Intangible resources and technology adoption in manufacturing firms

Research Policy 41 (2012) 1607–1619

Contents lists available at SciVerse ScienceDirect

Research Policy journal homepage: www.elsevier.com/locate/respol

Intangible resources and technology adoption in manufacturing firms Jaime Gómez 1 , Pilar Vargas ∗ Universidad de La Rioja, Departamento de Economía y Empresa, Edificio Quintiliano, Cigüe˜ na 60, 26004 Logro˜ no (La Rioja), Spain

a r t i c l e

i n f o

Article history: Received 9 November 2009 Received in revised form 5 March 2012 Accepted 28 April 2012 Available online 21 May 2012 Keywords: Intangibles Technology adoption Diffusion Complementarities

a b s t r a c t Our objective in this paper is to analyse the determinants of the use of advanced manufacturing technologies in manufacturing firms. We go beyond more traditional approaches and consider the role of complementarities in technology adoption at two levels. First, we adapt Teece’s (1986) framework to study the incentives to use new technology that stem from investments in R&D, human capital and advertising. Second, we analyse whether technology use is conditioned by a system effect that arises from the use of related technologies. We test our hypotheses on a representative sample of manufacturing firms in Spain. Our results fully support the idea that R&D investments increase the likelihood of technology use, but only offer partial support for human capital and advertising investments. Export intensity, being part of a business group and epidemic effects are also important determinants of adoption. © 2012 Elsevier B.V. All rights reserved.

1. Introduction Our objective in this paper is to understand the role of intangible resources in technology adoption. The idea that intangible resources contribute to the explanation of technology diffusion among firms has its roots in some analyses of the impact of information technology (IT) on performance. A part of this literature has argued that IT are strategic necessities in the sense that they have to be adopted in order not to suffer from competitive disadvantages, but that they cannot produce sustainable competitive advantages (Clemons and Kimbrough, 1986; Clemons and Row, 1991). Recent research on complementarities has shown that the performance of IT frequently depends on their interaction with organizational elements (Powell and Den-Micallef, 1997). In particular, the application of the resource-based view of the firm suggests that, in order to achieve inimitability, firms have to combine IT with resources protected by isolating mechanisms such as tacitness, causal ambiguity and time compression diseconomies2 (Barney, 1991; Dierickx

∗ Corresponding author. Tel.: +34 941 299 572; fax: +34 941 299 393. E-mail addresses: [email protected] (J. Gómez), [email protected] (P. Vargas). 1 Tel.: +34 941 299 373; fax: +34 941 299 393. 2 Time compression diseconomies make reference to the idea that the returns diminish when one input, in this case time, remains constant (Dierickx and Cool, 1989). Intangible assets, such as the ones created through investments in R&D or advertising are likely to be subject to them. Dierickx and Cool (1989, p. 1507) provide an example for R&D investments: “In the case of R&D, the presence of time compression diseconomies implies that maintaining a given rate of R&D spending over a particular time interval produces a larger increment to the stock of R&D know-how than maintaining twice this rate over half the time interval”. 0048-7333/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.respol.2012.04.016

and Cool, 1989). In other words, research on the impact of IT on performance suggests that firms possessing resources protected by isolating mechanisms should have higher incentives to adopt IT. Despite this, research on technology diffusion tends to focus on the role of some “classical” variables that characterise firms, such as size and investments in absorptive capacity (see, for example, Karshenas and Stoneman, 1993). In this paper, we shift the focus of our attention to the idea that the complementarities between resources and new technologies should increase the incentives of firms to adopt. Taking into account the arguments provided by the resource-based view (Barney, 1986; Wernerfelt, 1984), we examine the role of intangible resources through the framework provided by Teece (1986, 2006). Given the weak appropriability of new technologies, we focus on difficult-to-obtain assets to explain the different stimuli that firms may have to use new technologies. Accordingly, we develop hypotheses that relate investments in research and development, in personnel qualifications, and in advertising with the likelihood of a firm using a technology. These resources not only interact with technologies in order to create firm value, but they can also help the firm to appropriate the returns of its investment in them. We test our hypotheses on a representative sample of Spanish manufacturing firms. Our main focus is on the use of four process technologies (numerically controlled machines, robotics, computer aided design and flexible manufacturing) that have previously been considered as being part of what is termed Advanced Manufacturing Technologies (AMT). Using information on several technologies is in line with the literature, particularly the theory of Milgrom and Roberts (1990) that complementary technologies form a production system, which implies that adoption of the technologies should be positively correlated. This allows us to explore the hypothesis

1608

J. Gómez, P. Vargas / Research Policy 41 (2012) 1607–1619

that the use of any technology could also be explained by the use of other technologies with which it forms a system (Colombo and Mosconi, 1995; Sinha and Noble, 2008) and reinforces the idea that complementarities have a key role to play in the understanding of performance differences and adoption behaviour. To further check this system effect, we use a second sample in which we test whether the four process technologies are related to certain e-business technologies. More precisely, we use information on whether firms have adopted supplier-to-business, business-tobusiness and business-to-consumer technologies to estimate the complementarities arising within and among different groups of technologies. The use of different technologies will also help us to understand the degree to which our predictions can be generalized. From a methodological point of view, a problem introduced by the consideration of the role of complementarities is that the decisions to adopt several technologies could be related. In such a setting, the independent estimation of the models explaining the use of each technology would be inadequate, given that the decisions to adopt are potentially correlated. A solution to this problem is the estimation of a multivariate model, which takes the potential correlation into account. Testing for differences in correlations among different types of technologies makes it possible to evaluate the degree to which the system effect is present in our data. The rest of the paper is organized as follows. In the next section, we integrate Teece’s (1986) framework, the resource-based view and the literature on technology diffusion in order to develop four hypotheses on the role of complementary assets. A section that explains the methodology, describes the sample and defines the variables follows. After this, we present the results and perform some robustness tests in order to confirm them. The final section concludes and discusses the implications for technology management. 2. Intangible resources and technology adoption Our objective in this paper is to use the idea of complementarity to explain technology adoption by manufacturing firms. The concept of complementarity refers to the idea that if the investments in one activity are increased, the returns on investment of others would also increase (Milgrom and Roberts, 1995). The analysis of complementarities between organizational elements is receiving increasing attention in the economics and management literatures (Ennen and Richter, 2010). Different papers have studied the interactions arising between firm resources, practices, policies, strategies and environmental factors.3 The idea of complementarities may help us to understand the factors that underlie a firm’s technology adoption. In fact, it is surprising that, given the close relationship between innovation and technology adoption (Battisti et al., 2009; Koellinger, 2008; Santamaría et al., 2009), research on the two has progressed with relative independence. Research has shown the importance of new technologies on product and process innovation. For example, some AMT have been related to quality and flexibility strategies. These strategies are characterised by high innovation frequency and product R&D, among other features (Parthasarthy and Sethi, 1992, 1993). Similarly, Koellinger (2008) shows that process and product innovations are enabled by the use of IT. In this paper, we use Teece’s (1986) framework to integrate the two lines of research and to explain the different incentives firms

3 For example, Cassiman and Veugelers (1999) study the complementarities between external and internal R&D investments, whereas Belderbos et al. (2006) assess the complementarities arising in R&D cooperation strategies. A review of the literature on complementarities and on the types of factors considered may be found in Ennen and Richter (2010).

may have to use technologies. Following Teece (1986), the capacity of a firm to appropriate the returns from innovation mainly depends on two types of factors (Teece, 1986): the appropriability regime and the possession of complementary assets. The appropriability regime is defined by (1) legal and (2) technological factors. Additionally, the rents accruing to the innovator also depend on the possession of complementary assets. By complementary assets, Teece (1986) understands all the resources that need to be jointly used with the innovation in order to exploit it. They include, for example, the manufacturing, distribution and service resources that are needed to operate the different stages of the value chain. Teece’s framework could be easily adapted to understand the incentives of firms to adopt new technologies. The fact that new technologies are freely available in the market (there are no legal barriers protecting the adopter) focuses attention on complementarity assets.4 Despite the weak appropriability regime of new technologies, the returns from their use could be captured if the firm possesses complementary resources that are difficult to obtain, such as the ones created by investments in research and development and advertising. In fact, research has shown that the link between IT and profitability could be explained through the consideration of complementary resources under the control of the firm (Powell and Den-Micallef, 1997). However, their use as determinants of adoption has been scarce. The existence of difficult-to-imitate complementary resources could be an indicator of the stimuli that a firm has to adopt a new technology, given that they could provide the basis for competitive advantage. Selecting the types of resources more likely to be combined with new technologies to provide competitive advantages is a difficult task. Although both tangible and intangible assets play an important role in creating value, intangible assets frequently present time compression diseconomies and they are typically tacit and hard to codify (Conner and Prahalad, 1996; Kogut and Zander, 1992). Additionally, they are likely to be traded in imperfect factor markets (Barney, 1986) and are not consumed by their use (Collins and Montgomery, 1998). Teece (2000) suggests that a firm’s superior performance depends on its ability to defend and use the intangible assets it creates. Accordingly, we argue that technology adoption is conditioned by the possession of certain complementary resources that are difficult to acquire or copy, namely, technological, human and marketing resources.5 Similarly, we also analyse the presence of complementary relationships between technologies. As mentioned in the Introduction, we focus on production technologies and, in particular, the group of technologies that has been termed “advanced manufacturing technologies”.6

4 In fact, this has led some researchers to conclude that, in the long run, technology adoption might only provide competitive parity (Barney, 1991). Some authors have enunciated this as the strategic necessity hypothesis (Clemons and Kimbrough, 1986; Clemons and Row, 1991). Following this hypothesis, a focal firm would have incentives to adopt a valuable new technology. However, this would create incentives in rival firms to invest in the same technology in order not to obtain below normal profits. As long as new technologies are freely available in the market, the long-run result would be that rival firms would erode the rents accruing to first adopters, leading to competitive parity. 5 We are not claiming that difficult-to-imitate complementary resources are the only ones likely to interact with new technologies to provide competitive advantage. In fact, Rivkin (2000) points out that complex systems could be inimitable, even if their elements are not. 6 As suggested by an anonymous referee, we expect differences between the effects of each of the variables depending on the type of technology considered. For example, Rogers and Shoemaker (1971) argue that the diffusion of innovations depends on their attributes, namely, relative advantage, compatibility, complexity, trialability and observability. For example, we would expect the effect of technological resources to be more important in those cases in which the technology is more complex. Therefore, the magnitude of the associated effect should be higher for flexible manufacturing than for the other technologies. Similarly, we would expect a higher effect of the qualifications of human resources in those cases in which the

J. Gómez, P. Vargas / Research Policy 41 (2012) 1607–1619

2.1. Technological resources and technology adoption Investments in research and development contribute to the knowledge base of the firm, increasing its capabilities, and may also result in process and product innovations. Like other intangible resources, they tend to be tacit, idiosyncratic and deeply embedded in the firm (Winter, 1987), which makes it difficult for competitors to copy them. Imitation may be also complicated by the fact that this knowledge may be path dependent and subject to time compression diseconomies (Dierickx and Cool, 1989). There are different ways in which the knowledge base of the firm interacts with advanced technologies increasing the possibility to create complex combinations of resources (Denrell et al., 2003; Rivkin, 2000). On the one hand, the knowledge base of the firm has frequently been related to technology adoption through the absorptive capacity concept. A firm’s absorptive capacity is defined as the “ability of a firm to recognize the value of new, external information, assimilate it and apply it to commercial ends” (Cohen and Levinthal, 1990, p. 128). Although absorptive capacity tends to be related to adoption through the greater ability of firms to understand valuable information, we should emphasize its role in the exploitation of new technologies. In other words, once a new technology is adopted, its use would require relating the knowledge embodied in the innovation to the internal processes taking place inside the organization. Therefore, a better knowledge base could help firms to exploit new technologies more effectively and to capture the results. For example, this knowledge could be used to better understand and to refine a firm’s productive process. On the other hand, new technologies may also help the firm to take advantage of the results of the innovation process. In fact, research has shown that new technologies frequently enable process and product innovations (Koellinger, 2008). In this sense, several authors have considered that the use and appropriation of AMT also accelerate the resolution of more radical problems, opening the pace for a greater degree of novelty in the innovation (Amara et al., 2008; Wuyts et al., 2004). Therefore, new technologies may provide an adequate channel for converting the knowledge base of the firm into profitable products and services. Hence, they are more likely to be adopted when the firm possesses a larger knowledge base. Hypothesis 1. Firms with more technological resources are more likely to use a new technology. 2.2. Human resources qualifications and technology adoption The use of AMT could be also related to workers’ qualifications.7 The most common theory is the skill-bias technical change hypothesis, which states that there is a relationship of complementarity between new technologies and skilled workers because the latter are the only ones able to fully implement these technologies (Piva et al., 2005, p. 143). Doms et al. (1997) maintain that the positive correlation between technologies and the attributes of the workforce may be explained using three arguments. First, at a general level, AMT increase the level of automation in a factory. Workers using these machines must, at least, have reasonable language,

technology is more complex. However, the fact that we do not have information on all these five dimensions (or other that could be considered as important) precludes as from being more precise in our arguments. 7 Some papers in the agricultural economics literature have used education to explain adoption. However, its use seems to be more related to the concept of absorptive capacity. For example, Huffman (1974, p. 85) refers to an allocative effect, understood as “the human agent’s ability to acquire, decode, and sort market and technical information efficiently”. We attempt to emphasise the idea that these investments are important for exploiting and appropriating the returns to the technology.

1609

reading and basic math skills and they should also able to deal with higher levels of abstraction and to act quickly on complex information (Parthasarthy and Sethi, 1992). Thus, more automated plants will employ relatively more educated and skilled workers. Second, the introduction of a more technology-specific level may affect the organization of the workforce and will require skilled operators and technicians who replace skilled craftspeople but also less skilled workers. Finally, many of these technologies require qualified support staff to install and maintain them. It could be argued that these abilities have commodity-like character, given employee mobility, not constituting a difficult-toimitate resource. However, researchers have maintained that more qualified workers are more likely to invest in firm-specific investments. The reason is that their mobility opportunities are less than for employees with low qualifications and their activities tend to imply the use of their intellectual capabilities in idiosyncratic tasks (Vicente-Lorente, 2000). This makes it highly likely that more qualified workers invest more in firm-specific abilities than less qualified ones. More qualified workers also provide other abilities related to innovation and management that are important in the use of new technologies. First, as mentioned above, new technologies frequently enable process and product innovations (Koellinger, 2008). Education contributes to the innovation process by increasing a “person’s capacity to think systematically and creatively about techniques” (Wozniak, 1984, p. 71). Second, education may also be related to management skills, in other words, the increasing ability of educated workers to effectively integrate new technologies into the activities of the firm. In fact, given that technology-related knowledge may be contracted and is not firm- (but technology-) specific, it has been argued that management skills are the only likely source of competitive advantage: they are path dependent, they tend to be tacit and firm-specific and they may be socially complex (Mata et al., 1995). Hypothesis 2. Firms with more highly qualified personnel are more likely to use a new technology. 2.3. Marketing resources and technology adoption Complementarities are also present in marketing activities (Teece, 1986). The literature on management understands marketing investments as firm-specific assets (see, for example, Balakrishnan and Fox, 1993; Vicente-Lorente, 2001). Like technological resources, investments in marketing create intangible resources, such as reputation, brand image and closer relationships with customers. These resources are frequently difficult to imitate: they may be subject to time compression diseconomies (Dierickx and Cool, 1989) or be socially complex (Barney, 1991). Investments in marketing activities help the firm to develop a good reputation or a brand image. These assets are quasi-public, in the sense that they can continue to deliver services after being used. In other words, once they have been developed, they can be used to support product development, market development and diversification (Ansoff, 1965). To understand the complementarities between AMT and marketing investments it is first necessary to emphasize two characteristics of the former. First, AMT can be used to improve the information processing capabilities of firms (Kotha and Swamidass, 2000). Firms can use new technologies in order to cope with the uncertainty surrounding the activities that they perform, given that they help to collect and manage information. Those firms that follow a strategy that is intensive in information processing requirements will benefit most from the use of new technology. Second, AMT also improve the flexibility of firms to adapt to changing demands (Parthasarthy and Sethi, 1992). In particular, they

1610

J. Gómez, P. Vargas / Research Policy 41 (2012) 1607–1619

allow more frequent changes in the production line, which could be used to satisfy the diverse needs of consumers and provide the firm with a way to obtain scope economies. These two characteristics are key to understand the interdependence that arises between advanced manufacturing technologies and the marketing function (Blois, 1985; Kotha and Swamidass, 2000). Although the benefits of higher information processing capabilities and more flexibility may be important for a variety of firms, they are likely to be critical for firms investing more intensively in marketing assets. Both Kotha and Swamidass (2000) and Parthasarthy and Sethi (1992) argue that investments in AMT are more important in firms following a product differentiation strategy. These firms focus on offering a more varied set of products and present more product innovation than firms pursuing a cost leadership strategy (Porter, 1980). On the one hand, this means that the flexibility provided by AMT is of great importance. On the other hand, firms following a differentiation strategy are likely to have a higher level of complexity and they present more discontinuities in the production process (Kotha and Swamidass, 2000). Again, this will make the information processing capabilities of AMT critical. An important observation in this argumentation is that investments in marketing are a relevant part of a differentiation strategy (Hill, 1988). Advertising outlays may be used to build the reputation and brand image necessary to support differentiation. Therefore, although the complementarity between marketing investments and AMT may arise in any firm, they are more likely to appear in those cases in which marketing investments are more intense, providing incentives for adoption.8 Hypothesis 3. Firms with more marketing resources are more likely to use a new technology. 2.4. Complementarities between technologies As well as the complementarities between the use of new technologies and firm assets, complementarities between technologies may also arise (Stoneman and Kwon, 1994). According to Milgrom and Roberts (1990), the adoption of a cluster of technological innovations, which share some basic technological properties, is subject to significant complementarities. Under such circumstances, interdependences must be taken into account as they are likely to affect diffusion. Adopting powerful combinations of advanced technologies can leverage the technology gap a firm possesses over its competitors (Castellaci, 2002). The key advantage that arises from the use of multiple process technologies is that they can build complex manufacturing systems, which make manufacturing programmable and lead to timely information transfer across departments, employees, customers and suppliers. These systems are important since most manufacturing technologies are relatively available on the open market and any particular competitive advantage is difficult to defend. One way to achieve competitive advantage consists of building systems in which the number of elements and the degree of interaction among them is what makes them inimitable (Rivkin, 2000). In other words, while the adoption of individual technologies may no have effect on competitive advantage, certain technology combinations may help in building and sustaining it (Sinha and Noble, 2008) due to complementarities with past adoptions (Colombo and Mosconi, 1995). Previous work on complementary technologies has provided theoretical arguments and empirical evidence on the existence

of complementarities between different advanced technologies (Arvanitis and Hollenstein, 2001; Astebro et al., 2005; Colombo and Mosconi, 1995; Milgrom and Roberts, 1990; Stoneman and Kwon, 1994). From an empirical point of view, Arvanitis and Hollenstein (2001), Colombo and Mosconi (1995), Gómez and Vargas (2009) and Stoneman and Kwon (1994), demonstrate the importance of complementarities across adopted technologies. Beede and Young (1996) found enormous diversity in adoption patterns within the same industry, as well as important differences in the effect of various technology combinations on performance, suggesting the need to consider technology bundles in assessing adoption scenarios. More recently, Sinha and Noble (2008) found that firms that adopted several technologies were best positioned to survive, illustrating the importance of cumulative technology adoption. Hypothesis 4. The use of a technology is positively related to the use of the other technologies with which it forms a system. 3. Methodology, sample and variables 3.1. Sample description The dataset used for this study is drawn from the Survey of Business Strategies (ESEE). This is an annual survey on the activity of Spanish manufacturing firms and their business strategies financed by the Ministry of Industry and carried out by the SEPI foundation. Although it is not specifically designed to analyse technology adoption, it includes information on a number of technologies that are used by firms. The survey covers firms which have 10 or more employees and whose principal economic activity is listed in one of the two digit manufacturing industries of the NACE-Rev.1. In the base year, surveyed firms employing between 10 and 200 people were selected by means of a random sampling scheme, while firms with more than 200 were surveyed on a census basis. Although the survey has been administered annually to firms since 1990, questions about adoption behaviour were only included in 1994, 1998, 2002 and 2006. This does not permit us to establish causality relationships because the date of adoption of the technology is unknown. Fortunately, we do have lagged information on other firm and market characteristics that will be used as explanatory variables, as is explained below. Therefore, our objective is to find the combination of resources that is more likely to be used with new technologies. After selecting all the observations for which data on the independent variables is available, we are left with 4418 observations that will be used in the empirical analysis.9 Our main analysis is carried out on four process technologies. We focus on computer numerical controlled (CNC) machines, computer aided design (CAD), robotics and flexible manufacturing systems (FMS), given that these are the technologies for which a longer observation window is available.10 These technologies have been analysed in previous diffusion studies. In fact, and as we have mentioned before, they are part of what is termed Advanced Manufacturing Technologies. AMT is used as an umbrella term to describe “an automated production system of people, machines and tools for the planning and control of the production process, including the procurement of raw materials, parts and components and the shipment and service of finished products” (Pennings, 1987, p. 198). A characteristic of these technologies is that they are easy to integrate electronically. The key advantage of AMT is that

9

8

As mentioned by the two anonymous referees, the link between marketing resources and AMT is expected to be weaker than in the case of the other two elements (technological and human resources), given that their interaction is less direct.

The usual tests show that the sample is representative of the total population. In the survey, firms are asked whether their production process uses any of the following systems: (1) computer numerically controlled (CNC) machines, (2) robotics, (3) computer assisted design (CAD), (4) combination of some of the above systems through a central computer (FMS). 10

J. Gómez, P. Vargas / Research Policy 41 (2012) 1607–1619

1611

Table 1 Percentage of adopters and non adopters by technology and year. CNC

1994 1998 2002 2006

ROBOTICS

CAD

FMS

B2B

B2C

S2B

Yes

No

Yes

No

Yes

No

Yes

No

Yes

No

Yes

No

Yes

No

33.1 43.9 48.3 51.5

66.9 56.1 51.7 48.5

18.1 26.3 26.4 31.5

81.9 73.7 73.6 68.5

24.2 35.3 36.9 40.0

75.8 64.7 63.1 60.0

21.4 26.7 22.3 27.1

78.6 73.3 77.7 72.9

– – 5.8 9.9

– – 94.2 90.1

– – 4.1 5.1

– – 95.9 94.9

– – 13.8 25.9

– – 86.2 74.1

The expressions “Yes” and “No” indicate whether the firms have adopted or not

complex manufacturing systems may be built that make manufacturing programmable and lead to timely information transfer across departments, employees, customers and suppliers. This adds a new element to the study of complementarities: these technologies not only interact with the endowment of resources of the firm, but they are frequently used as a system. This reinforces the idea that complementarities have a key role to play in the understanding of performance differences and adoption behaviour. Although these four technologies will be the main focus of our analysis, the Survey also includes information on the use of e-business in 2002 and 2006, the last two periods for which information has been published. In particular, the available data refer to the adoption of business-to-business (B2B), business-to-consumer (B2C) and supplier-to-business (S2B) technologies. Although the main objective of these technologies is to manage information electronically, they are also used for enhancing communication between firms. In other words, similarly to Powell and Den-Micallef (1997), we distinguish between technologies that are used within firms (in-firm technologies, namely, CNC, CAD, robotics and FMS) and those used beyond their limits (beyond-firm technologies, namely, B2B, B2C and S2B). Following Hypothesis 4, our estimations should show a stronger relationship between in-firm technologies than between in-firm and beyond-firm technologies. Table 1 offers a first approximation to the data, showing the distribution of adopters and non-adopters by technology, and year. As can be seen, the number of adopters is different depending on the technology. CNC is the most-used technology (51.5% of the firms use it in 2006), followed by CAD (40.0% in 2006). Robotics and FMS have

the lowest figures (31.5% and 27.1%) among in-firm technologies. However, the adoption of e-business technologies presents much lower percentages. Looking at the figures for 2006, we can observe that B2B is adopted by 9.9% of the firms, B2C by 5.1% and S2B by 25.9%. Similarly, Table 2 presents the distribution of adopters by industry. It shows clear differences between the sectors included in the sample. Firms in the “Motors and autos” industry are the most active in the adoption of CNC machines, followed by the ones operating in “Other transport material” and “Furniture”. The adoption of robotics shows its highest values in the “Motors and autos” sector, with the “Beverages” industry and “Other transport material” following. Firms belonging to “Other transport material”, “Motors and Autos” and “Machinery for agriculture and industry” are the most frequent users of CAD. Finally, in the case of FMS, the most frequent users are “Motors and Autos”, “Other transport material” and “Paper”. In contrast, “Meat products” is the least active industry in CNC, CAD and FMS and “Other manufacturing” in the case of Robotics. The figures on the use of e-business technologies are, as mentioned above, generally lower. The most active sectors are “Paper” (in the case of B2B) and “Edition and graphical arts” (B2C and S2B). Contrarily, firms belonging to “Leather and footwear” (B2B and B2C) and “Wood” (S2B) are the least active in adoption. Finally, Table 3 shows the conditional aspects of technology adoption, revealing the complementarities that arise in the use of AMT. For any of the four main technologies studied, it shows the percentage of adopters that are also using none, one, two and three

Table 2 Percentage of adopters in 2006 by technology and industry. CNC

ROBOTICS

Meat products Food and tobacco Beverages Textiles and clothing Leather and footwear Wood industry Paper Edition and graphical arts Chemical products Plastic and rubber Non-metallic minerals Metallurgy Metallic products Machinery for agriculture and industry Machinery for offices, data processing, etc. Electrical material and accessories Motors and autos Other transport material Furniture Other manufacturing

28.6 41.5 37.5 36.9 34.8 47.5 56.4 37.9 34.8 52.5 46.1 62.0 62.1 64.5 46.2 57.4 75.4 75.0 67.2 42.1

21.4 27.7 43.8 16.7 17.4 25.0 41.0 17.2 18.8 42.6 35.5 42.0 34.7 32.9 30.8 38.9 62.3 43.8 21.3 15.8

Total manufacturing 2

51.5 72.2***

31.5 68.5***

* ***

Coefficient statistically significant at the 90% level. Coefficient statistically significant at the 99% level.

CAD 3.6 10.6 25.0 42.9 17.4 17.5 46.2 31.0 23.2 42.6 31.6 56.0 45.2 69.7 53.8 57.4 70.5 87.5 39.3 26.3 40.0 160.2***

FMS

B2B

B2C

S2B

7.1 20.2 37.5 21.4 8.7 12.5 38.5 24.1 31.9 29.5 22.4 38.0 29.8 31.6 23.1 31.5 45.9 43.8 19.7 15.8

10.7 5.3 6.3 8.3 0 2.5 23.1 15.5 15.9 11.5 9.2 14.0 7.3 7.9 15.4 13.0 4.9 6.3 13.1 10.5

3.6 2.1 12.5 6.0 0 0 5.1 22.4 4.3 4.9 2.6 4.0 2.4 5.3 0 5.6 6.6 0 6.6 5.3

21.4 17.0 37.5 25.0 13.0 10.0 23.1 39.7 24.6 23.0 17.1 36.0 29.0 30.3 30.8 25.9 32.8 37.5 29.5 21.1

27.1 44.0***

9.9 26.3

5.1 48.4***

25.9 29.3*

1612

J. Gómez, P. Vargas / Research Policy 41 (2012) 1607–1619

Table 3 Probability of adoption of AMT depending on the number of other technologies adopted and unconditional probability of adoption (2006).

CNN ROBOTICS CAD FMS

0

1

2

3

Unconditional probability

28.4% 16.9% 9.2% 10.0%

53.4% 30.0% 31.9% 16.8%

78.7% 49.1% 61.6% 39.4%

89.3% 64.3% 81.2% 60.7%

51.5% 31.5% 40.0% 27.1%

of the other technologies. The data clearly shows an increasing probability of using CNC, Robotics, CAD and FMS as firms increase the number of other AMT technologies in which they have invested. Therefore, this preliminary evidence seems to support the idea that the adoption of a given technology is favoured by the use of any of the other technologies with which it forms a system.

3.2. Main variables 3.2.1. Stock of technological resources We approximate the stock of technological resources by using the ratio of research and development capital to sales. For any given year (t) for which data on technology adoption is available, we compute research and development capital (K) through a partial inventory of past (3 previous years) and present annual internal R&D11 expenditures (R) with a constant depreciation rate, ı:

Kt =

3 

k

(1 − ı) × Rt−k

k=0

The annual depreciation rate (ı) was assumed to be 15% in accordance with Adams (1999), Griliches (1981), Griliches et al. (1981) and Villalonga (2004). The use of a depreciation rate can be explained by the decay of knowledge over time (Argote et al., 1990) and by the loss of economic value due to the development of new knowledge and technologies (Oriani and Sobrero, 2003).

3.2.2. Human resources The available information also allows us to calculate a measure of employee skills using the number of employees who possess a university degree (see, for example, Vicente-Lorente, 2000).12 With this information, we calculate the ratio of the number of employees with a university degree to the total number of employees of the organization. While this is not a perfect measure of employee skills, it has been used in previous papers (see, for example, Arvanitis and Louikis, 2009; Doms et al., 1997), and it is considered a reasonable proxy.

3.2.3. Stock of marketing resources Similarly to technological resources, the stock of marketing resources is measured as the ratio of advertising capital to sales. As with technological capital, for any given year (t) for which data on technology adoption is available, the value of marketing capital (P) is constructed using a partial inventory of past (3 previous years)

11 The survey collects data on (1) external expenses, (2) internal expenses and (3) total expenses ((1) + (2)) in research and development. To calculate the stock of technological resources, we have used internal R&D expenses. 12 Firms are asked to classify all the workers according to their qualifications. The options available are: (1) engineers and graduates, (2) middle level engineers, experts and qualified assistants and (3) other personnel. From this information we calculated the percentage of workers with a university degree.

and present annual advertising expenditures13 (A) with a constant depreciation rate, ı: Pt =

3 

k

(1 − ı) × At−k

k=0

Even though there is no consensus in the literature on the rate of depreciation, we follow Hirchey and Weygandt (1985) and Villalonga (2004) and use a depreciation rate of 45% going back three years. 3.3. Control variables 3.3.1. Firm size Firm size has traditionally played a prominent role in rank models of diffusion, usually presenting a positive effect on the probability of adoption. There are several explanations provided for this impact.14 First, firm size has been positively related to the existence of complementary assets within the firm. Hence, larger firms would obtain more profitability from the technology and would have more incentives to adopt it early (Colombo and Mosconi, 1995; Teece, 1986). A second explanation focuses on the concept of economies of scale. Larger firms would be able to spread the costs of the adoption of new technologies among a larger number of units. Firm size is measured as the total number of employees working for the firm. 3.3.2. Financial constraints The literature on diffusion attributes a role to financial constraints in the determination of a firm’s adoption behaviour. The arguments in favour of a negative relationship between financial constraints and adoption take three factors into account: uncertainty on the cash flows to be perceived from the adoption, information asymmetries between borrowers (the adopting firm) and lenders and, lastly, the frequent association between the diffusion process and investment in intangible and technologyspecific assets whose economic value could be difficult to recover (Canepa and Stoneman, 2005; Stoneman, 2001). As a consequence, we expect the availability of financial resources to have a positive impact on technology adoption. This variable is measured as the ratio of debts to total assets. 3.3.3. Corporate status The corporate status of a firm may also have an impact on the decision to adopt. By corporate status we mean whether the firm is a part of a larger business group or not. As noted by Baptista (2000), Bartoloni and Baussola (2001), Karshenas and Stoneman (1993) and Rose and Joskow (1990), the effect of this variable on the adoption decision is likely to be ambiguous. On the one hand, independent firms may be better positioned with regard to implementation, once the decision to adopt has been made. On the other, firms that are part of a larger institution may be better informed and bear less risk in adopting new technologies. Corporate status is

13 Firms provide some information on their accounts. For example, they are asked to report the expenses in publicity, advertisement and public relations. 14 Recent efforts have tried to explain this relationship (see, for example Astebro, 2002), although some confusion about the underlying mechanism still remains.

J. Gómez, P. Vargas / Research Policy 41 (2012) 1607–1619

measured through a dummy that takes a value of “one” when the firm is part of a larger corporate unit.

3.3.4. Ownership The fourth variable considered in the analysis is the type of ownership. The argument in favour of a positive effect is that the subsidiaries of a foreign firm may have access to the resources of the parent firm. Foreign investment may be a vehicle for the introduction of superior technology and scientific knowledge. Previous evidence on adoption models revealed a positive impact of foreign ownership (Baldwin and Diverty, 1995; Bosworth, 1996; Faria et al., 2003), although this relationship was only significant in the case of certain technologies. The presence of foreign capital in the focal firm is measured through a dummy that takes a value of “one” when the presence of foreign investors in the capital is higher than 30%.

3.3.5. Propensity to export The literature on international technology diffusion has argued that international trade is a channel through which firms can obtain information on new technologies (Keller, 2004). Exporting firms can also be expected to face more competitive international markets and be more likely to adopt new technologies in order to confront the higher levels of competition in the international arena. Previous empirical works have found either a positive relationship between exports and the adoption of new technologies (Cohen, 1975; Riedel, 1975) or no relationship at all (Lal, 1999, 2002). The ratio of total exports to sales is used to capture the effect of this variable.

3.3.6. Industry-specific rank effects We also consider two industry-specific influences on technology adoption. First, market structure has frequently been linked to the incentives of the firm to adopt a new technology, although with ambiguous influences (see, for example, Reinganum, 1981). We measure market concentration through the market share of the four largest firms (CR4 ). Second, one of the conclusions from the literature on technology diffusion is the importance of industry effects. Technology diffusion is largely determined by the technological characteristics of a given production process and this, in turn, is intimately linked to the sector of activity in which the firm is operating. Accordingly, we introduce 19 dummy variables to account for the 20 different sectors identified in the survey. Finally, and due to the fact that traditional models of diffusion have also taken other determinants of technology use into consideration, we introduce epidemic and stock effects into our model (Karshenas and Stoneman, 1993). Both these effects are taken into account through the inclusion of time variables. The passage of time has two opposite influences on technology adoption. On the one hand, the stock effect should mean that, as the number of adopters grows, the probability of adoption should be lower, given the lower returns. On the other, the epidemic effect acts through a learning-by-contact process. Hence, the number of previous adopters would also produce a positive effect on adoption as information flows increase in the industry. Such an effect could counteract the expected stock effects, making the net effect ambiguous (Baptista, 2000; Luque, 2002). We measure epidemic and stock effects through three dummy variables that take a value of “one” for the years 1998, 2002 and 2006, leaving 1994 as the base year.15

15

Appendix A contains some descriptive statistics of the covariates.

1613

3.4. Model and estimation strategy The model that we estimate is similar to other adoption models that have been used in the literature (see, for example, Barbosa and Faria, 2008; Battisti et al., 2009). Let A(t)ijk be the net expected profitability from the adoption (at time t) of technology k by firm i, that operates in industry j. The net expected returns from the technology depend on a set of firm-specific variables (xi ) that capture rank effects, a set of industry characteristics (zj ), epidemic, order and stock effects (st ) that are expected to vary over time, and an error term (εij ) that measures unobservable effects, in the following way: A(t)ijk = xi ˇ + zj  + st ω + εij The data set we use imposes two restrictions on this model. The first is that we do not have information on the date at which a firm adopts the technology, but only on whether it has adopted it at predetermined points in time. In other words, as in other papers in the literature (see, for example, Battisti et al., 2009), we cannot estimate a truly dynamic model. Second, the data describe adoption as a discrete choice, which means that we do not observe the returns to innovation, but only an indicator variable (Iijk ) that takes a value of “one” if firm i adopted technology k at time t (A(t)ijk ≥ 0), and “zero” otherwise. This suggests using qualitative dependent variable models (either logit or probit). In other words, we model the probability of adoption of the technology as a function of multiple explanatory variables capturing the rank, stock and epidemic effects described above. The probability of adoption may be expressed as a function of a vector of variables reflecting all three effects. Given the documented similarities between the logit and probit specifications and the considerations that we make below, we chose the probit link to perform our estimations. As mentioned when developing our empirical model, technologies may be complementary. In the case of the technologies analysed here, it has been argued that they form a cluster in which we could include numerically controlled machines, computer aided design, robots and flexible manufacturing. Recent research has extended the number of factors affecting adoption in order to take into account that technologies are complementary. The idea is that some technologies are difficult to use in isolation and, therefore, need to be adopted as systems that are jointly used in certain activities. From an empirical perspective, the assumption that technologies do not operate in isolation has an important methodological implication. If the profitability of adopting a given technology is related to the adoption of some other technology, this means that the two decisions are interdependent. Therefore, the estimation of any model of adoption should consider the fact that the firm may be using other complementary technologies. Given that our data on adoption provides us with information on the use of several technologies, we estimate the decision to use the technology through a multivariate probit model. This model can be seen as a generalization of the bivariate probit model, allowing more than two equations to be simultaneously estimated. The model captures the complementarities in the adoption of different technologies by allowing the disturbances of the different equations to be correlated. These correlations may be due to complementarities (positive correlation) or substitutabilities (negative correlation) between different technologies. Nevertheless, correlation could also be the result of unobservable firm-specific characteristics that affect adoption decisions but that are not easily captured by measurable proxies (Belderbos et al., 2004; Santamaria and Rialp, 2007). In fact, the arguments that we have used in this paper about the existence of complementarities between resources and technology could be extended to other assets that are unobservable. The multivariate probit model takes into account the

1614

J. Gómez, P. Vargas / Research Policy 41 (2012) 1607–1619

interrelationships arising in adoption, although it is not able to distinguish between the two sources of correlation described. However, if correlation exists, the estimates of the separate equations would provide us with inefficient estimates. The estimation is carried out using Stata’s mvprobit command, which applies the method of simulated maximum likelihood (SML) that uses the Geweke–Hajivassiliour–Keane (GHK) smooth recursive-conditioning simulator to evaluate the multivariate normal distribution. Following Cappellari and Jenkins (2003), the simulated probabilities are unbiased and bound within the (0, 1) interval. The variance–covariance matrix V of the cross-equation error terms has values of 1 on the leading diagonal, and the offdiagonal elements, correlations ␳jk = ␳kj, are to be estimated. The parameter ␳jk is the co-variance between the error terms of equations j and k. 4. Results Table 4 shows the results of estimating a multivariate probit on the 4418 observations available.16 Columns 1–4 estimate four models in which only control variables are included. They are used as a benchmark. A second set of estimations is presented in columns 5–8, which introduce the variables that help us to test our hypotheses. As we can see, all the models are globally significant, given the high values of the Wald statistic. Furthermore, their comparison favours the estimations presented in columns 5–8. As shown at the bottom of Table 4, the LR test is highly significant, pointing to the importance of complementary resources for explaining adoption. We should highlight that not only the coefficients but also the tratios accompanying them remain fairly stable across the two sets of estimations.17 Hypothesis 1 argued that firms with more technological resources would be more likely to use advanced technologies. This is, in fact, the case for the four technologies: the coefficients accompanying the “technological resources” variable are positive and significant. Therefore, the evidence confirms our hypothesis that the use of new technologies is favoured by the presence of higher levels of R&D-derived knowledge inside the firm. Hypothesis 2 stated that technology adoption should be also be conditioned by the qualifications of human resources. In this case, we observe clear differences between the four advanced technologies analysed. The variable “human resources” is highly significant in the case of computer aided design and flexible manufacturing, whereas the other two technologies do not show any impact. This result seems to resemble that of Dunne and Troske’s (2005) paper. These authors found that the effect of skilled labour on technology use varied across technologies, finding a stronger correlation between technologies associated with design and engineering functions and human resources (CAD) than those more closely associated with production activity (CNC and robotics).18

16 Following the comments of one of the anonymous referees, we also estimated an ordered probit model that considers whether the firm adopts 0, 1, 2, 3 or 4 manufacturing technologies. The results are highly consistent with the ones presented here: firm size, the availability of financial resources, technological and human resources, export propensity and integration in a business group are significant predictors of adoption. The estimates also show the relevance of epidemic effects. 17 We calculated the variance inflation factors (VIF) in order to detect multicollinearity problems. The maximum VIF was 5.02 and the average VIF was 2.47. Therefore, we do not have reasons to suspect that they could be present. 18 The variable used to test this hypothesis is calculated as the percentage of workers with a university degree. As one anonymous reviewer points out, this has the problem that some of these employees will be working on R&D activities and this may lead to double counting. Although we do have information on the number of employees working on R&D activities, the classification of the workers into different groups does not match the one for the whole firm. In any case, we rerun the estimations with a measure of human resources from which we subtracted the number of

Similarly, Hypothesis 3 stated that firms with more marketing resources would be more likely to use new process technologies. Again, we find clear differences between the technologies analysed. Whereas the variable “marketing stock” is significant for robotics and, to a lesser extent, for computer aided design marketing resources are not significant in the case of computer numerically controlled machines and flexible manufacturing. These findings are consistent with the ones of Kotha and Swamidass (2000). Using information on several AMT, they conclude that those firms following a differentiation strategy (an strategy associated with high marketing investments) are more likely to use product design technologies (including CAD) and high volume automation technologies (including robotics). Our analysis also included a set of control variables that are frequently used in the literature. Firm size is an important predictor of the likelihood of using a technology, as it shows a positive and significant effect in all the estimations. Having controlled for size, the importance of financial constraints is also significant for both robotics and flexible manufacturing, but not for CNC and CAD. Three other firm-specific variables present a significant impact on the adoption of new technologies. First, our results show that the propensity to export (ratio exports to sales), is positively associated with adoption in all cases. This result is interesting from the perspective of the international diffusion of technology (Keller, 2004). Several authors have argued that international trade is an important diffusion channel (see, for example, Eaton and Kortum, 2002; Grossman and Helpman, 1991). In the case of exports, the anecdotal evidence shows that firms may benefit from interacting with foreign customers, given that they tend to require higher standards and also provide information on how to meet them (Keller, 2004: 767). Furthermore, this result also seems to be consistent with the idea that exporters are different from non-exporters in terms of productivity (Bernard and Jensen, 1999) and innovation. In fact, the evidence from Spain shows that exporters are more productive and more likely to introduce process innovations than non-exporters. ˜ et al. (2009) find that not Using the same database, Mánez-Castillejo only the most productive firms have a higher probability of exporting, but also that process innovations increase the probability of exporting. Second, the data confirm that the integration of the firm into a business group has an impact on adoption. This result is in line with the idea that establishments that are part of a larger corporation will experience less uncertainty and less financial constraints when adopting new technologies. Third, we find that foreign ownership is only significant in the case of robotics.19 This result is partially consistent with previous evidence that points out that foreign-owned companies invest more in advanced manufacturing technologies and obtain higher operational performance (Beaumont et al., 2002). Third, our results are also generally consistent with the existence of an epidemic effect. The time dummies show that the use of all the four technologies is more frequent in 2006.20 Apart from assessing the impact of certain complementary resources on adoption, the use of the multivariate probit model

employees working for the R&D function that were (1) “graduates” and (2) “middle level technicians”. The results were similar to the ones presented here. 19 An anonymous reviewer suggested investigating on the sources of this capital. Unfortunately, our data do not allow us to identify the country in which these investments originate. We were, however, able to identify the country in which a firm invests for the two last periods of our sample. Our results are in line with the work of Griffith et al. (2006), suggesting that investments in OECD countries are significant for explaining Robotics and CAD adoption. 20 We also estimated a model that included both the percentage of adopters per industry and in the economy as a whole. The only variable that remains significant in this model is the first, offering support for the existence of an industry specific epidemic effect.

J. Gómez, P. Vargas / Research Policy 41 (2012) 1607–1619

1615

Table 4 The effect of technological, human and marketing resources on AMT adoption. CNC (1)

Robotics (2)

CAD (3)

FMS (4)

CNC (5)

Robotics (6)

Stock of technological resources

0.17*** (3.76) −0.08 (−0.87) –

0.38*** (7.77) −0.29*** (−2.92) –

0.19*** (4.47) 0.10 (1.11) –

0.48*** (9.35) −0.33*** (−3.41) –

Human resources

–

–

–

–

Stock of marketing resources

–

–

–

–

Export propensity

0.44*** (5.17) −0.38 (−1.32) 0.30*** (5.69) −0.02 (−0.01) 0.24*** (3.98) 0.34*** (5.69) 0.40*** (6.48) Yes −0.87*** (−4.32)

0.63*** 0.51*** (7.13) (5.74) −0.26 0.20 (−0.84) (0.64) *** 0.47 0.33*** (8.39) (6.01) 0.16** −0.04 (2.49) (−0.64) 0.23*** 0.35*** (3.36) (5.37) 0.27*** 0.38*** (3.95) (5.84) 0.39*** 0.45** (5.72) (6.84) Yes Yes −1.44*** −2.47*** (−6.30) (−9.28) 0.409*** (17.46) *** 0.457 (21.18) 0.292*** (11.80) 0.405*** (16.77) 0.297*** (11.69) 0.413*** (17.13) 4418 1907.70***

0.16*** (3.50) −0.07 (−0.83) 1.61*** (2.86) 0.12 (0.71) −0.22 (−0.91) 0.41*** (4.81) −0.35 (−1.22) 0.29*** (5.37) −0.00 (−0.08) 0.24*** (4.02) 0.34*** (5.66) 0.40*** (6.44) Yes −0.89*** (−4.38)

0.36*** (7.28) −0.28*** (−2.80) 2.42*** (4.12) 0.14 (0.73) 0.42*** (2.70) 0.60*** (6.65) −0.23 (−0.74) 0.45*** (7.95) 0.18*** (2.82) 0.25*** (3.64) 0.29*** (4.17) 0.42*** (5.98) Yes −1.49*** (−6.52)

Firm size Firm debt ratio

Market concentration Integrated in a business group Foreign capital Year 1998 Year 2002 Year 2006 Industry dummies Constant Rho2,1 Rho3,1 Rho4,1 Rho3,2 Rho4,2 Rho4,3 No. observations Wald test Comparison test LR test of Rho2,1 = Rho3,1 = Rho4, 1 = Rho3,2 = Rho4,2 = 0 Rho4,3 = 0 * ** ***

0.33*** (3.75) 0.31 (0.99) 0.39*** (7.10) −0.01 (−0.21) 0.14** (2.16) 0.02 (0.31) 0.13* (1.93) Yes −1.74*** (−7.32)

1047.75***

CAD (7)

FMS (8)

0.17*** (3.89) 0.12 (1.29) 3.66*** (6.32) 0.66*** (3.64) 0.32* (1.74) 0.44*** (4.92) 0.23 (0.75) 0.28*** (4.93) −0.01 (−0.22) 0.37*** (5.60) 0.39*** (5.87) 0.47*** (6.88) Yes −2.54*** (−9.52) 0.409*** (17.40) *** 0.459 (21.16) 0.290*** (11.67) 0.403*** (16.53) 0.294*** (11.48) 0.404*** (16.54) 4418 1992.44*** 111.98*** 1025.56***

0.46*** (8.82) −0.33*** (−3.36) 2.81*** (4.86) 0.71*** (3.88) −0.33 (−0.95) 0.28*** (3.13 0.32 (1.02) 0.35** (6.21) 0.00 (0.02) 0.14** (2.17) 0.01 (0.09) 0.11* (1.68) Yes −1.76*** (−7.36)

Coefficient statistically significant at the 90% level. Coefficient statistically significant at the 95% level. Coefficient statistically significant at the 99% level.

also allows us to estimate the correlations between the technologies once the variables included in Table 4 have been controlled for. These correlations are presented at the bottom of Table 4 and provide us with a way of testing Hypothesis 4. They are all positive and significant, suggesting that the arguments leading to Hypothesis 4 are correct. However, these correlations could be interpreted in two ways. First, a positive correlation could be due to the influence of firm-specific factors that are not included in our estimations and that determine the propensity of some firms to adopt all the technologies. Therefore, they provide us with a way of assessing the influence of unobservable firm effects. Second, a positive correlation could mean that the technologies form part of a system. Hence, the adoption of one technology would increase the probability of the adoption of the other technologies of the system.

One possibility to investigate Hypothesis 4 further is to reestimate our full model adding other technologies that we suspect are not closely related to the four process technologies analysed. In other words, the correlations between non-related technologies should capture the influence of non-observable firm-specific effects related to the adoption of all the technologies. Fortunately, our data set also contains information on the adoption of three e-business technologies. From 2000 on, the survey includes a question on whether the firm “buys goods or services (providers) through the Internet” (S2B), whether “it sells to final consumers through the Internet” (B2C) and whether “it sells to firms through the Internet” (B2B). Given that the three technologies supporting these systems are not obviously related to process technologies, a dramatic reduction in the value of the correlations should be observed. Unfortunately, this significantly restricts our observation window

Table 5 Correlation matrix between process and information technologies.

CNC Robotics CAD FMS B2B B2C S2B ***

CNC

Robotics

CAD

FMS

B2B

B2C

1 0.414*** 0.470*** 0.337*** 0.105*** 0.138*** 0.080***

– 1 0.396*** 0.357*** 0.090*** 0.130*** 0.067***

– – 1 0.436*** 0.075*** 0.009*** 0.131***

– – – 1 0.122*** 0.199*** 0.178***

– – – – 1 0.797*** 0.373***

– – – – – 1 0.355***

Coefficient statistically significant at the 99% level.

1616

J. Gómez, P. Vargas / Research Policy 41 (2012) 1607–1619

(to years 2002 and 2006) and the number of available observations (to 2367). The correlations resulting from estimating a full multivariate probit model with the seven technologies are presented in Table 5. The conclusion is that the correlations between process and ebusiness technologies are clearly lower than those within the groups. The highest correlation between the two groups is 0.199 (B2B and FMS), the value is non-significant in two of the cases (CAD-B2B and CAD-B2C) and three correlations are below 0.10. In order to make sure that these differences were significant, we compared all the correlations between in-firm and beyond-firm technologies.21 For a given correlation between two technologies belonging to the in-firm technologies (for example, CNC and Robotics), we tested whether the coefficient was different from that of the in-firm technology and each of the beyond-firm technologies (i.e., we compared it with CNC-B2B, CNC-B2C and CNC-S2B). The results (not shown) of all the comparisons are significantly different, showing that within-group correlations are higher than inter-group correlations. In other words, our results confirm that complementarities between related technologies exist and that they are very different from the ones corresponding to unrelated technologies, supporting Hypothesis 4. 5. Conclusions and implications In this paper, we have focused on the different incentives that firms have to adopt advanced technologies depending on the possibilities of appropriating their returns. We have argued that complementary resources are one of the mechanisms that firms use to profit from the use of new technologies. One implication of this hypothesis is that the use of AMT should be more likely in firms that possess more complementary assets. The evidence presented in this paper seems to confirm our conjecture: complementary resources are associated with a firm’s technology use. Clear differences in adoption behaviour are identified depending on the technologies analysed. More precisely, only technological resources are unequivocally related to the use of new technologies in manufacturing firms. The fact that we have studied the use of AMT could help us to understand this result. These technologies are related to product manufacturing and design. Therefore, investments in the knowledge base of the firm could have served to understand and refine both processes, detecting the need and the role of new technologies. Our results seem to suggest that more complex technologies benefit to a larger extent from these investments, as proposed by Rogers and Shoemaker (1971). Human and Marketing resources are also important, although they present a weaker relation and different effects depending on the technologies analysed. The quality of the services provided by human resources is positively related to computer aided design and flexible manufacturing, whereas marketing investments are only related to robotics and computer aided design. These results are consistent with previous literature (Dunne and Troske, 2005; Kotha and Swamidass, 2000). The interpretation of these results has, however, to be taken with some caution, given that we lack data on other characteristics of the technologies (see, for instance, Rogers and Shoemaker, 1971) and that, at the time of the study, the diffusion process had already started. We also explain the use of advanced technologies through the complementarities that arise within systems of technologies. In fact, the interrelations between some of the technologies are the main reason that we observe positive correlations between them, once we control for the determinants of adoption. This supports the

21 We followed Buis (2011) in order to get access to the correlations. We thank Maarten Buis and Stephen Jenkins for their advice.

view of those that contend that the diffusion of new technologies should not be analysed in isolation. Our results have implications for the study of technology diffusion and competitive advantage. With respect to the former, they support the move from epidemic to rank models undertaken by the literature on diffusion in recent years. Epidemic models contend that firms adopt new technologies once information on their characteristics reduces the uncertainty surrounding the innovation. However, although important, research has shown that information diffusion is not the only mechanism that explains why some firms are early users and others delay the adoption of the technology until the latest stages of the diffusion process or do not invest in it at all. Although the range of factors involved adds complexity to any explanation provided, the literature and our results show that complementarities play an important role. Complementarities are likely at two levels. First, they should be important to understand the firm-technology dyad. Both the resource-based view of the firm (Barney, 1991; Dierickx and Cool, 1989; Wernerfelt, 1984) and Teece’s (1986) framework provide support for the idea that the interconnectedness of resources should be understood in order to assess the quality of the services provided. In other words, firm resources cannot be analysed or understood in isolation. Accepting this argument and understanding new technologies as a resource, this implies that the endowment of resources of a given firm should determine its needs for or the adequacy of a given technology. Although research in the information systems and technology adoption literatures has argued that complementary resources are needed to appropriate value creation, it has mainly focused on the analysis of performance. This paper, thus, contributes by showing that the incentives of firms to adopt new technologies depend on the amount of complementary resources that they possess. This not only adds support to the hypothesis that complementary resources could help us to understand the link between technology adoption and competitive advantage, but also introduces a new, firm-specific element into the explanation of diffusion patterns. We have focused on the analysis of certain intangible resources, based on the argument that they tend to be more difficult to imitate. However, the range of firm assets likely to interact with and, therefore, to determine the use of a given technology is difficult to specify. As our results show, not all the technologies are expected to be equally related to firm resources. Second, complementarities should also be important in the technology–technology dyad. Several papers (Colombo and Mosconi, 1995; Milgrom and Roberts, 1990; Stoneman and Kwon, 1994) have argued that technologies cannot be analysed in isolation. On the contrary, they need to be adopted as systems that are jointly used in certain activities. Several papers have shown that multiple adoption behaviour can help the firm to build complex systems in which complementary relationships between the elements will arise. Moreover, the multiple adoption consideration opens new questions for research into the factors that affect the acquisition of new technologies because the technological trajectory and the resource endowments of a firm greatly affect adoption choices. These results also have implications for understanding firm heterogeneity and competitive advantage in the context of technology diffusion. The main argument for maintaining that, despite their positive effects on firm activities, new technologies are not related to sustainable competitive advantage is based on the idea that they can be acquired in the market. In other words, laggard firms could easily replicate the technology strategy of pioneers by buying units of the new technology. This would lead to a situation of competitive parity (Barney, 1991), providing no advantages to any firm. The literature on information systems has coined the term strategic necessity hypothesis to refer to this idea (Clemons

J. Gómez, P. Vargas / Research Policy 41 (2012) 1607–1619

1617

Table A1 Descriptive statistics of dependent and independent variables.

1. CNC 2. Robotics 3. CAD 4. FMS 5. B2B 6. B2C 7. S2B 8. Firm size 9. Firm debt ratio 10. Stock of technological resources 11. Human resources 12. Stock of marketing resources 13. Export propensity 14. Market concentration 15. Integrated in business group 16. Foreign capital

Mean

Observations

St. Dev.

Min.

Max.

0.45 0.26 0.35 0.24 0.08 0.05 0.19 0.23 0.56 0.01 0.11 0.03 0.19 0.41 0.33 0.18

4418 4418 4418 4418 2367 2367 2367 4418 4418 4418 4418 4418 4418 4418 4418 4418

0.50 0.44 0.48 0.43 0.27 0.21 0.39 0.62 0.23 0.04 0.13 0.13 0.26 0.13 0.47 0.39

0 0 0 0 0 0 0 0.001 0 0 0 0 0 0.20 0 0

1 1 1 1 1 1 1 14.20 1 0.65 1 4.32 1 0.95 1 1

and Kimbrough, 1986; Clemons and Row, 1991). However, recent research in strategic management has concluded that the process of technology diffusion among firms may not be as regular and widespread as expected. Greve (2009) has shown that social networks and cluster theory affect the diffusion of technologies. His findings indicate that there may be other sources of competitive advantage because resources that, in principle, are available to all will, in fact, be obtained by few firms, allowing them to get a longlasting competitive advantage. Our results contribute to this line of research by showing that the initial levels of heterogeneity of firm resources could produce very different results. In fact, they suggest that complementarities between firm resources and new technologies create varying incentives for firms to adopt, providing an additional dimension (the technology) through which to differentiate themselves. In other words, far from leading to resource similarity and competitive parity, the process of technology diffusion could create further heterogeneity, at least in the short run. The questions that remain are whether this increased heterogeneity is permanent and whether it can create competitive advantages. The answer, again, seems to depend on the balance between differentiation and homogenization. Although research on the interfirm dimension of diffusion tends to show that the process of adoption affects the majority of the firms operating in an industry (but see Greve, 2009), research on the intrafirm dimension shows that differences in the intensity of technology adoption increase over time (Fuentelsaz et al., 2009). However, the role of competitive imitation seems to be important when determining the effect of IT on performance. Although research on the diffusion of new technologies has recognized that a firm’s returns cannot be assessed in isolation when arguing in favour of stock and order effects (see, for example, Karshenas and Stoneman, 1993) and has also studied the influence of rivals’ adoptions on the diffusion of new technologies (Hannan and McDowell, 1987), the predominant view conceives competitors as information diffusers. Therefore, the role of competition, heterogeneity generation and competitive advantage creation should be more explicitly considered in technology diffusion studies. In the end, imitation would depend on the characteristics of the systems created and their inimitability and on the actions of competitors, which could progressively erode the rents of early adopters (Koellinger, 2008). In any case, the confirmation of the idea that complementary resources explain technology adoption and increase firm heterogeneity should not necessarily lead us to conclude that the net effect of technology on profitability is positive in the long run. Even if the relative positions of the firms are maintained or even if some firms improve their advantage over their rivals, the high investments necessary to acquire new

technologies could imply that all the firms reduce their initial levels of profitability if they are not able to compensate for them. Acknowledgements We acknowledge financial support from the Spanish Ministry of Science and Technology and FEDER (projects ECO2008-04129 and ECO2011-22947), the Regional Government of Aragón (project S09/PI138-08) and Universidad de La Rioja (projects API 19/08, API 11/25 and EGI 11/29). We are grateful for the comments and sugges˜ tions provided by Dolores Anón, Sergio Palomas, two anonymous referees, the Editor of this Journal, participants at the Conferences of ACEDE, IAMB, EURAM and at seminars held at the Universities of Zaragoza, La Rioja and Rey Juan Carlos. A previous version of this paper was included in the Working Paper Series of the Spanish Fundación de las Cajas de Ahorros (FUNCAS) (document number 505/2010). Appendix A. Descriptive statistics (See Table A1) References Adams, J.D., 1999. The structure of firm R&D, the factor intensity of production, and skill bias. The Review of Economics and Statistics 81 (3), 499–510. Amara, N., Landry, R., Becheikh, N., Ouimet, M., 2008. Learning and novelty of innovation in established manufacturing SMEs. Technovation 28, 450–463. Ansoff, I., 1965. Corporate Strategy: An Analytic Approach to Business Policy for Growth and Expansion. McGraw-Hill, New York. Argote, L., Beckman, S.L., Epple, D., 1990. The persistence and transfer of learning in industrial settings. Management Science 36, 140–154. Arvanitis, S., Hollenstein, H., 2001. The determinants of the adoption of advanced manufacturing technology. An empirical analysis based on firm-level data for Swiss manufacturing. Economics of Innovation and New Technology 10, 377–414. Arvanitis, S., Louikis, E., 2009. A comparative study of the effect of the ICT, organizational and human capital on labour productivity in Greece and Switzerland. Information Economics and Policy 21, 43–61. Astebro, T., 2002. Noncapital investment costs and the adoption of CAD and CNC in U.S. metalworking industries. Rand Journal of Economics 33 (4), 672–688. Astebro, T., Colombo, M.G., Seri, R., 2005. The Diffusion of Complementary Technologies: An Empirical Test (downloaded on 8 November 2009 from http://ssrn.com/abstract=690981). Balakrishnan, S., Fox, I., 1993. Asset specificity, firm heterogeneity and capital structure. Strategic Management Journal 14, 3–16. Baldwin, J., Diverty, B., 1995. Advanced Technology Use in Canadian Manufacturing Establishments. Research Paper, 85. Analytical Studies Branch. Statistics Canada, Ottawa (downloaded on 8 November 2009 from http://publications.gc.ca/collections/Collection/CS11-0019-85E.pdf). Baptista, R., 2000. Do innovations diffuse faster within geographical clusters? International Journal of Industrial Organisation 18, 515–535. Barbosa, N., Faria, A.P., 2008. Technology adoption: does labour skill matter? Evidence from Portuguese firm-level data. Empirica 35, 179–194.

1618

J. Gómez, P. Vargas / Research Policy 41 (2012) 1607–1619

Barney, J., 1991. Firm resources and sustained competitive advantage. Journal of Management 17 (1), 99–120. Barney, J.B., 1986. Strategic factors markets: expectations, luck and business strategy. Management Science 32 (10), 1231–1241. Bartoloni, E., Baussola, M., 2001. The determinants of technology adoption in Italian manufacturing industries. Review of Industrial Organization 19, 305–328. Battisti, G., Canepa, A., Stoneman, P., 2009. E-Business usage across and within firms in the UK: profitability, externalities and policy. Research Policy 38 (1), 133– 143. Beaumont, N., Schroder, R., Sohal, A., 2002. Do foreign-owned firms manage advanced manufacturing technology better? International Journal of Operations and Production Management 22 (7), 759–771. Beede, D.N., Young, K.H., 1996. Patterns of advanced technology adoption and manufacturing performance: employment growth, labor productivity and employee earnings. Research Paper: Economics and Statistic Administration, Office of Policy Development, Washington, DC. Belderbos, R., Carree, M., Diederen, B., Lokshin, B., Veugelers, R., 2004. Heterogeneity in R&D cooperation strategies. International Journal of Industrial Organization 22, 1237–1263. Belderbos, R., Carree, M., Lokshin, B., 2006. Complementarity in R&D cooperation strategies. Review of Industrial Organization 28, 401–426. Bernard, A.B., Jensen, J.B., 1999. Exceptional exporter performance: cause, effect, or both. Journal of International Economics 47, 1–25. Blois, K.J., 1985. Marketing strategies and the new manufacturing technologies. International, Journal of Operations and Production Management 6 (1), 34–41. Bosworth, D., 1996. Determinants of the use of advanced technologies. International Journal of the Economics of Business 3 (3), 269–293. Buis, M., 2011. Stata tip 97: getting at ␳s and ␴s. The Stata Journal 11 (2), 1–3. Canepa, A., Stoneman, P., 2005. Financing constraints in the inter firm diffusion of new process technologies. The Journal of Technology Transfer 30, 159–169. Cappellari, L., Jenkins, S.P., 2003. Multivariate probit regression using simulated maximum likelihood. The Stata Journal 3 (3), 278–294. Cassiman, B., Veugelers, R., 1999. Make and buy in innovation strategies: evidence from Belgian manufacturing firms. Research Policy 28 (1), 63–80. Castellaci, F., 2002. Technology gap and cumulative growth: models and outcomes. International Review of Applied Economics 16 (3), 333–346. Clemons, E.K., Kimbrough, S., 1986. Information systems, telecommunications, and their effect on industrial organization. In: Proceedings of the 7th International Conference on Information Systems, San Diego, CA. Clemons, E.K., Row, M.C., 1991. Sustaining IT advantage. The role of structural differences. MIS Quarterly 15 (3), 275–292. Cohen, B., 1975. Multinational Firms and Asian Exports. Yale Univ. Press, New Haven, CT. Cohen, W.M., Levinthal, D.A., 1990. Absorptive capacity: a new perspective on learning and innovation. Administrative Science Quarterly 35, 128–152. Collins, D.J., Montgomery, C.A., 1998. Creating Corporate Advantage Harvard Business Review 76 (3), 70–83. Colombo, M.G., Mosconi, R., 1995. Complementary and cumulative learning effects in the early diffusion of multiple technologies. Journal of Industrial Economics 43, 13–48. Conner, K.R., Prahalad, C.K., 1996. A resource-based theory of the firm: knowledge versus opportunism. Organizational Science 7 (5), 477–501. Denrell, J., Fang, C., Winter, S.G., 2003. The economics of strategic opportunity. Strategic Management Journal 24, 977–990. Dierickx, I., Cool, K., 1989. Asset stock accumulation and sustainability of competitive advantage. Management Science 35 (12), 1504–1511. Doms, M., Dunne, T., Troske, K., 1997. Workers, wages and technology. The Quarterly Journal of Economics 112 (1), 253–290. Dunne, T., Troske, K., 2005. Technology adoption and the skill mix of US manufacturing plants. Scott Journal of Political Economy 52 (3), 387–405. Eaton, J., Kortum, S., 2002. Technology, geography, and trade. Econometrica 70, 1741–1779. Ennen, E., Richter, A., 2010. The whole is more than the sum of its parts- or is it? A review of the empirical literature on complementarities in organizations. Journal of Management 36 (1), 207–233. Faria, A., Fenn, P., Bruce, A., 2003. A count data model of technology adoption. Journal of Technology Transfer 28, 63–79. Fuentelsaz, L., Gómez, J., Palomas, S., 2009. The effects of new technologies on productivity: an intrafirm diffusion-based assessment. Research Policy 38, 1172–1180. Gómez, J., Vargas, P., 2009. The effect of financial constraints, absorptive capacity and complementarities on the adoption of multiple process technologies. Research Policy 38 (1), 106–119. Greve, H., 2009. Bigger and safer: the diffusion of competitive advantage. Strategic Management Journal 30 (1), 1–23. Griffith, R., Harrison, R., Van Reenen, J., 2006. How special is the special relationship? Using the impact of U.S. Spillovers on U.K. firms as a test of technology sourcing. The American Economic Review 96 (5), 1859–1875. Griliches, Z., 1981. Market value, R&D and patents. Economic Letters 7, 183–187. Griliches, Z., Hall, B.H., Pakes, A., 1981. R&D, patents and market value revisited: is there a second technological opportunity factor? Economics of Innovation and New Technology 1, 183–202. Grossman, G., Helpman, E., 1991. Innovation and Growth in the World Economy. MIT Press, Cambridge, MA.

Hannan, T.H., McDowell, J.M., 1987. Rival precedence and the dynamics of technology adoption: an empirical analysis. Economica 54, 155–171. Hill, C.W.L., 1988. Differentiation versus low cost or differentiation and low cost: a contingency framework. Academy of Management Review 13, 401–412. Hirchey, M., Weygandt, J.J., 1985. Amortization policy for advertising and research and development expenditures. Journal of Accounting Research 23, 326–335. Huffman, W.E., 1974. Decision making: the role of education. American Journal of Agricultural Economics 56 (1), 85–97. Karshenas, M., Stoneman, P., 1993. Rank, stock, order, and epidemic effects in the diffusion of new process technologies: an empirical model. Rand Journal of Economics 24 (4), 503–528. Keller, W., 2004. International technology diffusion. Journal of Economic Literature 42, 752–782. Koellinger, P., 2008. The relationship between technology, innovation, and firm performance – empirical evidence form e-business in Europe. Research Policy 37, 1317–1328. Kogut, B., Zander, U., 1992. Knowledge of the firm, combinative capabilities, and replication of technology. Organization Science 3 (3), 383–397. Kotha, S., Swamidass, P.M., 2000. Strategy, advanced manufacturing technology and performance: empirical evidence from U.S. manufacturing firms. Journal of Operations Management 18 (2), 257–277. Lal, K., 1999. Determinants of the adoption of Information Technology: a case study of electrical and electronic goods manufacturing firms in India. Research Policy 28, 667–680. Lal, K., 2002. E-business and manufacturing sector: a study of small and mediumsized enterprises in India. Research Policy 31, 1199–1211. Luque, A., 2002. An option value approach to technology adoption in U.S. manufacturing: evidence from microdata. Economics of Innovation and New Technology 11 (6), 543–568. ˜ J.A., Rochina-Barrachina, M.E., Sanchis-Llopis, J.A., 2009. SelfMánez-Castillejo, selection into exports: productivity and/or innovation? Applied Economics Quarterly 55 (3), 219–241. Mata, F.J., Fuerst, W.L., Barney, J., 1995. Information technology and sustained competitive advantage: a resource based analysis. Management Information System 19 (4), 487–505. Milgrom, P., Roberts, J., 1990. The economics of modern manufacturing: technology, strategy and organization. The American Economic Review 80 (3), 511–528. Milgrom, P., Roberts, J., 1995. Complementarities and fit: strategy, structure, and organizational change in manufacturing. Journal of Accounting and Economics 79 (2-3), 179–208. Oriani, R., Sobrero, M., 2003. A meta-analytic study of the relationship between R&D investments and corporate value. In: Caldirini, M., Garrone, P., Sobrero, M. (Eds.), Corporate Governance, Market Structure and Innovation. Edward Elgar, Cheltenham, pp. 177–199. Parthasarthy, R., Sethi, S.P., 1992. The impact of flexible automation on business strategy and organizational structure. Academy of Management Review 17 (1), 86–111. Parthasarthy, R., Sethi, S.P., 1993. Relating strategy and structure to flexible automation: a test of fit and performance implications. Strategic Management Journal 14 (7), 529–549. Pennings, J.M., 1987. Technological innovations in manufacturing. In: Pennings, J.M., Buitendam, A. (Eds.), New Technology as Organizational Innovation: The Development and Diffusion of Microelectronics. Ballinger, Cambridge, MA, pp. 197–216. Piva, M., Santarelli, E., Vivarelli, M., 2005. The skill bias effect of technological and organisational change: evidence and policy implications. Research Policy 34, 141–157. Porter, M.E., 1980. Competitive Strategy. The Free Press, New York. Powell, T.C., Den-Micallef, A., 1997. Information technology as competitive advantage: the role of human, business, and technology resources. Strategic Management Journal 18 (5), 375–405. Reinganum, J.F., 1981. Market structure and the diffusion of new technology. The Bell Journal of Economics 12 (2), 618–624. Riedel, J., 1975. The nature and discriminants of export-oriented direct foreign investment in a developing country: a case study of Taiwan. Weltwirtschaftliches Archiv 111, 505–528. Rivkin, J.W., 2000. Imitation complex strategies. Management Science 46, 824–844. Rogers, E.M., Shoemaker, F.F., 1971. Communication of Innovations. The Free Press, New York. Rose, N.L., Joskow, P.L., 1990. The diffusion of new technologies: evidence from the electric utility industry. Rand Journal of Economics 21 (3), 354–373. Santamaría, L., Nieto, M.J., Barge-Gil, A., 2009. Beyond formal R&D: taking advantage of other sources of innovation in low- and medium-technology industries. Research Policy 38, 507–517. Santamaria, L., Rialp, J., 2007. La elección del socio en las cooperaciones tecnológicas: una análisis empírico. Cuadernos de Economia y Dirección de Empresas 31, 67–96. Sinha, R.V., Noble, S.H., 2008. The adoption of radical manufacturing technologies and firm survival. Strategic Management Journal 29 (9), 943–962. Stoneman, P., 2001. Technological diffusion and the financial environment, Working paper, no-01-3. The United Nations University, Institute for New Technologies (downloaded on 1 of March 2012 from http://www.intech.unu.edu/publications/eifc-tf-papers/eifc01-3.pdf). Stoneman, P., Kwon, M.J., 1994. The diffusion of multiple process technologies. The Economic Journal 104, 420–431.

J. Gómez, P. Vargas / Research Policy 41 (2012) 1607–1619 Teece, D., 1986. Profiting from technological innovation: implications for integration, collaboration, licensing and public policy. Research Policy 15, 285–305. Teece, D., 2006. Reflections on profiting from innovation. Research Policy 35 (8), 1131–1146. Teece, D.J., 2000. Managing Intellectual Capital: Organizational, Strategic, and Policy Dimensions. Oxford University Press, Oxford. Vicente-Lorente, J.D., 2000. Inversión en intangibles y creación de valor ˜ Economía Industrial 332, 109– en la industria manufacturera espanola. 123. Vicente-Lorente, J.D., 2001. Specificity and opacity as resource-based determinants of capital structure: evidence for Spanish manufacturing firms. Strategic Management Journal 22, 157–177.

1619

Villalonga, B., 2004. Intangibles resources, Tobin’s q, and the sustainability of performance differences. Journal of Economic Behaviour and Organization 54 (2), 205–230. Wernerfelt, B., 1984. A resource based view of the firm. Strategic Management Journal 5 (2), 171–180. Winter, S., 1987. Knowledge and competence as strategic assets. In: Teece, D. (Ed.), The Competitive Challenge. Center for Research in Management, Berkeley, CA, pp. 159–184. Wozniak, G.D., 1984. The adoption of interrelated innovations: a human capital approach. The Review of Economics and Statistics 66 (1), 70–79. Wuyts, S., Dutta, S., Stremersch, S., 2004. Portfolios of interfirm agreements in technology-intensive markets: consequences for innovation and profitability. Journal of Marketing 68, 88–100.