tructures. These sunk costs create a reluctance to
and other hardware and software components. switch to a new standard and related cluster of
As Tassey Ž1997. Žpp. 172–175. points out, in- technologies. Considerable resources are then allo-
dustries that have complex, measurement intensive cated to an inferior technology, which can extend
production processes must reengineer multiple stan- over long periods of time. 13
dards simultaneously for each new generation of the As pointed out by a number of economists, the
generic product technology. The semiconductor in- controller of the product-element standard captures
dustry is an excellent example. In the past, a single increasing returns to scale as initial market penetra-
dominant integrated circuit supplier has undertaken
the development of standards associated with transi-
tions between generations of semiconductor process
technology Žusually revolving around wafer size.. 13
For more than a decade, DOS — a clearly inferior operating
However, accomplishing the transition now re- system to alternatives such as Apple Computer’s system —
quires a complex and integrated standards develop- dominated the personal computing market because the
ment process, involving a large number of different MicrosoftrIBM combination was able to lock in the PC market.
 G. Tasseyr Research Policy 29 (2000) 587–602 597

tion begets dominance. This may be an acceptable standards for the same infratechnology, Ž2. poorly
price for rapid market growth from an economic designed standards, and Ž3. poorly timed standards.
welfare perspective, as long as either the standard- Multiple standards usually arise in the early phases
ized version of the technology element is the best of a technology’s development when innovating firms
one or sufficient opportunity exists for market forces are intent on achieving market penetration. At this
to periodically replace the standard. 14 point, the market is often not large enough to warrant
The tendency for modern technologies to have investment in infratechnologies such as sophisticated
systems structures increases the economic impor- test methods. Consequently, innovators rely on off-
tance of the lock-in phenomenon. In particular, most the-shelf or contrived methods that hopefully con-
technologies today have clusters of standards embed- vince customers to try the new technology.
ded in them. One, however, is frequently the domi- However, as the market expands, imitators enter
nant or driver standard. Thus, the controller of that with their own versions of the required test methods.
standard can have even greater impact on the evolu- Confusion among buyers soon takes over, as differ-
tionary development of the overall technology and ent vendors provide test methods that produce differ-
related markets. ent results. The cost to buyers of determining what
While the lock-in effect works on the demand test method most accurately measures performance
side, the ‘‘installed base’’ effect impacts the supply rises as their purchases increase. Moreover, because
side. Dominant suppliers of a technology, who either these multiple measurement and test methods are
control or adhere to the standard, have invested typically adapted from previous generations of the
substantial resources in developing and servicing the underlying infratechnology, they are frequently inac-
markets based on this standard. Their established curate and sometimes markedly so.
market positions promote evolutionary as opposed to These first two problems can be resolved by
revolutionary migration of customers, who demand additional research to develop better infratechnolo-
backward mobility to existing technology. Mobility gies, which can then be adopted as single industry
is more difficult and expensive, the more radical the standards. However, because a consensus process
new technology. If Microsoft, Apple, and IBM were involving numerous industry stakeholders promul-
freed from maintaining compatibility with existing gates most standards of this type, developing the
hardware and software, new generations of systems needed infratechnology and reaching a consensus on
software would likely evolve at a faster pace. the standard often encounters substantial obstacles. 15
The resulting delays can create significant costs in
executing market transactions, which raise prices and
5.2. Nonproduct standards
thereby retard market penetration.
The problems of multiple standards and the tim-
Because this class of standards consists of a set of ing of standardization also arise at the interface
‘‘tools’’ that make the process of developing, pro- between components of a system of technologies.
ducing, and marketing the core technology more The increasingly important objectives of compatibil-
efficient, they do not have the dominant or control- ity and interoperability are implemented through in-
ling effects of product-element standards. Thus, the terface standards. Interface standards typically have
potential negative impacts are not as obvious. How- both physical and functional elements and can be
ever, negative economic effects can occur. They extremely complicated to design, especially in sys-
typically result from three problems: Ž1. multiple tems technologies where multiple components must
interoperate.

14
Arthur Ž1996. provides an excellent analysis of the increasing
returns-to-scale phenomenon that characterizes many technology-
15
based industries and the consequent motivation for firms to set See US Congress and Office of Technology Assessment
and control product-element standards. Ž1992..
598 G. Tasseyr Research Policy 29 (2000) 587–602

The timing of interface standards is not only oped to facilitate industry’s adoption and implemen-
important, but also often essential for small firms tation of interoperability standards.
that supply a single component to a larger system.
Often, a standard interface must be in place for small
firms to even contemplate entering the market. Thus,
poorly designed or delayed interface standards can 6. Degrees of standardization
have decidedly negative effects on market struc-
ture. 16 From both the corporate strategy and public
policy points of view, standardization is not an all-
5.3. Cost or-nothing proposition. In complicated system tech-
nologies, such as distributed data processing,
The economic cost of inadequate standardization telecommunications, or factory automation, standard-
Žeither no standard or a poorly designed one. can be ization typically proceeds in an evolutionary manner
very high. A study by Brunnermeier and Martin in lock step with the evolution of both embodied and
Ž1999. estimated that interoperability problems asso- disembodied technology. The pattern of evolution is
ciated with sharing product and engineering data determined by several factors: the pace of technolog-
impose annual costs on the US automotive supply ical change embodied in each component category;
chain totaling approximately US$1 billion. Re- disembodied technology development, which deter-
sources expended to correct or recreate data files mines the overall system architecture and organiza-
because of software incompatibilities account for tion; and changes in market structure Žand, with it,
more than 80% of the total. Delays in the introduc- the incentives and ability to force the standardization
tion of new vehicles are responsible for almost 10% process..
of the cost. Other expenses include purchases of In the case of numerically controlled machine
several vendors’ versions of software designed to tools, early total standardization of data formats
perform similar tasks and spending for data-ex- would have severely compromised the range of per-
change services. Other large US industries, such as formance attributes desired by different users in the
aerospace, have similarly large electronic data trans- machine tools they purchased. Thus, a degree of
mission requirements within their supply chains, so standardization has been optimal, at least up to this
the total cost to the US economy of inadequate point in the technology’s evolution. In other words,
interoperability standards is much greater than the complete standardization too early in the technology
estimate for the automotive supply chain alone. life cycle can constrain innovation.
One potential solution is the evolving interna- As technology-based systems become increas-
tional Standard for the Exchange of Product Model ingly important and ‘‘windows of opportunity’’ for
Data ŽSTEP., or ISO 1303. STEP is a neutral file making successful investments in the associated mar-
format intended to support computer-to-computer ex- kets continue to shrink, the relevant standards will
changes of all types of product data, from initial have to be managed. If a standard is fixed, even if it
design to maintenance requirements. However, this is competitively neutral, it will eventually act to
standard is itself very complex and is requiring a stifle the introduction of new technology into the
huge investment worldwide to develop the generic system. Alternatively, if the standard is updated fre-
set of standards. In addition, infratechnologies such quently, then version consistency Župward mobility
as test methods and software tools must be devel- of current system components. can become a prob-
lem.
For example, efficient data processing and com-
16
munications networks are possible only if standard
Delays can also occur simply because all component tech- interfaces are provided on all the communication
nologies do not evolve at the same rate. Laggard components of
the system are often characterized by chaotic entry and exit as
paths in the network. Such interfaces need to be
well as by rapid technological change, which prohibits consensus defined between application programs, data formats,
being reached on needed standards. network protocols, printer control codes, humanr
 G. Tasseyr Research Policy 29 (2000) 587–602 599

machine interfaces, and so on. But, these system in Japan is an example of the latter. 17 R & D projects
elements are all evolving at different rates and thus that involve horizontal cooperation between suppli-
need updated interfaces at different points in time. ers of system components Že.g., computer hardware
Thus, with new technologies continuously being and software firms. can promote standardization of
introduced into ever expanding networks, the pres- key product elements at an earlier point in the tech-
sure on the standards infrastructure to adapt is sub- nology life cycle. Such joint research ventures with
stantial. One reason is the need to reflect changes in standardization as an objective are occurring more
the core distributed processing technology. Another frequently, especially in IT industries.
is the requirement to adapt system management tech- Finally, government R & D can establish and
nologies such as toleration of faults, rapid Žreal time. demonstrate a backbone infrastructure, which, in turn,
response to multimedia inputs and outputs Žvoice, promotes private-sector R & D investment in stan-
image, text, and others., and human friendly inter- dards to enable effective use of this infrastructure.
faces to a constantly changing network structure as No private entity would have funded several decades
nodes are added, removed, or replaced by different of the Internet’s development, as it slowly evolved to
systems. its current state where enormous private investment
is now taking place. Industry standards such as Java
are enabling a host of applications to be developed
and implemented, but Java would not have been
7. Policy implications possible without the US government’s persistent
funding of the Internet’s predecessors.
In most cases, market dynamics seem to affect This last example exemplifies a key issue for
standardization of product elements that drive net- government R & D policy. Success as an innovator
work externalities. As long as the private-sector orig- manifests itself in a large installed base. The lock-in
inator and hence controller of the standard does not effect that results from the role of increasing returns
use this control to gain advantage in related products, to scale can be accepted as a natural and even
long-run efficiency seems attainable. On the other desirable attribute of many technology-based mar-
hand, if monopoly control by a single firm becomes kets. However, this characteristic along with the
entrenched or, conversely, if no standard emerges installed-base effect can combine to prolong technol-
Ži.e., multiple standards persist. to the detriment of ogy life cycles at the expense of newer, more effi-
economic growth, some policy action may be re- cient technologies. The standard setter can maintain
quired. control by migrating the installed base through sev-
Cotrell Ž1994, pp. 163–164. discusses how the eral incremental generations of the technology within
Japanese government tried in several ways to remedy a longer life cycle. This supply-side incentive to
the multiple PC standards problem in its domestic resist major technological change is reinforced by
industry. The SIGMA ŽSoftware Industrial General- the fact that, once a large customer base is attained,
ization and Maintenance Aids. Project and a subse- user demand for backward mobility further promotes
quent five-year extension ŽSuperSIGMA. were major an evolutionary rather than a revolutionary approach
government efforts to achieve a Unix operating sys-
tem standard and thereby deal with the economic
inefficiencies of a heterogeneous Žnonstandardized.
product structure. The project attempted to promote 17
The TRON Project has been an ongoing effort between
automated software development and create stan- Japanese industry and universities since 1981. It seeks to develop
dardized Unix-based data structures, communications standardized operating systems, which in turn will drive compati-
interfaces, and applications platforms. ble hardware development. However, the installed base effect of
An important policy implication is that research the dominant US industry standards has at least partially thwarted
this effort, as one of the largest Japanese computer firms, NEC,
projects can be funded with a pure ‘‘technology has invested so much effort in marketing MS-DOSrWINDOWSr
development’’ objective or a combined ‘‘technology UNIX machines that the company has not participated in TRON.
and standardization’’ objective. The TRON project See Cotrell Ž1994, p. 166. and Kahaner Ž1991..
600 G. Tasseyr Research Policy 29 (2000) 587–602

to change. This phenomenon is a major reason why The critical point is that a range of infratechnolo-
smaller firms appear more innovative, especially with gies and associated standards are needed by industry
respect to radical innovations; they are not encum- over the technology life cycle. Individual standards
bered by the installed base. may have modest economic impact, but the large
One possible policy response is to require, through number of standards required by the typical high-tech
regulation, equal access to applications markets so industry means that the collective impact is substan-
that the firm controlling the infrastructure standard tial. Government research program management and
does not use this advantage to restrain competition associated funding must be sufficient to provide this
ŽArthur, 1996, p. 106.. Government funding of new complex technical infrastructure in the time frames
generations of technology-based infrastructure is an- dictated by industry R & D strategies in response to
other response to dealing with the negative effects of competitive trends.
this type of standardization.
The Internet provides a great example of the basis
for a new technology life cycle that may eventually
replace the ‘‘Wintel’’ standards of the current PC 8. Conclusions
technology life cycle. 18 In fact, the ‘‘open systems’’
nature of the Internet may preclude the type of Over the past decade, the infrastructure roles of
lock-in strategy observed so often in the 1980s. standards have increased in importance because Ž1.
Netscape gave away its Internet browser and attained many new technologies are systems or networks so
70% of the market. Its objective was to obtain that increasing returns to scale can generate huge
potentially larger and more lasting profits from the economic rewards for the version of a technology
substantial value added available from software ap- that becomes the standard; Ž2. the demand for qual-
plications accessed through the browser. Similarly, ity and reliability in technologically complex prod-
America Online gained an initial substantial lead of ucts and systems requires a range of standards based
more than 4.5 million subscribers by giving away on sophisticated infratechnologies; Ž3. the systems
free services. In these and other similar cases, how- nature of critically important technologies means that
ever, ease of entry implies achieving lock-in is prob- competition is greatly affected by the degree of
lematic at best. standardization within product structures and at the
Policy implications for nonproduct standards and interfaces between components of these systems; and
investment in the associated infratechnologies are Ž4. the shortening of the average technology life
subtler. Public sector infratechnology investment is cycle has on average increased the pressure on the
not resource intensive. The semiconductor industry, standards setting process with respect to timing.
for example, spends US$4–5 billion per year on The central strategic problem of managing the
measurement equipment ŽFinan, 1997.. However, timing and content of standards is a difficult one
government support for the infratechnologies Žmea- because Ž1. many types of standards are needed in
surement and test methods, for example. and associ- today’s typical technology-based industry; Ž2. they
ated standards required to attain the productivity interact to varying degrees with one another; and Ž3.
goals of this investment is considerably less. Never- nonproduct standards, as one type of technology
theless, tight R & D budgets in most industrialized infrastructure, derive from different sources than the
nations during the 1990s have constrained infratech- industry’s core technology and thereby often conflict
nology research support. 19 with corporate strategies.
Because the US domestic market is large with
considerable internal competition in most industries,
18
American economic philosophy has allowed the
‘‘Wintel’’ refers to the dual hardware and software standards competitive dynamics of the marketplace to set prod-
controlled by Intel and Microsoft, respectively.
19
For example, NIST infratechnology research funding relative
uct-element standards. Issues of unfair advantage in
to US industry R&D spending, which NIST is charged with related markets for the firm whose technology even-
supporting, is one-half the level of 25 years ago. tually wins out as the industry standard are raised
 G. Tasseyr Research Policy 29 (2000) 587–602 601

periodically. In other cases, several competing ‘‘lo- externalities and hence total market growth, although
cal’’ standards coexist for some time, resulting in market growth may be greater in the long run if this
complaints of inefficiency. In Europe, more coordi- situation allows superior technology to eventually
nated standardization efforts in areas such as wire- dominate. Therefore, especially because the process
less communications have been undertaken to gain of standardization within a technology life cycle is
first mover advantages. This latter approach is caus- often irreversible, these and other elements of the
ing US industry and government managers to rethink dynamics of this process are key policy variables.
their laissez faire standards strategies. Finally, standardization over a technology life
More generally, industry structures and hence cycle has a dynamic character, in that different de-
long-term economic efficiency can be significantly grees of standardization are optimal at different points
affected by which firms’ technologies become indus- in both the technology’s and the industry’s evolution.
try standards and the degree to which standardization Moreover, because standards interact with each other
enables or inhibits access to the markets for a tech- in imparting their economic impacts, the process of
nology. To the extent that monopolistic control of a standardization frequently must be managed as a
standard or the existence of multiple standards pro- system.
duces economic inefficiency, the globalization of
technology-based competition can accentuate these
problems. Acknowledgements
More national economies are now capable of
competing for the dominant product-element stan- The author is grateful to Albert N. Link for
dard in technology-based industries. National gov- comments on previous drafts.
ernments often support the domestic industry’s stan-
dard over alternatives. Moreover, multinational
alliances of firms frequently pursue their versions of References
a product-element standard in emerging technology-
American Electronics Association, 1997. Cybernation. The Ameri-
based markets. Thus, multiple standards can arise can Electronics Association, Washington, DC.
and persist for some time. In such cases, multilateral Arthur, W.B., 1996. Increasing returns and the new world of
efforts should be undertaken to at least harmonize business. Harvard Business Review 74, 100–109.
these competing standards, if not to select one as a Brunnermeier, S., Martin, S., 1999. Interoperability cost analysis
single international standard at an appropriate point of the U.S. automobile supply chain. NIST Planning Report
99-1. National Institute of Standards and Technology,
in the technology life cycle. Gaithersburg, MD.
In contrast, nonproduct standards have infrastruc- Cotrell, T., 1994. Fragmented standards and the development of
ture character and, thus, unlike product-element stan- Japan’s microcomputer software industry. Research Policy 23,
dards, evolve largely through nontraditional market 143–174.
processes. Such standards are themselves based on David, P.A., 1987. Some new standards for the economics of
standardization in the information age. In: Dasgupta, P.,
infratechnologies, which often derive from a differ- Stoneman, P.L. ŽEds.., The Economic Theory of Technology
ent science and engineering base than does the indus- Policy. Cambridge Univ. Press, London.
try’s core technology. If correctly configured, they David, P.A., Greenstein, S., 1990. The economics of compatibility
are competitively neutral and have their economic standards: an introduction to recent research. Economics of
impact by increasing efficiency for all domestic sup- Innovation and New Technology 1, 3–41.
David, P.A., Steinmueller, W.E., 1994. Economics of compatibil-
pliers and customers in the particular market for the ity standards and competition in telecommunications networks.
product or service. Information Economics and Policy 6, 217–241.
The unavailability of standards at different points Department of Commerce, 1998. The Emerging Digital Economy.
in a technology’s life cycle can result in large eco- US Department of Commerce, Washington, DC, April.
nomic inefficiency. Both types of standards can also Finan, W., 1997. Metrology-related costs in the U.S. semiconduc-
tor industry, 1990, 1996 and 2001 ŽPlanning Report 98-4..
cause economic losses if they are poorly structured. Technecon Analytic Research. Prepared for the Program Of-
Multiple standards may exist for prolonged periods fice, National Institute of Standards and Technology, Wash-
of time, limiting economies of scale or network ington, DC, May.
602 G. Tasseyr Research Policy 29 (2000) 587–602

Germon, C., 1986. La normalisation, cle d’un nouvel essor, la Schumpeter, J., 1950. Capitalism, Socialism and Democracy. Hap-
documentation francais. Report to the Organization for Eco- per and Row, New York.
nomic Cooperation and Development. OECD, Paris. Tassey, G., 1982. The role of government in supporting measure-
Kahaner, D., 1991. TRON Žthe real time operating system nu- ment standards for high-technology industries. Research Pol-
cleus.. Scientific Information Bulletin 16 Ž3., 11–19. icy 11, 311–320.
Kahin, B., Abbate, J. ŽEds.., 1995. Standards Policy for Informa- Tassey, G., 1992. Technology Infrastructure and Competition
tion Infrastructure. The MIT Press, Cambridge, MA. Position. Kluwer, Norwell, MA.
Link, A.N., Scott, J.T., 1998. Assessing the infrastructural needs Tassey, G., 1997. The Economics of R&D Policy. Quorum Books,
of a technology-based services sector: a new approach to Westport, CT.
technology policy planning. STI Review 22, 171–207. Tassey, G., 1999. R&D trends in the U.S. economy: strategies and
Link, A.N., Tassey, G., 1987. Strategies for Technology-Based policy implications. Planning Report 99-2. The National Insti-
Competition. Lexington Books, Lexington, MA. tute of Standards and Technology, Gaithersburg, MD.
Morris, C.R., Ferguson, C.H., 1993. How architecture wins the Ulrich, K., 1995. The role of product architecture in the manufac-
technology wars. Harvard Business Review 71, 86–96. turing firm. Research Policy 24, 419–440.
Putnam, Hayes, Bartlett, 1982. The Impact of Private Voluntary US Congress, Office of Technology Assessment, 1992. Global
Standards on Industrial Innovation, Vol. 2. The National Bu- Standards: Building Blocks for the Future ŽTCT-512.. US
reau of Standards, Gaithersburg, MD. Government Printing Office, Washington, DC.