2. Industry Characteristics Affecting Technology Adoption

This section presents the results of our research to develop a set of indicators that measure the likelihood of technology adoption. These indicators are subsequently used in our analysis of a specific ATP-funded technology, which was first developed for the automobile industry. The indicators developed in the report are derived from publicly available, broadly defined, industry-level data sets covering the manufacturing sector.

A significant amount of economic research has found that some industries adopt new technologies faster than others. For example, economic industrial organization studies based on the structure-conduct-performance (SCP) model have found that industry characteristics such as the number and size distribution of firms and quality of competition affect rates of technology adoption. Since the SCP model is widely used for organizing and analyzing industry characteristics and behavior, we apply it to develop a set of indicators that assess the likelihood of technology adoption.

According to the SCP model (Figure 1), market structure affects market conduct, which in turn affects market performance. Industry characteristics such as raw materials, technology, business attitudes, product substitutes, the ability of new firms to enter the market, and cyclical and seasonal demand determine the structure of industry, and in particular the numbers of sellers and buyers. This structure of sellers and buyers affects selling and buying conduct, such as production and pricing strategies, research and innovation, pricing behavior, advertising, investment, and legal tactics. Market conduct, in turn, affects the overall performance of the industry-that is, price and production levels. Public policy (such as taxes, subsidies, and regulations) affects both industry market structure and conduct. In the long run, these conduct and performance activities affect the market structure. While the conclusions drawn from early SCP models generally apply to all industries, recent SCP studies of specific industries augment the research with data specific to those industries.⁽¹⁾

Using the SCP model, different industry characteristics can be logically organized and analyzed according to their effects on technology adoption. A TAI is developed based on an industry characteristic with broadly defined data (available at 4-digit SIC or 6-digit NAICS and lower resolutions) that meet the criteria laid out in Section 1 of this report. The characteristics and developed indicators are grouped under market structure (Section 2.1), conduct (Section 2.2), public policy (Section 2.3), and history (Section 2.4). The discussion of each measure includes both a description of the theoretical rationale for its use in assessing the likelihood of adoption and recently published research on the measure. The sections also address proposed measures that were candidates for inclusion as TAIs.

Figure 1 . The Structure-Conduct-Performance Model

Source: Scherer and Ross (1990), p. 5.

The broadly defined comprehensive industry data are used to demonstrate the industry selection process in this case study. Using narrowly defined specific industry data, we were able to more accurately calculate and apply the TAI measures in the selected industry.

Section 3 applies the broadly defined TAI data to two case-study candidates as a screening mechanism. The narrowly defined TAI data are developed and applied to only the selected case study application. Industry concentration and patent measures are recalculated using narrowly defined industry data. Narrowly defined data on TAIs that are unique to the chosen case study, such as the history of adoption and public policy, are also utilized.

2.1 Market Structure

An exhaustive search of the business and economics literature generated an abundance of indicators connecting industry structure to technology adoption. According to our established criteria, industry structure measures developed in this report had to be quantitative, be publicly available, measure each industry in the manufacturing sector individually, be theoretically related to the motivation to adopt technology, and have empirical evidence in the literature to support the connection to technology demand. For example, we identified concentration measures as most useful because they are available at the 6-digit NAICS industry level and are related to technology adoption. Other measures of industry structure that have a close theoretical relationship with technology adoption would require the collection and development of new data for each industry and are thus not developed in this report.

2.1.1 Industry Concentration

Concentration-the number of sellers-is one definition of competition, and affects conduct.⁽²⁾   Examples of conduct are pricing behavior, R&D, plant investment, and technology innovation. Industry concentration therefore affects the demand for technology innovation and adoption and therefore the likelihood of technology adoption.

Industry concentration measures, as the name suggests, are used to benchmark industries by the market share (percentage of sales) held by the largest firms. Economic theory and empirical research suggest that highly concentrated industries, i.e., those in which most of the market is held by few firms (or by a single firm), face little pressure to adopt new technologies, thereby slowing innovation.⁽³⁾   On the other hand, low concentration is not an optimal environment to encourage innovation. As Scherer and Ross (1990, p. 637) explain: "Up to a point, increased fragmentation stimulates more rapid and intense support of R&D. . . . But when the number of firms becomes so large that no individual firm can appropriate quasi-rents sufficient to cover its R&D costs, innovation can be slowed or even brought to a halt."  Scherer and Ross note that industries characterized by mid-sized, competitive firms tend to adopt technology more than industries characterized by either many small firms by a few, very large firms (oligopolies). Small firms in highly atomistic industries do not have the capital or opportunity to adopt new technology, and severe oligopolies and monopolists face little pressure to adopt. In short, economic research suggests there is an optimal range of industry concentration in which technology adoption is most likely.⁽⁴⁾

There are two commonly used measures of industry concentration: the n-firm concentration ratio (CR), where n is the number of the largest firms included in the measure, and the Herfindahl-Hirschman Index (HHI).

The n-firm concentration ratio is the sum of the percentage market shares held by the n largest firms, or

. (1)

n is the number of firms (e.g. 4 or 8) and s_i is the market share of firm i. The market share is measured as a percentage of sales. The CR4 (n=4) is the sum of the market shares of the largest four firms, and the CR8 (n=8) is the sum of the percentage market shares of the largest eight firms. An industry with exactly four firms has a CR4 of 100; while an industry with 10 equally sized firms has a CR4 of 40.

The HHI is defined as the sum of the squared percentage market shares of all firms in the industry, or

(2)

N is the total number of firms in the industry and s_i is the market share of firm i. The HHI measure approaches 0 as all the firms in an industry approach zero market shares (theoretically perfect competition). The HHI has a maximum value of 10,000 when there is only a single firm-i.e., a monopolist. For a theoretical industry with 10 equally sized firms, the HHI would be 1,000 (= 10 x 10²).

The HHI is a more comprehensive and revealing measure of industry concentration. Because it uses the square of market share, and includes the share of every firm, it is able to show differences in concentration between industries even when the CR4 measures (or CR8 measures) are identical. For example, industry A consists of eight firms with the following concentrations: 65, 5, 5, 5, 5, 5, 5, 5. The CR4 is 80 and the HHI is 4,400. Industry B consists of eight firms with the following concentrations: 20, 20, 20, 20, 5, 5, 5, 5. The CR4 is 80 but the HHI is only 1,700. The HHI, unlike the CR4, captures the fact that Industry B is less concentrated.

According to Shepherd (1987), a CR4 between 40 and 60 identifies an industry with firm concentrations optimal for competitive behavior conducive to adopting new technologies. Scherer and Ross (1990) identify the bounds for the CR4 as 45 and 60. Scherer and Ross also report findings that the optimal CR8 for competitive industry is 70 (which roughly corresponds to a CR4 of 50 in the U.S. economy). We use a range of 10 points on either side of the optimal CR8 value as the range of values optimal for technology adoption.

The Federal Trade Commission (FTC) uses both the CR and the HHI to assess the extent to which a proposed merger will affect competition in that industry.⁽⁵⁾   According to the U.S. Department of Justice (DOJ), a market with an HHI less than 1,000 is considered unconcentrated, between 1,000 and 1,800 moderately concentrated, and over 1,800 highly concentrated.⁽⁶⁾   The DOJ is likely to challenge mergers that increase the HHI more than 100 points when the HHI index is greater than 1,800. A middle range-in which challenges depend on the increase in the HHI-occurs when the HHI is between 1,000 and 1,800. The FTC is unlikely to challenge mergers when the HHI is below 1,000. We use the DOJ definition of a moderately concentrated market (HHI between 1,000 and 1,800) to approximate the optimal industry concentration for technology adoption.

To plot and compare U.S. industry concentrations, we used the U.S. Census Bureau's 1997 Economic Census data on four (CR4) and eight (CR8) concentration ratios and the HHI index, for each six-digit NAICS code manufacturing industry. The CR4, CR8, and HHI values of the lawnmower-engine industry and the airplane engine industry are analyzed in Section 3.

2.1.2 Other Market Structure Measures

Industry concentration measures may be improved upon as indicators of technology adoption. First, they may be improved by more narrowly delineating the relevant industry. That is, by defining the industry more precisely than the six-digit NAICS level, more detailed economic information about technology adoption can be obtained. Analysis of some industries and markets may require a high level of resolution. Second, while concentration ratios are moderately correlated with technology adoption, measures with higher correlations could be developed. For example, entry and exit conditions-the ability of new firms to enter and leave the relevant industry-influence technology adoption even in highly concentrated industries. If entry is easy, even firms in highly concentrated industries are likely to adopt technology as a defense against new entrants.

New measures can be collected or constructed specifically for the industry under analysis. For example, according to Atkinson and Court (1998) at the Progressive Policy Institute (PPI), the relative number of new, fast-growing entrepreneurial companies in an industry is correlated with innovation and adoption of new technologies. The PPI researchers defined fast-growing companies as "companies with sales growth of at least 20% per year for four straight years," and notes that the number of initial public offerings (IPOs) reflects the increase in the number of fast-growing entrepreneurial companies. "Economic churn"-the replacement of old firms by new, more efficient firms-is also correlated with technology adoption. The U.S. Census Bureau provides a single measure of industry births and deaths for the U.S. manufacturing sector as a whole but not for individual industries. These measures-fast-growing companies, number of IPOs, and establishment births and deaths-could provide additional information about technology adoption, but require further data development to apply to specific industries.

2.2 Market Conduct

Industry conduct measures developed for this report had to be quantitative, be publicly available, measure each industry in the manufacturing sector individually, be theoretically related to the opportunity to adopt technology, and have empirical evidence in the literature to support the connection to technology adoption. The TAIs on market conduct could be thought of as addressing the supply of technology innovation.

Measures of industry conduct include both direct and indirect measures of technology adoption. For example, technology adoption itself is market conduct. Other market conduct measures also have an impact on technology adoption. Of these, several measures, such as the number of RJVs and the number of patents, currently have broadly defined industry data sets available at the 4-digit SIC and higher resolutions. It is possible to improve upon these measures by increasing their resolution and by developing data for new measures.

2.2.1 Patent Counts and Technology Adoption

The mechanism by which the number of patents influences technology adoption is less theoretically founded than connections with concentration ratios and HHI indices. In this report, we interpret as an indicator supply available to industry, influencing market conduct through amount industry.⁽⁷⁾

Griliches, Hall, and Pakes (1991) studied the relationships among R&D, patenting activity, and market value. Research continues on ways to use patent data to explain innovation. Recent research on "hot" patents and patent clustering has strengthened their empirical usefulness and could be incorporated into these indicators if broadly defined industry data become available (Breitzman 2001). (The duration of the payment of patent maintenance fees may relate to patent importance, and could be incorporated into patent indicators in the future.) The number of patents issued for potential application in an industry can be used as an indicator of technology adoption in that industry or in other "user" industries. The increase in innovation happens through a "supply-push" model of technology adoption, in which more patents result in the possibility of greater technology advances, as in the case of RJVs.

The mechanism by which the number of patents influences technology adoption is less theoretically founded than the connections of technology adoption with concentration ratios and HHI indices. Consequently, the interpretation of the number of patents is unclear. In this report, we use the total number of patents granted over the last five years as an indication of the amount of new technology available.⁽⁸⁾

The output of patents, publications, citations, and other technology developments can be measured using data from the U.S. Patent and Trademark Office (USPTO).⁽⁹⁾   The translation from the US Patenting Classification System (USPCS) to the SIC code system is straightforward, allowing industry-specific patent counts, though not at a high resolution. For the two industries under evaluation, patent data are available at the 3-digit SIC level. For any specific industry, it is possible to hone the patent-count measure to finer detail and greater accuracy by searching out patents directly applicable to that industry. The total number of patents as reported by the USPTO in SIC codes 13 to 39 from 1996 to 2000 are depicted in Appendix B.

The Patenting Trends database (USPTO 2001) provides two types of patent counts. The "whole" counting method matches the USPCS to all relevant SIC codes as explained in the Patenting Trends documentation:

The "fractional" counting method divides the USPCS patents equally among all the matched SIC fields. In other words, the patent found in three product fields would generate one-third of a patent in each SIC field. This has the effect of diluting the weight of broadly applicable patents and increasing the weight of patents dedicated to a single industry.

We use the number of patents reported in the USPTO database over the last five years to indicate the supply of innovation available to an industry. All else being equal, an industry with fewer patents is more attractive to developers and sellers of new technology. A developer of new technology would prefer to target an industry with higher demand for technology and fewer patents rather than an industry with less demand for technology and more patents.

2.2.2 Research Joint Ventures (RJVs), Inovation, and Technology Adoption

As was the case with patents, the mechanism by which the number of RJVs influences technology adoption is less theoretically founded than the connections of technology adoption with concentration ratios and HHI indices. Similarly, we interpret the number of RJVs as indicating the supply of technology available to an industry.

RJVs can increase the overall R&D in an industry in two ways. First, under certain conditions, economic theory shows that RJVs increase the level of research and development in an industry, thereby increasing the “supply” of new technology. These conditions are that the combined returns to R&D exceed the private returns of the joint venture members (Stenback and Tombak 1997). Under these conditions, the absence of a joint venture is a “market failure,” wherein the optimal level of R&D is not attained by the market. RJVs that address suboptimal levels of R&D in competitive markets are often justified on this basis.⁽¹⁰⁾ Second, the overall R&D in an industry can be increased when a legal restraint (such as anti-trust legislation) that hinders the formation of RJVs is lifted for certain RJVs that survive regulatory scrutiny. We use the survival of regulatory scrutiny as an indicator of RJVs that increase the overall level of R&D in an industry.

The Progressive Policy Institute (PPI) provided empirical evidence that collaboration and networks (such as RJVs) create more technology and innovation.⁽¹¹⁾ Causation could also run in the opposite direction—that is, industries characterized by innovation and technology adoption are conducive to RJV formation.⁽¹²⁾ Therefore, applying the theoretical and empirical observations that RJVs are conducive to the creation of new technology, we are able to use the number of RJVs in an industry as an indicator of technology adoption in that industry. The increase in innovation happens through a “supply-push” model of technology adoption, in which more and greater technology advances developed by the RJVs become so compelling that they attract firms to adopt them.

To measure this relationship between R&D and innovation, the nonprofit organization Council on Competitiveness has developed the Innovation Index, which measures the relationship between industry R&D (as measured by patents) and employment in research and development, expenditures on research and development, percentage of R&D expenditures funded by private industry, and percentage of R&D performed by universities. Unfortunately, the Innovation Index is available only at the national level and not for individual industries. Estimates of the Innovation Index at a national level indicate a strong connection between R&D activities and innovation (patents).

The Collaborative Research (CORE) database⁽¹³⁾ collects data on RJVs from Federal Register filings.⁽¹⁴⁾ Parties involved in RJVs who wish to gain protection under the National Cooperative Research Act and the National Cooperative Research and Production Act must file public notice of the RJV in the Federal Register. The CORE database consists of firms that have announced their intentions in the Federal Register, which indicates a belief that the RJV is likely to be one that would otherwise not be permitted in the market, and one that would contribute positively to the current level of R&D because it has survived government scrutiny. The number of RJVs from the SIC codes 20 to 39, as reported in the CORE database, are depicted in Appendix C.

We use the number of RJVs reported in the CORE database as an indicator of the supply of technology available to an industry. All else being equal, an industry with fewer RJVs is more attractive to developers and sellers of new technology. A developer of new technology would prefer to target an industry with higher demand for technology and fewer RJVs rather than an industry with less demand for technology and more patents.

2.2.3 Other Industry Conduct Measures

In addition to RJVs and patents, there are other statistics that measure key industry characteristics relating to innovation and technology adoption. For example, empirical research of stock market reaction to RJV formation indicates that the formation of certain types of RJVs—such as that between a large and a small company—result in appreciation of stock prices.⁽¹⁵⁾   Koh and Venkatraman (1991), McConnell and Nantell (1985), and Mohanram and Nanda (1995) demonstrate that the market reaction depends on the type of RJV and the characteristics of the involved companies. Empirical studies have addressed different RJV configurations and the impacts on R&D. For example, a RJV between a large and a small company combines the advantages of each firm size in technological innovation (Rothwell 1989). Technology adoption information obtained from RJV data could therefore be improved by controlling for the RJV configuration.

The amount of royalty revenues from technology licensing was used by Degnan (1999) as a proxy for the success of innovative activities resulting from U.S. R&D efforts. Data on U.S. royalty income were collected and compared to the national economic growth attributable to technological advances. Degnan found a strong correlation between R&D expenditures and innovation and economic prosperity. No industry-level R&D data about royalty revenues were found for inclusion as indicators of technology adoption.

2.3 Public Policy

As illustrated in Figure 1, public policy influences industry conduct via taxes, subsidies, regulations, and price controls. For example, a government subsidy would increase the production of the subsidized goods, whereas taxes would reduce the production of taxed goods. Regulations that raise the cost of the finished product act like taxes. Regulations that permit less-costly production alternatives would be similar to subsidies.

In the context of the supply-and-demand framework used in this report, public policy can act either as a demand or as a supply factor for technology innovation. Some policies (such as the limited suspension of antitrust regulations) can encourage the development of new technology supply. Other policies, such as specific regulations, tend to increase the demand for new technology. For example, air bag requirements in automobiles have required the implementation and refinement of sensing systems.

Regulation, as a public policy option, fits within the SCP model. It specifically affects structure and conduct. Regulations affect every industry, but each regulation varies in impact. There is no standardized, quantitative way to measure regulation or other aspects of public policy across industries. In a case study, specific interpretations of applicable public policies are required. The impact of regulation on this case study is discussed in Section 3.

2.4 Historical Patterns of Technology Adoption

The SCP model also takes into account industry-specific historical trends in market structure, as well as conduct, performance, and public policy. Industry-specific SCP trends can be used to assess the likelihood of technology adoption. In addition, the history of past technology adoption, as defined in an industry, provides valuable information about the likelihood of technology adoption in that industry in the future. There is no standardized way to account for past technology adoption with a broadly defined industry measure. In a case study, specific interpretations of what constitutes technology adoption, and time-trend data on those observable characteristics, are required. The impact of the history of technology adoption on this case study is discussed in Section 3.5.

2.5 Summary of TAI Measures Used

In this report, the HHI and four- and eight-firm concentration ratios (CR4 and CR8) are used as indicators of demand for technology adoption. The number of patents and RJVs are used as indicators of supply of technology adoption. Public policy, through regulation, is another driver of technology adoption in this report. An examination of the history of technology adoption adds empirical observations to the analysis. Section 3 describes the application of the available data on these TAIs to the case-study candidates.

____________________
1. This expansion of structure-conduct-performance analysis to include detailed industry-specific data has been
termed “new industrial organization.” For a description, see Mansfield (1977).

2. The concentration of buyers is also important, but in practice this is impossible to find out for all industries.

3. For example, see Shepherd (1987). A focus on market concentration leaves out the effect of entry and exit
conditions. The inclusion of such factors is explained in section 2.2.1.

4. In case studies, adding industry-specific information on the impact of concentration will improve the interpretation of the effect of concentration on the adoption of technology. For example, technology adoption should occur at lower concentrations in emerging industries driven by innovation (such as life-cycle work most suited to manufacturing). This is discussed in section 2.1.2.

5. Dr. Michael McFalls (1997) of the FTC provides an in-depth discussion of the ways in which the FTC assesses market power during an analysis of joint ventures.

6. Horizontal Merger Guidelines issued by the U.S. Department of Justice and the Federal Trade Commission, section 1.51 on the DOJ internet site www.usdoj.gov/atr/public/guidelines/horiz_book/15.html.

7. While the overall number of patents is an important indicator of technology supply, this measure does not address the marginal value of an additional patent.

8. In this report, the last five years of patents are combined to create the total number of patents available. Other ways of measuring patents include the number of highly cited patents and the rate of patenting.

9. U.S. Patent and Trademark Office, Patenting Trends in the United States 2000. CD-ROM. December 2001.

10. The assumptions made in a study—about the marketplace and the nature of the joint venture companies—are often the deciding factor in determining whether RJVs result in more or less R&D.

11. The Progressive Policy Institute (PPI) is a non-profit think tank that has researched the characteristics of the “new economy,” one of which is rapid technological innovation.

12. Robert P. Lynch (1989) lists several industry conditions that are conducive to JV formation. One important condition is “rapid changes occur in technology,” indicating technology adoption).

13. The CORE database is funded by the National Science Foundation (NSF) and maintained by Dr. Albert Link of the University of North Carolina.

14. Free internet access to the Federal Register is available at www.access.gpo.gov/nara. From this location, notices appearing in the Federal Register may be searched.

15. Stock prices increase based on an expectation of a successful joint venture leading to greater profit per share and/or increased dividends. Successful joint ventures produce new technology innovation and could, given procompetition RJVs, also produce demand for technological innovation.

Go to Section 3 or return to Table of Contents.

Date created: June 11, 2003
Last updated: August 3, 2005

ATP website comments: webmaster-atp@nist.gov  / Technical ATP inquiries: InfoCoord.ATP@nist.gov.

NIST is an agency of the U.S. Commerce Department
Privacy policy / Security Notice / Accessibility Statement / Disclaimer / Freedom of Information Act (FOIA) /
No Fear Act Policy / NIST Information Quallity Standards / ExpectMore.gov (performance of federal programs)