Distance to Customers, Absorptive Capacity, and Innovation in High‐Tech Firms: The Dark Face of Geographical Proximity

Abstract

This paper investigates the impact of both geographical and relational proximity on the innovative performance of the firm. We address the role of one firm characteristic—its absorptive capacity—as a specific contingency affecting the relationship between different proximities and innovation. Using data from 158 high‐tech firms located in the Tiburtina Valley in Italy, we studied the relationship between these firms and their key customers. Our findings support the need to downplay the role of geographical proximity in promoting innovation. Our results also show that relational proximity to key customers has a complementary relationship with absorptive capacity, which positively moderates its influence on innovative performance.

Introduction

A steady stream of literature on industrial clusters, also known as industrial districts (Piore and Sabel 1984) or hot spots (Pouder and St. John 1996), has long maintained the importance of geographical proximity between organizations for innovation processes (Almeida and Kogut 1999; Folta, Cooper, and Baik 2006; Scott 2006; Woolley and Rottner 2008). However, some scholars have recently questioned the relationship between geographical proximity and innovation (Owen‐Smith and Powell 2004; Stuart and Sorenson 2003), arguing that the organization's knowledge base is rarely tied exclusively to local networks, even in the most advanced industrial clusters (Felzensztein, Gimmon, and Carter 2010; Molina‐Morales and Martinez‐Fernandez 2010). In addition to these widely held viewpoints, two recent developments addressing the core issue of innovation in industrial clusters have emerged (Camisón and Villar‐Lopez 2012). First, some authors have argued that the concept of proximity is multidimensional and not exclusively associated with the physical or geographical dimensions (Boschma 2005; Knoben and Oerlemans 2006). Additionally, other scholars, explaining the determinants of innovation in industrial clusters, have highlighted the uneven distribution of local knowledge resources due to firms' internal organizational resources and absorptive capacities (Cepeda‐Carrion, Cegarro‐Navarra, Jimenez‐Jimenez 2012; Giuliani and Bell 2005; Li, Veliyath, and Tan 2013; McCann and Folta 2008; Munari, Sobrero, and Malipiero 2012). Although the literature on innovation in clusters has argued that some firm characteristics might moderate the relationship between proximity and knowledge acquisition in start‐ups (Presutti, Boari, and Majocchi 2013), empirical evidence has yet to be collected regarding the moderating role of a key characteristic, namely, absorptive capacity, in the relation between proximity and innovation.

Taking into account these more recent perspectives, questions arise about the actual importance of distance to key partners to supporting innovative performance and, within this context, about the role played by a firm's internal attributes, such as absorptive capacity. In this study, we aim to answer two interrelated questions:

• What is the direct impact of both geographical and relational proximities on the innovative performance of colocated firms?

• What is the effect of the interaction between geographical and relational proximities and absorptive capacity on the innovative performance of colocated firms?

To contribute to the debate on the role of proximity in a firm's innovative performance, our framework conceptualizes proximity as a multidimensional notion. First, we introduced the relational proximity dimension into the model (Morgan 2004; Oinas 1999). This dimension captures both social and cognitive inter‐organizational proximity (Leana and Van Buren, 1999; Moran and Galunic 1998; Nooteboom et al., 2007). Second, we verified the effects of the interaction between absorptive capacity (Hotho, Becker‐Ritterspach, and Saka‐Helmhout 2012) and both relational and geographical proximity on the innovation activity of colocated firms (Expósito‐Langa., Molina‐Morales, and Capó‐Vicedo 2011; Hotho, Becker‐Ritterspach, and Saka‐Helmhout 2012).

We tested our hypotheses through a survey of 158 firms located in a geographical cluster of small and medium‐sized high‐tech firms in the center of Italy, the Tiburtina Valley, which is close to Rome. We focus our attention on a high‐tech cluster for several reasons: in addition to the great importance of this type of cluster in the current environment (Fosfuri and Ronde 2004) and given the nature and relevancy of knowledge flows to innovation in high‐tech clusters, they constitute a suitable empirical context in which to address our research questions.

We found that geographical proximity between a firm and its key customers is always detrimental to innovative performance, regardless of whether there are high or low absorptive capacity levels. Moreover, we found that firms' innovation capabilities are positively influenced by high levels of relational proximity only when they also possess high absorptive capacity values.

These results are relevant, first, because they contribute to a better understanding of the importance of different dimensions of proximity in fostering innovation. Second, they contribute to evaluating the contingency role of internal resources, such as absorptive capacity, in the relationship between different forms of proximity and a firm's innovative performance.

We structure the remainder of the paper as follows. First, we propose the theoretical framework and justify our hypotheses. Then, we describe the empirical setting and method. We conclude with a discussion of the results and suggestions for further research.

Theoretical Framework

Innovation Processes, Dimensions of Proximity, and Absorptive Capacities

Over recent decades, there has been increasing interest in proximity and its potential effects on the competiveness and innovativeness of firms. Authors have emphasized the important role of geographical proximity in the success and development of areas characterized as Marshallian industrial districts (Becattini 1991; Brusco 1982). Districts and proximity have been considered the source of innovation and the foundation for developing winning regions (Piore and Sabel 1984). More recently, the Triple Helix perspective has used the notion of proximity as a central element of knowledge‐based systems. The knowledge‐based system transforms institutional conditions into a knowledge infrastructure or, in other words, a triple helix network of university‐industry‐government relations (Etzkowitz and Leydesdorff 2000). In a different context, the proximity notion has been applied to population density at the metropolitan level, demonstrating that the geographic proximity of individuals possessing high levels of human capital—skills, expertise, or creative capabilities—enables their interactions and that these interactions facilitate the spillover necessary for innovation (Knudsen et al. 2008). Similarly, authors like Florida (2002) argue that it is the density and diversity of these talents in the territory that promote a small number of fast‐growing firms that in turn are the drivers of local development (Birch 1987; Kirchhoff 1994). Finally, from a strategic perspective, Porter (1994) proposes the determinant role of proximity and location on competition and the competitive advantage of firms.

Considerable recent research has focused on analyzing those factors that affect a firm's innovation process (Yli‐Renko, Autio, and Sapienza 2001; Zahra and George 2002). According to these studies, firms' innovation processes derive more from interactions with external business actors than from a “solitary genius” (Burgelman and Hitt 2007; Davidsson, 2005; Smith and Cao 2007; Woolley and Rottner 2008). In other words, innovation is not a discrete process resulting from knowledge developed by an isolated firm but rather emerges through interactions (Ahuja, 2000; Slotte‐Kock and Coviello 2010) based mainly on external social relationships (Dyer and Singh 1998; Squire, Cousins and Brown 2006; Tsai and Ghoshal, 1998). This is the case for innovations driven by customers, and it was found to consistently hold even when drivers differ (e.g., for design‐driven innovations; Verganti 2011a).

These studies have also considered the location of the knowledge exchange process among partners (Audretsch and Lehmann 2006; Feldman 1994; Jaffe, Trajtenberg, and Henderson 1993), confirming the importance of geographical proximity given that knowledge spillovers are not without costs generated by geographical distance (Maskell and Malmberg 2007; Scott, 2006). Central to this argument is the traditional assumption in economics and geography that knowledge spillovers are geographically localized and locally bound (Alcacer and Chung 2007; Arikan 2009; Jaffe Trajtenberg, and Henderson 1993; Krugman 1991). Thus, stocks and flows of knowledge tend to be spatially concentrated and, to some extent, available only to the colocated actors (Bell, Tracey, and Heide 2009). In this context, innovation can be understood as a joint action among neighbors (Camisón and Villar‐Lopez 2012). Empirical studies confirm that spatially concentrated actors benefit from knowledge externalities, in that short distances literally bring people together, favor contacts for information exchange, and facilitate reciprocal exchanges of tacit knowledge (Gordon and McCann 2005; Maskell 2001; Waxell and Malmberg 2007).

Ever since Marshall (1920) presented his seminal work on industrial districts, most scholars have accepted this positive view of colocation. However, an increasing number of subsequent studies have considered this argument to be overly simplistic, if not unrealistic, and have argued that there is no obvious relationship between externalities and geographical proximity (Huggins and Johnston 2010; Owen‐Smith and Powell 2004). According to Oinas (1999), few empirical studies have proved that the process of knowledge exchange is actually local (Cooke 2001). Some studies even consider the role of physical or geographical proximity critically, arguing that it does not generate externalities per se (e.g., Boschma 2005). Instead, they suggest that the creation of new knowledge and interactive learning is more likely to occur inside nonlocal relationships and that the knowledge base is rarely tied exclusively to local networks (Owen‐Smith and Powell 2004; Saxenian 1994). Instead, global relationships provide crucial sources of knowledge that complement traditional local “buzz” as a dispenser of knowledge for the innovation process (Bathelt, Malmberg, and Maskell 2004; Presutti, Boari, and Majocchi 2013).

In summary, research suggests not only that geographical proximity is insufficient to account for all knowledge spillovers (Sorenson, Rivkin, and Fleming 2006) but also that the concept of proximity is not exclusively limited to physical or geographical proximity (Boschma 2005; Knoben and Oerlemans 2006). A way to explain these results would be to identify the different dimensions of proximity, investigating the role of relational, cognitive, and epistemic proximities in facilitating knowledge dynamics among entrepreneurs (Boschma 2005). This framework indicates that including the social and cognitive dimensions of proximity reduces the need for geographical proximity, especially in the case of tacit knowledge transfer. Specifically, relational proximity may capture both social and cognitive interorganizational proximity (Leana and Van Buren 1999; Nooteboom et al., 2007). Because other intangible dimensions of proximity are crucial to the transfer of knowledge in network relations, it is recognized that relational proximity between actors, and not geographical proximity, fosters knowledge spillovers. In other words, there is nothing inherently spatial about the innovation process (Breschi and Lissoni 2001).

To gain a complete picture of the current debate about the role of proximity in firms' innovation processes, the firms' internal resources must also be included. Giuliani and Bell (2005), for example, found that the uneven and selective distribution of local resources and of knowledge affects innovation activity. In the context of proximity, the individual firm's knowledge base is an additive and distinct attribute of its systemic resources and capacities. Consequently, firms vary in their capacity to exploit opportunities (Munari, Sobrero, and Malipiero 2012). The differences in the amount of internal knowledge held by a firm generate an uneven and selective distribution of resources and knowledge being transferred and received in a close environment (Cepeda‐Carrion, Cegarro‐Navarra, and Jimenez‐Jimenez 2012; Giuliani and Bell 2005; Sammarra and Biggiero 2008). Among a company's internal attributes, R&D efforts related to the firm's knowledge base, and thus its absorptive capacity (Hervas‐Oliver et al. 2012), are particularly significant. Previous research has associated a firm's absorptive capacity with organizational learning and the ability to apply information received from external sources (Cohen and Levinthal 1989; Valdaliso et al. 2011). In addition, empirical findings demonstrate that such absorptive capacity exerts a positive influence on innovation (Cassiman and Veugelers 2002; Hotho, Becker‐Ritterspach, and Saka‐Helmhout 2012; Huerg, 2006). The literature on innovation in clusters still needs to develop a focused analysis of the contingency role of absorptive capacity in the relation between proximity and innovation.

Hypothesis Development

The current study focuses on firms' relationships with their key customers. Key customers are defined as a firm's largest customers in terms of proportion of sales revenue. Supplier–customer relationships are universally recognized as being crucial for innovation (Von Hippel 1977) and have become an integral part of business‐to‐business operating strategies (Felzensztein, Gimmon, and Carter 2010; Langfield‐Smith and Greenwood, 1998; Squire, Cousins, and Brown 2009).

To address our research questions, we focus our hypothesis development on some key aspects. First, we evaluate the impact on innovation of two different dimensions of proximity. Moreover, as we believe that internal firm factors may affect the relation between proximity and innovation, we evaluate the moderating effect of absorptive capacity. More specifically, at the empirical level, we first consider the direct effects of geographical (Hp. 1) and relational proximity (Hp. 2) on the innovative performance of the high‐tech firms of our sample. Then, we consider the interaction effects of absorptive capacity with geographical and relational proximity on innovative performance (Hp. 3; Hp. 4).

Geographical Proximity and Innovative Performance

The role of geographical proximity is still hotly debated, with some current authors suggesting a positive effect of colocation but others arguing that geographical proximity has a negative impact on innovation (Gilsing et al., 2008; Hoppmann et al. 2013). Recently, an increasing number of scholars have challenged the view that physical proximity, in the context of industrial clusters, favors firm innovation (Antonelli 2000; Breschi and Lissoni 2001). Some authors maintain that the negative effects are particularly strong in the high‐tech context (e.g., Breschi and Lissoni 2001), where links with dispersed sources of knowledge that are distributed worldwide are essential (Friesl et al. 2008). Moreover, local interactions can generate spatial lock‐in, which is detrimental to learning interactions and, ultimately, to innovation (Capó‐Vicedo, Expósito‐Langa, and, and Molina‐Morales 2008). This perspective is sometime ascribed to the idea that clusters tend to become close networks (Boschma 2005), which might even isolate firms from external and profitable sources of knowledge and information through a lock‐in effect (Bathelt, Malmberg, and Maskell 2004; Romanelli and Khessina 2005). To avoid such spatial lock‐in, firms might pursue distant relationships that provide access to the outside world (Hendry, Brown, and De Fillippi 2000). Another perspective, though not subverting the previous one, is more focused on the diversity of accessed knowledge. Burt (2004) claims that when innovation is the key issue, the sources of knowledge might not be the focal question. Empirical research shows that what matters for innovation is how diverse and even contradictory the information and its possible interpretations are.

By combining these two perspectives and by focusing on the relationship with key customers, we argue that geographical proximity with key customers might have a negative impact on innovation.

Thus, we formulate the following hypothesis.

H1: For firms located in industrial clusters, the geographical proximity to key customers is negatively associated with their innovative performance.

Relational Proximity and Innovative Performance

The relational proximity dimension refers to both social and cognitive interorganizational proximity (Leana and Van Buren 1999; Moran and Galunic 1998; Nooteboom et al. 2007). According to reports on embeddedness (Granovetter 1985; Uzzi 1997), social proximity entails the level of trust between a local firm and its customers. Cognitive proximity focuses on similarities in how the actors perceive, interpret, and evaluate the world. High levels of social and cognitive proximity can reduce the knowledge distance between business partners (Boschma 2005), broadening their common knowledge base and expertise. Therefore, relational proximity facilitates knowledge acquisition and, particularly, tacit knowledge exchange. When relational proximity between partners is high, agents act similarly, and more knowledge can be transferred when practices, institutional legacies, and work cultures are comparable (Mowery, Oxley, and Silverman 1996). According to Dakhil and Clercq (2004), higher levels of common values and culture are typically associated with higher levels of innovation. When agents are involved in the same network and have a shared vision, reinforced by reciprocal trust, they tend to have similar perceptions about how to act, which promotes mutual understanding and the exchange of ideas and resources (Tsai and Ghoshal 1998). Moreover, when firms have similar referral structures, knowledge can be communicated, transferred, and acquired more effectively and efficiently (Knoben and Oerlemans 2006). In this context, parties tend to share goals and have a common understanding of which innovation is needed. In turn, a common attitude regarding how to perform research activities emerges, leading to an improvement in innovative performance (Krause, Handfield, and Tyler 2007). Shared vision can be regarded as a mechanism favoring knowledge integration between parties in a network (Inkpen and Tsang 2005); in this sense, it is a critical element in the innovation process. Finally, the presence of trust in business relationships makes reciprocal knowledge exchange more efficient, thus allowing it to substitute for the formal contracts, incentives, and monitoring mechanisms associated with typical business transactions. This, in turn, reduces the risk of reciprocal opportunism and, more generally, lowers transaction costs, promoting knowledge exchange among partners (Nahapiet and Ghoshal 1998).

Conversely, cultural conflicts and misunderstandings between firms and their customers can limit information and knowledge acquisition as well as interorganizational learning. When goals and cultures are incongruent, misunderstandings and conflicts between firms and customers are more likely to arise. Parties may become disappointed and restrict information exchange, which will negatively affect outcomes for the firms (Inkpen and Tsang, 2005; Krause, Handfield, and Tyler 2007). In summary, relational proximity facilitates interactions between actors and access to external knowledge, regardless of geographical distance. These arguments suggest a positive association between social and cognitive proximity and knowledge acquisition and, thus, innovative performance. This idea leads to our second hypothesis.

H2: For firms located in industrial clusters, relational (social and cognitive) proximity to key customers is positively associated with their innovative performance.

The Moderating Role of Absorptive Capacity

Firms vary in their capacity to understand, develop, and use certain external resources, particularly those related to knowledge and innovation. Previous studies have identified a firm's absorptive capacity as a key factor in improving its ability to benefit from external resources (Cohen and Levinthal 1989; Valdaliso, Aranguren, and Lopez 2011). This capacity is also considered to be a measure of the firm's capacity to innovate and to develop new knowledge (Zahra and George 2002). Extant research argues in favor of a positive relationship between absorptive capacity and innovative performance (Cepeda‐Carrion, Cegarro‐Navarra, Jimenez‐Jimenez 2012; Hotho, Becker‐Ritterspach, and Saka‐Helmhout 2012).

In the case of key customer relationships, the propensity to use knowledge depends on the firm's own absorptive capacity and on the level of firm integration into networks (Tether and Tajar 2008). Knowledge provided by key customers typically requires complex and intensive interaction between companies, in which both parties participate in interactive learning (Hertog 2000; Sundbo 2001). For example, studying the case of firms based in Hong Kong, Wu (2008) empirically demonstrates that repeated transactions among partners promote knowledge exchange, leading to a positive impact on firm performance. The literature has emphasized the complementarities between external knowledge provided by customers or knowledge providers and a company's resources and capabilities (Friesl 2012; Tether and Tajar 2008).

We share Zaheer and Bell's (2005) perspective, which claims that external and internal resources must be analyzed together to offer a complete explanation of a firm's performance, including innovative performance. External and internal resources can interact and affect each other. Different scholars have started to look at internal factors, such as absorptive capacity, as contingency factors in the firm's use of external alliances (Park, Chen, and Gallagher 2002) and in its capacity to extract value from its network position (Shipilov 2009).

Studies on firms located in clusters found that the uneven and selective distributions of local resources and of knowledge affect innovation activity, with absorptive capacity playing a direct positive role on firms' innovative performance (Giuliani and Bell 2005; Nooteboom et al. 2007). More recently, some researchers claim that the cluster literature overemphasizes the importance of geographical proximity while underestimating the role of network relations and the capacity of the firm (Ter Wal and Boschma 2009). In line with this perspective, other researchers suggest a contingency role for firms' internal resources, such as absorptive capacity, which might improve their ability to exploit external resources and thereby enhance their innovative performance (Hervas‐Oliver et al. 2012; Presutti, Boari, and Majocchi 2013). Accordingly, we expect that the exploitation of external resources, accessed through more or less proximate relationships with key customers, will be greater for companies with higher levels of absorptive capacity. In our research, we expect that a firm's absorptive capacity moderates the relationship between geographical and relational proximity and the firm's innovative performance. More specifically, higher levels of absorptive capacity should mitigate the negative effects of geographical proximity to customers on a firm's innovative performance. Moreover, we expect that higher levels of absorptive capacity will strengthen the positive effects of the cognitive and social proximity of customers on innovative performance.

H3: For firms located in industrial clusters, the effect of geographical proximity to key customers on innovative performance is moderated by the level of their absorptive capacity, such that geographical proximity will have a less negative effect if the levels of absorptive capacity are high.

H4: For firms located in industrial clusters, the effect of relational (social and cognitive) proximity to key customers on innovative performance is moderated by the level of their absorptive capacity, such that relational proximity will have a stronger positive effect if the levels of absorptive capacity are high.

Methods

Sample and Data

Over recent decades, the development of high‐tech clusters has become a popular strategy for firms and regions, and successful high‐tech clusters create jobs and attract capital, talent, and many other types of resources (Fosfuri and Ronde 2004; Longhi and Keeble 2000). In these contexts, innovation processes become fully developed; however, because these processes are so complex, firms also become much more dependent on external flows of knowledge resources (Bresnahan, Gambardella, and Saxenian 2001).

The field setting of this research consists of a geographical cluster of small and medium‐sized high‐tech firms located in one of the most important areas in central Italy, the Tiburtina Valley, which is approximately 14 km from the center of Rome. As in Silicon Valley in the United States, the high‐tech firms located in this restricted urban context have created a homogeneous agglomeration that currently represents the most typical example of a high‐tech metropolitan cluster in Italy. This cluster is also known for its high export rates due to the presence of several important foreign multinationals. The high number of specialized firms located in this small area has led to the development of new businesses and the rapid diffusion of knowledge among local actors (Majocchi, Odorici, and Presutti 2016). This urban cluster has already been the focus of previous international studies (Presutti, Boari, and Majocchi 2011).

At the beginning of our data collection (September 2013), the sectors represented in this area were (1) electronics (320 firms); (2) media (200 firms); and (3) new economy (e.g., manufacturers of new hardware, firms in the information services industry, Internet access providers, and telecommunication network managers; 522 firms). We focused on the electronics sector, which, according to the definition of the National Federation of Electronics Firms, consists of the computer industry, electronics in the strictest sense, and telecommunications. Specifically, we tested our hypotheses using survey data from 158 high‐tech firms in the electronics sector, defined according to standard industrial classification (SIC) codes, located within this cluster.

The data to test our hypotheses come from a direct survey using a specially designed questionnaire. The data collection process lasted approximately 5 months. The survey process was developed in four phases: (1) conducting literature reviews to develop the measurement scales; (2) developing the questionnaire (structured in a closed question‐answer form); (3) pretesting the questionnaire on three randomly selected sample firms; and (4) data collection. The firms in our sample belong to the “specialized suppliers” category according to the Pavitt classification in that they primarily produce and offer technology and services to their industrial customers. In this case, customers also access technology through the acquisition of products from suppliers (Yanez, Khalil, and Walsh 2010). Thus, for both customers and suppliers, firm knowledge acquisition and exploitation processes depend to a great extent on interactive learning between them. This situation encourages local firms to become increasingly specialized, resulting in strong differentiation in their internal knowledge bases. The R&D activity and the scope and extent of new products launched for these firms are strongly conditioned by knowledge acquired from their customers because new products need to be in line with customers' specific requirements (Dyer and Singh 1998).

To ensure that all firms in our sample were involved in innovation activities, we checked their business descriptions in the source database. Firms with no internal R&D activity or who only offered nontechnical services were excluded from the study. This process left a total of 275 firms, 158 of which accepted our request for a personal interview to complete the questionnaire (58% response rate). A comparison of differences in the mean values of the responding and nonresponding firms revealed no statistically significant response bias in terms of their 3‐year average sales revenues, firm age, or number of employees.

Our key informant was the entrepreneur. Firms were contacted by telephone to obtain the names of potential respondents (the entrepreneurs) and to determine whether they would agree to complete the survey through face‐to‐face interviews. The focus of our empirical research was on the vertical relationship between each firm and the customer that accounted for the highest proportion of sales revenue for the previous year (Yli‐Renko, Autio, and Sapienza 2001). Using a semistructured questionnaire, entrepreneurs were asked to use seven‐point Likert scales to express their opinions about all items. These items were selected to measure relational proximity with the firm's key customers. Moreover, entrepreneurs were asked to indicate the customer's location, the duration of their business relationship, the size and age of both firms, and the ratio of R&D expenditures of both firms to total sales. Finally, to capture innovation activity, firms were asked to indicate the number of new products developed as a result of this key customer relationship.

Overall, the firms in the sample developed an average of 7.3 new products per year (range 1–25). The average sales of the respondent firms amounted to €814,000, with an average age of 18.10 years. In addition, the average R&D expenditure in relation to total sales was 3.67% (range 0.01–15%). The key customers are similar in age to their suppliers (M = 18.16 years) but were on average smaller (M_sales = €530.63). The number of key clients located within the district is 54. Table 1 reports the descriptive statistics for the untransformed variables.

Table 1. Descriptive Statistics of the Untransformed Variables

	Average	S.D.	Min	Max
Innovation	7.3	6.1	1	25
Size (€0.000)	814.33	1116.75	100	10,000
CustSize (€0.000)	530.63	542.68	40	3,000
CustAge	18.01	12.86	2	90
FirmAge	18.16	13.44	2	80
AbsCap	3.67	3.04	0.01	0.15
RelProx	4.13	2.29	1	7

Measures

Dependent Variable: Innovative Performance

We measured innovation activity by asking the firms how many new products or services they had developed during the previous three years as a result of the relationship with their key customers. We consider both product and services without distinction (Walsh and Linton 2011) in order to focus our analysis on the all range of innovation activities the impact on firms' sales (McDermott and Walsh 2001).

Independent Variables

Absorptive Capacity (AbsCap)

Following other recent research on measuring absorptive capacity (e.g., Cohen and Levinthal 1989; Meeus, Oerleman, and Hage 2001), we used R&D spending as a proxy to measure the level of a firm's investment in absorptive capacity useful to its innovation activity. This parameter was measured by considering the log value of the average ratio between R&D expenditures and total sales for the previous 3 years.

Geographical Proximity Between Actors (GeoProx)

In accordance with previous studies (Boschma 2005; Rallet and Torre 2000), we distinguished between key customers located within the district and those located outside. We defined a dummy variable equal to 1 for customers located inside the cluster and 0 otherwise. The number of customers located within the district is 54 (34.1%).

Relational Proximity (RelProx)

We used multi‐item measures to analyze the relational proximity between the firm and its key customer because these measures provide considerable advantages over single‐item measures (e.g., Churchill 1979). In our research, we use the relational proximity dimension to capture both social and cognitive interorganizational proximity, in accordance with many significant studies (e.g., Leana and Van Buren 1999; Moran and Galunic 1998; Nooteboom et al., 2007). According to literature concerning embeddedness (Granovetter 1985; Presutti, Boari, and Majocchi 2011; Uzzi 1997), the social proximity dimension measures the level of trust between a local firm and its customers. In particular, we focus on three dimensions of interorganizational trust: lack of opportunistic behavior, creation of common investments (degree of commitment), and presence of informal relationships. We selected six items to measure these dimensions of trust that have been used in many previous studies on trust, networks, and social capital. The cognitive proximity dimension encompasses similarities in the ways that firms and customers perceive, interpret, and evaluate the world. Therefore, a high level of cognitive proximity (Boschma 2005) can reduce the distance between the knowledge bases of business partners, thereby broadening their common knowledge base and expertise. In line with previous studies (Nahapiet and Ghoshal 1998), we measured the cognitive proximity between business partners with six items using seven‐point Likert scales. These items capture two interconnected aspects of relational cognitive proximity. First, they reflect the extent to which the business relationship is characterized by the development of common goals, norms, and reciprocal expectations involving the goodwill and trustworthiness of the exchange partner (Tsai and Ghoshal 1998). Second, the selected items verify the overlap of personal and business interests among partners, that is, the extent to which the business relationship is characterized by personal social ties (Nahapiet and Ghoshal 1998).

To verify the unidimensionality of the relational proximity construct (Churchill 1979), we submitted the scales used to analyze both cognitive and social dimensions to a factor analysis. Our original selection of 12 items was unsatisfactory from a statistical viewpoint (Kaiser 1960) because some single factors had eigenvalues greater than 1.0, and the loadings of some items did not exceed the cut‐off point of 0.30. Moreover, our scale showed a low degree of internal reliability (Nunnally 1978). Thus, by submitting the items to factor analysis, we reduced the number of items from 12 to 6 (3 items for social and 3 for cognitive proximities) to measure relational proximity. With this reduced combination of items, the measurement model performed well. The selected constructs demonstrate good internal consistency and reliability (details are included in the next section). The measure is a reflective manifestation of the underlying constructs as defined by Edwards (2011) because it is the outcome of unobserved latent variables. Because we expected the effect of relational proximity to decrease for higher values of the variable, we considered a logarithmic specification of this variable.

Control Variables

We included some control variables to isolate the effect of the independent variables in the model. First, an important factor influencing innovation activity is the size of the involved partners (Acs and Audretsch 1991; Mowery, Oxley, and Silverman 1996; Tsai and Ghoshal 1998), in line with the idea that larger firms may invest more resources in R&D activities. As several studies suggest (e.g., Kogut and Zander 1992), both superior resources and economies of scale allow larger firms to exploit external knowledge successfully for their innovation process. We controlled for the effects of size by including the total sales of both the firms (Size) and their customers (CustSize).

Moreover, we included two additional control variables to evaluate the age of the firm (FirmAge) and of that its customers (CustAge). We assume that these elements will have an influence on the firm's learning ability and its knowledge exploitation process (Zahra, Jennings, and Kuratko 1999). We computed the age of the firms and of the customers as the number of years since their foundation. We used the log of these values to account for declining effects at higher values. Table 2 reports the descriptive statistics of the variables and their correlations.

Table 2. Descriptive Statistics and Correlations

	Average	S.D.	Size [1]	CustSize [2]	CustAge [3]	FirmAge [4]	AbsCap [5]	RelProx [6]	GeoProx [7]
[1]	814.33	1116.75	1
[2]	530.63	542.68	0.287^***	1
[3]	2.70	0.60	0.189 ^**	0.354^***	1
[4]	2.68	0.65	0.471^***	0.409^***	0.584^***	1
[5]	0.739	1.35	0.354^***	0.271^***	0.395^***	0.575^***	1
[6]	1.18	0.751	0.054	0.066	0.183 ^**	0.205^***	0.201 ^**	1
[7]	–	–	−0.281^***	−0.211^***	−0.382^***	−0.481^***	−0.584^***	−0.289^***	1

^***p < .01.

^**p < .05.

^*p < .1.

Reliability, Validity, and Data Analysis

Several precautions were taken to ensure that our data are reliable and valid. First, as mentioned previously, we pretested the survey with three entrepreneurs from our sample. Second, we used multi‐item measures to measure relational proximity because they provide considerable advantages over single‐item measures (Churchill 1979). We measured these items on the questionnaire using seven‐point Likert scales. As a first step of measure validation, to assess the unidimensionality of the research constructs (Churchill 1979), we factor analyzed the final scales with six items using the principal axis method, positing a single factor (exploratory factor analysis). After exploring the factor structure of the data, we submitted the data to confirmatory factor analysis. The results of this analysis verify that the measurement model performed well because the selected constructs demonstrate good internal consistency and reliability: the standardized factors are all above the recommended minimum of 0.40, and the average variances extracted are all above the recommended minimum of 0.50 (range 0.59–0.88). The composite reliabilities are all above the recommended minimum of 0.70 (Table 3). Finally, the CFA showed that the measurement models fit the data reasonably well, with a goodness‐of‐fit index superior to 0.9 (GFI).

Table 3. Factor Analysis on Relational Proximity

Factor name	Measurement item	Standardized loading	Cronbach's alpha
Relational proximity	Social dimension		0.822
	The firm sends the products to the customer before receiving the full payment.	0.88^**
	The transaction with the customer is based on standard procedures without formal agreements.	0.59^**
	The customer and the firm are strongly tied by common investments aimed to reinforce their growth.	0.65^**
	Cognitive dimension
	The firm has never had problems with the client because they have similar interests.	0.71^**
	The customer often comes to the firm's head office without reasons related to work.	0.75^*
	The purchase order was defined in a nonworking environment.	0.85^**
	Goodness‐fit‐index (GFI) > 0.91

^***p < .01.

^**p < .05.

^*p < .1.

Analysis, Techniques, and Results

To test our hypotheses, four models were analyzed using the ordinary least squares technique with Release 11 of the Stata software package. The estimated equation is: where α₀ is the constant intercept, β_i is the regression coefficient, and ε is the error term.

Table 4 presents the regression results for all models. The results are fairly constant across all specifications, and the explanatory power of the models is highly significant in terms of both R‐squared and adjusted R‐squared values. These results suggest that the models have been correctly specified and are stable. To determine whether our results are subject to multicollinearity, the variance inflation factors (VIFs) were also calculated for all models. The values of the mean VIF scores, which were well below the commonly used critical value of 10 (Cohen et al. 2003) and even the more restricted value of 5, suggest that in our study, multicollinearity is not a concern. The first column, Model 1, shows the estimated coefficients for the model containing our four control variables and the variable measuring absorptive capacity. The coefficients of the size variables were positive and statistically significant for the firm, its key clients, and absorptive capacity. These variables influenced the firm's level of innovative performance. In contrast, the variables on the age of both the firms and their clients were <0 but were not statistically significant. Therefore, our results did not support the view that age was a critical issue in determining a firm's innovative performance.

Table 4. Regression Results^a

	(1)	(2)	(3)	(4)
Size	0.00159^***	0.00154^***	0.00157^***	0.0014^***
	(0.000312)	(0.000296)	(0.000294)	(0.000294)
CustSize	0.00166^***	0.00172^***	0.00175^***	0.00177^***
	(0.000619)	(0.000588)	(0.000583)	(0.000561)
CustAge	−1.036	−1.348^**	−1.398^**	−1.422^**
	(0.629)	(0.601)	(0.596)	(0.576)
FirmAge	−0.0531	−0.400	−.477	−0.458
	(0.700)	(0.669)	(0.664)	(0.648)
AbsCap	2.774^***	2.271^***	2.263^***	2.257^***
	(0.273)	(.285)	(0.282)	(0.408)
GeoProx		−3.233^***	−2.948^***	−2.007^***
		(0.761)	(0.768)	(0.777)
RelProx			0.758^*	0.269
			(0.395)	(0.417)
GeoProx×AbsCap				−1.732^***
				(0.534)
RelProx×AbsCap				0.594^**
				(0.270)
Constant	6.029^***	9.291^***	8.600^***	8.266^***
	(1.709)	(1.793)	(1.813)	(1.767)
Observations	158	158	158	158
R‐squared	62.9%	66.9%	67.6%	70.40%
Adj. R‐squared	61.6%	65.5%	66.4%	68.60%
F(e;dof)	F(5, 152) = 51.50	F(6, 151) = 50.75	F(7, 150) = 44.80	F(9, 148) = 39.11
Prob > F =	0.0000	0.0000	0.0000	0.0000
VIF (mean)	1.61	1.67	1.60	2.16

^aStandard errors are in parentheses.

^***p < .01.

^**p < .05.

^*p < .1.

Models 2–4 include the independent variables, leading to a significant increase in the overall explanatory power of the models and confirming the significance of our independent variables. In Model 2, we tested for the impact of the dummy variable that gauges the location of clients within the cluster (GeoProx). The coefficient of the variable was negative and strongly significant (β = −3.23; p < .01), suggesting a relevant negative effect of being located nearby. These results fully support Hypothesis 1 in the sense that geographical proximity to key customers is detrimental to innovative performance. Next, Model 3 determined whether relational proximity had a positive effect on innovative performance (RelProx). In support of Hypothesis 2, the coefficient is positive and significant (β = 0.758; p < .10).

Next, we tested Hypotheses 3 and 4 using interaction terms to investigate the role of absorptive capacity in moderating the impact of relational and geographical proximities. Hypothesis 3 predicted that absorptive capacity would mitigate the negative effect of geographical proximity on innovation activity, whereas Hypothesis 4 predicted that it would reinforce the positive effect of relational proximity on innovation activity. Model 4 showed that although relational proximity loses its significance, there are significant interaction effects among the variables, as shown in Figures 1 and 2. To select high and low values, we followed the suggestions of Aiken, West, and Reno (1991) in using 1 and 0 for categorical variables, as in the case of our geographical variable, and in using values corresponding to 1 S.D. above and 1 S.D. below the mean for continuous variables, as in the relational proximity case. The Y‐axis represents the dependent variable (innovation activity), and the X‐axis represents low and high geographical proximity, that is, within and outside the cluster (Figure 1), and low and high relational proximity (Figure 2). Figure 1 clearly shows (as both slope lines decrease) that Hypothesis 3 is not confirmed. Innovative performance is negatively influenced by geographical proximity for both high and low absorptive capacity levels. Moreover, geographical proximity, that is, being located in the same cluster as clients, has a particularly negative impact when the absorptive capacity is higher.

Figure 1. Innovation Activity: The Interaction Effect of Geographical Proximity and Absorptive Capacity

Figure 2. Innovation Activity: The Interaction Effect of Relational Proximity and Absorptive Capacity

Figure 2 shows that innovative performance is clearly not influenced by a high level of relational proximity when the absorptive capacity of the firm is low. Only for high absorptive capacity values is innovative performance positively influenced by high levels of relational proximity to customers.

Discussion, Implications and Limitations

This paper adopted a multidimensional perspective on proximity to explore the effects of the geographical and relational (social and cognitive) proximities of high‐tech firms to their key customers in terms of innovative performance. Moreover, we investigated whether the firm's level of absorptive capacity moderates these effects.

First, our results suggest that the geographical proximity between firms and key customers reduces firms' innovative performance. Having customers located in the same cluster has a detrimental effect on firms' innovative performance. Geographical proximity with partners isolates firms from external sources that could be beneficial to their innovation activity through a lock‐in effect (Bathelt, Malmberg, and Maskell 2004; Romanelli and Khessina 2005). In this direction, other studies explain the negative impact of geographical proximity on innovation according to the network inertia problem (Boschma 2005; Uzzi, 1997), which limits the ability of a firm to exploit knowledge from different external sources of innovation when overembedded with geographically concentrated customers.

Regarding our specific context of high‐tech firms, the results could be ascribed to the fact that nearby business partners already have the same technological information as local firms, whereas exposure to many different external sources of technological knowledge is essential for exploring new innovation opportunities. This idea was supported by the characteristics of the larger and younger key customers of our sample; innovation opportunities could be better accessed through relationships with innovative and highly active customers located far from the cluster. The diversity associated with distant contacts can be essential for innovative contexts such as the investigated cluster (Friesl 2012; Hendry, Brown, and De Fillippi 2000). This idea is very much in line with what specifically happens in our empirical setting, where high‐tech firms require primary exploration activities to capture new ideas and improvements beyond the exploitation of the elements/resources/products they already have. These exploration activities are likely to be much better developed with distant rather than close partners, thanks to the diverse information and knowledge that can be accessed and combined (Gilsing et al., 2008; Hendry and Brown 2006; Hoppmann et al. 2013; Presutti, Boari, and Fratocchi 2007). Accordingly, our findings provide further support to the idea that diversity in accessed and combined information and knowledge could be much more important than a pure source (Burt 2004).

As expected, relational proximity was found to exert a positive impact on a firm's innovative performance. This result further supports the importance of following a broader multidimensional view of the proximity concept while also considering intangible proximity dimensions as important factors explaining the innovation process (Breschi and Lissoni 2001; Boschma 2005; Ter Wal and Boschma 2009; Torre 2008). Our results confirm that for high‐tech firms and their key customers, relational proximity seems even more relevant for knowledge transmission. Key customers can reinforce both the creation of new products and the technological distinctiveness of their primary suppliers. External knowledge is a particular requirement because high‐tech firms must leverage interorganizational customer relationships to broaden their stock of knowledge regarding both technology and R&D activity (Pirolo and Presutti 2010).

We found that a firm's absorptive capacity influences the role of geographical and relational proximity in innovation. In particular, and contrary to our expectations, we found that innovative performance is negatively influenced by geographical proximity to customers, regardless of whether the firm has high or low levels of absorptive capacity. Geographical proximity to customers has a particularly negative impact when absorptive capacity levels are high, thus suggesting that knowledgeable firms could be harmed by geographically proximate key customers.

Finally, with respect to the role played by absorptive capacity in the relation between relational proximity and innovation, we found that innovative firm performance was positively influenced by higher levels of relational proximity to customers only for firms with high absorptive capacity. In contrast, this proximity does not influence the innovative performance of firms with a lower level of absorptive capacity. This observation suggests that the combination of significant internal capacities and relational proximity is what matters most for innovative performance.

Our findings contribute to the literature on innovation in clusters, where the role of contingency factors in the relationship between proximity and innovation, although theoretically addressed (Presutti, Boari, and Majocchi 2013), still requires empirical evidence. In particular, our contribution focused on the moderating role of absorptive capacity.

Regarding the contingency role of absorptive capacity, our findings also illustrate that relational proximity has a complementary relationship with absorptive capacity, which positively moderates its influence on innovative firm performance. In contrast, absorptive capacity cannot substitute for a high level of geographical proximity by moderating its negative influence on innovative firm performance. A combination of high absorptive capacity and relational proximity to distant partners appears to be desirable.

By addressing the substitutive or complementary role of a firm's contingency factors, we also contribute to a growing focus in the literature on network advantage, which recognizes that individual actors might differently benefit from their network relations according to specific external contingencies (such as industry context and uncertainty characteristics) (Ahuja 2000) and internal contingencies (such as absorptive capacity, bargaining power, and protection against noncooperation of partners) (Shipilov 2009).

Our findings contribute to the more general understanding of the internal heterogeneity of knowledge distribution in clusters and its relation with innovation. Firm investments in high levels of absorptive capacity can help explain the access to external distant sources of knowledge, the impact on their innovative performance (Romanelli and Khessina 2005) and, consequently, the selective and uneven distribution of knowledge in the cluster (Giuliani and Bell 2005).

Finally, the insights obtained from this study contribute, even if indirectly, to the debate on the role of heterogeneous, that is, nonredundant, information and knowledge accessed through network relations and their impact on firms' innovative performance, a topic that is drawing greater attention in studies on geographical clusters and networks (Molina‐Morales and Martinez‐Fernandez 2006).

A large set of managerial implications arises from this study. Cluster firms should actively develop relationships with distant customers to favor knowledge acquisition useful to innovative performance (Presutti, Boari, and Fratocchi 2016). Firms must select those distant customers who can act as bridges to enable a local firm to enlarge its knowledge sources and innovation ideas. Thus, a well‐developed system of “global pipelines” connecting local firms to customers spread around the world would be extremely beneficial for a firm's innovation activity. In addition, our study highlights the importance of investing in absorptive capacity to benefit from relational proximity to customers. By combining high investments in absorptive capacity with relational proximity, firms are able to increase their innovative performance. Implications in terms of the role of nonredundant information accessed through relationally strong partnerships with geographically distant customers also arise from our study.

Some limitations of our study must be discussed to pinpoint opportunities for further research. First, our results refer to a high‐tech cluster. To generalize our theory, a test with more traditionally oriented clusters should be performed. Second, our definition of innovative performance is limited to the number of products and services developed by these firms in the past 3 years, regardless of the drivers of innovation (customer, process, design, etc.) and with no distinction between different types of innovation (i.e., disruptive and incremental). Further research could try to develop this point by addressing heterogeneity in innovation processes. Additionally, in this study, absorptive capacity was conceptualized only as an internal resource and as an indicator of capabilities. Further research could extend into other dimensions of strategic internal assets to gain a more complete picture of the importance of the mediating role of firms' internal resources and capabilities (Fernhaber, McDougall‐Covin, and Shepherd 2009). Researchers could also include other significant boundary conditions that might influence the importance of distant and near customer relationships for firms' innovation processes.

Finally, given that our study analyzes geographic and social proximity separately, further research could consider more complex relationships among these factors, such as combining effects, in line with the previous works of Parra‐Raquena, Molina Morales, and García Villaverde (2010) and Nooteboom et al. (2007).

Additional information

Notes on contributors

Manuela Presutti

Manuela Presutti is associate professor at the Department of Management, University of Bologna.

Cristina Boari

Cristina Boari is full professor at the Department of Management, University of Bologna.

Antonio Majocchi

Antonio Majocchi is full professor at the Department of Management, University of Pavia.

Xavier Molina‐morales

Xavier Molina‐Morales is full professor at the Department of Business Administration and Marketing, Universitata Jaume I.